IEE ELECTROMAGNETIC WAVES SERIES 28

Series Editors: Professor P. J. B. Clarricoats

Professor Y. Rahrnat-Samii
Professor J. R. Wait

Handbook of

ANTENNAS
Other volumes in this series:
Volume 1 Geometrical theory of diffraction for electromagnetic waves
G. L. James
Volume 2 Electromagneticwaves and CUN& structures L. Lewin,
D. C. Chang and E. F. Kuester
Volume 3 Microwave homodyne systems R. J. King
Volume 4 Radio direction-finding P. J. D. Gething
Volume 5 ELF communications antennas M. L. Burrows
Handbook of
Volume 6 Waveguide tapers, transitions and couplers F. Sporleder and
H. G. Unger
Volume 7 Reflector antenna analysis and design P. J. Wood
Volume 8 Effects of the troposphere on radio communications
M. P. M. Hall

ANTENNAS
Volume 9 Schumann resonances in the earth-ionosphere cavity
P. V. Bliokh, A. P. Nikolaenko and Y. F. Flippov
Volume 10 Aperture antennas and diffraction theory E. V. Jull
Volume 11 Adaptive array principles J. E. Hudson
Volume 12 Microstrip antenna theory and design J. R. James, P. S. Hall
and C. Wood
Volume 13 Energy in electromagnetism H. G. Booker
Volume 14 Leaky feeders and subsurface radio communications
P. Delogne
Volume 15 The handbook of antenna design, Volume 1A. W. Rudge,
K. Milne, A. D. Olver, P. Knight (Editors)
Volume 16 The handbook of antenna design, Volume 2 A. W. Rudge,
K. Milne. A. D. Olver. P. Kniaht (Editors)
Volume 17 ~ u ~ e i l l & cradar
e predichon P. Rohan
Volume 18 Cormaated horns tor microwave antennas P. J. B. Clarricoats
and A-D. Olver
Volume 19 Microwave antenna theory and design S. Silver (Editor)
Volume 20 Advances in radar techniques J. Clarke (Editor)
Volume 21 Waveguide handbook N. Marcuvitz
Volume 22 Target adaptive matched illumination radar D. T. Gjessing
Volume 23 Ferrites at microwave frequencies A. J. Baden Fuller Edited by
J R James & P s Hall
Volume 24 Propagation of short radio waves D. E. Kerr (Editor)
Volume 25 Principles of microwave circuits C. G. Montgomery,
R. H. Dicke, E. M. Purcell (Editors)
Volume 26 Spherical near-field antenna measurements J. E. Hansen
(Editor)
Volume 27 Electromagnetic radiation from cylindrical structures
J. R. Wait
Volume 28 Handbook of microstrip antennas J. R. James and P. S. Hall
(Editors)
Volume 29 Satellite-to-ground radiowave propagation J. E. Allnutt
Volume 30 Radiowave propagation
~ . - M. P. M. Hall and L. W. Barclay
(Editors)
Volume 31 Ionospheric radio K. Davies

Peter Peregrinus Ltd, on behalf of the Institution of Electrical Engineers

Published by: Peter Peregrinus Ltd., London, United Kingdom

o 1989: Peter Peregrinus Ltd.

All rights resewed. No part of this publication may be reproduced, stored

in a retrieval system or transmitted in any form or by any means- Contents
electronic, mechanical, photocopying, recording or otherwise-without
the prior written permission of the publisher.

While the authors and the publishers believe that the information and
guidance given in this work are correct, all parties must rely upon their Volume 1
own skill and judgment when making use of them. Neither the authors nor
the publishers assume any liability to anyone for any loss or damage Foreword xvii
caused by any error or omission in the work, whether such error or Preface xix
omission is the result of negligence o: any other cause. Any and aii such List of contributors xxi
liability is disclaimed. Introduction - J.R. James and P.S. Hall 1
1

1.I Historical development and future prospects

1.2 Fundamental issues and design challenges
1.2.1 Features of microstrip antenna technology
1.2.2 Fundamental problems
1.3 The handbook and advances presented
1.4 Glossary of printed antenna types
1.5 Summary comments
British Library Cataloguing i n Publication Data 1.6 References
Handbook of Microstrip Antennas 2 Analysis of circular microstrip antennas - L. Shafai and A.A. Kishk
1. Microwave equipment: Microstrip antennas
I. James, J. R. (James Roderick, 1933- 2.1 Introduction
II. Hall, P. S. (Peter S) Ill. Institution of Electrical 2.2 Formulation of the problem
Engineers IV. Series 2.2.1 Matrix formulation
621.381'33 2.2.2 Excitation matrix
2.2.3 Radiation fields
ISBN 0 86341 150 9 2.3 Application I: Circular patch antenna
2.3.1 Surface fields
2.3.2 Feed location
2.3.3 Effect of the substrate permittivity
2.3.4 Effect of the substrate thickness
2.3.5 Effect of the ground-plane radius
2.3.6 Effect of the ground-plane thickness
2.3.7 Circular polarisation
2.3.8 Effect of a central shorting pin
2.4 Application 2: Wraparound microstrip antenna
2.5 Application 3: Reflector antenna feeds
2.6 Concluding remarks
2.7 References
Printed in England by Short Run Press Ltd., Exeter
vi Contents Contents vii

3 Characteristics of microstrip patch antennas and some methods of 5 Microstrip dipoles - P.B. Katehi, D.R. Jackson and N.G. Alexopoulis
improving frequency agility and bandwidth - K.F. Lee and
J.S. Dahele Introduction
Infinitesimal dipole
Introduction 5.2.1 Analysis
Cavity model for analysing microstrip patch antennas 5.2.2 Substrate effects
3.2.1 lntroduction 5.2.3 Superstrate effects
3.2.2 Feed modelling, resonant frequencies and internal Moment-method techniques for planar strip geometries
fields 5.3.1 Basis functions
3.2.3 Radiation field 5.3.2 Reaction between basis functions
3.2.4 Losses in the cavity 5.3.3 Plane-wave-spectrum method
3.2.5 Input impedance 5.3.4 Real-space integration method
3.2.6 VSWR bandwidth 5.3.5 Point-dipole approximation
3.2.7 Qualitative description of the results predicted by 5.3.6 Moment-method equations
the model Centre-fed dipoles
Basic characteristics of some common patches 5.4.1 Single dipole
3.3.1 The rectangular patch 5.4.2 Mutual impedance
3.3.2 The circular patch EMC dipoles
3.3.3 The equitriangular patch 5.5.1 Methods of analysis
3.3.4 Annuiar-ring patch 5.5.2 Single dipole
3.3.5 Comparison of characteristics of rectangular, 5.5.3 Multiple dipoles
circular, equitriangular and annular-ring patches Finite array of EMC dipoles
3.3.6 Brief mention of other patches 5.6.1 Analysis
Some methods of improving the frequency agility and 5.6.2 Calculation of coefficients
bandwidth of microstrip patch antennas 5.6.3 Array design
3.4.1 Introduction Conclusions
3.4.2 Some methods of tuning MPAs References
3.4.3 Dual-band structures
3.4.4 Electromagnetic-coupled patch antenna (EMCP) 6 Multilayer and parasitic configurations - D.H. Schaubert
Summary
Acknowledgments 6.1 Introduction
References 6.2 Stacked elements for dual-frequency or dual polarisation
operation
4 Circular polarisation and bandwidth - M. Haneishi and Y. Suzuki 6.2.1 Antennas with separate feeds for each function
6.2.2 Antennas for multiple frequencies and increased
Various types of circularly polarised antenna bandwidth
4.1.1 Microstrip patch antennas 6.3 Two-sided aperture-coupled patch
4.1.2 Other types of circularly polarised printed antennas 6.4 Parasitic elements on antenna substrate
Simple design techniques for singly-fed circularly polarised 6.5 Summary
microstrip antennas 6.6 References
4.2.1 Rectangular type
4.2.2 Circular type Wideband flat dipole and short-circuit microstrip patch elements and
7
More exact treatment for singly-fed circularly polarised micro- arrays - G. Dubost
strip antennas
4.3.1 Analysis Flat dipole elements and arrays
7.1
4.3.2 Conditions for circularly polarised radiation 7.1.1 Elementary sources
4.3.3 Example 7.1.2 Array designs: losses and efficiencies
Some considerations on mutual coupling Short-circuit microstrip patches and arrays
7.2
Wideband techniques 7.2.1 Elementary source
4.5.1 Design of wideband element 7.2.2 Array designs
4.5.2 Technique using parasitic element 7.3 References
4.5.3 Technique using paired element
References
viii Contents Contents ix
8 Numerical analysis of microstrip patch antennas - J.R. Mosig, Edge-admittance and mutual-coupling networks
R.C. Hall and F.E. Gardiol 9.4.1 Edge-admittance networks
9.4.2 Mutual-coupling network
Introduction Analysis of multiport-network model
8.1.1 General description 9.5.1 Segmentation method
8.1.2 The integral equation model 9.5.2 Desegmentation method
Model based on the electric surface current Examples of microstrip antenna structures analysed by multiport-
8.2.1 Geometry of the model and boundary conditions network approach
8.2.2 Potentials for the diffracted fields 9.6.1 Circularly polarised microstrip patches
8.2.3 Green's functions 9.6.2 Broadband multiresonator microstrip antennas
8.2.4 Mixed potential integral equation (MPIE) 9.6.3 Multiport microstrip patches and series-fed arrays
8.2.5 Sketch of the proposed technique C A D of microstrip patch antennas and arrays
Horizontal electric dipole (HED) in microstrip Appendix: Green's functions for various planar configurations
8.3.1 The vector potential Acknowledgments
8.3.2 Scalar potential and the fields References
8.3.3 Surface waves and spectral plane k
8.3.4 Far-field approximations 10 Transmission-line model for rectangular microstrip antennas
8.3.5 Radiation resistance and antenna efficiency - A. Van rle Capelle
Numerical techniques for Sommerfeld integrals
8.4.1 Numerical integration oii the real axis Introduction
8.4.2 Integrating oscillating functions over unbounded Simple transmission-line model
intervals 10.2.1 Description of the transmission line model
Construction of the Green's functions 10.2.2 Expressions for G, and B,
Method of moments 10.2.3 Expressions for the line parameters
8.6.1 Rooftop (subsectional) - basis functions Improved transmission-line model
8.6.2 Entire domain basis functions 10.3.1 Description of the improved transmission-line model
Excitation and loading 10.3.2 Expression for the self-susceptance B,
8.7.1 Several microstrip-antenna excitations 10.3.3 Expression for the self-conductance G,
8.7.2 Coaxial excitation and input impedance 10.3.4 Expression for the mutual conductance G,
8.7.3 Multiport analysis 10.3.5 Expression for the mutual susceptance B,
Single rectangular patch antenna 10.3.6 Expressions for the line parameters
8.8.1 Entire-domain versus subdomain basis functions Application of the improved transmission-line model
8.8.2 Convergence using subsectional basis functions 10.4.1 Analysis and design of rectangular microstrip antennas
8.8.3 Surface currents 10.4.2 Comparison with other methods
Microstrip arrays 10.4.3 Comparison with experimental results
8.9.1 Array modelling 10.4.4 Design application
8.9.2 Mutual coupling Transmission-line model for mutual coupling
8.9.3 Linear array of few patches 10.5.1 Description of the model
Acknowledgments 10.5.2 Calculation of the model parameters
References 10.5.3 Comparison with other methods
Acknowledgements
9 Multiport network approach for modelling and analysis of microstrip patch References
antennas and arrays - K.C. Gupta 455
11 Design and technology of low-cost printed antennas - J.P. Daniel,
9.1 Introduction E. Penard and C. Terret
9.2 Models for microstrip antennas
9.2.1 Transmission-line model 11.1 Introduction
9.2.2 Cavity model 11.2 Analysis of simple patches and slots
9.2.3 Multiport network model 11.2.1 Rectangular and circular patches
9.3 2-matrix characterisation of planar segments 11.2.2 Conical antennas
9.3.1 Green's functions 11.2.3 Linear and annular slots
9.3.2 Evaluation of 2-matrix from Green's functions 11.3 Design of planar printed arrays
9.3.3 2-matrices for segments of arbitrary shape 1 1.3.1 Design parameters
x Contents Contents xi

11.3.2
11.3.3
Cavity model analysis of mutual coupling
Linear series array of corner-fed square patches
Volume 2
113.4 Two-dimensional cross-fed arrays
11.4 Synthesis methods for linear arrays Microstrip antenna feeds - R.P. Owens
14
11.4.1 Relaxation methods
11.4.2 Simplex method 14.1 Introduction
11.4.3 Experimental results
11.5 New low-cost low-loss substrate 14.2 Coupling to microstrip patches
14.2.1 Co-planar coupling to a single patch
11.5.1 Substrate choice
14.2.2 Series-array co-planar coupling
11.5.2 Fabrication procedure 14.2.3 Probe coupling
11.5.3 Electrical characteristics 14.2.4 Aperture coupling
11.5.4 Environmental tests
14.2.5 Electromagnetic coupling
11.5.5 Examples of printed antennas on polypropylene 14.3 Parallel and series feed systems
substrate 14.3.1 Parallel feeds for one and two dimensions
11.6 Concluding remarks
14.3.2 Series feed for one dimension
11.7 References 14.3.3 Combined feeds
14.3.4 Discontinuity arrays
12 Analysis and design considerations for printed phased-array antennas 14.4 Direct-coupled stripline power dividers and combiners
- D.M. Pozar 4 . 4 Simple three-port power dividers
14.4.2 Isolated power dividers/combiners
12.1 Introduction 14.4.3 Four-port direct-coupled power dividers
12.2 Analysis of some canonical printed phased-array geometries 14.5 Other feed systems
12.2.1 Some preliminaries 14.5.1 Alternative transmission tines
12.2.2 Infinite-planar-array solutions 14.5.2 Multiple beam-forming networks
12.2.3 Finite-array solutions 14.6 Acknowledgments
12.3 Design considerations for printed phased arrays 14.7 References
12.3.1 Introduction
12.3.2 Array architectures
12.4 Conclusion 15 Advances in substrate technology - G.R. Traut
12.5 Acknowledgments
12.6 References 15.1 Considerations for substrate selection
15.1.1 Impact of properties of various substrate systems on
13 Circularly polarised antenna arrays - K. Ito, T. Teshirogi microstrip antenna performance
15.1.2 Comparative list of available substrates
and S. Nishimura 15.1.3 Selection of metal cladding for performance
15.1.4 Thermal characteristics of PTFE
13.1 Various types of circularly polarised arrays 15.1.5 Anisotropy of relative permittivity
13.1.1 Arrays of patch radiators 15.2 Measurement of substrate properties
13.1.2 Arrays of composite elements 15.2.1 Stripline-resonator test method
13.1.3 Travelling-wave arrays 15.2.2 Microstrip-resonator test method
13.1.4 Other types of arrays 15.2.3 Full-sheet-resonance test method
13.2 Design of circularly polarised arrays 15.2.4 Perturbation cavity method
13.2.1 Arrays of patch radiators 15.2.5 Tabulated evaluation of methods for measuring
13.2.2 Arrays of composite elements relative permittivity and dissipation factor
13.2.3 Design of travelling-wave arrays
15.3 Processing laminates into antennas
13.3 Practical design problems
13.3.1 Mutual coupling
15.3.1
15.3.2
-
Handline incoming copper-clad laminates
Handling prior to processing
13.3.2 Unwanted radiation
15.3.3 Safetv considerations for PTFE-based substrates
13.3.3 Limitations and trade-offs
13.3.4 Non-planar scanning arrays 15.3.4 ~ e d i c i the
n ~ effects of etch strain relief
13.4 Wideband circularly polarised arrays 15.3.5 Machining of PTFE-based boards
13.4.1 Arrays of wideband elements 15.3.6 Bending etched antenna boards
13.4.2 15.3.7 Bonded-board assemblies
Arrays of dual-frequency stacked elements 15.3.8 Plating-through holes in' microstrip antenna boards
13.4.3 Wideband-array techniques
13.5 References
xii Contents Contents xiii

15.3.9 Device attachment on microstrip antenna substrates 17.2.12 Attenuation

15.4 Design considerations with selected materials 17.2.13 Higher-order modes and radiation
15.4.1 Environmental effects o n antenna substrates 17.3 Discontinuities: bends and junctions
15.4.2 Conductor losses at millimetre-wave frequencies 17.3.1 Definition
15.4.3 Multilayer circuit-board technology in microstrip 17.3.2 Models
antennas 17.3.3 TEM-line models
15.5 Special features and new materials developments 17.3.4 Variational techniques
15.5.1 Substrates clad on one side with thick metal 17.3.5 Fourier transform
15.5.2 Low thermal coefficient of K' in fluoropolymer 17.3.6 Dielectric Green's function
laminates 17.3.7 Integral equations for inductances
15.5.3 Microwave laminates with a resistive layer 17.3.8 Green's function and integral equation
15.5.4 Thermoset microwave materials 17.3.9 Green's function and electrostatic-inductance
15.5.5 Low permittivity ceramic-PTFE laminates computation
15.5.6 Very-low-dielectric-constant substrates 17.3.10 TLM (transmission-line-matrix) method
15.6 References 17.3.11 Waveguide model
17.4 Technological realisation: Materials and manufacturing process
16 Special measurement techniques for printed antennas - E. Levine 17.4.1 Introduction
17.4.2 Dielectric substrate
Introduction i7.4.3 Comment
Substrate properties 17.4.4 Inorganic substrates
Connector characterisation 17.4.5 Plastic substrates
Measurements of printed lines and networks 17.4.6 Semiconductor substrates
16.4.1 Measurement of printed-line parameters 17.4.7 Ferrimagnetic substrates
16.4.2 Measurement of printed networks 17.4.8 Metallisation
Near-field probing 17.4.9 Circuit realisation
Efficiency measurement 17.4.10 Etching
Concluding remarks 17.4.11 Metal deposition
References 17.4.12 Removal of photoresist
17.4.13 Under-etching
17 Computer-aided design of microstrip and triplate circuits - J.F. Ziircher 17.4.14 Thin and thick film
and F.E. Gardiol 17.5 Analysis and synthesis programs
17.5.1 Introduction
17.1 Introduction, definition of the structure EEsof: Touchstone
17.1.1 Outline CCC: The Supercompact Family
17.1.2 Microwaves CCC: CADEC +
17.1.3 Transmission lines for microwaves Acline
17.1.4 Balanced stripline or triplate Thorn '6: Esope
17.1.5 Microstrip RCA: Midas
17.1.6 Adjustments LINMIC
17.1.7 Multiple inhomogeneity High Tech. Tournesol: Micpatch
17.1.8 Measurement problems Spefco Software: CiAO
17.2 Basic relationships for uniform lines Made-it-associates: Mama
17.2.1 Uniform lines Ampsa: Multimatch
17.2.2 Conformal mapping Radar systems technology: Analop
17.2.3 Schwartz-Christoffel transform Microkop/Suspend
17.2.4 Zero-thickness balanced stripline -
Microwave software aoolications
A

17.2.5 Finite-thickness balanced stripline Planim

17.2.6 Equivalent homogeneous microstrip line DGS Associates: S/Filsyn
17.2.7 Characteristic impedance of microstrip Webb Laboratories: Transcad
17.2.8 Finite-thickness homogeneous microstrip 17.6 Layouts of circuits and cutting of masks
17.2.9 Microstrip-line synthesis for b = 0 17.6.1 Description
17.2.10 Dispersion in microstrip 17.6.2 CCC: Autoart
17.2.11 Effect of an enclosure 17.6.3 EFSOF: Micad
xiv Contents Contents xv

17.6.4 High Tech. Tournesol: Micros 20.2.3 Feeding the patch 1155
17.6.5 British Telecom: Temcad 20.2.4 Theoretical design method 1155
17.7 Insertion of components 20.2.5 Patch design 1158
17.7.1 Introduction Dual patch element 1161
17.7.2 Discrete components 20.3.1 Choice of design 1161
17.7.3 Mounting procedure 20.3.2 Location of patch phase centre 1161
17.7.4 Drilling holes in the dielectric substrate 20.3.2 Design and optimisation 1162
17.7.5 Deposited components Hybrid feeding network 1163
17.8 Examples 20.4.1 Overview 1163
17.8.1 Design of a broadband amplifier 20.4.2 Hybrid designs 1166
17.8.2 Bandpass filter design 20.4.3 90' bends 1168
17.8.3 Design of a miniature Doppler radar 20.4.4 Minimum track distance 1168
17.9 Conclusions 20.4.5 Feed-point terminations 1171
17.10 Acknowledgments 20.4.6 Track lengths 1171
17.11 References 20.4.7 Overall design 1172
Conical antenna array 1172
Substrate fabrication 1175
18 Resonant microstrip antenna elements and arrays for aerospace 1175
applications - A.G. Derneryd 20.6.1 Overview
20.6.2 Mask drawing and preparation 1175
20.6.3 Etching 1176
18.1 Introduction 20.6.4 Substrate preparation 1176
18 2 Circular antenna element 20.6.5 Triplate bonding 1177
18.3 Dual-band circularly polarised antenna element Forming the antenna 1177
18.4 Monopulse-array antenna 20.7.1 Bending the substrates 1177
18.5 Dual-polarised-array antenna 20.7.2 Attachment of components 1178
18.6 Concluding remarks 20.7.3 Final assembly 1181
18.7 References Antenna performance 1181
20.8.1 Grating-lobe suppression 1182
19 Applications in mobile and satellite systems -K. Fujimoto, T. Hori, 20.8.2 Axial ratio 1185
S. Nishimura and K. Hirasawa 20.8.3 Antenna gain 1187
20.8.4 Tracking slope 1188
Introduction Conclusions and future developments 1188
Mobile systems References 1191
19.2.1 Design considerations
19.2.2 Base stations
19.2.3 Wheeled vehicles 21 Microstrip field diagnostics - P.G.Frayne
19.2.4 Railways
19.2.5 Pedestrian Introduction
19.2.6 Radars Surface analytical techniques
Satellite system Scanning-network probe
19.3.1 Design considerations Theory of the monopole probe
19.3.2 Direct broadcasting reception Resonant microstrip discs
19.3.3 Earth stations Resonant microstrip triangles
19.3.4 Satellite borne Open-circuited microstriplines
References Antenna diagnostics
21.8.1 The rectangular patch
20 Conical conformal microstrip tracking antenna - P. Newham Linear element patch array
and G. Morris Circularly polarised patch antenna
Microstrip travelling-wave antenna
20.1 Introduction Acknowledgments
20.2 Single patch element References
20.2.1 Choice of array element
20.2.2 Choice of substrate
xvi Contents
22 Microstrip antennas on a cylindrical surface - E.V. Sohtell
22.1 Introduction
22.2 Theoretical models for a patch on a cylinder
22.2.1 Cavity model of the patch
22.2.2 Surface-currentmodel
22.3 Single patch application
22.3.1 Mechanical design
22.3.2 Measurements Foreword
22.3.3 Radiation-pattern comparisons
22.4 Array application
22.4.1 General
22.4.2 Theoretical treatment of finite and infinite arrays
22.4.3 Design of a phased array on C-band
22.4.4 Measured performance
22.5 Summary
22.6 References The Handbook of Microstrip Antennas could not have been written even five
years ago, for neither the technology nor the relevant analytical tools were
23 Extensions and variations to tho microstrip antenna concept P.S. Hall, sufficientlydeveloped. This text arrives when the field is at a rush of activity.
A. Henderson and J.R. James 1257 Fundamental mathematical tools are on hand to solve a variety of the important
problems, and practical engineering results are now finding applications. Poten-
Introduction tial future capabilities and applications now look more optimistic than at any
Radiation pattern control
23.2.1 Reflector feeds time in the history of this young technology. This new text describes vast
23.2.2 Spherical dielectric overlays developments in theory and practice. In two volumes, and representing the work
Wide-bandwidth techniques of over thirty authors, the text is presented with such authority that it is assured
23.3.1 Log-periodic structures a role as a key reference tool for many years.
23.3.2 Dichroic dual-function apertures Microstrip antennas are a new and exciting technology. Invented about
Millimetre-wave hybrid antenna
Novel use of materials twenty years ago for application as conformal antennas on missiles and aircraft,
23.5.1 Foam substrates for large direct-broadcast-satellite the microstrip antenna has found increasing use because it can be fabricated by
domestic receiving arrays 1288 lithographic techniques in monolithic circuits. Initially, microstrip patch anten-
23.5.2 Magnetic materials and beam scanning 1292 nas were used as individual radiators, but they soon found use in relatively large
23.5.3 Use of very-high-permittivity substrates in hyperthermia fixed beam (non scanning) arrays. More recently, they have progressed to arrays
applicators 1293
Summary comment 1294 for scanning in one or two dimensions. The advantage of this technology at
References 1295 microwave frequencies is its compatability with large scale printed circuit fab-
rication. Boards are fabricated lithographically and devices mounted by robot-
ics or automated production line techniques. Microstrip printed circuit arrays
are seen as an essential key to affordable antenna technology.
At millimeter wavelengths, the benefit of microstrip arrays are enormous and
so revolutionary as to create an entirely new technology; the monolithic inte-
grated antenna array. Such an array has transmission lines, amplifiers, phase
shifters and radiating elements, all on semiconductor substrates. Beyond this,
these monolithic subarrays will be compatible with the integration of various
solid state technologies on wafer size substrates. At these integration levels, the
antenna array design and monolithic integrated circuit design cannot be
separated, for the antenna architecture will need to optimise radiation, solid
state device integration, board layout and thermal design. And so is born the
antenna system architect!
xviii Foreword

Against this backdrop of energy and creativity, this timely and important
book is the first handbook entirely dedicated to presenting a detailed overview
of microstrip antenna development and theory. The vast scope of the text does
justice to the broad range of research and development being undertaken
throughout the world that is addressing a wide variety of microstrip elements
and arrays for radiating linearly and circularly polarised waves. The text
presents the work of a number of the most prominent and knowledgeable
authors and so documents the state of the art at many institutions and in several Preface
countries.
This monumental handbook is a milestone in the development of microstrip
antenna technology.
Robert J. Mailloux

Within two decades Microstrip Antennas have evolved as a major innovative

activity within the antenna field and for both of us it has indeed been a
fascinating and challenging experience to play a part in this vibrant research. In
so doing the opportunity to initiate this International Handbook has arisen and
this again has been a stimulating, meaningful objective that has also enriched
our personal experiences through contact with numerous colleagues worldwide.
It was around 1985 when it was apparent to us that the topic had raced ahead
so fast that our previous IEE book "Microstrip Antenna Theory and Design"
published in 1981 would soon need up-dating. Such is the vigour in Microstrip
Antenna research that neither of us felt that we could do justice to the topic, at
least across all its frontiers in a reasonable time scale, and it was at this point
that we conferred with colleagues worldwide and this multiauthored Handbook
was conceived.
As to the subject itself, it has been abundantly clear for years that it is system
driven and indeed continues to be so, and that its alarming pace has promoted
microstrip antennas from the ranks of a rather specialised technique to a major
type of antenna technology in itself. Historically one has always associated low
cost, low weight and low profile with Microstrip Antennas but this description
is simplistic and inadequate in the industrial atmosphere today where many new
systems owe their existence to these new radiators. In reality, the feasibility of
a low profile printed radiator has inspired the system creators and there is an
abundance of examples, not just in the Defence sector. For instance, we have
new generations of printed paper antennas, adaptive conformal antennas sitting
on the roofs of automobiles and printed antennas as true ground speed sensors
in many transport scenarios.
It is indeed a stimulating topic to be associated with and we hope that the
Handbook will portray this. For the in-depth researcher, however, the frontiers
to push forward carry the familiar headings of bandwidth extension techniques,
pattern control, minimisation of losses etc. but the scene has moved on in a
decade and industry is now thirsty for significant advances, all at low cost, to
meet the demand for higher performance and competitive costs. Research thus
xx Preface

addresses critical optimisation procedures and advances are hard won. The role
of substrate technology is now well appreciated and major developments have
taken place to design materials that withstand a wide range of operating
constraints, yet are affordable. As to the main thrust in research, it centres
around the continual quest for innovative electromagnetic printed structures
that satisfy the expanding system demands coupled with the ability to manufac-
ture them and it is in the latter area where computer aided design (CAD) forms
the cutting edge. Whether the manufacture of microstrip arrays can be fully List of contributors
automated via CAD in the immediate future is an open question that echoes
throughout the Handbook and at present, further research is necessary.
In organising the Handbook we have attempted to address all these aspects
giving a balanced viewpoint from both industry and research centres and the
overlap between chapters is intended to be sufficient to allow meaningful com-
parisons between contributors to be made. The broad theme adopted is to take N. G. Alexopoulos F. E. Gardiol
the reader through elements and arrays in the first volume followed by technol- University of California Ecole Polytechnique FCdkrale de
ogy and applications in the second volume but as may be expected, many USA Lausanne
authors include material covering more than one aspect. Look-up charts relat- Switzerland
ing items of interest to chapters and a Glossary of over one hundred different A. R. Van de Capelle
types of printed antennas form much of the Introduction to assist the reader to Katholieke Universiteit Leuven K. C. Gupta
efficiently select those parts that are of immediate interest. Belgium University of Colorado
Finally, we thank all authors for their creative contributions, splendid USA
cooperation, careful preparation of manuscripts and fellowship in the collective J. S. Dahele
aim to compile a worthy international text with many years9 useful life. In Royal Military College of Science
particular we thank Dr David Pozar and Dr Koichi Ito who helped us initially UK P. S. Hall
with communications in the USA and Japan respectively. We are also pleased Royal Military College of Science
to acknowledge the willing and professional cooperation of the publishers. J. P. Daniel UK
On a personal note, we have enjoyed the project and in particular the sincere UniversitC de Rennes I
experience of making new friends and acquaintances worldwide. France R. C. Hall
J. R. James Ecole Polytechnique FCdCrale de
P. S. Hall A. G. Derneryd Lausanne
Ericsson Radar Electronics Lab Switzerland
Sweden
M. Haneishi
G. Dubost Saitama University
UniversitC de Rennes I Japan
France

P. G. Frayne A. Henderson
University of London Royal Military College of Science
UK UK

K. Fujimoto K. Hirasawa
University of Tsukuba University of Tsukuba
Japan Japan
 List of contributors xxiii
xxii List of contributors
T. Hori S. Nihimura G. R. Traut J. E Zurcher
Nippon Telegraph and Telephone University of Osaka Rogers Corporation Ecole Polytechnique Fkdkrale de
Corporation Japan USA Lausanne
Japan Switzerland
R. P. Owens
K. It0 Thorn EM1 Electronics Ltd
Chiba University UK
Japan
E. Penard
D. R. Jackson Centre National D'Etudes de
University of Houston Elhmmunications
USA France
J. R. James
Royal Military College of Science D. M. Pozar
UK University of Massachusetts
USA
P. B. Katehi
University of Michigan D. H. Schaubert
USA University of Massachusetts
USA
A. H. Kishk
University of Mississippi L. Shafai
USA University of Manitoba
Canada
K. F. Lee
University of Toledo E. V. SohteU
USA Ericsson Radar Electronics Lab
Sweden
E. Levine
Weizmann Institute of Science
Israel Y. Suzuki
Toshiba Corporation
G. Moms Japan
Vega Cantley Instrument Co Ltd
UK C. Terret
Centre National #Etudes de
J. R. Mosig Elhmmunications
Ecole Polytechnique Fkdkrale de France
Lausanne
Switzerland T. Teshirogi
Radio Research Laboratories
P. Newham Ministry of Posts and
Marconi Defence Systems Telecommunications
UK Japan
 Chapter 1

Introduction
J.R. James
and P.S. Hall

1.1 Historical development and future prospects

The microstrip antenna is now an established type of antenna that is confidently

prescribed by designers worldwide, particularly when low-profile radiators are
demanded. The microstrip, or printed, antenna has now reached an age of
maturity where many well tried techniques can be relied upon and there are few
mysteries about its behaviour. The fact that you are now reading an historical
review is interesting in itself because all this has happened in a relatively short
time span of one or two decades; such is the rate of progress in contemporary
antenna technology. To imply that the topic of microstrip antennas is now static
would be grossly misleading because the opposite is true with the ever increasing
output of research publications and intensifying industrial R and D. The quest
now is for more and more innovative designs coupled with reliable manufactur-
ing methods. The driving force is the thirst for lower-cost, less-weight, lower-
profile antennas for modern system requirements. Lower costs, however, rely on
the ability of the designer to precisely control the manufacturing process, and
this in turn usually demands that the prototype innovative structures can be
adequately mathematically modelled and toleranced. It is in these latter respects
that the challenge to the antenna expert originates, and the search for the more
precise computer modelling of microstrip antennas is now the main preoccu-
pation of designers and researchers alike, as is reflected in this handbook.
The invention of the microstrip-antenna concept has been attributed to many
sources and the earliest include Greig and Englemann [l] and Deschamp [2]. At
that time the emission of unwanted radiation from the then new thin stripline
circuits was well appreciated and subsequently the dimensions of the substrate
and conducting strip were reduced to inhibit the radiation effects, thus creating
'microstrip'. Whether the advent of the transistor influenced the rapid develop-
ment of these planar printed circuits is debatable and the main interest was likely
to be the development of lower-cost microwave filters etc. Lewin [3] considered
2 Introduction
the nature of the radiation from stripline but there was apparently little or no
interest in making use of the radiation loss. Apart from a few references [4, 5,
61 the antenna concept lay dormant until the early 1970s [7,8,9] when there was
an immediate need for low-profile antennas on the emerging new generation of
missiles.
At this point in time, around 1970, the development of the microstrip-antenna
concept started with earnest and the research publications, too numerous to
itemise, started to flow. The period is perhaps most readily referenced by its
workshops and major works. The most significant early workshop was held at
Las Cruces, New Mexico, in 1979 [lo] and its proceedings were distilled into a
major IEEE Transactions special edition [I I]. At that time two books were
published by Bahl and Bhartia [I21 and James, Hall and Wood [I31 which
remain in current use today. Another more specialised and innovative develop-
ment was published as a research monograph by Dubost [14], and here the
flat-plate antenna was approached from the standpoint of flat dipoles on sub-
siraies that generally only partially filled ihe available .;o!ume.
The early 1980s were not only a focal point in publications but also a
milestone in practical realism and ultimately manufacture. Substrate manufac-
turers tightened their specifications and offered wider ranges of products cap-
able of working under extreme ambient conditions. Substrate costs were, how-
ever, to remain high. It was appreciated that analytical techniques for patch
elements generally fell short of predicting the fine pattern detail of practical
interest and the input-impedance characteristic to suficient accuracy. It was
also appreciated that the connection of feeders to patch elements in a large array
was fraught with problems and new approaches were necesary where the feeders
and elements are regarded as a complete entity. More recently the term 'array
architecture' has come into being as if to emphasise the importance of choice of
array topology and the fact that feeders cannot necessarily be freely attached to
printed elements, even if the latter are in themselves well optimised.
Recent system demands are, as previously mentioned, a dominant factor in
the development of printed antennas. Communication systems spanning wider
bandwidths are continually emerging and techniques for increasing the band-
width of microstrip antennas are a growth area. Controlling the polarisation
properties of printed antennas is another area of activity arising largely out of
the current awareness for making greater use of the polarisation properties of
waves, particularly in radar. In defence applications, systems that have an
electronic, as opposed to mechanical, beam-scanning facility are attracting
much research effort and the concept of 'active-array architecture' is now with
us where semiconductor packages and radiating elements are integrated into
planar apertures. The cost of such an array is very high and the whole concept
is state-of-the-art.
This brings us to the present and how we see the immediate future of printed
antennas. A seldom mentioned point is the fact that printed substrate tech-
nology is readily processed in University laboratories and continues to remain
Introduction 3
a rich source of complex electromagnetic problems; research publications will
thus continue to abound, and in parallel with industrial development will most
likely be dominated by two aspects:
The search for mathematical models that will predict practical antennas more
precisely and hence sharpen CAD techniques in manufacture.
The creation of innovative antennas to match the demand for new systems.
In this latter aspect it must be emphasised that a bulky conventional microwave
antenna may well out-perform its thin conformal printed counterpart. Many
new systems, however, particularly in aerospace, are only made feasible with the
existence of the printed antenna concept, and here lies a major driving force
where new systems arise solely from innovative antenna designs.
As to the distant future, one can but extrapolate the present trends towards
integrated electronically beam-scanned arrays. This leads to a vision of confor-
mal antennas distributed over the surface of vehicles, aircraft, ships, missiles
etc., thus replaciiig iiiary convcntional types of iadiatois, but the orgafiisatio::
and control of the radiation pattern co- and cross-polar characteristics is a
complex control problem that cannot be solved by software alone and demands
innovative physical concepts. Are we thus unconciously converging on the
concept of distributed sensors, so common in the insect and animal world, where
information is commonly gleaned in a variety of ways to best suit a particular
situation? Taking the comparison a step forward, we would therefore expect the
distributed conformal apertures to require a significant back-up from signal-
processing techniques, which amount to making use of temporal a priori in-
formation on signals and noise. Put this way these ideas are not so far-reaching
because many of these adaptive concepts can be recognised in some of our new
radar and communication systems, particularly for defence. In this light the
printed-antenna concept would therefore appear as a gateway to system com-
patibility and optimal deployment of sensors, embracing the numerous facets of
conformality, low costs, semiconductor integration, electronic radiation pattern
control and an opportunity to exploit signal-processing techniques to the full
using modern computing power. The prospects are indeed exciting and under-
line the importance of the microstrip-antenna concept, its continual evolution
and impact on electronic systems design.
1.2 Fundamental issues and design challenges
A handbook of this type is intended as an all-embracing treatment that is both
diverse and highly specialist. As such it is not possible to include comprehensive
background information and we anticipate that readers wishing to recap on
basic antenna theory, antenna mesurements and the rudiments of microstrip
technology etc. will have no difficulty in obtaining relevant literature. It is our
experience, however, that certain fundamental properties of printed antennas
4 lntroduction lntroduction 5

have been central to their evolution and limitations, and therefore embody the Table 1.1 Antenna desi.qners' checklist of performance factors
design challenges of the future as follows. Matching Input terminals matched to source feed
The microstrip antenna has many differences when compared with a conven-
tional antenna. Most of these stem from the planar construction in which for a Main beam Antenna gain and beamwidth properties
given substrate in the .uy plane there are only two degrees of freedom, allowing Sidelobes Constrained to desired envelope
the very thin printed-conductor topology to take any shape within the confines
of the .u and y co-ordinate directions. The first and most troublesome property Polarisation Cross-polar behaviour constrained to
is the issue of loss, principally in the thin conducting strip feeders connecting desired envelope
elements in large arrays. In some applications the loss in the radiating elements Circular Constraints on ellipticity
also creates dificulties. The radiating elements themselves have a restricted polarisation
bandwidth arising from the intrinsic high-Q resonator action in the thin sub-
strate. The generation of surface waves is equally important and cannot be Eficiency Wastage of power in antenna structure
avoided unless foam-type substrates are deployed allowing virtual air-spaced Aperture Relates to illumination distribution,
operation. The surface waves can corrupt radiation-pattern characteristics, eficiency gain and pattern characteristics
particularly when low sidelobe and cross-polarisation levels are demanded. In
many design specifications. problems can only be alleviated by compromising Bandwidth Frequency range over which all above
the manufacturing simplicity of the single coplanar printed assembly by employ- parameters satisfy specification -
ing overlaid element and feed concepts based on multilayer sandwich structures. commonly based on input terminal
Microstrip arrays generally require some sort of radome or weather shield, thus impedance charactericstics
increasing the structure depth, but in some cases a degree of radiation-pattern System Size, weight, cost
enhancement is obtainable. Last but not least, mention must be made of the demands
relatively high cost of substrates capable of providing the desired electrical and
mechanical stability in operation. The substrate cost is often an inhibiting factor
in what is otherwise a low-cost manufacturing process. The commonly upheld properties of microstrip antennas are listed in Table
These above issues are of a fundamental nature and we consider it important 1.3 and may be usefully compared with the general checklist of Tables 1.1 and
to highlight current understanding to identify aspects which may offer particular 1.2 to ascertain the suitability of microstrip for various operational roles.
scope for future advancement. Before addressing this we list, for completion, However, it is important to appreciate that the interpretation of Table 1.3 is very
some of the more commonly known properties of microstrip antennas in rela- dependent on the intended application. For instance, patch antennas on foam
tion to both contemporary antenna-engineering and modern electronic-systems
Table 1.2 O~erationaland manufacturing considerations
requirements.
Noise effects in receiving antennas
1.2.1 Features of microstrip antenna technology Power handling in transmitting antennas
The microstrip antenna is a newcomer to the world of antenna engineering and Creation of hazards for personnel in near-field
it is fitting to be reminded of features generally sought after when compiling an Robustness to lightning strikes
antenna specification. A typical checklist is given in Table 1.1 and it is appre- Electrostatic charge effects in space applications
ciated that it is unlikely that all the performance factors are relevant or indeed Effects of wind, vibration, ice, snow, rain, hail
critical in any given application. Equally demanding are operational and manu- Ambient conditions on temperature and humidity
facturing considerations such as those listed in Table 1.2 and these are very Exposure to sunlight
dependent on the application in mind. The generation of thermal noise in a Aerodynamic constraints, radomes and weather shields
receiving antenna is insignificant for most conventional antennas and is clearly Metal corrosion and creep
a new factor associated mainly with large lossy microstrip arrays. Likewise Mechanical and electrical stability of materials
power-handling and material effects are particularly relevant for microstrip Mechanical and electrical tolerances in manufacture
radiators, while the use of new materials such as carbon fibre necessitates careful Sensitivitiy of design to manufacturing tolerances
evaluation of electrical loading, intermodulation effects etc. Generation of intermodulation effects in materials
 6 introduction lntroduction 7
Table 1.3 Some commonly acknowledged properties of microstrip anten- Table 1 . 4 ~ Approximate performance trade-offs for a rectangular patch
nas
Requirement Substrate Substrate Patch
Advantages Disadvantages height relative width
Thin profile Low efficiency nermittivitv
Light weight Small bandwidth High radiation efficiency thick low wide
Low dielectric loss thin low -
Simple to manufacture Extraneous radiation from feeds, junctions Low conductor loss thick - -
and surface waves Wide (impedance) bandwidth thick low wide
Low extraneous (surface thin low -
Can be made conformal Tolerance problems
wave) radiation
Low cost Require quality substrate and good Low cross polarisation - low -
temperature tolerance Light weight thin low -
Strong thick high -
Can be integrated with High-performance arrays require complex
circuits feed systems Low sensitivity to thick low wide
tolerances
Simp!e arrays readily Polarisation piiriiy difficuit ro achieve
created Table 1.46 Approximate performance trade-offs for an array of circular
patches
substrates may have a less desirable thick profile but good efficiency and Requirement Substrate Substrate relative
reasonable bandwidth; in contrast a thin overlaid patch assembly with complex height permittivity
feed arrangements on a plastic substrate is likely to be more complicated to
manufacture and not necessarily low cost. The modelling and subsequent engin- High efficiency thick low
eering design of arrays for successful manufacture is often a factor that is Low feed radiation thin high
originally overlooked and ultimately pushes up development costs. There are Wide (impedance) bandwith thick low
many other examples where the commonly quoted properties of Table 1.3 need Low extraneous surface- thin low
qualifying, and recent experience from conferences and industrial contacts wave radiation
shows that academics have on occasions failed to convey a realistic impression Low mutual coupling thick low
to industry whereas industry itself has perhaps been too willing to implement the Low sensitivity to thick low
new technology without a sufficient design base that copes with the factors of tolerances
Table 1.2. We have already stressed the need for advances in CAD techniques
for manufacture and will specifically address this again later on, but now we tics of microstrip antennas and the fact that antenna volume-reduction benefits
return to the more general features of microstrip antennas such as the trade-offs must manifest themselves as cost factors which in turn demand a high standard
listed in Table 1 . 4 for
~ rectangular patch antennas. These are very approximate of engineering design to overcome.
and can be deduced from the basic patch equations [15]. An obvious deduction Finally we complete our discussion of general features with a list of appli-
which is nevertheless significant is that the use of thick low-permittivity sub- cations in Table 1.5 that have attracted the use of printed-antenna technology.
strates, giving essentially air spacing, gives many benefits. When the behaviour Almost without exception the employment of microstrip technology arises
of an array of patch elements (Table 1.4b) is considered, feeder radiation is seen because of a system demand for thin low-profile radiators. Conventional anten-
to increase for thicker lower-permittivity substrates [16, 171. With this exception, nas are clearly disadvantaged in such applications despite their often superior
any attempt to compact the antenna using a thin high-permittivity substrate will performance over microstrip antennas. In some cases the system has been
thus generally invoke all-round penalties in performance. These requirements created around the microstrip concept as mentioned earlier on.
are thus seen to be contrary to those for optimum operation of MICs, and this
imposes restrictions on the integration of antennas and associated front-end 1i2.2 Fundamental problems
circuitry. This perspective is valuable in emphasising the dominant characteris- In our vision of the future we have singled out reliable CAD techniques in array
manufacture and the system-led creation of innovative antennas as the major
8 Introduction Introduction 9
Table 1.7 Some generic types of bandwidth-extension techniques

'i
Table 1.5 Typical applications for printed-antenna technology
Aircrafr antennas Communication .and navigation Increasing antenna volume by incorporating
Altimeters
$1 parasitic elements, stacked substrates, use of
Blind-landing systems foam dielectrics
Missiles and telen?etr.y Stick-on sensors Creation of multiple resonances in input
Proximity fuzes response by addition of external passive
Millimetre devices networks and or internal resonant structures
Missile guidance Seeker monopulse arrays Incorporation of dissipative loading by
Integral radome arrays adding lossy material or resistors
Varactor and PIN dlode control grves a wlder effectrve bandwrdth and IS not Included In the above
Adaptive arrays Multi-target acquisition lrst
Semiconductor integrated
array thrusts. The problem areas will however centre around the fundamental issues
Batilefield communications Flush-mounted on vehicles listed in Table 1.6. These issues are un~versallyacknowledged and we will review
and surveillilnce some of them as follows to emphasise certain aspects which in our opinion are
I
worthy of clarification or perhaps need various points amplified, in particular
S ATCOMS Domestic DBS receiver to bridge the gap between academic research and industrial implementation.
Vehicle-based antenna
Switched-beam arrays 1.2.2.1 Bandwidth extension: The search for new microstrip configurations
Mobile radio Pagers and hand telephones with wider bandwith has been a dominant feature of the research literature and
Manpack systems much effort continues to be expended. No other type of antenna has been so
exhaustively treated as regards its bandwidth properties, yet the literature often
Reflector feeds Beam switching portrays an incomplete picture by not defining what is meant by bandwidth [18].
Remote Sensing Large lightweight apertures The many factors involved are listed in Table 1.7. A common and generally
realistic assumption is that the input-impedance characteristic of a resonant
Biomedical Applicators in microwave patch antenna behaves as a simple tuned circuit, in which case the 3 dB band-
cancer therapy width B is approximately (100/Q) percent, where Q is the Q-factor of the
Covert antennas Intruder alarms equivalent tuned circuit. If the antenna is matched at the resonant frequency of
Personal communication the tuned circuit, then away from resonance the input impedance will be
mismatched, creating a VSWR(> 1) of S, where

Table 1.6 Fundamental issues that will continue to be addressed

Bandwidth extension techniques
Control of radiation patterns involving sidelobes, beamshaping,
cross-polarisation, circular polarisation, surface-wave and
ground-plane effects
Reducing loss and increasing radiation efficiency
Optimal feeder systems (array architecture)
Improved lower-cost substrates and radomes
Tolerance control and operational factors
Use of a thicker and/or lower-permittivity substrate reduces Q and hence
1
1 increases B. An examination of numerous examples shows that, irrespective of
whether the permittivity or substrate thickness is changed, the main effect (Table
1.7) is that B increases with the volume of the antenna, i.e. the volume of
substrate between the patch and ground plane. Some examples are shown in Fig.
1.1, which also includes curves of radiation efficiency with and without al-
lowance for the power lost to surface waves. The first point of clarification is to
note that there are numerous ways of increasing the volume of a patch element
by employment of thicker substrate or stacking several substrates [19] or adding
 10 Introduction
parasitic elements 1201, but they all belong to the same generic type of bandwidth
extension technique.
A second generic technique (Table 1.7) consists of introducing multiple
Fig. 1.I Patch-antenna efficiency q and bandwidth B versus resonator volume for differen!
permittivities (Reproduced from Fig. 2 of Reference 78)
-x-x-x- is the radiation efficiency corrected for surface-wave action ( E , = 2 . 0 )
Fig. 1.2 Patch bandwidth extension using an external passive network
a Antenna without network
b Effect of matching network.
resonances in the input characteristic, as illustrated in Fig. 1.2 showing the
inclusion of a passive network in the input port; the presence of the network
invokes additional dissipative losses. The same bandwidth extension effects can
be brought about by introducing multiple resonances within the antenna itself
[I 81, which usually involves an increase in antenna thickness and hence volume.
The important point to note is that a multiple resonance input response does not
lntroduction 77
Table 1.8 Factors constraining the bandwidth of microstrip antenna ele-
ments and arrays
Element Array
--
Input-impedance Surface waves
characteristic
Side-lobe level Element mutual coupling
Cross-polarisation level Feeder radiation
Circular polarisation Corporate feed and mismatch
(axial ratio)
Pattern shape Scanning loss
(E- and H-plane symmetry)
Element gain
Efficiency
Feeder radition
obey the simple relationship of eqn. 1.1, and it is difficult to relate the various
multiple resonance bandwidth extension techniques that are reported in the
literature. Different researchers use different VSWR or insertion-loss criteria to
define the bandwidth and the insertion-loss curve shapes are likewise very
different.
A third much less common technique (Table 1.7) is simply to add lossy
material to the microstrip element. This technique would at first sight appear to
lead to unacceptable loss, but the manufacturing simplicity has definite appeal
and can outweight the other disadvantages.
We summarise the above three generic bandwidth extension techniques in
Table 1.7, but emphasise that from a system designers' standpoint the definition
of bandwidth based on the input-impedance characteristic is just one of many
factors listed in Table 1.8 that constrain the bandwidth of an antenna element
or array. For instance, the designer may decide to use a rectangular patch
accompanied by several parasitic elements to achieve an impedance bandwidth
specification, but then finds that the configuration fails to achieve adequate
cross-polarisation levels or perhaps E- and H-plane symmetry over the band. In
another instance it may be straightforward to meet all the bandwidth criteria for
a selected element only to find that, when the latter is connected in an array, the
bandwidth specification is not achieved because of mutual coupling or perhaps
feeder-line mismatches. Research workers seldom have the opportunity to
address the totality of problems in a system design, and it is a natural conse-
quence that they focus on the optimisation of a given property in isolation from
other requirements. In contrast, the industrial designer has to optimise many
parameters at the same time and bandwidth is a topic area where the gulf
72 Introduction
between isolated research and system design is at its widest. The challenge facing
researchers and industrial designers alike is to establish reliable designs for
elements and arrays that achieve bandwidth extension under a wide selection of
contraints as listed in Table 1.8. It is also highly desirable that the performance
of one type of element can be quantified in relation to the performance of any
other type of new element; the fact that there are in reality few generic types of
bandwidth-extension techniques (Table 1.7) [18] is an important guideline.
1.2.2.2 Pattern control: There is now ample evidence to show that the radia-
tion-pattern control of printed radiators is an order more difficult than with
reflector and aperture antennas. Even for modest performance levels of
sidelobes and cross-polarisation the printed-conductor topology presents many
variables to optimise for a given substrate thickness and permittivity. For
sidelobe and cross-polarisation levels of about -20 dB extraneous radiation
due to surface waves, feeder radiation and ground-plane edge effects is not
insignificant and computer models lose their precision. Surface-wave effects
decrease for lower-permittivity substrates but feeder radiation is then more
prominent [17]. There is evidence in the literature that much lower levels can be
achieved, but generally these are pattern cuts in certain preferred planes or
pertain to arrays fitted with lossy material or other special effects. A consensus
of opinion is that printed antennas are at present more fitted for applications
with less demanding pattern specifications. The challenge for the future thus
remains the lowering of the levels of extraneous radiation in printed arrays and
improved computer modelling of the overall patterns.
Some special mention needs to be made of circularly polarised elements and
arrays because considerable progress has been made in this respect and it is
likely to be an area for continued exploitation. It is well known that in principle
a linearly polarised antenna can be converted to perfect circular polarisation by
superimposing upon its radiation characteristics, those of its dual radiator
having transposed E- and H-field sources. For instance, a wire dipole (electric
source) would need to be combined with a wire loop (magnetic source), but in
reality it is physically impossible to construct or feed such an arrangement
precisely and compromises are made such as the employment of crossed-wire
dipoles which yields circular polarisation in a limited region of the hemisphere
and over restricted bandwidth. These and other techniques [21] are well esta-
blished for conventional antennas, and the point we make here is that they are
more difficult to translate to printed elements in view of the constrained planar
geometry and feeder requirements. It is therefore inspiring to note the inno-
vation that has been brought about whereby circular-polarisation characteris-
tics have been enhanced by sequential rotation of elements [22], incorporation
of finite substrate effects [23], novel feeder arrangements [24] and many more.
Creating improved low-cost radiators that provide circular polarisation over
wider bandwidths and larger sectors of the radiation-pattern hemisphere is a
goal towards which much international effort will continue to be directed.
Introduction 73
1.2.2.3 Eficiency and feeder architecture: The outstanding advantage of
microstrip - the simplicity of the printed conductor - is also the source of
one of its major disadvantages, which is the relatively high transmission-line
loss. The nature of the loss is well understood and arises from the high current
density at the strip edge and substrate losses. It is a fact that no worthwhile
reductions in transmission loss have been achieved since the inception of micro-
strip, and the simplicity of the structure offers little scope for innovation in this
respect. For patch elements the loss is less significant, and with an appropriate
low-loss substrate and strongly radiating patch, antenna efficiencies of 95% are
achievable. A conventional wire dipole antenna would have a better efficiency
than the patch but the order of loss of the latter is usually very small from a
systems standpoint. The main problem arises in large arrays having microstrip
or other forms of printed feeder lines because feeder losses limit the gain of the
aperture; in fact, beyond a certain critical aperture size the gain will actually
reduce. The beamwidth will, or course, also continue to narrow. The critical size
is dependent on the feeder topology, substrate etc. and a maximum gain around
35 dB is not uncommon. Fig. 1.3 shows typical computed and measured results
efficiency, %
100 50 j 0
/ I
gain
(dB)
u
O1 10 100
array size Dlho
Fig. 1.3 Patch-array gain
0 Calculated [17]; measured. with feed impedance + 100 a, x 120 A 200
[17] and indicates that at maximum gain an efficiency of about 10% can be
expected. Travelling-wave antennas show some economy of feeder loss over
corporate feeds but the frequency scanning loss for large travelling-wave aper-
tures is then the dominant limitation. Once again the simplicity of a printed
feeder system gives little scope for major design changes, and more recently
hybrid feeder systems are being considered incorporating more conventional
14 Introduction lntroduction 75

cables and waveguides for the longer feeder runs. We have already emphasised Table 1.9 Representative substrate list
in Section 1.2.2.2 the limitations on pattern control enforced by extraneous Material Supplier
E,
feeder radiation and any breakthrough in feeder architecture will need to
address the latter. However, for some applications the radiation-pattern specifi- Aeroweb (honeycomb) Ciba Geigy, Bonded Structures Div.,
cations are less critical than loss of gain and any improvements in feeder loss Duxford, Cambridge, CB2 4QD
would be a significant advance. The future challenge is to discover new feeder Eccofoam PP-4 Emerson & Cumming Inc, Canton,
architectures giving less loss, and if possible less extraneous radiation, with the (flexible low-loss Massachusetts, USA
knowledge that the already simplistic printed configurations offer little scope for plastic foam sheet) (Colville Road, Acton, London.
fundamental physical change. One possible avenue for advancement lies in the W3 8BU, UK)
integrated antenna concept whereby transistors are embedded in the feed struc-
ture to facilitate beam scanning. This may circumvent the loss problem but Thermoset microwave Rogers Corp., Bo 700, Chandler,
exacerbate extraneous radiation effects, and of course escalate costs. foam material AZ 85224, USA.
(Mektron Circuit Systems Ltd.,
1.2.2.4 Substrate technology: Substrate technology and marketing has been, 119 Kingston Road, Leatherhead,
and will continue to be, a key factor in the acceptance by industry of the Surrey, UK)
printed-antenna concept. Earlier microstrip antennas used plastic substrates or RT Duroid 5880 Rogers Corp.
in some cases alumina, but in recent, years the use of lower-permittivity sub- (microfiber Teflon
strates is common. The substrate role thus appears to be mechanical, enabling glass laminate)
the printed conductor to be suspended at a uniform height above the ground
plane. The use of lower permittivities also reduces surface-wave effects but Polyguide 165 Electronized Chemical Corp.,
feeder radiation is then more difficult to suppress. Antenna designers thus (polyolefin) Burlington, MA 01803, USA
require a wide range of substrates available having stable electrical and mech- Fluorglas 60011 Atlantic Laminates, Oak Materials
anical properties over the various ambient operating conditions. The major (PTFE impregnated Group, 174 N. Main St., Franklin,
problem has been, and is likely to be in the foreseeable future, a matter of glass cloth) M H 0323, USA.
substrate cost because the world demand is relatively small compared with that (Walmore Defence Components,
of some other plastic products. This has encourage some companies to manufac- Laser House, 1321140 Goswell Road,
ture their own substrates while in other cases the substrate costs have made some London, EClV 7LE)
large-array projects non-viable, and printed technology is then seen to be costly
in contradiction to the commonly upheld properties of Table 1.3. It is also noted Rexolite 200 Atlantic Laminates
that many microstrip antennas will require some sort of weather shield or (cross-linked
perhaps a radome, which again is a cost factor. Substrate technology thus offers styrene copolymer)
a challenge to material manufacturers to create lower-cost high-performance Schaefer Dielectric Marconi Electronic Devices Ltd.,
stable substrates. Clearly this is a somewhat circular problem which appears to Material, PT Radford Crescent, Billericay,
demand a higher-volume market to initiate an immediate advance; conversely (polystyrene with Essex, CM12 ODN, U K
such an advance would open up a larger-volume market. Such a situation is not titania filler)
uncommon, and with the considerable manufacturing interest in substrates that
is building up (Table 1.9) antenna designers should be optimisitic about the way Kapton film Dupont
substrate technology is likely to develop in the next few years. (copper clad) (Fortin Laminating Ltd., Unit 3,
Brookfield Industrial Estate,
1.2.2.5 Manufacture and computer-aided design (CAD): The microstrip Glossop, Derbyshire, UK)
antenna has been widely mathematically modelled for many years and yet from Quartz A & D Lee Co. Ltd., Unit 19,
the manufacturers' standpoint there is a dearth of ready-to-use design equa- (fuzed silica) Marlissa Drive, Midland Oak
tions, and hence reliable CAD packages. This situation has arisen partly owing Trading Estate, Lythalls Lane,
to the mathematical difficulties associated with practical geometries and partly Coventry, U K
 76 Introduction
6.0 RT Duroid 6006 Rogers Corp.,
(ceramic-loaded PTFE)
9.9 Alumina Omni Spectra Inc, 24600 Hallwood
Ct.
Farmington, Michigan, 48024, USA
(Omni Spectra, 50 Milford Road,
Reading, Berks, RGI 8LJ, UK)
10.2 RT Duroid 6010 Rogers Corp.
(ceramic-loaded PTFE)
II Sapphire Tyco Saphikin
(A & D Lee Co Ltd., Unit 19,
Marlissa Drive, Midland Oak
Trading Estate, Lythalls Lane,
Coventry, UK)
The brief details in the Table are intended to give readers an insight into the range and types of
materials available. Mention of any particular product does not imply our endorsement. Likewise
exclusion of a material does not imply adverse comment and we presume that some excellent
products have been omitted.
to the many varieties of patch antennas and the fact that designs must conform
to the vagaries of system requirements. Horn, wire and other conventional metal
antennas can be modelled to a high accuracy with well established formulas and
this is also true of electrically larger apertures such as the reflector antenna. In
contrast to these homogeneous electromagnetic systems the modelling complex-
ity arises largely from the presence of a finite-sized dielectric slab that gives rise
to the factors noted in Table 1.8 and elsewhere. This complexity is compounded
when the number of elements in an array increases and when the fine detail of
patch feeding or complex feed networks is required. Mutual coupling, surface-
wave effects and feed radiation manifest themselves as relatively small effects in
a small array but they quickly take charge of sidelobe and cross-polarisation
levels in the region of - 20 dB as the number of elements increase. When viewed
in the light of increasingly tight requirements, modelling accuracy is seen to be
the key parameter to successful implementation. As an example, the use of
inaccurate CAD may well be more expensive for the manufacturer than design
by hand in the long run. Likewise the range of applicability of the package needs
to be understood if inherently good models are not to be brought to bear on the
wrong problems. However, despite the unlikely possiblity of extending present-
day numerical analysis to arrays of patches in the immediate future, the problem
is not unsurmountable provided that a close liaison exists between the CAD and
antenna designer. Indeed there is a trend for manufacturers to evolve their own
CAD packages based on a mixture of simple closed-form expressions for the
radiation mechanism backed up by empirical results for the particular array in
question. Some degree of iteration is commonly included but with the decision
lntroduction 17
making and convergence under the designers control. Such an approach has
many merits since tolerance and operational effects can be added in gradually
to create a reliable manufacturing tool. A disadvantage of the approach is that
it must be re-established together with empirical data when a change is made in
the design, and furthermore the experience is confined to the particular man-
ufacturer. As we have mentioned already, some manufacturers have been sur-
prised by the need to underpin printed-array manufacture with positive modell-
ing in view of the commonly acknowledge property (Table 1.3) that the antennas
are 'simple and low cost'. There is ample evidence, however, showing that the
simplicity and low-cost properties are realisable once modelling has been acc-
omplished, and the latter is a one-off development cost and perhaps no more
than a few months of a printed-antenna specialist's time. The challenge for the
immediate future thus lies in the evolution of reliable interactive CAD packages
for printed-array manufacture that are capable of wider usage and of gaining
universal acceptance. In the long term one might expect some advances in the
rigorous analysis of microstrip-antenna geometries embodying practical feat-
ures which in turn will translate into more precise manufacturing techniques.
1.3 The handboook and advances presented
Many, if not most, of the international community of printed-antenna special-
ists have contributed to this handbook, which necessarily portrays the state-of-
the-art at the time of going to press. The contributions reflect the authors'
specialisation which in some cases is fairly wide ranging. This has meant that it
has not always been possible for the editors to maintain a full thematic flow
throughout the handbook. However, the chapters have been generally ordered
in the following way
element analysis and design
array aspects
microstrip technology
applications
To assist the reader we have already listed in Table 1.5 the general application
areas for microstrip antennas. In Table 1.10 the content of the Handbook is
resolved in more detail to identify with the various topic areas within the subject
of printed antennas. It can be seen that patch theory and design still concerns
many researchers, with current emphasis being on basic characterisation and
innovation for controlling in particular bandwidth and polarisation purity. The
same applies to arrays, with additional topics such as mutual coupling in
scanning arrays gaining more attention. Technology is addressed by several
authors with contributions on substrates, connectors, radomes and computer-
aided design and manufacturing. However, little on environmental factors has
 I
.
78 Introduction Introduction 79

"I. I
20 Introduction 1 Introduction 21
Table 1.11 Summary o f handbook chapters
Element analysis and design
2 Shafai and Kishk: Analysis of circular microstrip antennas: An analysis is
presented based on the equivalence principle involving both conducting and
dielectric boundaries. This allows substrate edges to be accounted for. The
I method is used to optimise a circular disc on a finite-sized circular ground
plane for low cross-polarisation and also a wrap around antenna.
3 Lee and Dahele: Characterisation of microstrip patch antennas and some
methods of improving frequency agility and bandwidth: The basic characteris-
tics of patches are reviewed here and conclusions on comparative perfor-
mance are made. Bandwidth is identified as being- crucial and methods of
overcoming the limitations by making the patches frequency agile are presen-
ted.
I
4 Haneishi and Suzuki: Circular oolarisation and bandwidth: The methods of
obtaining circular polarisation from patches are described in this chapler
I
together with design techniques. Again, as in the previous chapter, band-
I idth-extension methods are noted and, in particular, element pairing which is
described more fully in Chapter 13.
5 Katehi, Jackson and Alexopoulos: Microstrip dipoles: The analysis and
design of narrow-strip microstrip dipoles is presented here. For electromag-
netically coupled dipoles an improvement in feed radiation is noted together
with methods for offsetting mutual coupling in arrays. Bandwidth and super-
strate effects are also discussed.
6 Schaubert: Multilayer and parasitic confgurations: This chapter exhaustive-
ly reviews multilayer configurations and emphasises advances for wide band-
width, multiple frequency and dual polarisation. Such structures increase the
antenna thickness, and, as a contrast, antennas with coplanar parasitics are
also described.
7 Dubost: Widebandflat dipole and short-circuit microstrip patch elements and
arrays: Elements and arrays developed from the flat dipole concept are
described in addition to short-circuited quarter-wavelength patch elements
and arrays. The dipole work can perhaps be viewed as a parallel develop-
1 ment with microstrip and has produced antennas whose performance is
highly competitive.
8 Mosig, Hall and Gardiol: Numerical analysis of microstrip patch antennas:
An integral-equation formulation is solved by the moments method to give
solutions for arbitrary shaped wide patches, including input impedance,
radiation patterns and surface-wave effects.
9 Gupta: Multiport network approach for modelling and analysis of microstrip
patch antennas and arrays: Here patches possessing separable geometries in
whole or part are analysed using a planar model involving impedance
22 Introduction
Table 1.11 (Cont)
matrices. Radiation loading is included by means of edge admittances. The
author presents several illustrative examples and discusses the extension of
the work to CAD methods. Further progress in the application of analysis
such as this and others in this handbook is expected in the near future.
10 Van de Cupelle: Transmission-line model for rectangular microstrip anten-
nas: The application of the transmission-line model to patch analysis is
described. Various improvements to the basic model are noted, such as con-
nections for mutual coupling between the radiating edges, that enable good
agreement with measurements to be obtained. However, the attraction here
is the method's simplicity and easy adaption to CAD, an example of which
is given.
Array Aspects
I1 Daniel, Penard and Terret: Design and technology of low-cost printed
antennas: The design of elements and arrays with the emphasis on low-cost
technology is important for successful application in many areas. Design and
construction including array sythesis is described here. In addition, some
technology and substrate innovations are included which can be compared to
materials detailed in Chapter 15.
12 Pozar: Analysis and design considerations for printed phased-array anten-
nas: The effects of scanning of printed phased arrays are derived using
moments methods for both infinite and finite arrays of patches. Blind spots
due to surface-wave effects are noted to be particularly severe where high-
dielectric-constant substrates are used for millimetric integrated arrays. Some
alternative integration technologies are discussed that mitigate these and
other problems.
13 Ito, Teshirogi and Nishimura: Circularly-polarised-array antennas; Various
types of circularly polarised arrays are reviewed together with the possible
feeding arrangements. Some practical problems are considered including
pattern control and bandwidth. Some examples of practical arrays are also
highlighted.
Microstrip Technology
14 Owens: Microstrip antenna feeds: Printed antenna feeds are sometimes
given insufficientconsideration at the outset of an array design, thus degrad-
ing the array performance. Feed design is extensively reviewed here and
comparative examples drawn from the literature are used to give engineering
direction. Although considerable work has been done, further progress is
expected as the importance of good feed design for printed antennas is more
widely appreciated.
Introduction 23
15 Traut: Advances in substrate technology: Microwave substrates are one of
the important 'enabling technologies' in printed antennas. Advances in this
area are presented which given the reader some insight into manufacturing
and environmental factors that affect the antenna's progress from conception
to use. Progress here is determined to some extent by the volume of produc-
tion, and it is hoped that as the applications proliferate substrate technology
will continue to imvrove.
16 Levine: Special measurement techniques for printed antennas: Measure-
ment characterisation of connectors, lines and discontinuities, together with
analysis, form an important foundation to good antenna design. Such
measurement characteristics are described here with particular emphasis on
accuracy and applicability to design. Near-field probing can also form a use-
ful diagnostic tool, and examples are given together with a novel method for
efficiency measurement.
17 Zurcher and Gardiol: CAD of microstrip and triplate circuits: As noted
above, computer-aided design is likely to be an increasingly important factor
in printed antenna design. This Chapter is dedicated to CAD of microstrip
and triplate systems and highlights some of the important aspects such as
characterisation of components, materials, manufacturing, analysis and
synthesis.
Applications
18 Derneryd: Resonant microstrip antenna elements and arrays for aerospace
applications: Four examples of resonant microstrip antennas and arrays are
presented here for various requirements including dual frequency, monopulse
radiation pattern and dual polarisation. Important design freatures are noted
together with some environmental aspects.
19 Fujimoto, Hori, Nishimura and Hirasawa: Applications in mobile and
satellite systems: Mobile and satellite systems are an important area for low-
cost low-profile printed antennas. Various examples of such antennas are
given in this Chapter. It is likely that the explosive increase in information
systems in the future will accelerate the development of antennas to meet
these diverse needs.
20 Newham and Morris: Conical conformal microstrip tracking antenna: A
conical conformal microstrip tracking antenna is a severe requirement that
involves difficult manufacturing and fundamental electromagnetic problems
to be solved. Here the authors have described progress to date in this very
challenging area that is likely to require much more research and innovation
for some time to come.
21 Frayne: Microstripfield diagnostics: The near-field probing technique
noted in Chapter 16 is described here in some detail together with extensive
results both for microstrip patches and patch arrays with feed networks.
24 Introduction Introduction 25

22 SoizteN: Microstrip an~ennason a cylitirluiccrl .surfkce: Patch arrays on Relevant Chapter numbers are given together with References in brackets.
cylindrical bodies is likely to be an important future application of confor-
mal concepts. Modifications to the basic patch-design expressions due to the
curvature are presented here together with design and performance of a re- (a) Patches
presentative cylindrical array.
23 HUN, Henderson and Junzes: E.uretzsions and variations to the microstrip- The generic microstrip patch is an area
trntenncr concept: There are a wide variety of specialist applications that of metallisation supported above a
spawn innovative concepts in the use of microstrip antennas. This final ground plane and fed against the
Chapter highlights some of these, including applications where microstrip is ground at an appropriate point or
combined with other radiating or transmission structures to form hybrid
antennas. Operation over multi-octave bandwidths or at millimetric frequen-
cies are also design challenges that are covered, together with applications Principal shapes
involving very high and very low dielectric-constant materials. Such special- The freedom in the xy plane gives rise 2, 3,
ised requirements are likely to continue to lead antenna designers to further to the possibility of a multiplicity of 4, 8,
innovative progress in the microstrip field. possible shapes. Only a few have been 9, 10,
seriously examined such as the rectan- 11, 18,
gular or square patch and disc [25], ell- 21, 22
ipse [26], equilateral[27],or right-
appeared in the literature and Chapter 15 contains one of the few appraisals of angled isosceles [28] triangle, annular
this area of printed technology. A diverse range of applications is also noted that ring [29] and pentagon [30].
mirror the list given in Table 1.5. In terms-of printed-antenna techniques
microstrip is the predominant one discussed in the handbook although some
contributions on printed dipoles and ground-plane slots are presented. Although Characteristics of these principal
both these elements are likely to have similar electromagnetic properties to shapes are generally similar [Chapter
microstrip radiators, the increased complexity involved in manufacture seems to 31 with fundamental modes having
have been the overriding factor that has influenced designers away from them. broadside beam. Bandwidth and
This is particularly true for slot antennas in triplate stripline where shorting pins physical area vary between shapes.
or holes are needed to prevent parallel-plate mode excitation. It is thus clear that The annular ring gives increased
the conceptual simplicity of microstrip remains one of its most attractive feat- bandwidth, gain and sidelobe levels
ures.
As a final aid to the reader Table 1.1 1 highlights the advances described in
-
for higher-order modes but becomes
physically large.
each chapter with particular reference to the fundmental issues and challenges
identified in Section 1.2. Patches can be short-circuited along a 7, 1 I ,
null voltage plane to form the shorted 20

1.4 Glossary of printed antenna types

The various forms of printed elements and arrays are very numerous and it is
a l3ll
patch or hybrid microstrip antenna [31].
Impedance and resonant frequency re-
main the same as for a full-size patch
but for low dielectric constant the
bandwidth is increased.
useful for designers to have a check list at hand. With this in mind we have
composed the glossary giving an outline sketch, some key references and a few
supporting comments and an indication of which chapter in the handbook deals
with each type. Although the glossary is by no means exhaustive, the 70 entries
on radiating elements and the 37 on arrays reflect the wide range of flexibility
and scope for innovation that microstrip offers.
 26 Introduction
Variants on principal shapes
Circular sector and annular-ring-sector
[32] patches have been analysed using
the cavity model. Expressions for
impedance and resonant frequency,
but no indication of likely bandwidths
or pattern performance are given.
Star microstrip patches have been
theoretically investigated [33] as a
[331 radiator of higher-order modes with
good symmetry.
Rectangular-ring and H-shaped patch-
es have been investigated by Pala-
nisamy & Garg [34] and found to give
performance similar to principle
shapes.
Other possible shapes Many uninvestigated variants have
been suggested [28]. They are expected
to give performance similar to the
principal shapes. A selection is given
here.
Circularly polarised patches
Single point feeds
9 bandwidth. The above technique is
Multiple-point feed
135, 421
Introduction 27
Single-point feeding gives circular pol- 4, 9
arisation with constructional simplic-
ity. The feed excites two orthogonal
degenerate modes [35]. The 90" excita-
tion phase difference is obtained by
detuning the two modes by a variety
of geometrical distortions giving the
following types:
rectangular patch [36]
notched square patch [37]
slotted square patch [38]
notched disc [39]
truncated corner square [40]
ellipse [26]
penragon j3Fj
The input VSWR bandwidth is wider
than that of an isolated mode but the
axial-ratio bandwidth is much nar-
rower with axial ratio rising to about
6 dB at the edge of the 2: 1 VSWR
shown for the fundamental mode. Use
of internal slots in discs for higher-
order modes has also been shown [41].
Circular polarisation can be produced 4, 1 I ,
using multiple-point
- . feeding by: 13, 18,
19, 21
Two offset rectangular patches [35],
which is also used in arrays [42].
Here the offset phase centre leads
to a more rapid degradation of
axial ratio off boresight than the
following
28 Introduction Introduction 29
- - - - - - -.
90'
couplor
-- -
-
-.
-,
Single parches fed at two points [43,
44, 251
Shaped patches Patch shaping such as steps [521 or
conical depression [53] also yields wide
-. - [521 bandwidths
[43, 44, 251 Four-point feeding [45], which sup-
presses cross-polarisation generated
by higher-order modes within the ZBZZZ 1531
m [ 4 5 1 patch and this improves axial ratio
o... 110. Parasitics Multiple stacked patches with the 3
270' Short-circuit patches arranged to upper patches acting as electromagnet-
produce a crossed slot [46] which ically coupled parasitics can be design-
ed for extended bandwidths. Examples
using coaxial-probe feeding [54] and
microstrip-line feeding [55] have been
made with the latter designed for an
alumina (E, = 10) base substrate.
1551

A patch in a corrugated ground Parasitics can also be mounted copla- 3

plane [47], which also improves
low-angle performance 0
0
0 [561
nar as thin resonators [56], as addition-
al patches either gap or line coupled to
square [57] or triangular patches [58].
Alternatively many thin parasitics [59]

\ \ .:-- -- -- _-- ----

/ I
'--_ [47]
- ,' _
-
Wideband patches
Thick patches Use of low-dielectric-constant
(E, = 1.0), thick (h/& > 0.1) substrate
results in bandwidths > 10% [48].
Alternatively patches on thinner sub-
strates can be broadbanded (up to
30%) by e.uterna1 matching circuits
[49]. Use of thick substrates leads to
impedance-matching problems that
can be overcome by use of matching
gaps in the probe [SO] or patch [51].
n can be gap coupled to form a
wideband patch. Some of these con-
figurations exhibit variations in the
radiation pattern with frequency.
2, 4,
6, 7,
11, 18
Short-circuited quarter-wavelength par-
asitic~have also been applied to
square [60] or circular [61] patches. A
band-width increase of 2 is obtained
in the square-patch case. In the cir-
cular case cross-polarisation is sub-
stantially reduced.
 30 Introduction Introduction 31
Other wdeband forms The m~crostripspiral [62] gives about
40% bandwidth with limited effi-
ciency. The spiral 1s limited to less
than one turn, as further turns give
rise to radiation-pattern degradation.
p p

* . xledband notch
patch High-frequency patches have been lo-
cated within low-frequency patches to
Dual-frequency patches , give orthogonal polarisation [73] or
same circular polarisation [74] using
Multiple layers Multiple-layer patches having two- [I91 2, 3, [731 frequency-sensitive coupling stubs
and three- [63] frequency operation 4, 5,
use direct probe connection to the top 6, 18,
patch and gap coupling to the lower 19
ones. Direct connections to both
patches in two-frequency designs [64,
651 have also been made. Tuning of
the two frequencies is also possible by
an adjustable-height upper patch using [741
discs [66] and annular rings [67]. Other patch variants

In the coplanar stripline patch [75] the

input line is fed against the upper
ground plane. The overall perfor-
mance is similar to conventional
patches with reduced cross-polarisa-
751 tion and mutual coupling.

'm. [761
The electromagnetically coupled patch
[76] allows reduction in feed radiation
-
bv locating it closer to the -ground
plane than the patch. The effect of
Single layers Single-layer two-frequency patches 3, 4, 6 dielectric covers (superstrates) is noted
with orthogonal polarisations [68] using 1771

I /681
two feed points. Use of shorting pins
[69, 701 allows operation with the
same polarisation. The use of tabs in
[77], where, in addition to element
protection, enhanced gain can be
obtained.
rectangular [71] or circular [72] patches z The use of superimposed dielectric 23
also permits dual-frequency operation. spheres [78] on patches result in
[69, 701 improved gain and reduced cross-
polarisation.

patch resonator
32 Introduction Introduction 33
ground plane The groundplane dot [79] fed by a Outputs\
comparator
,ekments
microstrip line can be used as a bi-
directional element or as a unidirec-
tional one by the addition of a reflec
tor.

11n0 grwnd plane

[791
Folded dipoles can be operated close
to a ground plane by means of appro- substrate
priate matching circuits [14]. Many
variants are possible giving wide
bandwidth and low cross-polarisation.
Construction is more complex than Paiches have aiso been used in the re-
the basic microstrip patch although, ject array configuration [82] where
owing to the use of rr atching circuits, beam scanning is achieved by varying
wider impedance bandwidths may be the phase of the reflected wave by pin
possible. diodes. A single element is shown
here. Phase shift in the circularly pol-
arised system is obtained by varying
the angle of the short-circuit plane.

Microstrip patches and arrays can be 2, 23

maon combined with the reflector concept. Conformal antennas The flexibility of the microstrip con- 2, 7,
r~tlecto,
Their use as feeds [80, 811 allows in- cept allows use in conformal appli- 19, 20,
tegration with microwave integrated cations. Examples are: 22
4F
circuits, but with lower bandwidth and ground
plane - Spiral slot [83]
cross-polarisation compared to con-
\
ventional feeds.

SUppMt
34 Introduction Introduction 35
- Wrap-around antenna [84]: a single (b) Arrays
wide-quarter wavelength patch Feed structures
wrapped around a cylindrical body
Putch connection Patch elements for arrays can be 5, 6,
- Cylindrical 185, 861 or spherical [87] connected by through the substrate pin I I, 12,
patch arrays connections (through hole plating or 14, 16,
via holes), to one or more layers [89] 18, 19,
wide patch
of feed circuits located behind the 20, 22
ground plane in microstrip or triplate.
P I Mechanical simplification can be
achieved by aperture coupling to a
corporate parallel [90] or a perpendicular [91]
feed
microstripline. Feeding can also be
kcd pomni from a coplanar microstrip circuit [92]
which involves pattern perturbation
due to feed radiation. This can be
reduced by electromagnetic coupling
to overlaid patches [76]. Connector
effects give rise to fundamental limits
to array action [93] due to radiation
from the discontinuity in the guiding
structure. Pin connections to patches
give rise to higher-order modes [94]
that perturb the radiation pattern and
increase cross-polarisation levels.
---

Active patch Active devices can be integrated into Feed circuits Feed structures for many elements 4, 5,
patches. An example with a Gunn take various forms. Corporate feeds 7, I I,
diode [88] demonstrated the principle, [95, 961 for either one- or 12, 13,
but had high cross-polarisation and
low patch efficiency.
LJ LJ LJ LJ two-dimensional arrays give wideband 14, 18,
action, whilst series-fed arrays give - 19, 20,
narrow bandwidth with broadside 22
[951 beam when resonant or wide
bandwidth with a scanning beam
when travelling wave.
'4+"4-,-4'
36 Introduction
Wideband squintless operation is
obtained with the series-compensated
feed [97]
feed oolnt
The cross-fed arrangement [98] gives
narrowband action with a tapered
distribution and hence low sidelobes
with equal-width feed lines
Sequentially rotated feeding [22] gives
wide-bandwidth axial ratio and input
VSWR for circularly polarised patch
arrays
- -
Array structures
Two-dimensional arrays Microstrip patch arrays can be fed by
any of the feed structures noted
above. A typical linearly polarised
parch array [92] fed by a coplanar
microstrip corporate feed is shown
[921 here.
lntroduction 37
Dual polarisarion is obtained by dual
series interconnection of patches [99].
Chain type structures for linear
polarisation having rectangular [I001
and triangular or honeycomb shapes
[loll. Both the series array of patches
and chain arrays are resonant and
thus have a narrow bandwidth.
One-dimensional arrays: Early microstrip antennas were series 9, 10,
linear polarisation fed one-dimensional arrays of Gutton I I, 13,
and Bassinot [4] and Dumanchin [I021 14, 18,
in 1955 and 1959, respectively. Since 19
then many forms of series-fed array
have been developed.
Arrays can be formed using resonant
elements or meandering microstrip
lines, in which the radiation is
determined by radius of curvature or
line width.
Examples of arrays using resonant
elements are:
comb line [I031
parasitically coupled patch array
[ 1041
series-connected patches [I051
7, 1I,
12, 13,
14, 18,
19, 20,
22
 38 lntroduction lntroduction 39

Other array forms Various other forms of series fed 13. 23

linear arrays exist:

- - Pv Examples of arrays using meandering A A,/2 wide line can be made to

X *
U06, 1071 microstrip lines are: radiate by feeding with an asymmetric
serpent line 1106, 1071 step [I 131. Alternatively angled slots
triangle or tiapeioidi line 11061 can also force the line to radiate.
i y rampart line [I081
A combination of strip dipoles and
chain line [I091
slots can be used to form a circularly
[I 061 Franklin line [i l O]
polarised linear array [I 141.
The high losses in long microstrip
series arrays can be reduced by
replacing the line by a dielectric rod
[llj]; radiation occurs by coupling the
line energy to microstrip patches.

An omni-directional array has been

made by forming alternating
resonators in the line and ground
plane [I 161.

Multi-octave bandwidth operation can

be obtained by a series-fed log-periodic
arrangement of patches [I 171.
Circular polarisation Circular polarisation is obtained from 13, 14,
the rampart line [108], chain line [I 1I] 18
--.-<% and herringbone line [I 121.

L--.J

[lo71
40 lntroduction
1.5 Summary comments
The historical development and future prospects of the microstrip antenna are
reviewed to portray a n invention that is now reaching maturity while its sup-
porting research and development continues to expand unabated, driven by
system demands for conformal low-cost radiators. Future activity will be domi-
nated by both the creation of innovative designs to match system demands and
the search for improved CAD techniques in array manufacture. The concept of
distributed conformal sensors with integral signal processing is one projection
for the distant future. For completeness the more common features of micro-
strip antennas, their applications and typical antenna design criteria are listed
and briefly described prior to critically reviewing the outstanding design pro-
blems that are fundamental to microstrip antennas. The viewpoint of both
researcher and antenna manufacturer is usefully taken to identify knowledge
gaps and challenging issues that are vital to the advancement of the printed-
antenna concept. The importance of the contributions in the present handbook
in advancing the state-of-the-art is emphasied and each chapter briefly high-
lighted. Finally a glossary of microstrip antenna types is presented as an initial
guideline for the designer.
1.6 References
GREIG, D. D. and ENGLEMAN. H. F.: 'Microstrip - a new transmission technique for
the kilomegacycle range', Proc. IRE, 1952, 40, pp. 1644-1650
DESCHAMPS. G. A,: 'Microstrip microwave antennas'. 3rd USAF Symposium on Anten-
nas, 1953
LEWIN. L.: 'Radiatton from discontinuities in stripline'. Proc. IEE. 1960,107C, pp. 163-170
GUTTON. H. and BAISSINOT, G.: 'Flat aerial for ultra high frequencies'. French Patent
No. 7031 13, 1955
FUBINI. E. G. er a/.: 'Stripline radiators'. IRE Nat. Con. Rec., 3, 1955, pp. 51-55
McDONOUGH. J. A,: ' ~ k c e n tdevelopments in the study of pr~ntedantennas'. IRE Nat.
Conv. Rec., 5, Pt. 1, 1957, pp. 173-176
DENLINGER. E. J.: 'Radiation from microstrip resonators', IEEE Trans, 1969. MTT-17,
pp. 235-236
HOWELL, J. Q.: 'Microstrip antennas'. IEEE AP-S. Int. Symp. Digest, 1972, pp. 177-180
MUNSON, R. E.: 'Conformal microstrip antennas and microstrip phased arrays', IEEE
Trans., 1974. AP-22. pp. 74-78
Proc. of Workshop on Printed Circuit Antenna Technology, 17-19, Oct. 1979, New Mexico
State Univ., Las Cruces. New Mexico
IEEE Trnns, 1981, AP-29, (I)
BAHL. I. J. and BHARTIA. P.: 'Microstrip antennas' (Artech House, Dedham. Mass, 1980)
JAMES. J. R.. HALL. P. S. and WOOD, C.: 'Microstr~pantenna theory and design' (IEE.
Peter Peregrinus. 1981 )
DUBOST, G.: 'Flat radiating dipoles and applications to arrays' (Research Studies Press.
Antenna Series No 1, 1980)
JAMES, J. R., HENDERSON. A,, and HALL, P. S.:'Microstrip antenna performance is
v e ,Ve~en,s,1982, 12, (X), pp. 73-84
determined by substrate constra~nts',M i u u ~ ~ ~ oSwr.
Introduction 47
16 HALL, P. S., and JAMES, J. R.' 'Cross polarisation behaviour of series fed microstrip linear
arrays', IEE Proc., 1984, 131H, pp. 247-257
17 HALL, P. S., and HALL, C. M.: 'Coplanar corporate feed design effects in microstrip patch
array design', IEE Prac., 1988 135, H.
18 HENDERSON, A., JAMES, J. R., and HALL, C. M.: 'Bandwidth extension techniques in
printed conformal antennas'. Military Microwaves, MM 86, Brighton, June 1986, pp.
329-334
19 LONG. S. A., and WALTON, M. D.: 'Dual frequency stacked circular disc antenna', IEEE
Trans., 1979, AP-27, pp. 270-273
20 KUMAR, G., and GUPTA, K. C.: 'Non radiating edge and four edges gap coupled multiple
resonator broad band microstrip antennas,' IEEE Trans., 1985, AP-33, pp. 173-177
21 RUDGE, A. W., MILNE, K., OLVER, A. D., and KNIGHT, P.: 'Handbook of antenna
design' (IEE, Peter Peregrinus, 1982) pp. 24-28
22 TESHIROGI, T., TANAKA, M., and CHUJO, W.: 'Wideband circularly polarised array
with sequential rotation', Proc ISAP, Tokyo, Japan, Aug 1985. pp. 117-120
23 KISHK, A. A., and SHAFAI, L.: 'Effect of various parameters of circular microstrip
antennas on their radiation efficiency and the mode excitation', IEEE Trans., 1986, AP-34,
pp. 969-977
24 HORI, T., TERADA, N., and KAGOSHIMA, K.: 'Electronically steerable spherical array
antenna for mobile earth station'. IEE Conf. on Ant. and Prop.. ICAP 87, York, pp. 55-58
25 HOWELL, J.Q.: 'Microstrip Antennas', IEEE Trans, 1975, AP-23, pp. 90-93
26 SHEN, L. C.: 'The elliptical microstrip antenna with circular polarisation', IEEE Trans,
1981, AP-29, pp. 90-94.
27 LUK, K. M., LEE, K. F., and DAHELE, J. S.: 'Theory and experiment on equilateral
triangular microstrip antenna'. Proc 16th European Microwave Conference, 1986
28 Reference 12, pp. 139-153
29 CHEW, W. C.: 'Broadband annular ring microstrip antenna', IEEE Trans, 1982,AP-30, pp.
,."
41 51-07?
30 WEINSCHEL, H. D.: 'Cylindrical array of circularly polarised microstrip antennas'. IEEE
AP-S Int. Symp. Dig., 1975, pp. 177-180
31 PENARD, E., and DANIEL, J. P.: 'Open and hybrid microstrip antennas', IEE Proc., 1984,
131, H, (1)
32 RICHARDS, W. F., OU, J. D., and LONG, S.A.: 'Theoretical and exoerimental investiea-
tion of annular, annular sector and circular sector microstrip antennas, IEEE Trans, 1984,
AP-12. .oo.. 864-866
33 PARASNIS, K., SHAFAI, L., and KUMAR, G.: 'Performance of star microstrip as a
linearly and circularly polarised TM,, mode radiator', Electron. Lett., 1986.22, pp. 463-464
34 PALANISAMY, V and GARG, R.: 'Rectangular ring and H-shaped microstrip antennas:
Alternatives to rectangular patch antenna', Electron Left., 1985, 21, pp. 874-876
35 Reference 13, pp. 194-224
36 SANFORD, G. G., and MUNSON, R. E.: 'Conformal VHF antenna for the Apollo-Soyuz
test project'. IEE Int. Conf. on Antennas for Aircraft and Spacecraft, London, pp. 130-135
37 OSTWALD, L. T., and GARVIN, C. W.: 'Microstrip command and telemetry antennas for
communications and technology satellites'. IEE Int. Conf. on antennas for Aircraft and
Spacecraft, London, pp. 217-222
38 KERR, J.: 'Microstrip antenna developments'. Workshop on Printed Antenna Technology,
New Mexico State University, 1979, pp. 3.1-3.20
39 HANEISHI, M., et al.: 'Broadband microstrip array composed of single feed type circularly
polarised microstrip element'. IEEE AP-S Int. Symp. Dig, May 1982, pp. 160-163
40 SHARMA, P. C., and GUPTA, K. C.: 'Analysis and optimised design of single feed
circularly polarised microstrip antennas', IEEE Trans. 1983, AP-31, (6)
41 MARTIN PASCUAL, C., FONTECHA, J. L., VASSAL'LO, J., and BARBERO, J.: 'Land
mobile antennas for satellite communications'. IEE Conf. on Mobile Satellite Systems, 1984,
pp. 145-149
42 Introduction
HUANG, J.: 'Technique for an array to generate circular polarisation with linearly polarised
elements'. IEEE Trans, 1986, APZ-34, (9)
BRAIN, D. J., and MARK, J. R.: 'The disc antenna - A possible L band aircraft antenna'.
IEE Conf. Publ. 95, Satellite Systems for Mobile Communications and Surveillance, 1973,
pp. 14-16
MUNSON, R. E.: 'Conformal microstrip antennas and microstrip phased arrays,' IEEE
Trans, 1974, AP-22, pp. 74-78
CHIBA, T., SUZUKI, Y., MIYANO, N., MIURA, S., and OHMORI, S.: 'Phased array
antenna using microstrip antennas'. 12th European Microwave Conference, Finland, 1982,
pp. 475-477
SANFORD, G. G., and KLEIN, L.: 'Recent developments in the design of conformal
microstrip phased arrays'. IEE Conf. Publ. 160, Maritime and Aeronautical Satellites for
Communication and Navigation, London, 1978, pp. 105-108
BAILEY, M.C.: 'A broad beam circularly polarised antenna'. IEEE AP-S Symp. Stanford,
USA, 1977
CHANG, E., LONG, S. A,, and RICHARDS, W. F.: 'Experimental investigation of
electrically thick rectangular microstrip antennas', IEEE Trans, 1986, AP-34, (6)
GRIFFIN, J. M., and FORREST, J. R.: 'Broadband circular disc microstrip antenna',
Electron. Lett., 1982, 18, pp. 26-269
FONG, K. S., PUES, H. F., and WITHERS, M. J.: 'Wideband multilayer coaxial fed
microstrip antenna element', Electron Lett. 1985, 21, pp. 497-499
HALL, P. S.: 'Probe compensation in thick microstrip patches', Electron. Lett., 1987,23, pp.
606-607
PODDAR, D. R., CHATTERJEE, J. S., and CHOWDHURY, S. K.: 'On some broad band
microstrip resonators', IEEE Trans, 1983, AP-31, (I)
CHATTERJEE, J. S.: 'Conically depressed microstrip patch antenna', IEE Proc, 1983, 130,
H, pp. 193-196
SABBAN, A,: 'New broadband stacked two layer microstrip antenna'. IEEE AP-S Symp.,
Houston, 1983, pp. 63-66
HALL, P. S., WOOD, C., and GARRETT, C.: 'Wide bandwidth microstrip antennas for
circuit integration', Electron Lett. 1979, 15, pp. 458-460
SCHAUBERT, D. H., and FARRAR, F. G.: 'Some conformal printed circuit antenna
designs'. Proc Workshop of Printed Antenna Technology, New Mexico State University,
1979, pp. 5.1-5.21
KUMAR, G., and GUPTA, K. C.: 'Non-radiating edge and four edges gapcoupled multiple
resonator broad band microstrip antennas', IEEE Trans., 1985, AP-33, pp. 173-177
BHATNAGAR. P. S., DANIEL, J. P., MAHDJOUBI, K., and TERRET, C.: 'Hybrid edge,
gap and directly coupled triangular microstrip antenna', Electron Lett., 1986.22, pp. 853-855
AANANDAN, C. K., and NAIR, K. G.: 'A compact broad band microstrip antenna',
Electron Letl., 1986, 22. pp. 1064-1065
WOOD, C.: 'Improved bandwidth of microstrip antennas using parasitic elements', IEE
Proc., 1980, 127H, pp. 231-234
PRIOR, C., and HALL, P. S.: 'Microstrip disc antenna with short circuit annular ring,
Electron Lett., 1985, 21, pp. 719-721
WOOD, C.: 'Curved microstrip lines as compact wideband circularly polarised antennas',
IEE J. MOA, 1979. 3, pp. 5-13
MONTGOMERY, N. W.: 'Triple frequency stacked microstrip element', IEEE AP-S,
Boston, MA, June 1984, pp. 255-258
SENSIPER, S., WILLIAMS, D., and MCKONE, J. P.: 'An integrated global positioning
satellite antenna low noise amplifier system'. ICAP 87, York, IEE Conf. Publ. 274, 1987, pp.
51-54
JONES, H. S., SCHAUBERT, D. H., and FARRAR, F. G.: 'Dual frequency piggyback
antenna', US Patent No 4 162 499, 24 July 1979
Introduction 43
DAHELE, J. S., and LEE, K. F.: 'Dual frequency stacked microstrip antenna'. IEEE AP-S
Int. Symp. Dig. 1982, pp. 308-311
DAHELE, J. S., LEE, K. F., and WONG, D. P.: 'Dual frequency stacked annular ring
microstrip antenna', IEEE Trans, 1987, AP-35,
GRONAU, G., MOSCHURING, H., and WOLFF, I.: 'Input impedance of a rectangular
microstrip resonator fed by a microstrip network on the backside of the substrate'. 14th
European Microvwave Conference, Liege, Sept. 1984, pp. 625-630
ZHONG, S. S., and LO, Y. T.: 'Single-element rectangular microstrip antenna for dual
frequency operation', Electron Lett., 1983, 19. pp. 298-300
KUBOYAMA, H., HIRASAWA, K., and FUJIMOTO, K.: 'Post loaded microstrip anten-
na for pocket sue equipment at UHF'. Int. Symp. on Ant. &Prop., Japan, 1985, pp. 433-436
RICHARDS, W. F., DAVIDSON, S. E., and LONG, S. A,: 'Dual band reactively loaded
microstrip antenna'. IEEE Trans, 1985, AP-33, pp. 556-561
McILVENNA, J., and KERNWEIS, N.: 'Modified circular microstrip antenna elements',
Electron Lett., 1979, 15, pp. 207-208
KERR, J.: 'Other microstrip antenna applications', Proc 1977 Antenna Applications sym-
posium, Illinois, USA
SANFORD, G. G., and MUNSON, R. E.: 'Conformal VHF Antenna for the Apollo-Soyuz
Test Project'. IEE Conf. Publ. 128, Antennas for Aircraft and Spacecraft, 1975, pp. 130-135
GREISER, J. W.: 'Coplanar stripline antenna', Microwave J., Oct. 1976, pp. 47-49
OLTMAN, H. G.: 'Electromagnetically coupled microstrip dipole antenna elements'. 8th
European Microwave Conference, Paris, Sept. 1978, pp. 281-285
ALEXOPOULUS, N. G.. and JACKSON, D. R.: 'Fundamental superstrate (cover) effects
on printed circuit antennas', IEEE Trans, 1984, AP-32, pp. 807-816
JAMES, J. R.. HALL, C. M., and ANDRASIC, G.: 'Microstrip elements and arrays with
dielectr~coverlays' IEE Proc., 1986, 133, H. (6)
YOSHIMURA, Y.: 'A microstrip slot antenna', IEEE Trans, 1972. MTT 22, pp. 760-762
HALL. P. S., and PRIOR, C. J.: 'Microstrip feeds for prime focus red reflector antennas',
IEE Proc. 1987, 134, H, pp. 185-193
OLTMAN, H. G., WEEMS, D. M., LINDGREN, G. M., and WALTON. F. D.: 'Micro-
strip components for low cost millimeter wave missile seekers', AGARD Conf. Proc. 245,
Millimeter and Submillimeter Wave Propagation and Circuits, Munich, 1978, pp. 27.1-27.9
MONTGOMMERY, J. P.: 'Microstrip reflect array antenna element'. Proc 1978 Antenna
Applications Symposium, Illinois, USA
SINDORIS, A. R., SCHAUBERT, D. H., and FARRAR, F. G.: 'The spiral slot - A
unique microstrip antenna'. IEE Conf. Proc. Ant. & Prop., London, 1978, pp. 150-154
MUNSON. R. E.: 'Conformal microstrip antennas and microstrip phased arrays', IEEE
Trans. 1974. AP-22. pp. 74-78
SOHTELL, E. V., and STARSKI, J. P.: 'Cylindrical microstrip patch phased array antenna-
Chalscan C'. Military Microwaves Conf., Brighton, June 1986, pp. 317-322
LUK. K. M., LEE. K. F., and DAHELE, J. S.: 'Input impedance and Q factors and
cylindrical-rectangular microstrip patch antennas', ICAP 87, York, IEE Conf. Publ. 274,
1987. pp. 95-99
SEEHAUSEN, G: 'Polarisation control of conformal arrays consisting of linerly polarised
elements', ICAP 83, Norwich, IEE Int. Conf. on Ant. & Prop., 1983, pp. 154-157
THOMAS, H. J.. FUDGE, D. L., and MORRIS, G.: 'Active patch antenna';Proc. Military
Microwave Conf., June 1984, pp. 246-249
OWENS, R. P., and SMITH, A. C.: 'Dual band, dual polarisation microstrip antenna for
X band satellite communications', Military Microwaves Conf., Brighton, June 1986, pp.
323-328
POZAR, D. M.: 'Microstrip antenna aperture coupled to a microstrip line', Eleclron Lett..
1985, 21. pp. 49-50
BUCK, A.C., and POZAR, D. M.: 'Aperture coupled microstrip antenna with a perpen-
dicular feed, Electron Lett., 1986, 22, pp. 125-126
 Introduction
HALL, P. S., and PRIOR, C. J.: 'Radiation control in corporately fed microstrip patch
arrays'. JINA 86, Journeees Internationales de Nice sur les Antennes, 1986, pp. 271-275
HENDERSON, A,, and JAMES, J. R.: 'Design of microstrip antenna feeds - Pt 1: Estima-
tion of radiation loss and des~gnimplications', IEE Proc, 1981, 128H, (I), pp 19-25
LO. Y. T., SOLOMON, D., and RICHARDS, W. F.: 'Theory and experiment on microstrip
antennas', IEEE Trans, 1979, AP-27, pp. 137-145
Reference 13, pp. 116 and 161
Chapter 2
HALL, P. S., and JAMES. J. R.: 'Design of microstrip antenna feeds - Pt 2: Design and
performance limitations of triplate corporate feeds', IEE Proc., 1981, 128H. pp. 26-34
ROGERS, A,: 'Wideband squintless linear arrays', Marconi Rev., 1972, 187, pp. 221-243
Analysis of circular microstrip
WILLIAMS, J. C.: 'Cross fed printed aerials'. Proc 7th European Microwave Conference,
Copenhagen, Sept 1977, pp. 292-296
antennas
L. Shafai and A. A. Kishk
DERNERYD, A. G.: 'Linearly polarised microstrip antennas', IEEE Trans., 1976, AP-24,
pp. 846-85 1
TIURI, M.. HENRIKSSON, J., and TALLQUIST, S.: 'Printed circuit radio link antenna',
6th European Microwave Conference, Rome, Sept 1976, pp. 280-282
HILL, R.: 'Printed planar resonant arrays', [CAP 87, York, IEE Int. Conf. on Ant. &Prop.,
1987, pp. 473-476 2.1 Introduction
DUMANCHIN. R.: 'Microstrip aerials'. French Patent Application 855234, 1959
JAMES, J. R., and HALL. P. S.: 'Microstrip antennas and arrays - Pt. 2: New design Microstrip antennas are finding increasing popularity owing to their advantages
technique'. IEE J. MOA 1977, 1, pp. 175-181 in size, cost, conformity to the supporting structure and ease of fabrication [I]
CASHEN, E. R., FROST, R., and YOUNG, D. E.: 'Improvements relating to aerial
[2], [3]. To analyse their impedance and radiation properties many elaborate
arrangements'. British Provisional Patent (EM1 Ltd) Specification 1294024.
METZLER, T.: 'Microstrip series arrays', IEEE Trans., 1981, AP-29, pp. 174-178 analytical techniques are proposed and used. Numerical methods are also
TRENTINI, VON G,: 'Flachantenna mit periodisch gebogenem leiter', Freyuenz, 1960 14, developed and have received increasing attention in recent years, these being
pp. 230-243 primarily based on Sommerfeld-type integral equations. All these methods,
SKIDMORE, D. J., and MORRIS, G.: 'Design and performance of covered microstrip which are discussed in following Chapters of this handbook have one important
serpent antennas'. ICAP 83, Norwich, IEE Int. Conf. on Ant. and prop., 1983, pp. 295-300
HALL, P. S.: 'Microstrip liner array with polarisation control'. IEE Proc., 1983, 130H, pp.
assumption in common: they assume that the dielectric substrate and the
2 15-224 supporting ground plane are infinite in extent. The solutions are therefore valid
TIURI, M., TALLQUIST, S., and URPO, S.: 'The chain antenna', IEEE SP-S Int. Symp., for infinite geometries, or when the substrate and ground-plane dimensions are
Atlanta, USA, pp. 274-277 relatively large. The assumption does not introduce a severe difficulty in im-
NISHIMURA, S., NAKANO, K., and MAKIMOTO, T.: 'Franklin-type microstrip line pedance calculations since microstrip geometries are inherently resonant struc-
antenna', IEEE AP-S, Int. Symp., Seattle, pp. 134-137
HENRIKSSON, J., MARKUS, K., and TIURI, M.,: 'Circularly polarised travelling wave
tures and their impedance characteristic is primarily controlled by the printed
chain antenna'. Proc 9th European Microwave Conf., Brighton, 1979 elements. However, difficulty arises in predicting the radiation patterns, where,
JAMES, J. R., and WILSON, G . J.: U K Patent Specification N o 1529361, 18 Oct. 1978 for small antenna dimensions, diffraction effects alter the side and back radia-
MENZEL, W.: 'New travelling wave antenna in microstrip', Proc. 8th European Microwave tions. Consequently, the Geometrical Theory of Diffraction is occasionally used
Conference, Paris, 1978, pp. 302-206 in conjunction with other methods to improve the radiation-pattern predictions
ITO, K., ITOH, K., and KOGO, H.: 'Improved design of series fed circularly polarised
printed linear arrays', IEE Proc., 1986, 133, H , pp. 462466 [4, 51.
JAMES, J. R., JOHN, G., and HALL, C.M.: 'Millimetric-wave dielectric-microstrip anten- Accurate formulation of the electromagnetic problem of microstrip antennas
na array', IEE Proc., 1984, 131, H, pp. 341-350 is feasible. But, for finite substrate and ground-plane sizes the formulation must
HILL. R.: 'Twin line omni-directional aerial configuration', Proc. 8th European Microwave be solved numerically. In this Chapter we present a general formulation which
Conference, Sept. 1978, pp. 307-31 1 is based on the concept of equivalence principle, and provides integral equations
HALL, P. S.: 'Multi-octave bandwidth log periodic microstrip antenna array', IEE Proc.
1986, 133, H , pp. 127-137
for the field distribution on the surfaces of the conductors and dielectric sub-
strate. The formulation is exact and satisfies all boundary conditions. However,
since it involves the field distributions on the substrate and ground plane, the
numerical solution of the resulting integral equations is efficient only for small
antenna dimensions. The problem is considerably simpler for axisymmetric
geometries, where the surface distributions can be expanded in terms of the
azimuthal modes representing the physical modes of the structure. Consequent-
 46 Analysis of circular microstrip antennas
ly, since microstrip antennas support only a limited number of modes, the
numerical solutions for accurate field representations are readily obtainable. In
addition, the modal expansion of the fields reduces the problem to the solution
of matrix equations for each individual mode and simplifies computation con-
siderably. For this reason, all computed results in this Chapter are presented for
circularly symmetric configurations. The formulations can, however, be used
with additional labour for the investigation of other microstrip configurations
as well.
For instance, when the microstrip geometry is non-circular, or even arbitrary
in shape, one can use a surface-patch segmentation over the conducting and
dielectric surfaces. The current and surface distributions can then be represented
by appropriate basis functions over these patches to convert the integral equa-
tions to a matrix equation using a moment method. The solution of the resulting-
matrix equation gives the surface distributions over the conducting and dielec-
tric surfaces. However, since the segmentation is over the entire surface the
matrix size is normally large. In addition, the method gives numerical results for
the surface distributions and fails to provide information on the modal excita-
tion. This difficulty can be overcome by expanding the current distributions in
terms of patch eigen functions, in which case the procedure becomes similar to
the case of axisymmetric configurations discussed previously.
Since the existing solutions in the literature predict microstrip-antenna im-
pedance properties accurately, no attempt is made here to investigate the
antenna impedances. Instead, emphasis is put on predicting the radiation pat-
terns and investigating the effects of microstrip dimensions on them. Conse-
quently, to simplify the analysis, excitation sources are replaced by simple
electric dipoles. No significant effect is anticipated by this source simplification,
since the resonant nature of microstrip antennas controls their mode excitation,
and thus radiation patterns.
The generated equations are used to investigate the radiation properties of
three different antennas; namely, a circular microstrip patch antenna, a wrap-
around antenna and the reflector feeds. They are fundamentally different anten-
nas and selected to provide complementary analysis and design information.
For instance, the circular patch is one of the basic microstrip antennas. Its
radiation characteristics and mode excitations are studied in length and the
effect of the ground-plane size and other dimensional or material parameters on
its radiation patterns are investigated. The results, although computed for a
circular patch geometry, provide information for precise understanding of the
radiation properties of resonant patch antennas. The wrap-around antenna is
selected to show that the formulation can be used to investigate any axisymme-
tric antenna configuration. It is also shown that multiple source excitation can
be used to control mode excitation, and consequently the radiation patterns.
The last example, i.e. the reflector feed, is included to indicate the usefulness of
the method for design of precision antennas. The reflector feeds should not only
provide an efficient illumination function, but their cross-polarisation and the
Analysis of circular microstrip antennas 47
phase-centre location must also be controlled precisely. The numerical method
provided in this Chapter enables accurate and efficient generation of pattern
data, which, when coupled with optimisation algorithms, gives antenna-design
parameters to meet stringent performance requirements.
2.2 Formulation of the problem
The electromagnetic problem involving microstrip antennas deals with deter-
mination of the field components in the presence of conductors and dielectrics.
The boundary conditions to be satisfield are therefore of mixed type. This
requires vanishing of the tangential electric-field components on the conductors
and continuity of the tangential electric and magnetic components over the
dielectrics. Because practical geometries are finite h size, an exact analytical
solution cannot be found to satisfy all boundary conditions. A numerical
solution must therefore be utilised. In this regard, two formulation types can be
developed. One involves volume integral equations for the polarisation currents
in the dielectrics and the induced surface currents on the conductors. This type
of formulation is not convenient to work with, but is general enough to handle
inhomogeneous dielectrics. For homogeneous dielectrics a convenient formula-
tion can be developed in terms of the tangential field components over the
boundary surfaces. The resulting integral equation includes all boundary con-
ditions and the formulation is therefore an exact one. Thus the solution acc-
uracy will depend on the management of the problem thereafter and on the
numerical algorithms used to determine the unknown surface distributions. In
the following Sections we shall provide integral-equation formulations only for
the surface distributions, and present numerically generated data for several
known antenna configurations.
The formulations may be derived from the use of the equivalence principle [6,
71. To proceed we select a general electromagnetic problem shown in Fig. 2.la,
where a homogeneous dielectric material is sandwiched between two conducting
layers. The surfaces S,,, S,, and Sckrefer, respectively, to the boundaries between
the conductors and the exterior region, the conductors and the dielectric, and
the exterior region with the dielectric. Similarly, Ed, Rdand E', I;i' refer to the
field vectors within the dielectric and exterior regions, respectively. The dielec-
tric region has a volume V,, bounded by surfaces S, and S,, and its material
parameters are td and p,. The exterior region has a volume V, and its permit-
tivity and permeability are defined by E, and p,, respectively. The excitation
sources are provided by impressed electric and magnetic currents J', and ad,
within the dielectric.
We now may invoke thd equivalence principle to reduce the complex original
problem to two simpler ones [8], involving the exterior and interior regions. Fig.
2. I b shows the external equivalence. The combined volume of the conductors
and the dielectric is bounded by S,, and S,,, and supports equivalent currents I,,
 48 Analysis of circular rnicrostrip antennas
Analysis of circular rnicrostrip antennas 49
'
m.
:I
7, and These currents radiate in a homogeneous medium (E,, p,) and
produce (E',R')in V, and zero field within the bounded region. Here, J,, is the
electric current on S,, and Jdeand A? are the electrlc and magnetic currents on
S,. The internal equivalence is shown in Fig. 2.lc, where the volume V, is
enclosed by S,, and S,,. The equivalent currents are - I,,, - 1, and -A? and
together with J', and radiate in a homogeneous medium ( E,, p,) to produce
( E d , I?,) in V, and zero field elsewhere. Again, -3, is the electric current on
1 S,,, and - J, and -A? are the electric and magnetic currents on S,. Since, in
the original problem of Fig. 2. l a the surfaces S , and S , are perfectly conduct-
ing, they support only equivalent electric currents in Figs. 2.lb and 2 . 1 ~ The
.
1
negative-sign relationship between the aperature currents of Figs. 2. l b and 2.lc
is dictated by the zero-field stipulations and the continuity of the tangential
'ce
electric and magnetic field components across the aperature surface S, of Fig.
Fig. 2.1 a Original problem 2.la. However, the selection of the negative sign for -Jcd in Fig. 2.lc is not
mandatory and is made to match the negative sign of the aperature currents.
In the above example, the application of the equivalence principle reduced a

'ce -
I Jce -e y H-e
complex multi-region problem to two simpler ones involving homogeneous
regions. The field components in each region can therefore be found readily
from the equivalent currents. However, these equivalent currents are still unk-
nown and must be determined. This can be achieved by enforcing the boundary
conditions on the field vectors of the original problem in Fig. 2.la. The boun-
dary conditions to be satisfied are:

zero field

and the surface equivalent currents are

Fig. 2.1 b External equivalence

Again, the currents I,, I,, and 1, are the equivalent electric currents on each
respective surface and lii is the magnetic current on the interface surface
between the dielectric and the exterior region. The field components in eqns. 2.1
can be determined from these equivalent currents and provide the following field
relationships:

Fig. 2.1 c Internal equivalence

 50 Analysis of circular microstrip antennas Analysis of circular rnicrostrip antennas 51

- &""(J,, + J,, m - flP,(J', + J,, m formulations involve the surface distributions on the dielectric, their numerical
solution for arbitrarily shaped geometries may require excessive computer time
= - R&(Jid,0 ) on S,, (2.6) and storage. The problem is considerably simpler for axisymmetric geometries,
where E'(J, && and Ed(J,A?) represent the electric fields due to the currents 3
and A?, radiating in media characterised by E,, p,, and E,, p,, respectively. R'
(3, h?) and R q J , 12;1) are the associated magnetic fields. Note that, since the
equivalent currents are still unknown, the field eqns. 2.3 - 2.6 represent integral
equations for these currents. These integral equations can be generated using
appropriate vector potentials, in terms of which the field vectors are given by

where

and

The function Gv is the scalar Green's function and is given by

where R = (r - r'l is the distance between the field point r and the source point
r' onthe surface, k, = w ( ~ , p , ) "is~the propagation constant of the region and
q represents e or d.
2.2.1 Matrix formulation
The above formulations provide integral equations valid for any combination
of dielectric and conducting bodies of arbitrary shape. They can be solved for
the unknown currents by a non-linear optimisation routine or after linearisation
of the relationships by an application of a moment method. However, since the
Fig. 2.2 Geometry of the body of revolution
where the field vectors can be expressed in Fourier series of the azimuthal
co-ordinate. A solution can therefore be generated separately for each Fourier
component, resulting in reduced computation time and storage. This is par-
ticularly important in microstrip antennas, which are highly resonant and often
support only one of the azimuthal modes. For this reason, we shall restrict the
remaining material of this Chapter to the development of solutions for ax-
isymmetric geometries.
 52 Analysis of circular microstrip antennas Analysis of circular microstrip antennas 53
Fig. 2.2 shows a simple representation of an axisymmetric object, generally along the unit vectors ii, and i i , and expressed in the form [9 - 1I]
known as a body of revolution. The surface tangents can be defined along the
generating curve I and the azimuthal co-ordinate 4. They are shown in Fig. 2.2 J (7)= i,.S(1', 4') + k,J@ (t', 4') (2.25)
and form an orthogonal curvi-linear co-ordinate system on the surface of the
body. Because the geometry is rotationally symmetric the surface co-ordinates
can be represented conveniently in terms of p, 4 and z co-ordinates in a where S , and M', Mb are the current components along ii,. and ii,., respective-
cylindrical system, with the origin on the axis of the body. We define the ly.
orthogonal surface tangents by their unit vectors ii, and ii, and the outward The electric current J exists on both conducting and dielectric surfaces, but
normal by the direction of its unit vector ii given by A? exists only on the dielectric. If the electric and magnetic surface currents are
expanded into N, and N,,expansion functions, respectively, the surface currents
A = ii,xri, (2.16) can be represented by
On the surface of the body we define a field point by its co-ordinates ( I , 4) or
(e, 4 , z ) and a source point by (t', 4') or (Q', $', 2 ' ) . Their respective unit tangent
vectors are (i,, ii,) and (ii;, G).The unit vector ii, is orthogonal to the z-axis,
but ii, and i are at an angle v. This angle is assumed to be positive if ii, points
away from the z-axis. Similarly, v' is the angle between ii', and the z-axis at (t',
4'). The relationships among these unit vectors can be determined by an + M$ K,$ ( t ' , $')aq (2.28)
inspection of Fig. 2.2, and are given by
where Ji,, J$, Kh, K$ are expansion functions defined by
12, = sinv ii, + cow 2, (2.17)
ii,, = sinv' cos(4' - 4)ii, + sinv'sin(4' - 4)ii, + cosv'ri, (2.18)
The range - M , to + M , gives the total number of azimuthal modes. The
ii,, = -sin(@ - +)ii, + cos(4' - d)ii, (2.19) coefficients I;,, I$, MAl, M$ are the current coefficients to be determined by
solving the matrix equation which results when eqns. 2.27 and 2.28 are sub-
In addition, if the positional vectors of points ( t , 4) and (1', 4') are r and r ' ,
stituted via eqns. 2.25 and 2.26 into the integral eqns. 2.3 - 2.6. The procedure
respectively, then
involves taking the inner products of the field equations with certain testing
P = @ a , + zii, functions and integrating them over the surface. The testing functions are
defined by
and

F' = Q' Co~(4'- 4) + p'sin(4' - 4)i, + zf6; (2.21)

tTf = d4A (t) (2.31)
where
The inner product of two vectors P and is defined by their scalar product and
integrated over the surface of the body; that is

Also, by definition, the surface gradient of a scalar function @ on the body of

revolution is given by
and the surface divergence of a vector function @' is defined as
The unknown currents J a n d A? can now be decomposed into two components
The expansion and testing functions, i.e. [J,,,, R,,,]and [@,;I, as defined by eqns.
2.29,2.30 and 2.31 are orthogonal over the period 0 to 2a in 4. This means that
the inner products of k?;' and J{ ( p = t or 4) vanish for I # n and the
contribution of different azimuthal mode separates. The resulting equations
therefore involve only a particular mode of index n. This is the major simplifica-
tion that is introduced by the mode orthogonality in axisymmetric objects.
Accordingly we obtain a separate matrix equation for each mode.
For explicit evaluation of the matrix elements, one must choose f;(t). It is
known that subsectional expansions, using flat pulse or triangle pulse functions,
 54 Analysis of circular microstrip antennas Analysis of circular microstrip antennas 55
give rise to well conditioned matrices. Flat-pulse current expansions with point and are shown in Fig. 2.4, where the coefficient T', for i = 1, are given by
matching were first used to solve the scattering problems of conducting spheres.
However, with such expansions the moment-method solutions of the surface
currents do not converge rapidly to the exact solutions. The triangular-function
expansions, on the other hand, converge satisfactorily and provide accurate
solutions. For this reason triangular pulse functions are used here to represent
both current expansions and the testing functions [9]. A minor deviation from
the literature is to use d ( t ) instead ofJ;(t), which is defined by four impulses
given by

Now, all information needed to proceed with transferring the integral equations
to a system of linear matrix equations is known. Following procedures well
known in the application of moment methods the matrix equation for the nth
Fourier component of currents can be written as

where r, is a square matrix representing the impedance and the admittance

sub-matrices, 1, is a column matrix for the unknown expansion coefficients of
and fi,and p,,is the excitation column matrix. Each mode has a matrix

Fig. 2.3 Triangle function approximation

where 6(t) is the unit impulse function and its coefficients Tare defined in Fig.
2.3, which for i = 1 are given by

where q, = q,/q,; and V,d is the excitation sub-matrix due to the electric-field
sources in the dielectric and I ,is the excitation due to the magnetic-field sources
in the dielectric region, respectively. The sub-matrices Z and Y with superscripts
e and d denote the impedance and admittance matrices for the exterior and
Similarly, the derivative of ef;(t) is approximated by four impulses as
interior media, respectively. The first pair of suffixes identify field surface and
the second pair of suffixes identify the source surface where the Fourier mode
d
;ii[ d ( a = z4

p = l
T;+4;-46(t - 1,+21-2) (2.35) n is implied. I,,,,I ,,,,, Icle,,and M,, are the unknown expansion coefficients of the
electric and magnetic currents on S,,, S,, and S,,, respectively. In the above
 56 Analysis of circular microstrip antennas
Analysis of circular microstrip antennas 57
equations, each sub-matrix Y:: or Z::consists of four submatrices. They are given
( Y 3 ) , = &(z, - z,)G, (2.44)
by
(Y4)ij =

and

and
p' = p + 4 i - 4
r = I+4j-4
f = p + 2 i - 2

f = I+2i-2
where
Fig. 2.4 Derivative of the triangle function approximation

and their elements have the form

with
58 Analysis of circular microstrip antennas Analysis of circular microstrip antennas 59
and hf' is the spherical Hankel function of the second kind and zero order and
is the dipole moment in the =-direction. If the Hankel function is represented
by

at t = t ' , R is approximated by
with
G,,, = G-, =

I
r' > r
a,, hi2' (kqr')j,,(kqr)P~(cos9)P~(cosB'),
2.2.2 Excitation matrix I, F 0,
(2.66)
Microstrip antennas are normally excited by a transmission line or a coaxial
probe. To solve the problem numerically one must model the exciting source, f- a,, hL2' (kqr)j,,(kqr')P~(cos9)P,"(cosO'),
1, = "l
r' < r
from which the elements of the excitation matrix can be determined. However,
a precise modelling of either source, i.e. the junction between the transmission where a,, = (2n + 1) (n - m)!/(n + m)!; then using eqn. 2.61 for I?, its
line or the coaxial probe with the microstrip patch, although feasible is a difficult excitation matrix elements can be calculated from
task. On the other hand, microstrip antennas are highly resonant structures and
within their operating frequency band one of the Fourier components, i.e. the
modes, dominates. This means, one can represent the excitation source with a
simple elementary source, such as an electric dipole, without affecting the
solution accuracy. The representation of the source by a single electric dipole is
quite adequate if the substrate thickness is small, or the width of the trans-
mission-line feed is not excessively large. Otherwise, multiple dipole sources
must be used. For instance, when the substrate thickness is so large that the
current distribution along the coaxial probe is not constant, a linear array of where 1, is the t co-ordinate of the upper end of the generating curve. Similarly,
electric dipoles along the probe length may be used. In such a case, the excitation using eqn 2.62 for A'"',its matrix element can be calculated from
of dipoles must correspond to the current distribution of the probe. Similarly,
for simulating wide transmission-line junctions, multiple dipole exciations may
be used, where their excitation must be weighted by the field distribution under
the line. In the following analysis we consider only the case of thin substrates
or transmission lines, and represent the excitation source by a single electric
dipole. This simple form of source representation simplifies the excitation
matrix considerably.
Simulating the excitation by an electric dipole, its electric and magnetic fields 2.2.3 Radiation fields
can be computed from Once the induced currents J and A? on the surface are determined after the
Erncq = - j m ,$ - v ~q (2.61) solution of the matrix equation, the field components E, and E4 at a far-field
point (r,, O,, 4,) can be determined [I21 as

and
 60 Analysis of circular microstrip antennas Analysis of circular microstrip antennas 61

with and
1
Fl(Oo,4,) = L(.l i, +-
%
A - i4)e-"'o" ds (2.73)

where S is the exterior antenna surface, Po is a unit vector in the direction from
the origin of the co-ordinates to the field point, P' is the positional vector of the
+ -1 K ~ , [ s i n v ' c ~cos(4'
s~~ - 4,) - cosv sine,]
'I
source point (x', y', 2 ' ) on the antenna, and C and i, are unit vectors in the
, direction of increasing 0 and 4 , respectively. Referring to the field point, these

vectors can be written as

Changing the order of summations and integrations and using the integral
i, = cos 0, cos 4,cir + cos 8, sin 4,6? - sin B0cir (2.75)
representation for the Bessel function of the first kind, namely
i, = - sin q5,ci, + cos 4,4 (2.76)
and,the dot product in the exponential term of eqns. 2.73 and 2.74 can be shown jn(@) =
jn
2n
2~
le-jeco-e-~nada
0
(2.84)
to be
the integration in eqns. 2.82 and 2.83 involving the azimuthal co-ordinate 4' can
to P' = (x'cos 4, + y' sine,) sine, + z'cos 8, I be evaluated as
= ~ ' c o s(4' - 4,) sine, + z'cos0, (2.77) ~ ~d ~4' ( =
~ l r ~ - ~~o~coseo 4 ~2 i ~ j " d "J,
~~(kef sine,) (2.85)
where (Q', d', z') are the cylindrical coordinates of the point (x', y', 2'). If we
now substitute for J a n d fiin eqns 2.73 and 2.74, and evaluate the dot products
(fi,. i,), (fi,, i4) as
.
6,. i, = sinv' cos0, cos(4' - 4,) - cosv sine, (2.78)
fi,. i, = sinv' sin(& - 4,) (2.79)
and The far-field functions can then be represented by
fi4 5 = - sinO,sin($' - 4,)
fi4 . i4 = sin& + cos 4'cosq5,
= cos(4' - 4,)
then F,(flO,4,) and F,(B,, 4,) take the form

+ (R$,~M),k,?,j]
where
4

-12) cos &sin (4' - 4,)

(R?)j = C [j(J, + , - Jm- ,)sin v'cosOo - 2 cos v' sin eoJ,]
Y-I

(2.89)
+-v1 KAlsinv'sin(@ - 4,)

+-I?1 k$, COS(@- 4,) I

~ ' 4'
d dt'
 62 Analysis of circular microstrip antennas Analysis of circular microstrip antennas 63
can be generated from the above symmetric property. The second property is the
mode symmetry. This enables one to generate the Y and Z submatrices of the
with modes with negative indices from the positive ones, as

z?, = zp
The same relations are valid for the Y submatrices. Similarly, the elements of the
excitation for an electric dipole, generated using eqns. 2.67 - 2.70, satisfy the
following relationships:

The far-field component can finally be put in a compact form as

Thus the solution of the matrix equations, given by eqns. 2.38, gives the
following symmetric relationships satisfield by the current coefficients:

M", = M!
This completes the formulation of the problem and determination of the ra- Accordingly, one only needs to compute the coefficients of the positive modes,
diated fields from the computed equivalent currents. The external field near the
i.e. one half of the mode coefficients. This results in a major reduction of
antenna or the field within the dielectric substrate can also be determined, but
computation time and cost. The mode symmetry can also be used in calculating
are omitted here for brevity.
the far-field components from eqn. 2.98. The needed relationships are
Two different properties of the Y and Z submatrices are used to reduce the
computation time and cost. One is the matrix symmetry. This property is
evident from eqns. 2.42 - 2.49 and gives
(Z1l)
m B
. = (Z1')..
m 11

(Z')g = - (-Z!)ji @hJ = p.J

m

(Z?,, = (Z9ji which can reduce the summation over the modes to one from m = 0 to + M.
(Y")..
m v
= - (Y,y).,
(y3v = - (y?)Ji 2.3 Application 1: Circular patch antenna
(Y;,)v = - (Y;@)ji
The circular patch antenna is one of the fundamental microstrip geometries and
This means that only one half of the matrix needs to be created. The other half its impedance and radiation characteristics have been investigated extensively.
 64 Analysis of circular microstrip antennas Analysis of circular microstrip antennas 65
However, the methods used so far assumed an infinite size for the ground plane The geometry of a circular patch microstrip antenna is shown in Fig. 2.5,
and substrate. The solutions are therefore approximate and lack the influence where the excitation is simulated by an electric dipole immersed in the dielectric
of the finite substrate and ground-plane dimensions. Their accuracy therefore substrate under the conducting patch. The radius is selected as [2]
depends on the type of application. For instance, because microswp patch 2h na
geometries are highly resonant, their impednace characteristics are dominated a, = a[l +-
xu&,
(In - + 1.7726]'12
2h
by the patch dimensions. The ground plane size, provided it is reasonably larger
than the patch, has a negligible effect. Similarly, the radiation near the broadside
direction is determined primarily by the patch itself. Th finite size of the

c i r c u l a r patch

I= /-- dielectric

/'

, finite
ground p l a n e
Fig. 2.5 Microstrip antenna geometry

substrate or ground plane influences the radiation at wide angles, and par-
ticularly behind the antenna. Thus, when radiation patterns along the broadside
direction are necessary or their approximate form is adequate, the analytic
solutions can provide sufficient information. The numerical method presented Fig. 2.6 The computed electric and magnetic surface currents of the 7M,, mode on the
in this Chapter enables one to determine the radiation characteristics in all outside boundary (Reproduced from Reference 14 @ 1986 IEEE)
space. In addition, the accuracy of the generated results can be very high. Thus
it is a useful method for generating solutions for high-precision work, such as where a is the radius of the conducting patch, a, is the effective radius due to the
in a reflector-antenna feed design. In this Section we present a few representative spread of the fringing field from the patch edge to the ground plane, h is the
results for a circular patch antenna. dielectric thickness and E, is the relative permittivity of the dielectric substrate.
66 Analysis of circular rnicrostrip antennas
The effective radius is calculated from
where K,,,,, is the mth zero of the derivative of the Bessel function of order n. The
effective patch radius is therefore a function of substrate height, the dielectric
permittivity and the order of the excited mode. The effects of these parameters,
as well as the ground-plane size, o n the radiation characteristics of the patch are
investigated in the following Sections [14].
Fig. 2.7
The computed electric and magnetic surface currents of the TM,, mode on the
outside boundary (Reproduced from Reference 14 @ 1986 IEEE)
Analysis of circular microstrip antennas 67
2.3.1 Surface fields
In microstrip antennas the radiation is normally from the periphery of the
patch, where the fringing field is maximum. However, since the exciting dipole
launches guided modes of the parallel-plate region under the patch, it is desir-
able to compute the surface-field distributions on the conducting and dielectric
surfaces of the antenna to understand their behaviour. These distributions are
given by the equivalent currents 1 and A?. For two selected modes, i.e. the
dominant TM,, and the higher TM,, modes, the computed results are shown in
Figs. 2.6 and 2.7. In Fig. 2.6 the patch radius is a = 0.181I, where I is the
wavelength in free space and the patch resonates in the dominant T M , , mode.
The horizontal axis shows the length of the contour along the generating curve.
Since the geometry is rotationally symmetric only one half of the surface
contour is shown. The external surface currents are plotted with respect to their
locations on the surface, where points A to B correspond to the ground plane,
points B to C represent the dielectric substrate which supports both electric and
magnetic currents and points C to D correspond to the patch surface. An
examination of this Figure reveals that the electric current is the strongest on the
patch surface and has a negligible value o n the ground plane. Its equivalent
distribution on the dielectric, i.e. the tangential magnetic field on the dielectric,
is also small. However, it shows some slight increase near B, on the substrate
termination, which is an indication of surface-wave excitation. The distribution
of the magnetic current &?, i.e. the tangential electric field on the substrate, is
shown on the right side of the Figure. It increases progressively from B to C,
indicating strong fringing field near C. The contributions to the antenna radia-
tion are therefore mainly from on the patch and M4 on the substrate.
The surface distributions for the TMl, excitation are shown in Fig. 2.7. Again,
the currents on the ground plane are small, but J' shows stronger values on the
substrate termination near B. Here, both J' and J 4 are strong on the patch and
have rapid variations. The magnetic current A? again increases rapidly from B
to C, near the patch edge. The main radiation zones are similar to the T M , ,
mode case, being the upper patch surface, the dielectric surface near the patch
and its truncated end near the ground plane.
2.3.2 Feed location
For coaxial feeds, the location is usually selected to provide a good impedance
match. Since, we simulate the excitation by an electric dipole we ignore the
impdenace of the feed and investigate the effect of its location on the excitation
efficiency of various modes. Also, different modes have different radiation
patterns and affect the overall antenna pattern at different angular regions. For
this reason, rather than computing the magnitude of the Fourier coefficients of
the currents we compute the peak intensity of their radiation fields. Fig. 2.8
shows the effect of the feed position elon the excitation of the first three modes,
when the patch is resonant at the T M , , mode. The dominant mode has the
strongest excitation efficiency of the other modes, i.e. TM,, and TM,, modes,
 68 Analysis of circular microstrip antennas
increase progressively as the feed moves to the patch edge, but their peak
radiation level is always below - 15 dB. These modes radiate conical beams and
their effect will manifest a t a n angular range near 45' off the z-axis. However,
the T M , , mode has a broad beam and its pattern roll-off is about 4 d B near the
45" angle. Thus, other modes will not affect significantly the radiation of the
T M , , mode for 0 < 0 ,< 45O, and the co-polar patterns will be decided primari-
ly by the dominant mode. Their contributions will be significant for 0 > 45"
and, in particular, for determining the cross-polarisation which, from Fig. 2.8,
shows a peak at about - 25dB range. Here the cross-polarisation is computed
in 4 = 45' plane, in which it maximizes. The results also show that increasing
the substrate height generally increases the excitation efficiency of the other
modes
Fig. 2.8 The effect of the feed position on the excitation efficiency of TM,, mode (Re-
produced from Reference 74 @ 7986 IEEE)
The excitation efficiencies for a patch dominant a t the T M 2 , mode are shown
in Fig. 2.9. The results are plotted for two different substrate permittivities, and
show similar excitations. Again the dominant mode has the strongest excitation,
but its peak radiation increases for Q, > 0 . 6 8 ~and decreases thereafter. The
peak radiations of the other modes have more complex behaviour and minimise
Analysis of circular microstrip antennas 69
at = 0 . 7 5 ~Again,
. their contributions are below the - 25 d B range. However.
since the TM?, mode has a null along the z-axis, the contribution of the T M , ,
mode will cause a minor peak at this location. The results of Figs. 2.8 and 2.9
indicate that the resonance nature of a microstrip patch controls the excitation
of the azimuthal modes, and the resonant modes can easily be excited signifi-
cantly above the adjacent modes simply by selecting an appropriate location for
the feed. With this type of excitation the contributions of the adjacent modes
manifest themselves mainly in the cross-polarisation. They may be ignored if the
antenna cross-polarisation is not the main concern. Also, the substrate permit-
tivity seems to have a small effect on the mode excitation.
Fig. 2.9 The effect of the position on the excitation efficiency of TM,, mode (Reproduced
from Reference 14 @ 1986 lEEE)
We now present a few results for the radiation patterns. Fig. 2.10 shows the
computed patterns for the T M , , mode and Fig. 2.1 1 for the TM,, mode. In both
cases the feed location is selected to maximise the excitation of the dominant
 70 Analysis of circular rnicrostrip antennas
mode. In Fig. 2.10, the T M , , mode is dominant and the radiation pattern is
computed by including four modes, i.e. the T M , , mode along with adjacent
TM,, , TM,, and TM,, modes. The radiation peak is in the broadside with a
significant radiation level behind the ground plane, owing to its finite size. In
Fig. 2.1 1 the TM,, mode is resonant and the radiation patterns are generated by
including the first five modes, i.e. TM,, to TM,, modes. T o examine the accuracy
of the computed results, sample calculations are also compared with experi-
Fig. 2.10 The radiation patterns of a circular patch for the dominant mode excitation
t = 0,021
Ground lane thickness = 0.01 1
ment. Figs. 2.12 and 2.13 show the comparison for the T M , , mode. The
computed and experimental patterns are identical in the upper half plane and
deviate slightly thereafter, owing to the coupling between the antenna and its
support structure.
Analysis of circular rnicrostrip antennas 77
2.3.3 Effect of the substrate permittivity
Increasing the substrate permittivity reduces the patch size and consequently the
size of the radiation zone. One therefore expects to see a broadening of the
radiation pattern. This is shown in Figs. 2.14 and 2.15 for a T M , , mode patch
and in both E and H-planes. Since the ground-plane sizes are all the same, the
antennas have equal sizes. The results show that the broadening is taking place
only in the E-plane. The H-plane patterns are independent of the substrate
permittivity, but show an increase in the level of the back radiation, which is also
evident in the E-plane patterns. Note that, for the selected antenna dimensions,
Fig. 2.11 The radiation patterns of a circular patch for the TM,, mode excitation; data same
as Fig. 2.10
the small permittivity of E, = 2.32 gives nearly symmetric radiation patterns
with small cross-polarisation, Since increasing e, broadens only the E-plane
pattern, the pattern symmetry deterioriates by increasing the substrate permit-
 72 Analysis of circular microstrip antennas Analysis of circular microstrip antennas 73

-180.00 -135.00 -SB.00 -45.08 0.00 45.00 90.00 135.00 180

Angle, degrees

Fig. 2.1 2 Measured and computed data in H-plane of a circular patch excited with a coaxial
probe (Reproduced from Reference 1 4 @ 1986 IEEE) Fig. 2.1 4 E-plane radiation patterns of a circular patch with different substrate permittivities
E, = 2.54, g = 4.5crn, h = 0.159cm, f = 3.2GH,, feed at edge
(Reproduced from Reference 14 Q 1986 IEEE)
-measured
. . . . computed

Angle, degrees
Fig. 2.13 Measuredandcomputed data in E-plane of the case in Fig. 2.12 (Reproducedfrom
Reference 14 @ 1986 IEEE)
---measured. data same as Fig. 2.1 2 Fig. 2.15 H-plane and cross-polarisation patterns of Fig 2.14 (Reproduced from Reference
' ' ' ' computed 1 4 @ 1986 IEEE)
 74 Analysis of circular microstrip antennas Analysis of circular microstrip antennas 75
tivity. This means the antenna cross-polarisation will increase, which is evident
from the results presented in Fig. 2.15. Here, the cross-polarisations are com-
puted in the 4 = 45" plane, where it has the maximum magnitude.

1
11
=2.3:4A , ,

-- -32
-(rh,
-- -
---(
a, pi)
( h . a . f ,)
h, a, pi)
= (
= (
= (
0.02.
0.04.
0.06.
0.1806.
0.1732.
0.1675.
0.045
0.044
0.041
)A
)A
)A
E-plane
Fig. 2.17
' 9
H-plane and cross-polarisation patterns of Fig. 2.16 (Reproduced from Reference
14 @ 1986 IEEE)
---( h . a . p,) = ( 0.10. 0.1590. 0.039 )A
-40
-180 -135 -90 -45 0 45 90 135 180
8
Fig. 2.1 6 E-plane radiation patterns of a circular patch with different substrate heights (Re-
produced from Reference 14 @ 1986 IEEE)

2.3.4 Eflect of the substrate thickness

The bandwidth of microstrip antennas increases by increasing the substrate
height. It is therefore desirable to study its effect on the radiation patterns. For
the TM,, mode patch, representative results are shown in Figs. 2.16 and 2.17.
For h < 0.061 the beamwidth of the H-plane patterns decreases slightly by
increasing h, but increases to some degree in the E-plane. The relationship
reverses for h > 0.06 1. Consequently, the cross-polarisation increases initially
with h, but tends to decrease afterwards. Also, it is interesting to note that the
effect of the substrate may resemble a thinner one with a higher substrate
permittivity, which, from Fig. 2.14 may affect the E-plane patterns significantly.
However, the results of Fig. 2.16 show otherwise, where E-plane patterns are
relatively independent of h. This can be understood by considering the effect of
these two parameters. A higher permittivity reduces the patch size and the extent
of the fringing fields. Consequently, the radiation is due to a narrow magnetic
current ring around the patch periphery, which normally gives asymmetric
radiation patterns. A thicker substrate, on the other hand, does not reduce the
patch size significantly, but extends the zone of the fringing fields, thus resulting Fig. 2.1 8 E-plane radiation patterns of a circular patch with different ground plane diameter
in a broad radiation ring. for the dominant TM,, mode (Reproduced from Reference 14 @ 1986 IEEE)
 76 Analysis of circular rnicrostrip antennas Analysis of circular rnicrostrip antennas 77

2.3.5 Efect of the ground-plane r a d i u ~

Since the ground plane controls the back radiation, its size has a pronounced
effect on the patch radiation pattern. For the T M , , mode case, sample computed
patterns are shown in Figs. 2.18 and 2.19. Again, cross-polarisation are com-
puted in the 4 = 45" plane. In the H-plane, shown in Fig. 2.19, pattern
beamwidth decreases by increasing the ground-plane size. Consequently, the
infinite ground plane has the most rapid pattern roll-off. In the E-plane, on the
other hand, the beamwidth decreases initially by increasing the ground-plane
radius g, but increases for g > 0.71. The infinite ground plane gives the
broadest beam, which approaches -6 dB a t the horizontal plane. The cross-
polarisation therefore increases rapidly by increasing the ground-plane size
from its optimum radius. These results show that the assumption of an infinite
ground plane in approximate analysis of microstrip antennas will have a serious
effect on the correct prediction of the radiation patterns, particularly for angular
ranges beyond 45' off the main beam. The prediction of the cross-polarisation
will, in fact, be an impossible task.

Fig. 2.20 E-plane radiation patterns of a circular patch with different ground-plane diameter
for the dominant TM,, mode (Reproduced from Reference 1 4 @ 1986 IEEE)

Fig. 2.19 H-plane and cross-polarisation patterns of Fig. 2 . 7 8 (Reproduced from Reference
14 @ 1986 IEEE)

For a TM2, mode excitation the corresponding computed results are shown
in Figs. 2.20 and 2.21. The effect of the ground-plane size on the radiation
patterns is similar to the T M , , mode case. The beamwidth in the H-plane
decreases progressively with the ground-plane size, and for the infinite ground
plane the pattern roll-off is the largest. The angle for the peak radiation, which
is around 45O off the z-axis, is, however, almost independent of the ground Fig. 2.21 H-plane and cross-polarisation radiation patterns of Fig. 2.20 (Reproduced from
plane. The E-plane patterns also show similar behaviour to those of the T M , , Reference 14 @ 1986 IEEE)
 78 Analysis of circular microstrip antennas Analysis of circular microstrip antennas 79
mode, and their beamwidth initially decreases by increasing g, but increases for
larger ground planes. For an infinite ground plane the pattern remains relatively
constant beyond the peak of the pattern at about 45'. It is therefore evident that
the assumption of an infinite ground plane will not provide a meaningful pattern
shape for the TM,, mode, where the main feature of the patterns manifest itself
beyond the 45'.

Fig. 2.23 H-plane radiation patterns of the cases in Fig. 2.22

Fig. 2.22 E-plane radiation patterns of TM,,. T M , , and TM,, modes of a circular patch
a = 0.1806 1. g = 0.31, Q, = 0.051, t = 0.021, E, = 2.32,and the ground plane
thickness is zero.

The above results indicate that, the radiation characteristics of various modes
can easily be controlled by the ground-plane size. So far, the total patterns are
shown. It may be desirable to examine the effect of the ground-plane radius on
the mode-excitation efficiencies. To investigate this, the case of the TM,, mode
patch is selected and the mode patterns are computed for two ground-plane
radii of 0.31 and 0.42. This range of ground-plane radius gives the most
symmetric co-polar patterns, with minimum cross-polarisations. The computed
patterns for the 0.32 antenna are shown in Figs. 2.22 - 2.24. The E-plane
patterns in Fig. 2.22 are all in the 4 = 0 plane. The H-plane patterns are,
however, in the H-plane of each mode, being 4 = 90" and 4 = 45' for the
Fig. 2.24 Total radiation patterns of the case in Fig. 2.22
TMII and TM2, modes, respectively. The corresponding results for the 0.41
-E-plane
ground plane are shown in Figs. 2.25 - 2.27. These results indicate that the small --- H-plane
ground plane, with g = 0.31, all non-resonant modes are well below the cross-polarisation.
 80 Analysis of circular rnicrostrip antennas
Analysis of circular rnicrostrip antennas 81

dominant T M , , mode and their peak amplitudes are less than - 30 dB. Increas-
ing the ground-plane radius to 0.41 increases the excitation of both TM,, and
TM,, modes, and their peak amplitudes approach - 17 dB range. The generated
cross-polarisation is therefore larger in magnitude and increases from the - 25
dB level of 0.31 ground plane to more than - 20 dB for the 0.41 case.

0
Fig. 2.25 E-plane radiation patterns of TM,,, TM,,, TM,, and TM,, modes of a circularpatch
a = 0.18061.g = 0.41.p, = 0.051.t = 0.021.6, = 2.32,and the ground plane
th~cknessis zero

Fig. 2.27 Totalradiation patterns of the case in Fig. 2.25

-E-plane
--- H-plane
Cross-polarisation

2.3.6 Effect of the ground-plane thickness

The results of the previous Section indicate that the size of the ground plane
affects the excitation efficiency of non-resonant modes. In some applications
such as reflector feeds, the ground plane may not be infinitesimally thin but may
have a finite thickness. Since the thickness of the ground plane affects the
reflection coefficients of various modes, at its terminal edge, it may also affect
their excitation efficiency. This is investigated here for the TM,, excitation of the
patch and a ground-plane radius of 0.4 1.The results are shown in Figs. 2.28 -
2.30. Again, each of the modes of the patterns are generated in their respective
E- and H-planes. The overall patterns for the co-polarisation are in the q5 = 0"
and 90' planes, the principal planes of the TM,, mode and the cross-polarisation
are in the 45" plane. Comparing these results with those in Figs. 2.25 - 2.27 for
zero ground-plane thickness, one notes that increasing the thickness of the
ground plane to 0.05 1has reduced the excitation efficiencies of the higher TM2,
Fig. 2.26 H-plane radiation patterns of the cases in Fig. 2.25
and TM,, modes. The excitation of the TM,, mode, on the other hand, has
 82 Analysis of circular microstrip antennas I
I Analysis of circular microstrip antennas 83
remained the same. One therefore concludes that the thickness of the ground
plane can also be used to modify the mode excitation. In other words, in
designing microstrip antennas with small ground plane and for a high degree of
mode purity, one must optimise not only the resonant patch size but also the size
and thickness of the ground plane. This is, of course, valid for a given location
of the excitation source, which in practice is determined by the impedance
requirements. However, as shown earlier, the feed location has its own effect on
the mode excitation and can be used as a parameter if warranted.

B
Fig. 2.28 Same as Fig. 2.25 with the ground-plane thickness of 0 . 0 5 1

2.3.7 Circular polarisation

Many techniques have been proposed in the literature to generate circular
polarisation with a microstrip patch antenna. In most methods a geometrical
deformation is used to generate both symmetric and asymmetric modes to cause
a circularly polarised radiation. These methods are convenient to generate
circular polarisation when only one sense of polarisation is needed, and can be
implemented by a single feed. However, when a polarisation diversity is re-
quired, one must use two separate feed arrangements. In such a case, a symme-
tric patch with two separate feed points and an appropriate phase switch will be
sufficient. Also, to generate circularly polarised radiations with a low axial ratio,
one needs an antenna with a nearly symmetric radiation pattern. The results of
previous anlaysis indicated that the pattern symmetry can be controlled by
 84 Analysis of circular microstrip antennas Analysis of circular microstrip antennas 85
modifying the ground-plane size and thickness. To investigate the quality of the
circularly polarised radiation we present a few computed results for circular
patches, fed at two locations, with a 90' phase difference.
The antenna selected is resonant at the TM,, mode and has a ground-plane
radius of 0.41. The circularly polarised E, and E4 components, and their
difference as the cross-polarisation, are shown in Fig. 2.31. These principal-
plane electric vectors represent the envelopes of measurement data with a
rotating linearly polarised test antenna. The actual right- and left-handed
circular polarisation vectors are shown in Fig. 2.32. The peak of the left-hand
vector, which is the cross-polarisation, is below - 20 dB level in the upper half
plane, which is the main region of concern. Its pattern shape is identical to the
difference pattern of Fig. 2.31 and indicates a fairly good circular polarisation.
The computed data for the case of 0.3 1ground plane of zero thickness or for
the 0.41 ground plane of 0.05 2 thickness are not shown here. However, their
results can easily be determined from the co-polar and cross-polar patterns
already. .
provided. Since they have lower non-resonant mode excitations with
reduced levels of cross-polarisation, their generated circular polarisations will
Fig. 2.31 Radiation patterns of a circular patch fed by two dipoles for circular polarisation also be more superior. Thus, the foregoing results indicate that, by a proper
a = 0.18061, g = 0.4i.. e, = 0.051, t = 0.021, E, = 2.32, and the ground plane selection of the feed-point location or the size or thickness of the ground plane,
thickness of 0.1 1 circularly polarised patch antennas with extremely low axial ratio can be design-
----- E-plane
- - - H-plane ed. One only needs to optimise the antenna dimensions properly.
Cross-polarisation.
2.3.8 Effects of a central shorting pin
So far the computed results were presented for a standard circular patch
geometry. The patch dimension is therefore selected to resonate at a particular
mode. However, it is reported in the literature that, by using a central pin to
short the upper patch to the ground plane, one may improve the purity of the
resonant mode. The previous results show that this may not be the case. Since,
with a finite ground-plane size and thickness, the mode excitation can easily be
controlled by modifying their dimensions. An addition of a shorting pin acts as
an extra parameter to control the mode excitation. For a given antenna dimen- ,
sion, one can readily find a pin radius that minimised the non-resonant mode
excitations. This can be done easily with the current method, provided that the
pin does not destroy the rotational symmetry of the configuration. Since the
introduction of the pin increases the resonance size of the patch, perhaps the
most important property of the pin is to control the antenna gain by increasing
the patch size. This may be a useful parameter to use in the design of higher-gain
patch antennas.

2.4 Application 2: Wrap-around microstrip antenna

6'
Fig. 2.32 Co-polar and cross-polar circularly polarised radiation patterns of the example of The wrap-around antenna refers to a microstrip-ring conformal antenna that is
Fig. 2.37 embedded in a missile or cylinder body. Its various configurations are con-
 86 Analysis of circular microstrip antennas
sidered in literature and investigated [I 51, [3]. Analytic solutions, using cylindri-
cal Green's functions, can be obtained [I61 by assuming that the cylinder length
is infinite, so that boundary conditions on its surface can be satisfied. In practice,
however, the missile shape and the location of the antenna on the body influence
Fig. 2.33 Cross-section geometry of wrap around antenna for missile geometry
the antenna radiation patterns. The accurate determination of the radiation
patterns must again be determined numerically. Since the configuration is
rotationally symmetric, it can be analysed readily by the current method. In this
section, the radiation patterns for some of useful geometrical shapes are com-
puted. We have selected this antenna because of its complex shape. Although it
is a microstrip antenna, the conducting patch and the ground plane, i.e. the
cylinder, are not planar and the mode configuration are different. Here, the
modes under the patch form the azimuthal modes of the cylindrical zone and
their excitation is dependent on the cylinder radius. In practice, for a single
exciting source all modes are present, but their magnitudes decrease with the
Analysis of circular microstrip antennas 87
mode number. Thus, when a particular mode must be excited, the exciting
source must eliminate others, which can be achieved by the axial symmetry of
multiple source excitations. For each excited mode, the radiation patterns then
depend on the shape of the cylindrical surface. In most practical applications the
zero-order mode is used and the present study will provide the computed data
for its radiation patterns.
-B
-
Fig. 2.34 Rad~at~on patterns of wrap-around antennas for the zero and 4th order modes.
Zero mode
- - - 4th mode
W, = 1,/4, L, = 4 11, Wd = 0.61. L, = 0.1 A, a = 30', 6, = 2 32, a = 0.2571,
t = 0.021
Fig. 2.33 shows the cross-section of the antenna geometry that is investigated.
It consists of a conducting ring conformal to the cylinder surface and is suppor-
ted by a dielectric substrate, which is embedded in the cylinder. The radius of
the cylinder is selected to be a = 0.257 1,which represents that of a typical small
rocket. The excitation is due to four dipoles at the lower edge of the ring, which
are angularly separated by 90". Since the cylinder radius is small, the selection
of four excitations ensures that the azimuthal pattern is omnidirectional. The
azimuthal symmetry of the excitation means that only 4Kn modes are allowed
to be excited, where K is an integer. All intermediate modes cancel out. To
investigate the mode excitation we select a geometry and compute the radiation
patterns of the first two modes, i.e. K = 0 and 4. The results are shown in Fig.
2.34, where the dominant mode is for K = 0, the zero-order mode. The next
mode for K = 4 is weakly excited and its contribution is below - 30 dB range.
The next higher mode for K = 8 is far too weak to be shown on the plot. These
 88 Analysis of circular microstrip antennas Analysis of circular microstrip antennas 89
results indicate that, for the selected radius of the cylinder, only the zero-order same. This means that the separation distances of the antenna from the cylinder
mode has a significant value and all other modes can be neglected. Also, since ends can be used to control the radiation intensities in the forward and back-
E, is zero for the zero-order mode, all radiation patterns in this section will show ward directions. In the above examples the width of the ring was selected to be
the plots of the E,, component only. Here, we present the dependance of E, o n one half wavelength in the substrate. The effects of ring width on the radiation
the antenna parameters. Fig. 2.35 shows the computed patterns when the patterns are shown in Fig. 2.37, where the ring widths are 0.51,, 0.41, and
radiating ring is located at the base of the cone, i.e. L,, = 0. The computed 0.251,, respectively. Reducing the ring size reduces the broadside radiation.
patterns show the effect of the cone angle o n the radiation patterns, where cc is
the halfcone angle and cc = 90' refers to the geometry of a finite cylinder. Since
the patterns are all similar, they are progressively shifted down by 4 dB to
improve the clarity. The results show that, although the antenna is located at the
upper end of the cylinder, the main beam is in the backward direction and the
radiation towards the cone tip is small. Also, for the selected cone angles, the
effect of the cone is not significant. In this example, the substrate permittivity
is 2.32 and the width of the ring is 0.51,; i.e. one half wavelength in the
substrate. Other dimensional parameters are shown o n the Figure.

e
Fig. 2.35 Radiation patterns of wrap-around antennas with different nose angle a
-a= 90"
--- a = 60"
a = 45" 6
. . . . a = 30"
Fig. 2.36 Radiation patterns of wrap-around antennas with different tail lengths
L, = 4.1 i,
Wd = 0 . 6 L L, = 0.0. W,, = 1 d / 2 ,E, = 2.32, a = 0.2571 From up to down: L, = 5.1 1. L, = 3.1 1 and L, = 2.1 1, respectively. W, = 0.61,
L, = 0.1 1. W, = 1,/2, E, = 2.32, a = 0,2571. a = 30"
For G( = 30, Fig. 2.36 shows the effects of moving the antenna away from
the cone tip and changing the cylinder length, by retaining all other parameters Here, patterns are normalised by the main beam peaks. The quarter-wavelenth
constant. It is evident that increasing the separation from the cone base im- ring is a n end-fire antenna and radiates mainly in the back direction. The effect
proves the forward radiation. Also reducing the antenna separation from the of the substrate permittivity is shown in Fig. 2.38, where the width of the ring
cylinder base reduces the back-lobe level. Otherwise, the pattern shape stays the is again0.5 1,. Increasing the permittivity rapidly reduces the broadside radia-
 90 Analysis of circular microstrip antennas
tion. Finally, Fig. 2.39 shows the effects of moving the antenna towards the
cylinder base. It indicates that, by increasing the antenna separation from the
cone, the forward radiation increases.
Fig. 2.37 Radiation patterns of wrap-around antennas with different patch widths
-w, = 1,/2
--- W, = 0.4 1,
W, = 1,/4
L, = 4.1 1, a = 30'. E, = 2.32. a = 0.2571, t = 0.021
To complete this study, the radiation characteristics of the ring antenna on
a dielectric-coated cylinder with the antenna located symmetrically from two
ends. Fig. 2.41 shows the effect of the cylinder radius on the patterns. The
exciation is again due to four dipoles. For a small cylinder radius the pattern is
Analysis of circular microstrip antennas 97
symmetric and radiates mainly near the broadside direction, but the beam peak
is around 70'. Increasing the cylinder radius moves the beam peak initially
towards 90' and then towards the back direction. The effect of the substrate
thickness is also studied and shown in Fig. 2.42. Decreasing the substrate
thickness moves the beam peak towards the broadside and improves its sym-
metry. The effect of the substrate permittivity is shown in Fig. 2.43. Larger
permittivities broaden the radiation pattern, which is partly
. . due to the reduction
of the effective ring width.
Fig. 2.38 Radiation patterns of wrap-around antennas with different dielectric constants
-E, = 4
--- E, = 2.32
. E , = -10
Since the analytic solution of wrap-around antennas on infinite cylinders is
known, it is useful to generate their numerical solutions as well for comparison.
To handle the problem, the image theory of infinite cylinders is applied to
modify the originual geometry. Fig. 2.44 shows the original and equivalent
problems. In the originual problem a ring antenna, excited by four dipoles, is
supported on a dielectric-coated infinite cylinder. Applying the image theory
one can determine the image of dipoles and the conducting ring inside the
cylinder. The equivalent problem, in the cross-section of the cylinder, thus has
eight dipoles and additional central ring representing the image of the original
ring inside the cylinder. The two rings are separated by the dielectric substrate
and excited by eight dipoles. The numerical solution is obtained for the finite
geometry, i.e. a truncated cylinder of height 0.61, and the corresponding radia-
tion patterns are shown in Fig. 2.45 for different dielectric permittivities. The
 92 Analysis of circular microstrip antennas Analysis of circular microstrip antennas 93
computed results agree well with the published analytical data [16]. The radia-
tion is in the broadside direction, for the selected ring width of 0.5 l,, and the
pattern broadens by increasing the permittivity. These results show that the

Fig. 2.40 Geometry of a wrap-around antenna on a finite dielectric-coated cylinder

e
Fig. 2.39 Radiation patterns of wrap-around antennas with different L, and L,
From up to down (L, = 2.1 1, L, = 3.21). (L, = 1.1 1, L, = 4.1 1 ) and (L, = 0.1 1,
L, = 5.1 1 ) . respectively
W d = 0 . 6 1 ,W p = 1 , / 2 . ~ , = 2 . 3 2 , a = 0 . 2 5 7 1 , a = 3 0 '

numerical method presented here can also be used to investigate infinite struc- Fig. 2.41 Effect of the radius on the radiation patterns
-a = 0,451
tures and the accuracy of the generated results is satisfactory. The usefulness of
a = 0.351
the method, however, is in handling finite geometries where analytic methods --- a = 0.25L.
fail. w, = 1,/2, t = 0.021
 94 Analysis of circular rnicrostrip antennas Analysis of circular microstrip antennas 95

e Fig. 2.44 Geometry of wrap-aroundantenna of an infinite cylinder Left cross-section indica-

Fig. 2.42 Effect of coating thickness on the radiation patterns tes the original problem with four dipoles for excitation; right is the cross-section
-t = 0.051 using the image theory.
. . . . . t = 0.031
- - - t = 0.021

8
Fig. 2.43 Effect of dielectric constant on the radiation patterns
-a, = 10
Fig. 2.45 Effect of the dielectric constant on the radiation patterns
---- E, = 4
a, = 2.32
a, = 2.32 --- a, = 4.0
w, = &I2 -8, = 10
a = 0,251,t = 0.021
a = 0.2571,t = 0.021
 96 Analysis of circular rnicrostrip antennas Analysis of circular microstrip antennas 97
2.5 Application 3: Reflector antenna feeds ground planes, which reduces the blockage. There are, however, a few problems
to be overcome. The back radiation of a microstrip antenna with a small ground
In high-gain applications microstrip antennas may be used as feeds for ryflector plane is high and its bandwidth is normally narrow. The level of the back
antennas [I7 - 201. The merits of microstrip feeds, however, depend oq the type radiation can be reduced by incorporating peripheral chokes. Generally, adding
of application. In symmetric prime focus systems, the feed nomrally blocks the a single quarter-wavelength choke on the periphery of a waveguide feed reduces
central region of the aperture and causes reductions in aperture Jiiciency and its back radiation by about 10 dB [24]. Such a reduction of the back radiation
gain factor, raises the sidelobe levels and causes undesirable diffraction effects. in microstrip antennas is also expected. Additional chokes can further reduce
The rise of the antenna sidelobes also increases the antenna ~ o i s etemperature. the back-lobe level, but at the expense of increased aperture blockage. In
Because the size of the feed depends on the operating frequc .,y and the reflector microstrip feeds one should select one or perhaps two chockes, since a large
f / D , where f and D are the reflector focal length and diameter, the aperature number of chokes will increase the feed size. Microstrip antennas are small in
blockage is most severe in small paraboloid reflectors. A larger feed blocks a size and peripheral chokes will increase their relative size considerably, thus
larger portion of the reflector central region and also requires heavier support eliminating one of their main advantages. The limitation in the microstrip
structures. The latter further blocks the aperture, reducing the reflector perfor- antenna band-width can also be overcome by using any of the many methods
mance and limiting the cross-polarisation performance. A microstrip feed is which are avaiable in literature. However, broadening the bandwidth should not
normally smaller and reduces the central blockage and its subsequent degrading affect the pattern symmetry and shape.
effects. Furthermore, it is low cost and light weight. which reduces the complex-
ity of the supporting structure, and can be integrated readily with its associated
electronics.
The simplicity of microstrip elements also offers additional features with
other reflector configurations. In offset paraboloids and dual reflector systems
a small array can be used to control the reflector illumination and provides a
limited scan capability with reduced sidelobes and coma lobe difficulties. Such
arrays can also be used in non-paraboloidal reflectors, such as spherical reflec-
tors, to improve the aperture efficiency and reduce the abberation. Their main
advantage, however, is in the reduction of the system complexity. Microstrip
feed arrays can be integrated readily with their associated circuitry and electron-
ics, such as the power dividers, phase shifters and amplifiers. Here, we will only
address the design approach and determine the performance levels for wide-
angle feeds that are used with symmetric paraboloids. The array designs and
their associated problems are beyond the scope of this Chapter and are discuss-
ed in subsequent Chapters.
In symmetric paraboloid reflectors the system performance is controlled
primarily by the feed [21, 221. A desirable feed must illuminate the reflector 8
efficiently and cause small spillover. This means that the feed pattern must be
Fig. 2.46 Radiation patterns of a circular rnicrostrip patch, covered by a dielectric thickness
broad within the cone of the reflector and roll off rapidly thereafter. It should ofO.11
also have negligible back radiation. The shape of the feed pattern controls the a = 0.171. g = 0.41, Q, = 0.17 1 and E, = 2.32
reflector efficiency, but with a symmetric system does not affect the reflector -E-plane
cross-polarisation. Thus, for low cross-polarisation the feed must also have a --- H-plane
good polarisation property. From Ludwig's third definition, for minimum - .- .- Cross-~ o l a r
cross-polarisation. the feed pattern must be symmetric and have a unique phase
centre [23]. Here, we present a design example. The previous results for a circular micro-
A circular patch antenna is a good candidate as a feed for a symmetric strip patch indicated that the ground-plane size and thickness can be used as
reflector. Its pattern shape can be controlled readily by the size and thickness of parameters to equalize the E- and H-plane patterns. It was also shown that, for
the ground plane. Fortunately, symmetric patterns are achievable with small a ground plane radius around 0.4 1, the pattern symmetry is satisfactory. This
 98 Analysis of circular microstrip antennas Analysis of circular microstrip antennas 99

A
m

-180 -135 -90 -45 0 45 BO 135 180

e B
Fig. 2.47 Radiation patterns of a circular parch covered by a dielectric
Fig. 2.49 Radiation patterns of a covered circular patch with a conducting collar
g = 0.31,other data same as Fig. 2.46
g = 0.31; other data same as Fig. 2.48
-E-plane
--- H-plane -E-plane
--- H-plane
Cross-polar
Cross-polar

8
Fig. 2.48 Radiation patterns of a covered circular patch with a conducting collar
Data same as Fig. 2.46
-E-plane
--- H-plane Fig. 2.50 Radiation patterns of a two-layer stacked rnicrostr~p(Reproduced from Reference
Cross-polar 20 @ 1986 IEEE Diameters: 0.321.(up); 0,341(bottom; g = 0.4A)
 100 Analysis of circular microstrip antennas
results in a feed of diameter less than one wavelenth, which is considerably
smaller than commonly used waveguide feeds with chokes. To retain the geome-
trical symmetry and to increase the bandwidth one may use a stacked patch
configuration [25]. This means the resonant patch will be covered by another
dielectric-substrate which will alter its resonance frequency and the radiation
pattern. To investigate the latter we have shown the radiation patterns of the
new structure for ground-plane radii of 0.4 and 0.3 i in Figs. 2.46 and 2.47. The
symmetry of the patterns is satisfactory, but not perfect. To improve the
geometrical rigidity we then incorporate a peripheral collar around the substrate
and compute the new radiation patterns. They are shown in Figs. 2.48 and 2.49
for the previous configurations. The addition of the collar limits the radiation
from the substrate termination and considerably improves the pattern sym-
metry. The cross-polarisation is thus improved. Finally, we add the upper patch
to the configuration. The radiation patterns of the final design are shown in
Figs. 2.50 and 2.51, respectively for 0.4 /1 and 0.3 1 ground planes. It is interest-
0 ---- TMol + TMl, TM,, + TM,,.
Fig. 2.51 Radiation patterns of a two-layer stacked microstrip (Reproduced from Reference
20 @ 1986 IEEE)
g = 0.31;other data same as Fig. 2.50
-E-plane
--- H-plane
Cross-polar
ing to note that the pattern characteristics remain unchanged and excellent
pattern symmetries are found for both antenna geometries. The cross-polarisa-
tions for both cases are below - 24 dB, but the back radiations are high. The
latter will be reduced later by incorporating chokes. For the present, we inves-
tigate the sources of the cross-polarisation. We select the case of Fig. 2.48 where
Analysis of circular microstrip antennas 707
the ground-plane radius is 0.42 and compute the cross-polarisation of different
modes. Fig. 2.52 shows the contribution of the first four modes,to the cross-
polarisation. As expected, the TM,, and TM,, modes have the main contribu-
tions. However, since they have different azimuthal dependenmces their com-
bined cross-polarisation is asymmetric. Note that the feed cross-polarisation is
maximum in the 4 = 45" plane and all presented data are in this plane. Fig. 2.52
also shows that adding the contribution of the higher-order modes reduces the
cross-polarisation of the TM,, and TM,, modes. The overall cross-polarisation
is high at about - 24 dB, but within the small angular region of + 45" is below
the - 30 dB range.
-56
-90 -45 0 45 90
0
Fig. 2.52 Effect of different modes on the cross-polarisation of antenna in Fig. 2.48. (Re-
produced from Reference 20 @ 1986 IEEE)
.
. . . TM, + TM,,
TMo, + TM,, + TM,,
To reduce the back radiation we have used two different choke configura-
tions. In Figs. 2.53 and 2.54 the antennas of Figs. 2.50 and 2.51 are incorporated
with a choke behind the ground plane. Their pattern characteristics in the
forward directions remain unchanged, but the back radiation decreases to
around - 24 dB. This type of choke geometry is not as efficient as the peripheral
chokes in reducing the back radiation, but does not increase the feed diameter.
The results with a peripheral choke are shown in Fig. 2.55, where the back
radiation decreases to about - 30 dB range. Adding a second choke behind the
ground plane reduces the back lobe by an additional 2 dB. A second peripheral
102 Analysis of circular rnicrostrip antennas Analysis of circular rnicrostrip antennas 103

-40 A
-180 -135 -90 -45 0 45 90 135 ID0
0 0
Fig. 2.53 Radiation patterns of antenna in Fig. 2.51 with a 214 back choke (Reproduced for Fig. 2.55 Radiation patterns of antenna in Fig. 2.50 with a 114 side choke (Reproduced from
Reference 20 @ 7986 IEEE) Reference 2 0 @ 7986 IEEEJ
-E-plane -E-plane
--- H-plane - - - H-plane
Cross-~olar Cross-polar

Fig. 2.54 Radiation patterns of antenna in Fig. 2.50 with a 114 back choke (Reproduced for 9 (degrees)
Reference 2 0 @ 7986 IEEE) Fig. 2.56 Measured patterns of feed shown in Fig. 2.55
--- E-plane -E-plane
--- H-plane --- H-plane
Cross-polar Cross-polarisation at 45" plane
104 Analysis of circular microstrip antennas Analysis of circular microstrip antennas 105
I

choke can also be incorporated, but may not be necessary since the back lobe Table 2.2a Feed characteristics at the resonant frequency f,
is already low and a new choke will enlarge the feed size. The feed with the Case of Peak cross-pol. Gain Beamwidths., dee
peripheral choke was also fabricated and tested. Its measured E- and H-plane Fig. 0 < 0 < 900 (dB)
'2

patterns, as well as the cross-polarisation in the 4 = 45" plane shown in Fig. 3 dB 10 dB

(dB)
2.56, which are at the band centre frequency of 4.6GHz. The principal plane
patterns agree with the computed data, but the cross-polarisation is higher. The
measured peak cross-polarisation is about - 21 dB, which is about 7 dB higher
than the computed one. It also shows a central peak at the boresight, which
indicates the misalignment of the test set up. Also, within a bandwidth of 500
MHz, i.e. 11%, the co-polar patterns remained nearly symmetric. We expect
that, by improving the fabrication tolerances and a proper alignment of the test
range the measured cross-polarisation should approach the computed ones. The
return loss of the feed was also measured and is shown in Table 2.1. Further
improvement of these return losses can be achieved by modifying the feed-point
location.
Table 2.1 Mearurpd return lower o f l e e d ~
Feed 1, Fig. 2.51 Feed 2, Fig. 2.55 Table 2.2b Data of Table 2.2 at f = 1.05f,
Frequency, GHz Return loss, dB Frequency, GHz Return loss, dB Case of Peak cross-pol. Gain Beamwidths. deg
4.10 9.5 4.30 10.0
4.15 11.0 4.35 11.0
4.20 13.0 4.40 11.5
4.25 14.0 4.45 12.0
4.30 17.0 4.50 12.5
4.35 18.0 4.55 12.5
4.40 16.0 4 60 12.5
4.45 15.5 4.65 12.5
4.50 14.0 4.70 12.0
4.55 11.0 4.75 11.5
4.60 9.0 4.80 10.0

Table 2.2 summarises the performance of the above antennas, where the Table 2 . 2 ~Data of Table 2.2 at f = 0.95f,
beamwidths at 3 dB and 10 dB levels, as well as the peak cross-polarisation, are Case of Peak cross-pol. Gain Beamwidths, den-
provided. To evaluate the performance of these feeds on a reflector antenna the
Fig. 0 < 0 < 90' (dB)
data on the gain factor, spill-over efficiency and the corresponding aperture 3 dB 10 dB
(dB)
angles must be known. They are calculated and shown in Table 2.3. It shows
that the aperture angle varies from 60" to 71 and the gain factor rises from
72.5% to 74.24%. The feed performance is therefore reasonable. The computed
gain factors are somewhat smaller than those of waveguide feeds with a cor-
rugated flange. However, their aperture blockage is small owing to their small
size. Thus, when used on small reflectors, they should provide comparable
performance. These microstrip feeds are, on the other hand, light weight and
easy to fabricate and can readily be integrated with receiving electronics.
 106 Analysis of circular microstrip antennas Analysis of circular microstrip antennas 707
Table 2.3 also shows the location of the phase centre of each antenna cal- the phase centres are inside the dielectric and under the lower patch. In the
culated over its aperture angle, given in column 2 [26]. Their location is meas- remaining cases, all phase centres are outside the dielectric. From these results
ured from the ground plane, i.e. z = 0, and are all positive, indicating that the the following important conclusion can be drawn: In microstrip antennas, in
phase centres are above the ground plane. However, it is interesting to compare Figs. 2.46 and 2.47, the radiation is mostly from the aperture between the patch
and the ground plane. Incorporating the side collar raises the radiation zone to
Table 2.3a Reflector aperture angles, gain factors, spill-over eficiencies and the periphery around the upper patch.
phase-centre locations above the ground plane for various feeds, f = f , The performance of the above antennas listed in Tables 2.2a and 2.3a is also
Case of Fig. Aperture angle, Gain factor, Spill-over efficiency,phase centre, studied as a function of frequency. Within 5% frequency variation the
computed results are shown in Tables 2.26, 2 . 2 ~
and 2.36,2.3c. An examination
deg O h Yo I
of these results reveals that the feed performance remains relatively constant
2.46 68 72.93 85.43 0.075 within the band, in the magnitude of the peak cross-polarisation and the
2.47 71 73.85 86.70 0.0675 reflector gain factor. The feed gain, however, decreases to some degree, regard-
2.48 66 73.14 85.23 0.1 127 less of increasing or decreasing the frequency. This is primarily due to the
2.49 71 73.37 85.50 0.1167 increased excitation of the modes adjacent to the TM,, mode.
2.50 66 72.86 84.16 0.1202
2.51 71 73.47 85.48 0.1212
24 66 74.29 86.24 0 1186 2.6 Concluding remarks
2.55 60 73.83 84.67 0.1827
In this Chapter a general numerical method has been presented that enables one
Table 2.3b Data of Table 2.3 a t f = 1.05f, to solve antenna problems involving conductors and dielectrics. While the
formulation is applicable to arbitrary antenna shapes, the matrix formulation
Case of Fig. Aperture angle, Gain factor, Spill-over efficiency, phase centre, was provided only for axisymmetric configurations. The method was then used
deg Yo % I to investigate the radiation properties of three distinctly different antenna types.
2.46 67 72.1 84.9 0.08 The circular microstrip patch antenna was selected to study the radiation
2.47 71 73.9 88.1 0.09 mechanism of a typical microstrip antenna element. The wraparound antenna
2.48 67 72.3 86.6 0.144 was chosen to show that the method can be used to design or analyse complex
2.49 71 73.50 85.5 0.121 antenna candidates. The last example, i.e. the reflector feed, was included to
2.54 71 73.0 87.0 0.142 show the design steps involving high-precision antennas, with stringent am-
2.55 60 73.83 84.67 0.174 plitude and phase-pattern requirements.
The circular patch antenna was studied in some detail to show the effect of
Table 2 . 3 ~Data of Table 2.3 at f = 0.95 f, various material and dimensional parameters on its radiation patterns. For
instance, the results showed that the ground-plane size has a significant influence
Case of Fig. Aperture angle, Gain factor, Spill-over efficiency, phase centre, on the radiation patterns beyond 45' off the symmetric axis. In addition, it was
deg % % I shown that, by selecting an appropriate ground-plane size, nearly symmetric
2.46 69 73.2 85.8 0.065 radiation pattern with very low cross-polarisation can be obtained. On the other
2.47 71 74.1 86.2 0.068 hand, the feed-point location was shown to influence the excitation of non-
2.48 68 73.34 87.1 0.085 resonant modes, which also contribute to the cross-polarisation. The informa-
2.49 71 73.53 85.6 0.071 tion provided in this Chapter is intended to help the reader in understanding the
2.54 66 74.34 86.8 0.1 14 radiation mechanism of microstrip antennas and use of various parameters to
2.55 66 71.2 83.9 0.416 control them. While the results are computed for circular patch antennas, they
can also be used for square-patch configurations, and with judicious qualifica-
the cases of Figs. 2.46 and 2.47 with the remaining ones, which have the tions, for other patch geometries as well. Also, the results are valid only for
peripheral conducting collar. In the former cases the phase-centre location is single, i.e. isolated microstrip antennas. When antenna elements in a practical
just above the ground plane. Since the total thickness of the dielectric is 0.1 1, array environment are considered, their radiation characteristics will be affected
 708 Analysis of circular microstrip antennas Analysis of circular microstrip antennas 709
by their mutual coupling and the element location within the array. The main 18 HALL, P.S., and PRIOR, C.J.: 'Wide bandwidth microstrip reflector feed element.' 15th
effect of the mutual coupling will manifest itself in the mode excitation, which European Microwave Conference, Paris, 1985, pp. 1029-1044
is considered in a later chapter. The element location within the array affects its 19 PRIOR, C.J., and HALL, P.S.: 'Microstrip disc antenna with short circuit annular ring,'
ground-plane size, and thus its radiation patterns. For large arrays the ground Electron. Lett. 1985, 21, pp. 719-721
20 KISHK, A.A., and SHAFAI, L.: 'Radiation characteristics of a circular microstrip feed,'
plane is large and its effect can be neglected. For small arrays the peripheral Conference on Antennas and Comm., Montech 86, Montreal, Canada, 1986, pp. 89-92
elements will 'see' a smaller ground plane than the central ones and their 21 CLARRICOATS, P.J.B., and OLVER, A.D.: 'Corrugated horns for microwave antennas.'
radiation patterns will be affected accordingly. However, the ground-plane IEE Electromagnetic Wave Series 18 (Peter Peregrinus, 1984)
effect in array applications becomes significant mainly in phased arrays, where 22 RUDGE, A.W., MILNE, K., OLVER, A.D., and KNIGHT, P., (Eds.): 'The hand-book of
the beam must be scanned for low elevation angles. antenna design' Vol. 1. IEE Electromagnetic Wave Series 15 (Peter Peregrinus, 1982)
23 LUDWIG, A.C.: 'The definition of cross-polarisation,' IEEE Trans., 1973, AP-21, pp. 116-
119
2.7 References 24 SHAFAI, L., and KISHK, A.A.: 'Coaxial waveguides as primary feeds for reflector antennas
and their comparison with circular waveguides,' Archiv Fur EIektronik & Ubertragungstchnik,
1985, 39, pp. 8-15
I JAMES, J.R.. HALL, P.S., and WOOD, C.: 'Microstrip antenna, theory and design' (Peter 25 OLTMAN, H.G.: 'Electromagnetically coupled microstrip dipole antenna,' IEEE Trans.,
Peregrinus, 1981) 1986, AP-34, pp. 467-50
2 BAHL, I.J., and BHARTIA, P.: 'Microstr~pantennas' (Artech House, 1980. Dedham, Mass.) 26 SHAFAI, L., and KISHK, A.A.: 'Phase centre of small primary feeds and its effect on the feed
3 JOHNSON. R.C., and JASIK, H. (Eds.): 'Antenna engineering handbook' (McGraw-Hill, performance,' IEE Proc., 1985, 132, pp. 207-214
NY, 1984) 2nd edn., chap. 7
4 HUANG, J.: 'Finite ground plane effect on microstrip antenna radiation patterns,' IEEE
Trans., 1983, AP-31, 649-653
5 LIER, E.: 'Rectangular microstrip patch antennas.' Ph.D. Dissertation, University of Trond-
heim, Norway, June 1982
6 MAUTZ, J.R., and Harnngton, R.F.: 'Boundary formulation for aperture coupling problem,'
Archivfur Elekronik & Ubertrangungstechnik, 1980, 34, pp. 377-384
7 MEDGYESI-MITSCHANG, L.N., and PUTNAM, J.M.: 'Electromagnetic scattering from
axially inhomogeneous bodies of revolution,' IEEE Trans., 1984, AP-32, pp. 797-806
8 HARRINGTON, R.F.: 'Time harmonic electromagnetic fields' (McGraw-Hill, NY, 1961)
Sec. 3-5
9 MAUTZ, J.R., and HARRINGTON, R.F.: 'H-field, E-field and combined field solutions for
conducting bodies of revolution,' Archiv fur Elekrronik & ubertragungstecltnik, 1978, 32, pp.
175-164
10 MAUTZ, J.R., and HARRINGTON, R.F.: 'Electromagnetic scattering from a homogeneous
material body of revolution, Archiv fur Elektronik & Ubertragungstechnik, 1979,33, pp. 71-80
11 ISKANDER, K.A., SHAFAI, L., FRADSEN, A., and HANSEN, J.E.: 'Application of
impedance boundary conditions to numerical solution of corrugated circular horns,' IEEE
trans., 1982, AP-30, pp. 366-372
12 KISHK, A.A.: 'Different integral equations for numerical solution of problems involving
conducting or dielectric objects and their combination.' Ph.D., Dissertation, University of
Manitoba, Winnipeg, Canada, I986
13 YAGHJIAN, A.D.: 'Augmented electric and magnetic-field integral equations,' Radio Science,
1981, 16, pp. 987-1001
14 KISHK, A.A., and SHAFAI, L.: 'The effect of various parameters of circular microstrip
antennas on their radiation efficiency and the mode excitation,' IEEE Trans., 1986 AP-34, pp.
969-977
15 MUNSON, R.E.: 'Conformal microstrip antennas and microstrip phase arrays,' IEEE Trans.,
1974. AP-22, pp. 74-78
16 FONSECA, S.B.A., and GIAROLA, A.J.: 'Pattern coverage of microstrip wraparound anten-
nas.' Int. Conf. on Antennas and Propag., ICAP 83, Norwich, England, P. 1, 1983, pp.
300-304
17 KERR, J.L.: 'Microstrip antenna developments,' Proc. Workshop on Prmted Circuit Antenna
Technology, New Mexico State University, USA, Oct. 1979, pp. 3.1-3.20
 Chapter 3

Characteristics of microstrip patch

antennas and some methods of
improving frequency agility and
bandwidth
K.F. Lee and J.S. Dahele

3.1 Introduction

The develuptnerit of microstrip antennas arose from the idea of utilising printed-
circuit technology not only for the circuit components and transmission lines
but also for the radiating elements of an electronic system. The basic geometry
of a microstrip patch antenna (MPA) is shown in Fig. 3.1. A conducting patch
is printed on the top of a grounded substrate. The shape of the patch can in
principle be arbitrary. In practice, the rectangular, the circular, the equitrian-
gular and the annular ring are common shapes. The feed can be either a coaxial
cable (Fig. 3.la) or a strip line (Fig. 3.lb), which guides the electromagnetic
energy from the source to the region under the patch. Some of this energy
crosses the boundary of the patch and radiates into space. The MPA is a
relatively new form of radiator. In addition to compatibility with integrated-
circuit technology, it offers other advantages such as thin profile, light weight,
low cost and conformability to a shaped surface. The main disadvantage is its
inherent narrow bandwidth (typically a few percent) arising from the fact that
the region under the patch is basically a resonant cavity with a high quality
factor.
The MPA was first proposed by Deschamps in 1953 [I]. However, it was only
in the past 15 years or so that extensive research was devoted to this type of
antennas. This was motivated by the advantages mentioned above, which make
the microstrip antenna an attractive candidate for use in high-speed moving
vehicles such as aircraft, missiles, rockets and communication satellites. By
1981, two textbooks [2, 31 and a special journal issue [4] containing two review
articles [ 5 , 61 were devoted to the subject. A wealth of information is now
available about the microstrip patch antenna as a radiating element, primarily
for the case when the substrate thickness is much smaller than a wavelength.
This Chapter attempts to present some of this information, including some
developments since 1981.
The plan of the Chapter is as follows. In Section 3.2, the cavity model method
of analysing MPAs is described. The basic characteristics of common patch

7 72 Characteristics of microstrip patch antennas
shapes are presented in Section 3.3. Some methods of improving frequency
agility and bandwidth are discussed in Section 3.4. Section 3.5 contains conclud-
ing remarks.
3.2 Cavity model for analysing microstrip patch antennas
3.2.1 Introducrion
Let us consider the basic geometry of a microstrip patch antenna shown in Fig.
3.1, where the z-axis is perpendicular to the plane of the patch. Electromagnetic
waves are first guided along the coaxial or stripline and then spread out under
the patch. When they reach the boundary of the patch, some are reflected and
some radiate into open space. There are two lines of approach to deduce the
radiation fields. One is to find the current distributions along the antenna
structure and then obtain the radiation fields from these current sources. The
other is to find the fields at the exit region. These fields act as equivalent sources,
from which the radiation fields are obtained.
A - conductma
ground biane
feed
a
Fig. 3.1 Microstrip patch antenna with (a) coaxial feed and (b) stripline feed
Under the two approaches mentioned above, a number of methods of analy-
sis have been developed. The main ones are the transmission-line model, the
cavity model and the integral equation method. The transmission-line model in
its original form is limited to rectangular or square patches; however, extension
to other shapes is possible. The integral-equation method is perhaps the most
general: it can treat arbitrary patch shapes as well as thick substrates. However,
it requires considerable computational effort and provides little physical insight.
Both the transmission-line model and the integral-equation method are treated
elsewhere in this Handbook.
In this section, we shall introduce the cavity model. Most of the results
obtained using this model are for electrically thin substrates. For this case, the
cavity model offers both simplicity and physical insight. It also appears to yield
I
i Characteristics of microstrip patch antennas 113
results accurate enough for many engineering purposes. Our discussion will be
restricted to the thin-substrate case. At the time of writing, the extension of the
model to thick substrates is still in the early stage of exploration.
3.2.2 Feed modelling, resonant frequencies and internal fields
i The simplicity of the cavity model can be traced to the assumption that the
I thickness of the substrate is much less than a wavelength, i.e. t < A. The
following observations then follow from this assumption:
I
(i) The electric field E has only the z component and the magnetic field H has
only the transverse components in the region bounded by the conducting patch
and the ground plane.
(ii) The fields in the aforementioned region do not vary with z.
(iii) Since the electric current in the microstrip must not have a component
normal to the edge, it follows from Maxwell's equations that the tangential
component of H along the edge is negligible.
As a result of (i)-(iii), the region between the patch and the ground plane can
be considered as a cavity bounded by electric walls on the top and bottom, and
by a magnetic wall on the side. The fact that assumption (i) does not hold near
the edge because of the existence of fringing fields is taken into account by
extending the edge slightly. This model has long been used in the analysis of
microstrip resonators. However, the application to microstrip patch antenna
appears to be due to Lo et al. (71, Richards e t al. [8] and Derneryd [9, 101.
Writing Maxwell's equations for the region under the patch, we have
E in eqns. 3.2 and 3.3 is the permittivity of the substrate, the permeabtlity of
whlch is assumed to be p,. The current density J in eqn. 3.2 is due to the feed,
which is usually in the form of a coaxial cable or a stripline. The advantages of
the coaxial feed are that the desired impedance characteristic can be obtained
by proper location of the Inner conductor (see Section 3.3) and that the cable
can be placed under the ground plane to minimise coupling between the feed and
the antenna patch. The disadvantage is that the structure is not completely
monolithic and becomes more difficult to produce. This advantage is avoided in
a stripline feed, which, however, introduces some radiation of its own and offers
less flexibility in obtainmg the proper impedance. Usually, a quarter-wave line
with a proper characteristic impedance 1s necessary to transform the antenna
impedance to that of the stripline.
It is appropriate at this point to discuss the modelling of a feed current which
I
 174 Characteristics of microstrip patch antennas
has been used in the development of the cavity model. Consider first a coaxial-
line feed. It can be represented by a cylindrical band of electric current flowing
from the ground plane to the patch, plus an annular ring of magnetic current
at the coaxial opening in the ground plane [ l 11. The latter can be neglected with
little error, and the former can be idealised by assuming that it is equivalent to
a uniform current of some effective angular width 2w, centered on the feed axis.
For example, for a circular patch fed at a distance d from the centre, it is
illustrated in Fig. 3.2 and described by
1 V.E = 0
I Characteristics of microstrip patch antennas 115
be modelled by a :-directed equivalent current source of some effective width
2w. For the circular patch, it is of the same form as eqn. 3.5 except that d is
replaced by the patch radius a.
In both the coaxial and the stripline feed, the z-directed current is assumed to
be independent of z on account of the thinness of the dielectric region. Hence
V .J = -jwe = 0 and eqn. 3.3 reduces to
(3.7)
From eqns. 3.1, 3.2, 3.7 and 3.4, we obtain

where
where k , = is the wavenumber in the dielectric. The electric-wall con-
dition is automatically satisfied since E = Ezfwhile the magnetic-wall con-
dition implies that

on the sides of the cavity.

To solve eqn. 3.8 subject to the boundary conditions, we first find the eigen
functions of the homogeneous wave equation

I subject to the same boundary conditions. Let the eigen functions of eqn. 3.10 be
$I,, and the eigen values of k , be km,.
Assuming the eigen functions to be orthogonal, the solution to eqn. 3.8 is

where * denotes complex conjugate and

In eqns. 3.12 and 3.13, integration is over the domain of the patch.
, Fig. 3.2 Modelling of a coaxial feed by a current ribbon for a circular patch
The resonant frequencies are obtained from setting k: - k i n = 0 and are
given by
The effective angular width 2w is a parameter chosen so that good agreement
between the theoretical and experimental impedances is obtained. Usually, the fmn = kmn/2dZ. (3.14)
arc length 2wd is several times the physical dimension of the inner conductor.
If the feed is a stripline, it can be replaced by an equivalent current source 3.2.3 Radiation field
obtainable from the transverse fields in the plane where the stripline connects the To calculate the radiation field, consider a closed surface S shown in Fig. 3.3.
patch [12]. From uniqueness concepts, only the tangential magnetic field H The top face of S lies just outside the patch and the bottom face lies just outside
backed by a perfect magnetic conductor is needed. Hence the stripline feed can the ground plane. The vertical face of S coincides with the magnetic wall of the
 7 76 Characteristics of microstrip patch antennas
cavity. The fields exterior to S c a n be calculated from the equivalent sources on
Sand their images; the latter is necessary to account for the ground plane, which
is assumed to be infinite in extent for the purpose of analysis. Since the tangen-
tial electric fields on the top and bottom faces, as well as the tangential magnetic
field on the vertical surface, are zero, the only contribution to the equivalent
1
sources are the tangential electric field E, on the vertical surface of the cavity.
Together with its image, the total equivalent magnetic current is
1
Characteristics of microstrip patch antennas 7 17
given by Wood [I41 is more quantitative: t / l o < 0.07 for E, = 2.3 and t / l o <
0.023 for E, = 10 if the antenna is to launch no more than 25% of the total
radiated power as surface waves. More recent work by Fonseca et al. [I51
showed that the size of the patch is also a parameter. For simplicity, we shall
use Wood's criterion and assume that it is satisfied in subsequent discussions.
The dielectric loss P, and the conductor loss P, are calculated from the electric
field under the cavlty, while the radiation loss P, is calculated from the far-zone
electromagnetic field. They are given by

where ri is the unit outward normal.

P, =
1
lo loIEI2 ? sin OdO d 9
2n n

The quantity 6 In eqn. 3.20 is the loss tangent of the dielectric and R, in eqn. 3.21
is the surface resistivity of the conductors.
gro;nd plane
The radiation or antenna efficiency is the ratio of radiated power to input
power:
coax feed
Fig. 3.3 Application of the equivalence principle to calculafe the radiation from a microstrip
patch antenna
In calculating the losses, it is usual to make use of the resonance approx~mation
If the substrate thickness t is much less than the wavelength 1,its effect on the [8].This arises from the observation that, if the frequency is close to the resonant
radiation field is small and M can be assumed to radiate in free space. Using the frequency of a particular mode, the factor l/(k: - k i n )in eqn. 3.11 is very large
free-space Green's function, the electric potential F a t a point r is given by and the contribution to Ez, and hence to the radiation field E, is due mainly to
the resonant-mode term. The electric energy stored at resonance is

where integration is over the perimeter of the patch.

and the total stored energy at resonance is
The fields in the far-zone are given by
H(r) = -joF(r)
E d r ) = IoHd(r)
The effective loss tangent of the cavity, taking into account the three losses
= - IoHdr) P,, PCand P,, is given by
where io= a. 8, = P T / ( ~ ~ T ) (3.26)
3.2.4 Losses in the cavity where
The losses in the cavity under the patch comprise dielectric loss P,, conductor
+ PC + P,.
-
PT = Pd (3.27)
loss PC,radiation loss P, and surface-wave loss e,,..
James et al. [I31 estimated
that surface-wave excitation is not important if t / & < 0.09 for E , = 2.3 and A number of quality Q factors are defined as follows:
t/& < 0.03 for E , 10, where lois the free-space wavelength. The criterion
Dielectric Q: Q, = o WT/Pd = 116 (3.28)
 118 Characteristics of microstrip patch antennas Characteristics of microstrip patch antennas 119

Conductor Q: Q, = wW,/P, = e t (3.29) specific standard. In the case of the microstrip patch antenna which is basically
a strongly resonant device, it is usually the variation of impedance, rather than
Radiation Q: Q, = w WT/P, (3.30) pattern, which limits the standard of performance. If the antenna impedance is
Total Q: QT = wWT/P, = 1/6,8 (3.31) matched to the transmission line at resonance, the mismatch off resonance is
related to the VSWR. The value of VSWR which can be tolerated then defines
In eqn. 3.29, a is the conductivity of the patch and the ground plane.
the bandwidth of the antenna. If this value is to be less than S, the usable
3.2.5 Input impedance
bandwidth of the antenna is related to the total Q-factor by [I I]
The input impedance at the feed of the antenna is given by
Bandwidth (BW) = - % (s 2 1)
Z = R + jX = V/I = E,t/I (3.32) QT$

where E, is the average value of the electric field at the feed point and I is the
total current. For example, if the feed is modelled by eqn. 3.5, we have

and
I = -J(2wd) (3.34)
Unlike the calculations of 6@, it was found that non-resonant modes must be
included in the calculation of input impedance if good agreement between
theory and experiment was to be obtained. The appropriate equation for EZis
therefore eqn. 3.1 1, which contains the factor I/(k: - kin). To keep this term
finite at resonance, the permittivity of the dielectric must be considered complex.
If only the dielectric loss is considered, we have
E = E ~ E , (-
I j6) (3.35)
k: = a 2 & & = k$z,(1 - js), (3.36)
However, Richards et al. [16] found that better agreement with experiment was Fig. 3.4 Typical impedance characteristics around the resonant frequency of a mode
obtained if, instead of the loss tangent of the dielectric, the effective loss tangent
is used. Thus, in calculating the input impedance, eqn. 3.1 1 is modified to read
For S = 2, which is a common standard, the above equation reduces to

where
A typical impedance versus frequency curve is illustrated in Fig. 3.4. There is
usually some reactance at the resonant frequency of a mode due to the contribu-
tions from the non-resonant modes.
3.2.6 VSWR bandwidth
The bandwidth of an antenna is the range of frequencies within which the
performance of the antenna, with respect to some characteristic, conforms to a
While eqn. 3.39 is the most commonly used definition for bandwidth and is
the one we use in this Chapter, it should be pointed out that this is not a
universal definition. For example, some authors define the bandwidth as l/Q,.
3.2.7 Qualitative description of the results predicted by the model
In Section 3.3, the equations presented above will be used to obtain the specific
results for a number of microstrip patch antennas. It is perhaps instructive to
describe here the qualitative features which are common to MPAs. These
features follow naturally from viewing the MPA as a leaky cavity.
720 Characteristics o f microstrip patch antennas Characteristics of microstrip patch antennas 121

(i) There are an infinite number of resonant modes, each characterised by a width b. The electric field of a resonant mode in the cavity under the patch is
resonant frequency. given by
(ii) Because of fringing fields at the edge of the patch, the patch behaves as if Er = E,,cos (mnxla) cos (nnylb) - (3.41)
it has a slightly larger dimension. Semi-empirical factors are usually introduced
where m, n = 0, 1, 2 . . .
to obtain these effective dimensions. These factors vary from patch to patch.
The resonant frequency is
(iii) Each resonant mode has its own characteristic radiation pattern. The
lowest mode usually radiates strongest in the broadside direction. The pattern
of this mode is broad, with half-power beamwidths of the order of 100". where
(iv) For coaxial-fed antennas, the input impedance is dependent on the feed
position. The variation of input resistance at resonance with feed position
essentially follows that of the cavity field. For the lowest mode, it is usually large
when the feed is near the edge of the patch and decreases as the feed moves inside
the patch. Its magnitude can vary from tens to hundreds of ohms. By choosing
the feed position properly, an effective match between the antenna and the
transmission line can be obtained.
jvj Since the cavity under the patch is basicaily a resonaror, the rorai Q and the
impedance bandwidth are dependent on the thickness of the substrate t and its
permittivity E . For low values of E,, the bandwidth generally increases with
increasing t and decreases with increasing E,. This is presumably due to the fact
that the stored energy W, decreases with t and increases with E, while the total
loss P, is insensitive to these changes. However, detailed analysis (Section 3.3)
shows that the bandwidth and Q are complicated functions of frequency,
substrate thickness and the permittivity.
(vi) For thin substrates, the impedance bandwidth varies from less than one to
several percent.
In the next Section, the results obtained by applying the formulas of this
Section to rectangular, circular, equitriangular and annular-ring patches will be
presented.

3.3 Basic characteristics of some common patches

-c47
Fig. 3.5 Geometry for the rectangular patch
A number of canonical patch shapes can be analysed by straightforward ap-
plication of the cavity model. Of these, the rectangular, the circular, the equi- Eqn. 3.42 is based on the assumption of a perfect magnetic wall. To account for
triangular and the annular ring are the common shapes used in practice. They the fringing fields at the perimeter of the patch, the following empirical formula
will be considered in detail in this Section. An example comparing the charac- can be used for the effective dimensions [7]:
teristics of these patches will be given in Section 3.4, while other patch shapes
will be briefly mentioned in Section 3.5. a, = a + 112 (3.44)
b, = b + 112 (3.45)
3.3.1 The rectangular patch A more accurate but lengthy formula, suggested by James et al. [3] is

3.3.1.1 Introduction: The rectangular patch (Fig. 3.5) is probably the most
commonly used microstrip antenna. It is characterised by the length a and the
 122 Characteristics of microstrip patch antennas Characteristics of microstrip patch antennas 123
where

where f, is the resonant frequency given by eqn. 3.42. TMol mode

Eqn. 3.46 is found to yield resonant frequencies which are within 3% of
experimental values, while the perfect magnetic-wall model gives errors up to
20%.
The far-field, losses and Q, and input impedance can be calculated by apply-
ing the equations of Section 3.2. Since for the rectangular patch they are well
documented [ 3 , 7 , 8 ] ,they will not be reproduced here. In what follows, we shall
present numerical results based on these equations to illustrate the basic charac-
teristics of the rectangular patch. Experimental results will also be given.

3.3.1.2 Illustrative results: We present in this Section numerical, and in some

cases experimental, results illustrating the basic characteristics of the rectan-
gular patch antenna.

(a) Magnetic-current distribution TMIO mode

The electric-field and magnetic-surface-current distributions on the side wall for
b
TM,,, TM,, and TM, modes are illustrated in Fig. 3.6. For the TM,, mode, the
magnetic currents along b are constant and in phase while those along a vary
sinusoidally and are out of phase. For this reason, the b edge is known as the
radiating edge since it contributes predominantly to the radiation. The a edge
is known as the non-radiating edge. Similarly, for the TM,, mode, the magnetic
currents are constant and in phase along a and are out of phase and vary
sinusoidally along b. The a edge is thus the radiating edge for the TM,, mode.

(b) Radiation patterns

The modes of the greatest interest are TM,, and TM,, . However, the TM,, mode
has also received some attention. These three modes all have broadside radia-
tion patterns. The computed patterns for a = 1.56 and two values of E, are TMZO mode
shown in Figs. 3 . 7 ~ - fIn
. the principal planes, the TM,, and TM,, modes have
similar polarisations while that of the TM,, mode is orthogonal to the other two.
Fig. 3.6 Electric field and magnetic-surface-current distributions in walls for several modes
As will be discussed in Section 3.4.3.2,the TM,, and TM,, modes can be utilised of a rectangular microstrip parch antenna
to operate the rectangular patch as a dual-frequency antenna. The patterns do a TM,, mode
not appear to be sensitive to alb or t. However, they change appreciably with b TM,, mode
E, . c TM,, mode
124 Characteristics of rnicrostrip patch antennas Characteristics of rnicrostrip patch antennas 125
-9

OdB -10 -20 -30 -30 -20 -10 OdB

-80'
, L, , ,J,or

-30
,

-90'
OdB -10 -20 -30 -30 -20 -10 OdB

-8 l Eel
126 Characteristics of rnicrostrip patch antennas Characteristics of microstrip patch antennas 127

a = 1.5b. For E, = 2.32 and E , = 9.8, the results for three thicknesses are given.
In general, the efficiency increases with the thickness of the substrate and
decreases with increasing E, .
In using these curves, the criterion given by Wood for the avoidance of
excessive surface-wave excitation mentioned in Section 3.2.4 should be kept in
mind. For E, = 2.32 and E, = 9.8, the cut-off frequencies (below which the
surface wave is less than 25% of total radiated power) are 21/t and 6.9/t GHz,
respectively, where t is in millimetres. These correspond to 6.60 GHz for E, =

OdB -10 -20 -30 -30 -20 -10 OdB

Fig. 3.7 Helatlve fleld patterns for a rectangular patch with alb = 1.5, f,, = 1 GHz, and (I)
E, = 2.32, t = 0.318, 0.159, 0,0795cm; (ii) E, = 9.8, t = 0.127, 0.0635, 0.0254 cm
( a ) TM,,. 4 = O'
( b ) TM,,. + = 90"
( c ) TMOl , 4 = 90"
( d ) T M , , . 4 = 0"
( e ) TM,,, 4 = 90'
( f ) TM,,.4 = 0"
( g ) TM,, , 6, = 2.32

The patterns of most of the other modes have maxima off broadside. For
example, those of the T M , , mode are illustrated in Fig. 7g.
Fig. 3.8 shows the computed and measured radiation patterns of the TM,,
and TM,, modes obtained by Lo et al. [7] for a rectangular patch with a =
11.43cm, b = 7.6cm, E, = 2.62 and t = 0.159 cm. Both Eo and E4 were meas-
ured in each of the two cuts, q5 = 0" and q5 = 90". It was found that one
component of polarisation was negligible when compared to the other in each
case and is not shown. Fig. 3.8 Theoretical (x) and measured (solid or dashed line) radiation patterns in 4 = 0' and
C
$ = 90' planes of a rectangular patch antenna with a = 17.43 cm, b = 7.6 cm,
(c) Radiation eficiency E, = 2.62, t = 0.159 cm. (Reproduced from Reference 7 p. 140 @ 1979 IEEE)
( a ) and ( 6 ) at resonant frequency 8 0 4 MHz of (1, 0 ) mode
Let us first obtain some idea of the relative magnitudes of the power dissipated
( c ) and ( d ) at resonant frequency 1187 MHz of (0, 1) mode
in the metal, the power dissipated in the dielectric, and the antenna radiation
efficiency. These are described by the quantities P,/P,, P,/P, and P,/P,, respec-
tively. The cases of (i) a = 1,5b, E, = 2.32, t = 0.159 cm and (ii) a = 1.56, 2.32, t = 0.318cm and 5.43GHz for E, = 9.8, r = 0.127cm. For the other
E, = 9.8, t = 0.0635cm are illustrated in Fig. 3.9 for the TM,, mode. It is seen cases, the cut-off frequencies occur beyond 10 GHz.
that, for both E, = 9.8 and E, = 2.32, the loss due to the conductor is larger than
the loss due to the dielectric. The ratio P,/P, decreases rapidly as frequency (d) Directivity and gain
increases. The directivity D of an antenna is defined as the ratio of power density in the
The radiation efficiency e = P,/P, of the TM,,, TM,, and TM,, modes as a main beam to the average power density while the gain G = eD. For a rectan-
function of resonant frequent is shown in Figs. 3.10~-c for a patch with gular patch with a = 1.5b, the directivities as a function of resonant frequency
128 Characteristics of microstrip patch antennas Characteristics of microstrip patch antennas 129

for the three broadside modes are illustrated in Fig. 3.1 1. The directivity of the
TM,, mode is largest and that of the TM,, mode the smallest. It is not sensitive
to substrate thickness and resonant frequency. The gain, on the other hand,
increases with resonant frequency in the manner shown in Figs. 3.12L~-c.

".,
resonant frequency (GHz)
a

resonant frequency (GHz)

F i g . 3.9 Metallic (P,), dielectric (P,) and radiation (P,) losses for the TM,, mode as a
function of resonant frequency for (a) a = 1.56. c, = 2.32,t = 0.159cm and (6)
a = 1.56, e, = 9.8, t = 0.0635cm

(e) Total Q and bandwidth

For the three broadside modes, the variation of total Q with resonant frequency
is shown in Figs. 3.13~-c for the case a = 1.56. The bandwidth, as defined by
eqn. 3.43, is shown in Figs. 3.14~-c.It is seen that the TM,, mode has the lowest
Q and therefore the largest bandwidth compared to the other two modes. For
E, = 2.32, the bandwidth for a given mode increases with substrate thickness
except for frequencies below about 0.7 GHz. The behaviour is more complicated
for E, = 9.8. For this case, there appears to be a range of frequencies for which
a thinner substrate actually yields a larger bandwidth. resonont frequency (GHz)

(f) Input impedance

Richards et al. [8] have reported calculated and measured values of the input
impedance of a coaxial-fed rectangular patch with e, = 2.62 and t = 0.159cm.
The Smith chart plot for the TM,, mode is shown in Fig. 3.15 for three feed
positions. The variation of the input resistances at resonance of the TM,, and
TM,, modes is shown in Fig. 3.16. It is seen that the input resistance is largest
Fig. 3.10 Radiation efficiency as a function of resonant frequency for a rectangular patch
witho = 5 . 8 x 107S/m,6 = 0.0005, a = 1.5band(i) e, = 2.32,t = 0.318.0.159,
0.0795 cm; (ii) 8, = 9.8, t = 0,127, 0,0635, 0,0254 cm
 130 Characteristics of rnicrostrip patch antennas Characteristics of rnicrostrip patch antennas 137

T"03 a-1 5 b

resonant frequency (GHz)

C

when fed at the edge of the patch, but it can attain the convenient value of 50 R
when the feed position is chosen properly.
More detailed theoretical results for the input resistance at resonance are
shown in Figs. 3.17~-c for the TM,,, TM,, and TM,, modes for E , = 2.32,
t = 0.159cm and a/b = 1.5. The resistances are plotted as a function of feed
position parametric in the resonant frequencies. It is seen that the values vary
somewhat with the resonant frequency. It should also be noted that, for the
TM,, mode, the variation with feed position is not a monotonically decreasing
function, which is the case for the TM,, and TM,, modes.
It is clear from these illustrations that, for a coaxial feed, matching the
antenna impedance to the transmission-line impedance can be accomplished
simply by putting the feed at the proper location. There is less flexibility in the
case of a stripline feed. In this case, a quarter-wave transformer may be added 0 2 L 6 8 10
to effect matching. Alternatively, an insert into the patch can be made. resonant frequency (GHz)
To conclude this Section, we point out that a rectangular patch with a single
feed produces linearly polarised radiation. If circular polarisation is desired, the Fig. 3.11 Directivities (absolute value) of the TM,,, TM,, and TM,, modes as a function of
most direct approach is to use two feeds located geometrically 90' apart and resonant frequency for a rectangular patch with a = 7.56 and (ii) 6, = 2.32, t =
with a relative phase shift of 90". This arrangement excites two orthogonal 0,318, 0.159, 0.0795crn; (ii) E, = 9.8, t = 0.127, 0.0635, 0,0254 cm
modes, each providing a linearly polarised wave at right angles to each other
and at phase quadrature.
Circular polarisation can also be produced by a nearly square patch, where a and b must satisfy a/b = 1 + l/Q and the feed must be located along the line
one pair of sides resonates at a slightly higher frequency than the other pair. If y = f bxla. The plus and minus signs produce left-hand and right-hand cir-
the phase difference at the centre frequency between the pairs of sides is 4 2 , cular polarisations, respectively, in the direction normal to the patch.
circular polarisation results. Lo and Richards [17, 181 have shown that the sides
Characteristics of microstrip patch antennas Characteristics of microstrip patch antennas 133

I T"l~ a11.5b

resonant frequency (GHz)

a

resonant frequency (GHz)

C

Fig. 3.12 Gain (absolute value) as a function of resonant frequency for a rectangular patch
witho = 5.8 x 1O7Slm.6= 0.0005.a = 1.5band(i)c, = 2.32, t = 0~318.0.159,
0.0795cm; (ii) c, = 9.8, t = 0.127. 0,0635, 0.0254cm
(a) TM,,
( 6 ) TMo,
( c ) TMo3

resonant frequency (GHz)

b
134 Characteristics of microstrip patch antennas Characteristics of microstrip patch antennas 735
T"03 a=1.5b

resonant frequency (GHz) resonant frequency (GHz)

a
Fig. 3.13 Total Q factor as a function of resonant frequency for a rectangular patch with
rr = 5.8 x 107Slrn, S = 0.0005, a = 1.56 and (i) E, = 2.32, t = 0.318, 0.159,
0.0795cm; (ii) E, = 9.8, t = 0.1 17, 0.0635, 0,0254crn
( a ) TWO
( b ) TMOI
( c ) TM03

3.3.2 The circular patch

3.3.2.1 Introduction: The geometry of the circular patch or disc (Fig. 3.18) is
characterised by a single parameter, namely, its radius a. In this respect, it is
perhaps the simplest geometry since other shapes require more than one par-
ameter to describe them. The mathematical analysis, however, involves Bessel
functions.
The electric field of a resonant TM, mode in the cavity under the circular
patch is given by
0.318 crn
E, = E,,J, (k,, Q)cos n$ (3.49)
where Q and I) are the radial and azimuthul co-ordinates, respectively. E, is an
resonant frequency (GHz) arbitrary constant, J, is the Bessel function of the first kind of order n and
b
k, = Xm/a (3.50)
136 Characteristics of microstrip patch antennas Characteristics of microstrip patch antennas 137

resonant frequency (GHz) resonant frequency (GHz)

C
a
Fig. 3.14 Bandwidth as a function of resonant frequency for a rectangular patch with a =
5 . 8 x 107Slm. 6 = 0,0005. a = 7.56 and (i) E, = 2.32, t = 0.378, 0.759,
TMOI a:l 5b
30 0.0795cm; (ii) E, = 9.8, t = 0.1 17, 0.0635, 0,0254 cm
( a ) TMto
( b ) TMo,
( c ) TMo3

In eqn. 3.50, &, are the roots of the equation

where differentiation is with respect to x.

The first five non-zero roots of eqn. 3.51 are shown in Table 3.1. The resonant
frequency of a TM,,, mode is given by

where c is the velocity of light in free space.

Eqn. 3.52 is based on the assumption of a perfect magnetic wall and neglects
the fringing fields at the open-end edge of the microstrip patch. To account for
these fringing fields, an effective radius a,, which is slightly larger than the
physical radius a, is introduced [19]:

10 10
resonant frequency (GHz)

b
 138 Characteristics of microstrip patch antennas
Eqn. 3.53 is obtained by considering the radius of an ideal circular parallel-plate
capacitor which would yield the same static capacitance after fringing is taken
into account. Although the result is borrowed from the static case, it appears to
yield theoretical resonant frequencies which are within 2.5% of measured
values.
Characteristics of rnicrostrip patch antennas 739
3.3.2.2 Illustrative results: In this Section, we present graphical illustrations
of the circular microstrip patch antenna. They include the magnetic-current
distribution, radiation patterns, efficiency, directivity and gain, bandwidth and
Q,and input impedance.

x x measurement
- theory
T6 feed polnts
w
N

-
C

rectangular microstrip antenna

substrate rexolite 2200
1116"nominal th~ckness o o measured locus 0 0.1 0.2 0.3 0.4 0.5
x x computed locus
increment: 5 M H z y'lb for x'=5.33cm for (0.1) mode
(increasmg frequency is clockwise) x'lb for y'=3.81 cm for (1.0) mode
a b (x'y') IS location of feed point

Fig. 3.15 Impedance of the TM,, mode of a rectangular patch antenna of a = 1743cm.
b = 7.6 cm, 6, = 2.62, t = 0.759 cm. (Reproduced from Reference 8 p. 3 9 @ 198 1
IEEE)
a Feed placement for impedance measurements
b Comparison of measured (0) and computed (x) impedance loci

Table 3.1 The first five non-zero roots

of J, ( x ) = 0

Fig. 3.1 6 Variation of resonant resistance with feed position of the TM,, and TM,,, mode in
a rectangular patch antenna with a = 17.43cm. b = 7.62cm. &, = 2.62,
t = 0.159cm. (Reproduced from Reference 8 p. 42 @ 1987 IEEE)

(i) Magnetic current distribution

As with the rectangular patch, the expressions for the far field, losses and Q, The magnetic-current distribution around the edge of the disc for the nmth
input impedance etc. are well documented [2] for the circular patch and will not mode is proportional to cosn($ - a).This is illustrated in Fig. 3.19 for n = 0,
be reproduced here. The characteristics obtained from these equations will be 1, 2 and 3. It is independent of I(/ for modes with n = 0 and undergoes three
illustrated in the next Section. sinusoidal periods for modes with n = 3.
 140 Characteristics of microstrip patch antennas Characteristics of microstrip patch antennas 74 1

Fig. 3.1 7 Variation of resonant resistance with resonant frequency for a rectangular patch
antenna with alb = 7.5,E, = 2.32,t = 0.159cm
( a ) TMlo
( b ) TMoi
( c ) TM03
y'lb

(ii) Radiation patterns

Lo and co-workers [7] were among the first to obtain theoretical and experi-
mental radiation patterns of the circular disc. Some of their results are shown
below. In their experiment, a disc with a radius of 6.7cm and a dielectric
thickness of 1.5 mm was used. The relative permittivity was 2.62. For this disc,
the first resonance, i.e. that of the TM,, mode, was at 794MHz. The second
resonance, i.e. that of the TM,, mode, was at 1324MHz. The calculated and
measured radiation patterns in the Q, = 0" and Q, = 90" planes are shown in
Fig. 3.20. Qualitatively, the calculated and measured results showed reasonable
agreement. No attempt was made to take into account the effects of a finite
 142 Characteristics of microstrip patch antennas
ground plane in the theory, which was about 12 wavelengths on a side for
frequencies near 794 MHz.
The radiation patterns of the higher-order modes (0,2) and (3,O) also exhibit
nulls in the broadside direction. Since the higher-order modes are seldom used
in practice, only the characteristics of the lowest mode will be illustrated in
subsequent sections.
(iii) Eflciency
The radiation efficiency for the lowest mode TM,, as a function of resonant
frequency for various dielectric substrate is shown in Fig. 3.21. It is seen that the
efficiency increases with increasing substrate thickness and decreasing dielectric
constant.
Characteristics of microstrip patch antennas 143
(iv) Directivity and gain
The directivity versus resonant frequency is plotted in Fig. 3.22. A circular patch
on an alumina substrate has a directivity of about 3.5, which is almost indepen-
dent of substrate thickness and resonant frequency. If the substrate is Duroid,
it has a maximum directivity of about 5.3, which decreases with increasing
resonant frequency and dielectric thickness.
The gain of the lowest mode is illustrated in Fig. 3.23.
Fig. 3.18 Geometry of the circular patch antenna
(v) Total Q and bandwidth
The variation of total Q with resonant frequency for the lowest mode is shown
in Fig. 3.24. The bandwidth, as defined by eqn. 3.40, is shown in Fig. 3.25. Its
dependences on substrate thickness and E, are similar to the rectangular patch.
(vi) Input impedance
Richards et al. have reported calculated and measured values of the input
impedance of a coaxially fed circular patch as a function of radial feed position.
This is shown in Fig. 3.26 for the TM,, mode. It is seen that the input resistance
is largest when fed at the edge of the patch, but it can attain the convenient value
of 50R when the feed location is chosen properly.
 744 Characteristics of microstrip patch antennas I 745
Characteristics of microstrip patch antennas
Dahele and Lee [20] have studied experimentally the effect of substrate
thickness on the input im~edanceof a coaxially fed circular patch. Their results
are shown in Fig. 3.27.

Fig. 3.20 Theorettcal (x) and measured radtatlon patterns tn the q5 = 0' and q5 = 90' planes
of a circular patch w ~ t radius
h a = 6 . 7 5 ~ 1E,, = 2.32 (Reproducedfrom Reference
7 p. 141 @ 1979 IEEE)
(a) and ( b ) At 794 MHz of mode ( 1 . 1 )
( c ) and ( d ) At 1324 M H z if mode ( 2 , l )

Fig. 3.19 Surface magnetic-current distribution of the nmth mode in the circular patch
antenna

resonant frequency (GHz)

TO conclude this Section, We point out that, as in the case of the rectangular Fig. 3.21 Radiatton efficiency versus resonant frequency for the TM,, mode of the ctrcular
patch, feeding the circular patch at a single point results in linearly polarised patch with g = 5.8 x 1O7Slm, 6 = 0.0005 and (I) E, = 2.32, t = 0.318, 0.159,
radiation. Circular polarisation on boresight can be obtained using two feeds at i 0.0795cm; (;I) c, = 9.8, t = 0.127, 0.0635, 0.0254cm
 146 Characteristics of rnicrostrip patch antennas Characteristics o f rnicrostrip patch antennas 147

resonant frequency (GHzI

Fig. 3.22 Directivity versus resonant frequency for the TM,, mode of the circular patch: (i)
e, = 2.32, t = 0~318,0~759,0~0795cm;(ii) e, = 9.8, t = 0~127,0~0635,0~0254cm

Fig. 3.24
1
0.1 resonant frequency ( G H ~ )

Total Q versus resonant frequency for the TM,, mode of the circular patch with
a = 5.8 x 107S/m, 6 = 0.0005 and (i) E, = 2.32, t = 0.318, 0.759, 0.0795cm;
(ii) e, = 9.8, t = 0.727, 0.0635, 0.0254cm

0.1 10
01 1
resonant frequency (GHzl resonant frequency (GHz)

Fig. 3.23 Gain versus resonant frequency for the TM,, mode of the circular patch with a = Fig. 3.25 Bandwidth versus resonant frequency for the TM,, mode of the circular patch with
5 8 x 707S/m, 6 = 0,0005 and (i) E, = 2.32, t = 0,318, 0.159, 0,0795cm; (ii) a = 5.8 x 107S/m,6 = 0.0005 and (ii) E, = 2.32, t = 0.318, 0.159, 0~0795cm:
6, = 9.8, t = 0.127, 0,0635, 0.0254cm (ii) e, = 9.8, t = 0.127, 0.0635, O.0254cm
 748 Characteristics of microstrip patch antennas 749
Characteristics of microstrip patch antennas

- +
x x computed p a n t
2 5 M H z Increment (1, = t+b, and (1,, 90' and excited in phase quadrature. Alternatively, a slightly
measured locus
5.0 MHz Increme
elliptical patch with the right amount of ellipticity and fed at theappropriate
location can produce circularly polarised waves. This will be discussed further
in Section 3.3.6.

u computed polnt
Increment 5 OMHr

260r

2~0. 0 measured

220 -
+ Calculated P
200 -
--
C 180-
/
/

- i
-,a5 120-

loo- /
/
P
0 /
; 80-
60 - /
d
/

, , , , , ,
6 8
radlal source locat1on.d
Fig. 3.26 (a) Calculated and measured TM,, mode input impedance loci for several radial
feed locations. (6) Variation of TM,, resonant input resistance with radial feed
position for a circular patch with a = 6.7~171
on Rexolite 2200 substrate (E, = 2.62)
of thickness O.159cm (Reproduced from Reference 8 p. 41, @ 1987, IEEE)
Fig. 3.27 Measured real part (R) and imaginary part (X) of input impedance as function of
ial8 frequency for the TM,, mode of a circular patch with a = 6.8cm. 8, = 2.32 and
three dielectric thicknesses. The feed is at d = 6.5 cm (Reproduced from Reference
20 p. 359 @ 1983 IEEE)
3.3.3 The Equitriangular patch
Several triangular patch shapes are amendable to analysis by the cavity model.
These include the 45"-45"-90, 30'-60"-90, and the 60"-60"-60" equitriangular
 150 Characteristics of microstrip patch antennas Characteristics of microstrip patch antennas 151

(equilateral triangular) patches. However, unlike the rectangular and circular
patches which have been studied extensively, there are only a handful of inves-
tigation on the triangular patches [2, 21-24]. In this Section, the equitriangular
patch is treated in detail. The geometry, for the case of a coaxial feed, is shown
in Fig. 3.28. The presentation follows closely that of Luk et al. [24].
There were two suggestions for accounting for non-perfect magnetic wall effects.
Helszain and James [I71 suggested that the side length a in eqn. 3.55 be replaced
by the effective value
a, = a + t(~,)-'" (3.56)
On the other hand, Bahl and Bhartia [2] proposed that, in addition to a,
replacing a in eqn. 3.55, E, should also be replaced by the effective value

The question of which suggestion is appropriate can only be determined by

comparison with experiment. It was shown by Dahele and Lee [23] that the
suggestion of Helszajn and James [21] yielded much better agreement, and
-x top vlew
consequently, in the cavity-model theory of the equitriangular patch, the side
length a will be replaced by its effective value a, but E, will not be replaced by
E,. This is similar to the correction used in circular patch antennas.

(ii) Internal and radiation fields of the coaxial-fed antenna

In this Section, we present the formulas for the internal and radiation fields of
the coaxial-fed equitriangular patch antenna. These take the form of double-
infinite series comprising the various modes excited. If the characteristics of a
z particular resonant mode are desired, they can be obtained by examining the

patch, t term corresponding to this mode.

Let the coaxial feed be located on the bisector line at a distance d from the
tip of the triangle (Fig. 3.28). The x and y co-ordinates of the tip are - a/,and
0, respectively. Following usual practice, the feed is modelled by a uniform
current ribbon of some effective width 2w along the x-axis:

where
Fig. 3.28 Geometry of the equitriangular patch antenna

3.3.3.1 Formulas based on the cavity model: The solutions for the fields in an
equitriangular waveguide with perfect electric walls have been described by
Schelkunoff 1251. It follows from the duality principle of electromagnetism that The internal field for TM modes inside the cavity is assumed to be z-directed and
the TM modes with perfect magnetic walls are the same as those of TE modes is given by
with perfect electric walls. Starting with the solutions given in Schelkunoff, we m m 2nIx' 2n(m - n)y 2nmx'
obtain the following results for the equitriangular patch antenna. E, = jog 1 1 C,,,,,
"=om=" [cos -
,acos 3a
+ cos -$0

(i) Resonant frequencies

The formula for the resonant frequencies of z-independent TM modes satisfying
the perfect magnetic wall boundary condition is
X COS
2n(n - 1)y
3a
where the integers rn, n and I satisfy
2nnx'
+ cos -
a,
COS
2741 - m)y
3a I (3.60)
 752 Characteristics of microstrip patch antennas Characteristics of microstrip patch antennas 753
and
x' = x + a/$

4a
k,, = - (m2
3a
+ mn + n2)'l2 (3.63b)

sin
3
+ cos -
3
""I 111-

C",, =

I1
6
ifm=n=O
if ( m = 0 and n # 0 ) or ( m # 0 and n = 0 ) or ( m = n # 0 )
12 i f m # n # O

jo(x) = sin x/x

(3.65~)
(3.656)
In eqn. 3.60, the conditions on the indices m , n are m > n 3 0 because the eigen
functions ,$ = $, .
The far-zone electric field at a point P(r, 0, 4 ) is given by
Eo = -jw(,(FX cos 0 cos 4 + Fycos 0 sin 4 ) (3.66a)
E, = -jwl;,(-FXsin4 + Fycos$) (3.666)
where F, and Fy are the electric potential components given by
J X ~
+ cos 3 ,pi(/- m)2b2-
+jfi(a/2)z2

(1 - m)b sin 2(1 - m)n + cos

3 3

+ COS
n)nl 11
&%/2)x2
3

+ cos 2(m 3- - - JXZ

$[(m - n)2b2- x:]
~fi(012)~~

x T[ ( m - n)b
sin
2(m - n)n
3 + cos 3
 754 Characteristics of microstrip patch antennas Characteristics of microstrip patch antennas 755
3.3.3.2 Illustrative results
(i) Radiation patterns
The radiation patterns are not sensitive to the resonant frequency or the size of
( m - n)n
(m - n ) sin ---cos ( 7 4 3 ) - v the patch. The results to be presented are obtained using sidelength a = 10cm.
3 - For this length, the resonant frequencies of the first five modes for E, = 2.32,
X

(n
( m - n)n
COS -

-
3
sin (nv/3) +

(n - I)n
I (- 1)"3a
n[(n - 1)' - vZ]

I ) sin --- cos ( 7 4 3 ) - v

3 This feature is different from the circular and the rectangular patches, for which
x
( n - I)n .
cos ----- sin (nv/3)
3 I
t = 0.159 cm and E, = 9.8, t = 0.635 mm are shown in Table 3.2. The field-
strength patterns are shown in Figs. 3.29~-j.Notice that, in the $ = 0" plane,
only the component E, is present. In the 95 = 90" plane, however, both the Ee
and the E+ components are present, except in the broadside direction (0 = 0").
the principal-plane patterns contain only either E, or E+, but not both.
+ n[(l (- 1)"3a
- m)* - v2]

(I - m ) sin ( Icos (nv/3) - v

3

x cos ------
(' - m)n sin (nv/3)]}] (3.68)
3
In eqns. 3.67 and 3.68,

XI = kosin 0(cos $ + sin $/$)

x2 = kosin O(cos $ - sin dl$) (3.73)

(iii) Input impedance

The input impedance seen by the coaxial feed located at a distance d from the
tip of the triangle is given by

Z = R+jX= -jwhCC- 4 [
' cos ( 2-n l d ) lo
, (
.R
-;
)

rzd)(T) (z)
27a2 fia

+ cos (%)I
jo + cos jo

where deA.is the effective loss tangent.

Characteristics of microstrip patch antennas Characteristics of microstrip patch antennas 157
-0

OdB -10 -20 -20 -10 OdB

OdB -10 -20 -20 -10 OdB

158 Characteristics of microstrip patch antennas Characteristics of rnicrostrip patch antennas 159
-e For the TM,, and TM,, modes, radiation is strongest .in the ,broadside.
Although there is a slight dip at the broadside for the TM,, mode, the radiation
is still very strong in this direction. For convenience, we shall refer all three
modes as broadside modes. Reference to the Figures shows that the polarisa-
tions of the three modes are the same at 6 = 0". This suggests that the equitrian-

Table 3.2 Theoretical resonant frequencies of an equitriangular patch with

sidelength a = 10 crn
fm,GHz
(W n) E, = 2.32, t = 0.159cm E, = 9.8, t = 0.0635cm
(1, 0) 1.3 0.64
(1, 1) 1.84 0.90
(2, 0) 2.6 1.28
(2, 1) 3.44 1.66
(3, 0) 3.9 1.91

Fig. 3.29 Radiation patterns of the equitriangular patch at,,,f = 1GHz and (i) E, = 2.32.
t = O.159crn; (ii) e, = 9.8, t = 0.0635crn
(a) TM,,. d = 0'
(b) TM,,, 9 = 90'
( c ) TMii, d = 0'
( d ) TM,,,+ = 90' 2 4 6 8 10
(e) TM ., d = 0' resonant frequency (GHz)
( f ) TM, = 90' a
(9)TMa d = 0'
( h ) TM,, , 9 = 90' Fig. 3.30 Radiation efficiency versus resonant frequency for an equitriangular patch with
( i ) TM, 4 = 0' 0 = 5.8 x 1O7S/rn,d = 0.0005 and (i) E, = 2.32, t = 0.318, 0.159, 0.0795crn;

( j ) TM,. d = 90' (ii) E, = 9.8, t = 0.127, 0.0635, 0,0254crn

(a) TMlo
(b)
( c ) TM21
Characteristics of microstrip patch antennas Characteristics of microstrip patch antennas 761

4-
-.,
-g
-
?
U

3-
t -1.275rnrn. 0.635rnrn.0.254rnrn
er:9.8

2-

1-

0
7 2 4 6 8 10
resonant frequency (GHz)

resonant frequency (GHz)

resonant frequency (GHz)

Fig. 3.31 Directivity versus resonant frequency for an equitriangular patch with: (i) E, = 2.32,
l I ' " ' " b t t = 0,318, 0.159. 0,0795cm; (ii) E, = 9.8, t = 0.127. 0,0635, 0,0254cm
2 4 6 8 10
( a ) TWO
resonant frequency (GHz)
( b ) TM,o
C ( C ) TMz
 Characteristics of microstrip patch antennas 163
162 Characteristics of microstrip patch antennas
Tho
31

0 2 6 6 8 1 0
resonant frequency (GHz) 0
0.1 1 10
C resonant frequency (GHz)
b

0
0.1. 1 10
resonant frequency (GHz)

Fig. 3.32 Gain versus resonant frequency for an equitriangular patch with a = 5.8 x lo7 resonant frequency (GHz)
Slm. 6 = 0.0005 and (i) e, = 2.32,t = 0.318, 0.159, 0.0795cm; (ii) E, = 9.8, C
t = 0.127. 0.0635, 0.0254cm
( a ) TM,,
( b ) TMm
( c ) TMn
 Characteristics of microstrip patch antennas 165
164 Characteristics of microstrip patch antennas
TMZO
gular patch can be operated at the resonant frequencies of the three broadside I
modes with similar pattern and polarisation characteristics. We shall further
show in the next Section that it is possible to find a position for a coaxial feed
such that the input impedance of all three modes are in the range of 50-100 R.

(ii) Radiation eficiency, directivity, gain, total Q and bandwidth

The radiation efficiency, directivity, gain, total Q and bandwidth as a function
of resonant frequency for the three lowest broadside modes, i.e. TM,,, TM,,,
and TM,, are shown in Figs. 3.30-3.34. The results of e, Q,and BWare similar.
However, the directivities and gains are significantly different.

resonant frequency (GHz)

b

1'421

6,.9.8

0'

resonant frequency (GHz)

6,-2.32
1 59 mrn
3.18 mrn
Fig. 3.33 Total 0 factor versus resonant frequency for an equitriangular patch with o =
5 . 8 x 107Slm. S = 0.0005 and (i) 8, = 2.32, t = 0.318, 0.159. 0,0795cm; (ii)
E, = 9.8, t = 0.127, 0,0635, 0.0254cm
( a ) TMlo
10 " '2 k " j m6 m 8 10

( 6 ) TM*, resonant frequency (GHz)

( c ) TM21
 Characteristics of microstrip patch antennas 167
166 Characteristics of microstrip patch antennas

Compared to the rectangular and the circular patches, major differences are
found in the Q, and B W curves. While these parameters depend on E,, t and f
in a complicated manner for the rectangular and circular patches, their beha-
viour for the equitriangular patch is simple: Q , decreases (BW increases) with
decreasing E, and increasing t irrespective of frequency.

(iii) Input impedances and their variations with feed position

The input impedances and their variations with feed positions are important
characteristics of patch antennas. In the cavity-model theory for the equitrian-
gular patch, the coaxial is modelled by a current ribbon of effective width 2w
along the x-axis. Usually this is several times the physical diameter of the coaxial
inner conductor and the input impedance is not a sensitive function of 2w. For
a patch with a = IOcm, E, = 2.32 and t = 0.159cm, the input impedances of
the three broadside modes, TM,,, TM2, and TM2, are shown in Fig. 3.35. The
value of 2w used in the computation is 6 mm.

2 4 6 8 10
resonant frequency (GHz)

resonant frequency G H z )
a
Fig. 3.34 Bandwidth versus resonant frequency for an equitriangular patch with a = 5.8
107S/m,6 = 0.0005 and (i) &, = 2.32, t = 0~318,0~159,0~0795cm; (ii) e, = 9.8,
t = 0.127, 0,0635, 04254cm
( a ) TMlo
resonant frequency
c (GHz)
(b) W o
( c ) TMz1
168 Characteristics of microstrip patch antennas Characteristics of microstrip patch antennas 169

The variations of R and X at resonance with feed position d a r e shown in Fig.

3.36. It is seen that R decreases with increasing d only up to a certain value of
d. The distance d can be chosen so that, for a particular mode, it assumes the
value necessary to match the characteristic impedance of the feed. If the antenna
is to be used for more than one mode, it is desirable that the input impedances
for these modes are not too different, in addition to having the same polarisation
and similar radiation patterns. In this connection, it is interesting to point out
that, by placing the feed at an appropriate location, the input resistances at

Fig. 3.36 Resonant resistance versus feed position of the three broadside modes of the
equitriangular patch with sidelength a = 10cm. 6, = 2.32,t = 0.159crn

I
0
8
0
0
u
0
..-.
0
0
u
resonance of the three broadside modes can be made to fall in the range of
,;I r 50-100R. For example, if d is equal to 4,7cm, we obtain R = 100,50 and 60R
for modes TM,,, TM,, and TM,,, respectively.
Comparison of theory and experiment for the equilateral triangular patch is
relatively scarce in the literature. In Reference 24, a comparison was made on
resonant frequencies and input impedances, and reasonable agreements were
obtained.

3.3.4 Annular-ring parch

3.3.4.1 Introductory remarks: While the rectangular and the circular patches
are probably the most extensively studied patch shapes, the annular ring has
also received considerable attention [26-341. There are several interesting feat-
ures associated with this patch. First, for a given frequency, the size is substanti-
ally smaller than that of the circular patch when both are operated in the lowest
mode (see example in Section 3.3.5). In application to arrays, this allows the
 Characteristics of microstrip patch antennas
I Characteristics of rnicrostrip patch antennas 777
7 70
I

elements to be more densely situated, thereby reducing the grating-lobe pro- 3.3.4.2 Cavity-model theory
blem. Secondly, it is possible to combine the annular ring with a second (i) Resonant frequencies, internal and radiation jields
microstrip element, such as a circular disc within its aperture, to form a compact Consider an annular ring patch with outer radius b and inner radius a, as shown
dual-band antenna system [26]. Thirdly, the separation of the modes can be in Fig. 3.37. Assuming that only TM modes exist, the resonant frequencies are
controlled by the ratio of outer to inner radii. Finally, it has been found that, determined by
by operating in one of the higher-order broadside modes, i.e. TM,,, the im-
pedance bandwidth is several times larger than is achievable in other patches of
comparable dielectric thickness.
The annular ring has been analysed using the cavity model [2, 26, 301, the where k,,,, are the roots of the characteristic equation
spectral-domain technique in Fourier-Hankel transform domain [28] and the
J:, ( k b )Y', ( k a ) - J:, ( k a )Y:,( k b ) = 0 (3.77)
In eqn. 3.77, J,,(x) and Y,(x) are Bessel functions of the first and second kind,
I
order n, respect~vely,and the prime denotes derivatives with respect to x. Letting
I C = hla, eqn. 3.77 takes the form
I

J:, (c4,,,,1y:,(4,",1 - J:,(Kt",1Y'(c&,,,,

1 = 0 (3.78)
where
X,", = k,,d (3.79)
For the case C = 2, the roots of eqn. 3.78 are given in Table 3.3.

Table 3.3 Roots of the characteristic equation J',,(X,,C)Y,(X,,) - J',,

(X,,)Y',,(X,,,C) = 0,where C = bla = 2
m 1 2 3 4 5

s ~ d evlew

eed

The cases when n = 1 and n = 2 are also of particular interest. The roots are
Fig. 3.37 Geometry of the annular-ring patch antenna shown in Figs. 3.38 and 3.39. Note that the spacings between the roots are
dependent on alb. This parameter can therefore be used to control the frequency
separation of the modes. For the general equation 3.77, solutions presented in
use of the method of matched asymptotic expansion [27].In what follows, the the form of a mode chart are given in Reference 33.
results obtained using the cavity model will be presented, together with some To account for the fact that a small fraction of the field exists outside the
comparisons with experiment. dielectric, it is customary to use an effective permittivity E, in place of F, in eqn.
V
 7 72 Characteristics of microstrip patch antennas Characteristics of microstrip patch antennas 173

3.76. The formula for E, as given by Schneider [31] is To account for the fringing fields along the curved edges of the ring, it has
been suggested that the outer and inner radii be modified according to

where
W = (b - a)

where

po is the permeability and zo is the quasi-static characteristic impedance of a

microstrip line of width W.
A pair of empirical formulas for the modified radii, sufficient for many
engineering purposes, are given by [33]

Fig. 3.38 The roots X, of eqn. 3.78, for n = 7, as a function of C = bla

For the glven values of a and b, a, and b, are calculated. Then the characteristic
equation is solved by replacing a and b by a, and b,. After solving the charac-
teristic equation for k,,,, the resonant frequencies are determined from

It should be pointed out that the correction to the resonant-frequency for-

mula involves both the effective permittivity and the effective radii. This is
somewhat different from the cases of the circular, the rectangular, and the
equitriangular patches, which involve the effective dimensions only. Lee and
Dahele [30] had shown that, in the case of the annular ring, good agreement
between theory and experiment can be obtained only if both effective quantities
are used.
The electric field under the patch is given by
I

Fig. 3.39 The roots, X, of eqn. 3.78, for n = 2, as a function of C = bla The far-zone electric field is
 776 Characteristics of microstrip patch antennas Characteristics of microstrip patch antennas 7 77
The effective loss tangent, comprising the three kinds of losses, is given by

The total Q factor is the inverse of the effective loss tangent.

(iii) Coaxial-fed annular ring

For an annular ring fed by a coaxial line at a distance d from the centre, which
we modelled by a uniform current ribbon of effective width 2w (eqn. 3.9, the
expression for E in the cavity is

where
~ Z sin
X (2nw) cos nn[J,(k,d)Y',(k,,a) - Jk (k,a) Y,(knmd)]
Rum =

and E,, =
1 forn # 0
2 for n = 0

The 0 and C$ components of the far-zone electric field are

Fig. 3.41 Sketches of the rad~at~on
patterns of the TM,, and TM,, modes of the annular-r~ng
patch antenna w ~ t hbla = 2 2tk0up,, J -
e-Ikor
( a ) 9 = 0' E, = j"+l - cosnd
( b ) lp = 90'
X r n m knm

where the quantity I, is the integral: J,(koa sin 0) - J',(k0 b sin 0) -T--- (3.101)
Jn (k,,rnb)

E4 = -j""--- 2twp0J e-jkorcos e

~ 1 2n2cos20 J,,(koasin 0) - J,,(kob sin 0) J',(k,,,,, a)
= jo a x r sin 8 ., knm sinn)

b -I
J,,(koa sin 8) - J,,(kob sin 8) J',,(k,,a)
J'"(k",b)
(3.102)
 7 78 Characteristics of microstrip patch antennas Characteristics of microstrip patch antennas 7 79

The input impedance is addition to larger bandwidth, the TM,, mode also has a larger directivity.
Similar results were obtained in Reference 35.
The input impedance and bandwidth of the annular ring patch antenna have
also been determined by modelling the antenna as a section of radial line loaded
with wall admittances [36]. The results are in qualitative agreement with the
other methods mentioned above.

where

3.3.4.3 Broadside modes T M , ,and TM,,: The most interesting finding for the
annual ring patch is perhaps the relatively wide-band property of the TM,,
mode. This was first predicted theoretically by Chew [27] and by Ali et al. [28]
using the matched asymptotic expansion technique and the vector Hankel
transform, respectively. Experimental verification of the theoretical prediction
was first reported by Dahele and Lee [29]. Lee and Dahele [30] also obtain this
theoretically within the framework of the cavity model, i.e. using the formulas
of Section 3.3.4.2. For an annular ring patch with bla = 2, a = 3.5cm,
E, = 2.32, t = 0.159cm, the variation of input impedance with frequency of the
TM,, and TM,, modes is illustrated in Figs. 3.42 and 3.43 for two feed positions, -5001
590 600 610 620 630 640 650
one near the inner edge (d/a = 1.05 and the other near the outer edge (dl
f (MHz)
a) = 1.95. It is seen that, for the TM,, mode, the impedance is not sensitive to
the feed position and the impedance bandwidth is very narrow ( < 1%). On the
other hand, the input impedance of the TM,, mode is very sensitive to the feed Fig. 3.42 Theoretical input impedance of the TM,, mode of an annular-ring patch with b =
position. With the feed near the outer edge, the value at resonance is only about 7.0cm. a = 3.5cm. 8, = 2.32,t = 0.159cm. fed at two radial locations
20R. With the feed near the inner edge, it attains the convenient value of about
60R at resonance. The bandwidth is about 4%, which is several times that of
the TM,, mode. This is also larger than the bandwidth achievable with the 3.3.5 Comparison of characteristics of the rectangular, circular, equitriangular
rectangular, the circular, or the equitriangular patches with the same dielectric and annular ring patches
constant and thickness, as reference to the corresponding Figures in Sections It is instructive at this point to present an example comparing the characteristics
3.3.1-3.3.3 shows. The theoretical results agree with the conclusion of Chew of the rectangular, the circular, the equitriangular and the annular ring patches.
[27], and Ali et al. 2][! who analysed the problem using considerably more Let us take the operating frequency to be 2 GHz and fabricate the patches on
complicated methods. ~ x ~ e r i r n e n tthe
~ l above
~ , predictions had been verified a substrate material of thickness t = 1.59 mm and E, = 2.32. If the patches
by Dahele and Lee [29] and Lee and Dahele [30]. are designed to operate in the lowest mode, a rectangular patch with an aspect
Detailed theoretical results based on the cavity model for the characteristics ratio 1.5 has dimensions b = 3.28 cm, a = 4.92 cm; a circular patch has radius
of the TM,, and TM,, modes are shown in Figs. 3.44-3.48. Note that, in 4.92cm; an equitriangular patch has side length 6.57cm; and an annular ring
180 Characteristics of rnicrostrip patch antennas Characteristics of rnicrostrip patch antennas 181

with b / a = 2 has b = 1.84 cm, a = 0.92 cm. The characteristics of the lowest
mode for the four patches are shown in Table 3.4. It is seen that all are broadside
modes. The circular patch has the smallest beamwidth in both planes. The
annular ring patch has the smallest physical area. The circular patch has the
largest physical area but it also has the largest bandwidth, efficiency and gain.

I I
a
-0
a
2 -0
-%op
G a w

X
-30~7'5~' " a '
2800 * a ' 2850 '
4 2900
f (MHz)

Fig. 3.43 Theoretical input impedance of the TM,, mode of an annular-ring patch with
b = 7.0cm. a = 3.5crn. e, = 2.32, t = 0.159cm. fed at two radial locations

The picture changes somewhat if the annular ring with b = 2a is designed to

operate in the TM,, mode. This is shown in the last column in Table 3.4. The
beamwidth is much narrower while both the gain and the bandwidth are
considerably larger. However, these improvements are achieved at the expense
of increasing the size of the patch. For the TM,, mode to resonate at 2 GHz,
 782 Characteristics of microstrip patch antennas Characteristics of microstrip patch antennas 783
b = 8.9 cm and a = 4.45cm, yielding an area of 249cm2. This turns out to be
a specific example of a general principle: increase in bandwidth can only be
achieved at the expense of increasing the volume of the resonator.

resonant frequency ( G H z )
a

0 2 4 6 8 10
resonant frequency ( G H z )

Fig. 3.45 D~rect~v~tiesof the T M , , and TM,, modes for the annular-rlng patch wlth b = 2a.
(i) E, = 2.32, t = 0.318cm, 0.159 cm, 0.0795 cm, (ii) 8, = 9.8, t = 0.127 cm,
0,0635 em, 0.0254cm

3.3.6 Brief mention of other patches

Besides the rectangular, circular, equitriangular and the annular ring, a number
of other patch shapes have been studied in the literature. They include the
right-angled isosceles triangular patch [2, 7, 371, the annular sector [7, 371, the
0L circular sector [37], the rectangular ring [38], the H-shaped patch [38] and the
0.1 1 10 elliptical patch [38-411. The analysis of the rectangular ring and the H-shaped
resonant frequency ( G H z ) patch requires the segmentation method, while the other patches mentioned
b above can be analysed using the simple cavity model. However, except for the
elliptical patch which offers the'possibility of generating circularly bolarised
Fig. 3.44 Radiation efficiency versus resonant frequency for the annular-ring patch with waves using a single feed, the other shapes do not appear to contain any features
b = 2a: (i)E, = 2.32, t = O.318cm. O.l59cm, 0.0795cm (ii) 6, = 9.8, t = 0.727
cm, 0,0635 cm, 0.0254 cm
which are not obtainable from the rectangular, circular, equitriangular or
( a ) TMll
annular ring patches. For this reason, only the elliptical patch will be briefly
( 6 ) TM12 discussed below.
The geometry of the elliptical patch is shown in Fig. 3.49. Experimental study
 184 Characteristics of microstrip patch antennas Characteristics of microstrip patch antennas 185

0 ' ~ ~ ~ ~ ~ 6 ; ~ h l b
resonant frequency (GHz)

resonant frequency (GHz)

2 6 3' 10
resonant frequency ( G H z )
resonant frequency (GHz)
b

Fig. 3.46 Gain versusresonant frequency for the annular-ring patch with b = 2a. a = 5.8 x Fig. 3.47 Total O factor versus resonant frequency for the annular-ring patch with b = 2a.
707Slm. d = 0,0005, and (i) E, = 2.32, t = 0.318em. 0.759 cm, 0,0795 cm, (ii) a = 5.8 x 107Slm,6 = 0,0005and (i) E, = 2.32, t = 0.318cm. 0.159cm. 0,0795
E, = 9.8, t = 0.727 cm, 0.0635 cm, 0.0254 cm em, (ii) E, = 9.8, t = 0.127em. 0.0635cm. 0.0254cm
( a ) TMl, (a) TMll
( b ) TM12 ( 6 ) TM12
-
 Characteristics of microstrip patch antennas Characteristics of microstrip patch antennas 187

of this antenna was reported by Yu [39] and later by Long et al. [41]. Theoretical
studies were carried out by Shen [40] using the cavity model, by LovandRichards
[I81 using a perturbation method, and by Despande and Bailey [42] using
moment method. The main conclusions of these studies are summarised as
follows:
(a) The radiation in the direction perpendicular to the patch is in general
elliptically polarised. However, with proper selection of both the feed position
and the eccentricity of the ellipse, circular polarisation can be obtained.
(b) The desired circular polarisation is best achieved by limiting the eccentricity
of the ellipse to a range of 10-20%. This corresponds to a (semi-major axis) and
b (semi-minor axis) differing by only a few percent. The perturbation method of
Lo et al. [18] yields the formulas

0.11
0.1 1.0 10
semimajor axis a
resonant frequency (GHz)
semlmlnar axbs b
a
foci : x = t c
TMIZ bs2a c=(a2-b21'
eccentricity e,:+

Fig. 3.49 Geometry of the elliptical patch

where the quality factor Q can be assumed to be that of the circular patch of
radius a or b. For example, if Q = 46.35, b/a = 0.976.
(c) The feed point should be on a radial line making 45' relative to the semi-
I+
0.10.1 1.0 10 major axis, i.e. 4, = 45". The positive sign yields left-hand while the negative
sign yields right-hand circular polarisation.
resonant frequency (GHz)
( d ) To achieve an operating frequencyf, the semi-major axis should be chosen
to be
Fig. 3.48 Bandwidth versus resonant frequency for the annular-ring patch with b = 2a. o =
5 . 8 x 107Slm,6 = 0.0005 and (i) E, = 2.32, t = 0.318cm, 0.159cm, 0.0795cm.
(ii) E, = 9.8, t = 0.127cm. 0.0635cm. 0.0254cm
( a ) TM,,
( b ) TM,, where p is a constant ranging from 0.27 to 0.29.
188 Characteristics of microstrip patch antennas
( e ) To achieve an impedance of 50R, the feed point should be at a distance
from the centre on the 4, = +45" line, where Q, 2. 0 . 2 8 ~ .
3.4 Some methods for improving the frequency agility and bandwidth of
Characteristics of microstrip patch antennas 789
Q, frequency agility of MPAs. One line of approach is to consider methods where-
by the operating frequency of the antenna can be tuned over a range of values
so that the same antenna can be used for several adjacent channels. This is the
single-band tunable case illustrated in Fig. 3.50d. In another scheme, dual-
frequency antennas with resonant frequencies separated by a certain range have

1
microstrip patch antennas

3.4.I Introduction
As mentioned in Section 3.1, the microstrip patch antenna (MPA), being
basically a leaky cavity, is inherently narrow band. The pattern bandwidth is
j \ a deal

f
usually many times larger than the impedance bandwidth, which therefore is the
parameter controlling the frequency response of the antenna. For this reason,
our subsequent discussion on bandwidth will refer to impedance. For a single
patch operating at the lowest mode, typical bandwidth is from less than 1% to
-
several percent for thin substrates satisfying the criteria t / l , < 0.07 for E, 2.3
b typ~cal
MPA response

Rl
f
and t l l , < 0.023 for E , 2: 10. When these inequalities are satisfied, the effect of
surface wave is assumed to be unimportant. For comparison purposes, a half-
wave dipole with a radiusllength ratio equal to 0.01 has a bandwidth of about
16%, while a medium-length helix operating in the axial mode has a bandwidth
of about 70%.
f

.'
,,!7p
'\
C increased
absolute
bandw~dth

One way of obtaining a relatively wide bandwidth is to use an annular ring f

patch and operate it in the TM,, mode, as the results of Section 3.3.4 indicate.
The price one has to pay is that the size of the patch is considerably larger than d single band,
that of the rectangular, circular, equitriangular or the annular ring operated in tunable
the lowest mode. In this Section, we discuss a number of other methods which
have been developed for overcoming the bandwidth problem.
Let us illustrate our discussion of bandwidth with a series of frequency
response characteristics depicted in Fig. 3.50. Let the response in (a) represent
the ideal characteristics, in which the input resistance is constant over a wide
range of frequencies. A typical MPA response, however, is that shown in (b).
Since this is not satisfactory for most purposes, a great deal of attention has been
devoted in recent years to improving the bandwidth characteristics of MPAs.
n n c dual band
non -tunable

One line of attack is to widen the absolute bandwidth of the antenna as much
as possible, as illustrated in (c). This in principle can be achieved simply by
increasing the thickness of the substrate. However, this introduces several
problems. First, a thick substrate supports surface waves, which will produce
undesirable effects on the radiation pattern as well as reducing the radiation
efficiency of the antenna. Secondly, as the thickness of the substrate increases,
problems associated with the feeding of the antenna arise. Thirdly, higher-order
cavity modes with fields depending on z may develop, introducing further
distortions in the pattern and impedance characteristics. It is therefore of
interest to develop more sophisticated methods of improving the absolute
bandwidth of MPAs, and a great deal of research has been devoted to this effort.
There has also been a substantial amount of effort devoted to increasing the
f dual band
tunable
Fig. 3.50 Illustrating the various frequency-response characteristics
been developed (Fig. 3.50e). These duaLfrequency structures are useful in
situations where the antenna is required to operate in two distinct frequencies
which may be too far apart for a single antenna to perform efficiently at both
frequencies. Related to this is the dual-band tunable configuration, in which one
or both of the resonances are tunable. The case for which only the upper
resonance is tunable is illustrated in Fig. 3.505
 190 Characteristics of microstrip patch antennas
In the next two Sections, some of the methods that have been developed to
provide the characteristics illustrated in Fig. 3.50 will be described.
3.4.2 Some methods of tuning MPAs
We shall describe four methods of tuning the resonant frequencies of MPAs.
These utilise (i) varactor diodes, (ii) shorting pins, (iii) optically controlled pin
diodes and (iv) adjustable air gap. The advantages and disadvantages of these
methods will be discussed.
Fig. 3.51 Illustrating the use of varactor diodes for tuning
3.4.2.1 Varactor diodes: For a given set of patch dimensions, the resonant
frequency is primarily governed by the value of the relative permittivity E, of the
substrate. If some means is available to alter E,, the resonant frequency will
change. One method of achieving this is to introduce varactor diodes between
the patch and the ground plane, as shown in Fig. 3.51. The diodes are provided
with a bias voltage, which controls the varactor capacitance and hence the
effective permittivity of the substrate. Bhartia and Bahl [43] performed an
experiment on this method and the results are shown in Fig. 3.52. The resonant
frequency f , of the lowest mode of the rectangular patch increases with the bias
voltage, owing to the increase of the diode capacitance. It is seen that, in this
experiment, a tuning range of some 20% was achieved with a 10V bias. The
range increased to about 30% with a 30 V bias. Note that the curve ofj; versus
bias voltage is not a linear one.
Since the paper by Bhartia and Bahl [43], there appeared to be no further
reports in the literature on this method, either experimentally or theoretically.
3.4.2.2 Tuning using shorting posts (pins): The value of E, can also be chan-
ged by introducing shorting posts (pins) at various points between the patch and
Characteristics of microstrip patch antennas 191
the ground plane. These shorting posts present an inductance, and therefore
alter the effective permittivity of the substrate. In the context ,of microstrip
antennas, the method was first introduced by Schaubert et al. [44] in 1981. It is
illustrated in Fig. 3.53. Using two posts, the experimental results obtained are
shown in Fig. 3.54. It is seen that the resonant frequency is dependent on the
separation of the two posts and a tuning range of some 18% is obtained as the
separation varies between 0 and the whole width of the patch.
9l alpha D V H 6733-168-001 varactors
b ~ a svoltage ( V )
Fig. 3.52 Resonant frequency versus bias voltage for a varactor-loaded rectangular patch
antenna (Reproduced from Reference 43 p. 306 @ IEEE 1982)
Schaubert et al. [44] developed a theory of shorting pins based on the trans-
mission-line model, and the predictions (shown in Fig. 3.54) agree reasonably
well with experimental data. However, because the transmission line model is
not capable of predicting the variations in the inductive component of a load as
its position is varied within the element, it fails to predict certain trends in the
resonant frequency of a short-loaded patch as the shorting pin approaches the
patch edge. This model also cannot predict the impedance of the element very
accurately because the field distribution between the ground plane and the patch
of a loaded element is much too complicated to be adequately represented by
a single-mode transmission-line model. It should be noted, however, that the
transmission-line model has been further developed for rectangular as well as
for circular patches with shorting pins by Sengupta and co-workers [45, 461.
 192 Characteristics of microstrip patch antennas Characteristics of microstrip patch antennas 193

Lo and coworkers [47] have developed the cavity model for MPAs with
lumped linear loads in general and shorting pins in particular. They have also
applied the shorting-pin method to design dual-frequency structures. This will
be discussed in Section 3.4.3.2.

led

Fig. 3.55 Tuning using optically controlled pin diode

3.4.2.3 Optically tuned patch antenna: A method of tuning the resonant

frequencies of patch antennas utilising optically controlled pin diodes was
recently reported [48]. The scheme is illustrated in Fig. 3.55. A stub is connected
Fig. 3.53 Illustrating the use of shorting posts for tuning the resonant frequency of a patch to the patch by means of an optically controlled pin diode. When the diode is
antenna reversed biased, it acts as an open circuit and the patch resonates at the
frequency for which it is designed, say f,. When the diode is forward biased, it
acts as a short circuit and the resonant frequency becomes that of the patch and
the stub, i.e. f, - AJ In the experiment, f, was 10.285 GHz and f, - Af was
10.207GHz. These are the limits in the range of tuning. By illuminating the
diode with light, the diode impedance can be varied from a high value to a low
value. As a result, the resonant frequency is optically tuned. It was found in their
experiment that an illumination of 1 w/cmZ resulted in a 15 MHz downward
shift in the frequency. This method clearly needs further development as the
range of tuning reported was extremely limited.

Fig. 3.54 Resonant frequency versus separation of posts for a 6 . 2 x 9.0 cm rectangular patch
antenna with E, = 2.55,t = 1.6mrn (Reproducedfrom Reference 44p. 7 19 @ IEEE
79811
Discussion: The three methods described so far suffer from the following disad-
vantages:
(i) The design of the patches is complicated by the added components such as
varactor diodes, optically controlled pin diodes and their associated biasing
circuit. In the case of shorting pins, their precise positions are also important.
(ii) For high frequencies (say > IOGHz), the patch sizes are small and it is
difficult to accommodate the diodes and shorting posts underneath each patch.
(iii) The added complications in design multiply for an array consisting of a
large number of elements.
194 Characteristics of microstrip patch antennas
The potential advantage of the three methods is the possibility of electronic
tuning. For example, there were suggestions that the shorting pins could take
the form of switching diodes so that the frequency can be changed by electronic-
ally switching the diodes on and off. However, to the authors' knowledge, a real
demonstration of such electronic switching applied to MPAs has yet to be
reported in the literature.
In Section 3.4.2.4, we describe a somewhat different method of tuning the
resonant frequency of an MPA, i.e. utilising an adjustable air gap between the
substrate and the ground plane.
Characteristics of microstrip patch antennas 195
Experimental results: The first configuration studied by Dahele and Lee was the
circular patch. The radius of the patch was 5cm fabricated on Duroid material
of thickness 0.159 cm and relative permittivity 2.32. The width of the air gap is
controlled by using spacers between the substrate and the ground plane. In the
experiment, spacers of 0.5 mm and 1.Omm were used. The antenna was provided
with a coaxial feed near the edge of the disc at a distance d = 4.75 cm from the
centre. This feed position is chosen as it is well known that it yields a larger
resistance at resonance compared to a feed which is closer to the centre. The

3.4.2.4 Tuning using an adjustable air gap

Introduction: By introducting an air gap between the substrate and the ground
plane in a microstrip patch antenna the effective permittivity of the cavity will
change. This can be used to tune the resonant frequency of a microstrip patch
antenna as discussed below.

substrace I
I I conductma ~ a t c h

I
A airgap
;;50' ' ' ' 1300
' 1350 1400
spacer f (MHz)
T
Fig. 3.57 Measured input impedance of the T M , , mode of a 5 c m circular-disc microstrip
ground'plane I II /-coaxial feed
antenna for three values of air-gap width A. E, = 2.32,t = 0.159cm (After Re-
ference 51 pp. 455-460)

Table 3.5 Measured resonant frequencies and impedance bandwidths of

the first few modes of a 5cm-radius circular-disc microstrip
Fig. 3.56 Geometry of a microstrip patch antenna with air gap antenna for three values of the air-gap width

--

The geometry of a microstrip antenna with an airgap is shown in Fig. 3.56. fm % BW f,,,, %BW f,, % BW
Consider the cavity under the conducting patch. It is made of two layers: a TM,, 1128MHz 0.89 1286MHz 1.48 1350MHz 2.07
substrate of thickness t and an air region of thickness A. Compared to the case TM,, 1879MHz 0.85 2136MHz 2.15 2256MHz 2.61
with no air gap the effective permittivity of the cavity is evidently smaller. As a TM,, 2596MHz 0.77 2951MHz 1.63 3106MHz 2.02
result the resonant frequencies of the various modes will increase. Since the, E, fed at 4.75cm from the centre
= 2.32, t = 0.159~111;
effective permittivity decreases as A increases, tending towards the free-space
value E,, as A -+ a,it follows that the resonant frequencies can be tuned by
adjusting the air-gap width A. As a by product the bandwidth will also increase measured resonant frequencies are shown in Table 3.5. For the lowest mode
partly due to the increase in the height of the dielectric medium and partly TM,,, there is a tuning range of about 20% in frequency and a more than
because the effective permittivity is smaller. twofold increase in the bandwidth as A goes from 0 to 1.0mm. Similar beha-
Based on the above idea, Dahele and Lee [49-521 have carried out a series of viour is recorded for the other modes. The measured input impedances of the
experimental and theoretical studies on the microstrip antenna with air gaps. TM,, mode as a function of frequency are shown in Fig. 3.57. The upward shift
Some of their results are presented below. in the resonant frequency and the widening of the bandwidth are clearly seen.
196 Characteristics of microstrip patch antennas Characteristics of microstrip patch antennas 197

As for the radiation pattern it was found that the air gap did not have a effective permittivity of the two-layered medium:
significant effect on the pattern.
Another antenna studied was the annular-ring patch. The effect of an air gap
on the two broadside modes TM,, and TM,, are shown in Table 3.6 for an
annular ring of outer radius 7.0 cm and inner radius 3.5 cm fabricated on Duroid Eqns. 3.108 and 3.109 are valid for any patch shape. Note that, as the air gap
material of thickness 0.159cm and E, = 2.32. As in the circular patch there is width A increases, E," decreases and the resonant frequency increases. The
an upward shift in the resonant frequencies and a widening of the bandwidths. dependence ofJ;,,,(A) on A, however, is not a linear one.
It is significant that, for the TM,, mode, the bandwidth attains a value of 8.6%
when A is equal to 1.0 mm. Discussion:As in the other methods the adjustable air gap as a means of tuning
the resonant frequencies has advantages and disadvantages. The advantages
are:
Table 3.6 Measured resonant frequencies and impedance bandwidths of
the TM,, and TM,, modes of an annular-ring microstrip antenna (i) No costly components are added.
for three values of the air-gap width A (ii) It can be applied to patches of any shape. There is no need to know the
A = 0 A = 0.5mm A = l.Omm details of the fields in the cavity.
(iii) The method is particularly attractive for an array made up of a great
number of elements as illustrated in Fig. 3.58. If the elements are fed by
TM,, 626MHz 0.6 720MHz 0.7 778MHz 0.8 striplines, the resonant frequencies of all the elements, and therefore of the
TM,, 2757MHz 4.0 3040MHz 8.0 3240MHz 8.6 array, can be tuned by a single adjustment of the air gap width A.
Inner radius a = 3.5cm. outer radiur b = 7,0cm, 6, = 2.32, r = 0.159cm. The feed is placed a t
d/a = 1.05 where d is the distance from the centre

Theory: Lee and Dahele has developed the theory of the two-layered microstrip
antenna using the cavity model. The original assumptions of the model are
modified to account for the two layers as follows:
(i) Owing to the close proximity between the conducting patch and the ground
plane only transverse magnetic (TM) modes are assumed to exist. The z-com-
ponent of the electric field, however, is a function of z since the cavity is
two-layered.
(ii) The cavity is assumed to be bounded by perfect electric walls on the top and
on the bottom and by a perfect magnetic wall along the edge.
(iii) Across the dielectric-air interface the tangential electric field and the nor- spacer
mal electric flux density are continuous.
alr gap
Based on the above assumptions detailed analysis for the circular and annular-
Fig. 3.58 Tuning a microstrip antenna array by using an adjustable airgap
ring patch were carried out and good agreement between theory and experiment Stripline feeds are assumed
was obtained. In the interest of brevity, except for the resonant frequencies, the
theoretical formulas will not be included here. The formula for the resonant
frequency, however, is a very simple one and is given by The disadvantages are:
(i) The width of the air gap has to be changed mechanically. Electronic tuning
appears to be difficult.
(ii) The antenna is slightly thicker. This however is compensated for by an
where Jrf;,,(0)is the resonant frequency when there is no air gap and is the increase of the bandwidth.
 Characteristics of microstrip patch antennas 199
198 Characteristics of microstrip patch antennas

To end this Section we point out that it is possible to alter the resonant
frequency by inserting a piece of dielectric in the air region, as illustrated in Fig.
3.59. The relative permittivity of the inserted dielectric can be either the same
as that of the substrate or different. Both the thickness and the permittivity of
the inserted dielectric will determine the resultant resonant frequency.

,conduct~ng patch

s u b s t r a t e
-spacer

resonant frequency cdn be tuned by Inserting a piece of dielectric

In alr region (plug In u n ~ t )

'
.I
Fig. 3.59 Altering the resonant frequency by inserting a piece of dielectric in the air gap
20, - 3 78cm

3.4.3 Dual-band structures

There has been considerable interest in the development of dual-frequency
microstrip antennas. The characteristics of this class of antennas is illustrated in
Fig. 3.50e. They are useful when the antenna is required to operate in two 1.0 , 3.2 frequency (GHz)
distinct frequencies which are too far apart for a single antenna to perform
efficiently at both frequencies while the behaviour of the antenna in the range
of intermediate frequencies is of little or no concern. Several methods of obtain-
ing the dual-frequency characteristics have been developed. We begin with the
method of simply stacking two patches together.

\ d l Kconducting patch

?/ frequency ( (

Fig. 3.60 Non-tunable dual-frequency stacked microstrip antenna

Fig. 3.61 Real and imaginary parts of impedance of stacked circular patches etched on a
dielectric with E, = 2.47
3.4.3.1 Stacked circular-disc antenna: The first experimental report on a
( a ) 2a, = 3.70ci-n
dual-frequency structure using two stacked circular patches was that of Long ( b ) 2a, = 3.78cm
and Walton [53]. The geometry is shown in Fig. 3.60. The discs were photo-etch- ( c ) 2a, = 3.85 cm
ed on separate substrates and aligsxd so that their centres were along the same (Reproduced from Reference 53 p. 271 @ lEEE 1 9 7 9 )
200 Characteristics of rnicrostrip patch antennas
line. The sizes of the two discs and their spacings were varied and the resultant
behaviour of the antenna characteristics measured. The antenna was fed by
means of a coaxial line. The centre conductor passed through a clearance hole
in the lower disc and is connected electrically to the upper disc. If one considers
the two regions under the patch' as two resonant cavities it is clear that the
system behaves as a pair of coupled cavities. Since the fringing fields are different
for the upper and lower cavities, two resonant frequencies are expected even if
the diameters of the two discs are the same. While the qualitative explanation
is relatively simple the quantitative theory for this structure is still lacking. In
what follows the experimental results of Long and Walton are described.
upper d~scdlameter 2al (mm)
Fig. 3.62 Resonant frequencies versus upper disc diameter of stacked circular patches etched
on a dielectric with E, = 2.47, 2a2 = 3.78~171,t, = t, = 0.75mm. (Reproduced
from Reference 53 p. 271 @ IEEE 1979)
Fig. 3.61 shows the real and imaginary parts of the input impedance for
?a, = 3.78 cm, t , = t2 = 0.075 cm and three values of 2a,. The resonant fre-
quencies as a function of the upper disc diameter are shown in Fig. 3.62. Also
shown is the resonant frequency of the lowest mode for a single disc of diameter
2a and substrate thickness t = 0.075cm, taking into account the fringing field
through the effective diameter. It is seen that the lower resonant frequency is
relatively constant, remaining near the value of a single disc with 2a = 3.78 cm
and d = 0.075 cm. The upper resonance, on the other hand, is highly dependent
on the size of the upper disc. Radiation patterns were also taken by Long and
Characteristics of microstrip patch antennas 207
Walton which showed that they were similar to the radiation pattern of the
I lowest mode for the single circular patch.
While the results of Long and Walton [53] showed that it is possible to design
for the separation of the resonant frequencies by choosing the diameters of the
upper and lower discs, this is not very convenient in practice because of the lack
of formulas to predict the frequencies. Also once they are designed and etched,
it is not possible to alter or tune the separation of the two resonant frequencies.
The configuration as presented by Long and Walton was therefore a dual-band
I
I
! -
non-tunable antenna of the type illustrated in Fig. 3.50e.
conduct~ngpatch
ground 'plane
--I -I
Fig. 3.63 Tunable dual-frequency stacked microstrip antenna utilising the air-gap idea
Dahele and Lee 1541 have applied the air gap idea to study dual-frequency
stacked discs. The geometry is shown in Fig. 3.63 in which air gaps between the
lower substrate and the ground plane and/or between the two substrates are
introduced. Either of the air gap widths can be set to zero. Their experimental
results performed with two stacked discs of 7 cm radii, each etched on substrates
with E, = 2.32 and thickness 0.159cm are shown in Fig. 3.64 and 3.65. In Fig.
3.64, with the lower air gap set to zero the upper air gap is seen to increase the
resonant frequency of the upper resonance. In Fig. 3.65 the upper air gap is set
to zero and the effect of the lower air gap is studied. It is seen that the effect is
more complicated since it shifts not only the lower but also the upper resonance.
In both cases the bandwidth of the lower resonance is substantially broadened.
Dahele and Lee [55] have also studied a structure consisting of two stacked
annular ring patches as shown in Fig. 3.66. This structure was also found to
exhibit dual-frequency behaviour. As in the case of circular discs an upper air
gap was found to be a convenient method of altering the separation of the
frequency bands.
3.4.3.2 Single-element dual-frequency microstrip untenna: It is possible for a
single-element microstrip antenna to operate at many frequencies correspond-
202 Characteristics of microstrip patch antennas Characteristics of microstrip patch antennas

ing to the various resonant modes pertaining to the structure. However, for
most applications it is required that the radiation pattern, the polarization and
the impedance be similar if not identical in all the frequency bands of operation.

T M l r mode; d-6.5cm; A1=O

400
-R

I
substrate kb , conducting patches

Fig. 3.64 Measured input impedances of the TM,, mode of a pair of stacked circular discs of
7 c m radius for three values of the upper air gap: d = 6.5cm. A, = 0, e, = 2.32,
ground plane
Ill
11 /-coaxial feed
t = 0.159cm (After Reference 57 pp. 455-460)
Fig. 3.66 Geometry of the stacked annular-ring antenna

Fig. 3.65 Measured input impedances of the TM,, mode of the stacked circular discs of Fig.
3.63 for two values of the lower air-gap. d = 6 5 c m . A, = 0, e, = 2.32,
t = 0.759cm (After Reference 57 pp. 455-460)

This immediately rules out many modes. Furthermore, for a given geometry all
Fig. 3.67 Geometry of a rectangular patch antenna with six possible shorting pins anda short
the resonant frequencies are related in fixed ratios, providing no flexibility for matching stub. (After Reference 56 pp. 298-300)
the designer. All dimensions in centlmetres
204 Characteristics of microstrip patch antennas
If for a particular patch shape two modes can be found which produce similar
radiation patterns with the same polarisation, dual frequency is possible with a
single patch. For the rectangular patch the two modes (0, 1) and (0, 3) satisfy
this requirement. However, their resonant frequencies are related by a fixed
ratio of approximately 3, the exact value being dependent on the edge effect.
Suppose now shorting pins are placed on the nodal lines of the (0,3) model field,
there will be little effect on the (0, 3) mode but a strong effect on the (0, I) mode.
This offers a way of altering the separation of the two frequency bands. The
insertion of pins at proper locations can also be used to tune the input im-
pedance for the (0, I) mode while the feed location is chosen first for the desired
impedance for the (0, 3) mode.
The above idea has been successfully demonstrated experimentally by Zhong
and Lo [56]. A multiport-cavity-model theory has also been developed by Lo
and coworkers which appears to predict the effects of shorting pins on frequency
and impedance well. We limit here to a summary of the experimental results of
Zhong and Lo.
Table 3.7 Resonant frequencies for (0, 1 ) and (0, 3 ) modes against short-
ing pins used (After Reference 56)
The geometry of the rectangular patch in their experiment is shown in Fig.
3.67. It is made of 118 in copper-cladded Rexolite 2200 with six shorting-pin
positions. The effects of successively adding more and more pins (each approx-
mately 0.05 cm in diameter) at the positions indicated in Fig. 3.67 are shown in
Table 3.7. It is seen that the ratio of the two operating frequenciesf,,/f,, can be
varied approximately from 3 to 2. Since all these pins are located on the (0, 3)
nodal lines f,, remain constant at approximately 1865 MHz whilef,, is varied
from 613 to 891 MHz. In order for the impedances of the two bands to be close
to 50R at resonance it is necessary to attach a short capacitive stub of
0.6cm x 2.1 cm. With the stub added, the bandwidth with reference to 3:l
VSWR is about 2% for the low band and almost 8% for the high band. Typical
low- and high-band patterns in both E and H planes are shown in Figs. 3.68.
It is seen that, while the two modes radiate strongest in the broadside, the
Characteristics of microstrip patch antennas 205
directivities of the two modes are quite different. By increasing the number of
pins the two frequencies can be brought to a ratio of about 1.8. If a smaller ratio
is desired it is found that it can be achieved by introducing slots in the patch.
This, however, makes the fabrication of the patch somewhat complicated.
H - plane
E - plane
-70"
-80" 80'
-90" 90'
---- low-band. f - 8 8 3 M H z
-hlgh-band;f: 1848 MHz
Fig. 3.68 Typical radiation patterns in H- and E-planes for antennas shown in Fig. 3.67with
six pins inserted (After Reference 56 pp. 298-300)
The rectangular patch is not the only geometry capable of providing dual-
frequency operation. In Section 3.3.3, it was shown that the TM,,, TM,, and
TM,, modes of the equitriangular patch are all broadside modes with similar
polarisations in the broadside direction. Moreover, by choosing the location of
the feed properly, the impedances of these modes do not vary greatly. It thus
appears that it is possible to utilise the equitriangular patch for dual- or even
triple-frequency operation.
3.4.3.3 Dual-band microstrip antennas with reactive loading: A dual-
frequency microstrip antenna can also be obtained simply by loading it with a
reactive load. If the reactive load takes the form of a short-circuited length of
microstrip transmission line the low-profile characteristic of a microstrip patch
 206 Characteristics of microstrip patch antennas Characteristics of microstrip patch antennas 207

-
antenna is retained. Such a structure was suggested by Davidson et al. [57] and
was demonstrated to work experimentally.
Fig. 3.69 shows the dual-band rectangular microstrip patch antenna with a
monolithic load studied in the experiment of Davidson et al. The patch was of
The separation of the resonances can be varied by (i) changing the length of
the microstrip line and (ii) introducing an inset dimension S with an accompa-
nied gap spacing G between the line and the radiator, as shown in Fig. 3.71. The
results for the resonant frequencies are shown in Table 3.8.

sut
str rate ground 'plane

Fig. 3.69 Dual-frequency rectangular patch antenna with monolithic reactive loading (After Fig. 3.71 Geometry incorporating an insert dimension S and a gap spacing G (After Re-
Reference 57 p. 936-937) ference 5 7 pp. 936-937)

Table 3.8 Resonant frequencies of monolithic microstrip elements (after

Reference 57)
W L G S f~ fu
cm cm cm cm GHz GHz
0.33 4.0 1.O 1.5 2.356 2.494
0.33 4.0 0 0 2.275 2.666
0.33 8.4 0 0 2.339 2.628
0.33 4.0 0.7 1.5 2.437 2.494
0.33 4.0 0.3 1.5 2.47 1 2.514

-501
2200 2400 2600 2800
Discussion: In summary three methods of obtaining dual-frequency characteris-
frequency ( M H z )
Fig. 3.70 impedance of edge-loaded, 4 x6cm patch antenna with 1 = 4,Ocm.
tics for microstrip patch antennas have been described. The method using
W = 0.33 cm, E, = 2.77, t = 0.079 cm; coaxially fed near the edge and at the centre shorting pins use two different modes. As such the radiation patterns, while
of the 6cm side (After Reference 5 7 pp. 936-937) similar in the broad sense, do vary in detail as well as in directivity. The
separation of the resonances can be controlled by the number of pins, but it is
dimension 6cm x 4cm etched on a substrate with E, = 2.17 and thickness difficult to have them close together unless additional features such as slots are
0.079 cm. It is coaxially fed near the edge and at the centre of the 6cm side. For introduced in the patch. These additional design features appear to be difficult
a line length of L = 4.0cm and width w = 0.33 cm, the impedance characteris- to accommodate at high frequencies where the patch size is small.
tics is shown in Fig. 3.70. Good pattern performance was observed at each of The advantage of using shorting pins to realise dual-frequency characteristics
the resonant frequencies (2.275 GHz and 2.666 GHz, respectively). and to control the frequency separation is that it is a single-patch geometry,
 208 Characteristics of microstrip patch antennas
thereby retaining the low profile characteristic of microstrip antennas. This
advantage is also shared by the monolithic reactive-loading method. In the
reactive-loading method the two frequencies are separated by 10-20%. The
separation can be controlled by several parameters associated with the reactive
load. However, once a design is etched it is not possible to tune the antenna.
For the case when the separation is in the range of 10-20% it appears that
the stacked geometry discussed in Section 3.4.3.1 offers the advantages of
operating in the same mode and the flexibility of tuning the separation by means
of an air gap. This structure, however, is thicker than the single patch and the
low-profile characteristics of the microstrip antenna is slightly compromised.
3.4.4 Electromagnetic-coupled patch antenna ( E M C P )
3.4.4.1 Introduction: As mentioned in Section 3.4.1, a great deal of research
has been devoted to increasing the absolute bandwidth of MPAs. The methods
fall into three categories: electromagnetic-coupled patches (EMCP), use of
parasiticelements and log-periodic arrangement of an array of patches. We shall
discuss in this Section only the EMCP since it is related to the tunable stacked
geometry of Section 3.4.3.1. The use of parasitic elements and log-periodic
arrangement are covered in other Chapters of the Handbook.
As mentioned in Section 3.4.1 it is possible to increase the absolute bandwidth
of MPAs by simply using thicker substrates. This, however, introduces several
problems. The first is the excitation of surface waves, which distorts the normal
radiation pattern and introduces additional loss; the second is the excitation of
higher-order modes with z dependence, which introduces further distortions on
the pattern and impedance characteristics. The third is that the application of
common feeding techniques, i.e. direct feeding by either a coplanar microstrip
line or a perpendicular coaxial line, becomes increasingly difficult for the follow-
ing reasons. Consider first a coaxial feed. Since the probe (extension of inner
conductor of the coaxial line) introduces a series reactance almost proportional
to the substrate thickness, the lead inductance will become significant with
respect to the antenna radiation resistance for thick substrates and will therefore
prevent proper matching. Consider next a patch which is edge-fed by a coplanar
microstrip line. For a fixed impedance level the line width is almost propo~tional
to the dielectric thickness. Since the patch dimensions for a fixed resonant
frequency are only weakly dependent on the dielectric thickness (through the
fringing field) the width of the feed line will become non-negligible as the
substrate reaches a certain thickness. As a result the radiation pattern of the
antenna will be disturbed partly due to the covering of the radiating patch edge
by the line and partly due to increased radiation from the feed line.
In view of the above problems, electromagnetic coupling (instead of direct
coupling) has been studied as a possible feed technique for electrically thick
MPAs. In particular, promising results have been obtained for the stacked
dual-patch geometry which we now discuss.
Characteristics of microstrip patch antennas 209
3.4.4.2 Stacked dual-patch geometry: The basic geometry of the ,stacked
dual-patch electromagnetic-coupled microstrip antenna is shown in Fig. 3.72.
Each conducting patch is fabricated on an electrically thin substrate and
separated by a region of air or foam with E, 2: 1. The structure looks similar to
the tunable dual-frequency antenna of Section 3.4.3.1, but is different in two
aspects. First, the thickness of the air region is several times the substrate
thickness, while in the tunable version discussed earlier, it is a fraction of the
substrate thickness. Secondly, rather than being fed directly by a transmission
line, the top element is excited via electromagnetic coupling from the lower
element, which is located closer to the ground plane and is connected directly
to a feed line. The top and bottom patches are referred to as the radiating and
the feeding patches, respectively.
radlatlng patch
spacer----A I O L r reg10nt e e d l n g patch
ground plane
Fig. 3.72 Electromagnetic-coupled patch antenna
When the air region is small two resonances are expected, as in the case
discussed in Section 3.4.3.1. Experimental studies showed that, as the air region
exceeds a certain thickness, the lower resonance disappears and only one re-
sonance remains. The single resonance condition can also be obtained by
designing the diameter of the radiating element to be larger than the feeding
element.
Electromagnetic-coupled patches appeared to be first discussed by Sabban
[60]. Circular, annular-ring, rectangular and square patches in the S band
(2-4GHz), etched with substrates about 0.01 1 thick, were reported to yield
bandwidths ranging from 9% to 15%. His paper contained very little informa-
tion on the air-gap width and the relative sizes of the elements, other than the
statement that the radiating element was larger than the feeding element and
that the antennas exhibited a single resonance rather than dual resonances.
A more detailed experimental study was carried by Bhatnagar et al. [61]. The
elements were triangular patches operating in the S band, and foam material
with E , = 1 was introduced as the air gap between the dielectric layers (Fig.
3.73). The side lengths of both the top and bottom equitriangular patches were
37 mm fabricated on a substrate of thickness 1.6 mm and relative permittivity
2.55. The lower patch was provided with a coaxial feed at a distance
270 Characteristics of microstrip patch antennas Characteristics of microstrip patch antennas 27 7

F, = 13.5 mm. The width A of the air gap is controlled by using foam material and 20dB in the E-plane. It is interesting to note that the directivity of the
of uniform thickness. antenna was larger than that of an ordinary microstrip patch, the beamwidth of
The functional behaviour of the impedance characteristics is given in Table which was greater than 85'-90'.
3.9. The results for A = 3 mm showed an increase in the bandwidth at lower The configurations of EMCP studied by Sabban and Bhatnagar et al. can be
resonance and a sizable radiation resistance at the second resonance. so that the
structure may be operated as a dual-frequency antenna. Separation of the
resonances was about 12% at A = 0 and 18% at A = 1.5 mm.
For A > 3 mm the first resonance disappears. The second resonant frequency
increased and the real and imaginary parts of the impedance increased with A.
At t = 5mm the bandwidth was 595MHz, which was about 17.5% at the
centre frequency of 3.407 GHz.

-, p a r a s i t ~ cpatch

Fig. 3.73 Geometry of electromagnetic-coupled triangular patch antenna (After Reference

67 pp. 864-865)

Table 3.9 Characteristics of the stacked triangular patch antenna (after

Reference 61)
First resonance Second resonance
A f % BW Maximum f % BW Maximum
resistance resistance
mm GHZ a GHZ a
0 3.1 3.5 150 3.54 2.5 90
1.5 3.12 4.8 150 3.8 1 3.1 55
3 3.135 6.4 75 3.77 3.2 48
4 - - - 3.72 10.5 52
5 - - - 3.61 17.46 55 angle (degree)
6 - - - 3.56 14.8 62.5
9 - - - 3.45 8.6 105

Fig. 3.74 E-plane radiation patterns of the antenna of Fig. 3.73 with L = 37mm.
The radiation patterns in both the E-plane and the H-plane at various t , = t, = 7.6mm. A = 5 m m andF, = 73.5mm (After References 67 pp. 864-865)
-3.1 GHz
frequencies within the impedance bandwidth are shown in Fig. 3.74 and Fig. --- 3.3 GHz
3.75, respectively. The beamwidth varied from 75" to 85" in the H-plane and 55" 3.5GHz
to 65" in the E-plane. The cross-polar level was better than 16 dB in the H-plane 3.7 GHz
 212 Characteristics of microstrip patch antennas Characteristics of microstrip patch antennas 2 73

described as the 'normal' type. If the upper patch is fabricated on the underside
of the substrate an 'inverted' configuration is cobtained. Fig. 3.76 illustrates
these two types of configurations. The advantage of the inverted type is that
there is a protective dielectric cover for the upper conducting patch. It has been
studied by Chen et al. [62] and by Dahele et al. [63].
Further studies of the EMCP antenna were carried out by Lee et al. [64], using
rectangular patches etched on Cuflon substrates ( E , = 2.17). Exciting the TM,,
mode at about IOGHz, they recorded the variation of pattern shape, 3dB
beamwidth and bandwidth with the separation A, for A between 0 and 0.37 1,.
This is beyond the range studied by previous authors. It was found that,
depending on A, the characteristics of the antenna can be separated into three

radlatlng patch

coax probe

normal conf~guratlon ~ n v e r t e dconflguratlon

0 b

Fig. 3.76 Normaland inverted configurations of the electromagnetic-coupledpatch antenna

regions. Region 1 is associated with bandwidths exceeding 10%; region 2 has

abnormal radiation characteristics and region 3 is associated with narrow
beamwidth and high gain. The value of A separating these regions depends on
the dielectric material between the two layers. The gain in region 3 is 9-1 0 dB,
angle (degree) compared to 5.3 dB for the single patch. It begins at A = 0.3 1 2, for air dielectric
and at A = 0.21 1, for Teflon. The bandwidth in region 3, however, is only
about 1.3% for air and 0.85% for Teflon.
It is evident from the experimental results of the authors cited above that, by
operating in region 1, the EMCP offers a promising way of achieving band-
Fig. 3.75 H-plane radiation patterns of the antenna of Fig. 3.73 with L = 37mm. widths in excess of lo%, while reducing the problems encountered by simply
t , = t, = 1.6mm. A = 5mm and F, = 73.5mm (After Reference 61 pp. 864-865)
-3.1 G H z increasing the substrate thickness. If high gain rather than large band-
--- 3.3 G H z width is desired, the EMCP can be operated in region 3. Although analytical
3.5G H z methods are available [65], little work has been done to apply them to this
3.7 G H z interesting antenna. As such, the experimental results described above have not
been quantitatively explained, nor are there formulas available which would aid
the design in terms of resonant frequency, impedance, bandwidth and gain.
Such theoretical research is urgently needed.
214 Characteristics of microstrip patch antennas
3.5 Summary
This Chapter begins with introducing the simple cavity model for analysing
microstrip patch antennas with thin substrates. The formulas obtained from this
model for rectangular, circular, equitriangular and annular-ring patches are
then presented. The radiation pattern, efficiency, directivity, gain, quality factor
and imvedance of these. antennas are illustrated. mainlv for the broadside
modes. Experimental results are included or referenced where available.
After a brief mention of some other geometries, notably the elliptical patch,
the Chapter proceeds to discuss some methods of improving the frequency
agility of microstrip patch antennas. They include the use of varactor diodes,
shorting pins, optically controlled diodes, adjustable air gap, stacked geometries
and reactive loading. The advantages and disadvantages of these methods are
discussed. Finally, the method of increasing the absolute bandwidth by electro-
magnetic coupling of a fed and a parasitic patch in a stacked geometry is
discussed.
3.6 Acknowledgments
The authors wish to acknowledge the assistances of Dr. K.M. Luk and Mr. T.
Huynh. Dr. Luk provided the numerical data for the majority of illustrations for
the rectangular, equitriangular and annular-ring patches, while T. Huynh con-
tributed to some of the computations for the rectangular and circular patches.
3.7 References
1 DESCHAMPS, G. A.: 'Microstrip microwave antennas'. Presented at the 3rd USAR Sym-
posium on Antennas, 1953
2 BAHL, I. J., and BHARTIA, P.: 'Microstrip antennas' (Artech House, Mass., 1980)
3 JAMES, J. R., HALL, P. S., and WOOD, C.: 'Microstrip antenna theory and design' (Peter
Peregrinus, 1981)
4 IEEE Trans., Jan. 1981, AP-29
5 CARVER, K. R., and MINK, J. W.: 'Microstrip antenna technology', IEEE Trans., 1981,
AP-29, pp. 2-24
6 MAILLOUX, R. J., McILEVENNA, J. F., and KERNWEIS, N. P.: 'Microstrip array
technology', IEEE Trans., 1981, AP-29, pp. 25-39
7 LO. Y. T., SOLOMON, D., and RICHARDS, W. F.: 'Theory and experiment on microstrip
antennas', IEEE Trans., 1979, AP-27, pp. 137-145
8 RICHARDS, W. F., LO, Y. T., and HARRISON, D. D.: 'An improved theory for microstrip
antennas and applications', IEEE Trans., 1981, AP-29, pp. 38-46
9 DERNERYD, A. G.: 'Analysis of the microstrip disk antenna element', IEEE Trans., 1979,
AP-27, pp. 660-664
10 DERNERYD, A. G.: 'Extended analysis of rectangular microstrip resonator antenna', IEEE
Trans., 1979, AP-27, pp. 846-849
Characteristics of microstrip patch antennas 2 15
II DERNERYD. A. G.: 'Microstrio. array. antenna'. Proc. 6th European Microwave Conference,
1976, pp. 339-343
12 RICHARDS, W. F., LO, Y. T., and SOLOMON, D.: 'Theory and application for microstrip
antennas'. Proc. Workshop on Printed Circuit Antenna Technology', New Mexico University,
Las Cruces, 1979, pp. 8.1-8.23
13 JAMES, J. R., and HENDERSON, A,: 'High-frequency behaviour of microstrip open-circuit
terminations', IEE J Microwaves, Optics & Acoustics, 1979, 3, pp. 205-218
14 WOOD, C.: 'Analysis of microstrip circular patch antennas', IEE Proc., 1981 128H. pp. 69-76
15 FONSECA, S. B. A,, and GIAROLA, A. J.: 'Microstrip disk antennas. Pt. I: Efficiency of
space wave launching', IEEE Trans., 1984, AP-32, pp. 561-567
16 RICHARDS, W. F., LO, Y. T., and HARRISON, D. D.: 'Improved theory of microstrip
antennas', Electron. Lett., 1979, 15, pp. 42-44
17 RICHARDS, W. F., LO, Y. T., and SIMON, P.: 'Design and theory of circularly polarized
microstrip antennas'. IEEE AP-S International Symposium Digest, June 1979, pp. 117-120
18 LO, Y. T., and RICHARDS, W. F.: 'Perturbation approach to design of circularly polarized
microstrip antennas'. Electron. Lett., 1981, 17, pp. 383-385
19 SHEN, L. C., LONG, S. A,, ALLERDING, M. R., and WALTON, M. D.: 'Resonant
frequency of a circular disc, printed-circuit antenna', IEEE Trans., 1977, AP-25, pp. 595-596
20 DAHELE, J. S., and LEE, K. F.: 'Effect of substrate thickness on the performance of a
circular-disk microstrip antenna', IEEE Trans., 1983, AP-31, pp. 358-360
21 HELSZAJN, J., and JAMES, D. S.: 'Planar triangular resonators with magnetic walls', IEEE
Trans., 1978, MlT-26, pp. 95-100
22 KEUSTER, E. F., and CHANG, D. C.: 'A geometrical theory for the resonant frequencies
and Q factors of some triangular microstrip patch antennas', IEEE Trans.,1983, AP-31,27-34
23 DAHELE, J. S., and LEE, K. F.: 'Experimental study of the triangular microstrip antenna'.
IEEE AP-S International Symposium Digest, 1984, pp. 283-286
24 LUK, K. M., LEE, K. F., and DAHELE, J. S.: 'Theory and experiment on the equilateral
triangular microstrip antenna'. Proc. 16th European Microwave Conference, 1986, pp. 661-
666
25 SCHELKUNOFF, S. A,: 'Electromagnetic waves' (Van Nostrand, New York, 1943) Chap. 10
26 MINK, J. W., 'Circular ring microstrip antenna elements'. IEEE AP-S Int. Symp. Digest, June
1980, pp. 605-608
27 CHEW, W. C.: 'A broad-band annular-ring microstrip antenna', IEEE Trans., 1982, AP-30,
pp. 918-922
28 ALI, S. M., CHEW, W. C., and KONG, J. A,: 'Vector Hankel transform analysis of annular-
ring microstrip antenna'. IEEE Trans., 1982, AP-30, pp. 637-644
29 DAHELE, J. S., and LEE, K. F.: 'Characteristics of annular-ring microstrip antenna', Elec-
tron. Lett., 1982, 28, pp. 1051-1052
30 LEE, K. F., and DAHELE, J. S.: 'Theory and experiment on the annular-ring microstrip
antenna', Ann. des Telecomm., 1985, 40, pp. 508-515
31 SCHNEIDER, M. V.: 'Microstrip lines for microwave integrated circuits', Bell Syst. Tech. J.,
1969.48, pp. 1421-1444
32 OWENS, R. P.: 'Curvature effect in microstrip ring resonators', Electron. Lett., 1976, 12, pp.
356-357
33 WU, Y. S., and ROSENBAUM, F. J.: 'Mode chart for microstrip ring resonators', lEEE
Trans., 1973, MlT-21, pp. 487-489
34 DAS, A., DAS, S. K., and MATHUR, S. P.: 'Radiation characteristics of higher-order modes
in microstrip ring antenna', IEE Proc., 1984, 131, pp. 102-106
35 EL-KHAMY, S. E., EL-AWADI, R. M., and EL-SHARRAWY, E-B. A,: 'Simple analysis
and design of annular ring microstrip antennas', IEE Proc., 1986, 133H. pp. 198-202
36 BHA'ITACHARYYA, A. K., and GARG, R.: 'Input impedance of annular ring microstrip
antenna using circuit theory approach', IEEE Trans., 1985, AP-33, pp. 369-374
RICHARDS, W. F., OU, J. D., and LONG, S. A,: 'A theoretical and experimental investiga-
37
2 76 Characteristics of microstrip patch antennas Characteristics of microstrip patch antennas 2 17

tmn of annular, annular sector, and circular sector microstrip antennas', IEEE Trans., 1984. 62 CHEN, C. H., TULINTSEFF, A,, and SORBELLO, R. M.: 'Broadband two-layer microstrip
AP-12, pp. 864-866 antenna'. IEEE AP-S Int. Symp. Digest, 1984, pp. 251-254
PALANISAMY, V., and GARG, R.: 'Rectangular ring and H-shaped mlcrostrip antennas - 63 DAHELE. J. S., TUNG, S. H.. and LEE, K. F.: 'Normal and inverted configurations of the
Alternatives to rectangular patch antenna', Electron. Lett., 1985, 21, pp. 874-876 broadband electromagnetic coupled microstrip antenna'. IEEE AP-S Int. Sym. Digest, 1986,
YU. I. P.: 'Low profile circularly polarized antenna', NASA Report N78-15332, 1978 pp. 841-844
SHEN, L. C.: 'The elhptical microstrip antenna with circular polarization', IEEE Trans., 1981, 64 LEE, R. Q., LEE, K. F., and BOBINCHAK, J.: 'Characteristics of a two-layer electromagnet-
AP-29, pp. 90-94 ically coupled rectangular patch antenna', Electron. Lett.. 1987, 23, pp. 1070-1072; also IEEE
LONG, S. A., SHEN, L. C., SCHAUBERT, D. H., and FARRAR, F. G.: 'An experimental AP-S Int. Sym. Digest, 1988, pp. 948-951
study of the circular-polarized elliptical printed circuit antenna', IEEE Trans., 1981, AP-29, 65 RIVERA, J., and ITOH, T.: 'Analysis of an electromagneticallycoupled patch antenna'. IEEE
pp. 95-99 AP-S Int. Sym. Digest, 1983, pp. 170-173
BAILEY, M. C., and DESHPANDE, M. D.: 'Analysis of elliptical and circular microstrip
antennas using moment method', IEEE Trans., 1985, AP-33, pp. 954-959
BHARTIA, P., and BAHL, I.: 'A frequency agile microstrip antenna', IEEE AP-S Int. Symp.
Digest, 1982, pp. 304-307
SCHAUBERT, D. H., FARRAR, F. G., SINDORIS, A. R., and HAYES, S. T.: 'Microstrip
antennas with frequency agility and polarization diversity', IEEE Trans., 1981, AP-29, pp.
118-123
SENGUPTA, D. L.: 'Resonant frequency of a tunable rectangular patch antenna', Electron.
Lett., 1984, 20, pp. 614-615
LAN, G. L., and SENGUPTA, D. L.: 'Tunable circular patch antennas', Electron. Lett., 1985,
21, pp. 1022-1023
LO, Y. T., and RICHARDS, W. F.: 'Theoretical and experimental investigations of a micro-
strip radiator with multiple linear lumped loads', Electromagnetics, 1983, 3, pp. 371-385
DARYOUSH, A. S., BONTZOS, K., and HERCSFELD, P. R.: 'Optically tuned patch
antenna for phased array applications'. IEEE AP-S Int. Syrn. Digest, 1986, pp. 361-364
DAHELE, J. S., LEE, K. F., and HO, K. Y.: 'Mode characteristics of annular-ring and
circular disc microstrip antennas with and without airgaps'. IEEE AP-S Int. Sym. Digest, 1983,
pp. 55-58
LEE, K. F., HO, K. Y., and DAHELE, J. S.: 'Circular-disk microstrip antenna with an air
gap', IEEE Trans., 1984, AP-32, pp. 880-884
DAHELE, J. S., and LEE, K. F.: 'Theory and experiment on microstrip antennas with
airgaps', IEE Proc., 1985, 132H, pp. 455460
LEE, K. F., and DAHELE, J. S.: 'The two-layered annular ring microstrip antenna', Int. J.
Electronics, 1986, 61, pp. 207-217
LONG, S. A., and WALTON, W. D.: 'A dual frequency stacked circular disc antenna', IEEE
Trans., 1979, AP-27, pp. 270-273
DAHELE, J. S., and LEE, K. F.: 'A dual-frequency stacked microstrip antenna'. IEEE AP-S
Int. Sym. Digest, 1982, pp. 308-31 1
DAHELE, J. S., LEE, K. F., and WONG, D. P.: 'Dual-frequency stacked annular-ring
microstrip antenna', IEEE Trans., 1987, AP-35
ZHONG, S. S., and LO, Y. T.: 'Single-element rectangular microstrip antenna for dual-
frequency operation', Electron. Let?., 1983, 19, pp. 298-300
DAVIDSON, S. E., LONG, S. A., and RICHARDS, W. F.: 'Dual-band microstrip antennas
with monolithic reactive loading', Electron. Lett., 1985, 21, pp. 936-937
DAHELE, J. S., and LEE, K. F.: 'Top-loaded single and coupled microstrip monopoles'.
IEEE AP-S Int. Sym. Digest, 1983, pp. 47-50
MclLVENNA. J., and KERNWEIS, N.: 'Modified circular microstrip antenna elements',
Electron. Lett., 1979, 15, pp. 207-208
SABBAN, A.: 'A new broadband stacked two-layer microstrip antenna'. IEEE AP-S Int. Sym.
Digest, 1983, pp. 63-66
BHATNAGAR, P. S.. DANIEL, J.-P., MAHDJOUBI, K., and TERRET, C.: 'Experimental
study on stacked triangular microstrip antennas', Electron. Lett., 1986, 22, pp. 864-865
 Chapter 4

Circular polarisation and

bandwidth
M. Haneishi and Y. Suzuki

Microstrip antennas are widely used as an efficient radiator inmany commun-

ication systems [I]. One of the most interesting applications is their use for
transmitting or receiving systems ,required for circular polarisation [2-51. A
circularly polarised microstrip antenna can be classified into two categories, e.g.
single- or dual-fed types. The classification of an antenna is based upon the
number of feeding points required for circularly polarised waves. The singly-fed
antenna is useful, because it can excite circular polarisation without using an
external polariser. Therefore, it is important to understand the radiation mech-
anism of the antenna. However, one of the most serious problem in such an
antenna is the considerable narrowness of the bandwidth compared to ordinary
microwave antennas. This is a serious problem for the practical application of
this antenna. For this reason, it is also important to study some wideband
techniques.
In this Chapter, the various types of circularly polarised antennas are first
briefly introduced in Section 4.1. In Section 4.2, a simple design method for a
singly-fed antenna is described together with some useful design data. This
method is useful in understanding its radiation mechanism and to roughly
design it. However, if the general radiation mechanism and an accurate design
method are required, then the more exact treatment, developed in Section 4.3,
is necessary. In Section 4.4, some considerations of mutual coupling are des-
cribed. Finally, three kinds of wideband techniques are introduced in Section
4.5.

4.1 Various types of circularly polarised antennas

There are many types of circularly polarised (CP) printed antennas, which are
widely used as efficient radiators in many communication systems. Fig. 4.1
shows basic arrangements for various types of CP-wave printed antennas. In
this Section, we describe briefly techniques for designing such CP printed
antennas.
 220 Circular polarisation and bandwidth
4.1.1 Microstrip patch antennas
A microstrip antenna is one of the most effective radiators for exciting circular
polarisation. A circularly polarised microstrip antenna is categorised into two
types by its feeding systems: one is a dual-feed CP antenna with an external
polariser such as 3dB hybrid, and the other is a singly-fed one without a
polariser. The classification of antennas is based upon the number of feeding
point required for CP excitation.
(a-1) Dual fed patch
(a-2) Singly fed patch
( d l Microstrip printed slot
Fig. 4.1 Various types of circularly polarised printed antennas
(a) Dual-fed CP patch antenna: The fundamental configurations of a dual-
fed CP patch antenna are shown in both Fig. 4.1 (a-1) and Fig. 4.2 (a). The
patches are fed with equal amplitude and 90' out of phase by using an external
polariser. As shown in the Figure, these antennas are also divided into two
categories by the shape of an external polariser: one is the 3 dB hybrid type and
the other is an offset-feeding one.
As is well known, a 3 dB hybrid such as a branch-line coupler produces fields
of equal amplitude but 90' out of phase at its centre frequency. Therefore,
setting the outputs of such a hybrid to the edges of the patch, the antenna acts
as a CP radiator. It is necessary to note that each input terminal of a hybrid,
however, gives an opposite sense of circular polarisation. In the present case,
Circular polarisation and bandwidth 221
both the input VSWR and ellipticity bandwidth are broad, since a 3 dB hybrid,
in general, has a broadband nature.
The other category is the offset-feeding CP antenna. Here, offset feeding lines,
with one quarter wavelength longer than the other, are set at the edges of the
patch, as shown in Fig. 4 . 2 ~One
. of the most serious disadvantages of this type
of antenna is the narrow bandwidth, since the frequency dependency of an
offset-feeding line is greater than that of the usual hybrid.
-
RHCP LHCP
-
hybrid 8
RHCP LHCP
RHCP
( a ) Dual f e d CP patches ( b ) Singly f e d CP p a t c h e s
Fig. 4.2 Typical arrangements for circularly polarised microstrip antennas
LHCP: Left-hand circular polarisation
RHCP: Right-hand circular polarisation
(b) Singly-jed CP patch antenna: A singly-fed CP antenna may be regarded
as one of the simplest radiators for exciting circular polarisation. The typical
configurations of this antenna are shown in Fig. 4.2b. It is important to note that
the generated mode in this case is usually excited in an electrically thin cavity
region of the microstrip antenna. Accordingly, the operational principle of this
antenna is based on the fact that the generated mode can be separated into two
orthogonal modes by the effect of a perturbation segment such as a slot or other
truncated segment [6-7,10-1 I]. Consequently, by setting the perturbation seg-
ment to the edge of the patch, the generated mode is separated into two
 222 Circular polarisation and bandwidth Circular polarisation and bandwidth 223
orthogonal modes 1 and 2. The typical amplitude and phase diagrams after the present case, on setting the radiating elements such as the strip and the slot
perturbation are shown in Fig. 4.3, together with typical samples of antennas. to the maximum positions of V, and I,,these radiating elements radiate trans-
The radiated fields excited by these two modes are, in general, perpendicular verse and longitudinal electric fields E,, and E,,, respectively, in the boresight
to each other, and orthogonally polarised in the boresight direction. When the direction. The fields E,, and E,, can be excited in equal amplitude and at 90 out
amount of perturbation segment is adjusted to the optimum value, modes 1 and
2 are excited in equal amplitude and 90' out of phase at the centre frequency,
as shown in the Figure. This enables the antenna to act as a CP radiator in spite
of single feeding. This antenna has several advantages compared to dual-fed
ones and can excite CP radiation without using an external polariser. Design
techniques will be described in detail in Sections 4.2 and 4.3.

input- microstrip l i n e M :
;C%+ l1

(a) Radiating element for composite-type

circularly polarised printed antennaI81

( polarity)

input
E

input
matched - E

i-----
Frequency i matched

Fig. 4.3 Amplitude andphase diagrams for singly-fed circularly polarisedmicrostripantennas input

4.1.2 Other types of circularly polarised printed antennas

(b) Various arrangements of rampart line antennas[6]
In this section, we describe briefly the fundamental design procedures for others
types of CP antennas. Fig. 4.4 Typical arrangements for travelling-wave-type printed antennas

( a ) Composite type of CP printed antenna: Fig. 4 . 4 shows

~ the fundamental
configuration of a composite-type CP antenna [8]. The antenna is composed of
the combination of a half-wavelength-long strip conductor and a slot in the
ground plane. If the microstrip feeding line is short-circuited at 1 = 0, a stan-
ding-wave voltage V, and current I, occurs along the microstrip feeding line. In
of phase, if the strip and the slot are spaced one-quarter wavelength apart and
the coupling between the radiating element and feeder is controlled to be
identical in value. Therefore, this type of antenna acts as CP radiator without
using any external polariser. Details of design techniques for this antenna will
be discussed in Chapter 13.
 224 Circular polarisation a n d bandwidth
(b) Discontinuity type of C P printed antenna: A rampart-line antenna is a
typical radiator for the discontinuity type [6,9]. Fig. 4.4b shows typical rampart
line antennas that act as CP and LP radiators. Each radiator consists of a
microstrip meander line having a series of corner bends. The antenna also has
a matched load at the open end of the meander line. In this system, radiation
occurs mainly from the discontinuity section of the meander line such as a
corner bend. Therefore, both C P and LP rampart line antennas can be easily
fabricated by controlling the length L, width Wand period P of the meander
line. If L, Wand P a r e adjusted to 412, 3$/4 and ,Ig, respectively, for a unit cell
of the meander line (As is the wavelength of the travelling wave along the
meander line), the antenna acts as a CP radiator. When L = 2A8/3, W = 413
and P = 2Ag/3, the antenna radiates horizontal polarisation, while
L = W = 414, P = 1,/2 excites vertical polarisation, as shown in Fig. 4.46.
Details of the design procedure will be discussed in Chapter 13.
Circular polarisation a n d bandwidth 225
equations. while a perfect magnetic wall is assumed as a boundary condition at
the antenna peripheries (x = fa/2,y = 4 b/2);
4" = V, sin k x
4h = VOsin k y
where Vo = $/a and k = z/a.
The eigen function 4, is concerned with the field distribution of TM,, mode,
and 4, with that of TMo,, mode. By setting the perturbation segment As at an
appropriate position of the antenna, as shown in Fig. 4.5, two orthogonally
polarised modes are excited in a cavity region of the antenna.

4.2 Simple design techniques for singly-fed circularly polarised microstrip

antennas

This Section gives a brief description of design techniques for singly-fed radiat-
ors together with some useful experimental results. The approach is based on the
variational method, and is useful for understanding the mechanism of CP ( a ) Standard patch
radiation from such singly-fed radiators.

4.2.I Recfangular type

In general, microstrip antennas are divided into two types by the shape of
radiating element: rectangular type and circular type. However, since the rectan-
gular patch antenna is considered to be a fundamental device for exciting CP
radiation, the design techniques for this type are discussed first.

(a) Fundamental configuration of rectangular CP-wave antenna: The funda-

mental configurations of the antenna and its co-ordinate system are shown in
Fig. 4.5. In type A, the feeding point F is placed on the x- or y-axis, whereas in
(b) Singly fed circularly polarised patch
type B, F is placed on the diagonal axis. In both cases, the perturbation segment
As is set at an appropriate location in the patch element to excite CP radiation. Fig. 4.5 Fundamental configurations of singly-fed rectangular patches (From Reference I I )
Here, we describe briefly the sense of direction for the CP-wave. Right-hand or
left-hand CP radiation can be achieved by setting feeding points at appropriate The new eigen function 4' and the new eigen value k', after perturbation by
locations such as F(f Q,, 0) and F(0, -+eo), as shown in Fig. 4.6. the segment, are determined by the following equations [I 1, 12, 151:

(b) EfSecf of perturbation segment: The effect of perturbation segment As for 4' = P4, + Q 4 b I
the type-A antenna is described first, since this type of radiator is a basic device
for exciting CP radiation.
The eigen functions 4 , and &,, which are excited in an electrically thin cavity
region of the square patch, are generally given mathematically by the following
where P and Q are unknown expansion coefficients of the new eigen function 4'.
226 Circular polarisation and bandwidth Circular polarisation and bandwidth 227

The new eigen value k' of the antenna can be derived by employing the The eigen values used in eqn. 4.6 are assumed to be k,' = k,,' = k by means of
following matrix, since eqn. 4.2 is a variational-expression form: first-order approximation.

det I k2 + 41 - k 2 ( 1 + P I )
q12 - V2 P i 2 + 92
q12 - K2PI,
- kf2(1 + PI)
Furthermore, the turn ratios N,' and Nb', which correspond to the energy
distribution ratios for both the 4,' and 4,' modes after perturbation, are defined
as [I I]
In case of the type-A antenna, the parameters in eqn 4.3, such a s p , , p2, q , , q,, Nb = ($?/a)(sin kx - sin ky)
p , , and q,,, are expressed by the following equations [I I]: (4.7)
Nb = (@/a)(sin kx + sin ky)
In the case of the type-B antenna shown in Fig. 4.5, the eigen functions & ,
$6 and other parameters can also be derived by similar calculations employed
for type A. The equations obtained by these calculations are as follows:

Substituting eqn. 4.4 into eqn. 4.3, the new eigen values k,' and k,' for type A
are given as

where k,' and k,' correspond to the eigenvalues of the new orthogonal eigen
functions, 4,' and 4,', respectively.
Using eqn. 4.5, new sets of resonant frequencies for the 4,' and modes are
easily obtained as follows:

f, = for + Af: = h , ( l - 2 W S )
f b = hr + AfL. = A,

wheref,, is the resonant frequency for a normal square patch before perturba-
tion, and Af,' and Afb' are the shifts of resonant frequencies for the 4,' and 4,'
modes after perturbation.
Normalising the new eigen functions for the 4,' and 4,' modes, the unknown
expansion coefficients P and Q are determined as follows [I I, 121: where V, = I/a and k = nla.
For 4,' mode, Using eqns. 4.1-4.7, we can derive the equivalent circuit for the type-A
antenna. Furthermore, the equivalent circuit of the type-B antenna after per-
P, = (I/$) (1 - 2As/S) 2. (I/$) turbation can also be derived using the relations given in eqns. 4.8. The circuit
Qa = ( - 1 1 4 ) ( I - 2As/S) 2. (-I/$) for both the types of antennas is shown in Fig. 4.7. In this circuit, T', and T',
For 4,' mode
represent ideal transformers having turn ratios Nb and &, and 5 is input
voltage applied to the 1-1' terminal.
P, = Q, = (%'5)
(c) Condition requiredfor CP-wave radiation: In this Section, conditions for
Finally, using eqns. 4.1, 4.2 and the expansion coefficient, the new eigen func- exciting CP-wave radiation are determined by use of the preceding equivalent
tions 4,' and 4,' are given in a closed form by circuit. As is well known, the equivalent conductances Gh and Gb in the circuit
are expressed as the sum of the radiation, dielectric and copper losses. However,
in normal patches having adequate radiation efficiency above 90%, radiation
loss is dominant compared with the other losses.
 228 Circular polarisation and bandwidth Circular polarisation and bandwidth 229

Consequently, the equivalent conductances Gb and G ; are mainly caused by
the radiated fields resulting from the patch antenna. In other words, the induced
c,
voltages and generated on G:, and Gb can be assumed to correspond to the
radiated fields caused by the orthogonal 4: and 4; modes.
(Ijblt,) = + j is satisfied. Accordingly, these antennas act as a CP radiator by
setting the relative amplitude and phase between the two orthogonal modes at
I &/?,I = I and arg ( & I t ) = f90, respectively.
Applying the above conditions to eqn. 4.9, turns ratios are required to satisfy
the relation INf,/N:,I = 1. In addition, when this restriction is applied to the
type-A antenna, it is necessary to place the feeding point F on the x-axis for
(Nb/N:,) = 1. Contrariwise, the feeding point F is required to be placed on the
y-axis by another restriction (N',/K)= - 1.

1 Nb:l
dA- mode

( b l LHCP (1r.l ~ a / 2 1

I
~'=Klsinkx N ~ = K l r i n k xI
Fig. 4.6 Feeding locations required for circular polarisation -sin kyl
1-; L(1-1 *)N;=l<l sink* I;=b1(l-9+) N;*KI $inky)
RHCP: Right-hand circular polarisation + i n k
LHCP: Left-hand circular polarisation
eo = feed location
%=-2.ds
t,,
K=fi/a
{ 4, - d s i
I.,
f =~

Applying network analysis to the equivalent circuit, the complex amplitude

ratio t/cin the two orthogonal modes is given as follows:

Fig. 4.7 Equivalent circuit for rectangular circularly polarised patch antennas

Setting the feeding point at each location, the expression for the complex
amplitude ratio is shown as follows:
t,
where and 6 are input admittances for the orthogonally polarised 4: and 4:
modes, respectively. In addition, the unloaded Q factors in the above equation
are expressed as Q,, = Qob = Q, to first-order approximation, where Q,, and
Q,, are the unloaded Q factors of the 4: and 4: modes.
From eqn. 4.9, radiation of CP waves by these radiators may be expected if
 230 Circular polarisation a n d bandwidth Circular polarisation and bandwidth 237
By application of the CP conditions satisfying I ~ / E=I 1 and a r g ( c / < ) = (e) Radiation characteristics of CP antennas: In order to verify the validity
+
- 90" to the above equation, an important relation between Q, and (As/S) is of the above design procedures, typical samples of CP antennas were fabricated
obtained as follows: and tested at X-band. These antennas were fed with a coaxial probe to avoid the
influence of unwanted radiation from the feeding networks. The radiation

where M = (1 + mAs/S), N = (1 + nAs/S), and m and n are the constants in

f;,= fo,(l +
mAslS) andf, = fo,(l +
nAs/S).
In case of the type-A radiator, constants m and n are shown as m = - 2 and
n = 0, as mentioned previously. Substituting these values into eqn. 4.10, and
carrying out some modifications of the equation, the most important expression
for designing purposes is easily obtained as follows:

This expression is simple in form but very useful for actual design of the type-A
-*a,+.:-.
I'LUL'ZLUI .
Furthermore, in case of the type-B antenna, a basic equation for design can
also be derived by the use of similar techniques, and the expression is as follows:
I I I I I L I ' 0
0 2 4 6 8
substrate th~ckness(t/ho) (X~O-~)
The relations for both basic expressions are illustrated in Fig. 4.86 by solid lines.
(a) Unloaded Q(Qn) and radiation efficiency
They help to provide important design parameters such as the amount of
perturbation (As/S) required for CP radiation.

(d) Basic design procedures for CP-wave antennas: In designing CP-wave

antennas, it is necessary to estimate the value of unloaded Q(Q,) as a function
of substrate thickness t for an antenna. Therefore, theoretical values of Q, were
calculated for a typical sample (a = 9.14mm, t = 0.6 mm, E, = 2.55). employ-- -
ing a commonly used technique [6, 101. The theoretical values agreed well with
the experimental ones, as shown in Fig. 4 . 8 ~After
. determining the value of Q,,
the design of an antenna can be achieved by the following procedures:
(i) Using the relations shown in Fig. 4.8a, the unloaded Q(Q,) of the square
patch is chosen so as to ensure that the radiation efficiency q of the patch will
exceed 90%.
(ii) The amount of perturbation (AsjS) required for CP-wave excitation is 01 I I I I I
0 20 30 40 50
determined using the relation between Qo and (As/S) shown in Fig. 4.86. Unloaded Q(QO)
(iii) Finally, the input impedance of the test antenna is matched to that of the (b) Relations between Q0 and amount of perturbation
feed network by the offset loading technique of coaxial probe or using a segment ( A S I S )
quarter-wavelength transformer. Fig. 4.8 Fundamental design chart for singly-fed circularly polarised patch antennas
The approaches described above are performed without considering the effect
due to fringing fields. However, when the fringing effect is taken into considera- patterns measured by a spinning dipole are shown in Fig. 4.9. As seen from the
tion [6], the procedures help to provide more accurate design parameters re- Figure, the ellipticity of the test antenna is less than 0.5dB in the boresight
quired for CP radiation. direction. Furthermore, the ellipticity is within 1.5dB in the desired angular
region of 45'.
232 Circular polarisation and bandwidth Circular polarisation and bandwidth 233

Fig. 4.10 shows the measured impedance characteristics of the typical CP-
wave antenna. From these results, it is found that loop 1 in the impedance-plot
locus depends on the degree of mode separation; namely, loop 1 becomes larger
in area with an increase in mode separation, and converged to a point when the
mode separation is reduced. In any case, however, the best ellipticity can be
obtained at or near the peak of loop 1 in the impedance locus.

Fig. 4.10 Typical measured impedance characteristics for rectangular singly-fed circularly
polarised patch antennas (From Reference 7 I )
t/& = 0.018, E, = 2.55, tan6 = 0.0018 and ( e 0 l ( a / 2 ) )= 0.3
la) P a t t e r n of t y p e - A circularly polarlsed p a t c h

I substrate
180 135 90 45 0 45 90 135 180
(deg.)
(a) Standard patch
Ibl S~LLFCII of t y p e - 0 c ~ r c u l a r l y polarlsed patch

Fig. 4.9 Typical radiation patterns of rectangular singly-fed circularly polarised patches,
(From Reference 11)
[ t / i o= 0,018, E, = 2.55, tang = 0.0018. ( p o / ( a / 2 ) )= 0.3, !A( = W, = io (Fig.
4.5a), and X-band

4.2.2 Circular type

This Section gives a brief description of a design technique for a circular
antenna. The geometry and feed system of the antenna are shown in Fig. 4.1 1. ( b ) c i r c u l a r CP
( c l Feeding s y s t e m
In this antenna, the perturbation segment As is also located a t a specific location. elements
The degenerate mode is also separated from the dominant mode (TM,,,) in the Fig. 4.1 1 Fundarnentalconfigurationsfor circular singly-fed circularly polarised patch anten-
antenna into two orthogonal modes by the effect of the perturbation segment. nas
 234 Circular polarisation and bandwidth
The equivalent circuit after perturbation is useful for network analysis of the
radiator. Fortunately, the equivalent circuit for the antenna can easily be
obtained by employing the same procedures as in the previous Section. The
equivalent circuit for the antenna after perturbation is shown in Fig. 4.12 [7].
Fig. 4.1 2 Equivalent circuit for circular singly-fed circularly polarised patch antenna (From
Reference 7)
The CP-radiation condition for the circular patch can be determined using the
above equivalent circuit. Namely, by application of the design procedures
employed for the rectangular patches, the CP-wave condition for the circular
one is obtained as
where (ASIS), Qo and x,, correspond to the amount of perturbation, the un-
loaded Q and the eigen value of the dominant TM,,, mode, respectively.
Using the above equation, the relation that gives a CP-radiation condition for
the antenna is indicated by the dotted-line in Fig. 4.8b. This Figure helps to
Circular polarisation and bandwidth 235
provide important design parameters such as the amount of perturbation
required for CP-wave radiation. In actual design, however, it is important to
note that the Qo of the circular patch becomes equivalent to that of the rectan-
gular one, if both the patches are designed to have the same resonant
frequencies.
For this reason, the circular CP antennas can be designed as easily as rectan-
gular antennas by employing the relations shown in Fig. 4.8. However, the
fringing effect is disregarded in above approaches. If the effect of fringing fields
is taken into consideration [6], the experimental results agree well with theory,
as mentioned below.
In order to verify the validity of the design procedure, some circular CP
patches were fabricated and tested at X-band. These samples were fabricated
using a substrate consisting of copper-clad 0.6mm-thick Teflon glass fibre with
a dielectric constant of 2.55 and a loss tangent of approximately 0.0018. The
boresight ellipticity of the test antenna was about 0.5dB or less, and the
radiation patterns revealed as a high a level of performance as those of the
rectangular patch. In addition, the ellipticity bandwidth within 3 dB was about
1% with a substrate thickness of (t/lo) = 0.019, and about 2% with a thickness
of (tll,) = 0.037. These results indicate that, as the substrate thickness in-
creases, the ellipticity bandwidth also increases.
Furthermore, the trend of the impedance locus plot of the antenna coincided
with that of the rectangular one shown in Fig. 4.10.
4.3 More exact treatment of singly-fed circularly polarised microstrip antennas
The patch radiator can easily be modified from a circular, square or rectangular
shape so as to excite circularly polarised waves with a single feed as mentioned
previously. In addition to these shapes [4, 51, a specially shaped pentagonal [2],
triangular [I31 or elliptical radiator [3] can also radiate circular polarisation.
Furthermore, it has been known that the polarisation and resonant frequency
can be conveniently controlled by inserting posts at suitable locations within the
patch boundary [14]. However, it is not generally easy to analyse such antennas
accurately, so designers are often forced to use cut-and-try methods to realise
the desired characteristics.
In this Section, an analysis, based on variational method [IS] and modal
expansion technique [16], is briefly summarised for an arbitrarily shaped micro-
strip antenna with multi-terminals before starting the discussion concerning a
singly-fed circularly polarised antenna. Using the results of analysis the con-
ditions for producing circularly polarised waves are derived, and then it is
shown that a microstrip antenna, in general, can radiate circularly polarised
waves at two kinds of frequencies with a single feed. Finally, one design example
is given in order to confirm experimentally the several theoretical predictions
concerning the feed points and the operating frequencies to radiate them.
 236 Circular polarisation and bandwidth Circular polarisation and bandwidth 237
4.3.1 Analysis [ I 71 to this problem may be given by
The present method is based on the variational method applied to arbitrarily
shaped microstrip planar circuits with multi-terminals [I51 and the modal
expansion technique [16]. The following approach is more suitable and useful in where G(x, ylx,, yo) is a Green function generally expressed using the eigen
the analysis, and the design of a singly-fed circularly polarised microstrip values and eigen functions as
antenna than that based on the moment method [19]. In the present method, the
eigen values and orthonormalised eigen functions are derived from the Ray-
leigh-Ritz method under the Neumann boundary conditions [20]. The formula
for the mutual impedance is derived using the relations between the terminal In this equation, k"' and cp(') are the eigen value and eigen function for the I th
voltages, stored energies and radiated and dissipated powers. Also the equiv- mode, respectively, and can be derived by employing the Rayleigh-Ritz method
alent circuit applicable to the microstrip antenna with multi-terminals is ob- [20] for an arbitrarily shaped microstrip antenna.
tained.
( b ) Mutual impedance and equivalent circuit: In a multi-terminal microstrip
antenna, if the power is supplied only to the 9 t h terminal and the other
terminals are all open. the electric field Elq(x, y ) associated with the terminal
current I, located at (x,, y,) can be expanded in terms of series of eigen functions
as
N
El,(& Y) = 1, C F(l, X,, Y,)CP'"(X,Y)
I= I
(4.15)

Fig. 4.13 Structure of analytical model and co-ordinate system where the unknowns F(1, x,, y,) are functions of the mode number I and the
terminal location (xq, y,). Eqn. 4.15 implies that a mutual impedance can be
expressed as a superposition of that for each mode as follows:
(a) Green function: The geometry of an analytical model and the co-ordinate N
system employed are shown in Fig. 4.13. The arbitrarily shaped patch is located z,., = E,(x,, Y,)/I, = Cz 1! (4.16)
on the surface of the grounded dielectric substrate with thickness t and the I- I

dielectric constant 6,. Usually, the patch is fed either by microstrip feed lines or where Zz) is the mutual impedance between the p th and q th terminals for the
coaxial probes. However, microstrip feed lines lead to problems of coupling with I th mode and can be expressed by
the patch radiator and problems of radiation, though there is an advantage in
that they can be etched together with the patch radiator. Accordingly, the
following discussion is restricted to the case of coaxial-probe use, because we
want to separate the problem of the antenna itself from that of the feed network. In the above equation, V$) and Vf) are the terminal voltages, W$) and W i: are
In this Figure, C denotes a boundary line for the patch radiator, S is its area and the time-averaged electric and magnetic stored energies, Pj:' is the radiated
A is a unit vector normal and outward to the boundary C . In many practical power, and PSf and P$ are the powers dissipated in the conductor walls and the
applications, the substrate is electrically thin, so that only a Zcomponent of the dielectric, respectively. These parameters can be derived from the fields within
electric field and the X and Y components of the magnetic field exist in the region the patch boundary according to the perturbation theory. As a result, the
bounded by the patch radiator and ground plane. Assuming e'"' time variation, mutual impedance can be expressed as
the electric field E: associated with a current source Jz located at (x,, yo) must
satisfy
(V: + k2)E: = -jw&J:(x,,y,) (4.12)
where V, is the transverse part of the del operator with respect to the Z-axis, w
signifies angular frequency, and k2 = ~ , k &with k, being a free-space wave where
number. If the perfect magnetic wall is assumed on the boundary, the solution
 238 Circular polarisation and bandwidth Circular polarisation and bandwidth 239

M:' = flcp")(x,, y,) impedance'. When viewed from each feed point, it is generally defined as

C = -E, SEO
t

with R,being the real part of the surface impedance of the conductor walls and
tan 6 being the loss tangent of the dielectric substrate. Also, P,$'is written as

where

exp { jko(xcos + + y sin 9 )sin 6)dl (

0 G 0 G
3
- (4.29)

and R is unit vector of R-axis direction in polar co-ordinate, the asterisk means
complex conjugation and Re{.) means real part in the brace. Eqn. 4.18 implies
that an equivalent circuit for a multi-terminal microstrip antenna can be re-
presented by the network model shown in Fig. 4.14, where the first resonant Fig. 4.1 4 Equivalent circuit for multi-terminal microstrip antenna (From Reference 17)
circuit, for I = 1, has a(')= 0 and thus corresponds to the mode resonating at
zero frequency. This equivalent circuit is useful for network analysis of the
microstrip antennas. where Vy) is the input voltage to the q th terminal and Zo, is the characteristic
impedance of the q th terminal. I, is the current flowing in the q th terminal and
(c) Input impedance: When the microstrip antenna has multi-terminals, the found as a solution of the following matrix equation:
coupling among the terminals in the cavity must be considered in order to derive
the accurate input impedance. Such input impedance is called an 'active input [I] = [Z1]-' [v""'] (4.32)
 240 Circular polarisation and bandwidth
where [V'"']is the input voltage vector whose typical term is Vp' and [Z'] is the
impedance matrix whose typical term is given by
In practice, when using coaxial probes as feed lines, the correction is approxim-
ated by adding the following terms [21] to the self-impedance terms:
jx, = j & tan (kot) (4.34)
However, when using striplines, such a correction is not needed as the striplines
are regarded as a part of the patch radiator, and the eigen values and eigen
functions are derived for the patch boundary including the striplines.
(d) Radiation field, directive gain and radiation &icie.~cy: The total radiatior.
field can be calculated as a superposition of that for each mode. The radiation
field for the lth mode can be represented by
So the total radiation field is obtained from
where Qt' is an unknown coefficient, and can be determined for each mode from
the boundary condition with respect to the voltage at the q t h terminal as
follows:
Next, the directive gain U at O = 0 and the radiation efficiency q are defined by
where P,, P,. and P, are the total radiated power and the dissipated powers due
Circular polarisation and bandwidth 247
to the conductor and dielectric losses, and can be written as
tW
P,, = - E,E, tan 6
2
4.3.2 Conditions for circularly polarised radiation [I81
In general, ideal circularly polarised waves are obtained when the ratio of the
two orthogonally polarised radiation-field components is equal to &j. Solving
this relationship with respect to the frequency, two kinds of frequencies at which
the circularly polarised waves are radiated can be derived through an iterative
process. Also, all the corresponding optimum feed locations can be determined
numerically.
(a) Radiation field from singly-fed circularly polarised microstrip antenna: If
the contributions from the non-resonant modes are very weak and may be
ignored, except those for the two strongest orthogonal modes necessary to
radiate the circularly polarised waves, the total radiation field may be written as
where (x,, yo) is a feed point. In the above equation, the vth and (v + I)th
modes are chosen as the two wanted orthogonal modes. If a co-ordinate system
can be fixed for convenience so as to align the X-axis with the direction of the
vth field vector Et'(0, o)and the Y-axis with the direction of the (v + 1)th field
vector EC+l)(B,w) on the boresight, then the far field given by eqn. 4.44 can be
expressed on the boresight as
where
E, = E(0) . j = R('+l'(xo,yo)Etf "(0,0 ) . j . (4.47)
Eqn. 4.45 can also be modified as
E(0) = EL(f + j j ) + ER(3 - j j ) (4.48)
where E, and EL denote the right-hand and left-hand circularly polarised
 Circular polarisation and bandwidth 243
242 Circular polarisation and bandwidth
equations can be derived:
components, respectively, and they are written as
for RHCP
g'"(w) = + --- for LHCP
(4.54)

for RHCP
(4.55)
1 1 for LHCP
ER = - (JE, - Ey) = 2 { Q ( " ) ( Xyo)E$'(O,
~, a) f
2~ where
+ ,a@+
1 ) ( x 0 ,Y ~ ) E $ + I ) ( O , a).$I (4.50)
Therefore, from EL = 0 or ER = 0, the following equation can be obtained:

Q("+')(XY ~ ), E $ + I ) (wO) ., j = T j for EL = 0

(4.51)
R(')(xo,yo)E$)(O, o ) . f for ER = 0
By eliminating + Bjfi from eqns. 4.54 and 4.55, a biquadratic equation with
'#hen cqn. 4.5i is satisfied, the resultant fieids become the circuiar poiarisations respect to the CP operating angular frequency can be obtamed as follows:
and are expressed as
w4 + u 2 [ U 2 ( o )- { ( w ( ' + ' ) )+~ ( w ( ' ) ) ~ +
}] ) } ~0
{ W ( ~ ) . O ( " + ~= (4.58)
RHCP
yo)[E$)(O,o).f ] ( f T j j )
E(0) = Q(")(xo, where

where RHCP and LHCP mean right-hand and left-hand circular polarisation,
respectively.
( b ) CP operatingfrequency and optimum feed location: A microstrip antenna
may become singly-fed circularly polarised antenna when its dimensions are
adjusted to suitable values as mentioned previously. In addition, when the
operating frequency and feed point are chosen correctly, good circularly pol-
arised waves can be radiated. The frequency at which the ideal circularly
polarised waves are excited is called the CP operating frequency. This Section
indicates how the CP operating frequency and the corresponding optimum feed
location are derived.
Substituting eqns. 4.20 and 4.38 into eqn. 4.51, the following expression can
be derived for the CP operating conditions:
Through comparison between the coefficients of the real parts and the imaginary
parts on both sides of the above complex equation, the following simultaneous
In this case the significant roots of eqn. 4.58 are given by
where
D(w) = [{w") - w('+"I2 - U 2 ( o ) .] [{a(")
+ &+I) }2 - u 2 ( a ) ] 2 0 (4.61)
Eqn. 4.60 shows that eqn. 4.58 has two significant roots, provided that the
following CP operating condition, derived from an inequality of eqn. 4.61, is
satisfied:
low - w ( v + I ) ,1 > (4.62)
Physically, this implies that the microstrip antennas can produce circular pol-
arisation at two kinds of frequencies with a single feed. But, if the inequality of
eqn. 4.62 is not satisfied, good circularly polarised waves cannot be produced
from such antennas. However, eqn. 4.60 is not a closed-form expression, be-
cause the conductance components are, in general, a function of operating
frequency. So it is difficult analytically to find out the CP operating frequencies.
Accordingly, they are determined approximately through an iterative process;
 244 Circular polarisation and bandwidth
namely, the (p + I)th iterative solution is approximated by
using the pth iterative solution o,.Fortunately, satisfactory convergence for
eqn. 4.63 is usually obtained by about three iterations. Correct choice of the feed
point however, is, also very important for good circularly polarised radiation.
All the feed-location loci, consisting of obtimum feed points, are determined
numerically by substituting the convergence results of eqn. 4.63 into eqn. 4.54.
Next, let us consider the case when the equality in eqn. 4.62 is satisfied. Then
instead of eqn. 4.60, only one CP operating frequency is given by
rn = Jw(y)wo (4.64)
The theory developed here is quite adequate for obtaining initial design data
on the, CP operating frequencies and the corresponding feed-location loci.
However, if one wants to realise a near-perfect singly-fed circularly polarised
antenna, then experimental trimming may be necessary in the final design stages
to revise the errors due to the material-tolerances effects of the substrate used.
Examples of calculations based on this theory are presented in the following
Section.
Fig. 4.1 5 Plan view of pentagonal microstrip antenna
4.3.3 Example (13, 181
In this Section, an example is given of a singly-fed circularly polarised microstrip
antenna designed on the basis of the theory developed in the previous Section.
Several measured results are also presented for comparison with the calculated
ones. The antenna used in the experiments was made of copper-clad 3.2 mm-
thick Teflon glass fibre with a nominal dielectric constant E, = 2.55 and a loss
tangent of approximately 0.0018; it was fed by a coaxial probe to avoid the
degradation of ellipticity by unwanted radiation from the feed network.
Circular polarisation and bandwidth 245
Now, let us consider the patch radiator shown in Fig. 4.15. This Figure shows
a plan view of the patch radiator whose angles L E and L Fare right angles. The
shape of such a pentagon can be prescribed completely by two parameters c/a
and bla; i.e. the pentagon becomes a rectangle when cla = 0, an isosceles
triangle when c/a = 1, and it becomes a special pentagon proposed by Wein-
schel[2] when bla = 1.0603 and c/a = 0.3061. So it is interesting to investigate
the variation of CP operating frequency with respect to the aspect ratio cla in
the case of bla = 1.0603. This can be derived iteratively from eqn. 4.63; the
solid lines in Fig. 4.16 show the theoretical relations. In this Figure, the two
chain-dotted lines denote the resonant frequencies for the two orthogonal
modes contributing to the circular polarisation and the pair of dots indicate the
experimental results for the pentagon proposed by Weinschel. These results
show that antennas having dimensions satisfying the condition of eqn. 4.62 can
Fig. 4.16 Relations between CP operating frequencies and aspect ratio cla of pentagonal
microstrip antenna with bla = 7.0603, a = 700mm. t = 3.2mm. &, = 2.55,and
tan6 = 0.0018 (From Reference 18)
-calculated, a measured
always radiate circularly polarised waves at two frequencies. However, when
eqn. 4.62 is not satisfied, i.e. cla approaches 0 or 0.5 in this example, such
antennas cannot radiate any pure circularly polarised waves. In addition to the
above condition concerning patch dimensions and CP operating frequencies, in
order practically to obtain good circular polarisation, the antenna must be fed
 246 Circular polarisation and bandwidth
Circular polarisation and bandwidth 247
at a location so that the two orthogonal radiation fields, contributing to the
circular polarisation, are excited with equal amplitude. Such a feed location can
be determined numerically by substituting the corresponding CP operating
frequency into eqn. 4.54. The numerical results for several aspect ratios are
shown as feed-location loci in Figs. 4.17~-f[37]. These Figures show the various
pairs of feed-location loci when the aspect ratio cla is varied with bla = 1.0603
and a = 100mm. In these Figures, T, and T, show the loci when the CP
operating frequency isf;, and T,and T, show the loci when it isf,,, noting that
T, and T,, indicated by the solid line, correspond to the LHCP and T2 and T,,
indicated by the broken line, correspond to the RHCP. It is found from these
results that the triangular microstrip antenna can also radiate circularly pol-
arised waves at two frequencies. The triangular microstrip antenna has the
attraction that the area necessary for the patch radiator can be small; namely
one half to three quarters that of the nearly square one.
Fig. 4.18 Plan view of isosceles triangular microstrip antenna and feed-location loci in case
of bla = 0.98, a = 76mm. t = 3.2mm. 6, = 2.55, and tan6 = 0.0018 (From Re-
ference 13)
-RHCP, ---- LHCP

Fig. 4.17 Variation of feed-location loci with respect to aspect ratio c/a for pentagonal
microstrip an:enna with bla = 1.0603, a = 100mm. t = 3.2mm. E, = 2.55, and
tan6 = 0,0018 (From Reference 37)
---- RHCP, -LHCP
0 0 d2 ! ig60 aspect ratio bla
,A6
Let us therefore consider the triangular microstrip antenna in detail. Fig. 4.18
shows a plan view of the isosceles-triangular patch radiator with the loci T,to Fig. 4.19 Relations between CP operating frequencies and aspect ratio for isosceles trian-
r, of the theoretical feed location for each CP operating frequency, where gular microstrip antenna with a = 76mm. t = 3.2mm. E, = 2.55 and
a = 76 mm and bla = 0.98. In this Figure, T,and T2indicate the loci when the tan 6 = 0.0018 (From Reference 13)
- calculated, 0 measured
 248 Circular polarisation and bandwidth Circular polarisation and bandwidth 249
CP operating frequency is 1583.8 MHz and T, and T, indicate the loci when it good circularly polarised waves are excited at two different frequencies and that
is 1564.2MHz. In general, the shape of the isosceles-triangular patch shown in their bandwidths for 3dB axial ratio are about 0.5-0.6%. Although there are
Fig. 4.18 can be prescribed completely by introducing the aspect ratio bla as a frequency differences of about 20-25 MHz and slight different of feed point
parameter. The solid lines in Fig. 4.19 show the variation of CP operating between theory and experiment, both agrees well and also yield the excellent
frequency when bla is varied, with a =,76mm; the pair of dots represent the circular polarisation.
measured results when the aspect ratio is 0.98. From these results it can be noted
that the circularly polarised waves are always excited at two different frequen-
cies when bla is smaller than about 0.985 or greater than about 1.015 in this case.
Next, let us consider the axial-ratio characteristics. Fig. 4.20 shows the boresight
axial ratio with respect to frequency, when bla = 0.98 and a = 76mm. The

(a) Isolated patch

.o 2
a
1550 1560 1570 1580 1590 1600 1610
frequency. MHz

6 - b
\
m d
-D6 4 -
TI &
\ I
- 7 d
-
. 2- t f! i'
\ (b) Two elements array
a Fig. 4.21 Co-ordinate systems for circular patch and its array
0 I I I J
1570 1580 1590 1600 1610 1620 1630
frequency. MHz 4.4 Some considerations on mutual coupling
Fig. 4.20 Axial-ratio characteristics for isosceles triangular microstrip antenna with b /
In the design of an array, it is important to estimate the mutual coupling
a = 0.98.a = 76mm, t = 3.2mm. &, = 2.55,and tan h = 0.0018 (From Reference
131
between microstrip patch antennas [22-231. In this Section, we present a simple
a Calculated results for point B-fed case in Fig. 4.18 and measured results for point method for calculating the mutual coupling of patch antennas. The technique
6, -fed case based on the EMF method is simple and very effective for estimating the mutual
b Calculated results for point A-fed case in Fig. 4.1 8 and measured results for point coupling and the mutual admittance of antennas [22]. The geometry of the
A, -fed case analytical model and the co-ordinate system employed here are illustrated in
Fig. 4.2 1. Two patches are located on the same surface of a grounded dielectric
solid line in Fig. 4 . 2 0 ~represents the calculated results when selecting point B substrate having thickness t and the dielectric constant 8,. Using this co-ordinate
in Fig. 4.18 as a feed point, and the broken line the measured results with B, as system, and considering the field distribution of the dominant mode (TM,,,)
a feed point. In Fig. 4.20b, the solid line represents the calculated results when excited in a cavity region of the patch, the magnetic current J, due to the
selecting point A in Fig. 4.18 as a feed point and the broken line the measured dominant mode is given by
results with A, as a feed point. From these Figures it can be appreciated that
 250 Circular polarisation and bandwidth Circular polarisation and bandwidth 251

where J, corresponds to the z-component of the electric field at the periphery where
of the antenna, and J, is the maximum amplitude of an equivalent magnetic
current due to the dominant mode. Also, rT is an effective radius that contains
rn, (d, 4', +", JI) = dsin (4' - $) + iisin (4' - 4")
a fringing effect [6], S(z) is the delta function and i,. is the &-directed unit vector m2(d, 4', &", $) = dcos(4' - $) + acos(4' - 4")
at the point N(d, 4', 0) in the spherical co-ordinates.
R, = RI,=,, K, = Jml(j4n)
In order to simplify the following estimations, the equivalent magnetic cur-
rent J, is assumed to exist only in the xy-plane, as shown in Fig. 4.21. Then the e = {dZ+ 6' + 2ddcos($ - 4")}1/2
vector potential A, and the magnetic field H at P(e, 4, z) generated by this cos (4) = {dcos (JI) + ii cos (4")}/~
magnetic current J, are given mathematically by
sin (4) = {dsin ($) + ri sin(d")}le
and H: and H$are the components of the magnetic field Hqat the point Q.
By application of the EMF method to eqn. 4.67, the mutual admittance Y12
between the two patches can be easily obtained as follows:

- - 2a dsin (JI - 4")

- (4.68)
where J,($') = 2Jmcos(4'), o is the angular frequency, ko(= a=) is the Jmi(0) JUO)
free-space propagation constant and E , and po are the free-space permittivity and
permeability, respectively. R is the distance between the magnetic current J, and
the observation point P, and is given by
where J*,,(@) is the complex conjugate of m J@
2()' and i+.is the 4" directed unit
vector at the point Q.
Using eqn. 4.66 and the co-ordinate system shown in Fig. 4.216, the magnetic
Substituting eqn. 4.67 into eqn. 4.68, and carrying out numerical integration
field Hqat the point Q can be derived analytically by the following equation [22],
while it corresponds to the magnetic field at the periphery of patch antenna 2:
for eqn. 4.68, the values of mutual admittance c,
( = G I , + B,,) can be deter-
mined numerically, as shown by the solid lines in Fig. 4.22. It is important to

s = (K~I~P~) rn7{e
sin (4')
- ( d, 4,
note that the results shown in Fig. 4.22 are expressed in term of the normalised
admittance PI,(= Yl,/G,,) between the two patch antennas. Here, GI, de-
notes the self conductance for an isolated patch antenna, and is shown as
GI, = GI,/,,,. In order to verify the estimates of PI:,,,experimental work was
carried out at S-band using typical samples. The theoretical values based on eqn.
4.68 agree well with the experimental ones within the desired range, as shown
in the Figure.
In the design of an array, the mutual coupling IS,,I for a patch antenna is also
an important factor. The mutual coupling for an antenna is therefore described
here, together with the experimental results. Using the normalised admittance
z2 and the scattering S-matrix, we can express the mutual coupling by the
following equation:

where Poand g, correspond to the normalised self and mutual admittances,

respectively.
 252 Circular polarisation and bandwidth Circular polarisation and bandwidth 253
0.4/- 10.4 The mutual coupling ISl2)for a circular patch was estimated here using eqn.
4.69. The calculated values of coupling for a circular patch are shown in Fig.
4.23. As shown in the figure, the theoretical coupling for a rectangular patch
obtained by the same calculation coincides fairly well with that for a circular
patch. The samples of both rectangular and circular patches are designed to
have the same resonant frequency.
In this Section, we have presented a simple method for estimating the mutual
coupling of patch antennas. After estimating the coupling, the calculated values
were compared with the experimental ones. The calculated values agree well
with the experimental results in spite of neglecting the effects of dielectric
substrate.

4.5 Wideband techniques

As is well known, the handwidth of microstrip antennas c m basical!y be

Fig. 4.22 Normalised mutual admittance ?,
A, = 0.078 (From Reference 22)
increased by increasing the thickness of the substrate and decreasing its dielec-
tric constant [24, 271. Also, it is well known that it is quite effective to mount a
parasitic element on the original patch radiator [25]. On the other hand, in the
case of an array, the bandwidth can also be increased collectively by arranging
the antenna elements in a certain way [26].
In the following Sections, the wideband techniques for the first two methods
between two microstrip antennas, where t/
are described fully, and finally the last method is also described.

4.5.1 Design of wideband elements

The bandwidth of a microstrip antenna depends on the patch shape, the re-
sonant frequency, and the dielectric constant and thickness of the substrate. In
this Section, the relations between these parameters are derived and a design
method is described for the wideband microstrip antenna. The wider bandwidth
is usually obtained by employing a thick substrate with low dielectric constant
[24, 271. However, such an antenna, in general, has two major problems which
should be considered. One concerns the surface-wave radiation and the other
concerns the unwanted mode generation. The former may impose a limiting
factor on the maximum usable thickness of any substrate, because a practical
microstrip antenna is usually designed so as not to radiate any surface waves.
r;= 9.375 GHz
n However, the latter problem cannot be ignored for microstrip antennas having
a bandwidth greater than about 6%. Although a wideband microstrip antenna
has these problems, it has the advantage that no balun (or balanced-to-unbalan-
ced transformer) is required, so that the bandwidth can be accurately and
-
1--- circular
~atch
analytically estimated. This advantage makes it possible to design the microstrip
antenna taking into account not only the resonant frequency but also the
bandwidth.
In this Section, the relations between the design parameters are briefly sum-
element spaclng (dl+,)
marked first. Next, in order to make it easy to understand the design procedure,
Fig. 4.23 Comparison of IS,,I between circular and rectangular microstrip antennas a design example is given for a circular microstrip antenna whose bandwidth for
 254 Circular polarisation and bandwidth
Circular polarisation and bandwidth 255
a VSWR less than 2.0 is 8.75% [28]. Finally, two methods of cancelling out the
higher-order modes caused by lowering the antenna Q-factor is briefly des- when the following condition is satisfied for the coupling coefficient:
cribed.

(a) Relations between parameters necessary for design [ 2 9 ] : The antenna The results of eqns. 4.74 and 4.75 are illustrated by the broken and solid lines
bandwidth is, in general, represented as a function of the unloaded Q factor and in Fig. 4.24, respectively. Using this Figure, the unloaded Q necessary for the
the input VSWR [24]. Accordingly, if the requirements of the input VSWR and design of the antenna can be determined graphically if the requirements for the
the bandwidth are specified, the desired value for the unloaded Q factor can be bandwidth and input VSWR are specified. Also, the chain-dotted line in Fig.
determined. In this Section, the relation between them is first derived and it is 4.24 indicates the result obtained from eqn. 4.76. This relation is useful in
also shown that the product of the bandwidth and unloaded Q takes the determining the position of the feed point and the characteristic impedance
maximum value for the special characteristic impedance of feeder used. Gener- necessary for thefeeder used.
ally, the input admittance of the microstrip antenna may be approximated,
using the relative bandwidth denoted by B, and the unloaded Q denoted by Q,,
as follows:
Y, = g('y{l + jQoBr} (4.70)
where
g(')' = g ( O j ~ W

and g(') is the conductance component given by eqn. 4.24. Mg) is the turns ratio
for the 1th resonant circuit in Fig. 4.14, and is given by eqn. 4.19 or 4.20 for any
feed point. If the transmission line with characteristic admittance of Go is
connected to this antenna, the input VSWR is given by

Substituting eqn. 4.70 in the above, the equation giving the relation between the
bandwidth, unloaded Q, and input VSWR can be obtained by

QOB, = dPe - 1)(1 - Pie) (4.72)

where p is the coupling coefficient and is defined by
VSWR (P)
P = GO/g(ly
Fig. 4.24 Relationship between 0,.6,. Po and input VSWR (From Reference 29)
Eqn. 4.72 implies that, if an input VSWR of < Q is required, the product of the
Q factor and bandwidth necessary for the antenna is related only to the coupling
coefficient. As a special case, let us consider the case of P = 1, i.e. Go = g('); Next, the relation between the resonant frequency, the dimension of the patch
then eqn. 4.72 results in [24] radiator, and the thickness and dielectric constant of the substrate used is
derived. Let S be the physical area of the patch radiator and t be the substrate,
thickness. Now, if @ / t is introduced as a parameter, the product of the
However, it should be noted that eqn. 4.74 does not usually give the maximum resonant frequencyf (') and 4 can generally be expressed as a function of f i t
value for the product of the bandwidth and Q factor. It can be obtained from and the substrate dielectric constant 8,. Fig. 4.25 shows the relations between
f@' @ and @/Iwith E, as a parameter for a circular microstrip antenna. The
Q,B, = J(eZ - 1k1 - 1/e2>/:! (4.75) dots in the Figure show the measured results.
On the other hand, if the radiation conductance can be regarded as a domi-
 256 Circular polarisation and bandwidth Circular polarisation and bandwidth 257

nant factor compared with the other conductance components, the unloaded Q In eqn. 4.77, the second-order mode (TM,,,) is chosen as a dominant mode.
denoted by Q, can also be approximated as a function of E, and f i t . Fig. 4.26 From Fig. 4.24, the product of the maximum bandwidth and Q,, when Q = 2.0,
shows the relation between Q, and f i / t with E, as a parameter, for the circular is
microstrip antenna. In this Figure the dots show the measured results.
QoBr = 0.75 (4.79)
From the above equation and eqn. 4.78, it is found that Q, necessary for this
antenna is
Qo = 8.6 (4.80)
It follows from Fig. 4.26 that when a substrate of
E, = 1.21 (4.81)

Fig. 4.25 t/S

Relations between ff2) and @t with E, as a parameter, for circular microstrip
antenna (From Reference 29)

Although only one example of a circular microstrip antenna is given here, the
relationships shown in Figs. 4.25 and 4.26 are typical, and the above approach
is applicable to any shape of antenna including rectangular and triangular
microstrip.

( b ) Design example (291: In this Section, the specific design procedure is Fig. 4.26 Relations between unloaded 0 and $ I t with E, as a parameter for circular micro-
given for the example of a wideband circular microstrip antenna. Let us assume strip antenna (From Reference 29)
the following requirements for the frequency range and input VSWR:
Frequency range: 1530-1670 MHz is used, the f i r value necessary to get the Q, value of eqn. 4.80 is
Input VSWR: less than 2.0 &?/t = 5.84 (4.82)
In this case, the centre or resonant frequency and the relative bandwidth are Accordingly, in the case of a circular microstrip antenna whose fi/t value is
f'2' = 1600MHz (4.77) equal to the above, it can be seen from Fig. 4.25 that
fly2' = 114.86 (4.83)
 258 Circular polarisation and bandwidth
Circular polarisation and bandwidth 259
Eqns. 4.77 and 4.83 show that
mode, and their influence may become a serious problem, when the bandwidth
fl = 71.79mm is expanded without careful consideration.
and the radius of the patch radiator is 40.5 mm. From the above result and eqn. In this Section, the influence of lowering the quality factor is described. For
example, the antenna shown in Fig. 4.27 is a fairly wideband antenna whose
4.82, the thickness of the substrate to be used is calculated as - relative bandwidth is 8.75%, and the influence of the unwanted modes can no
longer be ignored. The mode closest to the wanted one is the TM,,, mode in this
In summary, the circular microstrip antenna to meet the proposed requirements case. The influence of the TM,,, mode may be most prominent when the antenna
has the following specifications: is used as a circularly polarised one with dual feeds. In that case, the TM,,, mode
gives rise to some coupling between two terminals. The measured results for this
Dielectric constant of substrate: 1.21 coupling are shown, together with the calculated ones, in Fig. 4.29, where the
Thickness of substrate: 12.3mm solid line shows the measured results and the broken line shows the calculated
ones. Also the chain-dotted line shows as a reference measured results for an
Patch radius: 40.5 mm antenna having a fairly high quality factor. In this Figure, the coupling is less
Such a microstrip antenna can be made using paper honeycomb materials as a
substrate. Fig. 4.27 shows a circular microstrip antenna manufactured accord-
ing to the above specification. Also, Fig. 4.28 shows the return-loss characteris-
tics for this antenna, where the calculated and measured results are indicated by
the solid and broken lines, respectively. Both results show the wideband perfor-
mance of about 8.75% for VSWR 5 2.0, which agrees well with the require-
ment in eqn. 4.78.
Circu l o r disk

Frequency ( GHz )
Fig. 4.28 Return-loss characteristicsfor circular microstrip antenna shown in Fig. 4.27 (From
Reference 29)

I/ I/ Paper honeycomb core

a = 40.5 mm, t = 12.3mm, and E, = 1.21

V C o a x i a l probes
Epoxy fiberglass skins
Fig. 4.27 Structure of circular microstrip antenna consisting of epoxy Fiberglass skins and
paper honeycomb core (From Reference 29)
( c ) Injhences of unwanted modes and countermeasures against them
[28]: The bandwidth of a microstrip antenna can be increased by employing a
thick substrate with low dielectric constant, as shown in Fig. 4.28. However, it
is expected that some unwanted modes will be generated as well as the wanted
than - 50 dB at the resonant frequency, which is small enough for the antenna
with a high quality factor, and the coupling increases to about - 28 dB for an
antenna with a low quality factor. These results suggest that the axial ratio may
be degraded owing to the influence of this coupling when the latter antenna is
used as a dual-fed circularly polarised antenna, although it is fed by a perfect
90' hybrid. Fig. 4.30 illustrates the axial-ratio characteristics for such an anten-
na, where the solid line shows the calculated results and the dots show the
measured ones. This Figure shows that the best axial ratio is only of the order
of 1.3 dB.
In order to improve the axial ratio it is necessary to investigate the mechanism
of the degradation. When a microstrip antenna has a low quality factor, the
 260 Circular polarisation and bandwidth Circular polarisation and bandwidth 261

Frequency (MHz)
Fig. 4.29 Coupling characteristics between orthogonal ports for circular microstrip antenna
with dual feeds (From Reference 28) Fig. 4.31 Equivalent circuit for dual-fed circular microstrip antenna having low-quality factor
and the corresponding inner-surface current flows on the patch radiator
a Equivalent circuit
b Current flows (e signs denote the feed point)

equivalent circuit can be approximated by Fig. 4.31~.This Figure implies that

the coupling may arise between two orthogonal terminals through the 4th
resonant circuit, in which the current due to the unwanted mode flows.
Therefore, the current flowing in the 4th resonant circuit must somehow be
,suppressed. Fig. 4.32 shows one method of cancelling out such a current by
adding more two terminals to the original two, and feeding from four terminals
90" out of phase with equal amplitude. Fig. 4.33 shows the axial-ratio charac-
teristics with such a method of feeding, where the solid line represents the
calculated results and the broken line the measured ones. As expected, the axial
ratio is improved remarkably compared with Fig. 4.30 for the case of the
antenna fed from only two terminals.
The above is particularly useful when the antenna is used only as a single
element. When it is used as an element comprising an array antenna, it is also
possible to cancel out the cross-polarised component radiated from each ele-
ment at any observation point in free space. For convenience, let us consider a
two-element array radiating elliptically polarised waves from each element, as
shown in Fig. 4.34. If they are both RHCP and the excitation ratio is l : ~ e ' the
~,
Frequency ( M H z ) total electric field radiated from them can be determined from Fig. 4.34 as
Fig. 4.30 Axial-ratio characteristicsfor dual-fed circularly polarised circular microstrip anten-
na
E = {a, + ydA(/?,C O S ~+ j/?,sinS)}f
a = 40.5 mm, t = 12.3 mm, and E, = 1.21
 262 Circular polarisation and bandwidth Circular polarisation and bandwidth 263

1
1
& , = 5 [@I - a21 + yBAe-j6(8l - PI)] (4.89)

with E, denoting the RHCP component. Therefore, in order to make the reverse
polarised components cancel out on the boresight direction, the following
complex excitation condition between two elements must be satisfied:

*
Fig. 4.32 Actual feeding methods for circularly polarised circular microstrip antenna with four
feed point (From Reference 28)
a Right-hand circular polarisation
b Left-hand circular polarisation

-
-:1.c
.-
0
+
"a.
d
Fig. 4.34 Two kinds of elliptically polarised waves radiated from two-element array
.-0 -
a Polarisation ellipse of no. 1 element
2 Col. b Polarisation ellipse of no. 2 element
I
f 30 1550 1600 1650 1700 LHCP: left-hand circular polarisation
RHCP: right-hand circular polarisation
Frequency (MHz)

Fig. 4.33 Axial-ratio characteristics improved by feeding from four terminals as shown in Fig.
4.32a (From Reference 28)
a = 40.5rnrn, t = 1 2 . 3 rnrn, and e, = 1.21

on the boresight direction, where 6 indicates the physical rotation angle of the
polarisation ellipse of the no. 2 element against that of the no. 1 element on an
XY plane as shown in Fig. 4.34. Dividing the above field into the two com-
ponents of co-polarisation and cross-polarisation, it can be expressed as
E = U f - + E,,,df + 9) (4.87) Frequency [MHz)

where Fig. 4.35 Axial-ratio characteristicsimprovedby employingpairedelemenrs (From Reference

28)
a = 40.5 mm, t = 12.3 mrn, and E, = 1.21
 264 Circular polarisation and bandwidth Circular polarisation and bandwidth 265
I

Since an ordinary array antenna consists of the same elements, the above tions, using an impedance matrix known as a matrix Green's function, as
relationship can be reduced to follows:

because it can be assumed that a, = p, and a, = B,. The radiation field for the
RHCP, being a co-polarised component in this case, can be expressed as
Em = ER = - j(a, + a,)d6sin6 (4.92)
parostic element
The above equation implies that the co-polarised component takes the maxim-
um value
IEmI = lERl = la! + (4.93)
when the following condition is satisfied for the rotation angle 6:
6 = 90" (4.94)
On the other hand, the excitation condition for LHCP, instead of RHCP as
above, can be similarly derived and is given by ?-
(a)
y p = - e-jd = &("-6) (4.95)
when a , = p, and a, = B,. The resultant LHCP component is represented as
a function of 6, as in the case of RHCP. So the same maximum value as in
RHCP case can be obtained, when the same condition (eqn. 4.94) is satisfied for
6. It is concluded from the above discussion that the axial ratio may be im-
proved, in case of an array antenna, by arranging for the paired elements to have
a rotation angle 6 = 90". Fig. 4.35 shows the measured results for the axial- air
ratio characteristics of such paired elements. As expected, the resultant axial
ratio is improved remarkably compared with the results of Fig. 4.30, and is of
the same order as that obtained by the previous four terminal-fed cases.

short
4.5.2 Technique using parasitic element [34]
The bandwidth of a microstrip antenna can also be increased by employing a for E- wove for H- wove
parasitic element [25, 30-321. In this Section, the type shown in Fig. 4 . 3 6 ~is (b)
analysed using the Hankel-transformed domain-analysis method [33], and it is Fig. 4.36 Circular microstrip antenna with parasitic element and its spectral-domain equiv-
shown from theory and experiment to achieve an increase in the bandwidth. In alent circuit (From Reference 34)
this Figure, the two substrates are stacked so that they are parallel and the two a Structure of analytical model and co-ordinate system
circular etched disc conductors are concentric. The upper one is used as a b Equivalent circuit for E- and H-waves
parasitic element.
where the sub-vectors [E(a)],,, and [i(cc)],,, consist of two elements, respectively,
(a) Characteristic equation in the Hankel-transformed domain: In general, and are
Green's function in the real domain is a very complicated convolutional integral
or summation form, as shown in eqn. 4.14. However, it is known that Green's
function becomes of simple algebraic form if it is expressed in the Hankel-trans-
formed domain. In this domain, it can be deduced from Reference 33 that the
electric fields on the boundary are related to the corresponding current distribu-
 266 Circular polarisation a n d bandwidth Circular polarisation a n d bandwidth 267
The tilde means the quantities in the Hankel-transformed domain, and each
element is defined from its tangential components, which consists of both for 1 = e (E-wave)
E-wave and H-wave, as
for 1 = h (H-wave)

with F,(r) and F6(r) being the tangential components and J,,,(x) being the
Bessel function of the first kind with (n + 1) or (n - I) th order. Also the for I = e (E-wave)
subscripts 1 and 2, being the order of elements in each sub-matrix of eqn. 4.97,
are referred to the lower and upper conductors. The sub-matrices in eqn. 4.96 for I = h (H-wave)
can be represented by
These elements can be derived from the spectral-domain equivalent circuit
shown in Fig. 4.366. Next, the unknown current distributions are expanded on
each circular conductor in the real domain as follows:

where the matrices [Z(a)], and [Z(a)], denote the impedance matrices for E-wave
and H-wave, respectively, and are written as

where A, and B, are expansion coefficients with i = 1 and 2 being referred to

the lower and upper conductors, whilef,,(r) and &(r) are basis functions. In
this case, the Hankel-transformed current distribution $*)(a) in eqn. 4.976 is
expressed as

They can be easily determined from the corresponding admittance matrices,

which consist of the following elements:
where [A,] and [B,] are unknown vectors with A;, and B,,, and [x(a)],,, and
[&(a)](+, are the transformed basis function vectors with x!:)(a) and &'(a).
Substituting eqn. 4.104 into eqn. 4.976 and then substituting the resulting
- Ysin (pd) sin ( r t ) ] (4.101~) equation and eqn. 4 . 9 7 ~into eqn. 4.96, the following matrix equation can be
-jYY1 obtained:
Y:2(a) = Y:! (a) =
Ycos (pd) sin (B't)+ Y'cos (p't) sin (pd)
[ A ~ l [ h f a ) l ( ++) j[~11[.?41fa)I(+)
(4.1016)
[A2l[X2fa)l1+, + i[~21[342fa)lc+)
) j[~~l&~fa)l(-)
[ A l l [ ~ l f a ) l ( --

- Y' sin (pd) sin (Pt)] (4.101~)

EJ-I (a) [A21[.f,2fa)l(-) - j [ ~ ~ l t f 4 ~ f a ) l ~ - )

with I = e or h and
p = ,/- Taking inner products of eqn. 4.105 with all the transformed basis functions
fi;)(a) and]$:'(a) according to Galerkin's method, the left-hand sides of all the
8. = J W ~ ~ , E , E , - a2 resulting equations vanish owing to the boundary conditions. Thus it is possible
 268 Circular polarisation and bandwidth Circular polarisation and bandwidth 269
to choose the following combinations in order to avoid divergence of integrals to exist. Therefore, the characteristic equation for this problem can be written
appearing in the characteristic equation: as
Q(w) = det[P] = 0 (4.1 12)
where w is the complex resonant angular frequency whose real and imaginary
parts correspond to the resonant angular frequencies and the damping factors,
respectively.

where a bracket ( . ) in the above equations is employed for the following

infinite integral:

Finally, eqn. 4.106 can be rewritten in a matrix form and is given by

p111

0.0
I
0.80 0.85 0.90 0.95 1 .OO 1.05
where the matrix [PI is defined as follows: real port Dr

Fig. 4.37 Contour of complex resonant frequencies of circular microstrip antenna with par-
asitic element (From Reference 34)
t = 1.6 mm and E, = 2.55.b = w/wo = b, + jh,

and the elements of the sub-matrices [P!]
j = I or 2 are given by
with
( b ) Electrical characteristics: In this case, two basis functions for each cur-
rent component provide satisfactory accuracy, so that the characteristic equa-
through [ P a ] with i = 1 or 2 and tion results in a form of determinant of size 8 x 8. Fig. 4.37 shows numerical
results for the complex resonant frequencies solved from eqn. 4.1 12, where the
thickness and dielectric constant for the substrate used are t = 1-6mm and
E, = 2.55, respectively. The Figure shows the contour map for the real and
imaginary parts of normarised complex resonant frequencies with a,/a, and d/al
as parameters. From this Figure, it is found that two dominant resonant modes,
which exhibit a double-tuned characteristic, exist in this antenna. So the input
VSWR characteristics can be calculated by considering a double-tuned perfor-
mance. In this case, two resonant resistances R, and R, can be determined
uniquely as follows:

From eqn. 4.108, the determinant of [PI must vanish for a non-trivial solution
 270 Circular polarisation and bandwidth Circular polarisation and bandwidth 277
because in some special applications [36]. In this Section, we briefly describe design
procedure of such a paired element.
Fig. 4 . 3 9 ~shows the fundamental arrangement of a microstrip paired-element
unit. The patch elements are rotated orthogonally on the coplanar plane and are
fed in uniform amplitude but 90' out of phase through the sequentially rotated
feeding points F, and F,.

1441c in Fig. 4.37

w0 = - (-n12) phase
a, shifter

where w,, and w,, are resonant angular frequencies for two dominant modes and
cis the velocity of light in vacuum. Fig. 4.38 shows the calculated and measured
input VSWR characteristics for the antenna with d = IOmm, a , = 20+3mm,
a, = 21.0 mm, t = 1.6 mm, and 6, = 2.55. In this Figure, the results for the
antenna without a parasitic element are also shown as a reference, and it is seen
that the effect of the parasitic element is considerable. ( a ) Paired element

EY

>
1.1 -
I

2.4 2.5 2.6 2.7 2.8

frequency ( G H r )

Fig. 4.38 Calculated and measured VSW R characteristics for circular microstrip antenna with ( b ) Principle of wideband
parasitic element (From Reference 34)
d = 1 0 m m , a, = 2 0 . 8 m m . a , = 2 1 . 0 m m . t = 1.6rnm. a n d & , = 2.55 Fig. 4.39 Paired element and its polarisation pattern

A sub-array composed of such paired elements demonstrates the broadband

4.5.3 Technique using paired element
Circularly polarised microstrip antennas including single-fed patches are widely
used as effective radiators in many communication systems [6, 101.
However, the most serious problem with such antennas is the narrowness of
the ellipticity and impedance bandwidth compared with ordinary microwave
antennas. Several techniques for the expansion of bandwidth have been repor-
ted in the literature [29-341. However, most of these broadband techniques,
including double-stacked CP patches, are applicable to isolated CP patch anten-
nas. Hence, other wideband techniques using sequential arrangements of anten-
na elements have been developed in recent years [26,35]. The simplest device for
such a sequential array is a paired element, and it is used as an effective radiator
nature in spite of using narrow-band patch elements [26]. In order to evaluate
the performance, the polarisation pattern of an antenna is briefly described.
In general, individual elements of a pair show elliptical patterns of polarisa-
tion, as shown in Fig. 4.396. The polarisation ellipses marked A and B corres-
pond to those of each element of a pair, while the Ex- and E,-axes correspond
to the horizontal and vertical components of the radiated electric field.
The polarisation ellipses for an individual CP patch vary rapidly with change
of frequency. However, if the CP patch of a pair is arranged orthogonally and
is fed uniformly in amplitude but 90' out of phase, the resultant polarisation
pattern due to the pair can be shown as a trace of perfectly circular polarisation
over a wide frequency range, as shown in the figure.
 272 Circular polarisation and bandwidth
In order to verify the performance, a 2 x 2 element sub-array unit having
two pairs was constructed and tested at X-band. With regards the feeding
system, the input impedance for each element of the pairs was matched to that
of the main feeder M, by means of A,/4 impedance transformers, T,, T,,T, and
T,, where 1, is the wavelength in stripline. Fig. 4.406 shows the typical measured
ellipticity bandwidth for the sub-array unit. The ellipticity bandwidth for an
isolated CP-patch antenna is also shown for comparison in the Figure. It is
shown that the sub-array unit using the paired element contributes enormously
to the improvement of the ellipticity bandwidth compared with a single CP-
patch element. The 3 dB ellipticity bandwidth obtained by this sub-array unit is
about five times the value obtained by an ordinary CP-patch element, as shown
in the figure.
A more detailed description of the sequential array is given in Chapter 13.
Fig. 4.40 Sub-array unit and its ellipticity bandwidth
4.6 References
1 MAILLOUX, R. J., McILVENNA, J. F., and KERNWEIS, N. P.: 'Microstrip array technol-
ogy', IEEE Trans., 1981, AP-29, pp. 25-37
Circular polarisation and bandwidth 273
2 WEINSCHEL, H. D.: 'A cylindrical array of circularly polarized microstrip antennas'. In!.
Syn7p. Dig. Antennas Propagat. Soc., June 1975, pp. 177-180
3 SHEN, L. C.: 'The elliptical microstrip antenna with circular polarization', IEEE Trans., 1981,
AP-29, pp. 90-94
4 RICHARDS, W. F., LO, Y. T., and HARRISON, D. D.: 'An improved theory for microstrip
antennas and applications', IEEE Trans., 1981, AP-29, pp. 38-46
5 KERR. J. L.: 'Microstrip polarization techniques'. Proc. Antenna Applications Symp., Aller-
ton Park, IL, Sept. 1978
6 JAMES, J. R., HALL, P. S., and WOOD, C.: 'Microstrip antenna'. (Peter Peregrinus, 1981)
chap. 7
7 HANEISHI, M., NAMBARA, T., and YOSHIDA, S.: 'Study on elliptical properties of
singly-fed circularly polarised microstrip antennas', Electron Lett., 1982, 18, pp. 191-193
8 ITOH, K.: 'Circularly polarised printed array composed of strip dipole and slot', Microwave
J., April 1987, pp. 143-153
9 JAMES, J. R., HALL, P. S., WOOD, C., and HENDERSON, A,: 'Some recent developments
in microstrip antenna design', IEEE Trans., 1981, AP-29, pp. 124-128
10 CARVER, K. R., and MINK, J. R.: 'Microstrip antenna technology*,IEEE Trans., Antennas
& Piopagi., :98:, AP-29, pp. 2-24
I1 HANEISHI, M., and YOSHIDA, S.: 'A design method of circularly polarised rectangular
microstrip antenna by one-point feed', Electron & Commun in Japan, 1981, 54, pp. 46-54
12 OKOSHI, T., and MIYOSHI, T.: 'Planar circuits', (Ohm Publishing (in Japanese), 1973)
13 SUZUKI, Y., MIYANO, N., and CHIBA, Y.: 'Circularly polarised radiation from singly-fed
equilateral-triangular microstrip antenna', IEE Proc. 1987, 134, pp. 194-198
14 SCHAUBERT, D. H., FARRAR, F. G., SINDORIS, A,, and HAYES, S. T.: 'Microstrip
antennas with frequency agility and polarization diversity', IEEE Trans., 1981, AP-29, pp.
118-123
15 OKOSHI, T., and MIYOSHI, T.: 'The planar circuits - An approach to microwave integrated
circuitry', IEEE Trans., 1972, MlT-20, pp. 245-252
16 CARVER, K. R.: 'A modal expansion theory for the microstrip antenna', Int. Symp. Dig.
Antennas Propagal. Soc., June 1979, pp. 101-104
17 SUZUKI, Y., and CHIBA, T.: 'Computer analysis method for arbitrarily shaped microstrip
antenna with multi-terminals', IEEE Trans., 1984, AP-32, pp. 585-590
18 SUZUKI, Y., and CHIBA, T.: 'Improved theory for a singly-fed circularly polarized micro-
strip antenna', Trans. IECE Japan, 1985, E68, pp. 76-82
19 For example, NEWMAN, E. H., and TULYATHAN, P.: 'Analysis of microstrip antennas
using moment methods', IEEE Trans., 1981, AP-29, pp. 47-53
20 MORSE, P. M., and FESHBACH, H.: 'Methods of theoretical physics: Pt. 11. (McGraw-Hill,
NY, 1953). pp. 1112-1 119
21 CARVER, K. R.: 'Input impedance to probe fed microstrip antennas', In!. Symp. Dig.
Antennas Propagat. Soc., June 1980, pp. 617-620
22 HANEISHI, M., YOSHIDA, S., and TABETA, M.: 'A design of back-feed type circularly
polarised microstrip antenna having symmetrical perturbation segment', Electron & Commun.
in Japan, 1981, 2, pp. 52-60
23 POZAR, D. M.: 'Input impedance and mutual coupling of rectangular microstrip antennas',
IEEE Trans., 1982, AP-30, pp. 1191-1 197
24 DERNERYD, A. G., and LIND, A. G.: 'Extended analysis of rectangular microstrip resona-
tor antennas', IEEE Trans., 1979, AP-27, pp. 846-849
25 ITAMI, H., and HORI, T.: 'Broad band circular polarized microstrip antenna'. In!. Conv. Rec.
lECE (in Japanese), 1982, p. 642
26 HANEISHI, M., YOSHIDA, S., and GOTO, N.: 'A broadland microstrip array composed of
single-feed type circularly polarized microstrip antennas', in In!. Symp. Dig. Antennas
Propagal. Soc., May 1982, pp. 160-163
27 MURPHY, L.: 'SEASAT and SIR-A microstrip antennas', Proc. Workshop on Printed
Circuit Antenna Technology, Oct. 1979, paper 18
274 Circular polarisation a n d bandwidth
28 CHIBA, T., SUZUKI, Y., MIYANO, N., MIURA, S., and OHMORI, S.: 'A phased array
antenna using microstrip patch antennas', 12th European Microwave Conference, Sept. 1982,
pp. 472-477
29 SUZUKI, Y.,and CHIBA, T.: 'Designing method of microstrip antenna considering the
bandwidth', Trans. IECE Japan, 1984, E67, pp. 488-493
30 WOOD, C.: 'Improved bandwidth of microstrip antennas using parasitic elements', IEE Proc., Chapter 5
1980, 127, pp. 23 1-234
31 LONG, S. A., and WALTON, M. D.: 'A dual-frequency stacked circular disc antenna', Int.
Symp. Dig. Antennas Propagat. Soc., June 1978, pp. 260-263 Microstrip dipoles
32 SANFORD, G. G.: 'Multiple resonance radio frequency microstrip antenna structure', US
Patent 4070676, Jan. 1978
33 ARAKI, K., and ITOH, T.: 'Hankel transform domain analysis of open circular microstrip
P.B. Katehi, D.R. Jackson and N.G. Alexopoulos
radiating structures', IEEE Trans., 1981 AP-29, pp. 84-89
34 ARAKI, K., UEDA, H., and TAKAHASHI, M.: 'Numerical analysis of circular disk micro-
strip antennas with parasitic elements', IEEE Trans., 1986, AP-34, pp. 1390-1394
35 TESHIROGI, T., TANAKA, M., and CHUJO, W.: 'Wideband circularly polarised array
antennas with sequential rotations and phase shift of elements'. Proc. Int. Symp. on Antennas 5.1 Introduction
& Propagt., Japan, Vol. 1, Aug. 1985, pp. 117-120
36 HANEISHI, M., HAKURA, Y., SAITO, S . , and HASEGAWA, T.: 'A low-profile antenna Microstrip dipoles have been studied extensively during the last 20 years. Tlley
for DBS reception'. Int. Symp. Dig. Antennas Propagat. Soc., June 1987, pp. 914-917 are planar elements which consist of a pair of collinear thin-strip conductcrs
37 SUZUKI, Y.: 'Analysis of microstrip antennas based on the planar circuit theory and its printed on the surface of a dielectric slab (Fig. 5.1). They resemble the free-space
applictions (in Japanese)'. Doctoral dissertation, Tokyo Inst. Technol., Nov. 1984
cylindrical dipoles in the sense that radiation results from a harmonically
varying dipole moment. Microstrip dipoles are attractive elements owing to
their desirable properties such as simplicity, small size and linear polarisation.

1 I I 1 Coaxial Excitation

Fig. 5.1 Excitation mechanisms for a microstrip dipole

They are well suited for higher frequencies in particular, where the substrate
may be electrically thick. In this case the bandwidth of the dipoles may be quite
significant. For thicker substrates it is also possible to alter the radiation
properties by the use of a superstrate layer, making dipoles a possible candidate
in a substrate-superstrate geometry.
When designing microstrip dipoles, the choice of feed mechanism is very
276 Microstrip dipoles Microstrip dipoles 277

important and should be made taking into consideration the following two complicated functions of 1, z, z', and are given in Reference 2. The time
factors: theoretical modelling and practical implementation. Fig. 5.1 shows the dependence is ef'"", and is suppressed here. The functionsf and g are of the form
most commonly used mechanisms: the coaxial feed, the twin-line feed and the
coupled-line feed (EMC dipole). In the twin-line feed a voltage is applied
directly to the arms of the dipole. In the coaxial feed the two dipole arms are
shorted together, with the dipole becoming essentially a narrow patch antenna
with a probe feed. The EMC dipole excitation is realised through electro-
magnetic coupling to the feed line, with no direct contact. Because of its
simplicity, the EMC feed represents the most desirable way to feed a dipole from where a and b are analytic functions of 1except for the branch-type singularity
a microstrip line. due to the wavenumber
Even if practical excitation mechanisms are employed in the design of micro- 112
strip dipoles, more ideal ones may be considered for their analysis. The reason k = (G -2)
lies in the fact that most of their radiation properties are independent of the
which appears in the expressions [2]. The functions D,(1) and Dm@)have zeros
excitation (i.e. bandwidth, efficiency, radiation pattern etc.). Throughout this
on the real axis at A,, producing poles in the integrand. The zeros of D,(1) are
Chapter the iiiiciosirip dip& is siiidied extensiveiy as a single radiator as well
as an array element. Furthermore, infinitesimally small, centre-fed and EMC
dipoles are presented separately, and the dependence of their properties on the
electric characteristics of the dielectric layers is discussed. The study of micro-
strip dipoles is concluded by presenting a design technique for a n array of EMC
dipoles which accounts for the mutual interactions between dipole elements.

5.2 Infinitesimal dipole

The simplest dipole structure which can be studied is the infinitesimal dipole. An
analysis of the infinitesimal dipole is important because all the radiation charac-
teristics of full-size dipoles may be obtained simply from this solution. Only for
near-field (impedance) calculations is it essential to analyse the full-size dipole
with a moment-method technique.

5.2.1 Analysis
A horizontal electric dipole (HED) is shown in Fig. 5.2. In general, the dipole
may be embedded within an arbitrary number of layers, although two layers are
sufficient to cover most cases of practical interest, including microstrip dipoles
with a protective top (superstrate) layer, or EMC dipoles with a transmission
line at the interface (z = b).
In the classical Sommerfeld solution, components of the magnetic vector
potential at x, y, z due to a source at x', y', z' are written as
Fig. 5.2 Substrate-superstrate geometry with horizontal electric dipole (HED) embedded
the TE-mode surface-wave propagation constants, while those of D,(1) corres-
pond to the TM-mode surface waves. These poles are in the region
k, < A, < k,, where k, = max (k,, k,). The path of integration goes
around the poles, as shown by contour C in Fig. 5.3.
By using symmetry properties [I] the integrations may be extended to (- co,
+ co) and the path deformed to an integral around the branch cut (contour C b )
plus integrals around the poles in the right-half plane. The potentials may then
be written as

with Q, q5 describing cylindrical coordinates [I]. The functions f and g are

 Microstrip dipoles 279

with P,a, and P,, the radiated and total surface-wave powers, respectively.

5.2.2 Substrate effects

The effect of substrate thickness on the efficiency of an infinitesimal dipole on
top of a single layer is shown plotted against the electrical substrate thickness
b / A d (where Ad = A,/JE,) in Fig. 5.4 for various substrates. Note that the
efficiencydecreases for increasing substrate thickness, and is lower for higher E , .

Fig. 5.3 Contours of integration

The residue contributions at the poles give the surface-wave fields. A steepest-
descent method may be used to find the far-zone radiation field, although a
reciprocity method is simpler [2]. The Poynting vector from the far-zone radia-
tion field may be integrated over a hemisphere to find the radiated power. This Fig. 5.4 Efficiency of HED versus substrate thickness for dipole on top of single substrate
layer
reduces to a one-dimensional numerical integration [2].
The surface-wave Poynting vector may be integrated over a large cylinder to
find a closed-form expression for the power in a surface wave. The surface waves The efficiency approaches 1.0 for thin substrates, but the radiated power of the
are orthogonal with each other and with the radiation field in the lossless case dipole then becomes very low, as seen from Fig. 5.5. This points toward one of
[3], so the total power is simply the sum of all the powers. The radiation the practical limitations of using resonant-length dipoles on thin substrates,
efficiency is defined in the lossless case as namely low input resistance. A patch antenna does not have this disadvantage
since the resonant resistance is fairly independent of substrate thickness [4].
 280 Microstrip dipoles Microstrip dipoles 281

,& Tad Normalized by 10 ' 1 1%

Another limitation of resonant-length dipoles on thin substrates is a very
small bandwidth -much lower than that of a patch (this is discussed in Section
5.4.2). Because of these limitations, dipoles find best application for thicker
substrates where bandwidth and resistance are no longer serious limitations,
and where the patch antenna becomes non-resonant [5].

5.2.3 Superstrate effects

It is interesting to note that a superstrate (cover) layer on top of a microstrip
dipole may significantly influence the radiation properties. For example, if the
substrate is thin enough, a superstrate layer may be used to eliminate surface-
wave excitation, resulting in an efficiency of 100% [2]. An example of this is
shown in Fig. 5.6 for a GaAs superstrate over a Teflon substrate. In a different
application, a substrate-superstrate geometry may be used to produce 'radia-
tion into the horizon', a phenomenon in which the far-zone radiation field

-E-PLANE PATTERN

blXd

Fig. Radiated power of HED (Watts) versus substrate thickness for unit-strength dipole
on top of single substrate layer

Fig. 5.7 Radiation pattern of HED demonstrating the radiation into the horizon effect (Re-
produced from Reference 6,@ 1985 IEEE)
The dipole is embedded within a single substrate layer here

Fig. 5.6 Efficiency of HED versus superstrate thickness, showing 100% efficiency at tlL, =
0.233 (Reproduced from Reference, 2, @ 1984 IEEE)
The dipole is at the interface (z, = b )
remains non-zero down to the layer surface. One result is the possibility of
producing very nearly omnidirectional patterns [6].An example of this is shown
in Fig. 5.7 for a dipole embedded within a single substrate layer of thickness b
(or, equivalently, the superstrate material is the same as the substrate).
A third application of a superstrate layer is in the production of high-gain
patterns about any desired angle 0, in space. By using one or more superstrate
layers of the proper thickness, narrow-beam patterns may be produced as the
superstrate E , ~becomes large [7].An example of this for 0, = 45" is shown in
 282 Microstrip dipoles Microstrip dipoles 283
Fig. 5.8 using a superstrate with E,, = 100. This narrow-beam effect is produced domain piecewise-sinusoidal basis functions of the form
by weakly attenuated leaky waves which exist on the structure [a].
All the above effects pertain to any type of microstrip element in a substrate- Bky) = J, = ~ 0 1 5(x)
) lyl < w / 2 (5.8)
superstrate geometry. However, except for the first effect (increased efficiency), 1x1 < d
these methods all require electrically thick layers, making dipoles an attractive
where
candidate.

E-PIANE PATTERN

Fig. 5.8 Radiation pattern of HED demonstrating the high-gain effect (Reproduced from
Reference 7, @ 1985 IEEE)
The dipole is embedded within the substrate with a superstrate layer on top

5.3 Moment-method techniques for planar strip geometries

The analysis of geometries comprising dipoles and transmission lines involves
strips which are usually narrow compared to a wavelength ( w 4 4). The
geometry of a strip is shown in Fig. 5.9. Because the strip is narrow, current may
be assumed in the f direction only. The narrow-strip assumption also allows for
certain techniques to improve the computational efficiency of the analysis. In
this Section methods for analysing strip structures are discussed.
Fig. 5.9 Planar strip in a layered geometry, divided into subsections.
Also shown is the basis function variation in x
Three useful choices for the transverse variation are the pulse function, the
modified Maxwell function [9] and the Maxwell function:

5.3.1 Basis Functions

For the moment-method solution of strip geometries, including dipoles and
transmission lines, the current on the strips may be represented using sub-
 284 Microstrip dipoles
The normalisation constants here are chosen to give a unit current at x = 0.
The Maxwell function more closely represents the true current on a narrow
strip. The modified Maxwell function is similar to the Maxwell function except
near the edges. Because it lacks the singular behaviour at the edges, the Fourier
transform of the modified Maxwell function decays faster for large values of the
transform variable, which makes it computationally more efficient for spectral-
domain analysis techniques.
5.3.2 Reaction between basis functions
The fundamental step in a Galerkin moment-method solution of strip problems
is the computation of reaction between two basis functions:
with
with S, and S, the basis-function surfaces. G,, is the f f component of the dyadic
Green's function for E,(x, y, 2,) due to a source at x', y', z,. This is given by
G,, = k Z n x
a2n + 2.
+ -+ a2n
ax axa~
If more than one strip is involved in the problem under consideration, it is
convenient to choose the x-axis offset between strips as an integral multiple of
the subsection length d, so that the impedance submatrices will all be Toeplitz.
Depending on the particular problem, d is typically in the range
0.01 A, < d < 0.05 A,. In this case 5(x) is close to a piecewise linear function,
and the value of k, in eqn. 5.8 is not critical. The choice of transverse distribution
~ ( y has
) some influence in the reaction values obtained, especially for the
self-term (no offset between basis functions). Experience has shown, however,
that the current amplitudes, obtained from the moment-method matrix solu-
tion, are fairly independent of the choice of qb).
For the solution of the strip problems, the starting point is the calculation of
arrays Z,(m) where the mth element is reaction between basis functions
separated by (m-l)d between centres in the %direction. The index i refers to the
particular submatrix in the Galerkin impedance matrix, corresponding to basis
functions on different strips. The y-directed offset may be different for each
submatrix if the strips have transverse offset. z, and z, may be different for each
submatrix as well.
Microstrip dipoles 285
5.3.3 Plane-wave spectrum method
Two distinct methods have been used in the literature for the calculation of
reaction: the Fourier transform (plane-wave spectrum) method and the real-
space integration technique. In the spectral method, eqns. 5.1 and 5.2 are first
transformed into rectangular (Fourier) transform form, and substituted into
eqns. 5.12 - 5.14. The resulting integrations in x, y, x', y' result in Fourier
transforms of the current [10,1I]. The result is
- z: (I,, A,) cos &A,) cos (44) d1d4 (5.15)
where
and
h(1, z,, z,) = - ag
A1 -
az
The separation between basis-function centres is denoted as Ax, Ay here. For the
basis function choices (eqns. 5.9 - 5.1 1) the transforms may be evaluated in
closed form [lo]. The integral on (0, co) in eqn. 5.15 is along the Sommerfeld
contour C in Fig. 5.3. A pole-extraction technique may be used to account for
the poles on the real axis [I 11. A simpler way is just to deform the contour to
go around the poles as shown by contour C, in Fig. 5.3 [12].
An advantage of the spectral approach is that self-term problems are avoided.
However, the convergence of the Sommerfeld integral in eqn. 5.15 becomes
worse as the separations Ax, Ay become large compared to the respective basis
function dimensions, for the case z, = z,. This is because the functions f and
h tend to constants as 1 + co in this case, resulting in a rapidly oscillating,
slowly converging integral. For z, f z,, as is possible for the reaction between
currents on different strips, the termsf and h decay exponentially in 1, and there
is relatively little trouble for most values of Ax, Ay. T o speed up the computation
for the case z, = z, = z, several numerical techniques may be employed. The
first is the use of a Filon integration method to account for the oscillatory cosine
terms [13]. A second technique is to subtract from the integrand a limiting-beha-
.viour term with constantsflco, z , z), h(co, z, z), so that the integral converges
much faster. The integral of the extracted term may be evaluated by identifying
 286 Microstrip dipoles
it as the reaction between currents in a grounded homogeneous half-space of
effective wave number k, [14]. This reaction may be evaluated using a free-space
Green's function with a real-space integration. In this case the pulse choice of
q b ) (eqn. 5.9) is most convenient, since the resulting real-space reaction integral
reduces to a one-dimensional integral [14].
Another type of extraction may be employed for the case z = z, = z, when
z is sufficiently far from a layer boundary, as for a strip embedded within a layer.
In this case, a term corresponding to the incident field in a grounded homoge-
neous space of wave number k, is extracted [15], where k, is the wave number of
the layer. The resulting integrand then decays exponentially. The reaction in the
grounded homogeneous half-space is computed as before. This extraction fails
for a strip at the interface of different media, and is therefore of more limited
use than the first type of extraction.
Another method for improving the convergence of eqn. 5.15 is to asymptotic-
ally approximate the transform Jxfor large 2 in a sufficiently simple form so that
the tail integral over ( A , co) may be performed analytically, for some large
number A. This is the most straightforward technique, but the tail integral must
be reformulated for each specific choice of qb).
Finally, as an alternative to trying to accelerate the convergence of eqn. 5.15
for large separation between basis functions, the reaction may be computed by
a different technique, as discussed in Sections 5.3.4 and 5.3.5.
5.3.4 Real-space integration method
The reaction Z,, may also be computed by integrating the electric field directly
in the spatial variables. To avoid a prohibitive amount of calculation, a 6-fun-
ction testing procedure is used at the strip centre instead of a Galerkin method,
so that impedance elements are defined as
A technique due to Katehi and Alexopoulos [16,17] computes this impedance
term by directly applying the electric-field operator (eqn. 5.14) to the Sommer-
feld form of potentials (eqns. 5.1 and 5.2). An integration by parts is used to shift
the derivatives to the current function J,, (x', y'), and various algebraic mani-
pulations are employed, including an analytical tail integral evaluation. The
resulting expression involves a single Sommerfeld-type integral of a rapidly
converging series. This formulation does not suffer from convergence difficulties
to the same degree that eqn. 5.15 does.
Although a comprehensive comparison of computational efficiency has not
been performed, it is felt that the real-space method is somewhat comparable to
the spectral method when one of the accelerating techniques mentioned
previously is used. Both methods have been used to generate the results of this
Chaptei.
Microstrip dipoles 287
5.3.5 Point-dipole approximation
The reaction between widely separated basis functions may often need to be
computed, especially for mutual-impedance problems. In this case, the most
efficient technique is to approximate the currents as point dipoles located at the
basis-function centres. Computing reaction is then equivalent to finding the Ex
field of an I-directed point source, which may be obtained efficiently. One way
to perform this calculation is by directly applying the electric-field operator
(eqn. 5.14) to eqns. 5.1 and 5.2. It is well known that the resulting Sommerfeld
integrals are nonconvergent when z = z', however, and therefore cannot be
evaluated directly. One technique for overcoming this difficulty is to extract
terms from the integrand to give convergence [18]. Another way is to extend the
integration contour to ( - co, + co) and deform around the branch cut, as
described in Section 5.2.1. The numerical integration around the branch cut
converges very rapidly for large radial separation @ between dipoles, owing to
the exponmtia! decay of the Hankel functions in eqns. 5.5 and 5.6 along the
imaginary axis. For separations larger than 0.25& this is usually the most
efficient of the two methods.
5.3.6 Moment-method equations
Consider an arbitrary set of x-directed strips having a 1V &gap voltage
source at some point on one of the strips. Let the basis functions be numbered
1, ...N with the source at the centre of basis-function numbers, at x = x,. The
current representation is then
with BJx, y) = B(x - x,, y - y,). Because q k ) in eqns. 5.9 - 5.11 is nor-
malised, I,, represents the current in amperes at x = x,. Enforcing Ex =
6 (x - x,) on the strips by applying Galerkin's method using eqn. 5.17 results
in the set of equations
which is then solved to find the currents on the strips.
5.4 Centre-fed dipoles
5.4.1 Single dipole
The analysis described in Section 5.3.6 may be applied to find the current
distribution for a centre-fed dipole, shown in Fig. 5.10. A variational expression
for the input impedance is [I91
 288 Microstrip dipoles Microstrip dipoles 289

1 CENTER - FED DIPOLES

Z, = - - j3Js, J,(i) G, (i,7 )Jr(?') dsds' .
LO
4; w = .05

Using eqn. 5.17 this reduces to

I 4--

z, = -xz
1; ,"" mn
I I
"
PATCH RESONATOR /
\
\

WITH Er= 2.45

which, in view of eqn. 5.18, reduces further to simply
1
Z," = -.
I",

Fig. 5.10 Centre-fed strip dipole

This simple formula for input impedance is thus accurate to second order owing
to the relation between Galerkin's method and the variational method [20].
From input impedance, the resonant length and bandwidth may be determined.
Approximate formulas for resonant dipoles may also be used [lo]. A dipole
on a substrate layer has a resonant length

provided h 9 w and w 4 1,. At resonance the resistance is

Fig. 5.11 Bandwidth (%) of centre-fed dipole versus substrate thickness for two different
substrates.
where P, is the total power (watts) produced by a unit-strength point dipole on The dashed lines indicate that no frequency is found for which X , = 0. In this case
the substrate. Owing to the behavior of P, (see Fig. 5.5) R, is very small for thin f, was chosen to minimise )Xj,,I, with X , (f,) then subtracted from all values. A
substrates. patch resonator is shown for comparison for h l l , < 0.1
290 Microstrip dipoles Microstrip dipoles 29 1
The substrate thickness has a dominant effect on bandwidth. The bandwidth
for different substrates with a dipole width of 0.05 1, is shown in Fig. 5.11. Here
the bandwidth is defined as

with f, the resonant frequency ( X , = 0) and f,, f, the frequencies at which

SWR = 2.0 on a feed line having a match at fa.The maximum bandwidth
occurs for h/& z 0.30, and increases for larger E , . For substrates thinner than
this, the bandwidth is relatively independent of E, for a given physical thickness
h/&. For comparison, the bandwidth of a microstrip patch resonator is shown
for the E, = 2.45 case. The patch has a much higher bandwidth for thin
substrates, a conclusion reached previously by Pozar [5].
The effect of dipole width is seen in Figs. 5.12a, and b for a substrate with
E, = 245. The resonant input resistance is insensitive to width, as is the re-
sonant length. The slope dXi,,/dL a t f , is lower for wider dipoles, indicating a
Fig. 5.12a Real part of the input impedance of a centre-fed strip dipole with e, = 2.45 and
greater bandwidth.
h = 0.21, However, the slope is not extremely sensitive to width. Only when the sub-
strate becomes thin does the width have a dramatic effect on bandwidth, owing
to the cavity-resonator effect.

5.4.2 Mutual impedance

Mutual impedance between centre-fed dipoles may be calculated in different
ways. In the moment method, dipole no. 1 is excited with no. 2 short-circuited.
After solving eqn. 5.18 for the currents, the formula [lo]
Xln
(Ohms)
800 . - - 0.0002
w 1,

with

may be employed. Alternatively, the classical reciprocity formula [21]

-1
Z,, = - Jx2 ds
Il(O)I*(0) J-s2
may be used, with piecewise-sinusoidal currents assumed on the dipoles. The
reaction may be evaluated using eqn. 5.15. For narrow strips, the dipoles may
be assumed filamentary and the integral (eqn. 5.26) evaluated directly using the
real-space technique, or by using the point-dipole approximation to find Ex
(Section 5.3.5). Results for filamentary dipoles obtained in this way are shown
in Figs. 5.13a,b for broadside and endfire dipoles on a substrate with E, = 10.
The slow decay of Z12in the endfire case is due to the dominant TMo mode,
Fig. 5.1 2b Imaginary part of the input impedance of a centre-fed strip dipole with e, = 2.45
which is strongest at 4 = 0 [22].
and h = 0.2 1,
292 Microstrip dipoles Microstrip dipoles

2.0
I
(Ohms)
1.5
1.5

1.0
1.0 (Ohms)

.5
.5

0
0

-.5 -. 5

I I I I I I I I I I
0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6 1.8
s IX, -1.0
Y

Fig. 5.1 3a Mutual impedance versus separation for resonant-length filamentary dipoles in
broadside configuration with E, = 10.0 (Reproduced from Reference 10, @ 1986
IEEE)
Fig. 5.13b Mutual impedance versus separation for resonant-length filamentay dipoles in
endfire configuration with 8, = 10.0 (Reproduced from Reference 10, @ 1986
IEEE)
 294 Microstrip dipoles Microstrip dipoles 295
5.5 EMC Dipoles

5.5.1 Methods of analysis

Fig. 5.14 shows several ways in which one or more dipoles may be electromag-
netically coupled to a transmission line. In each case the analysis involves a
moment-method solution together with simple transmission-line theory [23].
The transmission line is assumed close to the ground plane, so that a TEM-like
field propagates on the line. A 6-gap source is taken near the end of the line
farthest from the coupled end. The exact location is not critical. The source sets
up a standing-wave current on the line which is essentially a sinusoidal current,
except within a small region near the coupled end where the current is perturbed
by the dipoles. Away from this coupling region, the line may be regarded as
terminated by a equivalent self-impedance Z, = R, + j X, at some arbitrary
value of x, where

with

r(x) =
SWR - 1 e+,28(" - X,,")
Top v i e w SWR + 1
In eqn. 5.28 x,, is the position of a minimum and is the propagation constant
on the line, which may be determined from the distance between current minima
on an isolated line or by separate analysis [9,10].
In this formulation only the reaction between piecewise-sinusoidal-basis fun-
ctions is required. An alternative formulation using traveling-wave-basis fun-
ctions on the line may also be developed [24]. An advantage of this latter
formulation is the use of fewer unknowns for the line current, although it
requires different types of basis functions. For the results of this Chapter, only
piecewise-sinusoidal basis functions were used.

5.5.2 Single dipole

A single EMC dipole can be either overcoupled, matched, or undercoupled
according to the amount of coupling to the feed line (Fig. 5.15). Of all the
possible parameters which affect the coupling, the most critical is the distance
t between the line and dipole. If t is too large, then no choice of dipole length
or offset will yield an input match, and the dipole is said to be undercoupled. If
the line is sufficiently close to the dipole, an input match may be achieved by
varying the dipole length and either the longitudinal (x-axis) or transverse
(y-axis) offset, or both. The dipole is then said to be overcoupled. In this case
Top v i e w the locus of points for the centre of the matched dipole is somewhat elliptical
in shape, roughly centered about the end of the line [25]. For a given substrate,
Fig. 5.14 Various configurations of dipoles electromagnetically coupled to a microstrip line
this implies a maximum distance t,,, for which an input match may be achieved,
296 Microstrip dipoles Microstrip dipoles 297
where the locus collapses to a single point [25]. In this case the dipole may be
said to be critically coupled. Figs. 5.16 and 5.17 show how an input match may
be achieved for the overcoupled case by varying either the longitudinal
~ - - or
transverse offset, respectively, for E, = e,, = E,, = 2.35.
- matched
- undercoupled
overcoupled

Fig. 5.1 5 Current amplitude on the strip dipole ( I , ) andmicrostrip line (I,) (Reproducedfrom
Reference 2 3 @ 1984 IEEE)
1; is the incident current amplitude. The three cases correspond to h = 0.079&
(matched), h = 0.084A0(undercoupled), and h = 0,070d,, (overcoupled)

Fig. 5.17 Z,/Z, as a function of dipole length L, and transverse offset (Reproduced from
Reference 23, @ 1984 IEEE
The longitudinal offset is 50% of the dipole length. The impedance reference plane
is at the position of a current maximum on the line
Fig. 5.16 Z,/Z, as a function of dipole length L, and longitudinal offset (Reproduced from
Reference 2 3 @ 1984 IEEE)
Offset is measured by percent overlap as KO, = (AXIL,) x 100. The impedance
refaranre nlana is at the nnsitinn nf a cllrrant maxirn~~rn
nn the linn
298 Microstrip dipoles Microstrip dipoles 299

The bandwidth of an EMC dipole is almost identical to that of a centre-fed

dipole on the same substrate, provided the EMC dipole is matched. Hence, it
is desirable to keep the dipole significantly above the ground plane to obtain a
reasonable bandwidth. On the other hand, it is also desirable to keep the line as
close as possible to the ground plane, to minimise line radiation. However, for
a given substrate thickness h l l , corresponding to a prescribed bandwidth, the
line height should be chosen so that b > h - t,, to avoid being undercoupled,
which will reduce bandwidth. Hence, the height b = h - t,,, is a good trade-off
between bandwidth and line radiation. An improved coupling may be achieved
by using a top layer with E, > E , between the line and dipole. This improves
coupling by increasing the capacitance in between [lo]. However, a more pro-
nounced improvement is possible by using multiple dipoles.

53.3 Miiliipk d&i'~~

The bandwidthlline-radiation trade-off may be improved by using multiple
dipoles. A dominant theme in these schemes is the introduction of additional
capacitance between the line and the main radiating dipole. This is usually
accomplished by placing one or two coupling dipoles (parasitics) either in a
stacked fashion between the line and top dipole (Fig. 5.146) or coplanar to, and
near the end of, the line (Fig. 5.14~)[26].
For the stacked configuration, the bandwidth for SWR < 2.0 is shown
plotted against t p / t in Fig. 5.18 for two different values of h (3 mm and 4.5 mm)
with corresponding dipole lengths of 8.4mm and 8.7mm, respectively, at a
frequency of 10 GHz. These lengths are close to the input-match values for a
single EMC dipole. The transmission-line height b is constant a t 0.72 mm. Also
shown is the minimum achieved SWR as the frequency is varied for each value
of t,. From this Figure it can be seen that the thinner substrate has a lower
optimum bandwidth. Also, 'as the substrate thickness changes from 3 mm to
4.5 mm the range of tpfor SWR < 2.0 becomes smaller and shifts toward higher
values. Therefore, like the case of a single dipole, there is a maximum value h,,
such that, for h > h,,, the SWR is always larger than 2.0 for this value of b.
The addition of another dipole between the top dipole and line may further .
reduce the SWR and increase the bandwidth in this case.
When the parasitics are on the same level with'the line and of comparable
length to the radiating dipole, the bandwidth may be improved as shown in Fig.
5.19. Appropriate positioning of the parasitics may improve the bandwidth even
more. An advantage of this configuration is possibly fewer alignment difficulties
during fabrication. However, the improvement in bandwidthlline-radiation is
much more dramatic for the case of stacked dipoles.
300 Microstrip dipoles Microstrip dipoles 30 1
5.6 Finite array of EMC dipoles

5.6.1 Analysis
A finite array of EMC dipoles IS shown in Fig. 5.20. Away from the coupled
ends, the current o n the mth line may be written as the sum of incident and
backward waves, of the form
J,,(x,y) = 4,?Cv)e-'P'
Jh,(x,y ) = t,vb)e+'"' (5.29)

SIDE-VIEW OF ELEMENT #n

Fig. 5.20 Finite array of EMC dipoles

 302 Microstrip dipoles Microstrip dipoles 303

with x = 0 at the line end. The current on the mth dipole may be assumed To calculate B, and D m ,dipole n may be excited with a &gap source (line
sinusoidal as n may be absent), with line-dipole pair m passive.
Then
J;(x, Y) = I , d ? ( ~ ) m ) (5.30)
Here <(x) is given by eqn. 5.8 with d replaced by Lm/2.Because of linearity, the
line and dipole current amplitudes may be written as

where B,,, Dm,are back-scattering and dipole coefficient accounting for mutual
interaction between dipoles. To, is the voltage reflection coefficients at the end of
line m when all dipoles except number m are absent. E,, is similarly an excitation
coefficient in the isolated case. These equations may be written in matrix form
as

where [r], [El are diagonal matrices and [B], [Dl are zero on the diagonal. [Ib]
and [Ti] are column vectors.
Eqn. 5.34 yields
[fl = [El-' ([ul - [Dl) [PI
with [ V the identity matrix.
Substituting into eqn. 5.33 yields

Eqn. 5.35 gives the incident currents required to produce the desired set [PI
(which determines the radiation pattern). Eqn. 5.36 gives the resultant back-
scattered currents. From this the scattering matrix is found to be
[Sl = [ r l - [BI([UI - [DI)-'[Kl (5.37) Input Match
for the case of identical lines.
1-
5.6.2 Calculation of coefficients
The transmission-line analysis used to calculate r in Section 5.5.1 may be
extended to calculate E, as well as B,,,,, Dm,,between two line-dipole pairs. The
Fig. 5.21 Excitation coefficient E versus longitudinal offset for a single EMC dipole
formula for E is
E , = 2.2. E , = 2.35, b = 0.03 in, t = 0.06 in, L, = 0.367 in, and w = 0.059 in with
f = 1OGHz

where I, = j I, e-'8' is the incident-current amplitude, with I the line length.

with I, the line current at a maximum (x = x,,). The line may be excited as As for the calculation of r, a formulation using traveling-wave-basis functions
discussed in Section 5.5.1. could also be used here.
 304 Microstrip dipoles Microstrip dipoles 305
Results were calculated for the case E,, = 2.2, E,, = 2.35, b = 0,03in., used as an EMC dipole excited by line no. 1. The results for the EMC case
t = 0.06in, w = 0.059in at a frequency of 10GHz. The dipole lengths are depend on the feed-line length (TL,) to some extent owing to spurious coupling
taken as 0.367in, the value required for an input match in the isolated case with . between line no. 1 and dipole no. 2. The agreement is fairly good, however. In
only longitudinal offset. In Fig. 5.21 the excitation coefficient E is shown plotted Fig. 5.23, S,, results [27] are shown for the case measured by Stern and Elliott
0.25
T-IL 1-1

experimental - Ld-
: theoretlcz
0 . 3 8 4 5 rnch

Broadside Dipoles

k-
- JDIZ I k
............................ ....................................................

-
- - Source o n L i n e Feed
- .............. C e n t e r - F e d D i p o l e
-
-
-
-
....................................................................................
1-
I B12 1
-
- Fig. 5.23 Comparison o f theoreticalandmeasuredS,, for two line-dipole pairs. (Reproduced

- from Reference 27, @ 1987 IEEE)

The two dipoles are broadside with separation S, between centres. Measured
values taken from Reference 29: E, = 2.35, b = 0.037in. t = 0,0485in.
w = 0.059 in with f = 10GHz. The reference planes have been placed at the first
current maximum from the end of the lines

Fig. 5.22 Mutual coupling coefficients Dl,, E l , for two line-dipole pairs
Coefficients are calculated with dipole no. 1 excited in two ways: with a centre feed
(no line) and by coupling to a line of length TL, . Dimensions are the same as in Fig.
5.21 with dipole offsets chosen to give a match in the isolated case
against offset for a single line-dipole pair. The magnitude of E is a maximum
close to the input-match point (A, = 0.1 14in). In Fig. 5.22 results for B , , , D,,
between two pairs are shown. Dipole no. 1 is excited as a source in two different
ways: first, by a &gap feed at the centre (line no. 1 absent), and secondly, when
[28,29]. The dipoles were taken to be slightly shorter than the experimental
lengths since rounded dipoles were used in the measurement.
5.6.3 Array design
There are two possibilities for an array design: to have an input match on each
line in the active state, or to relax the match condition and simply put matching
transformers or stubs on each line. The second case is simpler because all dipole
lengths and offsets may be chosen the same (L,, = L,). Eqns. 5.35 and 5.36 give
the line currents, with the active reflection coefficient for line m given by
306 Microstrip dipoles Microstrip dipoles 307
As an example, a 4-element linear array with a 0.5 12, element spacing was
designed to give a broadside beam with a dipole excitation ratio of
From this the matching network may be designed. In particular, knowing
0.65: 1.0: 1.0:0.65. The board materials and line-dipole widths were the same as
determines the distance from the end of each line at which the impedance is
those in Figs. 5.21 - 5.22. The dipoles were chosen to be of resonant length at
purely real. At this point quarter-wave transformers may be placed to im-
the design frequency of 10.0GHz with an offset chosen to give a match in the
pedance match to the desired feed-line impedance. The input power on line m
is P, = + ZomIf,12(1 - lr;l2) where Z,,, is the impedance of line m, which
isolated case. A summary of the coefficient values and results obtained from the
equations in Sections 5.6.1 and 5.6.2 is given in Table 5.1.
couples to the mth dipole. The phase of fmdetermines the necessary phase delay,
Based on these results the ratio of powers into the feed lines is calculated as
and hence line length, of line m.
P2/Pl = P3/P4= 2.55. Because LT; = Lrt in this particular design, the
In the first case, an iterative procedure may be used to solve for each dipole
length and offset to give Tm= 0. This could be achieved by starting with an
initial length and offset to give an input match in the isolated case, and then
using eqn. 5.35 to find fm.The new length and offset are chosen to satisfy

The process then repeats. This is similar to the iteration scheme used by Elliott
[28] to design an array of EMC dipoles, which was based on an experimental
evaluation of coupling.
By utilising one of the design techniques described above, a complete array
of EMC dipoles may be designed, which will have prescribed dipole currents in
the presence of mutual coupling. The dipole currents directly determine the
array pattern, neglecting line radiation. The design equations 5.33,5.35 and 5.41
permit the direct design of the array feed network using standard corporate-feed
power dividers once the coupling coefficients have been obtained.

W i l k i n s o n Power S p l i t t e r s
Table 5.1 Summary of coefficient values and results
r = o

Fig. 5.24 Diagram of the feed network for a four-element array (not actual size)
The dipoles are shown displaced from the feed network for clarity

impedance is purely real and a minimum at approximately the same distance

from the end of the line on each feed line, at a distance s = 4.43 cm (A8 on the
feed lines is 2.02cm). At this location the active impedance is approximately
40Cl (from simple transmission line calculation). Hence, a feed network which
gives the required power split into impedances of 40Cl is required. Such a feed
network was designed using standard Wilkinson power splitters [30], and is
shown in Fig. 5.24 [31]. After a 1: 1 power split, each Wilkinson splitter further
 308 Microstrip dipoles
splits the power to a 2.55: 1.0 ratio. The 4 0 0 lines coming out of each Wilkinson
splitter meet the 62 !2 lines of width 0.05 1, at a distances from the ends, allowing
each Wilkinson splitter to feed into a matched load. The inner two lines were
made slightly shorter than the outer lines to account for a small unbalanced
phase shift through the Wilkinson splitters, which was observed experimentally.
The theoretical and measured radiation patterns for this array are shown in
Fig. 5.25. The theoretical pattern is found from a reciprocity method [2].The
H-Plane
9.9 GHz
Fig. 5.25 Theoretical (dashed line) and measured (solid line) patterns for the four-element
array
The theoretical pattern is at 10.0 GHz while the measured pattern is at 9.9 GHz.
E,, = 2.2. E,Z = 2.35, b = 0.03 in. t = 0.06 in. L, = 0,367 in and w = 0.059in. The
offsets are Ax = 0.1 1 4 in
measurements were performed at a frequency of 9.9 GHz since this was found
experimentally to be the optimum frequency for the array. At this frequency the
SWR on the 50R input feed line was approximately 1.4.
Finally, it should be mentioned that in large arrays of EMC dipoles, accurate
results may be obtained by using an infinite array analysis, involving a summa-
tion of Floquet modes [32]. Results for an infinite EMC dipole array have
recently been obtained [33].
5.7 Conclusions
Microstrip dipoles are generally characterized as being narrower than patch
antenna elements, and are not usually probe-fed like the patch. These features
Microstrip dipoles 309
allow the microstrip dipole to be a useful radiating element in many applica-
tions. In particular, due to the small size, the dipole may be useful when space
limitations are important. Dipoles have low radiation resistance and narrow
bandwidth for thin substrates, in comparison with the patch antenna. However,
the dipole may be used as a resonant element for thicker layers, for which the
bandwidth may be quite considerable, and the input resistance no longer a
limitation. Dipoles thus find best application for thicker substrates.
In this chapter a general method has been presented for analyzing strip
configurations, which include the microstrip dipole as well as the feeding micro-
strip lines. The analysis technique discussed is flexible, allowing for a wide
variety of different configurations, including the single dipole, the mutual cou-
pling between two dipoles, or the electro magnetically coupled (EMC) dipole.
One of the most practical methods for feeding a microstrip dipole is by
electromagnetic coupling to a microstrip line. If the dipole is sufficiently close
to the iine, an input match can always be achieved by varying the dipole length
and the offset from the line, in either the longitudinal or transverse directions.
In the EMC dipole it is usually desired to minimize the line radiation as much
as possible, while maintaining an input match. This implies a line height above
the ground plane for which the dipole is critically coupled. T o improve the
bandwidth-line radiation trade-off, multiple dipoles may be coupled to the line,
either in a stacked configuration, or coplanar with the line. The stacked con-
figuration gives the best improvement in bandwidth, for a given line height.
A design procedure for a finite array of EMC dipoles may be developed using
a matrix description of the line and dipole currents, together with a moment-
method solution for the necessary mutual coupling coefficients. Two design
procedures were discussed. One is an iterative procedure which yields an input
match on each line, but requires each dipole length and offset to be different. The
other design procedure allows each dipole length and offset to be the same, but
requires matching transformers on each line. This procedure is simpler, not
requiring any iterations. In both cases, the design equations allow for the direct
determination of the necessary feed network.
5.8 References
I SOMMERFELD, A,: 'Partial differential equations' (Academic Press, 1962)
2 ALEXOPOULOS, N. G., and JACKSON, D. R.: 'Fundamental superstrate (cover) effects on
printed circuit antennas', IEEE Trans., 1984, AP-32, pp. 807-816
3 COLLIN. R. E.: 'Field theory of guided waves' (McGraw-Hill. 1960)
4 CARVER. K. R. and MINK, J. W.: 'Microstrip antenna technology', IEEE Trans., 1981,
AP-29, pp. 2-24
5 POZAR. D. M.: 'Considerations for millimeter wave printed antennas', IEEE Trans., 1983,
AP-31, pp. 740-747
6 ALEXOPOULOS, N. G., JACKSON, D. R., and KATEHI, P. B.: 'Criteria for nearly
omnidirectional radiation patterns for printed antennas', lEEE Trans., 1985, AP-33, pp.
195-205
 3 10 Microstrip dipoles

7 JACKSON, D. R. and ALEXOPOULOS. N. G.: 'Gain enhancement methods for printed

circuit antennas', IEEE Trans., 1985, AP-33, pp. 976-987
8 JACKSON, D. R. and OLINER, A. A.: 'A leaky-wave analysis of the high-gain printed
antenna configuration', IEEE Trans., 1988, AP-36, pp. 905-910
9 DENLINGER, E. J.: 'A frequency dependent solution for microstrip transmission lines',
,..
lEEE Trans.. 1971. M'IT-19. DD. 30-39 Chapter 6
10 JACKSON, D. R., and ALEXOPOULOS, N. G.: 'Analysis of planar strip geometries in a
substrate-superstrate configuration', IEEE Trans., 1986, AP-34, pp. 1430-1438
I I UZUNOGLU, N. K., ALEXOPOULOS, N. G., and FIKIORIS, J. G.: 'Radiation properties
Multilayer and
of microstrip dipoles', lEEE Trans., 1979, AP-27, pp. 853-858,
12 NEWMAN, E. H., and FORRAI, D.: 'Scattering from a microstrip patch', IEEE Trans., 1987,
parasitic configurations
AP-35, pp. 245-251
13 FILON. L. N. G.: 'On a quadrature formula for trigonometric integrals', Proc. Roy. Soc. D.H. Schaubert
Edin., 1928, 49, pp. 38-47
14 POZAR, D. M.: 'Improved computational efficiency for the moment method solution of
printed dipoles and patches', Electromagnetics, 1983, 3(3-4), pp. 299-309.
15 YANG, H. Y.: 'Frequency dependant modelling of passive integrated circuit components' Ph. 6.1 Introduction
D. dissertation, University of California, Los Angeles, 1988
16 KATEHI, P. B., and ALEXOPOULOS, N. G.: 'Real axis integration of Sommerfeld integrals
with applications to printed circuit antennas', J . Math. Phys, 24, (3). pp. 527-533
. ..
A standard configuration for a microstrip antenna is a single patch of conductor
17 KATEHI, P. 8.: 'A generalized solution to a class of printed circuit antennas' Ph.D. Disserta- supported above a ground plane by a simple dielectric substrate and directly
tion, University of California, Los Angeles, 1984 contacting an appropriate transmission line in order to couple power between
18 JACKSON, D. R. and ALEXOPOULOS, N. G.: 'An asymptotic extraction technique for the resonant patch antenna and the transmitter or receiver circuit. This is a
evaluating Sommerfeld-type integrals,' IEEE Trans., 1986, AP-34, pp. 1467-1470
19 RUMSEY, V. H.: 'Reaction concept in electromagnetic theory', Phys. Rev., 1954, 94, pp. simple configuration that is rugged and relatively easy to fabricate, but it is
1483-1491 limited in its functional capabilities. The focus of this Chapter is antennas that
20 JONES, D. S.: 'A critique of the variational method in scattering problems', IRE Trans., 1956, consist of two or more metallic patches supported by one or more dielectric
AP-4, pp. 297-301 layers, or that consist of one metallic patch that is not directly contacting the
21 HARRINGTON, R. F.: 'Time harmonic electromagnetic fields' (McGraw-Hill, 1961) transmission line that feeds it. These configurations are more complicated to
22 ALEXOPOULOS, N. G. and RANA, I. E.: 'Mutual impedance computation between printed
dipoles', IEEE Trans., 1981, AP-29, pp. 106-111 design and fabricate, but they offer performance features that are not usually
23 KATEHI, P. B., and ALEXOPOULOS, N. G.: 'On the modeling of electromagnetically obtainable from the single-patch, single-dielectric configuration. These features
coupled microstrip antennas - The printed strip dipole', IEEE Trans., 1984, AP-32, pp. include increased bandwidth, multiple frequency operation, dual polarisation,
1179-1186 lower sidelobe levels, and ease of integration.
24 JACKSON, R. W. and POZAR, D. M.: 'Full-wave analysis of microstrip open-end and gap
The examples presented here do not represent all of the configurations that
discontinuities,' IEEE Trans., 1985, MTT-33, pp. 1036-1042
25 OLTMAN, H. G. and HUEBNER, D. A.: 'Electromagnetically coupled microstrip dipoles', have been successfully demonstrated, but they do represent many of the funda-
IEEE Trans., 1981, AP-29, pp. 151-157 mental methods that have been successful. One method often employed is to
26 KATEHI, P. B., ALEXOPOULOS, N. G. and HSIA, I. Y.: 'A bandwidth enhancement stack patch radiators one above the other with intervening dielectric layers. This
method for microstrip antennas', IEEE Trans., 1987, AP-35, pp. 5-12 allows two or more resonant patches to share a common aperture area. The
27 KATEHI, P. B.: 'A generalized method for the evaluation of mutual coupling in microstrip
arrays', IEEE Trans., 1987, AP-35, pp. 125-133
patches may be fed individually from microstriplines or coaxial probes, or only
28 ELLIOTT, R. S. and STERN, G. J.: 'The design of microstrip dipole arrays including mutual one or two may be fed directly while the others are coupled parasitically. Several
coupling Part I: Theory', IEEE Trans., 1981, AP-29, pp. 757-760 examples are presented in Section 6.2. Another method that is employed utilises
29 STERN, G. J. and ELLIOTT, R. S.: 'The design of microstrip dipole arrays including mutual a single resonant patch that is coupled electromagnetically to a microstrip feed
coupling. Part 11: Experiment', IEEE Trans., 1981, AP-29, pp. 761-765 line. This form of parasitic coupling usually involves two layers of substrate,
30 HOWE, H.: 'Stripline circuit design' (Artech House, 1974)
31 DINBERGS, A.: 'Analysis and design of an array of electromagnetically coupled microstrip
which may be on the same or opposite sides of the ground plane. If the feedline
dipoles'. Masters thesis, University of Houston, 1988 and patch are on opposite sides of the ground plane, a small aperture can be
32 POZAR, D. M., and SCHAUBERT, D. H.: 'Scan blindness in infinite phased arrays of printed used to efficiently couple power through the ground plane. This configuration
dipoles', lEEE Trans., 1984, AP-32, pp. 602-610 is presented in Section 6.3.
33 CASTANEDA, J. and ALEXOPOULOS, N. G.: 'Infinite arrays of microstrip dipoles with a The final method presented here consists of one or more patches on the
superstrate (cover) layer', IEEE Intl. Symp. Digest, 1985, Vol 2, pp. 713-718
surface of a single substrate layer. These patches may be coupled parasitically
 312 Multilayer and parasitic configurations
to each other, and possibly to the feed line. These configurations are generally
easier to fabricate than the multilayer configurations, buithe often
is not as desirable. Some examples are presented in Section 6.4.
Multilayer and parasitic configurations 313
efficiency of antennas designed for increased bandwidth. Decreased efficiency
has been reported for some configurations, but available experimental and
analytical data are not sufficient to quantify the relationship between bandwidth
and efficiency for the many configurations that have been demonstrated.

6.2 Stacked elements for dual-frequency and dual-polarisation operation 6.2.1 Antennas with separate feeds for each function
Stacked patches with separate feeds can take a variety of forms. Two of these
Stacked elements have the advantage of providing two or more metallic patches are shown in Fig. 6. lb and c. Another form is depicted in Fig. 6.2. In all these
within the same aperture area. This allows the antenna designer to obtain configurations, the outer conductor of the upper feed connects the lower patch
multiple frequencies with or without dual polarisation. Three typical configura- to the ground plane (Fig. 6.ld). This short-circuit connection, which actually
tions are depicted in Fig. 6.1, where case a represents a triple-frequency dual- presents a small inductive load to the antenna, can often be placed to have
polarised antenna [I], case b represents two linearly polarised elements operat- minimal effect on the antenna performance. However, it also can be placed to
ing at different frequencies [2], and case c represents two circularly polarised achieve desirable tuning effects, which will be described below.
elements operating at different frequencies. These examples are representative of The use of one-half-wavelength and one-quarter wavelength elements in the
stacked configurations, which may use a single feed for multiple frequencies and stacked configurations offers the designer considerable flexibility in selecting the
different feeds for each polarisation (case a) or separate feeds for each frequency operating frequencies of the antenna. Of course, the use of one-quarter-
and polarisation (cases b and c). The dielectric substrates may differ in thickness wavelength elements restricts the radiated field from that element to be linearly
or permittivity in order to control the bandwidths and sizes of the metallic polarised, and requires the fabrication of via connections to form a short circuit
radiator. along one side of the antenna. Nevertheless, this configuration is desirable for
This Section is organised into two Subsections. The first describes antennas many applications and the designer can stack a one-quarter-wavelength patch
that utilise separate feeds for each frequency and polarisation. The second and a one-half-wavelength patch of comparable sizes in order to obtain operat-
describes antennas that utilise a single feed to obtain multiple frequencies. ing frequencies that are separated by approximately one octave. Independent
Before presenting details of these stacked antennas, it is desirable to list some control of E,,,,, and q,,,, (Fig. 6. Id) provides another means of adjusting the two
of their general advantages and disadvantages. operating frequencies.
The piggy-back antenna depicted in Fig. 6.lb and described below is one
Advantages Disadvantages
example of an antenna that uses a one-half-wavelength lower element and a
Multiple functions share Stacked substrates must be smaller one-quarter-wavelength upper element that is tuned to operate at a
common aperture. aligned and bonded. frequency moderately close to that of the lower element. This configuration
leads to different beamwidths at the two closely spaced operating frequencies.
Stagger tuning increases Increased thickness and (Care must be exercised if extremely close operating frequencies of the same
bandwidth. weight of the antenna structure. polarisation are required because the stackedklements can exhibit strong mut-
Separately tuned radiators Fabrication of feed can be ual coupling, as described in Section 6.2.2.) As noted above, one of the principal
benefit from frequency and/or difficult, particularly when upper disadvantages of the one-quarter-wavelength antennas is the need to fabricate
polarisation isolation feed must attach to lower antenna. a short-circuit boundary by plated-through holes, rivets or soldered pins.
The shapes of the patches that are stacked is somewhat arbitrary, although
Many configurations are possible Increasing total substrate most designs that have been reported use similar shapes for the upper and lower
to meet a variety of needs thickness increases excitation of
patches. Square (or rectangular) and circular patches are the most commonly
surface waves, resulting in lowered used shapes, but the wedge shape depicted in Fig. 6.lb is useful for conformal
Different substrates may be
efficiency. antennas on small conical bodies.
selected for upper and lower
antennas.
Two-frequency antennas: The piggy-back antenna in Fig. 6. lb, or any similar
Most of these advantages relate to increases in performance, whereas most of configuration, has been found to work well for radiating or receiving two
the disadvantages relate to fabrication and mechanical concerns. One perfor- independent, linearly polarised signals at different frequencies from a common
mance parameter that has not received sufficient attention in the literature is the aperture. The input impedance and radiation patterns of the piggy-back antenna
374 Multilayer and parasitic configurations Multilayer and parasitic configurations 3 15

are typical of those obtained from ordinary patch antennas, although any
asymmetries of the stacked geometry may lead to slight asymmetries in the
patterns. Some asymmetries are evident in the E-plane patterns of Fig. 6.3. The
ground plane for the antenna is modest in size, so that some diffraction effects

_shortmg pin

feed points

section A A'

Fig. 6.1 Typical stacked-patch configurations

a Triple-frequency, dual-polarised antenna [After Reference 11
b Piggy-back antenna for two linearly polarised frequencies
c Two-frequency, circularly polarised antenna
d Cross-section of stacked patches that utilise upper feed as inductive post in lower
Wtch

are also evident. The operating frequencies and impedance bandwidths of each
element are approximately the same as they would be in the absence of the other,
top view
except that the lower element must be considered as a post-tuned antenna [3].
cross sectional vlew
(AA)
The feed for the upper element may not, in general, pass through the lower
element at a point where the electric field between the lower patch and the
ground plane is zero. Then, the effect is similar to an inductive post in a
 37 6 Multilayer and parasitic configurations Multila yer and parasitic configurations 377
/ A rectangular waveguide cavity. Several papers have analysed the post-tuned
antenna [4-61 and have succeeded in predicting quite accurately ,the resonant
frequency and input impedance. However, a first-order approximation is easily
obtained by using the transmission-line model [7] for a rectangular antenna as

Fig. 6.2 Orthogonally polarised version of piggy-back antenna with low-frequency, one-
quarter-wavelength patch on bottom

narmalised post spacing (sla)

E-plane H-plane

---- 1140MHz
quarter wave
Fig. 6.4 (a) Transmission-linemodel of patch symmetrically loaded with inductiveposts. (b)
- patch
990 MHz
Operating frequency (upper curves) and VSWR (lower curves) of post-tuned anten-
na on 1.6 mm Teflon fibreglass substrate
--- calculated.
-measured [taken from Reference 31

is done in Reference 3. Typical results taken from Reference 3 are shown in Fig.
6.4. The values used in the calculations for the post reactances were obtained
from

Fig. 6.3 Radiation patterns of piggy-back antenna

xp = x
377
tan
2zt
n ( t = substrate thickness) (6.1)
 318 Multilayer and parasitic configurations Multila yer and parasitic configurations 3 19
which is Carver's estimate 181 that the post's reactance should be the same as the
input impedance to a piece of short-circuited transmission line. The use of only
one inductive post yields similar results, but the amount of frequency increase - LH C

caused by the post is approximately one-half that caused by two posts (Fig. 6.5).

near feed
opposlte feed
horizontal feed,
\. +

Fig. 6.6 Rectangular patch antenna with feeds for horizontal and vertical polarisation at
different freyueritiies

post location (2dla)

Fig. 6.5 Measured operating frequency and VS WR of antenna with one post near the feed or
one post opposite the feed

Two-Polarisation Antennas: A simple rectangular patch antenna with two

feeds located along perpendicular centre lines of the antenna is an easy choice
for radiating or receiving two independent, orthogonal, linearly polarised sig-
nals. However, this antenna has the disadvantage that the width of the patch for
the horizontal mode is fixed by the desired operating frequency of the vertical
mode, which determines L, (Fig. 6.6). Therefore, the H-plane beamwidth of the
horizontal mode is fixed by the vertical frequency, and vice versa. This limitation
can often be overcome by using a stacked configuration such as the one in Fig.
6.7. The upper feed of the model shown passes through the lower element at a
voltage null so that the primary effect on the lower antenna is some loading due
to the upper substrate, but this effect is easily compensated in the design. Thus,
the patch widths and H-plane beamwidths can be controlled as long as the
performance requirements allow the designer to fit the upper vertically polarised
antenna within the boundaries of the lower, horizontally polarised antenna. If
the use of differing permittivities for the two substrates is permissible, a higher-
permittivity upper substrate helps to reduce the size of the upper patch.
The radiation patterns and input impedances of the individual antennas are
similar to those of comparable antennas operating in an isolated environment.
Some typical radiation patterns are shown in Fig. 6.8. The impedances of the
Fig. 6.7 Stacked patches for orthogonal linear polarisation at two independent frequencies
320 Multilayer and parasitic configurations Multilayer and parasitic configurations 32 1
two elements, when both are tuned to operate at 3.465 GHz, are shown in Fig.
6.9. The isolation between the feeds for the two orthogonally polarised elements
is greater than 18 dB over the operating band.

E plane E plane

Fig. 6.8 Typical radiation patterns of stacked orthogonal patches on a small ground plane
a Upper patch
b Lower patch start 3.000000000 GHz
stop 4.000000000 GHz

6.2.2 Antennas for multiple frequencies and increased bandwidth Fig. 6.9 Input impedance of stacked orthogonal patches
Each substrate is 1.6 mm thick, 8, = 2.2; upper patch is 2 9 x 23 mm and is fed
One form of the stacked patch antenna for multiple frequencies is depicted in 11.6 rnm from edge; lower patch is 27 x 36 mm and is fed 8.0 mm from edge
Fig. 6 . 1 0 ~This
. antenna is similar to the one in Fig. 6. l a [I], where three patches
permit three separate operating frequencies to be obtained at each of the
orthogonally polarised feed ports. This form is typical of that used for multiple-
frequency antennas where the feed probe passes through a clearance hole in the
lower patch and connects to the upper patch. This method of feeding couples
strongly to the resonance of each patch, even though the resonant frequencies
may be far part. This strong coupling is probably the result of currents on the
feed probe directly exciting the cavity of each patch antenna through which it
passes.
A second form of the stacked patch antenna is depicted in Fig. 6.10b, where
only the lower patch is fed directly and the upper patch is coupled parasitically.
This form of the antenna is widely used for increased bandwidth at a single
operating frequency [9-111, and it can be used for dual polarisation by inserting
a second, orthogonally polarised feed similarly to the multiple-frequency anten-
na in Fig. 6 . 1 ~ .
The characteristics of both forms of the stacked, single-feed patches have been b
investigated and compared to a single patch. For the data presented below, the Fig. 6.1 0 Single feed configurations for multiple frequencies or increased bandwidth
patches were circular and fabricated on separate sheets of Duroid 5870 a Top-feed model
b Bottom-feed model
 322 Multilayer and parasitic configurations Multilayer and parasitic configurations 323

Fig. 6.11 Input impedances of dual-frequency, top-fed circular discs

a Dimensions of the model
b DUD,, = 3.3 cm.
c D,,, = 3.45 cm
d D,,, = 3.55 cm
Fig. 6.1 1 Cont.
 324 Multilayer and parasitic configurations Multilayer and parasitic configurations 325

( E , = 2.3) substrate and then carefully aligned in the stacked configuration. The
input impedance of the multiple-frequency form is shown in Fig. 6.1 1 for three
different diameters of the upper patch. These and additional data are sum-
marised in Table 6.1, which tabulates the maximum value of input resistance
and the associated frequency and Q. As noted by Long and Walton [12], the
resonant frequency and resistance of the lower patch are relatively unaffected by
F i g . 6.12 Equivalent circuit to model input impedance of Fig. 6 . 1 1 ~
changes in the diameter of the upper patch. However, the resonant frequency
and resistance of the upper patch both decrease as the patch diameter increases.
In fact, increasing the upper patch diameter to 3.6cm results in an upper
resonance at 3.277 GHz with a resistance of 16ohms. It appears that the upper
patch in this configuration should always be smaller (or only slightly larger)
than the lower patch.
D
Table 6.1 Impedance characteristics of multiple-frequency stackedpatches
Upper patch R,.,, f.. GH7
.,,, Q
diameter cm ohms
3.3 75 3.185 61
93 3.465 60

Lower patch diameter: 3.5 cm

@"' = 8 P P " = 2.32
d""" = durnr = 04.79 cm = 0.030 in
Feed is 0.6crn from centre

An equivalent circuit has been used successfully to model the input impedance
of the stacked, dual-frequency antenna. This circuit, depicted in Fig. 6.12,
consists of two coupled parallel resonant circuits and a series inductor to model
the feed inductance that is usually observed in probe-fed patches. The values
shown in Fig. 6.12 are for the 3.45 cm-diameter upper patch. These values were
obtained by optimising the fit between the impedance of the measured data and
b
the model. The mutual inductance represents electromagnetic coupling between
Fig. 6.13 Input impedance of wide-bandwidth bottom-feed circular discs
the two discs. a Dimensions of the model
The wide-bandwidth form of stacked patches (Fig. 6.10b) has also been b D,,, = 3.45 cm
studied and three typical results are shown in Fig. 6.13. The best results for c D,,, = 3 . 5 0 cm
increased bandwidth are obtained when the two patches are nearly the same d DUDp,= 3 . 5 5 cm
 326 Multilayer and parasitic configurations
Multilayer and parasitic configurations 327
size. Since the feed probe does not pass through the cavity of the upper patch,
the coupling to that patch is very weak unless it is comparable in size, or larger
than the lower patch. Thus, the preferred configuration for multiple frequencies
is to feed the top disc and use a larger bottom disc, while the preferred configura-
tion for increased bandwidth is to feed the lower disc and use an equal or slightly
larger top disc. By properly adjusting the feed distance from the patch centre,
the impedance loop in Fig. 6 . 1 3 ~can be made to encircle the centre of the Smith
chart. This has produced a model fabricated from two sheets of 0.076 cm-thick
(0.030 in) Duroid 5870 (E, = 2.3) having a VSWR < 2 bandwidth of 5% at
3.3 GHz. This compares with approximately 3% bandwidth for a single patch
on 0.15 cm-thick substrate.

Fig. 6.14 Stacked discs for increased bandwidth [After Reference 101
a Antenna structure
b Return loss

Chen et al. [lo] have provided data on the operation of antennas in the form
of Fig. 6.10b for either wide bandwidth or dual frequencies. By using a relatively
thick, low-permittivity foam substrate between the patches, they have achieved
a 20% bandwidth for VSWR < 2 and a 10% bandwidth for VSWR < 1.22.
A typical result is shown in Fig. 6.14. They measured a gain of 7.9 dB at
4.25 GHz for the 10% bandwidth antenna. They also noted that stacked
patches with the lower patch on a relatively thin substrate exhibit lower cross-
polarised radiation than a single patch fabricated on a thick substrate to achieve
comparable bandwidth. This result is consistent with observations that direct
radiation from feed probes can be significant for antennas on thick substrates
[13]. For dual-frequency operation, Chen et al. present a graph showing the
relationship between the patch diameters, separations and operating frequen-
cies. However, they do not indicate if the results are limited to their particular
choices of dielectric substrates.
An analysis of stacked, circular patch antennas has been conducted by Araki
et al. by using the spectral domain Green's function [14]. They solve the
Fig. 6.13 Cont. eigenvalue problem to find the complex resonant frequencies of the structure,
and they calculate the input impedance. Their results are compared to measure-
 -
328 Multilayer and parasitic configurations Multilayer and parasitic configurations 329
ments in Fig. 6.15, which shows good agreement for a wide-bandwidth example. create a series capacitance, which adds a degree of freedom that can be used in
No results are presented for a dual-frequency example, but their analysis should conjunction with the inductance of the feed probe and the resonant patch
be valid for this case, as well. impedance to obtain increased bandwidth in the manner of Griffin and Forrest
[IS]. Alternative configurations suggested by Paschen are shown in Fig. 6.17.

ground plane-

-_-----
I---- \ /
/-
wlthout parasttic

antenna patch

m 1.2 with parasitic

> element

-1
---- measured
1 x x xJ calculated
2.4 2.5 2.6 2.7 2.8
frequency .GHz

b Fig. 6.16 Proximity coupling by means of series capacitance between patch and feed probe
Fig. 6.1 5 (a) Geometry of stacked circular patches for spectral domain computations a Paschen's design [After Reference 151
( 6 ) VSWR characteristics for h = 10.0, a, = 20.8 mm, a, = 2 1 . 0 mm, b Hall's design [After Reference 171
d = 1.6 mm, 6, = 2 . 5 5 [After Reference 141 c Fong et a1 design [After Reference 161

Another antenna configuration that resembles those of Fig. 6.10 has been
proposed by Paschen [15]. However, this antenna actually functions more like
that of Fong et al. [16], as is apparent from alternative configurations shown by
Paschen. Three antenna configurations (from Paschen [I 51, Hall [17], and Fong
et al. [16]) are shown in Fig. 6.16. The objective of these configurations is to
The cylindrical form of the capacitor has been used to fabricate an L-band
antenna covering the global-positioning satellite frequencies of 1227 MHz and
1575 MHz as well as the NDS frequency of 1381 MHz. Air dielectric was used
and a quadrature feed network for circular polarisation was implemented on a
thin circuit at the bottom of the antenna cavity (Fig. 6.18). By adjusting the
330 Multilayer and parasitic configurations Multilayer and parasitic configurations 331
probe diameter to control probe inductance and also adjusting the cylindrical radtotlng element

capacitor, the antenna was tuned to provide VSWR less than 2 for a bandwidth
of 25%. A typical radiation pattern is also shown.
ground p l m e

quadrature
feed network
'spherlcol orray (dome1

stand-off
/ connector ptn

Fig. 6.17 Alternative configurations for coupling by means of cylindrical capacitors

6.3 Two-sided aperture-coupled patch

Microstrip antennas have been popular elements for planar and conformal
arrays. A traditional means of fabricating these arrays that takes maximum
advantage of the economies of printed-circuit fabrication involves the layout of
radiating elements and a feed network on a single surface of a grounded
substrate. This minimises the size and weight of the total array and requires the
installation of only one coaxial connector to feed the array, thus reducing the
cost. However, the microstrip feed network radiates small amounts of power
that can degrade the sidelobe and polarisation characteristics of the array. Also,
the radiators and feed lines occupy much of the available area, leaving little
space for the phase shifters that are required for beam steering. Furthermore,
for monolithic phase shifters, the substrate must be GaAs or another appro- b

priate material, which is not a desirable substrate for the radiators [19]. Fig. 6.18 Circularly-polarised L-band antenna utilising air dielectric and cylindrical capacitor
coupling [After Reference 151
Most of these problems can be alleviated by using a two-layer structure that
a Antenna for use on spherical surface
has the radiating elements and their substrate on one side of a ground plane and b Typical radiation pattern at 1381 MHz
 332 Multilayer and parasitic configurations
the feed network and its substrate on the other. In the past, arrays fabricated
in this fashion have utilised a via connection probe at each element in order to
transfer power from the feed network to the radiators (Fig. 6.19). However,
these via connections are increasingly difficult to fabricate as the frequency
increases and excess probe inductance makes the antenna difficult to match.
Also, the use of GaAs as a feed substrate complicates the fabrication because
it is more difficult to drill the via hole.
r a d i a t h g patch
via connect
\ /,.
Fig. 6.19 Two-sided design with via connection between feed line and radiating patch
A useful alternative to the via connection probe is aperture coupling [20, 211.
In this configuration, power is coupled from the microstripline feed to the
radiating patch through an electrically small aperture in the ground plane (see
Fig. 6.20). No electrical connection is required and the performance is relatively
insensitive to small errors in the alignment of the two circuits. Single elements
have performed well at frequencies as high as K,-band [22].
Several additional advantages are obtained by the use of a two-sided con-
figuration. These include isolation of the feed network from the radiating
aperture of an array, which eliminates the spurious-feed-network radiation that
can degrade polarisation and sidelobe levels [13]. Also, the two-sided configura-
tion provides two distinct microstripline media so that the antenna substrate can
be chosen to optimise the performance of the radiating patches (e.g. low 6, to
improve radiation, increase bandwidth, and move scan blindness further from
broadside), and the feed substrate can be chosen independently to optimise feed
performance (e.g. high E, to reduce circuit size or the use of GaAs for active
integrated feeds). Furthermore, the feed substrate may be composed of many
diced wafers without introducing substrate discontinuities into the radiating
side of the array.
The aperture-coupled patch antenna resembles a traditional microstrip-array
element with a microstrip patch antenna on a substrate with relative permittivity
$ and a feed network on a substrate with relative permittivity These are4.
separated by a common ground plane. In the traditional configuration, a via
connection carries power from the feed network to the radiating patch. In the
Multilayer and parasitic configurations 333
new design, the via connection is replaced by an electrically small aperture (Fig.
6.20) that couples power efficiently to the patch and is easy to manufacture. An
open-circuited length of the microstripline extending beyond the aperture
provides an additional degree of freedom to be used for impedance matching
and bandwidth enhancement. This stub, together with the aperture length, can
be used to control the input impedance over a wide range of values, as illustrated
in Fig. 6.21. The calculated results were obtained by using the moment-method
technique of Reference 21, and have been found to be in good agreement with
measured results.
rad~atmgpatch
E:-L
Fig. 6.20 Microstrip radiator electromagnetically coupled to microstripline feed in two-sided
configuration [After Reference 2 11
The performance illustrated in Fig. 6.21 can be understood in terms of the
equivalent circuit [23] in Fig. 6.22, which is a two-port representation of the
microstrip feed line as it passes the aperture. The effect of the aperture and patch
is approximately a lumped series load on the microstripline of characteristic
impedance Z,. The input impedance of the patch, Z,n, can be obtained by
evaluating the series combination of the aperturelpatch circuit and a short stub
terminated in an open circuit. The series inductance represents the effect of the
electrically small (below resonance) aperture, and the impedance Z,, represents
the open-end effects of the microstrip stub. As the stub increases in length, the
input impedance at a fixed frequency approximately follows a constant-resis-
tance circle in Fig. 6.21a, with the reactance increasing according to the reac-
tance of the open-circuited stub. The effect of increasing the aperture size is
 334 Multila yer and parasitic configurations
similar to that of increasing the size of a hole coupling power from a waveguide
to a resonant cavity. When the aperture is small, the patch is undercoupled and
the resonant resistance is less than the characteristic impedance of the feed line.
As the aperture size increases, the coupling and the resonant resistance increase.
A wide range of resistance and reactance values can be achieved by adjusting the
aperture length and the stub length (It has been found that narrow slot apertures
offer the most effective coupling in this configuration.)
Fig. 6.21 lmpedance of aperture coupled patch E! = E: = 2.54,d, = d, = 0.16cm; patch
length = 4 cm; patch width = 3 cm, aperture width = 0.16 cm
a lmpedance for various stub lengths with aperture length = 1 . 1 2 cm
6 lmpedance for various aperture lengths with stub length = 2.0 cm
[Reproduced from Reference 211
Multilayer and parasitic configurations 335
Fig. 6.21 Cont.
The bandwidth and radiation patterns of the aperture-coupled patch antenna
are essentially the same as those of a probe-fed antenna on the same substrate.
The peak radiation from the aperture on the feed side of the ground plane has
been computed and measured to be at least 20 dB below the peak of the patch
radiation for the antennas that have been tested. This is an important charac-
teristic of the antenna that makes it useful in planar arrays, as compared to a
simple microstripline-fed slot that has a bidirectional radiation pattern.
Arrays of aperture-coupled patches can be built by using series feeding or
corporate feeding. Examples of E-plane arrays are illustrated in Fig. 6.23. Both
types of arrays have been built at C-band. An eight-element corporate-fed array
with patches on 0.159 cm E: = 2.2 and feed lines on 0.064 cm = 10.2 has
performed favourably, demonstrating the feasibility of using a substrate like
GaAs for the feed network and a lower-permittivity substrate for the radiating
 336 Multilayer and parasitic configurations
elements. The use of a low-E,antenna substrate will increase the angle at which
scan blindness occurs due to surface waves on the antenna substrate. However,
a blindness also will occur due to surface waves on the feed substrate, and
methods may be needed to control this phenomenon at very short wavelengths
where the substrate thickness may exceed 0.02-0.03&.
Multilayer and parasitic configurations 337
modification in the dimensions and would allow nionolithic phase shifters to be
integrated with the feed network. Mutual coupling levels in the array are given
by the S parameters in Table 6.3, and the input impedance of a typical interior
element is shown in Fig. 6.24. Radiation patterns of the array, fed by an external
coaxial power divider and appropriate lengths of line to steer the beam, are
shown in Fig. 6.25. These patterns agree well with expected results and the
radiation behind the ground plane is 20 dB below the main beam, despite the
relatively small ground plane (approximately 212, x 41,).

Table 6.2 Eight-element, E-plane array

Patch length = 1.78cm Element spacing = 3.0cm
Patch width = 2.54cm Stub length = 0.42 cm
Slot length = 0.83 cm Slot width = 0.056cm
Zin stub
&y' = 2.22 pd= 10.2
8"' = g.:5F i i ~ i dfwd= 0@j4cm
Fig. 6.22 Equivalent two-port network for aperture backed by a patch antenna and fed by a
microstripline
Table 6.3 Mutual coupling in array with d0/2spacing
IS,l, dB L Sij, deg
s4, - 30.7
- 79
SQ - 24.8 102
m~crostr~pfeed ltne aperture ~n resonant patch s
43 -21.6 - 46
on lower substrate ground plane on upper substrate - 52
s45 -22.1
s46 - 27.0 103
s4, - 30.3 -110
s4, -31.8 70

A variation of the aperture-coupled patch is shown in Fig. 6.26 [24]. This

configuration permits the feed network to occupy as much space as is needed in
the depth dimension. This use of depth has been essential in most microwave
li------------ij phased arrays that have been fabricated. Other variations with perpendicular
1I :------
L------
feed substrates have been built with non-contacting feeds [25], which offer
I significant advantages in fabrication.
b
Fig. 6.23 Aperture-coupled microstrip antenna arrays
a Series feed '6.4 Parasitic elements on antenna substrate
b Corporate feed
The microstrip antenna design is so appealing that engineers are inclined to
capitalise on its conformability, manufacturability, and ruggedness for applica-
The array dimensions in Table 6.2 could be scaled to yield operation at tions that are contrary to its inherent electrical characteristics. In particular,
25 GHz, which would lead to a feed substrate thickness of 0.005 in. The use of considerable effort has been devoted to increasing the operating bandwidth of
0.005-in GaAs (E, = 12.8) for the feed substrate would require only a slight the microstrip antennas. Several of these efforts have been described in previous
 Multilayer and parasitic configurations 339
338 Multilayer and parasitic configurations

Sections of this Chapter. However, a recent study [26] has shown that the
bandwidth obtainable from a microstrip antenna is approximately proportional
to its volume measured in 1;. This phenomenon is consistent with accepted
antenna theory [27]. Therefore, the bandwidth of a simple microstrip antenna

Fig. 6.24 Input impedance of element 6 in corporately fed E-plane array having dimensions
in Table 6.2.All other elements terminated in 50 R

can be increased by increasing its length, width, or substrate thickness. How-

ever, the length of a fundamental-mode antenna must be approximately one-
half wavelength in the dielectric substrate, so the antenna can be lengthened
only if the substrate permittivity is lowered. Unfortunately, feed lines and
probes radiate more on low-~,substrates, so this technique must be used with
care when cross-polarisation and sidelobe levels are important. Increasing the
antennas's width is fairly straightforward, but higher modes of the antenna can
be excited if the width is increased to one or two wavelengths. Also, elements
larger than approximately 1,/2 cannot be used in scanning arrays owing to
undesirable grating lobes.
In this Section, some alternatives to multiple-layer structures and to increas-
ing the substrate thickness, which can lead to increased feed radiation and
surface wave excitation, are presented. The fundamental approach here is to
 340 Multilayer and parasitic configurations Multilayer and parasitic configurations 34 1

create a double-tuned resonance by adding parasitic resonators on the same
substrate surface as the primary microstrip antenna. Many of the elements
described here have two disadvantages that must be weighed against the benefits
of increased bandwidth: (i) the physical area of the element is increased, and (ii)
the radiation pattern exhibits asymmetries that may change with frequency.
method [32] has been used to accurately predict the input VSWR. Examples of
these antennas are depicted in Fig. 6.29. All of the structures described in
References 29-31 exhibit multiple tuning, and an element like the one in Fig.
6 . 2 9 ~has been measured to have 25% impedance bandwidth (VSWR < 2) at
Teflon fiberglass (3.2rnrn thick)
/

rnicrostr~p antenna length

patch on substrate of parasltlcs
13.0 crn and 2.5Gcrn w ~ d e
front s ~ d e
feed 13.l crn
coupl~ng substrate
aDerture I

,',/ J,/ / contacts top /

/
resonant

1
d~rect~on
of patch
X ,of aperture
spacing = 08 crn

plane

Fig. 6.26 Aperture-coupled patch with perpendicular feed network. (Reproduced from Re-
ference 24)

The first example (Fig. 6.27) utilises narrow conducting strips adjacent to the
driven radiator [28] in order to alter its impedance and radiation properties.
These strips couple to the non-radiating edges of the antenna and significantly
modify its impedance. A square, edge-fed microstrip antenna with VSWR = 4
can be matched to VSWR = 1.2 by using the parasitic strips. The antenna
works best when the parasitic strips are slightly longer than the patch. The
performance of the antenna is strongly affected by the separation between the 11
775
I
800
700 725 750
strips and the patch, and the best performance has been obtained when the frequency (MHz)
separation is 2.5 to 3 times the substrate thickness.
b
The interaction between the strips and the patch changes the resonant fre-
quency of the patch by a few percent. However, it is possible to broaden the Fig. 6.27 Parasitically tuned antenna with narrow strips adjacent to nonradiating edges
impedance bandwidth of the antenna by stagger-tuning the strips. Fig. 6.276 [After Reference 281
shows the VSWR of the antenna with and without the strips. Radiation patterns a Dimensions for U H F model
b VSWR w i t h and without parasitics
of the parasitic-tuned antenna are shown in Fig. 6.28. The H-plane radiation
pattern is slightly skewed by the asymmetry of the stagger-tuned strips.
The coupled-resonator approach has been extended to include up to four 3.16 GHz on 0.125411 substrate, E, = 2.55. In order to achieve this bandwidth,
parasitic elements [29] and to provide for direct as well as electromagnetic the total element area is approximately one wavelength square. Also, the radia-
coupling to the parasitic elements [30]. Parasitic elements coupled to the radia- tion pattern changes with frequency within the band of operation as the seg-
ting edge of the antenna are described in Reference 29, where the segmentation ments contribute with differing amplitudes and phases. This is illustrated in the
 342 Multilayer and parasitic configurations Multilayer and parasitic configurations 343
E-plane patterns of Fig. 6.30, which shows the reported performance of the
antenna in Fig. 6.31. The VSWR < 2 bandwidth is 24% (about seven times that
of a single patch on the same substrate). However, the radiation pattern changes
appreciably at the four frequencies shown in these plots. No data on cross-

kEe
---- I \.- 3.72 1'
\
- E-plane
H- plane (in
\
+ ..plane),'
'

-90' \ \ / I 90'
Fig. 6.28 Radiation patterns of parasitically tuned antenna [After Reference 281
OdB 10 20 30 30 20 10 OdB

Fig. 6.29 Antennas with four parasitic elements

a Electromagnetically coupled parasitics [After Reference 291
b Directly coupled parasitics [After Reference 301

polarised radiation are presented. The usefulness of this, or the other parasiti- Fig. 6.30 Measured performance of antenna in Fig. 6.31 [After Reference 301
cally coupled antennas, will depend on the designer's ability to accept increased aVSWR ,
element area and pattern asymmetries and variations across the operating band. b E, in @ = 0' plane
c E+ in @ = 90' plane
 344 Multilayer and parasitic configurations Multilayer and parasitic configurations 345

An extension of parasitically coupled elements leads to a parasitically coupled

array [33]. In this configuration, a single driven patch is coupled to closely
spaced adjacent patches, which are coupled to additional patches to form a
linear or planar array (Fig. 6.32). Arrays of this type have been developed
experimentally to producelinear or circular polarisation. The distance between
patches is reported to be 0.1 to 2 times the substrate thickness and can be

Fig. 6.31 Dimensions of antenna used to obtain results of Fig. 6.30

L = 3.0cm W,, = W,, = 0.025 cm
W = 2.0cm F i g . 6.33 Three parasitically coupled arrays [Reproduced from Reference 331
W,, = W,, = 0.44cm
I, = 2.85cm 6, = 6, = 0.71 cm
I, = 2.635 cm a = 0.48 cm
1, = I, = 2.35cm e, = 2.55cm
h = 0.31 8cm

F i g . 6.34 Measured performance of 7-element, E-plane linear array of parasitically coupled

patches [Reproduced from Reference 331

patch
radiator adjusted to control the power distribution in rows and columns for sidelobe
minimisation. This distance also affects the input impedance, which is ap-
Fig. 6.32 Planar array of parasitically coupled microstrip elements [Reproduced from Re-
ference 331
proximately equal to the impedance of a single patch divided by the number of
patches.
 346 Multilayer and parasitic configurations Multilayer and parasitic configurations 347

The performances of the three antennas in Fig. 6.33 are reported in Reference (taken at band centre) illustrate the wide beamwidth and low cross-polarisation
33. The linear array operates at 10.8 GHz and has the characteristics shown in levels obtainable with the antenna. Maximum cross-polarisation within the
Fig. 6.34. The 5 x 3 array is reported to have 9 dB gain at 8.55 GHz and 10 dB beamwidth is reported to be -21 dB over the band 5.00-544 GHz.
- 26 dB sidelobes. The 2 x 2 array is designed to produce circular polarisation
by feeding the driven patch at a corner. The ellipticity was measured to be less
than 2 dB over a bandwidth of 130MHz at 5.83 GHz.
Another antenna that consists of a driven element and additional metallisa-
tion on the substrate surface is the microstrip disc antenna with a short-circuited
annular ring [34]. The antenna (Fig. 6.35) can be considered as a microstrip disc

Fig. 6.35 Microstrip disc antenna with annular ring

Centre frequency = 5.21 GHz, h = 3.18 mm, d = 19.5 mm, g = 2.25 mm.
r = 9 mm, E, 2.5 [Reproduced from Reference 341

with a parasitic annular ring, or as a cylindrical cavity with an annular slot. The
dimensions of the cavity produce a resonance at approximately the same fre-
quency as the resonance of the microstrip disc. The element produces a circular-
ly symmetric radiation pattern with 10 dB beamwidth equal to 160, which is
appropriate for illuminating a reflector with F/D = 0.3. By adjusting the slot
width g, VSWR < 2 bandwidths of 10% have been obtained, but significant Fig. 6.36 Radiation patterns of disc antenna with short-circuited annular ring [Reproduced
gain reduction is experienced at the upper portions of the operating band. It is from Reference 341
speculated in Reference 34 that the losses are due to a second resonance that is a Principal planes
contributing to the increased bandwidth. The radiation patterns in Fig. 6.36 b Diagonal planes
Table 6.4 Antenna characteristics 9
U)
Antenna type Section Performance characteristics Fabrication requirements
Piggy-back 6.2.1 Independent antennas Align and bond multiple
share aperture layers 3
Multiple frequency and/or Coaxial feed lines %
3
polarisation soldered to ground plane 2
Upper feeds form tuning and to lower patches
2
o
3
posts in lower antennas Q
Stacked patches 6.2.2 Multiple frequencies Align and bond multiple 2
with upper Can be dual polarised layers a2.
patch fed Upper patches smaller 2.
0
Stacked patches with 6.2.2 Increased bandwidth Align and bond multiple o
lower patch fed Can be dual polarisation layers %
G.
Reduces cross-polarisation Upper patch larger
May reduce efficiency or same size Sg.
Capacitive feed 6.2.2 Increased bandwidth Precise control of 2
tuning May reduce efficiency capacitor gap
No effect on pattern
Aperture coupling 6.3 Feed network isolated from No via connection
radiation Microstrip stub occupies
More parameters to space on substrate
control impedance Independent choice of
Can be used in most substrates
cases where probe Alignment not critical,
could be used but bonding is required

Patch radiation and

bandwidth not affected
Some radiation from
stub and aperture
Parasitic beside 6.4 Increased bandwidth Single-layer fabrication
driven patch Usually some pattern Direct or electromagnetic
asymmetry coupling to parasitics
Pattern may change with
frequency
Large bandwidth requires
greater than 1,/2 size fB
P
74,
3
Q
P
2
2.
P
0
ra
0
3
3
c
3
3
2
350 Multilayer and parasitic configurations
6.5 Summary
Microstr~pantennas composed of multiple conducting patches or feed lines
electromagnetically coupled to the resonant, radiating patch offer several ad-
vantages over the traditional single patch connected to a feed line or feed probe.
These advantages include increased bandwidth or multiple frequency operation,
dual polarisation, and control of input impedance. However, the antennas are
often more complicated to fabricate or they may require more surface area, so
that the designer may be required to sacrifice one desirable feature in order to
obtain another one. Table 6.4 contains a summary of several key features of the
antennas described in this Chapter, and can be used as a guide in choosing the
best configuration for a particular purpose.
Most of the structures in this Chapter have been modelled by using either the
cavity model or moment methods, and the segmentation method has also been
applied to some of the configurations. Ir? genera!, the analysis techniques
provide a good qualitative model for the antennas, thus providing the designer
with the insights needed to develop a functional antenna. Many of the existing
analyses also have shown good quantitative agreement for the cases appearing
in the literature, but the ranges-of validity are not generally known, and the
computed results may not be sufficiently accurate for many practical values of
substrate permittivity and thickness or for variations in patch geometry.
Nonetheless, multiple-layer microstrip antennas and antennas utilising parasitic
coupling to the feed or to other resonant patches offer distinct advantages for
many systems applications, and are likely to be used in antenna systems that
require the specialised features available from these antennas.
6.6 References
I MONTGOMERY, N. W.: 'Triple-frequency stacked microstrip element.' IEEE Ant. and
Prop. Intl. Sym., 1984, Boston, MA, pp 255-258
2 JONES, H. S. Jr., SCHAUBERT, D. H. and FARRAR, F. G.: 'Dual-frequency piggyback
antenna.' US Patent 4 162 499, 24 July 1979
3 SCHAUBERT, D. H., FARRAR, F. G., SINDORIS, A. R., and HAYES, S. T.: 'Microstrip
antennas with frequency agility and polarization diversity.' IEEE Trans. 1981, AP-29, pp.
118-123
4 RICHARDS, W. F., and LO, Y. T.: 'A wide-band, multiport theory for thin microstrip
antennas.' IEEE Ant. and Prop. Intl Syrn., 1981, Los Angeles, CA, pp 7-10
5 RICHARDS, W. F., and LO, Y. T.: ;Theoretical and experimental investigation of a micro-
strip radiator with multiple lumped linear loads.'Electromagnetics,1983, 3, pp. 371-385
6 SENGUPTA, D. L.: 'Transmission line model analysis of rectangular patch antennas,' Elec-
tromagnetic~,1984 4, pp. 355-376
7 MUNSON, R. E.: 'Conformal microstrip antennas and microstrip phased arrays,' IEEE
Trans., 1974, AP-22, pp. 74-78
8 CARVER, K. R.: 'Input impedance to probe-fed microstrip antennas.' IEEE Ant. and Prop.
Intl. Syrn., 1980, Quebec, Canada, pp. 617-620
9 SABBAN, A.: 'A new broadband stacked two-layer microstrip antenna.' IEEE Ant. and Prop.
Intl. Sym., 1983, Houston, TX, pp 63-66
Multila yer and parasitic configurations 35 1
CHEN, C. H., TULINTSEFF, A,, and SORBELLO, R. M.: 'Broadband two-layer microstrip
antenna.' IEEE AP-S Sym. Digest, 1984, Boston, MA, pp. 251-254
HOLZHEIMER, T., and MILES, T. 0.: 'Thick, multilayer elements widen antenna band-
widths,' Microwaves & RF, Feb. 1985, pp. 93-95
LONG, S. A,, and WALTON, M. D.: 'A dual-frequency stacked circular-disc antenna,' IEEE
Trans., 1979, AP-27, pp. 270-273
HALL, P. S., and PRIOR, C. J.: 'Radiation control in corporately fed microstrip patch
arrays.' Digest of 1986 Journees Internationales de Nice sur les Antennes (JINA '86), Nice,
France, pp. 271-175
ARAKI, K, UEDA H., and TAKAHASHI, M.: 'Hankel transform domain analysis of
complex resonant frequencies of double-tuned circular disc microstrip resonators/radiators,'
Electron. Lett., 1985, 21, pp. 277-279
PASCHEN, D. A,: 'Practical examples of integral broadband matching of microstrip antenna
elements,' Antenna Applications Syrn., Univ. of Illinois, Urbana, IL, I986
FONG, K. S., PUES, H. F., and WITHERS, M. J.: 'Wideband multilayer coaxial-fed
microstrip antenna element,: Electron Lett., 1985, 21, pp. 497-499
HALL, P. S.: 'Probe compensation in thick microstrip patches,' Electron lelt., 1987, 23, pp.
606-607
GRIFFIN, J. M., and FORREST, J. R.: 'Broadside circular disc microstrip antenna,' Elec-
won. Lett., 1982, 18, pp. 266-269
POZAR, D. M., and SCHAUBERT, D. H.: 'Comparison of architectures for monolithic
phased arrays,' Microwave J. 1988, 29, pp. 93-104
POZAR, D. M.: 'A microstrip antenna aperture coupled to a microstripline,' Electron Lett.,
1985, 21, pp. 49-50
SULLIVAN, P. L., and SCHAUBERT, D. H.: 'Analysis of an aperture coupled microstrip
antenna,' IEEE Trans., 1986, AP-34, pp. 977-984
SCHAUBERT, D. H., JACKSON, R. W., and POZAR, D. M.,' 'Antenna elements for
integrated phased arrays,' Antenna Applications Sym., Univ. of Illinois, Urbana, IL, 1985
SULLIVAN, P. L., and SCHAUBERT, D. H.: 'Analysis of an aperture coupled microstrip
antenna,' RADC-TR-85-274, Rome Air Development Center, Feb. 1986
BUCK, A. C., and POZAR, D. M.: 'Aperture coupled microstrip antenna with a perpen-
dicular feed,' Electron Lett., 1986, 22, pp. 125-126
POZAR, D. M., and JACKSON, R. W.: 'An aperture coupled microstrip antenna with a
proximity feed on a perpendicular substrate, IEEE Trans., 1987, AP-35, pp. 728-731
HENDERSON, A., JAMES, J. R. and HALL, C. M.: 'Bandwidth extension techniques in
printed conformal antennas,' Proc. Military Microwaves '86, Brighton, England, 1986, pp.
329-334
HARRINGTON, R. F.: 'Time harmonic electromagnetic fields' (McGraw-Hill, NY 1961) pp.
307-31 1
SCHAUBERT, D. H. and FARRAR, F. G.: 'Some conformal printed circuit antenna de-
signs.' Proc. Printed Circuit Ant. Tech. Workshop, New Mexico State Univ., Las Cmces, 1979
KUMAR G. and GUPTA, K. C.: 'Nonradiating edges and four edges gap-coupled multiple
resonator broad-band microstrip antennas,' IEEE Trans., 1985, AP-33, pp. 173-178
KUMAR G. and GUPTA, K. C.: 'Directly coupled multiple resonator wide-band microstrip
antennas,' IEEE Trans., 1985, AP-33, pp. 588-593
KUMAR G. and GUPTA, K. C.: 'Broadband microstrip antennas using additional resonators
gap-coupled to the radiating edges,' IEEE Trans., 1984., AP-32, pp. 1375-1379
GUPTA, K. C. and SHARMA, P. C.: 'Segmentation and desegmentation techniques for
analysis of planar microstrip antennas,' IEEE Intl. Sym. on Ant. and Prop., June 1981, pp.
19-22
ENTSCHLADEN, H. and NAGEL, U.: Microstrip patch array antenna,' Electron Lett.,
1984, 20, pp. 931-933
PRIOR, C. J. and HALL, P. S.: 'Microstrip disc antenna with short-circuited annular ring,'
Electron. Lett., 1985, 21, pp. 719-721
 Chapter 7

Wideband flat dipole and short-

circuit microstrip patch elements
and arrays
G . Dubost

This chapter comprises of two sections. Section 7.1 discusses a flat dipole which
is a wide- bandwidth hybrid radiating source, originated and developed i n ,
France, and used in flat arrays having several hundred or more of such elements.
Its low radiation resistance is advantageously matched to the characteristic
resistance of the stripline used to feed it. In each array, spurious radiation is
avoided because the feed network is completely shielded. Section 7.2 describes
the short-circuited patch acting at quarter-wave resonance. Its large beamwidth
in the E-plane and weak coupling in the H-plane are characteristics particularly
suitable for use in microstrip-phased arrays with beam steering over a large
angular sector. Because of their small thickness compared with wavelength they
can be used with advantage in flat arrays having omnidirectional radiation or
a directional deflected beam. These two radiating dipoles are studied theoretic-
ally by means of models which are equivalent to several lossy coupled trans-
mission lines operating in the quasi-transverse electromagnetic mode.

7.1 Flat dipole elements and arrays

Fig. 7.1 shows several models of the usual wide-bandwidth flat dipoles. The
specific properties of each model are given in Table 7.1.

7.1.1 Elementary sources

The flat dipole is used as a folded slot dipole symmetrically fed across a gap.
When isolated in Fig. 7.1 (2) or used in a large array in Fig. 7.1 (3), it is fed by
means of a Lecher line or a stripline network, respectively. It has already been
described in References 2 - 5. Fig. 7.2 shows the flat dipole which is linearly
polarized, parallel to, and at a distance H from, a reflector. A dielectric sheet is
placed between the radiating structure and the reflector, so that it can be
fabricated by means of printed circuits. The analysis and synthesis are presented
in sections 7.1.1.1 and 7.1.1.2. The first model Fig. 7.1 (I), which was asymmetri-
cal, has been described previously [I] and is used in arrays [3]. A broadband
Wideband flat dipole and short-circuit microstrip
Wideband flat dipole and short-circuit microstrip 355

Fig. 7.1 Usual wide-bandwidth flat dipoles

radiating
slots

/ '\

d~electric reflector plane

(Er)

Fig. 7.2 Geometry of directional flat folded dipole

356 Wideband flat dipole and short-circuit microstrip Wideband flat dipole and short-circuit microstrip 357

circularly polarised flat radiating source is shown in Fig. 7.1 (4) and described The input impedance is due to the transformation of two impedances; namely
in Reference 3. A dual polarised model in Fig. 7.1 (5) is described next. the radiation impedances of the two radiating slots. This transformation is
We present a flat radiating source which is able simultaneously to receive or achieved along a non-radiating slot line. Each half slot can be considered as two
transmit two frequencies of orthogonal linear polarisations, and subsequently a
circularly polarised wave of either the right-hand or left-hand sense. It is a new
flat dipole arrangement [(5) in Fig. 7.11, [9, 101. Circular polarisation can be
obtained from a trivial patch antenna when using two orthogonally phased
feeds, but it has a narrow circularly polarised bandwidth [l I] when non-isolating
power splitters are used. To broaden the bandwidth, we can make use of a thick
substrate of low dielectric constant, with a thickness of approximately one-tenth
of a wavelength, and isolating power splitters (Wilkinson) as in Reference 18.
The higher modes must be suppressed [12, 3, 13, 141 when symmetrical feed
structures are used.
The radiating source (Fig. 7.3) is an enlargement of the linear wide-bandwidth
fiat symmetrica! folded dipole [4, 51. The radiating structure is composed of two
symmetrical crossed and overlapped flat folded dipoles with orthogonal electric
moments. Each dipole is composed of two metallic plates which are fed in
opposite phase. The two dipoles, nos. 7, 8 and nos. 9, 10, are etched onto one
of the two metallic faces of the first printed circuit (1). The radiation in one
half-space is possible when the whole metallic face (1 1) of an identical second
printed circuit (2) is used. The other metallic faces of the two printed circuits,
(1) and (2), support the two stripline central conductors, (5) and (6), which are
insulated with a thin dielectric sheet (3).
Each dipole is coupled by its feeding line through a gap by means of an open
quarter-wave stripline section, so that the two edges are fed in opposite phase.
The model operates between 3.45 and 3.85 GHz [15] with a VSWR of less than
2, when directors are used to broaden the bandwidth.

7.1.1.1 Analysis in quasi-TEM approximation: Self and mutal impedances,

and far fields are next studied. We have shown previously that the symmetrical
flat dipole operates with the widest bandwidth at its third resonance [4], when
a model equivalent to several lossy and uncoupled transmission lines, operating
in a quasi-transverse electromagnetic mode, is taken into account.

Radiation impedance and bandwidth: Fig. 7.2 shows the flat dipole parallel to,
and at a distance H from, a reflector plane. A dielectric sheet is placed between
the radiating source and the reflector, so that it can be realised by means of
printed circuits. The stripline feeding network is shown. The coupling between
the radiating element and the feeding stripline is realised by means of a quarter-
wavelength open stripline. The two large plates are fed in opposite phase from
a gap AB. Thus the symmetrical plane perpendicular to the xx' axis and to the
reflector plane is at zero potential. It is then possible, but not necessary, to
short-circuit middle points of the folded arm at the reflector plane without
disturbance to the electromagnetics properties.
Fig. 7.3 Wide-bandwidth dual polarised microstrip antenna
transmission lines. To explainctheoretically the wider bandwidth, in conformity
with experimental work, it was necessary to introduce coupling between the two
transmission lines [5]; ie., across the two radiating slots (Fig. 7.2). The input
impedance Z, relative to the middle AB of the gap is due to the transformation
of two impedances 2Z', which are the radiation impedances of the two radiating

358 Wideband flat dipole and short-circuit microstrip
slots. This transformation is carried out along a non-radiating slot line of total
length Wand characteristic resistance R,.Each half slot, having a radiation
impedance equal to Z',, can be considered as two coupled transmission lines of
length I with characteristic impedances Z and Z', and C,, being the coupling
capacitance per unit of length. The length 1 is given by the expression
Wideband flat dipole and short-circuit microstrip 359
iMutua1 impedance and coupling: Fig. 7.5 shows the configuration of two
coupled f l a ~dipoles in a parallel position (or in the H-plane) separated by a
distance D and located above a perfectly conducting ground plane at a height
H. The mutual coupling may, in principle, be due to either guided waves, or

with

Ah, = 0.41 H
+
(E, 0,3)(s/H + 0.26)
(E, - 0.26)(s/H + 0.81)
where s is the width of the slot of total length 2h0,Ah, the increase in length due
to an end effect. The characteristic resistance Z and Z' are those of microstrip-
lines of equal H a n d thickness and widths, respectively, of W / 2 and (W-W)/2
- s. The conducting currents are strictly located on the edges of every slot.
Radiation is taken into account by introducing, for equivalent coupled lines,
attenuation constants per unit length cc and cc' given by the following equations:

where f is the frequency and tanh is the hyperbolic tangent.

The two coupled transmission lines are divided into N equal four-port sec- Fig. 7.4 Theoretical input admittance Y, = IIZ,and experimental points of a directional flat
tions. When the boundary conditions are applied, the potential and the electric- folded dipole (Y, = G, + jB,)
current distributions, and the input impedance Z, are deduced.
Fig. 7.4 shows the theoretical input admittance l/Z,, obtained with N = 250,
and experimental points in a wide frequency band with the following parameters
(in millimetres):

We deduced:

' I
substrate
reflector plane
Fig. 7.5 Configuration of two flat folded dipoles in H-plane
Owing to the coupling capacitance across the two slots, a fourth resonance
(1 1.2 GHz) was observed not far from the third one (10.3 GHz), which explains
the increased bandwidth. From another model [6] a bandwidth of 16% has been space waves, or both. In our theory, guided waves will be neglected. Then the
measured between 11.25 and 13.2 GHz for a VSWR lower than 2. The radiation mutal impedance Z , , and the coupling coefficient C(dB) are given by the
resistance is always located at about lOOR whatever the model parameters. following expressions [7,8]:
360 Wideband flat dipole and short-circuit microstrip Wideband flat dipole a n d short-circuit microstrip 361
Fig. 7.6 shows, for h, = 4.35mm, H = 3.2mm, W = 3.8 mm, s = 0.9 mm,
E, = 2.2 and f = 9.5 GHz, the theoretical mutual impedance and coupling
factor in terms of their distance D, together with some experimental points. It
was seen that, for microstrip patches, the E-plane coupling always exceeds that
of the H-plane coupling. It is the opposite for the flat dipole. Nevertheless the
coupling levels encountered, which are more critical for the H-plane (parallel
where R, is the normalisation resistance and E, is the equivalent relative dielec- position as in Fig. 7.5), should not involve any difficulty of array design. The
tric constant with k, = 2n/& and K = k,&. main coupling effect is via space waves [7, 81. In effect, for the arrays which are
described in Section 7.1.2, the ratio of the mean distance between two adjacent
flat dipoles to the wavelength in free space is greater than 0.8. Then the coupling
factor is always lower than - 25 dB.

Radiatedfields andpolarisation: We assume the flat-dipole reflector plane to

be infinite. The flat-dipole moment vector is parallel to the xx'-axis (Fig. 7.2).
By using electric l a n d polarization I, current distributions and applying the
volume equivalence theorem we can calculate the far field radiated by the
antenna in the E-plane from the following expressions [5]:

I,(x) = ~O(E +
- E,) W(COS~X tgkl sinklxl)V, i
? (7.8)
1x1
2V, is the potential which is applied between the two gap edges.
In the E-plane and after integration the co-polar pattern is deduced from eqn.
7.7:
s i n k (-i n ) sink,l(& + sin$)
E, = cos$ sin(k,Hcos4) +
kOl(J~,- sin$) k,l(&, sin4)+
2 kol
sin - (& - sin$) sin - (&
2
+ sin$)
- cots(k0 bc') + ko 1
- (&
2
+ sin$)
In the H-plane, the polarisation current distribution due to the substrate has no
effect, and the normalised far field radiated is given by the equation:
sin(k, Hcos0) W + s .
E" = cos [k, -s m ~ ]
sin k o H 2
Fig. 7.6 Mutual impedance Z,, = R,, + j XI, and coupling factor in terms of D l i , at
f = 9.5GHz in H-plane In eqns 7.9 and 7.10 no effect of the mutal coupling are taken into account, since
the equations apply to an isolated source and an infinite reflector plane.
The validity condition of eqns. 7.4 and 7.5 is:
7.1.1.2 Synthesis in quasi-TEM approximation: A theoretical model equiv-
alent to several lossy coupled transmission lines was used in section 7.1.1.1 to
362 Wideband flat dipole and short-circuit microstrip Wideband flat dipole and short- circuit microstrip 363

explain the bandwidth and the radiation resistance of the flat dipole. Conversely
it is useful to find the geometrical parameters of such an antenna and its
radiation resistance when the frequency band is known. The number of par-
ameters is reduced, and to obtain an attractive bandwidth their values are
bounded. From Fig. 7.2 we take [16]: W/L0 = 2(W/, + sll,), g/L, = 0.014,

Fig. 7.7 Optimal resistance R,,,

E, = 2.2. As 2h, ,< 21 ,< 2h we take 21 = 2h, with I given by eqn. 7.1. Then the
variable parameters are H/&, W/& and s/&, and using Figures. 7.7 - 7.9, we can
define the size of the whole antenna and its optimal resistance R,, when the
frequency band is given. In Fig. 7.8 the optimised bandwidth (B %) is due to
the input impedance Z,, related to R,, for a VSWR below 2; an example is
given. The frequency band is equal to 11.3 - 13.2GHz (B = 15.5%). With a
normalised thickness H = 3.2mm ( E , = 2.2), we obtain, for the mean fre-
quency (1, = 24.5mm), H/1, = 0.13. In Figs 7.8 and 7.9 we deduce, for
sll, = 0.02, that W/& = 0.1 (then W/I, = 0.24) and I/& = 0.204. Fromeqn. Fig. 7.8 Optimised bandwidth B(%) of input impedance 2, related to R,, (Fig. 7.7) for a
7.1 we deduce that Aholio = 0.029 and 2ho/l, = 0.33. The final antenna dimen- VSWR < 2
sions are: W = 2.45mm, W = 5.9mm, 2h = IOmm, s = 0.5mm, 2h, = W'/& = 2(W/L0 + s / i o ) ,g/&, = 0.014 8, = 2.21
8 mm, H = 3.2 mm, g = 0.34mm. For these parameters and from Fig. 7.7 we - - - flat dipole is acting at the 3rd and 4th resonances
obtain R,, = 140Q. Synthesis resulting (Fig. 7.7 and 7.8) indicate that the -flat dipole is acting at the 3rd resonance
 364 Wideband flat dipole and short-circuit microstrip
antenna bandwidth increases and radiation resistance decreases as its transverse
area (in relation to the square wavelength) and coupling between the two
equivalent radiating lines of each slot increase. Elsewhere, it was observed that
the flat dipole performs best when a substrate of low dielectric constant is used
[I 71, this is a general property [I 81. For these reasons we obtained, using e, = 1,
H = 5 mm, W' = 9.2mm, W = 7mm, g = 1,2mm, 2h = 9,6mm, 2h, =
7mm, s = 0,6mm, the correct behaviour for an experimental model acting
within one actave, i.e. between 8 and 16GHz [19, 251.
Wideband flat dipole and short-circuit microstrip 365
meshes. Each mesh is composed of four segments which are crossed cylindrical
conducting wires. The influence of the conducting-wire diameter on the radia-
tion impedance and bandwidth is considerable. The diameter chosen is equal to
one quarter of the mesh dimension. The scattering problem is reduced to a
flat-shaped bi-dimensional structure. In Fig. 7.11 we present the complete
j,, cut AB
IA

/ '
ref lector
plane

Fig. 7.10 Very large bandwidth flat folded dipole configuration

Fig. 7.9 Half-slot effective length I/>.,

7.1.1.3 Moments-method analysis: Recently we have studied a wide-band-

width flat dipole, fed through the gap by means of a Lecher line which is located
on a thin printed circuit parallel to a reflector plane and isolated by a sheet of
air (Fig. 7.10). The general scattering problem of an arbitrary shaped tri-dimen-
sional antenna solved by the moments method and the finite-difference ap-
Fig. 7.11 Average surface-current vector distribution
proach applied to integral equations has explained the antenna behaviour. The
wire-grid model is suitable for expressing the boundary conditions along the
distribution, which corresponds to the average surface current vector for each
various edges of the antenna, which are, in fact, the boundary conditions on the
mesh. The length vector is proportional to the current amplitude. In Fig. 7.12
thin wire surfaces. The antenna is equivalent to an array of conducting square
we present the theoretical and experimental input impedance related to the
366 Wideband flat dipole and short-circuit microstrip Wideband flat dipole and short-circuit microstrip 367

middle of the gap in terms of frequency (GHz). The impedance is normalised to

IOOQ, which is the Lecher-line characteristic resistance.: Table 7.2 gives the
parameters used in Fig. 7.12 (in mm)..

Fig. 7.12 Theoretical ( 1 ) and experimental (2) input impedance of the large bandwidth flat
folded dipole in terms of frequency (GHz)
1 = 100R

7.1.2 Array designs. Losses and eficiencies

A number of flat arrays with several hundred or more flat dipoles have been
designed and constructed. In every array each source radiates through a window
cut in one of the two metallic shields of the stripline feeding network (Fig. 7.2).

7.1.2.1 Large gain: Fig. 7.13 shows an array of 1024 radiating sources
operating between 11.7 and 12.4GHz with a measured isotropic linear maxim-
um gain of 37dB [20]. This high-gain array has been designed to receive
radio-broadcasting signals sent out by geostationary satellites. It is composed of
1024 (32 x 32) flat dipoles (section 7.1.1.1) fed by a stripline network (Fig.
7.14). Like the elementary source, the array is constructed using two large
compressed printed-circuit sheets with no direct connection between the feed
network and the radiating sources. The symmetrical feed network comprises
unequal stripline power splitters joined by stripline transmission lines, the
 368 Wideband flat dipole a n d short-circuit microstrip
characteristic impedance of which is 7 5 0 . The distance between two adjacent
sources is 0.89 i,,,where >., is the wavelength in air a t the mean frequency
( 12. I GHz). The square array of a 0.5 m' area was manufactured using two
printed-circuit sheets ( E , = 2.1 7) I .6 mm thick. Circular polarisation is
produced by a polariser embedded in a radome. Fig. 7.16 shows an exanlple of
the measured patterns in one diagonal plane at 12.1 GHz. Between 11.7 and
13.5GHz the experimental maximum linear isotropic gain was equal to
36.9 F 0.3 dB, and the efficiency is better than 48% between I 1.7 and 12.4 GHz.
Wideband flat dipole and short-circuit microstrip 369
0.85 x 0-40 x 0 4 5 m . The linear isotropic maximum measured gain a t
5.25 GHz is 26.5 dB. the efficiency is about 40% and the measured sidelobe level
is lower ~ h a n- 29 dB for a theoretical level of - 39 dB. Fig. 7.17 shows an
example of radiation patterns in E- and H-planes at 5.25GHz for an array
covered with a radome. The second array [23] (Fig. 7.18) is incorporated in a

Fig. 7.13 High-gainarray (11.7- 12.4GHIj offlatfoldeddipole.

The feeding arrangement is shown in Fig. 7.14 and the feeder dielectric a n d Fig. 7.14 High-gain array (7 1.7 - 12.4GHz): feeding arrangements
metallic losses are equal to 2.5 dB. The cross-polarisation level o n the axis is
lower than -30dB, and, together with the polariser, the measured array Table 7.3 Measured values for the array in Fig. 7.18
ellipticity ratio along the principal axis is better than I.5dB over the frequency 9.6 9.9 10.1
f, G H z 9.1 9.3
range. Fig. 7.15 shows the radiation patterns in E- and H-planes measured a t
12.1 GHz. Max. gain, d b 30.9 26.9 31.9 31.3 31.4
3dBbeamwidth,deg(H-plane) 2'40' 2'35' 2'40' 2'30' 2'30'
7.1.2.2 Low sidelohe level: Two passive arrays with flat dipoles showed a First sidelobe, dB (H-plane) < - 28 < -28 < - 36 < -30 < -29
sidelobe level lower than - 30dB. The first array [21, 221 operates between 5.25 The efficiency at 9.6 GHz is equal to 33.3% (or - 4.78 dB). The cross-polar~sationlevel is lower than
and 5.45 GHz in a system used to detect natural resources, and is carried o n a n - 35 dB.
aircraft. It is composed of 128 symmetrical flat dipoles which are fed by lines a n d
system used o n ground-surveillance radars. It is composed of 512 flat dipoles,
splitters realised using stripline technique by means of two printed-circuit sheets,
it uses stripline technology and features low sidelobe radiation: it operates
as has been shown for the high-gain array (Fig. 7.14); the dimensions a r e
between 9.4 and 10.1 GHz. The distance between two adjacent sources equals
 370 Wideband flat dipole and short-circuit microstrip Wideband flat dipole and short-circuit microstrip 371

Fig. 7.16 Large-gain array (see Figs. 7.13 and 7.14) (Reference 2 0 )
Pattern in a diagonal plane at 12.1 GHz

- e (DEGREES)
Fig. 7.17 Low-side lobe-level array.
A e (DEGREES)

Patterns in E- and H-planes at 5.25 GHz

Fig. 7.15 Patterns at 12.1 GHz. Large gain array (see Figs. 7.13 and 7.14) (From Reference
 372 Wideband flat dipole and short-circuit microstrip
0.9 E., in the H-plane and 0.8 E,, in the E-plane. Some measured values are given
in Table 7.3.
7.1.2.3 Vuriable r/irc.ctiiities: A flat array with four fixed beams of different
directivities and gains is considered [24]. This array supports 256 flat dipoles
separated by a distance of 0.85 /., together with 60 electronic switches, integrated
in a stripline structure (Fig. 7.19). The global efficiency is about 50%. For each
of the four states we can form sum (z)and difference (A) patterns in the two
Fig. 7.18 Low-side lobe-level array of flat folded dipoles (9.4 - 10.1 GHz)
orthogonal principal planes of the array. which acts between 11.7 and 12.4GHz.
Fig. 7.20 shows the ratio A / x for each of the four states in terms of the deviation
angle 8. The various beamwidths obtained at the mean frequency are 4"3', 7'7',
12'6' and 27', and the cross-polarisation level is always lower than - 25 dB. The
efficiency is given by the formula q% = 100 1'G,w/4ns, where s is the geometri-
cal area (0.12 m2)and G,, is the maximum isotropic linear gain. With a measured
value GM = 31 dB. we obtained y = 50%. For each of the four states the mean
switching losses are about 1.6dB.
7.1.2.4 Very large bunrtwidrh: This section discusses broad-angular-coverage
and wide-bandwidth antennas, which are increasingly used, for example in
Wideband flat dipole and short-circuit microstrip 373
airborne communication systems, electronic warfare systems or ECM. We show
[25]cylindrically shaped radiating elements working in octave bandwidth (Fig.
7.21). In azimuth, the theoretical angular-sector coverage can be equal to 360'.
In the different meridian planes or site planes, the beamwidth and sidelobe level
may be of any proportion. The antnenna consists of radiating elements arranged
in vertical arrays photo-etched on a printed circuit which is wrapped around a
cylindrical dielectric lens of 2a diameter. Each radiating element is a flat dipole,
used as a folded slot dipole symmetrically fed. When the substrate has air as the
Fig. 7.19 Multiple-beam flat array ( 1 1.7 - 12.4GHIJ
medium and if the thickness is increased, the flat dipole acts over wide band-
width (sction 7.1.1.3). Thus, when it is fed symmetrically with a Lecher line, the
operating range covers one octave. Each array, which is aligned along a genera-
ting line of the cylindrical lens, can be considered as a separate channel. In a
plane perpendicular to the axis the antenna acts as a Luneberg lens. In effect,
the n constant-index-of-refraction surfaces are cylinders of radius r such that
n ( r ) = [2 - ( r / ~ ) * ] "The
~ . dielectric lens was machined from a cylindrical Teflon
rod. It has longitudinal grooves which are uniformly distributed around the lens
axis and which are small compared with the vacuum wavelength [26].
In the meridian section of the lens, which contains one array the optical path
is deduced, by means of integration, from the continuous refraction law. Apply-
ing an asymptotic development of the Kottler formula, the radiated far field is
calculated from the electromagnetic field distribution along the lens outside
surface, as has been done previously, but for a small bandwidth using short-
circuited flat dipoles at quarter-wave resonance [27, 281.
The antenna is composed of 64 folded flat dipoles arranged in eight vertical
 374 Wideband flat dipole and short-circuit microstrip Wideband flat dipole and short-circuit microstrip

arrays, and it acts in the 8 - 16 GHz range. The diameter of the cylindrical lens
is equal to 86mm. The beamwidth in the frequency range lies between 24" and
10" in the azimuth plane, and between IS0 and 7" in the site planes. Whatever
the azimuth direction, the minimum isotropic linear gain which can be expected
over an angular coverage of 1lo0, and in the 8 - 16 GHz frequency range, is
18 dB. The maximum isotropic linear measured gain G,(dB) is compared with
the theoretical maximum directivity D,(dB) in Fig. 7.22. The impedance match-
ing of each array is given in Fig. 7.23. Fig. 7.24 shows the measured ratio A / C
between difference (A) and sum (I)
patterns expressed on a linear scale.

Fig. wad-angular-coverage and wide-bandwidth antenna

Fig. 7.20 Ratio A / C of sum (CI and difference (A) patterns for the four states (F, to F,) at
12.7 GHz (see Fig. 7.19)

7.2 Short-circuit microstrip patches and arrays

7.2.1 Elementary source

Self impedance and bandwidth: Fig. 7.25 shows the half short-circuited patch
configuration. It looks like a half short-circuited flat dipole acting at a quarter-
wave resonance. The relation between radiation conductance G, and Go can be
determined from [29]

Fig. 7.22 Maximum measured gain GM and maximum directivity DM

 376 Wideband flat dipole and short-circuit microstrip Wideband flat dipole and short-circuit microstrip 377
Eqn. 7.11 expresses the development of the input impedance of a lossy short-
circuited line whose length is assumed equal to H' = H + h. The losses com- r r
prise the radiation resistance of the patch. The main problem lies in determining
the line attenuation constant. In order to do this, the radiation resistance of the
half short-circuited patch, parallel to the perfect reflector plane, is identified
with that of a loop constituted by the half short-circuited dipole and its electrical -20 -10 0 10 20
degrees degrees
image in relation to the reflector plane [3].
When finite conductivity a and lossy dielectric substrate are taken into
account, as has been done for the opened flat dipole acting at half-wave
resonance (Patch) [30], the input conductance G, is given as follows: F= 8 GHz
0.05fdegree

o$ ; - degrees

Fig. 7.24 Measured ratio A / x between difference ( A ) and sum (I)

patterns in terms of
frequency. Broad-angular-coverage and large bandwidth antenna

shorting pins or
electric wall

Fig. 7.23 Broad-angular-coverage and wide-bandwidth antenna

Impedance matching of each array

where d, (nfha)-'I2 is the skin depth, 6 is the dielectric loss angle, 1, =

and Go = ( ~ ~ / p ~ ) " ~
The bandwidth B% related to VSWR is given by

-
Lastly, the radiating source efficiency at resonance is given by

_I
sub;trate (E,)
The quarter-wave resonance condiction is h + H 1. 114. Fig. 7.25 Half short-circuitedmicrostrip antenna configuration
378 Wideband flat dipole and short-circuit microstrip Wideband flat dipole and short-circuit microstrip 379
Mutual impedance and coupling: Explicit formulas are presented for the Feedthrough is used to excite each radiating element from the output of the
mutual impedance and coupling between two parallel or co-linear short- associated phase shifter by means of a short coaxial line. This 3 bit digital phase
circuited flat dipoles at resonance in air medium in terms of three dimensions shifter was recently described [34]. A total of only eight PIN diodes is required
with respect to wavelength [31]. Experimental and theoretical results are in good
agreement and show that these small microstrip antennas are particularly well
uncoupled, and therefore suitable for incorporation in a phased array with a
steering beam inside a large angular sector. Thus, for the following values (Fig.
7.25):

the coupling factor in t h e E- and H-planes is lower than - 20dB between

short-circuited flat dipoles separated by a distance 0.5 i,.

Radiatedfields andpolarisation: The far field radiated by this source has been
calculated [32j. With the w m e notation as in Fig. 7.25 we ~btaincd:
The Normalised electric field in the E-plane (4 = 0)

The normalised electric field in the H-plane (0 = n/2)

7.2.2 Array designs

These short-circuited patches acting at a quarter-wave resonance (section 7.2.1)
have been used in several arrays.
7.2.2.1 Phased array with steering beam: Eflect of mutual coupling: This
section concerns a Ku-band phased array acting in a large angular [33] at
around 15 GHz. The array has 64 short-circuited patches in air medium, located
on a 8 x 8 square lattice. The element spacing is lOmm (0.5 A,), such that the
coupling coefficient in the E- and H-planes is lower than - 20 dB. The measured
results on each radiating element at around 15 GHz are as follows:
Bandwidth for a VSWR < 2: 17%
+
Pattern in E-plane: 1dB in an angular sector of 160"
Beamwidth in H-plane at 3 dB:80
Fig. 7.26 3bit digitalphase shifter with eight PIN diodes in Ku band
for the whole phase shifter (Fig. 7.26) instead of ten for a conventional type. It
is arranged in such a way that no series capacitance is necessary to separate the
different bias [35]. So only three wires are used to apply the DC bias to the eight
PIN diodes. For each of the eight states, the phase-shifter insertion losses is
equal to 3 i I dB in a bandwidth of 15% at about ISGHz. The 64 phase
shifters are photo-etched on four fused-quartz substrate plates each of
5.08 x 5.08 cm (Fig. 7.27). So for every deflected beam we can form sum and
difference patterns. The feeding structure, which is composed of splitters,
380 Wideband flat dipole and short-c~rcuitmicrostrip Wideband flat dipole and short-circuit microstrip 381
branch lines, corporate feeds and DC bias. is also photoetched according to a
microstrip technique. Four plugs are connected by pliant and thin conductors
to the etched wires for DC bias feeding. The square feeding R F network is
divided into four parts. The adjacents parts are deduced from each other by
rotation symmetry (RS) so that the four outputs 1 , 2 , 3 . 4 are located on the four
array sides (Figs. 7.29 and 7.27). This rotation symmetry is only related to the
four parts of the square feed R F network, and does not concern the short-

Fig. 7.28 Phased array with steering beam i n Ku band; 6 4 radiating sources

Fig. 7.27 Phased array with steering beam i n Ku band: 6 4 3 b i t digital phase shifters and
feeding arrangements

circuited radiating dipoles which are linearly polarised (Fig. 7.28). The parasitic
signals at the input of the four quadrants, which are due to the radiating-dipole
Fig. 7.29 Square feeding network (From Reference 3 6 )
aerial alternately coupling in the E o r H planes, are out of phase. This concept S: output towards source ?:source electrical moment
382 Wideband flat dipole and short- circuit microstrip Wideband flat dipole and short-circuit microstrip 383
[36] has advantages as compared with an axial-symmetry network (AS). For 8 (degrees)
instance, in Fig. 7.30 we show the theoretical sum patterns in the H-plane -60 -3 0 0 30 60
deflected by 45' and related to (RS) and (AS). Furthermore, because of the
compensation of aerial mutual coupling, the risk of scan blindness is reduced
when the (RS) concept is used. Fig. 7.31 shows measured sum gains in the E, H
or diagonal planes and Fig. 7.32 shows the ratio A / C between the difference and
sum patterns for different deflection angles a. The measured efficiency q% is

8 (degrees)

Fig. 7.30 Deflected sum patterns in H-plane at f = 15GHz (From reference 36)
-(RS)
- - - (AS)

deduced from the following expressions:

Fig. 7.31 Measured sum gains G ( d 8 ) for different deflection angles or (degrees)
G , is the measured maximum isotropic linear gain (dB) and A is the antenna
geometrical area.
In conclusion, major advances have been made in lowering side-lobe level
[36], in the reduction of phase-shifter complexity [35] as well as in the manufac-
turing procedures developed for building the array.

7.2.2.2 Omnidirectional radiation array for a mobile telecommunication sta-

tion: This section discusses a very flat antenna with circular or elliptical pol-
arised directional radiation over a large conical angular sector, and linear
omnidirectional radiation in the reflector plane [37,38]. It can be advantageous-
ly used for communication between mobile stations. The antenna comprises
four short-circuited half dipoles (section 7.2.1) acting at quarter-wavelength
resonance and fed in phase quadrature (Fig. 7.33 a and b). The ends of the four
short-circuited half dipoles (1, 2, 3, 4) are screwed (7) on the edges of a cavity
Fig. 7.32 measured ratio
deflection angles a
AIC between difference (A) and sum (u patterns for different

1 : E-plane
7. U-nlann
384 Wideband flat dipole and short- circuit rnicrostrip Wideband flat dipole and short-circuit rnicrostrip
b

tw -I

Fig. 7.33 Omnidirectional radiation array configuration

1 I

Fig. 7.34 Linear directivity in E-plane (f = 1.2GHz)

11 ):
\ , GTD
( 2 ) : GO
( 3 ) : Measurements ( m m ) 2h = 115/W = 5 0 / H = 7,5/e = 1.6
386 Wideband flat dipole and short-circuit microstrip Wideband flat dipole and short-circuit microstrip 387
in air medium (6). A printed circuit (8) of thickness e supports, on one face, the 7.2.2.3 Log-periodic array: This antennas is a microstrip travelling-wave
four half dipoles and, on the other face, the feed microstrip lines (lo), with antenna of large bandwidth [19,41,42,43]. The log-periodic array is composed
quarter-wave transformers (9) which are connected to coaxial lines (1 I). The of flat short-circuited half dipoles acting at quarter-wave resonance and fed in
four quarter-wave-transformer ends are joined and short-circuited to the bot- series by means of a high-characteristic-resistance microstrip line, R, (Fig. 7.38).
tom of the cavity by 12. Another feed concept is possible (Fig. 7.33 c) by which
the feed-point location is chosen to match the radiation resistance of the coaxial
line.
With the following dimensions (in mm): 2h = 115, W = 50, H = 7.5,
e = 1.6,2 d = 300 (2 d is the dimension of the square reflector); the bandwidth
is equal to 4.35% for mismatch losses lower than 0.5 dB, while the theoretical
bandwidth given by eqn. 7.13, is equal to 4.3% for a VSWR < 2 and E, = 1.
The mean frequency is equal to 1.2 GHz.
Fig., 7.34 shows the linear directivity in the E-plane. The E-plane is the
symmetrical plane related to the two opposite short-circuited half dipoles (Fig.
7.33 b ar cf which Ire !inear!y po!arised. Curve (!) gives the directivity when
GTD on the reflector edges is taken into account. Fig. 7.35 shows the linear
isotropic gain measured for different site angles 0, with horizontally or vertically half short-circuited
polarised waves. We can see that the antenna is circularly polarised along its axis
(0, = 0), which is perpendicular to its plane, and quasi-linearly polarised in all
directions (8, = 90") of the reflector plane. cut AB
Another configuration is shown in Fig. 7.36. The antenna, which is vertically
polarised with omnidirectional radiation, is composed of four short-circuited Fig. 7.36 Vertically polarised antenna with omnidirectional radiation

half-dipoles acting at quarter-wavelength resonance in air medium and fed in

phase [39]. To match the radiation resistance of each half-dipole, a correct
feed-point location along the symmetrical axis is found [40] and soldered to the
central conductor of a coaxial line. The four coaxial lines are fed in parallel by
means of a standard divider. A model was calculated and tested with the
following parameters (in mm): 6, = 1, I = 40.5, W = 52, H = 2.6, h = 34.5,
2 d = 400. The measured resonance frequency was found to be 1.90GHz while
the theoretical one (f, given by eqn. 7.17) is equal to 1.93 GHz.
f;' = [h
4ph'2~A'2 + H + 0.72H(W/H + 0.26)(W/H + 0.81)-'1
(7.17)
The measured bandwidth for a VSWR of less than 2 is equal to 4.7% while the
theoretical bandwidth, given by eqn. 7.13, is 4.62%. Fig. 7.37 shows the mea-
sured E-plane, co-polar and cross-polar gain patterns, and the theoretical one
when diffraction corrections, obtained with GTD applied to the square reflector
edges, are taken into account. Maximum isotropic linear directivity of 6.46dB 8 (degrees)
occurs for 0 = 37", while the measured maximum gain of the antenna with its
Fig. 7.37 Measured E-plane directivity pattern (copolar and cross-polar)
divider is equal to 54dB. For a constant angle 0 around the antenna, the 4 = n / 2 . O: variable
measured radiated far field is practically constant, with variations of f l dB E, = 1 , I = 40.5mm, W = 5 2 mm, H = 2.6 mmlh = 34.5mrn, 2d = 400 mm

when 0 < 0 < 90'. The bandwidth can be improved when the height H is
increased, and the maximum site directivity angle may be altered by when Each half dipole has a large E-plane pattern and weak coupling, especially in the
adjusted the I parameter. This model is particularly thin since H/1, is equal to parallel position (H-plane), and so the log-periodic array can be used for a
0.017. progressive wave, when a squinted beam in the H-plane and large pattern
388 Wideband flat dipole and short-circuit microstrip
-
coverage in the E-plane are suitable. Fig. 7.39 shows the geometrical parameters
W,,,H,,.and /I,, for the half-dipole (D,,)of the nth order. acting at a quarter-wave
resonance in air medium with /I,, + H,, (4hv%)-', where f;, is the re-
sonance frequency. The various radiating parts are located on one metallic face
Fig. 7.38 Log-periodic array of short-circuited patches
( 1 ) : Printed circuits and complete array
(2): Cross-sect~on
Wideband flat dipole and short-circuit rnicrostrip 389
of a printed circuit (I). They are fed in series across a gap (G,,) by means of a
stripline (M) etched on a second printed-circuit sheet (J). The classical log-
periodic parameters are:
,
I,, being the distance along the stripline between the gaps (G,,) and (G, + ,) of
+
(D,,) and (D,,, ,) separated by a distance 4+,. To limit the variation of the
VSWR in the frequency band, it is necessary to increase the expansion para-
meter T. The distance I,, must be lower than As/2 to compensate for the various
reflections appearing at each gap. The following parameters have been adopted:
r = 0.95, a = 0.4, 4,/A, = 0.2, H,,/A, = 0.1, I,,/i, = 0.32, W,/1, = 0.166,
R,,= 180R. With these parameters, a 50" 3 dB beamwidth and 45" deflected
beam are calculated. With 50 radiating sources the antenna dimensions are
1.1 x 0.1 x 0.07m, which corresponds to a theoretical frequency band of 0.9
- 6GHz bounded by a VSWR < 2 and an efficiency of 95%.
LA cut AB
Fig. 7.39 Geometrical parameters of the log-periodic structure
The antenna performs well between 0.75 and 4.5 GHz. Fig. 7.40 shows the
measured H-plane radiation patterns at three frequencies. The average values
are 42" for the deflection angle and 50" for the 3dB beamwidth over all the
experimental bandwidth.
 390 Wideband flat dipole and short-circuit rnicrostrip Wideband flat dipole and short-circuit microstrip 391
7.3 References

I DUBOST, G., and ZISLER, S.: 'Antennes a large bande. Theories et applications' (Masson.
Paris, 1976)
2 DUBOST, G., and VINATIER, C.: 'Doublet replie symetrique en plaques fonctionnant a t*s
hautes frequences et a large bande'. Brevet Europeen 0044779 BI, 13 Nov. 1985; USA, 284702,
20 July 1981; Japan, 114526, 23 July 1981
3 DUBOST, G.: 'Flat radiating dipoles and applications to arrays' (John Wiley, 1981)
4 DUBOST, G. BEAUQUET, G., and VINATIER, C.: 'Theoretical radiation admittance of a
large bandwidth flat symmetrical folded dipole', Elec~ron.L e ~ t 1984.20,
, pp. 252-253
5 DUBOST, G., and RABBAA, A.: 'Analysis of a slot microstrip antenna', IEEE Trans., 1986,
AP-34, pp. 155- I63
6 DUBOST, G.. and VINATIER, C.: 'RCseau dedoublets replies symetriques en plaques, a large
bande autour de 12GHz'. L'Onde Electrique, 1981, 61, pp. 34-41
7 DUBOST. G., and GUEHO, S.: 'Impedance mutuelle et couplage entre deux doublets repli6s
plans paralleles en fonction de leur ecartement', CR Acad. Sci. Paris, 1985, 301, ser. 11, pp.
79-82
g DUBOST, G.: 'Mutual ooupling between flat folded dipoles in terms of frequency'. Int.
Symposium Antennas and EM Theory, Beijing, China, Aug. 1985, pp. 706-711
9 DUBOST, G., and FRIN, R.: 'Antenne plaque a double polarisation croisee'. (Brevet 36 05
0 . degrees 990, 23 April 1986
10 DUBOST. G.: 'Large bandwidth dual polarized multilayer microstrip antenna'. AP-S Intern.
(A) Symp., Philadelphia. USA, June 1986, pp. 455-458
11 PALANISAMY, V., and GARG, R.: 'Analysis of circularly polarized square ring and crossed-
strip microstrip antennas', I E E E Trans., 1986, AP-34, pp. 1340-1345
12 DUBOST, G.: 'Broadband circularly polarized flat antenna'. Int. Symp. Ant. &Prop., Japan,
1978, pp. 89-92
13 CHIBA T., et ai.: 'Suppression of higher modes and cross polarized component for microstrip
antennas'. IEEE AP-S, 1982, p. 285
14 HUANG, J.: 'CP microstrip array with wide axial ratio bandwidth and single feed LP
elements*. IEEE AP-S, 1985, pp. 705-708
15 DUBOST, G., FRIN, R.: 'Antenne a double polarisation associee des directeurs'. Brevet
Europe, USA and Japan, April 1987
16 DUBOST, G., and RABBAA, A,: 'Synthbe de I'antenne plaque a double fente', L'Onde
Electrique, 1987, 67, pp. 72-79
17 DUBOST, G., and RABBAA, A.: 'Substrate influence on flat folded dipole bandwidth',
Electron. Lett., 1985, 21, pp. 426-427
18 JAMES, J. R., HALL, P. S., and WOOD, C.: 'Microstrip Antenna' (Peter Peregrinus 1981)
IEE Electromaghehe Waves Series 12.
19 DUBOST, G.: 'Comparison between flat radiating source bandwidth'. IEE Coloquim on
Antenna bandwidth extension techniques, London, 28 Oct. 1985.
20 DUBOST, G., and VINATIER, C.: 'Large bandwidth and high grain array of flat folded
dipoles acting at 12GHz' 3rd Int. Conf. on Ant. and Ptopag. ICAP, Norwich, April 1983
21 DUBOST, G.: 'Improvements in printed-circuit radiating sources and arrays'. IEE Colloquim
on receiving antennas for satellite broadcasting', London, April 1984
8 . degrees 22 DUBOST, G., BEGUIN, D., CHAPUIS, E., AURIOL, A.: 'Reseau plat a grand gain et a
faibles lobes secondaires a 5 GHz'. Conference Internationale sur le Radar, Paris, May 1984
23 MARCHAND, M.: 'Antenne plane a reseau en bande X', Rev. Tech. Thornson-CSF (France),
Fig. 7.40 A Measured H-plane directivity patterns a t 4.5 GHz 1985, 17, pp. 83-109
B Measured H-plane directivity patterns 24 DUBOST, G., POTIER, P.: 'Rkseau plat a commutation tlectronique de faisceaux dans la
-.-.- Copolar at 0 4 GHz
bande des IZGHz', L'Onde elecwique, 1985, 65, pp. 56-61
-Copolar at 2.9GHz 25 DUBOST, G.. and NICOLAS M.: 'A braod angular coverage and large bandwidth antenna'.
---- Cross-polar at 2.9 GHz
17th European Microwaves Conf., Rome, Sept. 1987
392 Wideband flat dipole and short-circuit microstrip
26 DUBOST G.. NICOLAS, M., VALLEE, P.: 'Antenne a pas de balayage reduit dans un large
secteur angula~re'.Brevet 85 07 348, 15 May 1985
27 DUBOST, G.: 'Antenne symetrie de revoloution associant une lentille dielectrique a une
source plaque'. IEEE Symposium AP-S, Vancouver, Canada, June 1985, pp. 587-590
28 DUBOST, G.: 'Flat linear radiating array applied on a cylindrical lens'. Melecon'85 Mediter-
ranean Electrotechn~calConference, Madrld, Oct. 1985, pp. 215-218
29 DUBOST, G.: 'Short-or open-circuited dipole parallel to perfect reflector plane and embedded
in substrate and acting at resonance', Electron. Lett., 1981, 17, pp. 914-916
30 DUBOST. G.: 'Transmission-line model analysis of a lossy rectangular microstrip patch'
Electron. Lett. 1982, 18, pp. 282-282
31 DUBOST. G.. and RABBAA, A.: 'Mutual impedance between two short-circuited flat re-
sonant dipoles'. IEEE Trans., 1981. AP-29, pp. 668-671
32 DUBOST. G.: 'Far field radiated by short-circuited microstrip antenna acting at a quarter
wavelength resonance' Electron. Lett. 1983, 19, pp. 737-739
33 DUBOST, G., GUEHO, S., and BEGUIN, D.: 'Ku band phased array in a large angular
sector', 5th Int. Conf. on Antennas and Propag., ICAP 87, 1987, University of York
34 DUBOST, G., and GUEHO, S.: 'A 3 bits digital phase shifter in Ku band for microstrip
- ,.**.d
y..,,,, buAuquu~a --
,".,,..'.-., Q , h ,-,,I,.",.:- vu x
'r,:"v ~ --.., a vr ~ r ~ r ~ :...
r ~ r u n c A:-..
u r u c d r r unug.
35 DUBOST, G., GUEHO, S., BEGUIN, D.: 'Dephaseur Clementaire en ligne microruban et
dephaseur a commande numerique en faisant application', Brevet 86 11 923. 21 Aug. 1986
36 DUBOST, G . , GUEHO, S., and BEGUIN, D.: 'Antenne reseau carrk, monopulse a balayage
C.lectronique'. Brevet 8720774, 13 Jun. 1987
37 DUBOST, G., ALEXIS, R.: 'Antenne formee a partir d'une cavitt resonnante comportant une
face rayonnate'. Brevet 8408.392, 19 May 1984
38 DUBOST, G.: 'Antennes plaques pour les te1&communications entre stations mobiles' L'Onde
Electrique, 1985), 65, pp. 41-49
39 DUBOST, G.: 'Vertically polarized flat antenna with omnidirectional radiation". Int. Sym-
posium on Antennas and Propagation, ISAP, Aug. 1985, Kyoto, Japan, pp. 109-112
40 DUBOST, G.: 'Influence of feed-point location on radiation resistance of a short-circuited flat
dipole', Electron. Lett., 1984, 20, pp. 980-981
41 DUBOST, G., BIZOUARD, A,: 'Antenne periodique plane'. Brevet 83.19.924 13 Dec. 1983
42 DUBOST, G., GUEHO, S., and BIZOUARD, A,: 'Log-periodic short-circuited dipole array
with a squinted beam', Electron. Lett., 20, pp. 41 1-413
43 DUBOST, G., and GUEHO, S.: 'Theory of a large bandwidth microstrip plane array with a
deflected beam'. IEEE Int. Symposium on Ant. Prop., June 1984, Boston, Mass., USA
Chapter 8
Numerical analysis of microstrip
patch antennas
J.R. Mosig, R.C. Hall and F.E. Gardiol
A..-
u, ~ 7< 8"..
7m uuddpe~i,
~ . Hu~rya~y 8.1 Introduction
8.1.1 General description
Microstrip patch antennas are thin and lightweight radiating elements, formed
by a substrate, including one or several dielectric layers, backed by a metallic
sheet (the ground plane).
Thin metallic patches (the radiating elements) are located on the air-substrate
interface and, possibly, between the dielectric layers. Microstrip antennas are
manufactured by the photolithographic process developed for printed circuits.
Their low profile, low weight and mechanical ruggedness make them an ideal
choice for aerospace applications. They can be mass-produced, and could thus
provide inexpensive receiver antennas for direct reception of microwave signals
from satellites (television, mobile communications). Finally, they are ideally
suited to be combined in large arrays, the individual patches sharing the same
substrate. Thus directive antennas can be obtained in spite of the inherent low
directivity of a single patch.
The remarkable practical advantages offered by microstrip antennas are
offset, to some extent, by their inhomogeneous nature, and a rigorous analysis
was long considered to be a hopeless task.
An accurate model should take into account the three inhomogeneities of a
microstrip structure (Fig. 8.1):
( a ) The presence of at least two dielectrics (often air and substrate)
( 6 ) The boundary conditions on the interfaces between different layers are
inhomogeneous since thin metallic plates making up the radiating elements and
feeding the structure can partially fill the interfaces
(c) Any microstrip structure is finite in dimensions; i.e. its ground plane and its
dielectric substrate are bounded in the transverse directions. The edges may,
however, be located at a very large distance, in which case this third inhomoge-
neity may be neglected (the structure is then assumed, mathematically, to extend
to infinity).
 394 Numerical analysis of microstrip patch antennas
Models used to study microstrip patch antennas range from very simplified
ones, such as the transmission-line model, through cavity models, planar circuit
analysis, segmentation techniques, and up to quite sophisticated approaches
based on an integral-equation formulation. In the framework of the integral-
equation model, many different approaches exist, depending on the use of
spectral or space quantities and on the geometries to be included.
Fig. 8.1 The three inhomogeneities of a microstrip structure.
a Dielectric media of unequal permittivity,
b Infinitely thin conductors introducing surface current between the dielectric media
c Finite transverse dimensions of substrate and ground plane
Detailed surveys of these models are available (e.g. Reference I), and several
of them are treated elsewhere in this book.
8.1.2 The integral equation model
The purpose of this Chapter is to provide a rigorous treatment of microstrip
antennas, free from over-simplifying assumptions. Among its principal features,
the proposed model is able to handle patches of arbitrary shapes where no
educated guess of the surface-current distribution is possible. Also there is no
limitation in frequency and substrate thickness. The model automatically takes
into account mutual coupling between elements and can predict the perfor-
mance of a patch embedded in an array environment. Surface waves are in-
cluded as well as dielectric and ohmic losses. Thus the model allows accurate
prediction of quasistatic behaviour, dominant and higher modes of resonance,
and input impedances, coupling coefficients, radiation patterns, gain and effi-
ciency at any frequency.
The model relies upon the identification of a microstrip antenna as a par-
ticular case of a stratified medium. The pioneer work on electromagnetic-wave
propagation in stratified media must be ascribed to A. Sommerfeld, who inves-
tigated the radio-wave propagation above a lossy ground as early as 1909.
The microstrip antenna is modelled by an integral equation where the main
unknown is the electric surface current density on the patches. The Green's
functions forming the kernel of this equation include the effects of the layers,
and are obtained in the form of inverse Hankel transforms by the systematic use
Numerical analysis of microstrip patch antennas 395
of stratified media theory. The first Sections of this Chapter are devoted to the
construction of the integral equation and of the pertinent Green's functions.
Considerable attention is paid to the development of efficient numerical techni-
ques for evaluating the Green's functions.
The integral equation is directly formulated in the space domain using a
vector and a scalar potential. The resulting mixed potential integral equation
(MPIE), similar to that obtained for wire antennas, is better suited for numerical
analysis than the customary electric-field integral formulation.
The MPIE is solved by a method of moments. In the general case, rooftop
subsectional-basis functions are used. For particular geometries, it can be more
efficient to use entire domain-basis functions corresponding to the eigenvalues
of the equivalent cavity. Finally, standard circuit analysis is used to deal with
multiport antennas, loaded antennas and arrays.
The final Sections of the Chapter present numerical results for the input
impedance, coupling coefficients and radiation patterns of several microstrip
antennas and arrays of practical interest.
8.2 Model based on the electric surface current
8.2.1 Geometry of the model and boundary conditions
In order to present the theory in the clearest possible manner, the electric
surface-current model will be developed for the simple microstrip structure of
Fig. 8.2 with a single dielectric layer and a metallic patch. The generalisation to
multilayer antennas and to multiple patches (arrays) will be considered later.
The substrate is assumed to extend to infinity in the transverse directions and
is made of a nonmagnetic, isotropic, homogeneous material which can be lossy.
The ground plane also has infinite dimensions, and the upper conductor (the
metallic patch) has zero thickness. Both the ground plane and patch are allowed
to have ohmic losses.
The direction perpendicular to the ground plane is selected as the z-axis (Fig.
8.2). The patch extends over part of the z = 0 plane, denoted by the surface So.
The remainder of the z = 0 plane, i.e. the surface separating the two dielectric
media, is denoted S and called hereinafter the interface.
Indexes 1 and 2 are associated, respectively, with the infinite dielectric extend-
ing above the substrate, usually the air, and with the substrate itself. Thus, we
have
El = Eo, Z > 0 (8.1)
E~ = E~E, = ~ ~ ~ i ( I - j t a n 6 )- h < z < O
where h is the substrate thickness and
everywhere.
 396 Numerical analysis of microstrip patch antennas Numerical analysis of microstrip patch antennas 397
The excitation is provided by a time-harmonic electromagnetic field. Complex The induced currents in turn create diffracted or scattered electromagnetic
phasor notation is used throughout this Chapter. Any complex scalar quantity fields. These fields, denoted E" Hd, add to the excitation fields to yield the total
C represents an instantaneous quantity C(t) given by: fields E, H existing in the entire space.
c(t)= J2 Re [Cexp (jwt)] On the air-dielectric interface (the plane z = 0, excluding the surface of the
patch So)the boundary conditions are:
ez x (El - E l ) = 0 (8.4)
e r x (HI-Hz) = 0 (8.5)
On the metallic surfaces the boundary conditions will be inhomogeneous
owing to the presence of the currents. Assuming that the patch and the ground
plane are perfect conductors (this restriction will be removed later) we have on
the upper side of the patch S,(z = 0+):
e Z x El = 0; e, x H, = 4, (8.6)
and, similarly on the lower side z = Ow:
e, x El = 0; e, x Hz = - J,2

Combining the above pairs of equations, we obtain

e, x (El - E,) = 0 (8.8)
e, x (HI - Hz) = J,, + J, = J, (8.9)
which apply to the fields on both sides of the patch. Eqns. 8.4-8.9 are written
in terms of the total fields. However, since the excitation fields are assumed to
be continuous, the boundary conditions (eqns. 8.4-8.5) and (eqns. 8.8-8.9) also
hold for the diffracted fields. Hence the diffracted tangential electric field is
continuous across the patch, while the jump in the diffracted tangential magnet-
ic field equals the total surface current on the patch.
Finally, the boundary conditions on the ground plane z = - h are
Fig. 8.2 General view of a microstrip antenna and vertical cut in the y = 0 plane
Superscripts e and d refer to the incident fields (excitation) and to the scattered ez x El = 0 (8.10)
(diffracted) fields. For a infinitely thin patch the currents J, and J,, can no longer
be distinguished and the meaningful quantity is the total current J, = J,,+ J,, e, x Hz = 4 (8.1 1)
flowing on the patch.
It must be pointed out that the microstrip antenna problem can be completely
solved without actually knowing the exact distribution of surfacepxrent on the
ground plane. In fact, image theory can be used to remove the ground plane and
The excitation fields are denoted by Ee and He. They can be the fields of a
the boundary conditions (eqns. 8.10-8.11) will then be included automatically
plane wave coming from infinity (receiving antenna) or the local fields created
in the Green's functions.
by a finite source located within the microstrip structure (transmitting antenna).
In either case, the excitation fields induce surface currents on the upper side
8.2.2 Potentials for the diffracted field
of the ground plane and on both sides of the metallic patch. However, since the Since no volume sources are considered in this model, the diffracted fields satisfy
patch is assumed to have zero thickness, the model cannot differentiate between the homogeneous Maxwell's curl equations:
the currents flowing on its upper and lower side. Indeed, the patch is modelled
as a sheet of current 4 whose value at any point is the algebraic sum of the upper V x E L-jwp,,Hd (8.12)
and lower surface currents 4, and 4, existing at z = 0+ and z = 0- (Fig. 8.2). V x Hd = j w ~E d (8.13)
 398 Numerical analysis of microstrip patch antennas
The resolution of antenna problems can in many cases be simplified by introduc-
ing the scalar and the vector potentials for the diffracted fields:
subject to Lorentz's gauge:
Introducing the above expressions into Maxwell's equations 8.12 and 8.13 and
combining, we obtain two homogeneous Helmholtz's equations for the poten-
tials:
Numerical analysis of microstrip patch antennas 399
effects of the dielectric substrate and of the ground plane. Therefore, the Green's
functions must satisfy the boundary conditions (eqns. 8.4-8.5) and (eqns. 8.10-
8.11). Moreover, the boundary condition for the magnetic field (eqn. 8.9) is
automatically built into the formulation of the Green's functions. These func-
tions are dyadics, and unfortunately do not have a closed analytical expression.
However, once they are formulated and numerically evaluated, the only
unknown which remains is the true electric surface-current distribution on the
conducting patch.

where k, = ~ ( j i ~ is~the , ) wavenumber

"~ in medium i.
It must be recalled here that the choice of the couple A , V is not unique [2].
Indeed, any vector A* = A + gradY can be used as a new vector potential
provided that the scalar potential is replaced by V * = V - joy. Moreover,
if Y is a solution of the homogeneous Helmholtz's equation, the new potential
will still satisfy Lorentz's gauge (eqn. 8.16). We shall discuss later some con-
venient choices for the potentials.

8.2.3 Green's functions Fig. 8.3 Horizontal electric dipole ( H E D ) o n a microstrip substrate
T h e fields and potentials of such an elementary source give the Green's functions
Let us consider an arbitrarily oriented Hertz dipole of moment I dl located at associated with a microstrip antenna.
the point r' (Fig. 8.3). In general, the vector potential a t the point Y due to this
dipole is given by the linear relationship
Keeping in mind the linearity of Maxwell's equations, the vector potential of
a given current distribution can be written as a superposition integral involving
w h ~ E,e is a three-dimensional dyadic Green's function. The physical meaning the corresponding dyadic Green's function
of G, is evident: the scalar component Git gives the s-component of the vector
potential existing at the point r created by a t-directed Hertz dipole located at
the point r'. The scalar potential V is obtained by introducing the above expression in the
If the source and the observer are surrounded by an infinite homogeneous Lorentz gauge with the result:
medium, the dyadic G, is diagonal and can be exp~essedas the product of a
scalar Green's function GA times the unit dyadic a.In this case, the vector
potential is always colinear with the source dipole.
For a microstrip antenna it is possible to use a scalar free-space Green's The Green's function G, associated with the scalar potential must be carefully
function of the vector potential. However, if we do, we then need to use fictive defined. In fact, the uniqueness of G, is guaranteed only if the divergence of GA
electric and magnetic surface currents on the air-dielectric interface in order to is an irrotational vector. Thus we can write [3]
satisfy the boundary conditions. These currents are also unknowns in the
integral-equation formulation of the problem and add to the complexity of the
numerical solution. The preferred solution is to include in the Green's functions
where V' acts on the primed co-ordinates. Expression 8.21 is now easily transfor-
 400 Numerical analysis of microstrip patch antennas
I Numerical analysis of microstrip patch antennas
med into
where dS, is the perimeter of the patch and n is the outwards-pointing normal
unit vector (Fig. 8.2). The edge condition guarantees that the normal com-
ponent of the surface current vanishes on the perimeter of the patch. Hence, the
line integral in eqn. 8.23 can be eliminated.
We can now introduce the associated surface-charge density q, via the con-
tinuity equation:
Finally, we can express the scalar potential as:
It is worth mentioning that the edge condition can be applied even if So is a
portion of a larger patch. Such a situation may arise when solving the problem
with a method of moments using subsectional-basis functions. In this case the
line integral in eqn. 8.23 can still be eliminated, but since 4 does not necessarily
vanish on the boundary of So, the continuity equation must be interpreted
according to the theory of distributions, and delta functions corresponding to
line charges in the boundary of So, will appear in the expression for q,.
The Green's function G,can be viewed as the scalar potential created by a point
charge, even if isolated time-varying point charges do not exist in the real world.
Thus, owing to the lack of a sound physical interpretation, it is better to consider
G, only as a useful mathematical device. Only when the frequency goes to zero,
does this function become the familiar electrostatic potential of a point charge.
8.2.4 Mixed potential integral equation ( M P I E )
The diffracted fields derived from the potentials of eqns. 8.20 and 8.25 satisfy
Maxwell's equations and the boundary conditions of the problem (eqns. 8.4-
8.5) and (eqns. 8.8-8.9). The last step is now to relate these fields to the
excitation fields via conditions (eqns. 8.6-8.7). If the total tangential electric field
is forced to vanish on the patch surface, we get the standard electric field integral
equation. This equation can be slightly modified to account for the ohmic losses
on the patch. The total tangential electric field is now proportional to the total
surface current, and we can write
or, introducing the potentials
where the proportionality constant Z, is a surface impedance (measured in
407
ohms) which accounts for the finite conductivity of the patch. An accurate value
for Z, can only be obtained by measurement since Z, must include technological
data such as the thickness and roughness of the metallic patch. However, in
most cases the patch is thick compared with the skin depth 6, and the classical
expression
Z, = (1 + j)/a*6 (8.27)
represents a good approximation. In the above expression a* is an effective
conductivity that includes roughness effects and can be several times lower
- than
the values of conductivity found in standard tables.
Finally, introducing the integral form of the potentials (eqns. 8.20-8.21) in
eqn. 8.26 we get the final expression for the mixed potential integral equation
(MPIE):
The validity of this equation depends on the possibility of defining the Green's
function G, by eqn. 8.22.
Eqn. 8.28 is a Fredholm integral equation of the second kind. However, the
term Z, J, is usually small and the iterative techniques commonly used for
Fredholm integral equations of the second kind that arise, for example, when .
using the magnetic field integral equation [4], do not apply here.
The unknowns in the integral equation 8.28 are the surface current J, and the
surface charge q,. However, they are not independent, and are related through
the continuity equation.
8.2.5 Sketch of the proposed technique
The successive steps in solving the microstrip antenna problem are now clear.
We start with the theoretical determination of the required Green's functions GA
and G,. In general, the Green's functions are available as definite integrals over
semi-infinite intervals and they must be numerically evaluated for distances
ranging from zero to the maximum linear dimension of the patch. The construc-
tion of accurate numerical integration algorithms to evaluate the Green's fun-
ctions is a crucial step of the overall problem.
Once the Green's functions are computed, the unknown surface current is
expanded over a set of basis functions and the integral equation is tested against
a set of test functions using the so-called method of moments (MOM). Here, the
correct choice of these sets of functions is essential for the quality of the final
results. In this way, the integral equation is discretised and transformed into a
set of linear equations. The complex eigenvalues of the matrix equation provide
the unperturbed resonant frequencies of the patch and its unloaded quality
factor.
The next step is the construction of the excitation fields. These fields depend
strongly on the physical nature of the excitation. In many cases (coaxial pin,
microstrip line) the computation of the excitation fields calls for the same or
 402 Numerical analysis of microstrip patch antennas Numerical analysis of microstrip patch antennas 403
related Green's functions that were calculated for the MOM matrix. Testing the Finally, integration of the far fields over the upper half-space will give the
excitation fields yields the independent vector of the matrix equation. directivity of the antenna and an estimation of its efficiency and gain.
The solution of the matrix equation gives a numerical estimation of the These steps are summarised in Table 8.1.
unknown surface current. Computing the voltages at the excitation points
allows the determination of the impedance and scattering matrices of the
antenna. Standard circuit analysis may be used to account for any load or for 8.3 Horizontal electric dipole (HED) in microstrip
multiple excitations. Resonant frequencies and loaded quality factors are easily
derived from the input impedance. The construction of the Green's functions requires the determination of the
fields created by a horizontal electric dipole (Hertzian dipole) located on the
Table 8.1 Essential steps (-) and main results (*) of the proposed numeri- air-dielectric interface of a microstrip structure (Fig. 8.3).
cal technique The first investigations of a HED embedded within a stratified medium are
due to Sommerfeld, who published in 1909 the exact solution for a dipole over
an imperfect ground. The Hankel integral transforms which appear within such
from the boundary conditions a problem are often known as Sommerfeld integrals.
The vertical dipole over a conducting plane covered with a dielectric layer was
studied by Lo and Brick in two articles which appeared almost simultaneo~sly
-construct the pertinent
15, 61. The problem is quite similar to the one arising in microstrip antenn, s,
except for the dipole orientation. However, even though the first microstrip lines
I
were introduced around the time these articles were pubIished, no connection
-find the pole and the residue
between the two fields was made - microstrip antennas were to be developed
J I associated to the GFs when
expressed in the spectral domain 1 I A DIPOLE ) some 20 years later. This may explain the different approaches to the two
I
problems, in particular the absence of a detailed study of the near field, which
-evaluate numerically the GFs is essential when the source and the observer are both located on the surface of
for the potentials in the GFs for the fieldsin the the substrate.
I near field region I far field region
The general theory of dipoles - either electric or magnetic, horizontal or
1 vertical, located within an arbitrary stratified medium - was developed later,
mainly by Brekhovskikh [7], Wait [8], Felsen and Marcuvitz [9], and Kong [lo].
AND UNLOADED a However, as was done in previous publications, the emphasis was put on the
to the intearol eountion
study of radiated fields, for which approximate asymptotic analysis is sufficient.
For the accurate study of microstrip radiation, however, precise knowledge
complex determinant
of the surface currents on the patch, and hence of the near fields on the dielectric
interface, are required. For this reason, the fields created by a HED located on
the air - dielectric interface will be determined.
+SURFACE CURRENT 4
DISTRIBUTIONsINPUT
IMPEDANCE AND
LOADED Q
4- solve the matrix equation]
I
8.3.1 The vector potential
Let us consider a HED of moment Idx equal to unity placed at the point r' = 0
-

*RADIATION PATTEF as indicated in Fig. 8.4.

POLARIZATION. In order to ease the mathematical development required for solving the
DIRECTIVITY. GAIN
AND EFFICIENCY Helmholtz equation 8.17 satisfied by the vector potential, we will replace the
space variables x , y by their spectral counterparts k,, ky according to the double
To calculate the radiation pattern we need to go back to the Green's functions Fourier transform:
and obtain their asymptotic forms in the far field. This can be done analytically
and the radiation pattern of the antenna, including polarisation and phase
information, is obtained by using a superposition integral over the patch.
 404 Numerical analysis of microstrip patch antennas Numerical analysis of microstrip patch antennas 405
If the dipole is embedded in an infinite homogeneous medium of permittivity E,,
the vector potential is parallel to the dipole and exhibits a spherical symmetry:
1
-jfmmf(kX,
2n
= k,)exp(jk,x +jk,y)dk,dk, (8.30)
In the spectral domain the homogeneous Helmholtz equation becomes:
On the other hand, it is well known [I I] that two cartesian components of A are
needed to satisfy the boundary conditions in an inhomogeneous structure such
as a microstrip antenna. Here, we shall adopt Sommerfeld's choice and pos-
where tulate an additional vertical component for the vector potential. Hence
A = e,A, +ezA,.
Choosing now for A, and A, general expressions of the form of eqn. 8.32, we
obtain after satisfying the boundary conditions the expressions:

i +
PC
2 = -
2n
(, - 1) J cosh u(z h)
(DmD, cosh uh)

where D, = u, + ucoth uh, D, = E,U, + u tanh uh, and the upper and lower
expression inside the symbol { ) correspond, respectively, to the upper semi-
infinite medium (z > 0) and to the substrate (- h < z i0).
Fig. 8.4 Co-ordinate system for the study of an x-directed horizontal electric dipole (HED)
It can be easily shown that if er = 1 and h 4 co, the vertical component A,
on a microstrip substrate vanishes and A, becomes the free-space vector potential given by eqn. 8.34.
The air-dielectric interface is at z = 0 and the ground plane is at z = -h. The
continuity equation implies that there are two point charges of value * I dxljw at both 8.3.2 Scalar potential and the fields
extremities of the dipole The continuity equation applied to an electric dipole implies the existence of two
According to eqn. 8.1 we shall adopt from now on the shorthand notation
+
point charges q = Ijjw at both ends of the dipole.
Since the product Idx has been assumed to be equal to unity, the moment of
u: = U: = ki - % a n d ui = u2 = ki - E,%. this pair of charges is qdx = l/jw. The scalar potential associated with the
The general solution of eqn. 8.31 for a cartesian component of A is (s = x, dipole is given directly by the Lorentz gauge. Introducing eqns. 8.35 and 8.36
Y, 2): in eqn. 8.16 we get:
j ~~xP(-w)/(DTEDTM)
p = -2 (8.37)
where the unknown factors a, b may be functions of the spectral variables.
The fields are obtained by usingeqns. 8.14-8.15 and the Lorentz gauge, which
~~JWNsinh
I u(z h)/(DrEDTM +
sinh uh)
also holds in the spectral domain. Since no external excitation fields are con- with N = u, + u tanh uh.
sidered here, the fields of eqns. 8.14-8.15 are total fields. These fields must satisfy In the space domain, the scalar potential V of an electrostatic dipole of
the boundary conditions eqns. 8.4-8.5 on the interface and eqn. 8.10 on the moment lljw is related to the scalar potential V, of a single unit point charge
ground plane. In particular, the HED is equivalent to a surface current density by the well known expression:
in the plane z = 0 given by: I av,
v = ---
jw ax
 406 Numerical analysis of microstrip patch antennas Numerical analysis of microstrip patch antennas 407
or, in the spectral domain by

Hence, the surface waves appear as poles of the integrands in the complex
Comparing eqn. 8.39 with eqn. 8.37, we can deduce by analogy that the plane k, = 1 + j v . It can be shown [I31 that D , has no zeros if
potential of a unit point charge on the air-dielectric interface of a microstrip kOh(c;- 1)1'2< 4 2 and D , has only one corresponding to the dominant
antenna is given by zero-cutoff TM surface wave. This condition is equivalent to the restriction:

i
For the sake of simplicity we shall assume from now on that this inequality
Nsinh u(z + h)
holds. Higher frequencies would add new poles, but the analysis made for the
DTEDTM sinh uh single-pole case remains qualitatively valid.
Now, the construction of the fields is straightforward, for we have in the For substrates with moderate losses the pole 1, + jv, lies slightly below the
spectral domain: real axis (v, < 0) and its real part is bounded by 1 < Ap/k0 < ~ f " (Fig. 8.5).

It can be easily demonstrated that the vertical component E, shows the

expected jump discontinuity when crossing the air-dielectric interface.

8.3.3 Surface waves and the spectral plane k,

It can be shown [I21 that the equations DTE = 0, D , = 0 are the characteristic
equations for the TE and TM surface-wave modes propagating on a dielectric-
coated conducting plane. Hence the zeros of D , and DTM give the phase Pole (lossless)
constant of the surface waves existing on a rnicrostrip structure.
The terms D, and DTMdepend on k, and k, only through the radial spectral
+
variable $ = k2, $. For functions exhibiting such a kind of dependence, the
Fig. 8.5 Topology of the complex plane k, with the original integration path from zero to
inverse Fourier transform 8.30 can be written as
infinity
g-'[.&)l = jom~ d ~ ~ ddkQ~ ~ 3 ( ~ ~ ) (8.42~) The Figure also shows the geometrical locus of the pole as a function of dielectric
losses.
and
More precisely, a good approximation for electrically thin substrates [I41 is
where Q, 4 are polar co-ordinates and Jois the Bessel function of zeroth order
and first kind. The replacement of the double Fourier transform by Hankel
transforms is of great utility when performing the numerical evaluation of the
fields and potentials. For instance, if the observation point is also on the
interface ( z = 0) we have:
The integration path of the Sommerfeld integral eqn. 8.43-8.44 is, in general, the
real positive axis 1. But, if we consider the theoretical case of a lossless substrate,
then v, = 0 and the pole is on the real axis. Since, by continuity, the integration
path must remain above the pole, the integral from zero to infinity in the lossless
 408 Numerical analysis of microstrip patch antennas Numerical analysis of microstrip patch antennas 409
case is interpreted as: On the other hand, it can be shown [9] that this replacement will give correct
results even for the broadside direction Q = 0".
The final expression of the integral in the w-plane is
where the symbolfstands for Cauchy's principal value and R is the residue of
the integral at the pole k, = 4.
Finally, it should be mentioned that the function u; = ki - ki introduces a
branch point at k, = ko (Fig. 8.5). However, the integral remains bounded here
and no deformation of the integration path is needed.

8.3.4 Far-Jield approximations

Far-field approximations are essential for the evaluation of the radiation pattern
of a microstrip antenna. They can be obtained by using standard asymptotic
techniques, such as the steepest-descent method [9]. We shall outline briefly the
application of these techniques to Sommerfeld integrals. We begin by replacing
the Bessel functions in these integrals by Hankel functions. This is done recalling
the identity
F i g . 8.6 The integration path C closing at infinity in the k, plane and the new branch cut
introduced by the Hankel function
A portion (dotted line) of the path C is in the lower Riemann sheet

which holds if.f'is an even function of k, [ll]. In this way, the integration path
C in the complex plane k, closes at infinity (Fig. 8.6) while the topology of the
plane remains unchanged except for an additional branch point introduced by
5
the Hankel function (Fig. 8.6). For the sake of clarity, the pole k, = is located
on the real axis (lossless substrate, see Fig. 8.5).
Let us consider now a typical integral in the air:

where C is the path shown in Fig. 8.6.

To obtain an asymptotic expansion for I, the plane k, is transformed into a
new complex plane w defined by the relation

The transformed path C* in the plane w is shown in Fig. 8.7. The pole is now Fig. 8.7 The new complex plane w (k, = k,sin w ) showing the transformed path C*,the

+
located at w, = 4 2 jarcosh (lp/ko)and the branch cuts associated with the steepest descent path ,
C and the new location of the pole
Also the correspondence with the four quadrants of the plane k, is given, either in
points k, = $. ko disappear owing to the new choice of variables. Introducing the upper (U)or in the lower (L) Riemann sheet.
spherical co-ordinates r, 8 (z = r cos 8, Q = r sin 8) the argument of the Hankel
function becomes

Since we want an expression useful in the far field (k,r p l), we can replace the F(W) = j" ( nR sin"
0 sin w )'I2 k, cos w/(ko sin w)
Hankel function by its first-order asymptotic expression. The integration path
can be deformed far from the origin w = 0 in order that sin w never vanishes. with R = kor and q(w) = -jcos(w - 8).
 410 Numerical analysis of microstrip patch antennas
The saddle point w, of function q is given by w, = 0. The steepest-descent
path CsD through w, is defined by Im(q) = - 1. Its particularities are easily
derived (Fig. 8.7):
(i) Cm is at a 45" angle with the real axis.
(ii) Cs, intersects the line Re(w) = n/2 at the level Im(w) = cosh-'(l/sin 0).
(iii) As a result, the pole w, is located between CsDand the positive real axis only
when 0 > 0, = sin-'(k,/&).
(iv) CsD possesses two vertical asymptotes at Re(w) = 0 +
n/2.
Contributions from the integrals joining the two paths C* and CsD at infinity
can be eliminated since the term exp (Rq) vanishes in these regions. The path C*
can then be deformed into the path CsD, provided the contribution of the pole
is added for angles greater than the critical angle 0,.
This is written symbolically as
where U is the Heaviside unit step function and Cp is a patch surrounding the
pole wp (Fig. 8.7).
The first-order approximation for the integral along CsD can now be obtained
by standard techniques [9]. The integral around the pole is evaluated using the
residue theorem and is expressed in the original k, plane.
Finally, the asymptotic expression of the integral I(eqn. 8.49), valid in the far
field region, is
I = 2j"+l cotan 0f (k, sin 0) exp(-'kor) [I + ~ ( r - I ) ]- U(B - B,)
where the residue R is given by
4)f(k,)
R = lim (k, -
kg - AP
(8.55)
and the Landau notation O(f) indicates the behavior at infinity of the first term
neglected in the asymptotic expansion.
It must be noted here that the asymptotic approximation eqn. 8.54 is only
valid if the pole rZ, is located far enough from the saddle point 1 = k,, sin 0, i.e.
(k, sin O - AP)r % I. Otherwise, the contributions of the saddle point and of the
pole cannot be separated and a modified saddle-point method must be used [9].
From the asymptotic point of view, the fields and potentials are the sum of
two terms. The first term, due to the saddle point, represents a spatial wave with
a complex factor depending on angle 0 and corresponds to the geometrical
optical fields. The second term, due to the pole at Ap, represents a cylindrical
wave decreasing exponentially away from the substrate that corresponds to the
surface wave.
Numerical analysis of microstrip patch antennas 41 1
The surface wave is only relevant for angles 0 > OP, i.e. near the interface.
However, its field dependence on e-'I2 can make it the dominant term of eqn.
8.54 when fields on the substrate surface at z = 0 are evaluated.
Radiated electricjeld: An immediate application of the asymptotic relation-
ship eqn. 8.54 is the computation of the radiated field. It is assumed here that
0 < OP,so that the surface wave can be neglected. In a real situation, a substrate
always has finite dimensions and the surface wave can be observed directly only
close to the substrate. For angles near 0 = n/2 (grazing angles) but far from the
substrate, diffraction effects of the surface wave on the edges become significant.
The radiation field is obtained by transforming, for z > 0, the rectangular
components of E (eqn. 8.41) into spherical components and then applying eqn.
8.54 to the resulting integrals. The final expressions are
where 2, is the free-space impedance, I , the free-space wavelength, not to be
confused with the spectral variable 1 = Re[k,], and
go(@ = Tcos 0/(T - j&,cosOcotan k,hT)
g,(0) = cos 0/(cos0 - jTcotank,hT)
These asymptotic expansions have also been derived, with a different approach,
by several other authors [15, 161. The result for E, shows that this component
decreases faster with distance than Ilr, and is thus not a radiated component.
Figs. 8.8 and 8.9 give the polar radiation patterns, respectively, in the E-plane
and H-plane. In each Figure, four substrate thickness k,h = 0.05,0.1, 0.2 and
0.5 have been considered, and, for each thicknesses, curves corresponding to
four dielectric constants E, = 1, 2, 5 and 10 have been plotted.
The presence of a dielectric layer increases, in general, the half-power beam-
width in the E-plane, especially for thin substrates. On the other hand, the
H-plane pattern is almost independent of the substrate parameters, except for
very thick substrates just on the threshold for the generation of a second surface
wave.
Potentials at the interface: The Green's functions appearing in the integral
equation 8.28 can be obtained from the potentials A, and V,. Solving the
integral equation requires the knowledge of these potentials only at the inter-
face. If we apply eqns. 8.54 to eqns. 8.43 and 8.44, transformed according to eqn.
8.48, we obtain with z = 0 and O = x/2:
 472 Numerical analysis of microstrip patch antennas
4m0 % - -2zjRHi2)(&~)
where the residue R is given by eqn. 8.55 with f = k,N/DTEDTM.It is here
apparent that the Q - ' contribution of the saddle point vanishes in both ex-
pressions. If higher-order terms in Q-' are desired for A,, which has no surface-
wave component, the whole C,, integration path in the complex plane k, must
be considered.
Numerical analysis of microstrip patch antennas 413
(8.58) in all situations, but the calculations are rather complex, requiring error func-
tions of complex arguments [9],and they will not be carried out here.
Asymptotic expressions 8.58 and 8.59 will be used in the following develop-
ment to check the results obtained using numerical integration of the potentials.
H plane: kh=0.05 H plane: kh=O.l

E plane: kh=0.05 E plane: kh=O.l

H plane: kh-0.2 H plane: kh=0.5

E olane: kh=0.2 E plane: kh=0.5

Fig. 8.9 Polarplot of the H-plane radiation pattern (in dB) of a HED on a microstrip substrate
for four normalised substrate thicknesses
Fig. 8.8 Polarplot of the E-plane radiation pattern (in dB) of a HED on a microstrip substrate ( a ) koh = 0.05 (6)koh = 0.10
for four normalised substrate thicknesses ( c ) koh = 0.20 ( d ) koh = 0.50
( a ) k,h = 0.05 ( b ) koh = 0.10 A: 8, = 1 B:E, = 2 C: &, = 5 D: E,= 10
( c ) koh = 0.20 ( d ) koh = 0.50
A:E,= 1 B:E,= 2 C:c,= 5 D : t , = 10

8.3.5 Radiation resistance and antenna efficiency

It can be shown that the asymptotic behaviour of the integral is mainly
determined by the discontinuity of the derivative of the integral at the point
k, = k,,. The dominant term in the asymptotic expansion is
with the parameter A defined as A = k , h ( ~ ,- 1)'''.
It is possible to replace eqn. 8.58 by a uniform asymptotic development valid
If the cylindrical components of the fields are expressed in terms of the poten-
tials by transforming eqns. 8.41, it can be shown that the term DTMappears only
in the denominator of components E,E, and H 4 . Therefore, these are the
components that include a surface-wave term, and thus the Poynting vector S
associated with the surface wave has a radial and a vertical component, respec-
tively, S, = - E,H4 and Sz = E,H4. Consideration of the general expression
8.54 shows that the vertical component decreases exponentially with z , and
consequently does not contribute to the radiated power. On the other hand, the
414 Numerical analysis of microstrip patch antennas Numerical analysis of microstrip patch antennas 415
radial component can be written according to expression 8.54 as and we get
S, = (Zo/2nko)(~,
- 1) ~ o s ~ ~ ~ ~ ( l d l ) ~ L ~ F ( z ) / @ (8.60)
with
Here again, the above formula becomes for E, = 1 the well known radiation
exp (- uO~)/~TE resistance of a horizontal dipole above a ground plane [40]. It is worthwhile to
F(z) =
cosh u(z + h)/D,cosh uh compute numerically the integral 8.64 and plot the values of the radiation
) ~a, function of the parameter A =
resistance R,, normalised to n ~ ~ ( d l / l , as
where the asymptotic expansion of the Hankel function has been used, and R
is the residue of l/D, at the pole.
koh m, which is proportional to the substrate thickness and to the fre-
quency. This has been done and the results are given in Fig. 8.10 for three values
It can be shown that the integral of this component over a cylindrical surface of the dielectric constant. Initially, the radiation resistance increases with the
of radius Q extending from z = - h to infinity has a non-zero value independent square of the thickness, as predicted by eqn. 8.66. But as the normalised
of Q Hence, there is a net amount of power carried away by the surface wave. thickness increases, the values of the resistance oscillate and show a discon-
The surface integral must, in general, be numerically evaluated, essentially tinuous derivative at the points where A is an odd multiple of 4 2 . These values
because there is no analytical formula for computing the pole Ap. However, for correspond to the generation of higher-order surface waves [17].
thin substrates we can use the approximation given in eqn 8 46 and estimate the
residue appearing in eqn. 8.60 as:

Then we define a radiation resistance R,,,, associated with the surface wave as

In a similar fashion, we can introduce the radiation resistance R, associated

with the space wave. Starting with the asymptotic expansions 8.56 for the fields
and using the fact that in the far-field zone we have Z o H = e, x E we can
demonstrate that the Poynting vector is radial, its modulus being given by

Integrating this expression over the upper (z > 0) half-sphere of radius r and
equating the resulting power to 1 2 R , we obtain the expression of the radiation 0 0.2 0.4 0.6 0.8
resistance of the space wave: normalized thickness
Fig. 8.10 Normalised radiation resistance O = R,l[nZo(dl/io)Z] of a HED on a rnicrostrip
substrate as a function of the parameter A = kohJ(E,- 1). Z, % 120n is the
free-space wave impedance.
As before, this surface integral cannot be analytically evaluated except for the A: E, = 2
8:E, = 3
case E, = 1 and h = co,where we recover the classical result for the radiation C: E, = 4
resistance of a Hertzian dipole radiating into free space.
However, for thin substrates we can again estimate the surface integral 8.64 We can now evaluate the ratio between the power carried away by the surface
by using the approximations: wave and the power radiated by the space wave as:
jk0h(a, - sin20) Power (surface wave) = n2
g, = ; g, = jk,h cos 0, (8.65) (a, - 1)'h/& (8.67)
E, = Power (spatial wave) 2 4
-E;(E, - 1) -E,
3 15
+
 476 Numerical analysis of microstrip patch antennas
which is proportional to the normalised thickness and shows a rather com-
plicated dependence on the permittivity of the substrate. Finally, the radiation
efficiency of a HED on a thin lossless microstrip substrate is given by
a = l/(l + 4)
Numerical tests have shown eqn. 8.67 to be accurate for h C 0.05 Lo. Fig.
8.1 1 gives the theoretical values of the efficiency for several dielectric constants.
Fig. 8.11 Radiation efficiency (space- wave radiated-powerltotalradiated-power) of a HED
on a lossless microstrip substrate as a function of the normalised substrate thickness
A: E, = 2
8: E, = 5
C:E , = 10
Finite sizepatches: The above considerations refer to an elementary Hertzian
dipole, but can be easily extended to finite-size microstrip patches by using
superposition. The patch is excited by a unit current, and the input impedance
Z,, of the antenna is obtained with techniques to be described in the following
Sections. Once the radiated fields are known, a radiation resistance R, is
calculated using the techniques outlined in this Section. The overall antenna
efficiency is given by the ratio R,,/Re (Z,,,,). If the antenna has been analysed
assuming a lossless substrate and a perfect conductor, the conservation of power
implies that R, +
R,,, = Re (Z,,,,). This is a useful check on the accuracy of the
numerical calculations. Otherwise Re (Z,N) is greater than R, +
R,, and the
difference gives the power dissipated in the antenna in the foim of ohmic and
dielectric losses.
Numerical analysis of microstrip patch antennas 47 7
The gain of the patch can now be defined in a customary way, as the product
of the efficiency times the directivity.
(8.68)
8.4 Numerical techniques for Sommerfeld Integrals
When a microstrip antenna is analysed by an integral-equation technique, it is
necessary to evaluate the interaction between points separated by distances
ranging from zero to several wavelengths. For most of these distances the
accuracy of near- and far-field approximations is not sufficient and the poten-
tials must be numerically evaluated. For a single-layer microstrip antenna the
source and the observation point are both on the interface. Hence z = 0 in the
integral expressions for the fields and the potentials, and the exponential fun-
ction which ensures convergence of the integrands disappears. This is the most
difficult case numerically, and we will concentrate on it in this Section.
Even though many deformations of the original path C of Fig. 8.6 have been
tried [14, 181, we feel that the integration along C (the real positive axis A of the
complex plane k,) provides the most efficient algorithm for evaluating the
Sommerfeld integrals appearing in microstrip problems.
8.4.1 Numerical integration on the real axis
The functions to be integrated oscillate on the real axis due to the Bessel
functions. The square root uo = (A2 - k;)'l2 introduces a discontinuity of the
derivative at 1 = k, that corresponds to a branch point in the complex plane.
If the integrand contains D, in the denominator there is a pole just below the
real axis (or on the axis itself for a lossless substrate) that produces very strong
variations of the integrands. Finally, many of the oscillating integrands have an
envelope which converges very slowly (A,, V,) or even diverges at infinity (A,,
V). When the envelope diverges, the integral diverges in the Riemann sense since
the area under the curve representing the integrand fails to converge to a finite
value as the upper bound goes to infinity. However, the integral can be inter-
preted in the Abel sense as
lomF(1) d l = lim
Z-o lomF(1) exp ( - u, z) d1
and the exponential guarantees convergence. This means, physically, that the
potentials at the interface can be considered as the limiting case of the potentials
in the air when the height of the observation point above the interface goes to
zero.
All these facts are clearly depicted in Fig. 8.12, which shows the integrand of
the scalar potential V, for E,, = 2.55, tan6 = 0, k,h = 0.3 x, and k,e = 3. In
this Figure, the substrate has been chosen quite thick for the sake of pictorial
clarity. Electrically thinner, more common substrates will exhibit a pole very
close to the branch point.
478 Numerical analysis of rnicrostrip patch antennas
The integration interval is decomposed into three subintervals [O, k,], [ko,
koJE,] and [k,J&,, co]. In the region [0, k,] the infinite derivative in ko is
eliminated with a change of variables 1 = k,cos t. The resulting smoother
function is integrated numerically. In the interval [k,, k,,/~,], the singularity is
first extracted if the integrand has D , in the denominator. By writing the
function under the integral sign in the form F(1) = J.(~Q)f(L), we have
Fig. 8.12 Normalised values of the integrand associated with the scalar potential V of a HED
on rnicrostrip
~ , = 2 , 5 5koh=0.3x k o e = 3
A: Discontinuities in the derivative
8 : Sharp peaks due to the pole
C: Oscillatory and divergent behaviour at infinity
-Real part ---- Imaginary part
where
Here 4 + jvp is the complex pole (v, < 0) and R the residue of F at the pole.
The function F,,, is integrated analytically as
Numerical analysis of microstrip patch antennas 419
It is worth mentioning that, in the lossless case (v, = O), the above integral
becomes
and therefore the principal-value formulation (eqn. 8.47) of the lossless case is
included' as a limiting case in this numerical technique.
Fig. 8.13 Realpart of the integrand of Fig. 8.12 for the lossy case in the interval[k,, k, J(E,)]
A: Before the extraction of the singularity
8 : After the extraction of the singularity
C: After the change of variables: 1 = k,cosht
Fig. 8.13 depicts the real part of the original function F(l) (solid line, A) and
the difference F(1) - Eing(l)(dotted line, B) after the singularity has been
extracted. There is still an infinite derivative in the curve B at I = k,; however,
with a change of variables 1 = k,cosh(t) one finally obtains a very smooth
integrand (the dashed line C in Fig. 8.13), which is integrated by a Gaussian
quadrature. The same procedure is applied to the imaginary part of F(1) to
eliminate in a similar way the sharp peak and the infinite derivative.
Finally, in the region [~,JE,, co] we first extract the static term defined by F(d,
ko = 0). Fig. 8.14 depicts the integrand F(1, k,) (curve A) and the difference
F(1, k,) - F(d, 0) (curve B). It can be shown that the static term has the form
and hence it can be integrated analytically.
420 Numerical analysis of microstrip patch antennas
The remaining part is a slowly convergent oscillating function over a semi-
infinite interval that is integrated with specially tailored techniques.
8.4.2 Integrating oscillating functions over unbounded intervals
Sommerfeld integrals, as given by eqn. 8.42, can be grouped in a more general
class of integrals defined by:
I(@) = Jom g @ d f( 4 d l (8.74)
where
(a) g(i@)is a complex function whose real and imaginary parts oscillate with a
strictly periodic behaviour (sin, cos), or behave asymptotically as the product of
a periodic function and a monotonic function. A typical example of this class
of functions, which will be termed from now on as quasi-periodic, are the Bessel
functions of the first kind.
(b) f (A) is a smooth, non-oscillating function which behaves asymptotically as
2 exp (- lz). Therefore, for points on the interface (z = O), the function f ( i )
decreases very slowly or even diverges at infinity if a > 0. Here, we shall discuss
the most interesting case, z = 0, which is also the most difficult from a numeri-
cal point of view. For the sake of simplicity, f(A) is assumed to be real. Complex
functions can be handled by working alternately with their real and imaginary
parts.
(c) The lower integration bound a has been chosen conveniently to ensure that
the interval [a, co) is far enough from any possible singularities of f(L). For
instance in our problem, we shall take a = JE,.
It is worth mentioning that the general expression 8.74 includes many integral
transforms such as Fourier and Hankel transforms. Hence, the following techni-
ques can be applied to many other problems in numerical analysis.
The classical problem involving Sommerfeld integrals is the problem of
radio-wave propagation above a lossy ground, where the comprehensive mono-
graph of Lytle and Lager is the classical reference [19]. These authors have
found an iterative Romberg integration satisfactory, since here the integrand
displays an exponential convergence and its poles have been removed from the
real axis. In microstrip problems, Romberg integration has also been used, but
its effectiveness decreases considerably in the absence of a well-behaved inte-
grand.
In recent years, there has been a considerable amount of work published on
the numerical evaluation of Fourier transforms, which are included in eqn. 8.74
as a particular case. The involved techniques can be classified in three groups.
(a) The decomposition [a, co] = [a, A] + [A, co]. Here, Filon's algorithm is
applied to the finite interval [a, A], while an asymptotic expression of the
integrand is used to estimate the integral's value over [A, co] [20]. The most
serious drawbacks of this approach are the choice of A and the analytical work
Numerical analysis of microstrip patch antennas 42 7
required - two features which are difficult to incorporate in an automatic
computation routine.
(b) Another approach applies if g(lp) is a strictly periodic function. The follow-
ing decomposition is then used:
where P is the period of the function g. The infinite sum under the integration
sign can be evaluated using standard techniques such as Euler's transformation.
Recently, a more involved technique using theoretical Fourier-transform con-
cepts has been described in connection with a problem on quantum-mechanics
impact cross-sections [21]. These methods work very well for large values of Q
and an exponentially decreasing integrand. Unfortunately, their extension to
quasi-periodic diverging integrands seems problematic.
(c) The third group of techniques, introduced by Hurwitz and Zweifel[22], are
defined by the decomposition
The integration over each half-cycle is performed prior to the series' summation.
As before, an accelerating device, such as the nonlinear transformations of
Shanks [23] and Sidi [24] can be used to sum the infinite series.
We feel that the decomposition 8.76 is particularly well suited for the Som-
merfeld integrals encountered in microstrip problems and we have used it
extensively throughout this work. However, instead of the sophisticated non-
linear techniques mentioned above, we have devised a new simple technique
based on the concept of a weighted average between the half-cycle integrals [14].
This accelerating device has proved to be faster and more accurate for these kind
of integrals.
8.5 Construction of the Green's functions
Once the potentials of a HED are known, the potentials created by an arbitrary
surface current J, existing on the plane z = 0 can be determined by using the
superposition integrals 8.20 and 8.25. The Green's functions arising in these
expressions are closely related to the potentials of a HED and can be easily
obtained if the symmetry properties of microstrip structures are taken into
account.
We shall restrict ourselves here to the case where the source and observer are
both on the air-dielectric interface (z = z' = 0). Then, according to the
translational invariance of the microstrip structure in any plane perpendicular
 422 Numerical analysis of microstrip patch antennas Numerical analysis of microstrip patch antennas 423
to the z-axis (eqn. 8.43), we have dipole. However, since only horizontal surface currents are considered we do
not need to compute these expressions.
It is now a matter of straightforward algebra to show that the transverse
divergence of the dyadic

is given by
where the angle a is given by a = tan-'{0, - yf)/(x - x')). In short, we can
say that the components G F and GF are given by the Sommerfeld integrals 8.43
with the polar co-ordinates Q, cp replaced by R = lr-r'l and a.
and, correspondingly, is derived from a scalar potential according to eqn. 8.22.
0.4 Therefore, provided that no vertical currents are considered, it is possible to
A
define a Green's function associated with the scalar potential as
G,(rlrf) = eqn. 8.44 with Q replaced by Ir - r'l = R. (8.81)
This concludes the derivation of the Green's functions needed to solve the
I integral equation 8.28. A method-of-moments solution, described briefly in the
next section, is used to numerically solve the equation and calculate the micro-
strip antenna parameters of interest.

8.6 Method of moments

The integral equation 8.28 will be numerically tackled with a method of mo-
ments (MOM). This technique [25] transforms the integral equation into a
matrix algebraic equation which can be easily solved on a computer. The
method of moments is among the most widespread numerical techniques in
electromagnetics and a detailed account of the underlying principles will not be
given here. For the problem of the microstrip antenna two particular versions
of the MOM deserve attention: the subsectional-basis functions approach and
Fig. 8.1 4 Real part of the integrand of Fig. 8.12 for the lossy case in the interval [k,J(&,), m]
A: Before the extraction of the static term
Galerkin's method with entire domain-basis functions.
B: After the extraction of the static term
8.6.1 Rooftop (subsectional)-basis functions
If no a priori assumptions about the shape of the patches are made, a successful
In a similar way, since a microstrip substrate shows a symmetry of revolution technique must decompose the patch into small elementary cells and define
around the z-axis, we can write simple approximations for the surface current on each cell. The most commonly
I GP (rlr') = G7(rlr1) used shapes for the elementary cells are the triangle [26] and the rectangle. Even
though the triangular shape is more flexible, rectangular cells involve simpler
calculations and suffice for many microstrip antenna problems. Concerning the
basis functions to be defined on these rectangular cells, a comparison of avail-
able possibilities [13] led to the selection of rooftop functions for the surface
current J,, that have been successfully used in similar problems [27]. To imple-
To evaluate G;', G r , Gy we would need the potentials of a vertical electric ment these functions, the patch's boundary is replaced by a Manhattan-type
424 Numerical analysis of microstrip patch antennas Numerical analysis of microstrip patch antennas 425
polygonal line (Fig. 8.15). As most commonly used antennas exhibit this kind
of geometry anyway, this requirement is easily satisfied.
The patch's surface is then divided into rectangular cells, called charge cells,
which, for the sake of clarity, will be assumed to be of identical size. This is not
an essential requirement for the theory of the algorithm, but the use of different 4 = f Z I,,T,(r - r,])
/=I
cell sizes considerably increases the computation time.
where the basis funct~onsare rooftop-type funct~onsshown In Flg 8 16:
- Ixl/a 1x1 r a, I yl < b/2
T,( 4 =
elsewhere

Fig. 8.16 Subsectional basis functions defined over pairs of adjacent cells (a), associated
F i g . 8.15 Decomposition of the upper conductor in elementary cells showing the discretisa-
constant charge distribution on each cell (6). and razor testing functions (c)
tion of the current and the test segments. After [38].

Two adjacent charge cells, sharing a common border perpendicular to the
x-direction (y-direction) will form an x-directed ( y-directed) current cell (Fig.
8.16). An automatic overlapping of current cells is obtained in this manner so
that a particular charge cell may belong to up to four different current cells. The
number of charge cells is thus related to the number of current cells, though the
relationship is not a simple one, since it depends on the shape of the patch.
However, for rectangular patches with m x n charge cells, the number of
x-directed current cells is M = (m - l)n, and there are N = m(n - 1) y-direc-
ted current cells.
Every current cell supports one rooftop basis function and there is one
associated test segment joining the centres of the two charge cells making up the
A similar expression is obtained for T, by interchanging a tt b and x ++ y in the
above equation.
The introduction of factors I/a and llb in eqn. 8.83 yields unknown coef-
ficients I, and I,,having the dimensions of a current. Moreover, every coefficient
gives the total current flowing across the common boundary of two charge cells.
The associated surface charge density is obtained from eqn. 8.83 by using the
continuity equation, yielding
N

current cell. The centre of the segment C,r, associated with the j t h x-directed + 2I I t J [ n ( r- r;)
J=
- n(r - r i )I (8.85)
current will be denoted by the vector r,,, and its ends by r,; and r; (Fig. 8.16).
These three vectors are related through where H(r) is a two-dimensional unit pulse function defined over a rectangle of
dimensions a x b , centered at r = 0.
The charge density within every elementary cell remains constant, justifying
A similar relationship is written for y-directed segments C,(j = 1, 2,. . . N ) . the name charge cell. For the charge cell of Fig. 8.15, with four test segments
ending at its centre, the surface charge density is simply given by
Basis functions: The Cartesian components of the surface current are expan-
ded over a set of basis functions T,, T,:
426 Numerical analysis of microstrip patch antennas
The charge density is discontinuous on the borders between charge cells. How-
ever, the scalar potential remains bounded, while the electric field becomes
singular, since q, does not satisfy a Holder condition [13]. This means that the
test functions must be selected carefully, avoiding the locations where the
electric field is singular.
Discrete Green's functions: The notation and the computational task can be
simplified by introducing 'discrete Green's functions', that have as a source a
complete basis function, rather than the traditional elementary point source.
The vector potential T, is created by a rooftop distribution of surface current,
whereas rvis the scalar potential resulting from a rectangular distribution of
unit surface charge. In practice it is convenient to deal with dimensionless
quantities and in a normalised space where physical lengths are replaced by
electrical lengths. The following dimensionless expressions are therefore in-
troduced that define the discrete Green's functions:
A similar expression may be written for T:?. In these formulas rq(ro,) denotes
the centre of a test segment and S,(Soj) the surface of a current (charge) cell.
The discrete Green's functions exhibit the same properties of translational
invariance and symmetry as the conventional Green's functions. In general, the
surface integrals in eqns. 8.87 and 8.88 must be evaluated numerically. When the
observation point r belongs to the source cell, some difficulties arise in the
integration process. It is then recommended that the singular part of the Green's
function which corresponds to the dominant term of their static value be
extracted, i.e. G = G, + (G - G,) where the static value G, is given by:
for the vector potential and
for the scalar potential.
The singular part G, can be analytically integrated over the cell's surface. For
-
instance, the singular part of eqn. 8.88 is
2n(e, + l)Tv(OIO) 2k0a In tan
(; + 3
- - 2k0b In tan ( 4 2 ) (8.91)
with tan cr = bla.
Numerical analysis of microstrip patch antennas 427
When the observer is located many cells away from the sources, the sources
can be concentrated at the centre of the cell. The following approximations may
then be used:
Test functions: The last step of the solution with the moment method is the
selection of a suitable test function. Previous work [I31 has shown that the best
choice, compatible with the basis functions selected, is the use of unidimensional
rectangular pulses. The use of these test functions, also called razor test fun-
ctions (see Fig. 8.16), is equivalent to integrating the boundary condition (eqn.
8.26) along the segments linking the centres of adjacent cells (test segments), and
therefore the testing procedure yields equations of the type:
where C,v,is the x-directed test segment extending from r,; to :r and VE1 is the
excitation (impressed) voltage along the segment. A similar relationship is
obtained for y-directed test segments. It is worth mentioning that this choice
eliminates the need for computing field values near the edges where field sin-
gularities can adversely affect the performance of the moment method.
Eqn. 8.93 is well suited for numerical treatment since all derivatives have been
removed. The integration of J,, can be done easily using the expansion given by
eqn. 8.83 with the result
The last approximation is valid for a reasonably smooth current distribution.
The matrix equation: Introducing the expansions 8.83 and 8.85 in eqn. 8.93
and using the discrete Green's functions defined above, the following matrix
equation is obtained:
 428 Numerical analysis of microstrip patch antennas
The elements in the submatrices are given by
where 6, is the Kronecker delta. The expression for C$yis obtained by interch-
anging the couples ( x , y), (a, b) and (M, N ) within eqn. 8.96. Finally, it is easily
shown that C? = CT.
For distances Ir,, - r,l much greater than the dimensions of a cell, the
integrals in eqn. (8.96) can be replaced by
In principle, this approximation is not valid for short distances between cells.
For these situations, however, the contribution of the vector potential is over-
shadowed by that of the scalar potential, so that the approximation of eqn. 8.98
still suffices. As a matter of fact, eqn. 8.98 may be used everywhere but on the
diagonal terms. This approximation has been confirmed by extensive numerical
tests.
A last point worth mentioning concerns the number of discrete Green's
functions that must be calculated. For a rectangular patch with rn x n charge
cells, the number of matrix elements is (M +N)2, with M = ( m - 1)n and
N = (n - I)m. When all the cells have identical sizes, only rn x n values of T,,
M values of Py and N values of Ti;Y are needed in order to completely fill the
matrix. This is the great advantage of using cells of equal size, and it is generally
more convenient to use a larger number of identical cells rather than fewer cells
of different sizes.
Interpolation among Green's functions: The evaluation of the matrix in eqn.
8.95 requires a large amount of computation. For a rectangular patch divided
into 10 x 10 cells, the order of the matrix is 180; hence the number of elements
in the matrix is 1802 = 32400. Even when a simple 4 x 4 Gaussian quadrature
is used to evaluate the discrete Green's functions, the number of Sommerfeld
integrals that need to be evaluated would exceed half a million.
Numerical analysis of microstrip patch antennas 429
Fortunately, for a given structure these integrals depend only upon the
distance from source to observer. It is thus possible to tabulate the integrals for
a small number of distances, and then to interpolate between the tabulated
values. The distances to be considered range from zero to the maximum linear
dimension of the patch. Several interpolation schemes have been tried [13]. The
best solution was obtained by separating the Green's functions according to
eqn. 8.89, and then using a simple parabolic Lagrange interpolation for the
regular part.
For a square patch with 10 x 10 cells, at frequencies for which the patch's
length is less than a free-space wavelength, the error obtained when interpolat-
ing from 25 tabulated values is hardly noticeable: less than 0.5%, even though
the computation time was reduced by a factor of loo!
8.6.2 Entire domain basis functions
If the microstrip patch has a simple regular shape (circular or rectangular) we
can consider the equivalent electromagnetic cavity obtained when the patch is
enclosed by a lateral magnetic wall. If the eigen modes have a simple analytical
expression it is reasonable to use them as a set of entire domain basis functions.
For thin substrates, the surface-current distribution on a microstrip patch at
resonance follows closely the behaviour of the corresponding eigenmode except
for a slight disturbance due to the antenna's excitation. Therefore, meaningful
results can frequently be obtained by using a very small number of entire
domain basis functions. This fact enables the analysis of microstrip arrays
having hundreds of elements. The size of the linear system to be solved will be
equal to two or three times the number of patches.
It is clear that if only one basis function is allowed per patch, we cannot use
a poor testing procedure such as point matching and match the boundary
condition only at the centre of the patch. The best alternative is provided by
Galerkin's method where the test functions are identical to the basis functions
and the inner product includes a surface integration over the patches [25].
To be clear, let us consider a single rectangular patch of dimensions L, W. The
eigen modes of the corresponding cavity are [28]
mxx nny mnx . nxy
1; = e,sin-cos-
L W
+ e,cos-sin-
L W
where the co-ordinate origin is in the lower left corner of the patch. The vectors
correspond to the modes TM,,, of the equivalent cavity. The choice of the
modes TM,,no in the expansion of J, is rather arbitrary and depends on the
problem considered. For instance, they can be ordered according to their
resonant frequencies, or we can consider only the TM,, subset if variations
along the co-ordinate y can be neglected. In any case, the relation between the
integers m, n and the ordinal number j in eqn. 8.99 should be clearly defined.
We assume now for the unknown surface current density the following
 430 Numerical analysis of microstrip patch antennas
expansion:
and consequently
Notice that here the unknown coefficients a, are interpreted as being amplitudes
of the surface current distribution [A/m] while, when using subsectional basis,
the unknowns were the total currents [A] flowing across contiguous cells.
If we test now the boundary condition 8.26 in the Galerkin sense, we get the
set of equations
The integral involving the scalar potential can be rewritten as
+
where the identity V . ( f ; V ) = f; . V V W .f; has been used.
Finally, introducing expansions 8.100 and 8.101 into eqn. 8.102 we get the
linear system
where
and
The last term in the expression of c, vanishes if i # j due to the orthogonality
of the eigenmodes f;.
It can be seen that, essentially, each matrix element requires the computation
of two fourfold integrals. Fortunately, two of the four integrations can be
performed analytically, by introducing the new variables u = x - x',
U' = x + x', v = y - y', v' = y + y'. As in the subsectional basis case, the
elements c,, will include a singularity when r = r' in the Green's functions.
Again, a decomposition of the type of eqn. 8.89 or a change to polar co-ordina-
tes may be used to eliminate the singularity.
Numerical analysis of microstrip patch antennas 437
The above techniques can easily be generalised to the analysis of an array of
patches. In this case, the domains of the ith and j t h modes do not necessarily
coincide. If the distance between the centres of the patches is greater than the
largest linear dimension of a single patch, we can use the approximation
where vi, rj denote the centres of the patches.
More accurate approximations can be obtained by expanding the Green's
functions in a Taylor series. These mathematical tools and powerful numerical
integration routines enable the total computation time to be kept within reason-
able limits.
8.7 Excitation and loading
In practice, microstrip antennas are excited by many different techniques. A
good survey of these has been recently given by Pozar [29]. This Section will
present a brief description of those most commonly found, keeping in mind that
the method of moments will be used to analyse the antenna.
8.7.1 Several microstrip-antenna excitations
A very common technique used for feeding a microstrip antenna consists of a
microstrip line directly connected to the patch. A thorough treatment of this
excitation requires the analysis of the incident and reflected quasi-TEM current
waves existing on the line, which is assumed to extend from the patch to infinity.
A more realistic model assumes that the microstrip line has a finite length and
introduces a mathematical excitation (for instance, a filament of vertical current
or a series voltage generator) at the end of the line. The vertical filament of
current is a good model for the coaxial excitation and on a physical basis is
preferred to the series voltage generator. The microstrip line is then cut at a
point where uniform line conditions can be assumed. In this way the discontinu-
ity created by the line-patch junction, and possibly discontinuities of the line
itself, are included in the analysis. If a vertical filament of unit current is used
as excitation, the input impedance is simply the voltage at the insertion point.
When using other mathematical excitations, the section of line included in the
analysis must be long enough to support a standing-wave pattern that can be
used to estimate the input impedance of the patch.
Finally, it is worth mentioning that replacing the microstrip line by a coaxial
probe at the edge of the patch is a first-order approximation that gives sur-
prisingly good results in many practical cases.
Microstrip line under the patch: A more sophisticated excitation, designed to
minimise the spurious radiation coming from the feed line, uses a microstrip line
under the patch to couple energy to it electromagnetically [29, 301. The presence
 432 Numerical analysis of microstrip patch antennas
of two dielectric layers adds an additional degree of freedom to the design. The
comments made above for the line directly connected to the patch also apply
here. The microstrip line can be cut far from the patch and terminated in a
coaxial probe or excited by some other mathematical device.
Slot in the ground plane: A slot in the ground plane can be used to couple
energy to the patch from, presumably, a triplate transmission line [31]. The
mathematical treatment replaces the slot by an equivalent distribution of sur-
face magnetic current. The excitation field E''' results from this current. In the
numerical procedure, this modifies only the independent terms b, (eqn. 8.106).
8.7.2 Coaxial excitation and input impedance
A coaxial line attached to the bottom of the patch (Fig. 8.17) is also a practical
way of feeding microstrip patches. From a theoretical point of view, the coaxial
excitation is of great interest because simple yet accurate mathematical models
are available. While the models presented here are constant current or voltage
sources and not constant power sources, that are actually used in practice, the
calculated values of the input impedance agrees closely with measured results.
Fig. 8.1 7 Microstrip patch antenna excited by a coaxial probe
The most rigorous model of this configuration considers the probe as belonging to
the patch and the whole structure excited by a frill of magnetic current. Simpler
models reduce the coaxial line to a filamentary current entering the patch.
The most accurate treatment of a coaxial probe [32, 331 assumes that the
portion of the inner coaxial conductor embedded in the substrate belongs to the
patch. The whole structure is then excited by a frill of magnetic current existing
Numerical analysis of microstrip patch antennas 433
between the inner and outer conductors of the coaxial line on the ground plane
(Fig. 8.17). For thin substrates, however, replacing the coaxial probe by a
vertical filament of current gives results accurate enough for engineering pur-
poses.
If the Galerkin technique is used, we can model the probe as a filament of zero
diameter ending in a point charge at the junction with the patch. The calculation
of the excitation fields would normally require a knowledge of the fields created
by vertical electric dipoles embedded in the substrate; however, the reciprocity
theorem allows the evaluation of the terms bi (eqn. 8.106) using only formulas
related to horizontal electric dipoles. Thus, we have
where E, is the field created by the surface current density J, = f, of eqn. 8.99
and E' is the field produced by the excitation current density Je [A/m2]entering
the feed point (x,, y,). The total excitation current is normalised to IA, i.e.
Je = e,6(x - x,)6(y - y,) -h < z < 0 (8.109)
and its domain V, reduces to a segment of length h in the case of zero-diameter
filament. Consequently:
Finally, in terms of the Green's functions we have
an expression which can be cast in an easier form as
with the modified scalar Green's function being given by
G*, = u,tanhuh/u DTM.
Now, once the vector of excitation terms b = (b,) is known, the amplitudes
of the basis functions are obtained by solving the linear system of equations
(eqn. 8.104). The input impedance is finally given by
When the subsectional basis functions are used together with razor test fun-
ctions, the coaxial probe must be modelled more carefully in order to avoid
mathematical singularities in the excitation terms. The total current entering the
probe must be spread over a region of the patch surrounding the insertion point.
A simplified attachment mode in which the current is spread over a cell with a
linear dependence has been developed and successfully tested [38]. In the frame-
 434 Numerical analysis of microstrip patch antennas
work of this model, the expression 8.108 for the excitation terms is still valid but
the excitation Ee is created by the currents belonging to the probe and to the
attachment mode. Also, the expression 8.1 13 for the input impedance can still
be used, but the effect of the attachment mode on itself must be added in order
to obtain accurate predictions for the reactance values. Finally, owing to the
discretisation inherent in this approach, the excitation point (x,, y e ) can only be
located at the centre of a charge cell.
The reactance of theprobe: Expression 8.11 3 gives the impedance at the patch
level, i.e. z = 0. To obtain the impedance at the ground plane level, z = - h,
we need to add in series the self-impedance of the coaxial probe. Assuming now
that the inner conductor of radius r, carries a current I evenly distributed on its
surface, we have
Numerical analysis of microstrip patch antennas 435
of the M-port. Therefore, the complete determination of the impedance matrix
requires the solution of M linear systems, but, fortunately, the matrix C of the
system remains unchanged when exciting each port.
The elements Z, have been termed 'input impedances' previously (eqns. 8.1 13
and 8.117); however, it must be pointed out that these are input impedances
corresponding to a single-port excitation since the remaining M-1 ports are
open-circuited. These impedances may be quite different from the true input
impedances for which an expression will be given now.
Let us consider that each port is connected to a voltage generator U,with an
internal impedance Z, (Fig. 8.18). This arrangement includes the case of a
passive load Z,, by allowing U, = 0.

For thin substrates, we can approximate the magnetic potential due to these I
currents by

u1
array or
multiport
and the self impedance is given by patch

UZ Ui

m
- a

This term is mainly inductive. Thus, finally, a better estimate of the input
impedance of a coaxial-fed antenna is
Z, = eqn. 8.113 + Z,,, (8.117)
Fig. 8.18 Equivalent circuit of a microstrip antenna array considered as a multiport device
8.7.3 Multipart analysis The voltage generators are replaced by short circuits at the ports terminated with a
In many practical situations the microstrip antenna is excited simultaneously at passive load
M points, for instance, in the case of a microstrip array. In this case the antenna
can be considered as an M-port device and standard circuit theory may be
applied to completely characterize the antenna. We define a vector U with elements U, and a diagonal load matrix Z , with
The first step is to solve the linear system Ccc = b, obtained by application of elements Z,,. The equations relating currents I, and voltages v a t each port (Fig.
the method of moments, M times for M different excitation vectors bi. These 8.18) are
vectors correspond to a physical situation in which a unit current is entering the u = Z,I + v = ( Z , Z ) I + (8.1 18)
j t h port while the remaining M-1 ports are open-circuited. After solving the
matrix equation, we get the vector a, = C-I b, containing the amplitudes of the where Z is the matrix of impedances previously calculated. The vector of port
N basis functions. Then, by computing the voltage at each port we get the currents is then given by
quantities Zv,i = 1, 2 . . . M, which is the j th column of the impedance matrix I = (Z, + z)-'U (8.1 19)
 436 Numerical analysis of microstrip patch antennas
,
and the vector of port voltages is V = Z(ZL + z)-I U . The new vector
N several resonances and the far-field radiation pattern.
a* = 1 I'ai (8.120)
,=I
contains the amplitudes of the basis functions for the real working conditions
of the antenna, i.e., with all the loads and excitations simultaneously present.
This is the vector to be used in the computations of the radiation pattern and
when studying the surface current distribution. Finally, the input impedance at
each port is
It is clear that, for a single-port antenna, this input impedance equals the
parameter Z , , which is directly given by eqns. 8.1 13 or 8.1 17.
Fig. 8.1 9 Geometry of a single rectangular patch
6, = 2.55 tan 8 = 0.002
h = 1.28 mm 2, = 0.9e-7
8.8 Single rectangular patch antenna
In previous Sections the mathematical theory and numerical procedures have
been developed for the analysis of general microstrip structures. This Section
will concentrate on a single, rectangular, coaxial-fed patch to illustrate how the
theory is applied, present computed results for a simple common structure and
answer some of the questions that remain. Remaining questions include the
convergence of the method-of-moments procedure and the advantages and
disadvantages of the various choices of basis functions presented in Section 8.6.
Results will be presented for a single patch as shown in Fig. 8.19 showing the
Numerical analysis of microstrip patch antennas 437
input-impedance loci on the Smith chart, the surface current distribution at
8.8.1 Entire-domain versus subdomain basis functions
Section 8.6 presented several choices of basis functions that could be used in a
MOM procedure. Among the choices were wide-triangle or rooftop subdomain
basis functions used with razor testing and entire-domain cosine basis functions
used in a Galerkin procedure, i.e. tested with cosine testing functions. Computer
programs have been written using these two choices of basis and testing
functions, and a comparison of the results obtained will be presented in this
Section.
The calculation of the far-field radiation pattern will be considered first. If
subdomain basis functions are used the patch is reduced to an array of
horizontal electric dipoles (HED). In this sense, each rooftop basis function is
equivalent to a HED whose moment is given by the product of the total current
flowing across the common border of two cells times the distance between the
centres of the cells.
Now, the far field for a horizontal dipole, which can be thought of as the
element pattern in this procedure, is multiplied by the array factor resulting
from the segmentation of the patch to give the total far field pattern. Mathemat-
ically, the far field is thus given by
M
Em = GF(rI0) aIx,exp(jk0e;g~)
i=I
+ GY(40) C bI, exp Woe,. Q;)
j= 1
(8.122)
where a = 0, cp, Li and I, ark the MOM current coefficients and G, represents
the far fields due to a HED. The pattern can then be integrated to calculate the
directivity, gain and efficiency.
When properly chosen entire domain basis functions are used, the far field for
each basis function can be calculated analytically, and the total far field is a
simple sum of the fields generated by each basis function. For other choices of
basis functions costly numerical integration techniques may be required. Fig.
8.20 shows the far-field pattern for a single rectangular patch. The subdomain
rooftop and entire-domain cosine basis functions yield identical co-polar pat-
terns at resonance while only the rooftop basis functions yield an estimate of the
cross-polar pattern. The cross-polar pattern is due mostly to currents on the
patch in phase with the excitation spreading out from the coaxial probe, and to
the currents in the probe. In general, the cross-polarised pattern is very difficult
to calculate accurately and is sensitive in practice to the size of the ground plane;
an item not included in the numerical model. The model can be used, however,
to study the effect of the placement of the coaxial probe on the cross-polar
pattern. Figs. 8 . 2 0 ~and 8.206 show the far-field pattern of a single patch when
excited by a coaxial probe located at different points.
 438 Numerical analysis of microstrip patch antennas Numerical analysis of microstrip patch antennas 439
The entire domain functions could, in theory, model accurately the patch at
frequencies away from resonance and its cross-polar behaviour, however, in
practice, this is costly. Near resonance only one or two entire domain functions
are needed to model surface current, and are therefore very well suited to
standard geometries near resonance. The cost to include additional basis fun-
ctions needed away from resonance or to calculate the cross-polar pattern is,
however, relatively high compared to the subdomain basis functions. Thus it
may be more efficient to use subdomain basis functions when studying a single
patch or small array away from resonance or when cross-polar pattern is
needed.

theta (degrees)
10 ,

Fig. 8.21 Input impedance for the single rectangular patch shown in Fig. 8.19.
Frequency range: 1.52-1.58 GHz.
Frequency increases clockwise with a 0.01 GHz step
Rooftop basis functions
0 Entire-domain cosine basis functions

The input impedance for a single patch is given on the Smith chart of Fig.
8.21. The two choices of basis functions yield approximately the same results
theta (degrees) with a slight shift in frequency. The resonant frequencies obtained differ by
Fig. 8.20 Radiation pattern for the single rectangular patch shown in Fig. 8.19. Fre- 0.77%, which is often much less than what arises due to uncertainties in the
quency = 1.565GHz.
A Coaxial feed at x = 16.66 mm. y = 16.66 mm
manufacturing process and material parameters. Fig. 8.22 shows the variation
B Coaxial feed at x = 16.66 mm. y = 20 mm of the real part of the input impedance as a function of position for three choices
. . . . .E-plane co-polar of widthlaspect ratios. In each case, the coaxial line was centered in the non-
H-plane co-polar resonant direction on the patch and then moved inward from the edge of the
---- E-plane cross-polar
--- H-plane cross-polar
patch to the centre. Note that the results were calculated at the resonant
frequency for each patch.
 440 Numerical analysis of microstrip patch antennas
8.8.2 Convergence using subsectional basis functions
The question of convergence must always be dealt with when using a moment
method. If a numerical result has not converged there is virtually no hope of its
being correct. The factors affecting convergence include: the choice of basis and
testing functions, frequency, antenna shape, the dielectric used and even the
numerical precision of the computer. Since this list includes nearly all the
parameters of the antenna and affects nearly all of the decisions made during the
development of the computer program to some degree, it is difficult to give rules
that guarantee that a particular result has converged. However, several rules of
thumb are applicable and can be used as a base when studying convergence.
This Section will briefly demonstrate how the MOM solution using subsectional
basis functions converges.
10
Fig. 8.22 Real part of the input impedance as a function of the position on the patch. The
coaxial is centered along the non-resonant direction and moved from the edge of
the patch ( x l L = 0 ) to the centre of the patch ( x l L = 0.5).
For case ( a ) the antenna parameters are given in Fig. 8.1 9 where L = 60mm and
W = 40mm. For cases ( 6 ) and ( c ) the width W is varied.
( a ) . . . . . L I W = 1.5 f,,,,,, = 1.555 GHz
(6)----LIW=l.O f,,,,,,=1.543GHz
( c ) --- L I W = 0.667 f,,,,,, = 1.535 GHz
The general rule of thumb given in the published literature is that, when using
subsectional basis functions, of the order of 10 basis functions per wavelength
are needed to obtain good results. This rule also holds for microstrip antennas
operating near the first few resonances when calculating the input impedance.
It is interesting to note, though, that the number of basis functions in the
Numerical analysis of microstrip patch antennas 441
direction transverse to the resonance direction, i.e. parallel to the H plane at the
first resonant frequency, may be reduced without a significant penalty, resulting
in large savings in computation time. Fig. 8.23 shows how the input impedance
of a rectangular patch converges when the number of basis functions in the
H-plane direction is varied at frequencies near the first resonance. Note the
resonant frequency changes by only 0.3% when using three basis functions in
the transverse direction as opposed to using 7, however, the input impedance
changes by approximately 20%. Thus, a rough study of the antenna's resonant
frequency and radiation pattern can be performed quickly at low cost. The final
analysis can then be performed using additional basis functions.
Fig. 8.23 Input impedance for the single rectangular patch shown in Fig. 8.19 with the
coaxial centered vertically (y = 20mm) and at x = 76.66mm versus the number of
basis functions in the H-plane direction, i.e. along the y-axis
Frequency range is from 1.52 GHz increasing clockwise to 1.60 GHz with a step of
0.01 GHz
+: 9 by 7 cells
0: 9 by 5 cells
': 9 by 3 cells
8.8.3 Surface currents
The subsectional basis functions can be used to model virtually any current
distribution owing to their flexibility. As an example, a rectangular patch was
analysed and measured at the first four resonances, TM,,, TM,,,, TM,,, and
TM,,. The patch dimensions are 60 x 40mm and the dielectric is a standard
low-frequency printed-circuit substrate with high losses in the microwave range
(E, = 4.34 and tans z 0.02). The excitation point has been selected at
 444 Numerical analysis of microstrip patch antennas
induces currents in its neighbours affecting the element's radiation pattern and
input impedance. Thus, elements in an array environment have to be studied in
the actual array environment to properly account for mutual coupling [34, 351.
However, when the coupling between array elements is less than 20 or 30dB, it
may be possible to neglect mutual coupling and still obtain acceptable results.
Fig. 8.25 Input impedance near the two first resonances of the patch of Fig. 8.24.After [38].
0 Theory (9 x 6 cells)
Measurements.
This section will present results for several small arrays and show how the
element factor and input impedance are affected when an isolated element is
incorporated in an array.
8.9.1 Array modelling
To accurately model an array the model should not assume any particular
Numerical analysis of microstrip patch antennas 445
current distribution on the elements. The method of moments is well suited for
the task because any current distribution can be calculated to the desired
precision and coupling coefficients are easily calculated. However, practical
limitations exist when applying the MOM to anything but small arrays. The use
of subsectional basis functions, while very flexible in modelling arbitrary geome-
tries and current distributions, comes with a high price in the number of
unknowns. To accurately model an element of the array approximately 50
unknowns are needed; thus for a linear array 10 elements long, 500 unknowns
are needed. It can be seen that the capacity of even the largest currently available
computer is quickly surpassed.
There are, however, several techniques that can be used to study larger arrays
without simply using a larger computer or more computer time. These techni-
ques include:
(a) Entire domain basis functions
(b) Infinite array techniques
Additionally, specialised numerical techniques may be applied with success in
certain cases. The method of conjugate gradients has been proposed as a method
that would allow the solution of larger systems of linear equations [36, 371.
However, normally the MOM matrix is not sparse and the slow convergence of
iterative routines applied to fully populated linear systems precludes their use.
Of the two techniques discussed above only the use of entire-domain basis
functions will be discussed since infinite array techniques are included as a full
Chapter of this handbook. Using entire-domain basis function only one or two
basis functions are typically needed at resonance per element; so arrays having
up to several hundred elements can be studied easily with today's computers.
However, the study of circularly polarised elements or the cross-polarised fields
requires the use of additional higher-order basis functions. This considerably
increases the computation time and reduces the size of largest array that can be
studied.
8.9.2 Mutual coupling
A 2 x 2 element array was build on a lossy, inexpensive substrate and the four
elements of the scattering matrix were measured and compared with the results
obtained using the numerical procedures presented previously in this Chapter.
The scattering matrix is defined by
where Z i s the impedance matrix with elements calculated using eqn. 8.1 17 and
I is the unit dyadic.
The array consisted of four identical patches, each 60 mm along the E-plane
and 40 mm along H-plane. The substrate thickness was 0.8 mm and the dielectric
constant was 4.34 with a loss tangent of 0.02. The elements were coaxially fed
with the coaxial line centered along the H-plane and located lOmm from the
446 Numerical analysis of microstrip patch antennas Numerical analysis of microstrip patch antennas 447

-1 8
1.10 1.14 1.18 1.22 1.26 1.30
-44 1
1.10 1.14 1.18 1.22 1.26 1.:
frequency (GHz) frequency (GHz)

1.10 1.14 1.18 1.22 1.26 1.30

frequency (GHz)
-60 I , , , , , , , , ,
1.10 1.14 1.18 1.22 1.26 1.:
Fig. 8.26 Scattering parameters for a 2 x 2 microstrip array: measured versus theory frequency (GHz)
Patch size: 60 mm by 40mm, h = 0.8 mm
E, = 4.34 tan 6 = 0.02
E-plane separation 20mm between patch edges
H-plane separation 1 6 mm between patch edges
(8) S I ~ ( 6 ) s12
( c ) S13 ( d ) Sll
---- Measured . . . . .Theory
 448 Numerical analysis of microstrip patch antennas
edge. The separation between the patches was 20mm along the E-plane and
16 mm along the H-plane, corresponding, respectively, to 0.08 and 0.064 free-
space wavelengths at the isolated patch's resonance of 1.2 GHz. The reslllts are
shown in Fig. 8.26. The two algorithms, which use different basis fmctions,
yield nearly identical results across the frequency band. As can 'se seen, the
agreement between the measured and calculated results is good.
Fig. 8.27 Measured (Jedlicka, Poe and Carver [35]) and calculated mutual coupling bet-
ween two coaxial-fed rnicrostrip antennas versus the separation between the patch
centres D measured in free-space wavelengths
W = 105.7mm L = 65.5mrn h = l . 6 m m
E, = 2.53 f, = 1.414GHz
Measured E-plane (From Reference 35)
0 Measured H-plane (From Reference 35)
. . . . . Calculated E-plane
---- Calculated H-plane
--- Calculated 45' plane
Fig. 8.27 gives the E- and H-plane coupling results as a function of the
distance between two patches measured by Jedlicka, Poe and Carver [35], and
compares these results with those calculated using the theory presented above.
Entire-domain basis functions were used. In addition, the calculated coupling
results between two antennas located along a diagonal are also presented. It is
readily apparent that the H-plane coupling is stronger for small separations.
This is the kind of coupling found in standard microstrip coupled line filters,
and it is mainly due to quasi-static terms and to the space wave. However, the
Numerical analysis of rnicrostrip patch antennas 449
H-plane coupling decreases faster than the E-plane coupling, which is sustained
by the surface wave and becomes dominant for greater separations.
It must be finally pointed out that the values of the coupling (s,, parameter)
are dependent on the input impedance (z,, parameter) and are usually given for
a couple of patches matched to 500. The small differences with measurements
in Fig. 8.27 can be due to a slight mismatch of the patches.
8.9.3 Linear array of four patches
The last example in this Chapter will illustrate, from a theoretical point of view,
the relevance of mutual coupling in the computation of input impedances and
mutual coupling, The selected configuration is shown in Fig. 8.28 and consists
of four rectangular identical patches fed by coaxial probes and working in the
lowest-order mode. The substrate parameters are h = 0.787 mm, and E, = 2.23.
The patches are coupled by their non-resonant sides (H-plane coupling) and
they are excited uniformly. On the other hand, their spacing is non-uniform in
order to obtain a lower first-sidelobe level [39].
H
3rnm
t 4 4;;1 t
10.3
rn rn
10.3
rnrn
-
h= 0.787rnrn
f = 11.9GHz
relative perrnitivity=2.23
Fig. 8.28 Linear array with four identical patches and nonuniform spacing
The substrate has a permittivity of E, = 2.23 and a thickness of h = 0.787mm. The
nominal resonant frequency is f,, = 11.9GHz.
Fig. 8.29 shows the normalised real and imaginary parts of the input im-
pedance presented by an inner patch. For each of the quantities, two curves are
given, corresponding to theoretical calculations without mutual coupling (the
patch is considered as an isolated element) and with mutual coupling (the patch
is embedded in an array environment with the three other patches terminated
by 50R loads). It can be seen that, for this array, mutual coupling raises the
maximum resistance from 130R to 154R. This significant change (18%) shows
clearly that mutual coupling should be taken into account for properly matching
a microstrip array.
The influence of mutual coupling on radiation patterns is even stronger. Fig.
8.30 shows the H-plane radiation patterns of an isolated patch and of an inner
patch in an embedded configuration. The asymmetries in the geometrical en-
vironment of an inner patch (coupled to two patches on one side but only to one
450 Numerical analysis of microstrip patch antennas Numerical analysis of microstrip patch antennas 45 7

o! , , , , , , , , , I
11.0 11.4 11.8 12.2 12.6 13.0
frequency (GHz)

Fig. 8.30 Calculated isolated-element and embedded-inner-element far-field radiation pat-

terns for the array of Fig. 8.28
I n the embedded case all the other elements are loaded with 50R
--- Isolated element
. . . . . Embedded element

-0.75 i---7-TT
, , , , , ,
11.0 11.4 11.8 12.2 12.6 13
frequency (GHz)

Fig. 8.29 Calculated isolated-element and embedded-inner-element input impedance versus

frequency for the array of Fig. 8.28. Values are normalised to 50R
a Real part
theta (degrees)
b Imaginary part
. . . , . Isolated element Fig. 8.31 Calculated H-plane far-field pattern of the array shown in Fig. 8.28 with and
--- Embedded element without mutual coupling
. . . . . With mutual coupling (MOM result)
First sidelobe = -1 3 S d B
--- Without mutual coupling
First sidelobe = - 18.6 dB
452 Numerical analysis of microstrip patch antennas
on the other side) are clearly reflected in the asymmetrical pattern for the
embedded situation.
Finally, Fig. 8.31 gives the overall radiation pattern of the array in the
H-plane. A first theoretical prediction neglects mutual coupling and computes
the array pattern as the product of the isolated element pattern times the array
factor. The first-order sidelobe is then at - 18.6dB, well below the value of
- 13.2dB which can be obtained with four equally spaced, uniformly fed
elements [40]. However, if the integral-equation model presented in this Chapter
is used, mutual coupling is automatically taken into account. Theoretical predic-
tions show then how mutual coupling deteriorates the radiation pattern, raising
the first-sidelobe level to only - 13.8dB.
8.10 Acknowledgments
The authors would like to thank the members of the Laboratoire d'Electro-
magnQisme et dlAcoustique (LEMA) for their aid and support. Specifically,
Anja Skrivewik, Lionel Barlatey and Bertrand Roudot performed many of the
calculations and measurements presented in this Chapter. Also special thanks
are due to Mrs. Mary Hall for typing the manuscript.
8.11 References
1 BAHL, I. J., and BHARTIA, P.: 'Microstrip antennas' (Artech House, 1980).
2 STRATTON, J. A,: 'Electromagnetic theory' (McGraw-Hill, NY, 1941).
3 MICHALSKI, K. A.: 'On the scalar potential of a point charge associated with a time-
harmonic dipole in a layered medium', IEEE Trans., 1988, AP-36.
4 POGGIO, A. J., and MILLER, E. K.: 'Integral equation solutions of three-dimensional
scattering problems', In M1TTRA;R. (Ed.), 'Computer Techniques for Electromagnetics'
(Pergamon Press, 1973).
5 LO, Y. T.: 'Electromagnetic field of a dipole source above a grounded dielectric slab', J. Appl.
Phys., 1954, 25, p. 733.
6 BRICK, D. B.: 'The radiation of a Hertzian dipole over a coated conductor', Proc. IEE, 1955,
102C, pp. 103-121.
7 BREKHOVSKIKH, L. M.: 'Waves in layered media' (Academic Press, NY, 1960).
8 WAIT, J. R.: 'Electromagnetic waves in stratified media', (Pergamon Press, 1962).
9 FELSEN, L. B., and MARCUVITZ, N.: 'Radiation scattering of waves', (Prentice Hall, New
Jersey, 1973).
10 KONG, J. A,: 'Theory of electromagnetic waves', (Wiley, NY, 1975).
I I SOMMERFELD, A.: 'Partial differential equations in physics' (Academic Press, NY 1949).
12 HARRINGTON, R. F.: 'Time harmonic electromagnetic fields' (McGraw-Hill, NY, 1961).
13 MOSIG, J. R., and GARDIOL, F. E.: 'A dynamical radiation model for microstrip struc-
tures'. in HAWKES, P. (Ed.) 'Advances in electronics and electron physics', (Academic Press,
NY, 1982) pp. 139-237.
14 MOSIG, J. R., and GARDIOL, F. E.: 'Analytic and numerical techniques in the Green's
function treatment of microstrip antennas and scatterers', IEE Proc., 1983, 130H, pp. 175-182.
15 SHASTRY, S. V. K.; Ph.D. Dissertation, Indian Institute of Science, Bangalore, India, 1979.
Numerical analysis of microstrip patch antennas 453
UZUNOGLU, N. K., ALEXOPOULOS, N. G., and FIKIORIS, J. G.: 'Radiation properties
of microstrip dipoles', IEEE Trans., 1979, AP-27, pp. 853-858.
MOSIG, J. R., and GARDIOL, F. E.: 'Dielectric losses, ohmic losses and surface wave effects
in microstrip antennas.' Int. URSI Symposium, Santiago de Compostela, Spain, 1983, pp.
425-428.
MICHALSKI, K. A,: 'On the efficient evaluation of integrals arising in the Sommerfeld
halfspace problem', IEE Proc., 1985, 132H, pp. 312-318.
LYTLE, R. J., and LAGER, D. L.: 'Numerical evaluation of Sommerfeld integrals'. Report
UCRL-52423, Lawrence Livermore Lab., Univ. of California, 1974.
PANTIS, G.: 'The evaluation of integrals with oscillatory integrands', J. Comp. Phys., 1975,
17, pp. 229-233.
BORIS, J. P., and ORAN, E. S.: 'Evaluation of oscillatory integrals', J. Comp. Phys., 1975,
17, pp. 425-433.
HURWITZ, H., and ZWEIFEL, P. F.: 'Numerical quadrature of Fourier transform integrals',
Math. Tables Aids Cornput., 1956, 10, pp. 140-149.
ALAYLIOGLU, A., EVANS, G., and HYSLOP, J.: 'The evaluation of oscillatory integrals
with infinite limits,' J. Comp. Physics, 1973, 13, pp. 433-438.
SIDI, A,: 'The numerical evaluation of very oscillatory infinite integrals by extrapolation',
Math. Comp., 1982, 38, pp. 517-529.
HARRINGTON, R. F.: 'Field computation by moment methods' (MacMillan, NY, 1968).
RAO, S. M., WILTON, D. R., and GLISSON, A. W.: 'Electromagnetic scattering by surfaces
of arbitrary shape', IEEE Trans., 1982, AP-30, pp. 409-418.
GLISSON, A. W., and WILTON, D. R.: 'Simple and efficient numerical methods for problems
of electromagnetic radiation and scattering from surfaces', IEEE Trans., 1981, AP-29, pp.
593-603.
COLLIN, R. E.: 'Foundations for microwave engineering', (McGraw-Hill, NY, 1966).
POZAR, D. M.: 'New architectures for millimeter wave phased array antennas', Journees
Internationales de Nice sur les Antennas (JINA), Nice, 1986, pp. 168-179.
KATEHI, P. B., ALEXOPOULOS, N. G.: 'On the modelling of electromagnetic coupled
microstrip antennas - The printed strip dipole,' IEEE Trans. Antennas Propagat., AP-32, pp.
1179-1 186, 1984.
SULLIVAN, P. L. and SCHAUBERT D. H., 'Analysis of an aperture coupled microstrip
antenna,' IEEE Trans. Antennas Propagat., AP-34, pp. 977-984, 1986.
POPOVIC, B. D., DRAGOVIC, M. B., and DJORDJEVIC, A. R.: 'Analysis and synthesis
of wire antennas', (Wiley, NY, 1982).
HALL, R. C., MOSIG, J. R., and GARDIOL, F. E.: 'Analysis of microstrip antenna arrays
with thick substrates', 17th European Microwave Conf., Rome, Italy, 1987.
ALEXOPOULOS, N. G., and RANA, I. E.: 'Mutual impedance computation between pr~nted
dipoles', IEEE Trans., 1981 AP-29, pp. 106-1 11.
JEDLICKA, R. P., POE, M. T.. and CARVER, K. R.: 'Measured mutual coupling between
microstrip antennas', IEEE Trans., 1981, AP-29, pp. 147-149.
JACOBS, D. A. H.: 'A generalization of the conjugate-gradient method to solve complex
systems', IMAJ of Numerical Analysis, 1986, 6, pp. 447-452.
PETERSON, A. F., and MITTRA, R.: 'Method of conjugate gradients for the numerical
- -
solution of large-body electromagnetic scattering problems', J. Opt. Soc. Am. A, 1985, 2, pp.
971-977.
MOSIG, J. R., and GARDIOL, F. E.: 'General integral equation formulation for microstrip
antennas and scatterers', IEE Proc., 1985, 132H, pp. 424-432.
GRONAU, G., and WOLFF, I.: 'Spectral domain analysis of microstrip antennas', Proc. of
Workshop 'Analytical and numerical techniques for microstrip circuits and antennas', Mon-
treux, Switzerland, March 1988.
BALANIS, C. A,: 'Antenna theory: analysis and design', (Harper & Row, NY, 1982).
 Chapter 9

Multiport network approach for

modelling and analysis of
microstrip patch antennas and
arrays
K. C. Gupta

9.1 Introduction

The multiport network approach for microstrip patch antennas is based on the
use of segmentation and desegmentation methods for analysis of planar struc-
tures. Segmentation and desegmentation techniques were developed originally
for analysis of two-dimensional (planar) circuit components [I-7.Since micro-
strip patch antennas on thin substrates can be treated as two-dimensional planar
components, segmentation and desegmentation techniques have been employed
for the analysis of microstrip antennas also [8, 91. This approach has been used
successfully for the analysis and design of several types of microstrip patch
antennas [lo-171 and arrays [18], and promises to be an appropriate methodol-
ogy for computer-aided design [19,20] of microstrip patch antennas and arrays
in hybrid as well as in monolithic configurations.
This Chapter describes the multiport network approach, segmentation-
desegmentation techniques, and their applications to design of microstrip anten-
nas. Relevant aspects of various models for microstrip antennas are presented
in Section 9.2. The multiport network model [19], which is an extension of the
well known cavity model [21,22] for microstrip patches, is discussed. Evaluation
of multiport impedance matrices from the Green's functions for various types
of segments is described in Section 9.3. Modelling of external fields (including
fringing, radiation and surface wave) by edge-admittance networks is detailed
in Section 9.4.
Segmentation and desegmentation methods for analysis of planar electro-
magnetic structures (and for multiport networks) are discussed in Section 9.5.
Various examples of microstrip antenna configurations, which have been
analysed and designed using multiport network approach, are reviewed in
Section 9.6. These include various types of single-feed circularly-polarised
microstrip patch configurations, such as a diagonally-fed nearly square patch,
.a square patch with truncated comers, a square patch with a diagonal slot, a
pentagonal-shaped patch, a square ring patch, and a cross-shaped patch. The
second group of antennas analysed by the multiport network approach consists
456 Multiport network approach for modelling Multiport network approach for modelling 457
of broadband multi-resonator microstrip antennas. Three-resonator and five- ports are located along the non-radiating edges, transmission from port 1 to
resonator configurations (coupled to a central patch either by a capacitive-gap port 2 can be controlled by suitable choices of distances x, and x,. Again, the
coupling or by short sections of microstrip lines) are included in this group. The two models shown in Figs. 9.2a and b do not incorporate the parasitic reactan-
third category of microstrip antenna configurations, discussed in Section 9.6.3, ces associated with the feed-line-patch junctions.
consists of two-port rectangular, two-port circular patches, and series-fed arrays
making use of these two-port patches.
Discussion related to the development of CAD procedures for microstrip
patches and arrays is contained in Section 9.7. It is pointed out that multiport
network modelling and segmentation/desegmentation methods of analysis are
ideally suitable for implementation of CAD procedures for microstrip antennas.

9.2 Models for microstrip antennas

The transmission-line model [23, 241 and the cavity model [21, 221 are the two
most widely used network models for analysis of microstrip antennas. We will a unloaded patch
discuss these models briefly before introducing the multiport network model
suitable for implementing segmentation/desegmentation methods.

9.2.1 Transmission-line model

In this model, a rectangular microstrip antenna patch is viewed as a resonant
section of a microstrip transmission line. A detailed description of the trans-
mission-line model is given in Chapter 10. The basic concept is shown in Fig.
9.1 which illustrates the transmission-line models for (a) and unloaded rectan-
gular patch; (b) a rectangular patch with a feed line along the radiating edge; and
(c) a rectangular patch with a feed line along the non-radiating edge. Z,, is the
characteristic impedance of a microstrip line of width W,, and E , is the corres-
ponding effective dielectric constant. Be and G, are capacitive and conductive b feedline along the radiating edge
components of the edge admittance Y,. The susceptance B, accounts for the
fringing field associated with the radiating edge of the width W,, and G, is the
conductance contributed by the radiation field associated with each edge. Power
carried away by the surface wave(s) excited along the slab may also be represen-
rn
ted by a lumped loss and added to G,. In Fig. 9. lb and c, Zof and are the
characteristic impedance and the effective dielectric constant for the feeding
microstrip line of width Wf. In both of these cases, the parasitic reactances
associated with the junction between the line and the patch have not been taken
into account.
Transmission-line models may also be developed for two-port rectangular
microstrip patches [14]. These configurations are used in the design of series-fed c feedline along the non-radiating edge
linear (or planar) arrays [25, 181. Models for two types of two-port rectangular Fig. 9.1 Transmission-line models for three rectangular microstrip patch configurations
microstrip patches are shown in Fig. 9.2. Fig. 9.2a illustrates the equivalent
transmission-line network when the two ports are located along the radiating There are several limitations inherent to the concept of the transmission-line
edges, and Fig. 9.26 shows the transmission-line model [26] when the two ports model for microstrip antennas. The basic assumptions include: (i) fields are
are along the non-radiating edges. It has been shown [14,26] that, when the two uniform along the width W, of the patch; and (ii) there are no currents transverse
 458 Multiport network approach for modelling Multiport network approach for modelling 459

to the length I of the patch. Detailed analysis of rectangular patches has shown
[27] that, even at a frequency close to the resonance, field distribution along the
radiating edge is not always uniform. Also, the transverse currents are caused
by the feeding mechanism and are invariably present. Moreover, the circularly
polarised rectangular microstrip antennas (whose operation depends upon the
excitation of two orthogonal modes) cannot be represented by the transmission-
line model discussed above. Clearly, a more accurate method for modelling of
microstrip antennas is needed.

a feedlines olong radioting edges

b feedlines olong non-radioting edges

Fig. 9.2 Transmission-line modes for two-port rectangular microstrip patch antennas

9.2.2 Cavity model

A planar two-dimensional cavity model for microstrip patch antennas [21, 221
offers considerable improvement over the one-dimensional transmission-line
model discussed in the previous Section. In this method of modelling, the
microstrip patch is considered as a two-dimensional resonator surrounded by a
460 Multiport network approach for modelling Multiport network approach for modelling 461
perfect magnetic wall around the periphery. The fields underneath the patch are
expanded in terms of the resonant modes of the two-dimensional resonator.
This approach is applicable to a variety of patch geometries. These geometries,
the corresponding modal variations denoted by $ ., and the resonant wave
numbers k,, are shown in Table 9.1 (from Reference 21). E and H fields are
related to ,$, by
Ern" = *d (9.1)
K . = i x V,rCl,liw (9.2)
where i is a unit vector normal to the plane of the patch. Resonant wave
numbers k,,, are solutions of
(Vf + en)*,= 0 (9.3)
with

on the magnetic wall (periphery of the patch). V, is the transverse part of the del
operator and p is perpendicular to the magnetic wall.
The fringing fields at the edges are accounted for by extending the patch
boundary outwards and considering the effective dimensions to be somewhat
larger than the physical dimensions of the patch. The radiation is accounted for.
by considering the effective loss tangent of the dielectric to be larger than the
actual value. If the radiated power is estimated to be P,,the effective loss tangent
6, may be written as

where Pd is the power dissipated in the dielectric substrate and 6, is the loss
tangent for the dielectric medium. The effective loss tangent given by eqn. 9.5
can be modified further to incorporate the conductor loss. The modified loss
tangent 6, is given by

The input impedance of the antenna is calculated by finding the power diss-
ipated in the patch for a unit voltage at the feed port, and is given by

where P = P, + PC + P,, + P,.WEis the time-averaged electric stored energy,

and W, is the time-averaged magnetic energy. The voltage V equals Ezd av-
eraged over the feed-strip width (d is the substrate thickness). The far-zone field,
and radiated power are computed by replacing the equivalent magnetic-current
 462 Multiport network approach for modelling
ribbon on the patch's perimeter by a magnetic line current of magnitude Kd on
the ground plane (xy plane). The magnetic current source is given by
where ii is a unit vector normal to the patch's perimeter and iE(x, y) is the
component of the electric field perpendicular to the ground plane.
A cavity model for microstrip patch antennas may also be formulated by
considering a planar two-dimensional resonator with an impedance boundary
wall all around the edges of the patch. A direct form of network analogue
(DFNA) method for the analysis of such a cavity model has been discussed in
Reference 28.
9.2.3 Multiport network model
The multiport network model of microstrip patch antennas [19, 291 may be
considered as an extension of the cavity model discussed above. Electromagnetic
fields underneath the patch and outside the patch are modelled separately. The
patch itself is analysed as a two-dimensional planar network [I], with a multiple
number of ports located all around the edges as shown in Fig. 9.3. Each port
represents a small section (of length v ) of the edge of the patch. is chosen
so small that the fields over this length may be assumed to be uniform. Typically,
for a rectangular patch, the number of ports along each radiating edge is taken
to be 4, and along each non-radiating edge the number is taken to be eight.
Fig. 9.3 Multiport representation of a rectangular patch
Thus, a 24 x 24 matrix is typically adequate for the characterisation of the
interior fields of a rectangular patch.
For patches of regular shapes (rectangles, circles, rings, sectors of circles and
of rings, and three types of triangles), this multiport planar network model can
be analysed by using two-dimensional impedance Green's functions available
for these shapes [I]. A multiport Z-matrix characterisation representing the
fields underneath the patch can be derived from the Green's function as:
Z, = w1y [&JY
- ~ ( x lY, ~ I ~j)dsids,
X~, (9.9)
where x,,, y,, denote the locations of the two ports of widths
and y,
Multiport network approach for modelling 463
respectively. Green's function G is usually a doubly infinite summation with
terms corresponding to various modes of the planar resonator (rectangular or
circular or triangular) with magnetic walls.
Fig. 9.4 (a) A cross-shaped microstrip patch. (6) Multipart-network model of the cross-
shaped microstrip patch
For patches of composite shapes (such as a cross shape shown in Fig. 9.4a),
a multiport network model can be written by treating the composite shape as a
combination of the elementary shapes for which Green's functions are available.
The cross shape of Fig. 9 . 4 ~can be considered as a combination of three
rectangular segments as shown in Fig. 9.46. Segmentation and desegmentation
methods [I] are used for finding the 2-matrix of a composite shape from those
of the elementary segments. If the patch or one of the segments of a composite
patch is of an irregular shape for which Green's function is not available, a
technique called the contour integral method [I] can be used to evaluate the
Z-matrix.
In the multiport-network modelling of radiating microstrip patches, the fields
outside the (namely, the fringing fields at the edges, the surface-wave fields
and the radiation field) are incorporated by adding equivalent edge admittance
networks (EAN) connected to the various edges of the patch. This representa-
tion is shown in Fig. 9.5 for the case of the rectangular patch shown in Fig. 9.3.
EANs are multiport networks consisting of parallel combinations of the capa-
citances C (representing the energy stored in the fringing field) and the conduc-
tances G (representing the power carried away by radiation and surface waves)
as shown in Fig. 9.6a. Each capacitance-conductance pair is connected to a port
of the planar equivalent circuit of the patch. EANs at the non-radiating edges
may be simplified to consist of capacitances only, as shown in Fig. 9.6b. Values
of capacitance and conductance in the edge-admittance networks may be ob-
tained from the various analyses reported in the literature [30-321.
The flexibility of the multiport network model leads to several advantages
when compared with the conventional cavity model discussed in Section 9.2.2.
For example, the parasitic reactances at the junction between the feed line and
the patch can be incorporated in the multiport network model by considering
 464 Multiport network approach for modelling Multiport network approach for modelling 465
a small section of the feed line as an equivalent planar circuit connected to the current sources at the two corresponding sections of the edges. Similar MCNs
patch at a finite number of (typically five) ports. Solution of this network may also be included between the non-radiating edges, between a radiating edge
problem (depicted in Fig. 9.7) is equivalent to the expansion of the fields (in the and a non-radiating edge, or between two edges of different patches in an array.
feed line as well as in the patch) in series of eigen
- functions and matching the
fields at the interface.

NR-EAN

NR-EAN NR-EAN

Fig. 9.7 lncorporation of feed-junction reactance in multipart-network model of a rectan-

gular patch

Fig. 9.5 Edge-admittance networks (EANs) connected to the multiport representation of a

rectangular patch

Fig. 9.6 Edge admittance networks for (a) radiating edges; (b) non-radiating edges
Also, the multiport network model discussed above can be extended to
incorporate the effect of mutual coupling between the two radiating edges [33]
by inserting a mutual coupling network (MCN) as shown in Fig. 9.8. The
edge-admittance terms associated with various ports at the edges constitute the
diagonal terms of the admittance matrix for MCN. The non-diagonal terms of
this matrix are obtained from the 'reaction' between the equivalent magnetic
R-EAN
I
Fig. 9.8 Incorporation of mutual coupling in multipart-network model of a ~ectangularpatch
It may be noted that, in the multiport network model, the characterisation of
fields underneath the patch is conceptually similar to that used in the conven-
tional cavity model [21, 221. In both of these models, the fields under the patch
 466 Multiport network approach for modelling
are considered two-dimensional with no variations of fields perpendicular to the
substrate. For this reason, the limits of the applicability of the technique in
terms of substrate permittivity and thickness are similar to that for the cavity
model. Also, both of these methods will not be accurate when applied to
narrow-width microstrip dipoles rather than to the wide microstrip patches
discussed in this Chapter.
y 4 Y A
Fig. 9.9 Various geometries of the planar segments for which Green's functions are available
Green's functions used for evaluating impedance-matrix characterisation for
patches of various shapes are discussed in Section 9.3. Derivation of the Z-ma-
trix is also included therein.
Multiport network approach for modelling 467
9.3 Z-matrix characterisation of planar segments
9.3.1 Green's functions
For practical microstrip antennas, the thickness of the substrate is much smaller
than the wavelength. Therefore fields underneath the patch do not vary in the
z-direction (perpendicular to the substrate). Electric field has a z-component
only. Since aE,/az = 0, we may define a voltage V(x,y) given by
(9.10)
V(x,Y ) = - Ez(x, y)d
where d is the substrate thickness. When a magnetic-wall boundary condition is
assumed at the edges of the patch, V(x,y) satisfies the boundary condition given
by eqn. 9.4, i.e.
If we consider a z-directed electric current source Jx(xo,yo) located at (xo,) ,),
the voltage V(x,y) is related to the source current through a two-dimensional
impedance Green's function G(x, ylx,, yo) defined by
microstrip patch antenna
I I
top view
&coax
connector
side view
Fig. 9.10 Microstrip patch antenna with a probe feed perpendicular to the substrate
where the source current J, is distributed over a region D in x, y plane. These
Green's functions are known [I] for several regular shapes shown in Fig. 9.9.
Expressions for these Green's functions are listed in Appendix 9.8.
When a microstrip antenna is excited by a probe feed perpendicular to the
 468 Multiport network approach for modelling Multiport network approach for modelling 469
substrate, as shown in Fig. 9.10, the current density may be related to the axial and involves integration of G(x, ylx,, yo) over the extent of the two-ports
current through the z-directed probe. For patches excited by a microstrip line corresponding to the specific element of the Z-matrix. For a microstrip line feed
feed, the current J,,flowing into the patch can be expressed as an equivalent with an effective width y, the integral is carried out over the width 4. For a
the patch (as shown in Fig. 9.1 I),
-
z-directed electric current sheet J, as follows. At the magnetic wall surrounding probe-feed type of external port, the integration is carried out over a circular
path corresponding to the cylindrical surface of the probe. Alternatively, the
J, = B x H,
circular probe maybe replaced by an equivalent strip and the integration carried
(9.13)
over this equivalent width.

Z-matrix for rectangular segments: Green's functions for various geometries,

discussed in Appendix 9.8, appe8r as double-infinite summations. In numerical
computations of the 2-matrix elements, order of integration and summations
could be interchanged. For rectangular segments, the integrals involved may be
carried out analytically. When sides of the rectangle are oriented along x- and
y-axes and for the two ports (say, port p and port q), the impedance-matrix
element Z,, may be written in the following form [37]:

. dJm,(~q,~,)l(k2x+ k: -
4'
where, for ports oriented along the y-direction,

feedline
Jin

H, :!tj, patch
6, (x, y) = cos (k,x) cos (kyy) sinc (Y)
and for ports oriented along the x-direction

Fig. 9.11 Equivalence between the port current and the z-directed fictitious current density
dJmn(x,y) = cos ( k p ) cos (k,,y) sin (v)
\ L /
at the junction between a microstrip-line feed and a patch
The function sinc (z) is defined as sin (z)/z, and
mn nn
and for the planar waveguide model of the microstrip line feeding the patch k = - k = -
I a ' Y b
Jm = LxH, (9.14)
Thus (J,I = I Ji,J.If the effective width of the microstrip line is (for the j t h
port), the input current at the port j may be written as
4 = TI JzI (9.15)
6 = loss tangent of the dielectric
The length of rectangle is a, its width is b, and height of the substrate is d. The
9.3.2 Evaluation of 2-matrix from Green's functions points (x,, y,) and (x,, y,) denote the locations of thep and q ports, respectively.
Green's functions discussed above may be used to find the 2-matrix charac- It has been shown [37] that the doubly infinite series in (eqn. 9.16), along with
terisation of various planar segments of the shapes shown in Fig. 9.9 with
eqns. 9.17 and 9.18, can be reduced to a singly infinite series by summing the
respect to specified locations of external ports. These external ports may be
inner sum. The choice of summation over n or m depends on the relative
either of the probe-feed type (Fig. 9.10) or the microstrip-feed type (Fig. 9.11) locations of the ports p and q, and also on the aspect ratio of the rectangular
or a combination of these. Evaluation of the Z-matrix is based on relation 9.9 segment. We consider two different cases
 470 Multiport network approach for modelling Multiport network approach for modelling 477
Case I: When both the ports ( p and q) are oriented along the same direction (x The sign of yl is chosen so that Im (y,) is negative. w, and w, are widths of ports
or y). We may write Z,, as p and q, respectively. Also, we use
Y > = max(y,, Y,) Y < = min(yp9 Y,)
and a similar notation for x, and x, when I = n. The choice of the integer L
in eqn. 9.19 becomes a trade-off between fast computation and accuracy. A
compromise is to select L so that (y,F) is less than or equal to 100.
Case 2: When the two ports ( p and q) are oriented in different directions (x and
y), various elements of the Z-matrix may be
1
Z, -CF- 1a,cos(k,u,)cos(k,u,)cos(y,z,)
. f
I=L+I
cos (k,u,) cos (kp,) sinc (F) =
'I /=o

where 1 "
-CF - 2
'I l = L + I
cos (kp,) cos (k,u,)

and . sinc (y) ( -.iyl

exp (v> - v < - T)) (9.20)
C = jopd/(ab) dwj
Choice of Iis made by noting that, for convergence of the last summation in the
When the two ports are oriented in the y-direction we choose I = n, and when above equation, we need
they are along x-direction I is put equal to m. Also,
(v, - V, - wj/2) > 0 (9.21)
We choose the index of the inner summation so that this condition is satisfied.
This condition may be written more explicitly as
I = m, if {max(y,, y,) - min ( y,, y,) - wj/2) > 0 (9.22)
and
I = n, if {max (x,, x,) - min (x,, x,) - w,/2} > 0 (9.23)

When both of these conditions are satisfied, any choice of I will ensure conver-
gence.
If I = n, wi corresponds to the port oriented along y-direction and w, corres-
ponds to the port along the x-direction. On the other hand if I = m, wi is for
the port along the x-direction and wj for the port along the y-direction.
and
Z-matrix for circular segments: For circular-shaped patches, the impedance
Green's function is given by eqns. (A9.10) and (A9.11) in Appendix 9.8. When
ports are located along the circumference of the circle (as shown in Fig. 9.12),
 472 Multiport network approach for modelling
various elements of the Z-matrix [38] may be written as follows. For any port
i, the Z-matrix element Z,, may be written as
Fig. 9.12 Various parameters for ports located at the circumference of a circular segment
Off-diagonal terms of the impedance matrix are found to be
{COS[n(Ai - A,)] - cos ["(Ai + A,]} cos (n+d)
where
Similar expressions for computation of the Z-matrices for planar segments of
other geometries shown in Fig. 9.9 have not been reported so far.
9.3.3 Z-matrices for segments of arbitrary shape
When we come across segments of arbitrary shapes, for which Green's functions
do not exist, the impedance matrix can be found by a method known as contour
integral method [39]. The contour integral method is based on the Green's
theorem in cylindrical co-ordinates. The R F voltage at any point M(s) inside the
periphery of an arbitrarily shaped planar segment shown in Fig. 9 . 1 3 ~is given
by
Multiport network approach for modelling 473
where HA') is the zeroth-order Hankel function of the second kind and r is the
straight-line distance between the point M(s) and the source point on the
periphery (given by L(so)). The integral on the right-hand side of eqn. 9.26 is
carried out over the entire periphery. The RF voltage at any point just inside the
periphery can be derived from the above relationship. We obtain
2jv(s) = fc {k cos OH$ (kr)v(so) + jopdJ,, (so)~h2)(kr)}
ds, (9.27)
Fig. 9.1 3 (a) Configuration of a planar segment for analysis by contocr integralmethod; (b)
Division of the periphery in N sections for the analysis
where HI2)is the first-order Hankel function of the second kind, and J, denotes
line current density flowing into the segment at so.The variables s and sodenote
(9.25) distances along the contour C and r is the distance between the two points M
and L (specified by s and so) as shown in Fig. 9.13~.The angle 0 is the angle
made by the straight line joining points M and L with the normal to the
periphery at L. Line current density J,, flowing into the segment at a coupling
port, is given by
For the numerical calculation of the impedance matrix, we divide the periphery
into N sections having arbitrary widths W, , W ,. . . W, as shown in Fig. 9.13b.
The periphery is divided in such a manner that each coupling port contains an
integral number of such sections. For greater accuracy, wider coupling ports
may be divided into a multiple number of sections. We set N sampling points,
one at the centre of each section, and assume that each section is a straight edge.
It is further assumed that the widths of the sections are so small that magnetic
and electric fields can be considered constant over each section. Under the
assumptions outlined above, the line integral in eqn. 9.27 can be replaced by
 474 Multiport network approach for modelling Multiport network approach for modelling 475
summation over the N sections. The resulting expression is given by
be deleted from ZN.If each coupling port covers only one section, the matrix
N
thus obtained (after deleting rows and columns corresponding to the open
2jv, =
",=I
{kv,G, + jwpd i,&,} sections from Z,) is the required impedance matrix. If some coupling ports
extend to more than one section, the sections in these coupling ports are like
where v, is the voltage over the Ith section and i,(= J, W,) is the total current
sub-ports and the procedure detailed in Reference 1 can be used to obtain the
flowing into the m th section. The matrix elements G, and fi,,, are given as
overall admittance matrix at the coupling ports (and hence impedance matrix,
if desired).

to> otherwise 9.4 Edge-admittance and mutual-coupling networks

and
As discussed in Section 9.2.3, the multiport network modelling approach ass-
umes that the fields outside the patch may be represented by an equivalent
network model. The concept of edge admittance associated with the radiating
edges has been widely used in conjunction with the transmission-line model [23,
24,261. The same concept can be extended to the multiport network modelling
approach. For the multiport network model, the two-terminal edge admittance
(used in the transmission-line model) is replaced by the multiport network
In eqn. 9.31, y(= 0.5772.. .) denotes the Euler's constant. In the above discuss-
shown in Fig. 9 . 6 ~or b. If an edge of the patch is divided into n sections, the
ion we assume that the current can be fed into the planar circuit from all the N
edge-admittance network (EAN) is an n-port network with the common termi-
sections and ,i denotes the current fed from the m th section. This yields the
nal being the ground return.
impedance matrix for the N-port circuit. This matrix can be used to obtain the
In this Section, we discuss various methods for evaluating parameters of
impedance matrix for any specified number and location of ports on the planar
edge-admittance networks. Also, the network modelling approach can be exten-
circuit being analysed. Eqn. 9.29 is written for each section I on the periphery
ded to incorporate the effect of mutual coupling between the radiating edges of
of the planar circuit. All these equations combined together in matrix form
a single patch, as well as among the edges of adjacent patches in an array
become
environment.

9.4.1 Edge-admittance networks

where v and i are the voltage and the current vectors at each section. A and B
Edge admittance associated with a radiating microstrip patch consist of two
denote N by N matrices, determined by the shape of the circuit. The elements
components: (i) a susceptance representing the energy stored in the fringing field
of these matrices, obtained from eqn. 9.29, are
associated with the edge; and (ii) a conductance G representing the power
a , = - k G , for I f m transmitted to the radiation field as well as the power carried away by the
surface waves excited along the dielectric substrate. For incorporating edge-
a , = 2j admittance networks in the multiport network model for microstrip patch
and antennas, a capacitance-conductance pair is connected to each of the ports of
the equivalent planar multiport representation of the patch. The conductance G
bh = j w d & (9.34) consists of two parts: a radiation conductance G,and a surface-wave conduc-
From eqn. 9.32, the impedance matrix for the N sections, considered as ports, tance G,.Radiation conductance associated with an edge of a microstrip patch
is obtained as is defined as an ohmic conductance (distributed or lumped), which, when
connected to the edge (continuously or at discrete ports), will dissipate a power
equal to that radiated by the edge (or by an equivalent magnetic current source)
for the same voltage distribution. If the edge has width W and the power
In practice, the coupling ports are connected to onlv a few of the N sections
-... - ..- .
-
radiated for a uniform voltage distribution is P,,, the radiation conductance per
Rows and columns corresponding to the sections which are open-circuited can unit length of the edge is given by 2PJ W, where P,,,is calculated for a unit
 478 Multiport network approach for modelling Multiport network approach for modelling 479
Two other formulas, useful for design and based on Wiener-Hopf charac- e, < 2.65.
Accuracy of formula 9.42 is 2.6% for 0.2 < kod < 0.6 and 2.45
terisation of an infinitely wide microstrip patch edge, are given by Kuester et al.
More rigorous formulations for edge conductance of open-ended microstrips
[32] and Gogoi et al. [30].
have been reported by James and Henderson [41], and more recently by Katehi
The formula of Kuester et a[. (32) for electrically thin substrates (kod < 1)
and Alexopoulos [42] and by Jackson and Pozar [43]. Analysis techniques and
may be written as
numerical results for a limited set of parameters have been given in References
41-43. James and Henderson [41] have pointed out that their rigorous results
agree with eqn. 9.37a and a previous estimation by Lewin [44] for d/Ao < 0.09
when E, = 2.32, and for d/& < 0.03 when e, = 10. Results based on full-wave
analysis using the method of moments [43] agree with results in Reference 41 for
substrate thicknesses up to 0.1 &.

Edge susceptance: As in the case of edge conductance, several different results

are available for edge susceptance also. One of the formulas for edge suscep-
tance B is based on the parallel-plate waveguide model of a microstrip and is
given by Reference 45 as:

where Z, and e, are the characteristic impedance and effective dielectric con-
y = 0.57721 (Eurler's constant)
stant of a microstrip line of width a. Expressions for Zoand e, are well known
[I]. cis the velocity of waves in free space ( = 3 x 108m/s).Another formula for
The formula given by Gogoi et al. [30] is as: B, which is based on open-end capacitance of a microstrip line [46] is given by

Accuracy of eqn. 9.41 is 1.1% for 0.05 < k,d < 0.6 and 2.45 < e, < 2.65. where
These two formulas are accurate for wide patches (large value of b in Fig. 9.11).
Results based on these formulas are also plotted in Fig. 9.14.
Fig. 9.14 shows a comparison of the formulas 9.37-9.41 for the following set
of data: frequency f = 7.5 GHz; dielectric constant e, = 2.48; and thickness of Other formulas, which are based on the Wiener-Hopf formulation and which
substrate, d = 1/32 in. It is inferred from Fig. 9.14 that formulas 9.37 and 9.38 can be used for wide patches, are given in References 30 and 32. From Reference
yield close results as expected. The difference between formulas 9.37 and 9.41 32,
increases with decreasing value of the width. This is because of the fact that
formulas 9.40 and 9.41 are valid only for wide patches. Formulas 9.40 and 9.41
are both based on the Wiener-Hopf formulation and therefore yield identical
results for all values of the width. From this limited discussion, it seems reason-
able to use formula 9.37 owing to its simplicity. where ~ ( 0 is) as given earlier for eqn. 9.40. According to Reference 30,
Although the power coupled to surface waves is very small compared to the
radiated power, the conductance corresponding to the surface waves should be
added to GR.This conductance may be expressed as [30]
where
 480 Multiport network approach for modelling Multiport network approach for modelling 481
The accuracy of expression 9.48 is 2% for 0.1 < kod < 0.6 and 2.45 < E, .: with
2.65. Formulas 9.43-9.48 are compared in Fig. 9.15. Expressions 9.43, 9.44 and
9.47 give close results for all practical values of the resonator width
(0.25 1, < w < 0.6 1,). Eqn. 9.44 predicts an end-susceptance value of one-half
of that computed using other formulas. For all practical values of interest,
formula 9.43 may be used, since there is no restriction on the width of the patch
for this formula. When the width is large, both expressions 9.46 and 9.47 can
be used. 0.5274 arctan [0.084(w/d)l94"'52]
53 = l + &0.9236
re

T4 = 1 + 0.0377 arctan [ 0 . 0 6 7 ( ~ / d ) " ~ ~ ~ ]

x ( 6 - 5exp(0.036(1 - 8,)))
(, = 1 - 0.218exp(-7.5w/d) (9.49)
Expressions in eqns. 9.49 make use of the effective dielectric constant formula
of keference 50, ;.e.

where u = wld.
Accuracy df results given by eqn. 9.49 is claimed to be better than 2.5% for
the range of normalised widths 0.01 < wld < 100 and E, < 50.

Non-radiating edges: For rectangular patches radiating linearly polarised

I I I electromagnetic waves, radiating and non-radiating edges can be distinguished
0.5 1.0 1.5
normalised width ( b / h o ) clearly. As shown in Fig. 9.66, the non-radiating edges can be modelled by EAN
consisting of capacitances only. An equivalent approach is to extend the width
Fig. 9.15 Edge susceptance of a rectangular microstrip antenna versus normalised width of the patch by moving the non-radiating edges outwards so that the edge
(From Reference 27) capacitance is accounted for by the increased capacitance of the wider patch.
The concept of radiating and non-radiating edges has been studied [59] by
studying the total and partial reflections from the end of a parallel-plate
More rigorous characterisation of microstrip open-edge susceptance has been waveguide with an extended dielectric slab. It has been pointed out [60] that the
reported in [41, 43, 47, 481. Results of the analysis in References 47 and 48 are radiating or non-radiating nature of the edge depends on the angle at which the
available [49] as a closed form expression obtained by curve fitting of numerical wave underneath the patch is incident on the edge. For grazing incidence there
data. Normalised outward extension of the radiating edge (Auld in eqn. 9.44) is is no radiation, whereas for normal incidence the edge radiates. In between there
given by is a critical angle where the transition from non-radiation to radiation takes
place.
It has been recognised that the so-called non-radiating edges of a rectangular
 482 Multiport network approach for modelling

patch (operating at the resonance of 1,O mode) do contribute to a cross-

polarised radiation field. Compared to a single radiating edge, cross-polarised
radiation from a single non-radiating edge is typically 10-15 dB lower. When we
combine the radiated fields (in the broadside direction) from the two non-radiat-
ing edges, the total field is much smaller. This is caused by the fact that
,
E 1('
I
I
L-,
--,
Multiport network approach for modelling

i
I 'tL' X
patch
483

E1 10 E-f ield
equivalent magnetic currents corresponding to the fringing fields at the two 0I
\
iE
distribution
non-radiating edges are in the opposite direction and tend to cancel each other. I '%- -4+
However, the non-radiating edge-admittance network (NR-EAN), shown in I I
I I
Fig. 9.6b can be used only in cases where approximate results (obtained by
ignoring the cross-polarised radiation from NR edges) are considered satis-
factory. In more general cases, especially when the antenna operation is not at
1, 0 mode resonance (as for the circular polarised radiator discussed in Section
9.6.1), the EANs at all edges are similar to that shown in Fig. 9 . 6 ~ .
In spite of the numerous results for edge admittance that have been reviewed
in this Section. the lack of an accurate characterisation for the edne admittance
u Fig. 9.1 6 Modelling of the fringing field at the patch edges in terms of magnetic-current line
is one of the major shortcomings in the design information required for precise sources
design of microstrip patches and arrays.

9.4.2 Mutual-coupling network

As shown in Fig. 9.8, the mutual interaction between the fringing fields asso-
ciated with any two edges can be expressed in terms of a network model. The
basis idea of using an admittance element to represent mutual coupling between
two radiating edges was initially presented in Reference 51 in co&n&on with
the transmission-line model of microstri~antennas. The reader is referred to
Chapter 10 for details of this concept. The multiport mutual-coupling network
shown in Fig. 9.8 is an extension of this concept suitable for incorporating in the
multiport network model of microstrip patch antennas discussed in this Chapter.

Evaluation ofmutual-coupling networks: For computation of external mutual

coupling between various edges of a microstrip patch antenna on a thin sub-
strate, the field at the edge may be modelled by equivalent line sources of
magnetic current placed directly on the ground plane at the location of the
edges. This is illustrated in Fig. 9.16. Magnetic currents M are given by

where d is the height of the substrate. Product ( - E d ) is the voltage V(x, y)

defined in eqn. 9.10. 4 denotes a unit vector normal to the ground plane as
shown. For the E-field in the direction indicated in this Figure, both magnetic
current sources M a r e in y-direction, i.e. directed out of the plane of the paper.
The coupling between two magnetic current line sources is evaluated by
dividing each of the line sources in small sections, each of length dl. The
magnetic field produced by each of these sections (on the line source 1) at the
locations of the various sections on the line source 2 can be written by using Fig. 9.17 (a) Two arbitrarily spaced magnetic current elements; and ( 6 ) the co-ordinate
fields of a magnetic current dipole in free space [52]. The configuration and system used for computation of fields
 484 Multiport network approach for modelling Multiport network approach for modelling 485

co-ordination system is shown in Fig. 9.17. We have usually larger (typically 12). However, while using MCN for antenna analysis
(by segmentation), a small number of ports along radiating edges (typically 4)
. koMdl sin8 is sufficient. Thus the original mutual-admittance matrix (48 x 48 for 12 ports
H, = J
4n~or

Here ko is the free-space wave number and r is the distance between the point
P and the magnetic current element Mdl. When the two edges (say i and j ) are
oriented arbitrarily, as shown in Fig. 9.18a, the magnetic field H at (xi, y j )
produced by a source dliM a t ( x i , y,) may be written as
H = fH, + j H y (9.54)
with .

Hy = Hy,cos8, - H,. sin 0, (9.55)

H, = Hy,sin8, + H,cos 0, (9.56)
where
e,= H, cos 0 + H, sin 8
Hy. = -H, sine + H,cos0
Co-ordinate systems (x, y) and (x', y') are illustrated in Fig. 9.18b. Mutual
admittance between sections j and i may be written in terms of the electric
current density 4 induced in the upper surface of the edge segment j. We have
4 = ri x H = (-H,cosaj - Hy sinaj)j (9.59)
The current density induced on the surface of the edge section j underneath the
patch is -4. The mutual admittance between sections i and j is given by the Fig. 9.18 (a) Configuration showing sections i and j of two patch edges; and (b) two
different co-ordinate systems used for mutual coupling calculations
negative of the current flow into section j (underneath the patch) divided by the
voltage at section i, i.e.

The second minus sign in eqn. 9.60 accounts for the fact that the direction of
current for defining the admittance matrix of a multipart network is directed
into the network as shown in Fig. 9.19. The two edges shown in Fig. 9 . 1 8 ~may
be the edges of the same radiating patch or those of the different patches in a n
Fig. 9.1 9 Representation of the mutual coupling by an admittance matrix
array environment. When coupling between two adjacent patches in an array
environment is being computed, several individual edges of the two patches
contribute to the mutual coupling network (MCN). An MCN configuration along each edge) is reduced to a smaller size (16 x 16 as shown) by paralleling
taking the four radiating edges into account is shown in Fig. 9.20. Here, the the ports in subgroups of three each. Contributions of non-radiating edges can
MCN is connected to four ports along each radiating edge. In practice, the also be incorporated in MCN. This is not shown in the Figure. Detailed
number of sections considered on each edge for mutual-coupling calculations is computations [53] point out that the mutual-coupling contribution by non-
 486 Multiport network approach for modelling Multiport network approach for modelling 487
radiating edges is usually much smaller and may be ignored as a first-order mond points) is seen. Also shown in these Figures are results of Van Lil et al.
approximation. [51], based in transmission-line theory.
Mutual coupling computations based on the above formulation have been 1t may be noted that the preceding method of evaluating mutual coupling
verified [53] by comparison with the available experimental results [54]. Some of (based on the equivalent magnetic current model shown in Fig. 9.16) is valid
only for electrically thin substrates where the effect of surface waves along the
EAN EAN substrate is negligible. This modelling approach has been recently extended to
edge 1 edge 3
microstrip patches on thin substrates, but covered by a relatively thicker dielec-
tric cover layer [61,62]. The equivalent magnetic current model used in this case

1 patch 1 I I MCN I I patch 2

a=6.698 cm
b=10.56 crn
x = 1.84crn
f GHz

Fig. 9.20 A mutual coupling network (MCN) representing the coupling between two adja-
cent patches in an arrav

-15-
a =6.698 crn
b=10.56cm
-20 -

-25 -
m network model
-.-30 -
0

;;
-
v,
Fig. 9.22 Comparison of theoretical and experimental results for H-plane mutual coupling
between two rectangular microstrip patches (Reproduced from Reference 53)

Fig. 9.21 Comparison of theoretical and experimental results for E-plane mutual coupling
between two rectangular microstrip patches (Reproduced from Reference 5 3 )
these results are shown in Fig. 9.21 for E-plane coupling and in Fig. 9.22 for
H-plane coupling between two probe-fed rectangular patches. A very good
agreement between the computation (solid line) and experimental results (dia-
is shown in Fig. 9.23. The basic approach is similar to that for the case without
a cover layer discussed earlier. Eqns. 9.52 and 9.53 are replaced by H, and H,
in the presence of the cover layer. These field components are now dominated
by the effect of surface waves in the thicker cover layer.
When the substrate thickness is increased, equivalent magnetic current
models for Figs. 9.16 and 9.23 become more and more inaccurate. Conceptually,
multipart-network modelling of mutual coupling between two patches is still
possible if more rigorous analytical/numerical techniques [63, 641 could be
extended to arrive at a network representation of mutual coupling.
488 Multiport network approach for modelling
9.5 Analysis of multiport-network model
The most outstanding advantage of the multiport-network model is the fact that
various analysis and optimisation techniques available for multiport networks
can now be used for analysis and optimisation of microstrip antenna elements
and arrays. Most widely used techniques for planar networks are segmentation
[I-3, 6-81 and desegmentation [4, 5, 81 methods. These two network-analysis
techniques are reviewed in this Section. Examples of various microstrip antenna
configurations where these techniques have been used are reviewed in Section
9.6.
Multiport network approach for modelling 489
(shown by rectangular boxes in the figure) can be considered as a 'segment' for
application of the segmentation method. Essentially, the segmentation method
gives us overall characterisation or performance of the multiport network, when
the characterisation of each of the segments is known. Originally the segmenta-
tion method was formulated [2] in terms of S-matrices of individual segments;
however, it was found subsequently [3] that a 2-matrix formulation is more
efficient for microwave planar circuits (also for microstrip antennas). In this
Section, we will describe the procedure based on Z-matrices.

cover
layer
< \substrate

Fig. 9.24 Segmentation of a ring-shaped structure into four rectangular segments

Fig. 9.23 Magnetic-current source model for computation of mutual coupling in case of
microstrip patches on a thin substrate but covered with a thick dielectric layer

9.5.1 Segmentation method

The name 'segmentation' had been given to this network-analysis method when
it was used for planar (two-dimensional) microwave circuits by Okoshi and his
colleagues [2, 6, 35, 391. The basic idea is to divide a single large planar circuit
into simpler 'segments' which have regular shapes and can therefore be charac-
terised relatively easily. An example of such a segmentation is shown in Fig. 9.24
where a ring-shaped geometry is broken down into four rectangular segments
for which Green's functions are available. For the multiport-network model of
a two-port microstrip antenna shown in Fig. 9.8, each of the components
Fig. 9.25 Two connected multiport networks A and 8
For illustrating the procedure, we consider a multiport network consisting of
only two segments A and B, as shown in Fig. 9.25. Various ports of these two
segments are numbered as shown. The external (unconnected) ports of segment
A are called p,-ports (which may be more than one). Similarly, the external
 490 Multiport network approach for modelling Multiport network approach for modelling 491
!
unconnected ports of segment B are called p, ports. Connected ports of the It may be noted that the size of ZA, is (p,+ p,) x (p, + p,). The second term
segment A are named q-ports and the connected ports of the segment B are on the right-hand side is a product of three matrices of the sizes: (pa + p,) x q,
designated as r-ports. q- and r-ports are numbered such that q , is connected to
r , , q2 to r2, and so on. As a result of these interconnections, we can write:
+
q x q, arid q x (p, p,), respectively. From the computational point of view,
the most time-consuming step is the evaluation of the inverse of a matrix of size
& = V, and iq = -i, (9.61) (q x q), where q is the number of interconnected ports.
In order to illustrate the above procedure for combining Z-matrices of two
Z-matrices of segments A and B may be written as segments together, let us consider an example of two lumped resistive networks
connected together as shown in Fig. 9.260. Z-matrices of the individual com-
ponents A and B may be written as

where Zpa,Zpaq,Zqpa,z,,, Zpb,Z,,,, Z,pb,Z,, are sub-matrices of appropriate

dimensions. As we are dealing with reciprocal components

2-matrices of the segments A and B can be written together as

and

where
In terms of the notations of eqn. 9.64, we have

and

Superscript t indicates the transpose of a matrix, 0 denotes a null matrix of

appropriate dimensions.
It may be noted that interconnection conditions 9.61 have not been used for
writing eqn. 9.64 which represents a rearrangement of individual matrices ZA
and Z, given in eqn. 9.62. Relations 9.61 can now be substituted in the eqns. 9.64
to eliminate V,, 5, i,, and i,. The resulting expression may be written as
V, = [Z,,] ip,where
 492 Multiport network approach for modelling Multiport network approach for modelling 493
Substituting all these sub-matrices in eqn. 9.65 we get

which may be evaluated as

The resultant matrix ZABin eqn. 9.68 may be verified by rewriting the circuit
shown in Fig. 9 . 2 6 ~as the one shown in Fig. 9.266.

Fig. 9.27 Segmentation as applied to two sections of a transmission line

Let us consider another example. Two transmission-line sections of electrical

lengths 0, and O2 are connected in cascade as shown in Fig. 9.27. Z-matrices of
individual sections A and B are given by

and

In terms of notations of eqn. 9.65, we have

Fig. 9.26 (a) Two lumped networks considered for illustrating the segmentation procedure;
(b) Circuit simplification for writing Z-matrix of the combination of the networks
A and 8
 494 Multiport network approach for modelling Multiport network approach for modelling 495
which Green's functions are known. In cases like this, an alternative method
called desegmentation [4] is useful. The concept of desegrnentation can be
explained by considering the example of a rectangular patch with a circular hole.
Referring to Fig. 9.28, we note that, if a circular disc called segment fl (Fig.
9.28~)is added to the configuration of Fig. 9.28a, the resulting configuration y

Substituting all these submatrices in eqn. 9.65 we get

1(is equivalent t o )

a (desegrnented with)

(9.71~)
Substituting for z , and 2, and using trigonometric formulas for sin (8, + 8,) and I

cos (8, + 02), eqn. 9.71a may be expressed

Fig. 9.28 Concept of desegmentation

which is a 2-matrix for a uniform transmission line of length (0, +

8,) and
illustrates the validity of eqn. 9.65.
When the segmentation method is applied to the multiport network model of
microstrip antennas (such as the one shown in Fig. 9.8), we are interested in
2-matrix with respect to external ports (1 and 2 in Fig. 9.8) and also in the
voltages at the ports connecting R-EAN to the patch. This voltage distribution
at the radiating edges is expressed in terms of equivalent line source of magnetic
current. The radiation field (and associated characteristics like beamwidth, SLL F i g . 9.29 Port nomenclature used in desegmentation procedure
etc.) are obtained from the magnetic current distribution by using far-field (Ilr
variation) term of eqn. 9.52 and integrating over the various radiating edges.
Referring to Fig. 9.25, voltages at the connected ports (q-ports) may be is a rectangular segment shown in Fig. 9.286. Green's functions are known for
obtained by (Reference 1, p. 357) both the circular (Fig. 9.28~)and rectangular (Fig. 9.286) shapes, and therefore
2-matrices for characterising both of these components may be derived. The
desegmentation method allows us to derive the Z-matrix of the configuration a
where ip is the current vector specifying the input current(s) at the external shown in Fig. 9 . 2 8 ~when the 2-matrices of the rectangular segment in (6) and
port(s) of the antenna. the circular segment in (c) are known. For deriving a relationship among the
2-matrices of three shapes we consider a generalised configuration (shown in
9.5.2 Desegmentation method Fig. 9.29). Here, region fl is not included in the a segment. Ports p,, p, etc. are
There are several configurations of planar components which cannot be external ports of a. Characterisation of a is required with respect to these ports.
analysed by the segmentation method discussed above. For example, the con- In general p-ports may also be located on the part of the periphery of cc where
figuration shown in Fig. 9.28 cannot be partitioned into regular segments for the segment jis connected. An example of this is portp4 shown in Fig. 9.29.The
 496 Multiport network approach for modelling Multiport network approach for modelling 497

Z-matrices of /I and y segments are known, and may be written as Using eqn. 9.74 we get Z, as

As in the case of segmentation, ports q (of a) and ports r (of P ) are numbered which may be verified to the correct value. If the characterisation of the
such that q, is connected to r , , q, to r,, etc. Ports d a r e unconnected (external) a-network is needed with respect to ports 1 and 3, it becomes necessary to
port of the segment /I. Evaluation of Z, is simplified when the number of d-ports include port 3 in the y-network also. For this purpose the y-network is re-written
is made equal to the number of q (or r ) ports. The number of q (or r ) ports as in Fig. 9.32. Now 2, becomes
depends upon the nature of field variation along a-/3 interface and, as in the case
of segmentation, is decided by iterative computations. On the other hand, the
number of d-ports is arbitrary and can be always made equal to that of q- (or
r-) ports after that number has been finalised. Under these conditions, the
impedance matrix for the a-segment can be expressed (Reference 5 ) in terms of
the Z-matrices of /I- and y-segments as

Fig. 9.30 A resistive network considered for itlustrating the desegmentation procedure

It may be noted that the size of Z, is ( p x p) since all the specified ports of a Fig. 9.31 Reconfiguration of the network in Fig. 9.30 for eva/us?ionof the matrix I, used in
desegmentation procedure
segment have been numbered as p-ports. Let us consider an example for illu-
strating the implementation of eqn. 9.74. Consider the resistive network shown
in Fig. 9.30. Let us say that sub-network a is the one whose Z-matrix is to be
determined, the Z-matrix of the fl segment is known, and the combination of Now Z, is obtained by using eqn. 9.74 as follows:
these two is the y-segment whose 2-matrix is also known. We have = zm - zpd zdp
{z&- zd'}-I

If the characterisation of cc is to be found with respect to port 1 only, network

y can be considered as a 2-port network with one p-port and one d-port. The
Z-matrix of y is obtained by re-writing the network as in Fig. 9.31. We have
which again may be verified to be the correct result.
Let us also consider another example of two transmission-line sections of
electrical lengths PI, and PI, connected in cascade as shown in Fig. 9.27. Say we
want to find ZAwhen Z , and ZA, are known. When we need characterisation
of A only with respect to port 1, the segment y is a two-port network with Z,
 498 Multiport network approach for modelling Multiport network approach for modelling 499

This can be simplified to

These two examples illustrate the applications of the desegmentation eqn. 9.74.
It may be noted that, for implementing the desegmentation method, d-ports
of the j-segment need not be located on the periphery of the j-segment. In fact,
in the case of the rectangular patch with a circular hole (Fig. 9.28a), no region
is available on the periphery of the circular P-segment for locating d-ports. AS
and shown in Fig. 9.33, d-ports may be located inside the circular region. In this case
three d-ports dl, 4, d, are shown located inside the /?-segment. Since the Green's
functions are valid for any point on the segment (or on the periphery), the
desegmentation procedure remains unchanged.

44
Fig. 9.33 A configuration where the desegmentation procedure recluires location of d-ports
Fig. 9.32 Modification of the y-network for two-port characterisationof the a-segment in
inside the 8-segment
desegmentation procedure

The Z-matrix (1 port) of segment A may now be written using eqn. 9.74 as 9.6 Examples of microstrip antenna structures analysed by multipart-network
ZPI = zpP/ - ZpdIZddy - z d d p } - l z * approach

The multiport-network modelling and analysis approach discussed above has

been used for the analysis, design and optimisation of a variety of microstrip
radiators [8, 10-181. We will discuss these applications in three groups: (i)
,circularly polarised microstrip patches; (ii) broad-band multiresonator micro-
strip antennas; and (iii) multiport microstrip antennas and series-fed arrays.

9.6.1 Circularly polarised microstrip patches

Single-feed circularly polarised microstrip patches analysed by multiport-
network approach include: diagonal-fed nearly square patch [lo], truncated-
corners square patch [lo], square patch with a diagonal slot [lo], a pentagonal-
shaped patch [8], square ring patch [ l q , and cross-shaped patch [17]. The
desegmentation method has been used for analysing a truncated-corners square
patch and a square patch with a diagonal slot. Implementation of the deseg-
mentation procedure in these two cases is illustrated in Fig. 9.34. For a trun-
cated-corners patch, the a-segment (c.f. Fig. 9.29) is shown in Fig. 9.34~.For
 500 Multiport network approach for modelling Multiport network approach for modelling 501
p-ports d-ports implementation of the desegmentation procedure in this case, we use two
triangular-shaped P-segments P, and P,. Addition of P, and P2 to a-segments
results in a perfect square-shaped y-segment shown in Fig. 9.346. Configuration
of the u-segment for a square with a diagonal slot is shown in Fig. 9 . 3 4 ~In
. this
case the P-segment is a rectangle of the size of the slot, and d-ports are located
inside the rectangle as shown in Fig. 9.34d (this configuration may be compared
with Fig. 9.33). Detailed results for these two configurations are given in

feed Y-network
a b

feed
c d
Fig. 9.36 Radiation pattern for circularly polarized truncated-corners square antenna (Re-
Fig. 9.34 (a) and (b) Desegmentation method applied to a corners-truncated antenna; (c) produced from Reference 10 @ IEEE 1983)
and (d) desegmentation method applied to a square antenna with a diagonal slot Thickness: 118 in; E, = 2.52; frequency 3.176 GHz

Reference 10. Results based on multiport-network analysis have been compared

[lo] with experimental results. The type of agreement observed for the case of
a truncated-corner square antenna is illustrated in Figs. 9.35 and 9.36. Theoreti-
cal and experimental values of VSWRs and axial ratios are plotted in Fig. 9.35,
whereas radiation patterns are compared in Fig. 9.36. A more quantitative
comparison of the theoretical and experimental performance is contained in
Table 9.2 (from Reference 10). Reasonable agreement between the theoretical
and experimental results verifies the validity of the multiport-network approach.
Corresponding results for a square antenna with a diagonal slot are depicted in
Figs. 9.37 and 9.38 and in Table 9.3 (again from Reference 10). The multiport-
network approach is seen to perform equally well in this case also.
Square-ring patch and cross-shaped patches have been analysed [17] by the
experiment
__---- segmentation method. Segmentation of a square ring in four rectangular
segments is shown in Fig. 9.24. Segmentation of a cross-shaped patch in three
rectangular segments is illustrated in Fig. 9.39. Detailed results for these two
configurations are given in Reference 17. Comparison for five different shapes
3170 3180 3190 (square ring, crossed strip, almost square patch, corner-chopped square patch,
frequency, MHz
and square with a diagonal slot) indicates that the maximum axial-ratio band-
Fig. 9.35 Theoreticaland experimentalresults for axial ratio and input VSWR for a truncated- width (about 5.2% for f = 3.0GHz, E, = 2.5, and h = 0.159cm) is obtained
corner square antenna (Reproduced from Reference 10 @ IEEE 1983)
by using a square-ring configuration.
Table 9.2 Performance of corners-chopped square patch antennas (Reproduced from Reference 70 @ IEEE 1983) f
I Parameters Antenna I Antenna I1 3
1 Thickness, E, 1/8", 232 1/16, 2.51 i
I1
2 Dimensions a x a cm2
3 Truncation bla
Performance
1
(see Fig. 9.34)

Theoretical
2.73 x 2.73
0.04578
Experimental
2.86 x 2.86

Theoretical
0.0573
Experimental
%
3
3

1 Centre frequency f,(GHz) 3.1758 3.1750 3.1756 3.1753 3

2 Resonant frequencies of orthogonal
modes (GHz)
3.1340
3.2 155
3.1325
3.2125
3.1370
3.2340
3.1 343
3.2298
9
3
3 Axial ratio at centre frequency (dB) 0.02 0.0 0.12 0.1 5 Q
4 Bandwidth (MHz) for axial 26.4 29.4 14.0 14.4 3
5
ratio < 6 dB (0.831 %) (0.925%) (044%) (0.4535%) r~
5 Input VSWR at centre frequency 2.26 2.26 1.6 1.8
6 Beamwidth for 3 dB difference
between lE,l and lE41 129' 152' 129' 138"

Table 9.3 Performance of square-patch antenna with a diagonal slot (Reproduced from Reference 10 @ IEEE 1983)
Theoretical Experimental
3.130
c
1 Centre frequency f, (GHz)
2 Resonance frequency of orthogonal modes (GHz)
3.130
3.063
3.212
3.060
3.210
5.
2

3 Axial ratio at f, 0.198

4 Bandwidth for axial ratio less than 6 dB 35.5 MHz
(1.134%)
5 Input VSWR at chosen feed location 2.9 2.9 %
6 Beamwidth for 3 dB difference P0
between lEsl and IEJ 116" 124' aI
3
Substrate thickness = 1/8in, e, = 2.52. Dimensions of square patch = 2 . 6 0 2 ~ ~x 12 , 6 0 2 ~ ~ 1
Dimensions of slot = 2 . 8 9 ~ ~x 10 . 4 7 ~ ~ 1 . 9
2
3
P
2
 504 Multiport network approach for modelling Multiport network approach for modelling 505
Another interesting antenna configuration analysed by the multiport-
network approach is a pentagonal-shaped patch originally proposed in Referen-
ces 55 and 28, and analysed in Reference 8. The antenna configuration is shown
in Fig. 9 . 4 0 and
~ two different methods of analysis are depicted in Fig. 9.406 and
c. Desegmentation with two triangular segments PI and P, yields a 90'-60'-30'

-VSWR

0L I I I I I I I
3100 3120 3140 3160 3180
frequency, M H z

Fig. 9.37 Theoretical and experimental results for a square antenna with a diagonal slot
(Reproduced from Reference 10 @ IEEE 1983)
Thickness = 118 in: e, = 2.52; frequency = 3.1 30 GHz (Reproduced from Re-
ference 10 @ IEEE 1983)

Fig. 9.39 Segmentation of a cross-shaped patch into three rectangular segments

Fig. 9.40 (a) A pentagonal-shaped microstrip patch

(b) Desegmentation procedure applied to the pentagonal shape
(c) Combination of segmentation and desegmentation procedure for the pen-
tagonal patch

triangular segment for which Green's function is available. The second ap-
OdB -10 -20 -30-30 -20 -10 OdB proach illustrated in Fig. 9 . 4 0 ~employs desegmentation with one segment PI to
Fig. 9.38 Radiation pattern for a circularly polarised square antenna with a diagonal slot yield a kite-shaped geometry. The kite shape is then segmented into two identi-
Thickness = 118in; E, = 2.52; frequency = 3.1 30GHz (Reproduced from Re- cal 90"-60"-30' triangles as shown. Both of these approaches yield identical
ference 10 @ IEEE 1983) results.
 506 Multiport network approach for modelling Multiport network approach for modelling 507
9.6.2 Broadband muItiresonator microstrip antennas
Another group of microstrip antenna configurations which have been analysed
using the multipart-network approach are broadband microstrip antennas
using coupled resonators [I 1-13]. All these configurations use multiple resona-
tors with slightly different resonant frequencies. These configurations are shown
in Figs. 9.41 and 9.42. Different resonators in any of these configurations are
coupled to each other, and only one (usually the central one) is connected to the
feedline. Two different coupling mechanisms have been used. The three con-
figurations shown in Fig. 9.41 use capacitive coupling across the gaps between
the closely spaced edges, whereas the three configurations of Fig. 9.42 employ
short sections of microstrip lines for providing the necessary coupling.
Analysis procedure for a gap-coupled multiresonator antenna configuration
is illustrated in Fig. 9.43. Coupling gaps are modelled by the multiport lumped
Fig. 9.41 (a) Configuration of a radiating-edge gap-coupled microstrip antenna (REG- RC network shown in Fig. 9.436. Values of C,, C, and C, are obtained from
COMA) coupled microstrip transmission-line analysis. G represents the radiation con-
(6)Configuration of a non-radiating-edge gap-coupled microstrip patch antenna ductance and is obtained by treating the gap fields as a line source of equivalent
(NEGCOMA) magnetic current. Since the feed point is located along the centre line XX (Fig.
(c) Four-edges gap-coupled microstrip patch antenna (FEGCOMA)
9.43a), symmetry of the configuration may be used to simplify the computations
and only one-half of the antenna configuration, shown in Fig. 9.43c, need be
analysed. The multiport-network model is shown in Fig. 9.43d. RCs represent
edge-admittance networks and GCs are two networks modelling the coupling
gaps. Mutual coupling networks are not shown in this Figure because the effect
of mutual coupling was not incorporated in the results presented in Reference
11. Component REs are planar network models for the three patches. Ports 1,
2, and 3 on the central patch are the three locations investigated for locating the
probe feed. Experiments were performed to verify the theoretical results ob-
tained by using the multiport-network approach, and a comparison is shown in
Fig. 9.44. In this case, a O.159cm-thick copper-clad substrate (6, = 2.55) was
used. The experimental bandwidth of the antenna is 225 MHz (6.9% at centre
frequency f, = 3.27GHz), which is slightly more than the theoretical value
(207MHz), possibly because of the dielectric, conductor and surface-wave
losses ignored in the computations of the results reported in Reference 11. For
cornpa&on, the corresponding bandwidth of a single patch is 65 MHz. Thus the
microstrip antenna configuration shown in Fig. 9.44 yields a bandwidth nearly
3.5 times that of a single patch.
A multiport-network for a directly coupled three-resonator antenna con-
figuration is shown in Fig. 9.45. In this case also, one can make use of geometri-
cal symmetry, and only one-half of the antenna configuration (with a magnetic
wall placed along the plane XX) needs to be analysed. The multiport-network
model is drawn in 9.45~.Interconnecting microstrip line sections are also
Fig. 9.42 (a) Configuration of a radiating-edge directly coupled microstrip patch antenna modelled by two planar rectangular segments RE,. We have nine edge-admit-
(REDCOMA) tance networks denoted by RCs. The segmentation formula 9.65 is used for
(b) Configuraton of a non-radiating-edge directly coupled microstrip patch anten-
na (NEDCOMA)
finding the input impedance at the feed port and eqn 9.72 for evaluating the
(c) Configuration of a four-edge directly coupled microstrip antenna (FEDCOMA) voltage distribution at the edges of the radiating patches.
 508 Multiport network approach for modelling Multiport network approach for modelling 509
Fairly wide impedance bandwidth for microstrip antennas can be achieved by antenna is
using the multiple resonator configurations shown in Figs. 9.41 and 9.42. microstrip
Typical values for the six configurations fabricated on substrates with E, = 2.55 microstrip

1.01 I I I I I J
3.1 3.2 3.3 3.4
frequency, GHz
6
Fig. 9.44 Theoretical I----) and experimental (-) results for a gap-coupled triple-
Fig. 9.43 (a) Three-patchgap-coupled antenna configuration
resonator antenna (Reproduced from Reference 11 @ IEEE 1984)
(6) Multipart-network modelling of the gap betkeen two patches
(a) Impedance locus on Smith Chart
(c) Half-section of the antenna configuration with a magnetic wall along the plane
( 6 ) VSWR variations.
of symmetry XX
(d) Multiport network model of the half-section of the antenna
antenna (configuration shown in Fig. 9.46) is being currently developed as a
are summarised in Table 9.4. Various acronyms (REGCOMA etc.) are defined compact broadband microstrip patch [57].
in Figs. 9.41 and 9.42. The factor M gives the bandwidth BW as a multiple of
the corresponding value for a single rectangular patch antenna;f, is the centre 9.6.3 Multiport microstrip patches and series-fed arrays
frequency and d is the thickness of the substrate. Series-fed linear arrays of microstrip patches employ two-port radiators as basic
 570 Multiport network approach for modelling
Multiport network approach for modelling 51 7
Table 9.4 Typical impedance bandwidth values for microstrip antennas
using multiple coupled resonators (Based on Reference 56) building blocks. For this application, both two-port rectangular patches [14,29]
as well as two-port circular patches [15] have been analysed by using the
Configuration d(cm) f(GHz) BW(MHz) BW(%) M
multiport-network modelling approach.
REGCOMA 0.159 3.29 33 1 10.0 5.3
NEGCOMA 0.3 18 3.1 1 480 15.4 4.0 Two-port rectangular patch: The multiport-network model of a rectangular
FEGCOMA 0.3 18 3.16 815 25.8 6.7 patch with two microstrip-line ports along the non-radiating edges is shown in
REDCOMA 0.3 18 3.20 548 Fig. 9.8. Segments labelled FLN (feed line network) are rectangular planar
17.1 5.0 segments representing small sections (typically 118 long) of microstrip lines
NEDCOMA 0.3 18 3.31 605 18.3 5.5
FEDCOMA 0.3 18 3.38 810 connected to the two ports. Widths of FLNs are equal to the effective widths of
24.0 7.36

radiating
edges
/
Fig. 9.46 Configuration of a broadband coupled microstrip line radiating patch

RE c Fig. 9.47 Variations of the power transmitted to the port 2 (as a percentage of the input
power) with the changes in the locations of the two ports (Reproduced from
Fig. 9.45 (a) Three-patch antenna configuration with direct (microstrip-line) coupling bet- Reference 27)
ween the parches
(6) Half-section of the antenna configuration in (a) with a magnetic wall along the
plane of symmetry XX the two lines, respectively. Multiple interconnections between FLNs and the
(c) Multiport-network model of the antenna half-section shown in (b) patch ensure that the parasitic reactances associated with the feed-line-patch
junctions are taken into account. Radiating edge admittance networks (R-EAN)
 572 Multiport network approach for modelling Multiport network approach for modelling 513

and non-radiating edge-admittance networks (NR-EAN) are obtained by Thus we conclude that the multiport-network model and the analysis approach
modelling fringing fields at the edges as discussed in Section 9.4.1. The mutual discussed here are well suited for S-parameter characterisation of the radiating
coupling network (MCN) represents the external interaction between two ra- patches.
diating edges as discussed in Section 9.4.2.
As mentioned earlier, for two-port patches with ports along the non-radiating
edges, transmission from port 1 to port 2 can be controlled by su:1.able location
of the ports (distance x, and x, in F:g. 9.2b). An example of this feature is
presented in Fig. 9.47.This Figure shows the variation of th, power transmitted
to port 2 with the relative locations of external ports. Vlr .x of x , are chosen
to ensure match at the input port (S,, = 0). When the p i t locations are altered,
the associated change in the junction reactances causes the patch resonance
frequency to shift slightly. The corresponding change in the resonant dimension
a is also plotted in this Figure. The results shown are for a substrate with
6, = 2.48, d = 1/32in, tans = 0.002 and for a resonant "requency of 7.5GHz.
0
A comparison of theoretical and experimental results for S,, of a two-port patch II -
n
is shown in Fig. 9.48. Design parameters of the two-port patch are also listed
in this Figure. Apart from the magnitude of S,, which determines the amplitude -80-
-
A theoretical
experimental

-looO I I I I I i I I I
7.0 7.2 7.4 7.6 7.8 8.0
frequency, GHz

Fig. 9.49 Comparison of the theoretical and experimental values of the transmisson phase
angle for a two-port rectangular patch (Reproduced from Reference 29)

2 -5
!- / -
A theoretical
experimental
circular rectangu/lar
planer
circular
[patch )
I l l I l l (
7.0 7.2 7.4 7.6 7.8 8.0
frequency, GHz

Fig. 9.48 Comparison of the theoretical and experimentalresults for transmission coefficient
of a two-port rectangular patch (Reproduced from Reference 29) Fig. 9.50 Analysis of a circular two-port microstrip patch antenna by combining multiport
network-modelling approach with the cavity method

variation along the linear series-fed array, another parameter of interest in the
two-port patches is the phase angle of the transmission coefficient from port 1
to port 2. Theoretical computations of the phase angle of S,, (based on the
multipart-network approach) have been compared with experimental values
obtained from measurements using an automatic network analyser. Fig. 9.49
shows these results. The excellent agreement obtained demonstrates the useful-
ness of multiport-network model for computations of the phase angles also.
Two-port circularpatches: Circular microstrip patches with two ports located
along the circumference have been analysed [I51 by using a cavity model for the
circular patch ;nd a multiport modelling approach for the input/output micro-
strip feed-line junction. This approach is illustrated in Fig. 9.50. The physical
radius a of the disc and its loss tangent 6 are replaced by effective values a, and
6,. The effective radius a takes into account the fringing capacitance around the
 514 Multiport network approach for modelling 515
Multiport network approach for modelling
circumference [45]. The effective loss tangent 6, includes 'loss' due to the power
radiated from the patch. Relation 9.5 or 9.6 may be used for this purpose. Power is not exactly the multipart-network approach discussed in Section 9.2.3 but is
radiated 8,power dissipated in the dielectric P, and the power lost because of a hybrid combination of the conventional cavity method (Section 9.2.2) and the
multiport-analysis technique. The multiport-network approach itself can, in
finite conductor conductivity PCmay be evaluated as illustrated in Reference 46,
principle, be applied to circular patches also, but no such efforts have been
pp. 92-94.
Approximate results [15], using the dominant mode only and ignoring the reported to date.
feed-junction reactances, point out that, for a match at the input port (S,, = O),
the impedance Zo of the feed line at the input port is related to the Z , , element
of the Z-matrix by C,=2.2
d = 1132 inch
f = 7.5GHz
a=0.7815cm

d = 1/32 inch
f =7.5GHz
a-0.7815crn
N
V)

5-10 A theoretical data

measured data

-201 I I I I L
7.0 7.2 7.4 7.6 7.8 8.0
frequency, GHz
frequency, GHz
Fig. 9.51 Comparison of the theoretical and experimental results for transmission coefficient
Fig. 9.52 Theoretical and experimental results for transmission phase angle of a two-port
of a two-port circular patch (Reproduced from Reference 58)
circular patch (Reproduced from Reference 58)

where 4,, is the angular separation between the two ports and
Series-fed microstrip arrays: The two-port rectangular or circular patches

(3
Z , , = 1.674 - 1016,

where d is the substrate thickness. The corresponding transmission coefficient

S2, is given by
Eqn. 9.86 suggests that, for the dominant mode, the transmission coefficient S2,
varies from 1 to 0 as the angular separation #+, is changed from 0 to 90". Also
we note that, for high values of S,, close to unity, the input-port characteristic
impedance becomes very small and makes the design impracticable.
Results based on the above method have been verified experimentally and a
sample comaprison of theoretical and measured S,, values is presented in Fig.
9.51. Also, a comparison of the transmission phase in this case is shown in Fig.
9.52. Again, a reasonably good agreement is obtained. It may be recalled that,
for this example of a two-port circular patch, the method of analysis followed
discussed above may be cascaded together to form a series-fed linear array as
shown schematically in Fig. 9 . 5 3 ~and b. The rnultiport-network modelling
approach has been used [la] for design and sensitivity analysis of series-fed
arrays. Each unit cell of the array is characterised in terms of a 2-port Z-matrix.
A two-port representation of a typical unit cell of a series-fed array is shown in
Fig. 9.54. This is a simplification of the rnultiport-network representation shown
in Fig. 9.8. Edge-admittance networks associated with non-radiating edges are
not shown. Multiple interconnections between various components are shown
symbolically by a single multiconnection symbol defined in the inset of the
Figure. Ports 1 and 2 are two external ports of the unit cell. el and e2 are the
two radiating edges. Typically, there are four interconnections (at each of these
edges) among the R-EAN, the patch, and the mutual coupling network (MCN).
Analysis based on the segmentation procedure allows us to determine voltages
(or currents) at the input/output ports of each of the cells of the array. This
information is used in conjunction with another set of impedance matrices
(called Z, matrices) for each cell. The Z, matrices (obtained from eqn. 9.72)
 516 Multiport network approach for modelling
relate the voltages at various ports on radiating edges e l and e2 to the input
currents at the input/output ports 1 and 2 of each cell. Thus we have a multiport-
network representation of the series-fed array. This representation may be
extended to include MCNs (mutual coupling networks) representing the coupl-
ing between adjacent cells. In this case, the two-port representation of each cell
input "
I.
Fig. 9.53 (a) Series-fed linear array of rectangular patches
(6) Series-fed linear array of circular patches
R-EAN
---@---symbol for multiple
lnterconnections
Fig. 9.54 Multiport network representation of a typical unit cell of a series-fed linear arrav
needs to be extended to multiple n-port representation, the additional (n - 2)
ports being the fictitious ports to account for the mutual coupling. A schematic
network representation of such a unit cell of a series-fed array is shown in Fig.
9.55. The network MCN2 denotes a multiport mutual coupling network re-
presenting the coupling between the two adjacent patches (i.e. between the patch
shown and the next patch on the right-hand side). In the model shown, the
coupling to the third patch (on either side) has been ignored. The n-port network
Multiport network approach for modelling 517
shown here represents a unit cell of the series-fed array when a non-negligible
mutual coupling is present between the adjacent cells. Such a network re-
presentation can be used to evaluate the effect of mutual coupling on the array
performance and to iteratively modify the array geometry to compensate (as far
as possible) for the undesirable effects of the mutual coupling.
For analysing multiport-network models of large arrays, a circuit analysis
technique called the sub-network growth method [I] is very convenient. In this
method, only two adjacent components are combined together (at any stage in
the iterative loop) to form a larger sub-network. Consequently, the size of the
matrices to be processed is restricted to the number of ports in the two com-
ponents and does not increase proportionately to the size of the array.
At present, research efforts are in progress for using multiport-network
modelling techniques for the automated design of one-dimensional and two-
dimensional arrays of microstrip patches.
in-.
. -1- ports
---@---symbol for multiple
Interconnections
Fig. 9.55 Incorporation of mutual coupling between adjacent patches in the network re-
presentation of a unit cell shown in Fig. 9.54
9.7 CAD of microstrip patch antennas and arrays
Increasing interest in the use of microstrip antenna technology in phased-array
systems, and the potential of fabricating millimeter-wave arrays monolithically
on GaAs wafers, have made it necessary to develop CAD techniques for the
accurate design of microstrip patches and arrays. The multiport-network ap-
proach presented in this Chapter is well suited for CAD (i.e. for modelling,
analysis and optimisation) of microstrip patch antennas and arrays [20].
Basic aspects of CAD methodology are well known [l] and are common to
the computer-aided-design process in various other disciplines. A generic flow
chart for CAD is shown in Fig. 9.56. Starting with a given set of specifications,
 578 Mulriport network approach for modelling Multiport network approach for modelling 519
synthesis methods and available designs (pre-stored in computer) help us to design for calculating the changes in the designable parameters. Iterations in the
arrive at the initial design. A model of this initial design is analysed by a optimisation loop are carried out until the specifications are met or the optimum
computer-aided-analysis (simulation and performance evaluation) package. performance of the design (within the given constraints) is achieved. The anten-
na design so obtained is now fabricated and experimental measurements are
carried out. As indicated in the lower part of Fig. 9.56 (portion inside the dashed
specifications rectangle), some modifications may still be necessary if the modelling and/or
design data and analysis has not been sufficiently accurate. The modifications, hopefully, should
synthesis
be small and the aim of CAD is to minimise these experimental iterations as far
as is practicable.
The three main aspects of the computer-aided-design process are modelling,
initiol analysis and optimisation. Modelling and analysis approaches suitable for CAD
of microstrip patches and arrays have been discussed in this Chapter. The most
attractive features of the multiple-network approach is the extension of the
I network-analysis methods to microstrip patches and antennas. Also, the techni-
ques for sensitivity analysis and optimisation, which have been developed
extensively for multiport networks [I], can now be extended to network models
of microstrip patches and arrays. Use of gradient optimisation techniques
analysis
involves calculation of gradients of antenna performance with respect to various
designable parameters. The adjoint network method of sensitivity analysis has
been used extensively for calculating gradients for circuit optimisation. This
method can now be used for sensitivity analysis of microstrip patch antenna
configurations also. Since the network models involved are passive and recipro-
cal, the adjoint network is identical to the original network model itself, and
thus a single network analysis is sufficient to yield the sensitivity information
also. Among other things, this would yield the sensitivity of the voltage distribu-
tion along the radiating edges with respect to the various designable parameters
of the antenna configuration. Sensitivity of the radiation characteristics with
respect to the antenna parameters can be calculated therefrom.
It is expected that the techniques reported in this Chapter will lead to
implementation of accurate CAD procedures for microstrip patches and arrays.
I measurements
I 9.8 Appendix: Green's functions for various planar configurations
I
In this Section Green's functions for some planar shapes, shown in Fig. 9.9, are
given. In the expressions that follow, ai is given by

rJi =
Fig. 9.56 Typical flow chart for computer-aided design methodology for microstrip antennas 2, otherwise

The performance characteristics obtained are compared with the given specifica- (a) A rectangle: The Green's function for the rectangle shown in Fig. 9 . 9 ~
is
tions. When the specifications are not met, the designable parameters of the given as [2]
antenna configuration are modified and the analysis is repeated. These analysis,
modification and comparison steps constitute a single iteration in the optimisa-
tion loop. Several optimisation strategies include sensitivity analysis of the cos (kxxo)cos (kyyo)cos (k,x) cos (kyy)
r ~ ~ r ~ , ,

kt+g-k2
 520 Multiport network approach for modelling Multiport network approach for modelling 521
where (d) A right-angled isosceles triangle: The Green's function for the right-
ml~
k, = - and k, =
n~
- angled isosceles triangle shown in Fig. 9.9d is given by
a b

(b) A 30"-60" right-angled triangle: The Green's functions for the triangle
shown in Fig. 9.9b is given as [34]: where
mnx nny nnx mny
T(x, y) = cos -cos - + (- I)"+" cos -cos - (A9.9)
a a a

(e) A circle: The Green's function for the circle shown in Fig. 9.9e is given by
[351
where
Tl(x, y) = (- 1)'cos (z) cos [2n(m3a ")Y]

+ (- I)mcos (-)
2nmx
$a
cos [2n(n - I ) y
3a ] where J,,(.) represents Bessel's function of the nth order, and k,, satisfies

with the condition that The subscripts m in k,, denotes the m th root of eqn. A9.11. For the zeroth-order
Bessel's function, the first root of eqn. A9.11 is taken to be the non-zero root.
I = -(m + n)
(n A circular sector: The Green's function for circular sectors are available
(c) An equilateral triangle: The Green's function for the equilateral triangle only when the sector angle a is a sub-multiple of R. For the circular sector shown
shown in Fig. 9 . 9 ~
is given as [34] in Fig. 9.9f for which a = all, the Green's function is given as [36]

where TI(x, y) is given by eqn. A9.4 and

T,(x, y ) = (- ~)'COS- sin ($) ['"",a n)~]

+- c o s (-1
2smx
$a
[ sin
2n(n -
3a ]
1)y (g) An annual ring: The Green's function for the annular ring shown in Fig.
9.9g is given as [36]

+ (- 0 s
$a
( ) [ sin
2n(l - m) y
] (A9.6)

As for Tl(x, y), the integer I in T2(x, y) is given by eqn. A9.5.

522 Multiport network approach for modelling
In the above relations N,(.) denotes Neumann's function of order n and J:(.) and
N',(.) denote first derivatives with respect to the arguments.
(h) An annular sector: As in the case of circular sectors, the Green's function
for annular sectors are available only if the sector angle a is a submultiple of n.
For the annular sector shown in Fig. 9.9h for which ir = n/i, the Green's
function is given as [36]
where n, = nl, and F,,,(.) is defined in eqn. A9.15. The values of k,,,,,, are
obtained from eqn. A9.16.
9.9 Acknowledgments
Most of the material discussed irr this Chapter is based on the theses and
publications of several of my students and colleagues: Rakesh Chadha, P. C.
Sharma, Girish Kumar, Yinggang Tu and Abdelaziz Benalla. Their contribu-
tions are gratefully acknowledged.
9.10 References
1 GUPTA, K. C., et al.: 'Computer-aided design of microwave circuits', (Artech House, USA,
1981) chaps 8 and I I
2 OKOSHI, T., and TAKEUCHI, T.: 'Analysis of planar circuits by segmentation method',
Electron. Commun. Japan, 1975, 58-B, pp. 71-79
3 CHADHA, R., and GUPTA, K. C.: 'Segmentation method using impedance-matrices for
analysis of planar microwave circuits', IEEE Trans., 1981, MlT-29, pp. 71-74
4 SHARMA, P. C., and GUPTA, K. C.: 'Desegmentation method for analysis of two-dimen-
sional microwave circuits', IEEE Trans., 1981, MTT-29, pp. 1094-1098
5 SHARMA, P. C., and GUPTA, K. C.: 'An alternative procedure for implementing deseg-
mentation method', IEEE Trans., 1984, MTT-32, pp. 1-4
Multiport network approach for modelling 523
6 OKOSHI, T.: 'Planar circuits for microwaves and lightwaves' (Springer-Verlag. 1985) chap. 5
7 GUPTA, K. C., et al.: 'Two-dimensional analysis for stripline/microstrip circuits'. 1981 IEEE
MTT-S International Microwave Symp. Digest, pp. 504-506
8 GUPTA, K. C., and SHARMA, P. C.: 'Segmentation and desegmentation techniques for
analysis of two-dimensional microstrip antennas'. 1981 IEEE AP-S International Antennas
and Propagation Symp. Digest, pp. 19-22
9 GUPTA, K. C.: 'Two-dimensional analysis of microstrip circuits and antennas', J. Inst.
Electron. Telecommon. Engrs. (India), 1982, 28, pp. 346-364
10 SHARMA, P. C., and GUPTA, K. C.: 'Analysis and optimised design of single feed circularly
polarised microstrip antennas', IEEE Trans., 1983, AP-31, pp. 949-955
I I KUMAR, G., and GUFTA, K. C.: 'Broadband microstrip antennas using additional resona-
tors gap-coupled to radiating edges', IEEE Trans., 1984, AP-32, pp. 1375-1379
12 KUMAR, G., and GUPTA, K. C.: 'Non-radiating edges and four-edges gap-coupled multiple
resonator, broadband microstrip antennas', IEEE Trans., 1985, AP-33, pp. 173-178
13 KUMAR, G., and GUPTA, K.C.: 'Directly coupled multiple resonator wideband microstrip
antennas', IEEE Trans., 1985, AP-33, pp. 588-593
14 GUPTA, K. C.: 'Two-port transmission characteristics of rectangular microstrip patch radiat-
ors'. 1985 IEEE AP-S International Antennas Propagat. Symp. Digest, pp. 71-74
15 GUPTA, K. C., and BENALLA, A,: 'Two-port transmission characteristics of circular
microstrip patch antennas'. 1986 IEEE AP-S International Symp. Antennas Propagat. Digest,
pp. 821-824
16 PALANISAMY, V., and GARG, R.: 'Analysis of arbitrary shaped microstrip patch antennas
using segmentation technique and cavity model', IEEE Trans., 1986, AP-34, pp. 1208-1213
17 PALANISAMY, V., and GARG, R.: 'Analysis of circularly polarised square ring and crossed-
strip microstrip antennas', IEEE Trans., 1986, AP-34, pp. 1340-1346
18 BENALLA, A., and GUFTA, K. C.: 'A method for sensitivity analysis of series-fed arrays of
rectangular microstrip patches'. National Radio Science Meeting (URSI), Boulder (CO),
USA, Jan. 1987, Digest, p. 65
19 GUPTA, K. C.: 'Multiport-network modelling approach for computer-aided design of micro-
strip patches and arrays'. 1987 IEEE AP-S International Symp. Antennas Propagat., Black-
sburg (VA), USA, June 1987
20 GUPTA, K. C., and BENALLA, A,: 'Computer-aided design of microstrip patches and
arrays'. Int. Microwave Symp./Brazil, July 1987, Symp. Proc. Vol. 1, pp. 591-596
21 LO, Y. T., e t a / . :'Theory and experiment on microstrip antennas', IEEE Trans., 1979, AP-27,
pp. 137-145
22 RICHARDS, W. F., et a[.: 'An improved theory for microstrip antennas and applications',
IEEE Trans., 1981, AP-29,pp. 38-46
23 MUNSON, R. E.: 'Conformal microstrip antennas and microstrip phased arrays', IEEE
Trans., 1974, AP-22, pp. 74-78
24 DERNERYD, A. G.: 'Microstrip array antenna'. Proc. 6th European Microwave Conf., 1976,
pp. 339-343
25 CAMPI, M.: 'Design of microstrip linear array antennas by computer'. Proc. Antenna Ap-
plications Symp., Robert Alerton Park, Univ. of Illinois, Urbana, USA, Sept. 1981
26 BENALLA, A., and GUPTA, K. C.: 'Transmission line model for 2-port rectangular micro-
strip patches with ports at the non-radiating edges', Electron. Lett., 1987, 23, pp. 882-884
27 BENALLA, A., and GUPTA, K. C.: 'Two-dimensional analysis of one-port and two-port
microstrip antennas'. Electromagnetics Laboratory, Scientific Rept. 85, Univ. of Colorado,
May 1986, p. 48
28 COFFEY, E. L., and LEHMAN, T. H.: 'A new analysis technique for calculating the self and
mutual impedance of microstrip antennas'. Proc. Workshop Printed Circuit Antenna Technol-
ogy, New Mexico State Univ., 1979, pp. 31.1-31.21
29 BENALLA, A., and GUPTA, K. C.: 'Multiport-network model and transmission characteris-
tics of two-port rectangular microstrip antennas', IEEE Trans., Oct 1988, AP-36, pp. 1337-42
 524 Multiport network approach for modelling Multiport network approach for modelling 525
30 GOGOI, A., and GUF'TA, K. C.: 'Wiener-Hopf computation of edge admittances for micro- 56 KUMAR, G., and GUPTA, K. C.: 'Broadband microstrip antennas using coupled resona-
strip patch radiators', AEU, 1982, 36, pp. 247-251 tors'. 1983 IEEE AP-S Int. Antennas Propagat. Symp. Digest, pp. 67-70
31 VAN DE CAPELLE, A,, et a/.: 'A simple accurate formula for the radiation conductance of 57 GUPTA, K. C., and BANDHAUER, B.: 'Coupled line model for multiresonator wide band
a rectangular microstrip antenna'. 1981 IEEE AP-S International Symp. Antennas Propagat., microstrip antennas'. National Radio Science Meeting (URSI), Boulder, Jan. 1988
Digest, pp. 23-26 58 BENALLA, A,: Unpublished experimental results, 1986
32 KUESTER, E. F., et 01.: 'The thin-substrate approximation for reflection from the end of a 59 CHANG, D. C., and KUESTER, E. F.: 'Total and partial reflection from the end of a
slab-loaded parallel plate waveguide with application to microstrip patch antenna', IEEE parallel-plate waveguide with an extended dielectric slab', Radio Sci., 1981, 16, pp. 1-13
Trans., 1982, AP-30, pp. 910-917 60 CHANG, D. C.: 'Analytical theory of an unloaded microstrip patch', IEEE Trans., 1981,
33 GUPTA, K. C., and BENALLA, A.: 'Effect of mutual coupling on the input impedance and AP-29, pp. 54-62
the resonant frequency of a rectangular microstrip patch antenna'. National Radio Science 61 TU, Y.: 'Edge admittance and mutual coupling in rectangular microstrip patch antennas with
Meeting (URSI), Boulder, Jan. 1986, Digest, p. 226 a dielectric cover layer', Ph.D. Thesis, Univ. of Colorado, 1987, pp. 62-84
34 CHADHA, R., and GUPTA, K. C.: 'Green's functions for triangular segments in planar 62 TU, Y., GUF'TA, K. C., and CHANG, D. C.: 'Mutual coupling computations for rectangular
microwave circuits', IEEE Trans., 1980, MTT-28, pp. 1139-1143 microstrip patch antennas with a dielectric cover layer', URSI National Radio Science Meet-
35 OKOSHI, T., eta/.: 'Planar 3-dB hybrid circuits', Electron. Commun. Japan, 1975.58-B, pp. ing, Boulder, Jan. 1987, Digest p. 66
80-90 63 POZAR, D. M.: 'Input impedance and.mutual coupling of rectangular microstrip antennas',
36 CHADHA, R., and GUPTA, K. C.: 'Green's functions for circular sectors, annular rings and IEEE Trans., 1982, AP-30, pp. 1191-1 196
annular sectors in planar microwave circuits', IEEE Trans., 1981, MlT-29, pp. 68-71 64 JACKSON, D. R., et a/.: 'An exact mutual coupling theory for microstrip patches', 1987 IEEE
37 BENALLA, A., and GUPTA, K. C.: 'Faster computation of Z-matrices for rectangular AP-S Int. Symp. Antennas Propag. Digest, Vol. 2, 1987, pp. 790-793
segments in planar microstrip circuits', IEEE Trans., 1986, MTT-34, pp. 733-736
38 GUPTA, K. C., and ABOUZAHRA, M. D.: 'Analysis and design of four-port and five-port
microstrip disc circuits', IEEE Trans., 1985, MTT-33, pp. 1422-1428
39 OKOSHI, T., and MIYOSHI, T.: 'The planar circuit - An approach to microwave integrated I
circuitry', IEEE Trans ., 1972, MTT-20, pp. 245-252
40 JAMES, J. R., et a/.: 'Microstrip antenna theory and design', (Peter Peregrinus, 1981), p. 23
41 JAMES, J. R., and HENDERSON, A,: 'High-frequency behaviour of microstrip open-end
terminations', IEE Microwaves, Optics & Acoustics, 1979, 3, pp. 205-218
42 KATEHI, P. B., and ALEXOPOULOS, N. G.: 'Frequency-dependent characteristics of
microstrip discontinuities in millimeter-wave integrated circuits', IEEE Trans., 1985, Mm-33,
pp. 1029-1035
43 JACKSON, R. W., and POZAR, D. M.: 'Full-wave analysis of microstrip open-end and gap
discontinuities', IEEE Trans., 1985, MTT-33, pp. 1036-1042
44 LEWIN, L.: 'Radiation from discontinuities in stripline', Proc. IEE, 1960, lWC, pp. 163-170
45 WOLFF, I., and KNOPPIK, N.: 'Rectangular and circular microstrip disk capacitors and
resonators', IEEE Trans., 1974, MTT-22, pp. 857-864
46 BAHL, I. J., and BHARTIA, P.: 'Microstrip antennas' (Artech House, 1980) chap. 2
47 JANSEN, R. H., and KOSTER, N. H. L.: 'Accurate results on the end effect of single and
coupled microstrip lines for use in microwave circuit design', AEV, 1980, 34, pp. 453-459
48 JANSEN, R. H.: 'Hybrid mode analysis of the end effects of planar microwave and millimeter-
wave transmission lines', Proc. IEE, 1981, 129, pp. 77-86
49 KIRSCHNING, M., e l a/.: 'Accurate model for open end effect of microstrip lines', Electron.
Lett., 1981, 17, pp. 123-125
50 HAMMERSTAD, E., and JENSEN, 0.: 'Accurate models for microstrip computer-aided
design', 1980 IEEE M l T - S Int. Microwave Symp. Digest, Washington, 1980, pp. 407-409
51 VAN LIL, E. H., and VAN DE CAPELLE, A. R.: 'Transmission line model for mutual
coupling between microstrip antennas', IEEE Trans., 1984, AP-32, pp. 816-821
52 BALANIS, C. A,: 'Antenna theory analysis and design', (Harper and Row, 1982). p. 169
53 BENALLA, A,, and GUFTA, K. C.: 'Multiport network approach for modelling mutual
coupling effects in microstrip patch antennas and arrays', IEEE Trans., Feb 1989, AP-37, pp.
148-52
54 JEDLICKA, R. P., et a[.: 'Measured mutual coupling between microstrip antennas', IEEE
Trans., 1981, AP-29, pp. 147-149
55 WIENCHEL, H. D.: 'A cylindrical array of circularly polarised microstrip antennas'. IEEE-
APS Int. Symp. Antennas Propagation Digest, 1975, pp. 177-180
 Chapter 10

Transmission-line model for

rectangular microstrip antennas
A. Van de Capelle

List of sumbols

length of patch
width of patch
thickness of patch or co-planar strip conductor
conductivity of patch or co-planar strip conductor
RMS surface error of patch or co-planar strip conductor
length of substrate
width of substrate
thickness of substrate
relative permittivity of substrate
loss tangent of substrate
conductivity of ground plane
RMS surface error of groun9plane
thickness of ground plane
width of strip conductor of a microstrip line
length of microstrip line
Y,, = characteristic admittance of microstrip line
y, = propagation constant of microstrip line
Y, = characteristic admittance of transmission line representing a rec-
tangular microstrip antenna
y, = propagation constant of the above
Y, = self-admittance representing the open-end terminations of a
microstrip antenna
G, = real part of Y,, self-conductance
B, = imaginary part of Y,, self-susceptance
A, = extra length of microstrip line by open-end effect
y, = admittance per unit length of a slot with infinite length
g, = real part of y,
b, = imaginary part of y,
E,, H, = electric, magnetic field in equivalent slot apertures
528 Transmission-line model for rectangular microstrip antennas 529
Transmission-linemodel for rectangular microstrip antennas
= excitation voltage of a slot i
i,, i,, i, = unit vectors of x, y , z co-ordinates 10.1 Introduction
S = width of equivalent slots
go,&! = Fourier transform of E,, H, Microstrip antennas have a physical structure derived from microstrip trans-
k,, k,, k, = components of k mission lines. Therefore a transmission-line model is the first and most obvious
k = propagation vector choice for the analysis and the design of microstrip antennas. However, the
p = real part of the complex radiated power per unit length transmission-line model is often regarded as a simplified and somewhat dated
q = imaginary part of the complex radiated power per unit length theory. This is true for the original, simple transmission-line model; but the
q = wave impedance in half-space above antenna accuracy of the improved transmission-line model is comparable to that of other
k = propagation constant in half-space above antenna more complicated methods. Even mutual coupling between rectangular micro-
p = permeability of half-space above antenna strip antennas can be calculated in a fairly accurate and very efficient way with
E = permittivity of half-space above antenna
the transmission-line approach.
ym = mutual admittance per unit length The practical design of a microstrip antenna or a microstrip array, including
g, = real part of y, matching and feeding networks, has to be done by means of a CAD software
b, = imaginary part of y, package. Existing programs represent the network components by equivalent
J = Bessel function of the first kind transmission lines. If the antenna elements are modelled by the same trans-
Y = Bessel function of the second kind mission-line approach, the incorporation in the available CAD software is
s = normalised slot width straightforward.
C, = Euler's constant The concept of the transmission-line model can be applied to any microstrip
E , ~= effective relative permittivity
antenna configuration for which separation of variables is possible. In this
Kfl = effective width Chapter we will devote our attention entirely to rectangular (and square)
a = attenuation constant microstrip antennas.
/I= phase constant The transmission-line model does not include surface waves. Therefore, the
qo = wave impedance of free space
application is limited to antenna configurations where the thickness and the
k,, = propagation constant in free space substrate permittivity are sufficiently small to avoid considerable excitation of
1, = free-space wavelength those surface waves. But, in practice, this is not a severe limitation. However,
We = length of equivalent slot research is going on to also include surface waves in the transmission-line model.
w = normalised slot length
L, = centre distance between equivalent slots
F, = auxiliary coupling function for the mutual conductance 10.2 Simple transmission-line model
E, = relative error on the radiation conductance G,
Fb = auxiliary coupling function for the mutual susceptance 10.2.1 Description of the transmission-line model
Y, = mutual admittance The transmission-line model will be discussed for rectangular (and square)
Gm = real part of Y, microstrip antennas. The antenna consists of a conducting patch, a dielectric
Bm = imaginary part of Y,
substrate and a conducting ground plane. The antenna is fed by a microstrip line
Kg = correction function for the mutual conductance (as shown in Fig. 10.1) or by a coaxial probe (Fig. 10.2).
Kb = correction function for the mutual susceptance The patch is characterised by the resonant length L (resonant for the funda-
D = directivity in broadside direction mental mode), the width W, the thickness t , the conductivity a, and the RMS
G = gain in broadside direction surface error Ap .
q = antenna efficiency
In the analysis, the dielectric substrate is supposed to have infinite dimensions
Q = antenna quality factor in the plane of the patch. In practice, it has a length L,, a width W, and a
thickness h. Electrically, it is characterised by a relative permittivity E, and a loss
BW = impedance bandwidth
tangent 6,. It is supposed that the substrate consists of one homogeneous layer.
A multilayer substrate can be replaced by an equivalent homogeneous layer with
equivalent relative permittivity and loss tangent.
 Transmission-line model for rectangular microstrip antennas 537
530 Transmission-linemodel for rectangular microstrip antennas

The conducting ground plane has the same dimensions as the substrate: L,,
K i n practice, infinite of extent for the analysis. It is further characterised by a
conductivity u,, a RMS surface error A, and a thickness t,.
In the case of an antenna fed by a co-planar microstrip line, the strip
conductor has a width W, and a length L,. The other parameters of the
microstrip line ( t , h, t,, a,, a,, Ap, A,, E,, 6,) are the same as for the antenna. The
cross-sectional geometry of the microstrip line is characterised by the aspect
ratio W,/h. Likewise, the microstrip antenna can be considered as a microstrip
line with a very large aspect ratio W/h.

Fig. 10.2 Rectangular microstrip antenna fed by a coaxial probe

L = length of patch
L,,,=length of feedline
W = width of patch
WP, width of feedline
E,= relative permittivity of substrate
t = thickness of conducting patch and feedline

- = thickness of conducting groundplone

thg thickness of substrote

Fig. 10.1 Rectangular microstrip antenna fed by a microstrip line

In the case of a microstrip antenna fed by a microstrip line (Fig. 10.1), the
introduction of the transmission-line model is straightforward (Fig. 10.3):
The microstrip feed line is represented by a transmission line with a charac-
teristic admittance Y, (mainly determined by the aspect ratio W,/h and the
relative permittivity E,), a propagation constant y, and a physical length L,. I b
The rectangular microstrip antenna is represented by a transmission line with
a characteristic admittance Y , (mainly determined by the aspect ratio W/h and
WL- L -38'

the relative permittivity E,), a propagation constant y, and a physical length L. Fig. 10.3 Rectangular microstrip antenna with transmission-line model
 532 Transmission-line model for rectangular microstrip antennas
0 At the cross-sections AA' and BB', in Fig. 10.3a, the microstrip line with
aspect ratio Wlh has an open-ended termination, which can be represented by
a parallel admittance (Y, = G, +
jBJ.
In the case of excitation with a coaxial probe, the equivalent transmission-line,
model has to be modified as shown in Fig. 10.4.
Fig. 10.4 Rectangular microstrip antenna with coaxial feed and equivalent transmission-line
model
Fig. 10.5 General three-port equivalent transmission-line model
A general transmission-line model, which can be applied in both cases (micro-
strip line feed or coaxial feed), consists of a three-port circuit (Fig. 10.5). The
general three-port model has to be completed with a piece of transmission line
Transmission-line model for rectangular microstrip antennas 533
at port 1 and port 3 has to be left open (Fig. 10.3) in the case of a microstrip
line feed. In the case of a coaxial-feed probe the model has to be completed with
an inductance at port 3.
The main step in the modelling of a microstrip antenna by a transmission-line
equivalent, is the representation of the open-ended terminations by a parallel
admittance Y,. An open-ended microstrip line does not perform as a perfect
open circuit (see Fig. 10.6):
0 The field lines do not stop abruptly at the end of the strip conductor: there
is a stray field extending beyond the end of the strip; this can be interpreted as
an electrical lengthening A1 of the line, which implies an amount of stored.
energy; on the other hand, the stray field is also source of power radiated in the
space above the antenna and launched as surface waves along the substrate;
The real part G, of the parallel admittance Y, represents the radiation effect
(and surface waves), and the imaginary part B, models the stored energy in the
extra line length.
Fig. 10.6 Open-ended microstrip line with aspect ratio Wlh
10.2.2 Expressions for G, and B,
The accuracy of the transmission-line model depends strongly on the choice of
expressions for G, and B,. In the original transmission-line model, proposed by
Munson [I], a simple but very approximate expression for Y, has been proposed:
Y, = WY, (10.1)
534 Transmission-line model for rectangular microstrip antennas Transmission-linemodel for rectangular microstrip antennas 535

with y, = admittance per unit length of a uniformly excited slot with infinite where i,, i,, i, = unit vectors of the x, y, z co-ordinates; V,, V , = excitation
length and width h, in an infinite, perfectly conducting plane. voltage of slot I and slot 2, respectively; S = h = width of the equivalent slots;
This expression is not accurate enough, but it is important because of the L, = L + h = centre distance between the equivalent slots.
concept behind it:
The radiation of a rectangular microstrip antenna can be explained as
originating from the tangential electric field in the plane of the patch.
In the fundamental mode, only the contribution from the two open ends is
important.
The source of radiation can be limited to two narrow zones along the two
open ends of the patch.
The field in these two narrow zones can be thought of as the field of two
rectangular slots in an infinite, perfectly conducting plane.
For the fundamental mode of the microstrip antenna the tangential field in
these two slots can be considered to be uniformly distributed.
A slot with a uniform excitation field can be considered as a cut from an
infinitely long, uniformly excited slot.
The idea of representing the microstrip antenna by equivalent slots in an infinite,
perfectly conducting plane is very powerful. The inaccuracy of eqn. 10.1 is
mainly due to the last simplification, where the edge effects of finite-length slots Fig. 10.7 Two-slot model
are neglected.
We now have two concepts available to explain the radiation of a microstrip
antenna: The spatial Fourier transform, with respect to y, of this aperture is given by
the open end concept
the equivalent slot concept.
where k,, k,=components of the propagation vector k ;
k,,
These two concepts can be used to derive expressions for the parameters in the Ikl = k = w@E)'I~ =propagation constant.
transmission-line model. In the following Sections the equivalent, slot concept As the field of eqn. 10.2 has only a y-component, the Fourier transform is
will be applied where possible, and the open-end concept where necessary.
In this Section we want to derive suitable expressions for the parameters of = 8,iY
the simple transmission-line model of Fig. 10.5. We proceed with eqn. 10.1 and with
derive expressions for the real part g, and the imaginary part b, of y,.
A configuration of two equivalent slots, as shown in Fig. 10.7, is considered.
The slots are pieces of length W, taken from infinitely long, uniformly excited
slots. The tangential electric field in the slot apertures can be written as:
The complex radiated power ( p + jq) per unit slot length, in terms of the spatial
Fourier transform, is given by [Reference 2 pp. 61-68]:

E, =
{<
V
iy for
L, - S
2 <-y<-
Le
>
'II (10.2)

(0 elsewhere
 536 Transmission-line model for rectangular microstrip antennas Transmission-line model for rectangular microstrip antennas 537
From a network point of view, the two-slot configuration of Fig. 10.7 can be From eqns. 10.9-10.17 the following expressions for g, and b, can be derived
considered as a symmetrical two-port with a self-admittance per unit length
y,(=g, + jb,) and a mutual self-admittance per unit length y,(=g,,, +
jb,).
Expressed in these quantities, the complex radiated power per unit length is
given by

where the terms in s4,s6etc. have been neglected. The maximum truncation error
Taking V, = 0, g, and 6, follow from eqns. 10.5-10.8: of eqns. 10.18 and 10.19 is not larger than 0.1% for s < 1. Expressions 10.18
and 10.19, combined with expression 10.1, completely determine the parallel
admittance Y,.

.stria conductor

The single integrals in eqns. 10.9 and 10.10 can be written (see Reference 3
appendix) as double integrals of a Bessel function of the first kind J, and the
second kind Y, respectively: 7 \dielectric substrate
conducting groundplane
a

perfect conducting
Weft *delectric walls

where s = kS is the normalised slot width

I'
\ +@! b magneticconducting
walls

J:(s) = lIi Jo(v)dvdu

Fig. 10.8 Planar- waveguide model for microstripline
Y/(s) = &(v) dvdu a Cross-section of microstripline with aspect ratio Wlh
b Cross-section of planar-waveguide model
By twice integrating the series expansion of Joand Y, [4], the following series are
obtained 10.2.3 Expressions for the line parameters
To derive expressions for the characteristic admittance Y , and the propagation
constant yp(=ap + j&) of the equivalent transmission line representing the
antenna, and for the characteristic admittance Y,and the propagation constant
y, of the microstrip feed line, the planar-waveguide model is used, see [5].
A microstrip line with aspect ratio W/h (cross-section see Fig. 10.8a) and with
a dielectric substrate of relative permittivity E, is modelled by a planar waveguide
(cross-section see Fig. 10.86). The top and bottom walls of this planar
waveguide are electrically conducting, while the side walls are perfect magnetic
conductors. The guide is of height h equal to the microstrip substrate height, but
it has an effective width W ,. larger than the physical width W of the strip. The
where X = ln(s/2) + C,; C, = Euler's constant = 0.577216. guide is filled with a dielectric that has an effective relative permittivity E , ~ .
 538 Transmission-linemodel for rectangular microstrip antennas
The characteristic impedance Z, and the phase constant /3 of the fundamental
mode propagating in the microstrip line are given in terms of the planar-
waveguide parameters:
a = k0& (10.21)
where 4, = && = wave impedance in free space; ko = w 6 = propaga-
tion constant in free space.
In the quasi-static approximation, Reference 6 gives convenient expressions for
Kf(0) = tahlln (hF/ W' + $ + (2h/ w')'}
W = W + {I + ln(4/J(t/h)' + ( l / a ) 2 / ( ~ / t+ 1.1)')
and for
The attenuation a and the frequency dependence of Z, and are neglected in the
simple transmission-line model. Indeed, as will be explained in the improved
transmission-line model, the simple model has important shortcomings, so that
it makes no sense to take into account these second-order effects.
10.3 Improved transmission-line model
10.3.1 Description of the improved transmission-he model
The simple transmission line model has important shortcomings:
(i) The expressions for Y, are inaccurate for the usual patch widths (i.e. for
W < lo;& = free-space wavelength);
Transmission-linemodel for rectangular microstrip antennas 539
(ii) The mutual coupling between the two equivalent slots is neglected;
(iii) The radiation from the side walls is not taken into account.
Derneryd [ I , 81 has partly eliminated the first two shortcomings:
(i) To determine G, = Re(Y,), he considers the two main slots with an identi-
cal excitation and a negligible width. He finds an integral expression for G,, for
which an approximate analytical solution has been derived by Lier [9]. Der-
neryd's model corrects the first two shortcomings of [Reference 11 for the real
part of Y,,but it still neglects the influence of the side walls on G,.
(ii) To determine the susceptance B, = Im(Y,), Derneryd makes this par-
ameter equal to the open-end self-susceptance of the microstrip line formed by
(10.22) the patch. This corrects the first shortcoming of [Reference 11 for the imaginary
part of Y,.
(10.24)
Fig. 10.9 improved transmission-linemodel represented as a three-port
An improved transmission-line model, proposed by Pues and Van de Capelle
[lo], will be discussed here. This model corrects the three shortcomings of
[Reference 11for the real as well as for the imaginary part, and has a broad range
of validity. The circuit representation of the present model is shown in Fig. 10.9.
In this network Y, is the self admittance of the open-end terminations of the
patch, and Y, is their mutual (radiation) admittance. The mutual coupling is
formally taken into account by voltage-dependent current sources.
The admittance matrix of this three-port model is given by
[YI =
I
Y , + Y,coth(y,L,)
- Y,
- Y, csch (y, L,)
-Y,
Y, + Y, coth (y,L,)
- Y, csch (y, L2)
- csch (Y,L,
- Y ,csch (y, L,)
Y,coth (y,L,)
1
+ Y,coth (y, L2)
where cothz and cschz are the complex hyperbolic cotangent and cosescant
(10.29)
 540 Transmission-linemodel for rectangular microstrip antennas Transmission-line model for rectangular microstrip antennas 541
functions of argument z, respectively. The copper and dielectric losses of the been spent in comparing available formulas and deriving new ones where
antenna are taken into account by the attenuation constant a,, the real part of needed. It was a primary goal to combine accuracy with numerical efficiency.
the complex propagation constant y,. We have tried to obtain analytical expressions for all the model parameters.
If there is only one feed point, an input admittance can be defined. Assuming The imaginary part of Y,, the self-susceptance B,, is determined by means of
I, = I, = 0, it follows from eqn. 10.29 that the open-end-effect concept. Indeed, in the equivalent-slot concept, the self-
susceptance depends strongly on the aperture field and there is no information
available on an appropriate choice of this field. The real part of Y,, the self-
+ + 2Y,Y,coth(yPL) - 2YmY,csch(ypL)
Yf + Y:
Yf Y: - Yi
- Yi) coth (ypL) + (Yf - Y: + Yi) cash (2ypA)csch (ypL) + 2Y, Y, I
(10.30)
conductance G,, is modelled as the radiation conductance of an equivalent slot.
The mutual admittance Y, is also determined from the equivalent-slot concept.
For the line parameters, the attenuation and the frequency dependence of Y, and
y are included.
where
10.3.2 Expression for the self-susceptance B,
For the self-susceptance B,, the correct transmission-line formula is used:
B, = Y, tan (PAI) (10.36)
L, and L, are defined in Fig. 10.9.
In the case of a microstrip-line-fed antenna, this corresponds to I, = I3= 0. It where Y,, p, A1 are, respectively, the characteristic admittance, the phase con-
follows from eqns. 10.29 and 10.30 that stant and the open-end extension of a microstrip line with aspect ratio W/h, as
formed by the patch.
Y, =
~f + Y: - Y; + 2Y, Y,coth (ypL) - 2YmY,csch (y,L) The most appropriate expression for A1 is given in Reference 11:
Y, + Y ,coth (ypL)
(10.32) where
To model the parasitic effects of the feed line on the antenna behaviour, the
t, 6';
= 0.434907 60,,,
+ 0.26 (W/h)0'85" + 0.236
(10.38)
self-admittance of the open-end termination facing the feed line is reduced by ,f - 0.189 (W/h)0'85M+ 0.87
a factor

0.5274 arctan { 0 ~ 0 8 4 ( ~ / h ) " ~ ~ " " ~ }

where W, = width of feed line; W& = effective width of patch. &0.9236 (10.40)
This reduction takes into account the partial covering of the open-end termina- eff
tion by the feed line. The reduction of the self-admittance at terminal l can be 0.0377 arctan {0.067(W/h)1'456)
considered as an addition of a parallel admittance
{ 6 - 5 exp [0.036(1 - E,)]} (10.41)

The antenna input admittance is given by

Expressions for Y,, p and eeffare given in Section 10.3.6.

10.3.3 Expression for the self-conductance G,

The accuracy of the improved transmission-line model depends strongly on the
accuracy of the expressions for the model parameters. Therefore much effort has
For the real part of Y,, the self-conductance G,, the equivalent-slot concept
explained in Section 10.2.2 is applied. The model is similar to that of Derneryd
[7], except for the dimensions of the equivalent slot. The open-end terminations
of the patch are replaced by uniformly TE-excited narrow rectangular slots of
length W, = Kf (instead of Win Reference 7) and width A1 (instead of h in
Reference 7).
 542 Transmission-line model for rectangular microstrip antennas Transmission-line model for rectangular microstrip antennas 543
To calculate the self-conductance G,, one such equivalent slot is considered,
This complex power can also be written in terms of network parameters:
as shown in Fig. 10.10. The electric field in the slot aperture is assumed to be
uniform:

Equating expressions 10.48 and 10.50, expressions for G, and B, follow

where y = excitation voltage of the equivalent slot; S = AZ = width of the
equivalent slot.

(10.52)
Fig. 10.10 Equivalent slot radiator in an infinite, perfectly conducting plane
Using expression 10.46 for gY,we obtain for G,:

The spatial Fourier transform of the aperture field is defined as

4 =
+m
j-, I-m
+m
En dkyydx dy (10.44)
The aperture field has only a y-component, so that the Fourier transform
The inner integral can be written as a double integral of the Bessel function of
4 = &J,,
the first kind and order zero 131.Expansion of the Bessel function in a Maclaurin
where series and double integration term by term give:

&y = yW, sin (k, W,12) sin (k,,S/2)

x K1 (kyS/2)
The complex power radiated by this slot may be found by integrating the
complex Poynting vector over the aperture surface A:

P +jQ = f jl, E. x H $ . i z dxdy (10.47) The first two terms of this series expression are used in eqn. 10.53 to obtain
finally
with Ha the magnetic field in the slot aperture A.
1 sin w
Expressed in terms of the Fourier-transformed aperture field
G, % -{[wSi(w)
ntl +-
w
+ cosw - 21

where
x (1 -;)+q;+-p--&)} cos w sin w

= + (k2 - kt - k$)'I2 for k2 2 k: +g where w = kW, = normalised slot length; s = k S = normalised slot width;
kz (10.49) sin u
= -j(k2,+g-kl)'12 for k l < e + g si(x) = j0
-du
u
 544 Transmission-line model for rectangular microstrip antennas Transmission-line model for rectangular microstrip antennas 545
As explained before, expression 10.52 is not used to calculate B,, as it is where the superscript (2n) denotes the 2n th derivative. Truncation of these
impossible to define a suitable aperture field. One has to fall back on the series, maintaining the first two terms, gives
open-end-effect concept (see Section 10.3.2).

10.3.4 Expression for the mutual conductance G,

The expression for the mutual conductance G,,, of finite-length slots will be
derived from the mutual conductance between infinite slots. Therefore, an Using the identity
auxiliary coupling function is defined 1
Ji2'(l) = 2 {J2(l) - Jo(l))

where g, and g, are the per-unit-length self-conductance and mutual conduc- we finally obtain for g,:
tance, respectively, of two infinite-length TE-excited slots in a perfectly conduct-
ing infinite ground plane, as shown in Fig. 10.7. The aperture field was given in
eqn. 10.2. An analytical expression has been derived for x1 in Section 10.2.2:

The maximum truncation error of eqn. 10.63 is about 0.1 % of g, for s < 1.
Using eqn. 10.18 found for g, and eqn. 10.63 for g, the auxiliary coupling
function F, = g,/g, can be expressed as
To obtain an expression for g,, the complex radiated power per unit slot length
+
( p jq) has to be expressed in terms of the Fourier-transformed aperture field
(eqns. 10.6 and 10.7), and in terms of the slot voltages V, and V, (eqn. 10.8).
Setting 6 = V,and equating the real partp of the radiated power, the following
expression for gmis found: The auxiliary coupling function F, has been introduced to calculate a first
approximation of the mutual conductance G, of the finite-length slots by
putting
G,,, = GSF, (10.65)
where L, = L + A1 = centre distance between the two slots; S = Al = width with G, the self-conductance as given by eqn. 10.55.
of the equivalent slots. The results of this approximation are compared to the following reference: the
Notice that radiation conductance of the four-slot equivalent system shown in Fig. 10.11.
cos ( / c , ~ ~ =
1
2 sin2 (ky
) - F) This four-slot system consists of two main slots and two side slots. Themain
slots have a length W, = wfl
and a width 81, as used before, and have a centre
distance L, = L + Al. The side slots have a length L,, a width A1 and a centre
distance W,. The tangential electric field in the aperture plane z = 0 is:
W L -A1 Le + A1
Similarly to the derivation of eqn. 10.1 1 from eqn. 10.9, we obtain the following for 1x1 < 2;
2 2
< IYI < - 2
expression for eqn. 10.57 using eqn. 10.58:

E, =
v sin ($ ix for
L we- A1
(yl<$;-<x<-
2
W + A1
2
L W,- Al
- W , +7
-
sin @) i)i
A1
with Jf(s) as defined in eqn. 10.13 and 1 = kL,. Expanding J,'(I s) in a -3
Al
for I -2
Taylor series around 1 leads to
elsewhere
(10.66)
 546 Transmission-linemodel for rectangular microstrip antennas Transmission-line model for rectangular microstrip antennas 547
This aperture field is an acceptable approximation of the true tangential electric Extensive numerical investigation of this quantity for a large number of par-
field in the plane z = 0 of the microstrip antenna excited in the fundamental ameter values in the ranges w > 0.1, 1 < 3.2 and s < 1 shows the surprising
mode. As shown in Reference 12, it allows an accurate computation of the far result
field and the radiation conductance. The computation is straightforward using
Krf z 1 (10.71)

Therefore, the simple expression G, = G, F, can be considered as valid within

the given parameter ranges. The validity of this expression for both small (> 0.1)
and large values of w can be understood as some kind of compensation: the
influence of the side slots (which is not taken into account in the calculation of
G, and G,) and the influence of the finite length of the main slots (which is
neglected in the calculation of the factor F, in G,) appear to cancel each other
out almost perfectly. To illustrate this effect, Table 10.1 lists the quantities GI
(= 2G,, i.e. 2 x self-conductance of a main slot), G; (= radiation conductance
of the two-slot system consisting of the two main slots, see Fig. 10.12), G:
( = G?, i.e. the radiation conductance of the four-slot system consisting of the
two main slots and the two side slots, see Fig. 10.11) and G y d(i.e. the radiation
conductance found with the transmission-line model of Fig. 10.9 and calculated
from expression 10.69), as a function of w for 1 = 2 and s = 0. The influence
Fig. 10.11 Four-slot radiation model of the side slots can be deduced from a comparison between G; and G:; the
influence of the finite length of the main slots can be seen from a comparison
the plane-wave spectral method. The resulting integral expression for the radia- between Gf"""and Gf. In Reference 9 it is argued that the influence of the side
tion conductance is too complicated for analytical integration, but it can be slots on the radiation conductance can be neglected, but according to Reference
10 a distinctly better correspondence with experiment is obtained if the side slots
evaluated numerically without any difficulty. We call this numerically evaluated
quantity, the reference conductance G/:' and we will use it to verify the accuracy are taken into account as described above.
of the radiation conductance G y d as predicted by the transmission-line model
(Fig. 10.9). The conductance G y d is given by Table 10.1 Radiation conductance for 1 = 2 and s = 0
Gyd = 2(G, + G,) (10.67) w G! G; G,* GYd
where G, is calculated from eqn. 10.55 and G, is expressed to a first approxima- (mS) (ms) (mS) (mS)
tion by eqn. 10.65. Poor correspondence between G r d and Gyf is expected; 1 0.55 0.75 0.69 0.68
therefore we add a correction function Kg to compensate for eventual influence 2 2.1 1 2.84 2.63 2.58
of the side slots and of the finite length of the main slots. Instead of eqn. 10.65, 3 4.40 5.86 5.48 5.38
we use: 4 7.80 9.33 8.79 8.67
5 9.87 12.84 12.23 12.08
G, = GsFgKg (10.68) 6 12.61 16.19 15.58 15.43
The correction function Kghas to be determined by comparison of GYdand GYf: 7 15.26 19.39 18.82 18.68
8 17.86 2254 22.00 21.86
GYd = 2(G, + G,,,) = 2GS(1 + <K,) (10.69)
To obtain a good correspondance between ~ Y " a n dG:'/, thecorrection function
Kg has to be a good approximation to the numerical reference quantity The accuracy of expression 10.65 can also be shown in a systematic way by
computing the relative error:
 550 Transmission-linemodel for rectangular microstrip antennas
we finally obtain for b,
The maximum truncation error of eqn. 10.80 is about 0.1% of b, if s < 1.
In Section 10.2.2 a closed-form expression for b, was derived:
Using eqns. 10.80 and 10.19 in eqn. 10.74 enables to write the auxiliary coupling
function Fb as
We consider the product (B,F,) as a first approximation of the mutual suscep-
tance B, and introduce a correction function Kb so that:
One could expect to apply the same method to determine Kb as is used for K,;
i.e. numerical evaluation of the aperture susceptance of the four-slot system of
Fig. 10.11 and to consider this quantity as a reference for deriving a suitable
expression for Kb. But following problems occur:
(a) It is not clear how a reference susceptance evaluated for a four-slot system
has to be related to the quantity K, of the transmission-line model. We cannot
assume conservation of reactive power passing through two different reference
planes: the aperture plane z = 0 in the equivalent slot model and the input port
plane (e.g. port 1 or 2) in the transmission-line model.
(b) The susceptance of a radiating aperture is much more sensitive to the precise
form of the aperture-field distribution than is the conductance. Consequently
the approximate field distribution of eqn. 10.66 is not appropriate for comput-
ing a reference susceptance. Besides, this field distribution does not even meet
the required edge behaviour.
(c) It is much more difficult to compute the susceptance of a radiating aperture
than its conductance. Using the plane-wave spectral method, the conductance
is given by a surface integral over a finite part of the wave-number plane,
whereas the susceptance requires a surface integration over an infinite domain.
Because of problem (a) we are unable to use a four-slot system to determine the
correction function Kb, but must use the two-slot system of Fig. 10.12 consisting
of the main slots. Hence we neglect the influence of the side slots on the
susceptance.
Transmission-line model for rectangular microstrip antennas 551
Because of problem (b), an aperture field that meets the edge conditions [3]
is required: the tangential components of the electric field perpendicular to the
edge must decrease with distance d from the edge, and the components parallel
to the edge must decrease with d2from the edge. There is no further information
availableto select the appropriate field distribution. Different distributions give
different values for the aperture susceptance. But a detailed numerical evalua-
tion shows that the dependence of the aperture susceptance on the aperture
distribution comes from the dependence of the self-susceptance, not from the
mutual susceptance. For the configuration of Fig. 10.12, it can be concluded
that the uniform aperture-field distribution predicts the mutual susceptance
within 1% of the values obtained using appropriate tapered distributions.
Hence our reference correction function K:/ is defined as
where BZis the mutuai susceptance of the two-slot system of Fig. 10.12 with a
uniform aperture field, and b,,,We is the mutual susceptance of an equivalent
system of length W, taken from two infinite-length slots having the same width
A1 and the same centre distance L,. Numerical investigation shows that K;Pf is
nearly independent of s or I. It can be concluded that an expression having only
the variable w can represent the correction function K,:
K, = 1 - exp ( - 0 . 2 1 ~ ) (10.84)
This equation has the correct asymptotic behaviour for w F m, as the influence
of the finite length of the main slots disappears and Kb has to approach unity.
10.3.6 Expressions for the line parameters
In Section 10.2.3 we discussed the planar waveguide model and the correspond-
ing expressions for the line parameters. Eqns. 10.20-10.28 describe the charac-
teristics of a microstrip line in a quasi-static approach. The frequency depen-
dence of the model parameters can be taken into account through frequency
dependence of and Kff.
A convenient expression for eCffcan be found in References 13 and 14:
where ~ ~ ( is0given
) in eqns. 10.25-10.28 and
P = P,P2{(0.1844 + P,P4)f,)'5763 (10.86)
PI = 0.27488 + (0.6315 + 0.525/(1 + 0.0157S,)20)~
- 0.065683 exp (- 8.7513 u) (10.87)
P2 = 0.33622{1 - exp(- 0.03442~,)) (10.88)
P, = 0.0363 exp(-4.6 u){l - exp [- (f,/38.7)4'97]} (10.89)
 552 Transmission-linemodel for rectangular microstrip antennas Transmission-line model for rectangular microstrip antennas 553

P, = 1 + 2.751 (1 - exp [-(~,/15.916)~]} (10.90) The dielectric losses are given in Reference 17 as
f, = F [ i n GHzmm] = 47.713 kh (10.91) a, = 0.5P -6, & g ( f ) - 1 tan 8 (10.103)
&@(f E r - 1
u = {W + (W' -
W)/&,}/h (10.92)
For &(A an expression has been proposed in Reference 15: The conducting losses are given in References 17 and 18 as
a, = %R,FA,~ (10.104)
a~g = a"R~gF~f (10.105)
with W&(O) as given in eqns. 10.22-10.24; according to Reference 15, K can be with
equal to 1 and
Rm = ,hmz (10.106)
4 = J3& (10.107)

where c, is the free space velocity of light.

However, it has been shown by Pues and Van de Capelle [I61 that a better
asymptotic behaviour for E, b 1 is obtained if

and a better accuracy, particularly for high frequencies, if

FA, = 1 + -x2 arctan {~.~(R,A,u,)~} (10.110)

F, = 1 + -x2 arctan {1.4(R,Agug)2} (10.11 1)

Substitution of eqns. 10.95 and 10.96 in eqn. 10.93 gives a cubic equation from
which Kff(f) has to be solved. There is one real solution:

which gives, with eqn. 10.24 for W',

where

10.4 Application of the improved transmission-line model

R, = (P; + Q;)'I2 (10.100) 10.4.1 Analysis and design of rectangular microstrip antennas
All parameters of the improved transmission-line model have been given in
s, = c,2 (10.101) terms of closed-form expressions. This enables one to program the model very
4f2[&es(n - 11 easily for analysis as well as for design purposes.
The attenuation constant a can be divided into dielectric losses in the substrate The input admittance xn is expressed in eqn. 10.30 and the resonance con-
(a,), conducting losses in the strip conductor (a,) and in the ground plane (a,): dition is defined as
a = ad + am + acg (10.102) Im(&,) = 0 (10.1 14)
 554 Transmission-linemodel for rectangular microstrip antennas
For the case of a microstrip-line-fed antenna the input admittance is given in
eqn. 10.35. The usefulness of the transmission-line model is illustrated by the
simple expressions that can be derived for several important antenna charac-
teristics. For the radiation conductance we find
G, = G,(r + IvI2) - 2G,Re(v) (10.115)
where r is defined in eqn. 10.33; v = voltage-excitation ratio of the main slots
The antenna efficiency follows from
The directivity in the broadside direction is given by
and the antenna gain by
G = qD (10.1 19)
The resonant input conductance is defined as
Gre* = Re ( Kn)l,=,,e, (10.120)
where f,, is the resonant frequency which follows from eqn. 10.1 14.
The antenna input admittance, given by eqn. 10.35, can be modelled fairly
accurately by the resonant input conductance G,, connected in parallel with a
lossless open-ended half-wavelength transmission line with characteristic admit-
tance Y,:
Consequently, the unloaded antenna quality factor is given by
and the impedance bandwidth by using Reference 19,
where S is the maximum value of the voltage standing-wave ratio that is allowed
on the feed line.
Transmission-line model for rectangular microstrip antennas 555
10.4.2 Comparison with other methods
To verify the usefulness of the improved transmission-line model, we compare
it with other published results, theoretical as well as experimental. Fig. 10.14
shows the input impedance of a rectangular microstrip antenna excited by a
microstrip line. The Figure compares measured results of Lo et al. [20], cal-
culated results published by Deshpande and Bailey [21] and calculated results
obtained with the improved transmission-line model. Eqn. 10.35 was evaluated
for: W = 144mm, L = 76mm, W, = 4.3 mm, h = 1.59mm, E, = 2.62,
tan6 = 0.001, t = 0.035mm, a, = as = 0.556 x loSS/mm, A, = As =
0.00 15 mm.
Fig. 10.14 Input impedance of a rectangular microstrip antenna fed by a microstrip line
The moment-method results of Reference 21 agree somewhat better with the
experimental results 1201, than does the transmission-line model. However,
detailed comparison with Reference 21 proves that the transmission-line model
is more accurate than the calculated results of Lo et al. obtained with a cavity
model [20] and of Newman and Tulyathan obtained by a moment method [22].
The discrepancy between the transmission-line model and experiment can be
 556 Transmission-linemodel for rectangular microstrip antennas
explained by the tolerances on the structural parameters. For example, almost
perfect agreement with experimental results was obtained using E, = 2.64 in-
stead of 2.62, and if the losses were somewhat less. Observe that copper losses
are neglected in the calculations of Deshpande and Bailey [21].
10.4.3 Comparison with experimental results
Transmission-line model for rectangular microstrip antennas 557
models (see Fig. 10.16):
Model for the coaxial-microstripline transition [23],
Transmission-line representation of the feed line
Model for the step discontinuity in the microstrip lines
Transmission-line representation of the quarter-wave transformer
Improved transmission-line model for the microstrip antenna

-
The improved transmission-line model is used to analyse a square microstrip
antenna, shown in Fig. 10.15. The antenna is matched to 50 R by a quarter-wave
transformer. The structure has been photo-etched on a RT/Duroid 5880 sub- - - --- - -
strate of 0.03 1 in = 0.787 mm thickness and connected to an OSM-215-3 connector line step line antenna
connector. The measured dimensions of the copper pattern are: e-- - -

L = 33.147 mm, W = 33.165 mm, W, = 0.473 mm, L, = 18.713 mm, Fig. 10.16 Schematic representation of complete analysismodelfor the antenna of Fig. 10.15

-40L I I I
2.90 2.95 3.00 3.05 3.1 0
frequency, GHz
Fig. 10.17 Return loss of antenna shown in Fig. 10.15

Fig. 10.15 Square microstrip antenna fed with quarter-wave transformer
The parameters of the substrate are:
Fig. 10.17 shows the measured reflection diagram and three calculated curves.
To model the antenna element, we have used eqn. 10.35, but with different
values of Y, and Y,,. To simulate Derneryd's model [7] we set Y, = 0; and to
simulate Munson's model [I] we set Y,, = 0 and Y , = Wey,. One can clearly
observe the effect of neglecting the mutual coupling between the equivalent slots
(by comparing Derneryd's model with the improved transmission-line model)
and of neglecting the influence of the finite length of the slots (by comparing
Munson's model with Derneryd's model).

10.4.4 Design application

The antenna has been analysed completely by a cascade of transmission-line The design of microstrip antennas by the improved transmission-line model is
 558 Transmission-line model for rectangular microstrip antennas Transmission-line model for rectangular microstrip antennas 559
very powerful if the model is combined with CAD packages for microstriplines
or striplines. To illustrate this application we discuss the design of a rectangular
microstrip antenna, combined with a broadband impedance-matching network.
The antenna shown in Fig. 1 0 . 1 8 ~ has the following dimensions:
W, = 119.83 mm, L, = 31.68 mm, W, = 4.87mm, L2= 83.32mm. It has
been etched on a RT/Duroid 5880 substrate of 200mm x 150mm x 1.5748
mm. The other parameters (E,, tan6, t , a,, a,, A,, As) are the same as in the
previous Section. Fig. 10.19 shows the calculated and the measured return loss
of this antenna. The best match occurs at 3.025GHz (-21.5dB) and the
improved transmission-line model predicts best match at 3.040GHz
(- 26.96 dB).

Sll log MAG

REF O.OdB
2.5dBi

L I I I I s 1 1 I l l ,
start 2.600000000 GHz
stop 3.600000000 GHz

Fig. 10.19 Return loss of the antenna of Fig. 10.18a

-measured
---- calculated

The impedance bandwidth can be increased with a reactive matching network

Fig. 10.18 Rectangular microstrip antenna
a Fed by a 5052 microstrip line
b Fed by a co-planar impedance-matching network
[24]. A broadband-matching design procedure developed by Pues (US Patent
4445122) can be applied. The design has been carried out combining the im-
proved transmission line model for the antenna element with suitable design
models for the microstrip-network components. The design result is shown in
Fig. 10.18b. The dimensions are as follows:
 560 Transmission-line model for rectangular microstrip antennas Transmission-line model for rectangular microstrip antennas 567
W, = 4.46 mm, L, = 35.95 mm, W6 = 2.65 mm, L6 = 20.50 mm, antenna. It can be observed that the co-planar matching network does not
disturb the radiation characteristics.
This design procedure can also be applied to combine the antenna element
~ b) have been realised by the same etch process
Both prototypes (Figs. 1 0 . 1 8and with a stripline matching network in a multi-layer structure [25].
on pieces of substrate cut from the same sheet. The return loss of the impedance-
matched antenna is shown in Fig. 10.20. Within the band of operation, the worst
match occurs at 3.035 GHz (- 843 dB). The bandwidth at this reflection level 521 log MAG
REF -28.0 dB
(VSWR = 2.14) has been increased by a factor of 3.2 up to a value of 275 MHz 2.OdBl

sll Log MAG

REF O.OdB
2.5 dB1

start 2.600000000 GHz

stop 3.400000000 GHz

Fig. 10.21 Comparison of transmission performance of the antennas of Fig. 10.78

start 2.600000000 GHz
stop 3.600000000 GHz
Fig. 10.20 Return loss of the antenna of Fig. 70.18b
- measured
---- calculated
or 9.1%, whereas the theoretical maximum bandwidth-enlargement factor for
this degree of matching equals 4.0. For further illustration of this application,
Fig. 10.21 compares the transmission performance of the two antennas, which
is proportional to the realised gain. The impedance-matched antenna is a more
efficient radiator, including dissipation losses in the matching network, over the
2432-2.988 GHz and the 3.055-3.174GHz band, whereas the antenna without
impedance matching is more. efficient between both frequency bands. The
maximum gain difference equals 0 6 1 d B and occurs at 3.026 GHz. The co- and
cross-polar radiation patterns have been recorded in E- and H-plane at three
different frequencies: 2.9, 3.0 and 3.1 GHz. The results are shown in Fig. 10.22
for the unmatched antenna and in Fig. 10.23 for the impedance-matched
a Antenna of Fig. 10.188
b Antenna of Fig. 10.18b
10.5 Transmission-line model for mutual coupling
10.5.1 Description of the model
Although more rigorous methods have been developed to calculate the mutual
coupling between microstrip antennas, the transmission-line model provides a
numerically efficient alternative. The mutual coupling is caused by the simul-
taneous effect of:
Interaction through free-space radiation
0 Interaction through surface waves.
The influence of surface waves can be neglected if we confine our attention to
antennas with substrates of small electrical thickness and low permittivity. In
practice, this limitation is not very restrictive.
To develop the transmission-line model for mutual coupling, the following
procedure is used [26]:
 562 Transmission-line model for rectangular microstrip antennas Transmission-line model for rectangular microstrip antennas 563

I I I
-90' O0 90
H-plane, f = 2 . 9 GHz a

Fig. 10.22 Radiation patterns of the antenna of Fig. 10.1Ba

564 Transmission-line model for rectangular rnicrostrip antennas Transmission-line model for rectangular rnicrostrip antennas 565

%
cross

cross

I
-90 O0 90
E-plane, f= 3.1 GHz f
566 Transmission-line model for rectangular microstrip antennas
Transmission-line model for rectangular microstrip antennas 567

I I I
-90 0" 90
H-plane, f=3.1 GHz e

-3 0
cross

I cross

-b
90 0"
1
90
E-plane, f = 3 0 GHz d
 568 Transmission-line model for rectangular microstrip antennas Transmission-linemodel for rectangular microstrip antennas 569
Each microstrip antenna is represented by its improved transmission-line Y: for antenna 2. The mutual admittance between equivalent slots within one
model as described in Section 10.3. antenna is given by Y,!,and Yi, respectively. The mutual admittances between
To model the mutual coupling between different antennas, each antenna is equivalent slots of different antennas are denoted by Y,, Y,, Y,/, Y,,, respec-
replaced by a two-slot system, consisting of the two main slots (Fig. 10.246); tively.
hence the influence of the side slots on the mutual coupling is neglected.
The aperture field in the equivalent slots is assumed to be uniform; the slots
have length Ct: = W&, width S = A1 and a centre distance L, = L Al. +
The transmission line model of each antenna is completed with voltage-
dependent current sources representing the mutual coupling between equivalent
slots of different antennas (Fig. 10.25).

Fig. 10.25 Transmission-line model for mutual coupling between rectangular microstrip
antennas

Consider the simplest case of feeding directly at the edge of each microstrip
antenna (i.e. L,!,= Li = 0 in Fig. 10.25), and assume port 1 of each antenna
is the respective input port. The input admittances Y" of antenna 1, Y22of
antenna 2 and the mutual admittances yL2= y2' between antennas 1 and 2 are
obtained through elimination of the voltages (Vi and v:)at ports 2 of both
antennas:

y" = (y:l)2y2 - y:t(ymr)2 + 2yt!ttymrymf

N
- (y:O2y~ - (ymf)2y:,
Fig. 10.24 Two-slot model for the mutual couplingbetween rectangular microstrip antennas

Fig. 10.25 shows the complete transmission-line model, including mutual coupl-
ing, for the case of two antennas (1 and 2). The self-admittance in the trans-
mission-line model of each individual antenna is denoted by Y,' for antenna 1,
 570 Transmission-line model for rectangular microstrip antennas Transmission-line model for rectangular microstrip antennas 571

y12 = y21 = y, Kl y,: - ym,( ymr)2 + y,: y,: ym, v c

and impressed on the equivalent slots i and j, respectively, and in terms of
N the self and mutual admittances:

with
N = Y,: Y,: - (Ym,)2
YA = Y,' +
Y,'coth(y;l')
Y,!,, = YA - Y,' csch (y;ll)
Y,: = Y: + Y: coth ($I2)
Y
,: = Y; - Y: csch (yi12)
In these expressions a superscript (e.g. i in Y') denotes the number of the
antenna (i = 1 or 2); an exponent is denoted as (Yi)2, which means Yi to the
power 2.
L
.
+*
$ I slot I

10.5.2 Calculation of the model parameters

The transmission-line model for mutual coupling (Fig. 10.25) contains:
Parameters depending on only one antenna Fig. 10.26 Geometrical configuration of two arbitrarily chosen slots i and j
Parameters expressing the mutual admittance between equivalent slots of
different antennas where Y,,, Y;, are the self-admittances of slots i and j, respectively. Y;,, Ti is the
For the first kind of parameters, the expressions derived in Section 10.3 are mutual admittance between slot i and slot j. Owing to the reciprocity theorem
valid: the mutual admittances are equal:
Self conductance Gj = expression 10.55
Self-conductance B: = expression 10.36
Eqn. 10.133 can be developed in terms of the fields E, H or in terms of the
Mutual conductance GL = expression 10.65
Fourier-transformed fields 8, Zf. The last one is suitable for deriving analytical
Mutual susceptance BA = expression 10.82
series expressions (as used in Section 10.2.2, 10.3.3, 10.3.4 and 10.3.5). The
Line parameters, see Section 10.2.3 and 10.3.6
integral expression in terms of the fields E, H is more suitable for direct
For the second kind of parameters, the mutual admittance between equivalent numerical evaluation. Both methods will be considered.
slots of different microstrip antennas has to be calculated. The geometrical To develop eqn. 10.132 in terms of the E, H fields, the aperture fields have to
configuration of two arbitrarily chosen slots i and j is shown in Fig. 10.26. The be expressed in their impressed and induced (by mutual coupling) field com-
slots have length Wj, Wi respectively; width Ali and All, respectively; centre ponents:
distance AxS in the x-direction, and centre distance Ayo in the y-direction.
We start from the expression for the complex radiated power:

where A is the aperture surface of the slots i and j, and E and H a r e the electric
and magnetic fields, respectively, in the aperture plane. where E,, 4, 4 , H; = total tangential fields, slots i, j; E,,,, Ha,, E,, H, = im-
The complex radiated power can be written in terms of the impressed voltages pressed tangential fields, slots i, j; E,,, Hij, El,, Hli = induced tangential fields,
 572 Transmission-line model for rectangular microstrip antennas Transmission-linemodel for rectangular microstrip antennas 573
slots i,j. Eqn. 10.133 is written in terms of impressed voltages; consequently eqn. where hji is a dyadic Green's function giving the magnetic field at a position r,
10.132 has to be developed in terms of impressed electric fields in order to enable on slot j caused by a unit magnetic current source located at position ri on slot
identification of terms. Consequently: i; and K,, is the equivalent magnetic surface current replacing the aperture field
Ei. = 0 (10.139) Eaiby a current source in free space:
E,, = 0 (10.140)
and The impressed aperture fields Eaiand E, occurring in the two-slot system of Fig.
10.246 have only a -y-component. Consequently, K,, and K,,,, have only an
x-component:
K,, = -2i, x i,E,, = 2E,,ix (10.151)
+ illAJEd x
1
(H: + H $ ) i , dxdy Km~ = -i, x i,E, = E a1. i* (10.152)
Applying eqns. 10.148-lO.l52 in eqn. 10.147 enables to write

We denote the xx-component of the dyadic Green's function as e.

This
depends only on the distance lr, - r,l between observation point and source
(10.141)
point. The impressed field distributions were assumed to be uniform in the
Equating eqns. 10.133 and 10.141, and identification of terms, gives: equivalent two-slot system of Fig. 10.26; consequently:
Y$ F V) = JJA,
(i, x E,,) .H*, dx dy (10.142)

After substitution:

Y$~v:= JJAJ (i, x ~ , ) . ~ $ d x d y (10.145)

Eqns. 10.142 and 10.143 determine the self-admittances Y,and The mutual x,. the integration with respect to v can be performed analytically and the four-
dimensional integral can be reduced to a two-dimensional one, which improves
admittance F, = can be calculated from eqn. 10.144 or 10.145:
the numerical efficiency and makes this expression very suited for direct numeri-
cal evaluation.
In order to avoid numerical evaluation of integral expressions, and to further
improve the efficiency, analytical series expressions have been derived [27].
These expressions are obtained starting from the expressions for the mutual
admittance in terms of the Fourier-transformed aperture fields.
We proceed with eqn. 10.147. The vector product
10.5.3 Comparison with other methods
In order to check the validity of the transmission-line model for mutual coupl-
where K, is an equivalent magnetic surface current. ing, comparison with published theoretical and experimental results will be
The induced magnetic field q., in slot j is caused by the impressed field E,, in discussed.
slot i. This induced field can be expressed as First we compare with the experimental results published by Jedlicka and
Carver [28]. It Gals with the mutual coupling between two identical microstrip
qi = A, hji . K,, dx dy (10.149) antennas in the E-plane (Fig. 10.27) and in the H-plane (Fig. 10.28). The
 574 Transmission-line model for rectangular microstrip antennas Transmission-line model for rectangular microstrip antennas 575

coupling is measured as a function of the distance between the patch edges,
which is normalised with respect to the free-space wavelength 1,.Neither the
dimensions or the location of the feed probe, nor the permittivity of the sub-
strate are mentioned in Reference 28. As the microstrip antenna is matched at
1.405 GHz and the permittivity was estimated in Reference 29 to be 2.50, we
obtain a probe diameter of 0.3 mm and a distance between the feed point and
the edge of the patch equal to 20.0mm. To account for the probe inductance,
Harrington's formula (Reference 30 pp. 378), has been used. The calculated
results obtained with the transmission-line model are shown by the solid line in
model, it sometimes gives better agreement with experiment than the trans-
mission-line model, except for very small distances ( < 0.1 &), where Penard's
model does not give satisfactory results owing to neglecting the slot width.
In Fig. 10.31 the transmission-line model is compared with theoretical and
experimental results published by Malkomes [321. In this case the mutual
000 Carver (Exper.)
.... Pozar (Theor.)
- Transrn. (Theor.)

Figs. 10.27 and 10.28. In the E-plane, the correspondence between the trans-
mission-line model and the experimental results is quite good; the largest
difference occurs at a small distance of 0.2 & . In the H-plane, the correspondence
is less good, as can be expected from neglecting the side slots.

000 Carver (Exper.)

.... Pozar (Theor.)
- Transrn. (Theor.)

ho
Fig. 10.28 Mutual coupling between two rectangular microstrip antennas in the H-plane
Wx=66mm Wv=105.6mm f=1.405GHz
h = 1.5785mm e, = 2.50 W
,, = 20.0 mm

000 Penard (Exper.)

x x x Penard (Theor.)
- Transrn. (Theor.)

Fig. 10.27 Mutual coupling between two rectangular microstrip antennas in the E-plane
W, = 66 mm Wv = 105.6 mm f = 1.405 GHz
X$ -20 ...
0
h = 1.5785 mm e, = 2.50 W,, = 20.0 mm A
.......O 0 0
2 2.. .....X. ......
.............Y ........... ............0
In order to compare the transmission-line model with more complicated -30 -
models, Figs. 10.27 and 10.28 also show by dotted line the results published by
Pozar [29]. It is a moment-method solution, which uses the grounded-dielectric-
slab Green's function to account for the presence of the substrate and surface
waves. This method gives, as expected, excellent agreement with experiment,
even though the calculations were made with at most three expansion functions.
Another theoretical method, due to Penard [31], is based on a cavity model Fig. 10.29 Comparison of cavity model and transmission-line model for mutual coupling in
where the mutual coupling is considered as the coupling between two current the E-plane
loops around the patch surface. This method takes into account all the equiv-
alent slots around the patch, but neglects the width of the slots, the surface
W,=40rnm
h = 1.52mm e, -
W,=6Omrn
2.55
f=1.548GHz
W,, = 12.5mm

waves, the variation of the ideal field distribution in the slots versus frequency coupling is given as a function of the centre distance between the microstrip
and the higher-order cavity modes. Figs. 10.29 and 10.30 compare Penard's antennas. The antennas are fed by microstrip lines and the measurements are
results with the transmission-line model for a severe test case, i.e. mutual done at a frequency of 4.77GHz. Malkomes' theoretical results have been
distances smaller than 0.5 1,. Because of the presence of the side slots in Penard's obtained by taking into account all the cavity modes and the four equivalent
 578 Transmission-line model for rectangular microstrip antennas
10 PUES, H., and VAN DE CAPELLE, A.: 'Accurate transmission-line model for the rectan-
gular microstrip antenna', IEE Proc., 1984, 131H, pp. 334-340
11 KIRSCHNING, M., JANSEN, R., and KOSTER, N.: 'Accurate model for open end effect
of microstrip lines', Electron. Lett., 1981, 17, pp. 123-125
12 HAMMER, P., VAN BOUCHAUTE, D., VERSCHRAEVEN, D., and VAN DE
CAPELLE, A.: 'A model for calculating the radiation field of microstrip antennas', IEEE Chapter 11
Trans., 1979, AP-27, pp. 267-270
13 KIRSCHNING, M., and JANSEN, R.: 'Accurate model for effective dielectric constant of
microstrip with validity up to millimeter-wave frequencies', Electron. Lett., 1982, 18, pp. Design and technology of low-cost
272-273
14 WHEELER, H.: 'Transmission-line properties of a strip on a dielectric sheet on a plane', IEEE printed antennas
Trans., 1977, MlT-25, pp. 6 3 1 4 7
15 OWENS, R.: 'Predicted frequency dependence of microstrip characteristic impedance using J.P. Daniel, E. Penard and C. Terret
the planar-waveguide model', Electron. Lett. 1976, 12, pp. 269-270
16 PUES, H., and VAN DE CAPELLE, A.: 'Approximate formulas for frequency dependence
of microstrip parameters', Electron. Lett., 1980, 16, pp. 870-872
17 HAMMERSTADT, E., and BEKKADAL, F.: 'Microstrip handbook', ELAB Report STF44
A74169, Trondheim, Norway, Feb. 1975
18 BAHL, I., and GUPTA, K.: 'Average power-handling capability of microstrip lines', IEE J. 11.1 Introduction
Microwaves. Optics & Antennas, 1979, 3, pp. 1 4
19 VANDESANDE, J., PUES, H., and VAN DE CAPELLE, A,: 'Calculation of the bandwidth In the last decade printed-array antennas have received increasing attention for
of microstrip resonator antennas'. Proc. 9th European Microwave Conf., Brighton, Sept. 1979,
pp. 116-119
applications in various communication and navigation systems. Microstrip
20 LO, Y., SOLOMON D., and RICHARDS, W.: 'Theory and experiment on microstrip patches can be very efficient candidates for inexpensive antennas when narrow
antennas', IEEE Trans., 1979, AP-27, pp. 137-145 bandwidth (typically less than 5%) and medium gain are required (15-25 dB).
21 DESHPANDE, M., and BAILEY, M.: 'Input impedance of microstrip antennas', IEEE However, divergence in substrate parameters and manufacturing tolerances
Trans., 1982, AP-30, pp. 645-650 means that a wider frequency bandwidth and a better control of radiation
22 NEWMAN, E., and TULYATHAN, P.: 'Analysis of microstrip antennas using moment
methods', IEEE Trans., 1981, AP-29, pp. 47-53
characteristics are necessary in the mass production of printed antennas.
23 PUES, H., and VAN DE CAPELLE, A.: 'Computer-aided experimental characterisation of Thus simple but accurate investigations of radiating elements are necessary to
microstrip-to-coaxial transitions'. Proc. 14th European Microw. Conf., Litge, Sept. 1984, pp. obtain the design requirements. Analysis of normal-shaped patches and slots
137-141 can be developed using both known models (transmission-line or cavity models)
24 PUES, H., VANDENSANDE, J., and VAN DE CAPELLE, A.: 'Broadband microstrip or more elaborate theory (spectral-domain-apprbach).
resonator antennas'. Int. IEEE/AP-S Ant. &Prop. Symp. Digest, Washington, May 1978, pp.
268-27 1 On the other hand, the design of planar arrays requires a thorough knowledge
25 PUES, H., VAN LIEBERGEN, H., THISSEN, L., NAUWELAERS, B., and VAN DE of typical properties of printed linear sub-arrays such as directivity versus
CAPELLE, A.: 'Broadband multi-layer microstrip antenna'. Proc. MIOP '87 Conf., Wies- spacing, mutual-coupling effects, losses etc; for this purpose, simple formulas
baden, May 1987 and analyses have been developed. Two-dimensional arrays, with non-identical
26 VAN LIL, E., and VAN DE CAPELLE, A.: 'Transmission line model for mutual coupling sub-arrays, such as cross-fed structures, are well suited for low-cost antennas.
between microstrip antennas', IEEE Trans., 1984, AP-32, pp. 816-821
"'
NAUWELAERS, B., VAN DE CAPELLE, A,: 'Formulas for the calculation of mutual
Design equations and curves are included.
coupling between rectangular microstrip antennas'. Proc. Int. Conf. Ant. & Prop., April 1985, Microstrip patches exhibit different E-plane and H-plane radiation patterns;
Coventry, pp. 99-102 analytical synthesis methods (Fourier, Chebyshev etc.) are not always suitable
JEDLICKA, R., and CARVER, K.: 'Mutual coupling between microstrip antennas'. Proc. for a small number of sources or when the pattern is specified by a given outline.
Workshop on Prmted Circuit Antenna Technology, Las Cruces, pp. 4-114-19, Oct. 1979 Two new numerical synthesis methods, taking into account the directivity
POZAR, D.: 'Input impedance and mutual coupling of rectangular microstrip antennas',
IEEE Trans., 1982, AP-30, pp. 1191-1 196 pattern of sources with equal or unequal spacings, are proposed.
HARRINGTON, R.: 'Time-harmonic electromagnetic fields', (McGraw-Hill, NY, 1961) Numerical programs can be implemented on conventional personal com-
PENARD, E., and DANIEL, J.: 'Mutual coupling between microstrip antennas', Electron. puters.
Leu., 1982, 18, pp. 605-607 Cost reduction will necessitate common microstrip laminates or new polymer
MALKOMES, M.: 'Mutual coupling between microstrip patch antennas', Electron. Lett., substrates exhibiting good mechanical and electrical properties (typically a low
1982, 18, pp. 520-522 dielectric constant of about 2-2.5 and losses of tan6 z A new low-cost
DANIEL, J., VAN DE CAPELLE, A,, and FORREST, J.: 'Microstrip patch arrays for
satellite communications'. ESAICOST 204 Phased Array Antenna Workshop, ESTEC, polypropylen! whose fabrication process is quite simple has been developed at
Noordwijk, June 1983, pp. 9-14
 580 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 581

CNET Lannion, France. This substrate can be made as a multi-layer structure OMA with four magnetic walls (as in Fig. 11.1~)
or with thick metal backing. HMA with three magnetic walls and one electric wall (as in Fig. Il.lb)
The treatment considers perfect magnetic or electric walls and groups all anten-
na losses together in an effective dielectric loss tangent determined by an
11.2 Analysis of simple patches and slots
iterative process (Fig. 11.2). In this process it is necessary to calculate the electric
The microstrip antenna designer needs a method of analysis (not too time and magnetic energies We and 4,stored in the cavity.
consuming) to calculate, as nearly as possible, the parameters of interest:
resonant frequency, Q-factor, input impedance, pattern etc. Moreover, he
should be able to evaluate the surface-wave effects and to take into account the
superstrate applied to the antenna as a protective layer.

11.2.1 Rectangular and circular patches

Two simple techniques can be used for the analysis of microstrip antennas:
transmission-line model [ I , 2, 31 and cavity model. More sophisticated analysis
will be presented for multi-layered structures.

11.2.1.1 Cavity-model analysis: Cavity-model analysis [4] allows modal field

description and gives good results depending on the effective parameters ob-
tained from previous microstrip line formulas, as long as the substrate thickness
is thin compared with the wavelength. This method has often been used with
magnetic walls to calculate the electrical properties of open microstrip antennas
(OMA) with simple shapes [5, 61 (cf. Chapter 3). Its application to the rectan-
gular hybrid microstrip antenna (HMA) [A permits a comprehensive com-
parison between OMA (Fig. 1 1 . 1 ~and) HMA (Fig. I l.lb) antennas [8]. It can
also be easily applied to multi-port microstrip antennas [9, 101.
In the cavity method, according to the Lo and Richards model, the total
interior field with a unit z-directed excitation current at (x,, xo) is given by

(b)
Fig. 11.I (a) OMA with four magnetic walls. (b) H M A with three magnetic walls and one
electric wall
with

sin (x) These quantities are, of course, evaluated by integrating the electric E and
J,(x) = - magnetic H fields over the cavity volume V. In fact, the exact expression of We
X
and Wh can be obtained in a very simple manner without performing any
integration at all, as explained in the following [I I]. The total power P i s injected
into the cavity as
km, is the wave resonant number. The eigen functions $
,, are solutions of the P = VIT = V = -tEl(xo, yo) (1 1.2)
Helmholtz equation with different boundary conditions: where V is the voltage at the source point. On the other hand, the same power
 582 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 583
can be expressed as a function of W, and W,:

Which leads to the following relationships: Where Re and Im stand for real and imaginary parts, respectively.
Note that w, is determined without using the H field. Eqns. 11.4-11.5 are
obtained independently of the mathematical method employed to derive the
expressions for the ELfield or of the geometry of the cavity section, i.e. the patch
shape. Thus eqns. 11.4 and 11.5 are general formulas valid for planar microstrip
antennas of arbitrary shape. However, the cavity thickness t must be quite small
compared to ,Ie in order to ensure the validity of expression 11.2.
Read :
The analysis yields input impedance, resonant frequency and Q-factor; the
radiation pattern is found from the knowledge of the interior field. Two meth-
ods are available:

(a) Electric-current method

The radiated field is found from the electric currents J, flowing on the patch and
on the electric walls. If there is a dielectric, volume polarisation currents J, must
be added to the cavity region (Fig. 11.3) [12, 13, 14, 151.

I 5-
0 / b Y
ground plane

Fig. 11.3 Electric currents and polarisation currents in an HMA

(6) Aperture-radiation method

One application of the equivalence theorem [16] is to choose a perfect closed
electric surface S with null field inside and equivalent magnetic sources outside:
Far f i e l d Ee,EO M, = E x n; n being the normal outward from S. S defines the limit of an
calculation homogeneous exterior medium and must be carefully selected.
As the rectangular OMA structure has been extensively described, only the
HMA is considered here; e.g., the choice of S for HMA is indicated in Fig. 11.4.
The surface S, around the cavity (Fig. 11.4~)is not permitted, except in the
Far f i e l d keep case of a vacuum medium ( E , = 1). It seems more convenient to choose S, (Fig.
11.4b) lying on the z = t plane, because it is always available whatever the
medium, i.e. E, = 1 or E, # 1.
In practice, electric walls are not perfect conductors, but the electric-field
(+\ distribution in the vicinity of the electric wall (Fig. 11.4~)suggests that antenna
radiation occurs mainly from the magnetic walls. This assumption is well
Fig. 11.2 Flow chart of the cavity method
confirmed by experiment. It may be assumed that HMA radiation is entirely due
584 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 585
to the magnetic walls, i.e. three-slots array. Experimental and theoretical results
for an HMA are given in Fig. 11.5.
The impedance loci in the case of the dominant mode (0, 1) are given in Fig.
11.5~.All the modes have been taken into account in the computation. The
antenna was edge-fed by a coaxial-line probe at a position x, = 3.5cm, yo = 0.
Two methods were used for the radiated field (equivalence theorem and electric
current plus polarisation current) and both agree well with experimental data.
The experimental patterns in planes 4 = 0' and 4 = 90" are plotted in Figs.
11.56 and 11.5~.

\
wall electric w a l l

ground plone /
C

Fig. 11.4 Application of the equivalence theorem to HMA -60 -30 0 30 60 -60 -30 0 30 60
a First choice of surface S, b 0 , degrees c 0 , degrees
b Second choice of surface S2
Fig. 11.5 Experimental and theoretical results i n the case of an HMA
c Local geometry of the electric field of an HMA near the shon circuit
a = 6.00cm. b = 4.00cm, 8, = 2.56
t = 0.146 crn, tan 6 = 0.001, mode(0, 1)
The uniform aperture lying along the Ox-axis gives an omnidirectional pat- (a) Impedance loci
tern in the plane 4 = 90' (Fig. 11.5c), the cross-polarisation level being very Feed point xo = 3,5cm, yo = 0
0-0 experiment
low ( < 30dB). The discrepancy between the patterns computed from the electric
0 magnetic currents
currents alone (broken line) and from the aperture (large broken line) appears X electric currents and polarisation currents
clearly in this Figure; it confirms the importance of the dielectric slab. Polarisa- (b) Plane 4 = 0"
tion currents must be included in the far-field calculation. They are z-directed -experiment
and therefore contribute only to the E, components. Therefore, the electric- ---- theory
current approach gives a useful physical insight into the radiation properties, (c) Plane 4 = 90"
-experiment
particularly regarding the major influence of the dielectric constant on the E, ---- theoretical pattern: electric currents only
component. --- theoretical pattern: magnetic-currents method
586 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 587
In the plane 4 = 0, the radiated field of the two slots lying along the y-axis
leads to a high level of cross-polarisation radiation with a null in the broadside
direction (the magnetic currents are each 180" out of phase). The diffraction [17]
at the edge of the ground plane appears particularly in the oscillations of the
principal component E, in the planes 4 = O0 (notably for 0 = 30' and
0 = 90"). Nevertheless, the theoretical results are again well confirmed by
experiment.

-60 -30 0 30 60
a 8 . degrees 8,degrees

Fig. 11.6 Experimental and theoreticalpatternsin for the HMA with one electrical wall
a = 6.00 cm, b = 4.00cm. &, = 2.56, t = 0.148 cm, mode(1 .l)
-experiment
---- theory
(a) Plane 4 = 0"
(b) Plane 4 = 90"

The high-level cross-polarisation can be reduced when the exciting mode is (1,
I). The radiation pattern in the 4 = 0" plane (Fig. 11.6~)is therefore due to an
array of two slots directed along the y-axis and has a cosine field distribution.
The cross-polarisation due to the x-slot is smaller than for mode (0,1) because
of the phase inversion along the x-apertures, and is cancelled in the plane
4 = 90" (Fig. 11.6b, theoretical results only). The important reduction of
beamwidth is worthy of note here.

Feedposition: As in the OMA case, it is possible to match the HMA by properly

selecting the feed position along the y-axis.
When considering the dominant mode alone and its resonant frequency
+
(given by k2 = k2mn), the input impedance Z(= R, j&), which becomes real
at resonance, can be simply expressed as a function of the feed position xo,yo.
For the HMA with one electrical wall

where R,(O, 0) is the input impedance for xo = 0, yo = 0 (corner feed).

For example, in the case of the one-electric-wall antenna and the mode (I, I),
588 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 589
the impedance loci are given in Figs. 1 1 . 7 ~and 11.76 for two feed positions
xo = 3.5 cm, yo = 0 and xo = 4.5 cm, yo = 0 . It can be seen that the impedance
decreases as the feed moves along the x-axis toward the centre of the edge. The
discrepancy between measurement and the theoretical results is due first to the
experimental error in the feed position and secondly to the feed model, which
does not take into account the disturbance of the internal field distribution near
the probe.

( c ) Comparison between HMA and OMA characteristics

The different parameters, Q-factor, input impedance and radiation patterns of

the OMA and the HMA, can be compared for the same electrical dimensions.
Both act in the dominant mode (0, 1 ) and at the same resonant frequency. The
wave number ko in free space is given by

Table 11.1 Comparison of theoretical impedance and Q-factor

HMA OMA
Experiment Theory Experiment Theory

Table 11.2 Comparison of cavity method with Woo& method [7]

OMA HMA
Wood Cavity Wood Cavity
method ' method
f,,MHz 1275 1257.2 12750 1278
AJ MHz 26 22.4 22.5 23

with b' the OMA dimension along the y-axis and b the HMA dimension along
the y-axis; therefore b' = LJ2 and b = 6,/4 (A, is the wavelength in the dielec-
tric).

Vacuum medium, E, = 1: The theoretical impedance and Q-factor are com-

pared in Table 11.1 with experimental data for both antennas. In this case the
Q-factor of the HMA is smaller but the impedances have approximately the
same value. The slight discrepancy between experiment and theory is due to the
failure of the magnetic-wall model when E, = 1.
 Design and technology of low-cost printed antennas 591
590 Design and technology of low-cost printed antennas
Dielectric medium case, E, = 2.5: Theoretical results for the bandwidth and
resonant frequency are compared with Wood's measurements [7] in Table 11.2.
There is an important difference compared with the previous case: here the
bandwidths of the two kinds of antenna are very similar.

(d) Theoretical results for different antenna lengths all,

Q-factor, input impedance: In Fig. 1 1 . 8 the

~ variation of the input resistance at
resonance for the HMA and OMA is given as a function of the parameter all,
for E, = 1 and E, = 2.1. The Q-factor variation is given in Fig. 11.8b; for all the
different cases considered it can be concluded that, for E, = 1 the Q-factor of
the HMA is lower than that of the OMA (the bandwidth is therefore larger).
However, the impedances of OMA and HMA are very similar.
For E, = 2.1 the HMA and OMA have approximately the same Q-factor (and
therefore the same bandwidth), but the input impedance of the HMA is twice
that of the OMA. These important differences between the vacuum medium and
the dielectric medium can be explained by the contribution of the polarisation
currents. Their effect should be more important for the HMA than for the OMA
owing to the cosine distribution of the internal field, which has no change of
phase along the y-axis (Fig. 11.4~).For a full comparison of the two antennas,
the cross-polarisation level must also be considered.

Cross-polarisation level as a function of all,: Theoretical results are given in Fig.

11.9 in the plane 4 = O0 for different values of 0 (0 = 20, 60') and for
E, = 2.1. The cross-polarisation Cp is defined here by

It may be seen that, for the HMA, the cross-polarisation remains very high, and
therefore this antenna is not suitable for use in an electronically scanned array
(except when very small scanning angles are used).
In conclusion, the HMA is shorter (lJ4 long) than the OMA (lJ2 long), each
having the same resonant frequency for the mode (0, 1). A closer comparison
of these two kinds of antenna shows that the dielectric plays a fundamental role
in the far-field pattern, the input impedance and the bandwidth.
When E, = 1, the impedance of the HMA and OMA have the same values,
whereas the HMA bandwidth is twice as broad. On the other hand, with the
usual dielectric (E, = 2), the bandwidths have the same value, but the impedance
of the HMA is twice that of the OMA.
Fig. 11.9 Comparison of the cross-polarisat~on level of the OMA and HMA for different
11.2.1.2 Spectral-domain approach (SDA): As the frequency of operation is values of & / a . Results are given for the plane q5 = 0"
increased, more complex analyses are necessary to rigorously account for the a OMA
0 HMA
effect of the dielectric substrate, which may couple surface waves. Recently, ---- E, = 1
quite efficient approaches based on integral formulations and numerical resolu- -E, = 2.1
( a ) 8 = 20"
( b ) 0 = 60"
 592 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 593
tions were proposed [18-231. These methods use the exact Green's function for E; = E,E,, (i = 0, 1, 2) is the complex permittivity
the grounded dielectric slab, and hence the results depend on the numerical yf = a2 + f12 - k;; k, = E,, ki = W ~ ~ , , & E , ;
techniques used to calculate these functions accurately.
(a exp jut time variation is assumed)
(a) Resonant frequency and quality factor
Among these approaches, the SDA [20] is particularly suitable for determining
the resonant frequency and Q-factor of patch antennas embedded in dielectric
substrates. The structure and the co-ordinate system employed are shown in Fig.
11.10. In SDA, the Fourier transform of the dielectric field E,(a, P), E,(a, P) at
2 = d, in region 2 is related to the current distributions J,(a, P), J,(a, P) on the

The zeros of dte and dtm define the surface-wave poles in the composite layer,
the dominant mode of which, TM,, is always above cut off regardless of slab
thicknesses.
The matrix M can easily be obtained in terms of an equivalent transmission-
line circuit as presented in Fig. 11.11, by generalising the method of Reference
20. Using the moments method and a modal representation of J, a matrix
equation is derived:

where the elements of matrix K are given by double integrals on a and P. With
the transformation a = kQsin 8, /I= k, cos 0, we have, for example

The condition det [K] = 0 is only satisfied by a complex frequency f = f ' +

6'that
' leads to the resonant frequency f' and the quality factor Q = f ' / 2 f nof
this radiating open resonator. In this free regime, ki is complex with Im [k,] > 0,
and a proper Riemann surface must be defined for evaluation of yi = ($ -
k f ) ' I 2 . The sign of y,, y, does not affect the value of the integrals, as the terms
Fig. 11.10 Patch antenna with dielectric layer involving the radicals are even functions of y,, y,. For this reason, only the
d, = protective dielectric-layer thickness branch-cut contribution by the radical yo has to be considered, the branch cuts
d, = dielectric-substrate thickness being shown in Fig. 11.12. It can be shown that this condition involves an
(W, I) = width and length of the patch integration path in the kQcomplex plane and not only on a real axis as in
Reference 20. Furthermore, it is worthy of note that the surface-wave poles lie
conducting patch by within the range k, + max [k,, k,].
A possible position of an integration path in the kQ-planeis shown in Fig.
'
ws, dte dtm ~ ( y , . 4 ) [ 8 ]
1 1.12, with A % Im [k,], taking into account the presence of the poles.
In Table 11.3 and Fig. 11.13, the present method of analysis is compared with
available theoretical and experimental results on resonant frequency and quality
where y, is the propagation constar~tin the z-direction in the i t h region and factor.
594 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 595
Generally, the resonant frequency& of a microstrip antenna is defined as the
frequency giving the input reactance X equal to zero. For electrically thin
substrates,& is also very close to the frequencyS, where the input resistance R

Table 11.3 Measuredandpredicted f, and O of a rectangular patch antenna:

TM,, and TM,, modes
Method & Q A Q
TM 10 TMo,
SDA [24] 1.5458 102.7 2.2616 44.68
Cavity [25] 1.548 112 2.272 47
Dubost [3] 1.52 105 2.3 36.2
EXP 1251 1.53 97 *
10 2.237 44 +-
6
W = 60 mm; 1 = 40 mm; d, = 1.46 mm; E, = 2.56; tg A, = lo-' (d2 = 0)

Table 11.4 Comparison between the present method and results predicted
by Bahl[26]

" 4 (mm) f, GHz Q (%)

T.M. woves T.E.waves

Fig. 11.11 Trahsmission-linemodel in the spectral domain SDA 4.092 33.68 0

0
Experience 4'1 04 0
SDA 3.986 34.52 2.62
0.8
Experience 4.008 2.34
SDA 3.9336 33.49 3.9
1.59
Experience 3.934 4.14
SDA 3.8767 31.82 5.09
3.18
Experience 3.895 5.09
W = 19mm; I = 22.9mm; dl = 1.59mm; e,, = 2.32 tg A, = lo-'; e,, = 2.32; tg A, = lo-'

-
f,
= -
f*'
f#
- f, = resonant frequency without protection

reaches a maximum, R,,,. In coaxial-feed excitation, as the substrate thickness

increases,& lies farther fromf, owing to an inductive shift of the reactance curve.
Furthermore, the bandwidth, defined in terms of impedance, is affected by this
' \ inductive shift. In order to remove the disturbance of the feed, and to compare
Fig. 11.12 Branch-cuts position and path integration in the k, plane measured and predicted results, the resonant frequency and Q-factor were taken
596 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 597

and

It should be noted that the SDA method gives

f = f' +JY"
with

The effect of a protective layer on the resonant frequency and Q-factor of a

rectangular microstrip antenna is shown in Table 11.4. In this case the Q-factor
varies slightly, while the resonant frequency clearly decreases.
Additional experiments were carried out on a rectangular patch antenna for
several different thicknesses of superstrate and three different values of the
superstrate dielectric constant: E,, = 2.17, 2.55 and 3.6. The results are presen-
ted in Fig. 11.13. It should be noted that, for d J l , < 0.015, the theoretical
Q-factor increases slightly. This corresponds to a better matching of the antenna
to free space and a maximum efficiency of radiation. As d, increases, so does the
coupling of energy with the superstrate by the surface waves, and radiation
efficiency decreases while the bandwidth increases. It is worthy of note that, for
a given thickness, Af,/A and bandwidth increase with increasing E,,.

( b ) Input impedance
The input impedance can be determined using Richmond's reaction equation in
spectral domain. If the real axis is used as the integration path, the residue
contribution of the surface-wave poles has to be taken into account [21]. In our
approach this is avoided by using the previous integration path in the complex
plane (Fig. 11.12). Fig. 11.14 shows our theoretical results compared with those
of Bailey et al. for an antenna without (with) dielectric cover.

Fig. 11.1 3 (a) Effect of dielectricloading on resonant frequency and Q-factor of a rectangular
microstrip antenna
I = 22.9 rnrn, W = 19 mm, E,, = 2.1 7, dl = 1.58 rnrn
tan A, = 6.0 x 10-4, 6 , = 2.17, tan A, = 6.0 x 10-4
(b) Effect of dielectric loading on f, and Q-factor of a rectangular patch antenna
1 = 22.9 mm, W = 19rnrn. E,, = 2.1 7, dl = 1.58 rnrn
tan A, = 6.0 x 10-4, E, = 2.55, tan A, = 10-3
(c) Effect of dielectric loading on f, and Q-factor of a rectangular patch antenna
I = 22.9rnrn. W = 19 mrn, E,, = 2.1 7, dl = 1.58rnrn
tan A, = 6.0 x lo-', ,&, = 3.6,tan A, = 2 x 10-3
 598 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 599
(c) Radiation efficiency [27, 281 region 0 are given by
The SDA can also be used to calculate the influence of surface waves on the
radiation efficiency of a rectangular patch antenna. Once the problem is solved
for the resonant frequencyf,, far-field radiation may be obtained by using the
inverse Fourier transform. Hence the electric-field components in each point of
. x (Z - d)) exp (- j(ax + by) dadp
q = (x, Y , Z) (d2 = 0,d, = d )
with
yo = ,/a2 + j2- GRe[yO] > O,Im[yo] > O,ko = 0&(11.13)
and w real.
To compute the far field, eqn. 11.13 is first transformed into spherical co-
ordinates and secondly transformed to specify the (a, b) plane in terms of a
spherical-polar angle (u, v). Branch points of yo have thus disappeared, and the
resulting integral can be evaluated by the saddle-point method, provided that
the contribution of the integrand poles is correctly taken into account.
The integral along the steepest-descent path corresponds to the space-wave
component, while the summation of the residues of these poles is related to the
TM- and TE-mode surface waves. Only the dominant lowest-order TM mode
which may be excited for the usual values of d/loand E, has been considered here.
After integration, the electric far field may be expressed as

I
where Eel,Emland E,, are expressions for the space and surface waves, respec-
tively. Note that the space-wave components are the Fourier-transformed inter-
face (z = d ) fields, and may be expressed as
I
Eel a sin W&
,, B) + cos 4E&, b)
(11.15)
Em,a cos O[cos 4EN(a, 8) - sin q5Ed(a, p)]
with
a = kosinOcosq5
p = kosinOsin+
The efficiency of space-wave launching can be calculated as the ratio of
radiated power P, to total (radiated plus surface-wave) power:

Fig. 11.14 Input impedance of microstrip antenna without (with) dielectric cover compared
with Bailey's results (Reference 2 1 )
9--0 Bailey et a/. with
X X Our results
A-A Bailey eta/.
0 0 Our results
600 Design and technology of low-cost printed antennas
zo = Jmand 5, = d m -= k,sinup
Q = rsin9 pole in the 5 plane
E, having been deduced from EC Furthermore, the gain may be defined by
Fig. 11.15 Radiation efficiency and gain against dl& for rectangular patch antenna
W = l 5 m m ; I = 10mm. d = l.58mm. E, = 2.35, f, = 8.58GHz
Fig. 11.15 shows numerical results for q and G against dielectric thickness d
normalised to 1, for a rectangular patch antenna fed along the x-direction. An
optimum value of gain for d x 0.151, with q x 95% is observed. For
d > 0.151, the increase of P, causes a slight reduction of the gain, but the first
TE-mode surface waves excited for d x 0.2161, are not considered here. Identi-
cal results have been obtained for microstrip disc antennas [27] for large values
of dielectric constant.
(d) Surface-wave effects on radiation patterns [24, 28, 291
Radiation into the horizon: The radiating Eipfield for printed antennas with an
infinite dielectric layer tends to zero along the horizon (9 -+ n/2). However, in
many cases the dielectric layer should be considered truncated after a certain
distance a (Fig. 11.16), and the surface waves radiate some of their energy when
they reach this discontinuity. If the radius a is sufficiently large, the far-field E,
of the surface wave may be used to calculate the corresponding far-field E,".
Fig. 11.17 shows the radiation pattern in the E-plane ( 4 = 0") for the total
+
(Eip E,") electric field of the rectangular patch antenna with a truncated
dielectric radius a x 3.51,. Note that, for 9 = 90, the far field is due only to
the components of the surface wave. In the H-plane ( 4 = 90), ET = 0, and the
total field varies as cos 0.
Design and technology of low-cost printed antennas 601
11.2.2 Conical antennas
In some cases, the narrow bandwidth of microstrip patch antennas continues to
be the main constraint. The widest bandwidths are likely to be achieved with
thicker substrates 1301, but higher modes and surface waves are limiting factors.
Replacing a conventional circular disc by a solid conical patch of the same
radius, a new type of 'microstrip antenna' [31] is obtained. This structure
grouhd plane
Fig. 11.16 Microstrip patch antenna with a truncated dielectric layer of radius a
0 (degrees)
0" 7.
Fig. 11.17 Radiation pattern in the E-plane for the total electric field E, = EiP + Er of a
rectangular patch antenna with a truncated dielectric layer
W = 1 5 m m . I = 1Omm. d = 1 . 5 8 m m , ~=
, 2 , 3 5 , f , = 8 3 8 G H z , a = 3.51,
---- Theory
-Experiment
facilitates radiation owing to better matching of the internal field of the micro-
strip cavity to free space, and it provides a broader bandwidth 1321.
11.2.2.1 Resonant frequency: The cavity model is applied to the volume
bounded by the ground plane, the cone surface and the spherical magnetic wall
602 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 603
(Fig. 11.18). By analogy with the circular disc, we can define an effective
spherical radius Re given by

where 8, and a are, respectively, the angle and the base radius of the conducting
cone, and E, is the relative dielectric constant. The fields within the cavity
corresponding to radial TEw-modes may be derived from a scalar electric
potential $(r, 0, 4) which must satisfy the wave equation

where k = wfi.

Table 11.5 Theoretical and experimental resonance frequencies of the

TE,, -mode for several values of 0, and E,
&, 1 1 2.1 2.1
80 74O 85' 72' 75'
J (GHz)
(theoretical)

For 0, near 212, applying boundary conditions on the conducting walls, $ can
be written as

w , 0, 4 )
$z
-2 Jm+ 112(kr) cos m 4

where J,+ll,
is the Bessel function of the first kind and

Then the resonance frequency can be obtained from the boundary condition on
the magnetic wall:

The radiating mode of interest is the TE,,-mode corresponding to

a, x 0.618. Solving eqn. 11.21 numerically, we find
 Design and technology of low-cost printed antennas 605
604 Design and technology of low-cost printed antennas
where c is the velocity of light in free space. In Table 11.5 the theoretical and
experimental resonance frequencies of the TE,,-mode are presented for several
values of 0, and E,; the radius of the cone base is fixed at 18.5mm, while the
substrate thickness is h = 6 mm.

11.2.2.2 Quality factor: The resonant frequencies, Q-factor and correspond-

ing bandwidth ( B W ) of conical-patch and circular-disc antennas with the same
parameters a, and E, are given in Table 11.6. It should be noted that in an
air-dielectric (e, = I), the same conical patch leads to: f, = 4-86GHz, Q = 5
and a BW 20%.

Table 11.6 Resonant frequencies and bandwidths of conical-patch and

circular- disc antennas
Antenna
Disc
Conical
structure

Fig. 11.19 Input-impedance locus of matched conical patch antenna Fig. 11.20 Radiation pattern of a conical patch with two excitations

11.2.2.3 Irnpedancernatching: As the substrate is relatively thick (h z O.lR,),

the effect of the feed probe introduces an effective reactance and causes a provides series-reactance compensation of the feed probe. Furthermore, by
clockwise rotation on the Smith chart of the input impedance seen at the feed shifting the probe away to a, = 10mm, we obtain the curve of Fig. 11.19 and
port. An open coaxial transmission line situated inside the conducting cone a bandwidth (VSWR < 2) of 12%.
606 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 607
11.2.2.4 Radiationpatterns: They are found to be relatively stable for linear radiating slot
I .electric field lines,
polarisation over a wide-frequency range. Dissymetries in the E-plane and
cross-polarised components in the H-plane, due to the generation of higher
modes, can be suppressed as usual by using two radially opposite feed probes groun
(Fig. 11.20). The necessary 0-180' phase difference can be obtained with a plane
rat-race (3 dB coupler).

11.2.2.5 Circularly polarised conical-patch antennas: It can be observed in icrostrip conductor

Fig. 11.20, contrary to the case of the classical disc antenna, that the patterns
are identical in both E- and H-planes. Thus, by feeding the conical patch as
shown in Fig. 11.21, a good broadband, circularly polarised antenna can be
a
\ dielectr~csubstrate
realised. The axial ratio versus elevation angle 0 is presented.
Fig. 11.22 (a) and (b): Printed slot fed by a microstrip line at first resonance and second
It can be concluded that flat conical-patch antennas are more advantageous resonance
in some respects than traditional microstrip antennas. Infact, they present a
broader bandwidth, and their structure lends itself to a simple way of impedance
matching. 'Moreover, they are well adapted to circular polarisation.

0; I0 o: a.
9 degrees
410 o: 6b rb
Fig. 11.21 Axial ratio of a microstrip conical patch measured at 3.5GHz versus elevation
angle 6

11 J.3 Linear and annular slots

Printed slots have received somewhat less attention than patch antennas in array Fig. 112 3 VSWR versus OIL of M S A at first resonance
design, since care must be taken to suppress undesired modes such as the - L = 13cm F = 1.15GHz
parallel-plate mode excited between the ground planes in a typical stripline-fed XXXX L = 15.5cm F = O.94GHz
slot 133-351. However, the feeding and matching techniques remain very simple.
000 L = 9cm F = 1.53GHz
11.2.3.1 Linear slot [36, 371: A microstrip line offers the possibility of +=4.4 w=5.0mm h=ll.6rnm
608 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 609
feeding the slot antenna at the first or second resonance without excessively Table 11.7 Bandwidth at resonance
disturbing the field distribution in the slot (Fig. 11.22). The strip conductor is BW = 14% for L/2W E 16
I st resonance
connected through the dielectric substrate to the edge of the slot. B W = 20% for 10 < L/2W < 19
2nd resonance

Fig. 11.24 Radiating slot length normalised to A, versus frequency measured at the first and
second resonance
W = 5mm. E, = 4.4, h = 1.6mm
X X X first resonance
000 second resonance

Fig. 11.25 Influence of W on the first and second resonant length of a MSA
W = 5 mm, 6, = 4.4, h = 1.6 mm
X X X first resonance
000 second resonance

A good match to the 50R microstrip line can be obtained by choosing a

centre-fed slot or an offset-fed slot for second-resonance or first-resonance Fig. 11.26 Input impedance of MSA fed by a microstrip line versus L/1,at different frequen-
operations, respectively. Fig. 11.23 gives the VSWR of the microstrip slot cies [40]
antenna (MSA) at first-resonance operation as a function of D/L, where L is the W = 0.6 mrn, E, = 2.1 7, h = 0.8mrn
610 Design and technology of low-cost printed antennas
length normalised to I, of the slot of width W 1,. D is the distance between
the centre of the strip conductor of width W, and the small side of the slot, and
I, is the free-space wavelength. From the Figure it can be seen that the slot is
matched to the feed line for D/L % 0.25.
Fig. 11.24 shows the length normalised to 1, of a radiating slot of 0.5cm
width, measured at the first and second resonance, versus frequency. The
influence of the width W o n the first- and second-resonant length is shown in
Fig. 11.25. At the first resonance L = 0.491,, where I, is the free-space
wavelength, while at the second resonance L increases with L/2 W. These varia-
tions are identical to those obtained for a cylindrical dipole [38, 391. For the
previous slot antennas, an optimum value of bandwidth (VSWR < 2) has been
obtained (Table 11.71.
ground plane
coplanar waveguide
Fig. 11.27 Printed slot fed by co-planar waveguide
The large bandwidth obtained at the second resonance is due to the weak
variation of the input impedance of the MSA, as shown in Fig. 11.26 [40].
Despite various attempts, an exact theoretical study of the input impedance of
the MSA with limited ground plane does not really exist [34, 411.
An expression for the complex admittance at the first resonance of a radiating
slot in the ground plane of a microstripline was found by Das [42] from the
complex radiated power and discontinuity in the modal voltage. The concept of
a complementary dipole [43] in an uniform medium of effective permittivity
8, = 2&,/(1 + E,) has been used to calculate the radiated power.
Experimental studies have been carried out over the frequency range 0.8-
2 GHz using a 5 0 n coplanar waveguide [37] (Fig. 11.27); a bandwidth
(VSWR < 2) of about 30% has been obtained. Nesic [44,45] used coplanar-fed
slots in primary radiators in a phase-scanned antenna.
Design and technology of low-cost printed antennas 61 1
11.2.3.2 Annular slots (Fig.11.28)
( a ) Model of an annular slot (without reflector plane) [46]: Analysis of radiat-
ing slots has been developed using a lossy transmission-line model (Fig. 11.29);
+
it requires the computation of a propagation constant a j p and a characteris-
tic impedance Z,. and Z, are obtained using Cohn's method [47], and a is the
solution of the numerical equation P,(a) = P,(a) where P, (a) is the power
delivered to the lossy loop and P,(a) is power radiated from the annular slot.
annular slot on
rnetalllc Dlane
reflector plane
(distance = d ' )
Fig. 11.28 Printed annular slot above a reflector plane
Fig. 11 2 9 Transmission-line model of the annular slot
(ai) Guided wavelength and impedance [ 4 8 ]
Theory: To date, no result of Cohn's method for calculating the slot parameters
on a low-permittivity substrate has appeared in the literature. The use of Cohn's
formula, for magnetic walls parallel to the slot and a relative permittivity of
217, gives the ratio of guided wavelength, ,If2,,
to wavelength in free space, 1, and
 612 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 613
the characteristic impedance defined in Reference 47. The results of computa- [52]. Comparison of that method with theoretical results obtained by the
tion are plotted in Fig. 11.30 in the manner they were presented earlier. A simple spectral-domain technique in a recent paper [49] gives good agreement as shown
formula for &/A has been found like that in Reference 51: in Table 11.8.

Fig. 11.31 Resonant-ring transmission-line model

a b
Fig. 11.30 (a) & / A versus d l l and w l d
(b) Characteristic impedance versus d l l and w / A
e, = 2.17

Table 11.8 Some theoretical results of A,/A (from Reference 49) and
Cohn's method
Frequency Cohn's Spectral-domain
@Hz) method technique [4]
2 0.8890 0.8885
2.5 0.8840 0.8834
30 0.8796 0.879
3.5 0.8756 0.875

Fig. 11.32 Theoretical and experimental results for guided wavelength for slots of different
width w
The following set of parameters is considered: d e x p e r i m e n t (6, = 2.17, substrate thickness = 0.78mm)
---- theory
E, = 2.17
0.01 < wld < 2.0 Experiment: Experiments were carried out to verify Cohn's method. The meth-
0.01 < d/A < 0.1 od of the resonant slot ring was chosen, because it has given very good results
on substrates of higher permittivity ( E , = 9.6) [SO]. Three rings were etched. The
The previous formula fits the theoretical results with better than 2% accuracy. outer radius of the rings was equal to 38 mm. The measured widths were 0.3,
The upper limit of d / l corresponds to appreciable excitation of surface waves 0.99 and 3.3 mm. The thickness of the substrate was 0.78 mm and the relative
 614 Design and technology of low-cost printed antennas
dielectric constant was 2.17. Measurements were performed on an automatic
network analyser HP 8510. The output and input lines were cut until the signal
level weakened at about - 40 dB below the reference set to 10 dBm (Fig. 11.3 I).
dielectric
-thickness
- er=2.17
Fig. 11.33 a solution of P,(a) = P , (or) for X-band annular slot
The difference between the theoretical and experimental curves of Fig. 11.32
is due to ohmic, dielectric and radiation losses. In the case of a slot on a dielectric
substrate of low permittivity, the radiation loss is important: it increases when
the radius of curvature decreases and remains significant for a straight slot. It
can be noted that, for quasi-straight slots, the experimental ratio AJA always
appears lower than the theoretical one, as in Reference 49.
(aii) Attenuation coe8cient and antenna design
The method which is used to find cr requires numerical computation.
Design and technology of low-cost printed antennas 615
P,is the power delivered to the lossy loop (analytically known)
0 P, is the radiated power.
then a is the solution to the equation
It will be noticed that P,increases with a while P,decreases; thus the solution
is unique. An example is given in Fig. 11.33; the loop was designed at X-band
from the dimensions given in Fig. 11.33. The variation of a near the resonant
frequency is plotted in Fig. 11.34, and it appears that a has a linear variation.
Fig. 11.34 a variation versus frequency (near the resonant frequency)
r = 4,14rnrn, W = 154pm
The design of antennas matched to 50Q has been considered using the
following model shown in Fig. 11.35; a quarter-wavelength line (length I,) is
used to feed the loop and a second transformer (length I,) is used as a matching
section. The equivalent transformer from the microstrip feed line and slot are
taken into account. Impedance curves are drawn in Fig. 11.36 for frequencies
varying from 9 to 11 GHz. The theoretical results agree well with the experi-
mental values.
 676 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 67 7
To facilitate the design of antennas, analytical formulas for a have been
derived in the vicinity of the resonant frequency:
A
c! = - + -Bd
I.
c! = attenuation, N/m
1 = free-space wavelength
d = thickness of dielectric
where

A x

Fig. 11.35 Annular slot and its equivalent circuit

=11 z
REF 1.0units
4 200.0m units1
V 169.38 n-40.523 n (Ay+
- 18005.3 - (
570.8 log,, 200 -:.)"(:)" . .
The accuracy of the above formula is better than 2%; it can be used for the
following parameters:

(b) Annular slots with rejector planes (Fig.11.28)

The analysis is very similar to the previous one; a modification of Cohn's
method makes it possible to obtain the two main parameters, guided wavelength
and characteristic impedance of the slot line in front of the reflector plane Figs.
11 . 3 7 and
~ b show the variation of and 2, for the two distances d' (d'jd = 3
start 9 . 0 0 0 0 0 0 0 0 0 ~ ~ ~
stop 11.000 and d'ld = 11) as a function of dl1. The dielectric constant is 2.17 and the slot
a :theory b : experiment is very thin ( Wjd = 0.01). It appears that ,Ig and 2,do not differ from the values
for slot lines without a reflector when it is sufficiently displaced. Detailed
Fig. 11.36 Impedance of the printed annular slot analysis of the mathematical expressions [46] shows that it is possible to find the
r = 4.1 mm, W = 140pm, W, = 0.76mm
limit distance d',im,which leads to identical values of Ag and 2, with or without
678 Design and technology of low-cost printed antennas 1 Design and technology of low-cost printed antennas 679
a reflector plane: (a) Radiation occurs only in the upper half space.
(b) No energy couples to the space between the two metallic planes (slot plane
1 and ground plane).

The following assumptions are then necessary to obtain the value of a.

I
Fig. 11.38 u variation versus frequency for slot above reflector plane
r = 4.41 mm, w = 154prn

Table 11.9 Dimensions of the two annular slots of Figs. 11.40a and b
Antenna r W 1, I2 13 w K w

Fig. 11.37 (a) Variation of &/,I and Z, for slot above reflector plane ( d / d = 3)
(b) Variation of 4 1 1 and 2, for slot above reflector plane ( d l d = I I )
Then, for a EMF given generator, the total radiated power is half the value
obtained without a reflector plane; a is also reduced by a factor of 2, and the
impedance is twice that without a reflector plane. Fig. 1 1.38 shows the frequency
variation of a with or without the reflector plane.
To avoid guided waves between the two metallic planes, four cylindrical
metallic posts are positioned as shown in Fig. 11.39a; the post spacing equals
one half-wavelength, and their height is 8.5mm. In Fig. 11.396 impedance
curves have been plotted either for theoretical results (assuming radiation in the
upper half-space only) or for experimental results with and without posts.
 Design and technology of low-cost printed antennas 627
Theory and experiment agrees well when the posts are taken into account.
Input impedances have been computed and measured for two different slots
and feed lines with the same ground-plane distance d' = 8.5 mm (Fig. 11.40);
the reference plane is n, (Fig. 11.35) and the dimensions are given in Table 11.9.

START .9.000000000GHz
STOP 11 00000000 GHz

experiment

START ~9.000000000GHz
STOP 11.00000000 GHz

T ~ Y experiment b
Fig. 11 3 9 (a) Location of the metallic posts near the annular slot
Fig. 11.40 (a) lnput impedance of annular slot no. 1: 6 = 8.5mm
(b) lnput impedance of the annular slot (reference plane n,) ( 6 ) lnput impedance of annular slot no. 2: d = 8.5mm
---- theory
- experience with posts
Conclusion: It has been proved that the transmission-line model of annular slots
--- experience without posts
yields good results in terms of impedance variation versus frequency. However,
some problems occur when a ground plane is added to produce a directional
 622 Design and technology of low-cost printed antennas
antenna; further work is necessary to understand the parasitic effects of guided
waves between the ground plane and the metallic plane of the antenna.
11.3 Design of planar printed arrays
Design and technology of low-cost printed antennas 623
and a,,, is the excitation coefficient (relative to a reference element) of the (m,
n) source located at (x,, y,,) (Fig. 11.41). When the radiating elements are
regularly spaced along two orthogonal directions (Fig. 11.42), simple ex-

Photo-etching techniques offer great flexibility in designing one- or two-dimen-

sional arrays of microstrip antennas. Many parameters could be considered; e.g.
position of elements, spacings, amplitude and phase distribution, feeding lines
(on the same side or behind), connection with active devices etc. In fact, each
structure is developed in response to a given class of problem. Only the usual
design parameters and the realisation of pencil beam antennas are developed
here; shaped-beam pattern design will be considered, together with the synthesis
methods, in Section 11.4.

X
/
Fig. 11.42 A m y of linear sub-arrays

( \ r n , n ) ~element
~
Fig. 11.41 Co-ordinate system and location of the (m, n ) th element

11.3.1 Design parameters

With the assumption of identical elements, the array gain is decomposed into the
product of array factor, isolated element gain pattern, and a factor displaying
the impedance effects of mutual coupling. When mutual coupling is negligible,
the overall radiation pattern becomes [53]
40, $1 = f (0, $1 g(0, $1 (1 1.25)
where
g(0, $) = radiation pattern of one element Fig. 11.43 Equivalent orthogonal linear arrays of a planar array

f(0, $) = array factor pressions can be derived for the two main planes, 4 = '
0 and $ = 4 2 . For any
$-direction, the planar array can be analysed using the pattern-multiplication
method if the various sub-arrays are identical (same number of elements and
 624 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 625

same distribution); then the planar-array factor remains the product of two Losses have been plotted (Fig. 11.44a, b, c) on a dB scale (loss (dB) = P,,,,
linear-array factors (Fig. 1 1.43). - P,,,) for a number N varying from 2 to 40. Three main parameters have been
As planar arrays and linear arrays have much in common, some basic results considered: Zo, the characteristic impedance with typical values of 5 0 0 and
for linear series microstrip antenna arrays are given first: (i) directivity estima- 100Q dielectric loss tangent, typically lo--'; and R,equal to 300 or 50R. In each
tion taking into account the dielectric substrate, (ii) line losses depending on the case the usual dielectric-constant value for printed circuits of Teflon fibre
element impedance value, and (iii) beam width and limitation on gain. substrate is equal to 2.2, and the different curves have been obtained for a
frequency of 10 GHz. When N increases, the efficiency is reduced, as might be
R, =300Cl R,=300fl =
?,I 5011
expected; the highest value of 3dB is obtained only for a large number of
z0. 5 o n zO=loon z 0 = loon elements. However, the slope of the curves depends on the antenna resistance
linear serles array llnear serles array h e a r s e r ~ e sarray and characteristic impedance; it will be noticed that small R, (typically 50R) and
3.0r 3.0r large Zo (here 100R) lead to high losses, even with a moderate number of
elements.
In order to reduce line losses, the designer should choose a high input
impedance R, and a low characteristic impedance 2,. However, the last con-
straint often conflicts with the necessarily small width of the input feed line
compared with the width of the antenna itself.

11.3.1.2 Directivity of unijbrm linear arrays: The simplest type of linear array
is the uniform one. Analytical expressions for directivity have been proposed for
-. - . - - . .- . - - .. typical elements such as ideal isotropic sources, co-linear short dipoles and
T-delta-0001 T-delta - 0 6 1 T-delta :O 001 parallel short dipoles [53-561. Rectangular printed antennas radiate from fring-
thickness (MMS):0.78 thickness(MMS)=078 thickness (MMS) = 0.78
freq (GHz) = 10 freq ( G H z ) :I0 freq (GHz) = 10 ing fields around the edges; the fields along the two radiating edges are ap-
proximately uniform. Thus each antenna can be considered as an array of two
Fig. 11 .Q4 Line losses of linear series array uniformly excited identical slots; its directivity is about 5-8 dB depending on the
( a ) R, = 3000,Z, = 500 dielectric substrate. Two linear-array structures are to be considered, namely the
( b ) R, = 3000,Z, = 1000 H-plane and E-plane (Fig. 11.45); the slots of each patch are defined by their
( c ) R, = 50R.Zo= 1000
length Wand their spacing do; d equals the distance between each element. The
array radiates in the half-space defined by x > 0. Elements are located along Oz
to get simple expressions for slot patterns and simple integrals of directivity.
11.3.1.1 Line losses, eflciency of linear series array: Typical linear resonant
arrays of microstrip antennas can be considered as a transmission line loaded
( a ) Directivity expression
periodically by shunt resistances whose values depend on the elements themsel-
The element pattern of each patch g(0, 4 ) can be easily expressed in H- and
ves (rectangular, circular, triangular with proper feeding points), or on the
E-plane configuration as:
equivalent sub-array attached to the main line. For the sake of simplicity, let us
consider a transmission line with a characteristic impedance Z,, and a complex
+
propagation constant y = a ja. The periodic loads (radiating elements) are
H-plane: g(0, 4 ) = (Y )
sin 0 cos k - s ~ 0n sin 4
located at each guided wavelength in order to get a uniformly excited array. The
resonant structure is then represented by a cascade of lossy transmission-line
sections. The radiated power P, and input power P), are readily computed, and
efficiency can be deduced:
E-plane g(0, 4) =
( -2
(1 - sin20 sin24)li2cos k cos 0
1
and the directivity expression of an N-element array in the broadside direction
may be written
2
- -
2 (1 1.28)
D = .
where Ri,, = input resistance; R, = resistance of each radiating element and
= potential at node j.
 626 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 627
where where

8 N-l
N-m 3 cos a

and
a, = kd, a = 0.24774379
a =; mkd b = 0.01294148
a, = mkd +k4
a, = mkd - kd,
The a and b constants which occur in the H-plane-configuration expression
appear in the 0 integration of !2,

a, = 1; cos2 (Fsin 0 sin 4) 2 (0) f2 (9) sin 8 d0 dm

H - plane E -plane
The kernel of the Bessel function J, remains between 0 and 2.5 for most printed
Fig. 11.45 E-plane and H-plane linear arrays of equivalent slots of microstrip antenna antennas (E, 3 2); then J,(x) can be approximated by a simple polynomial
expression:
The coefficient 2 arises because the antenna radiates in the half-space, X > 0.

,+
With d n = sin 0 d0 d 4 and using the finite-series expression [55, 591 off 2(0):
The use of two points x, = 1 and x, = 2.402 leads to the previous values a and
1 2N-1
f2(0) = - 1
N2 ,,,=I
( N - m) cosm$ b. J,(x), and the approximate polynomial are plotted in Fig. 11.46. There is
good agreement within (x, x,).
the results are
( b ) Results
H-plane configuration:
Figs. 11.47 and 11.48 show the variations of D for linear arrays of printed
antennas in H- and E-plane configurations for two kinds of dielectric: PTFE
(E, = 2.2 and thickness = 1.6mm) and alumina (E, = 9.8 and
thickness = 0.625 mm). The dimensions of the antennas were computed using
E-plane configuration: the following expressions:
D = d - -- 2Ad0 and Ad, = length extension [6]
2 O - 2&f
(N - m)[2S + S, + S2- ( ~ S C+ SC, + SCJ
I
 628 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 629
The curves show that D is an increasing function of dl1 when dl1 is less than 1
and N is large. The optimum distance d for large directivity and no grating lobes
lies between 0.71, and 0.91,. For usual substrates ( E , = 2 to 2.5) and with

00 1
0.20.4
' ,0 6. 0 8. 1.0
, 1.2
. 1.4
. 00% d.6 0:8 1'0 1' 2 114
d~rectivity Dllambda d~rectivity Dllarnbda
linear array H-plane linear array E-plane
printed antennas die1 const ~ 2 . 2 printed antennas die1 const =2 2

Fig. 11.47 Directivity of an N-element linear array of printed antennas as a function of

uniform element spacing
Dielectric constant = 2.2.

-0.3
approximate polynomial of the 0-order bessel 0<X<2.401

Fig. 11.46 Approximation of the zero-order Bessel function

directivity Ollarnbda directivity D l lambda
linear array H-plane linear array E-plane
one-1,-long straight feeding lines between neighbouring elements, the distance prlnted antennas die1 const.9.8 printed antennas die1 const.9 8
d is approximately 0.751,. For small values of N, the E-plane arrays exhibit a
Fig. 11.48 Directivity of an N-element linear array of printed antennas as a function of
larger directivity than H-plane ones. On the other hand, the H-plane arrays uniform element spacing
present similar directivity variations for E, = 2.2 and 9.8. Dielectric constant = 9.8.
 630 Design and technology of low-cost printed antennas
11.3.1.3 Non-uniform linear arrays: The reduction of the sidelobe level re-
quires tapering of the amplitude distribution; some care must be taken in the
element spacing; the optimum distance d z 0.75& for printed antennas on a
PTFE substrate does not always fulfil the spacing requirement of the Chebyshev
design for small number N and low sidelobes. Fig. 11.49 gives (d/l0),, versus
sidelobe level for various N; and example of an array factor with N = 6 and
0.5
10 20 30 LO 50
-Rde
Fig. 11.49 ( d l l ) , versus sidelobe level for Chebvchev taper
Fig. 11.50 Chebpshevpattern of a six-isotropic-source linear array
(Spactng = 0.751,)
(dl&) = 0.75, designed for - 40 dB sidelobe level, is shown in Fig. 11 S O . A
strong effect appears on the sidelobe level, which reaches - 23 dB at 0 = 90 +
instead of -40dB. On the other hand, directivity D decreases and the half-
Design and technology of low-cost printed antennas 631
power beamwidth (HPBW) increases. Directivity can be computed for isotropic
k~ementsusing the following formula [%I:
where A, is the normalised amplitude of the n t h element. However, the excita-
tion efficiency q, = DID, (where Do is the directivity of uniform array of
isotropic elements) is available either for Chebyshev, Taylor or Villeneuve
distribution [53, 54, 57, 58, 601; assuming that q, of printed antennas remains
very similar to q, of isotropic sources, a quick estimate of D can be obtained
from Do given previously.
11.3.2 Cavity-model analysis of mutual coupling
Since the experimental work of Jedlicka and Carver [61, 621, many theoretical
efforts have been made to calculate the mutual interaction between microstrip
antennas [25, 63-68]. Mutual-coupling effects are caused by radiation through
free space and by surface waves which propagate along a dielectric substrate.
The theoretical method proposed by Penard and Daniel [65, 251 is restricted to
the first effect; this is acceptable as long as the dielectric constant is low and the
thickness is small compared to the wavelength [67]. The method uses the cavity
model of the antenna, and it is assumed that mutual coupling does not disturb
the field distribution in the cavity. In the cavity method, by using the equivalence
principle, one can relate the internal field in the cavity to the magnetic-current
loop radiating in the upper half-space. Then the mutual-coupling coefficients
can be derived from the interaction between magnetic current loops.
The mutual impedance 2, between two elements is deduced from the reaction
theorem
where C, = contour of antenna j; M, = magnetic line source; Hi = magnetic
field set up by antenna i on antenna j; and 4 = current feeds of the two
patches.
Only the fundamental mode is usually considered in evaluating integral 11.32 in
order to reduce the computation time. The coupling coefficient Sv can easily be
calculated by applying the impedance-matrix equations of network theory.
Provided the internal electric field is known, printed antennas of any shape can
be considered.
11.3.2.1 Rectangular patch antennas: In the H-plane configuration (Fig.
11.51), the mutual impedance between identical OMA (open microstrip anten-
632 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 633
nas) is given by [65] where C is the width of the excitation and

The expression takes all the equivalent slots around the patch into account. The
contribution of the two OX-edge radiating slots appears in the integral R,.

t
....
...:.....::..
...... :.: ..............................
....
.............
. . . . . . ... .. ... .. ... .. ... .. ..... . . . ............ .. .............. .
I.....
'.X

ground plane/
Fig. 11.51 Geometry of two rectangular microstrip antennas in H-plane configuration

freq ~ 2 2 9 8GHz
Er ~2.55 Fig. 11.53 Contribution of OX edges and OY edges in mutual-impedance calculation in
d. cm E-plane configuration; mode(0, I )
increment -0.43cm

The effect of the OY slots appears in R, and R, and cannot be ignored, as shown
I
in Fig. 11S2.
0.6v1
-,4.08
I
yI real
ax~s.crn1
-3 -2 -1 0 1 24' 3

Fig. 11.52 Contribution of OX edges and OY edges in mutual-impedance calculation in

H-plane configuration; mode(0, I )

where t is the substrate thickness with

It should be noted that R, is cancelled out when the dielectric slab is replaced
by a vacuum medium. Similar relations can be found for two patches in the
634 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 635
E-plane configurations [25]

In this case, the effect of the two OX-directed slots (R')) is stronger, but the
contribution of the longitudinal slots (R; and R ; ) is not negligible (Fig. 11.53).
Fig. 11.54 shows calculated values and experimental data of the coupling
between two identical patches for various spacing in E-plane and H-plane.
When the distance d becomes smaller than 0.08&, the discrepancy between
theoretical and experimental results in the H-plane increases. In this case, the
coupling disturbs the internal field distribution, whieh clearly appears on the
measured value of the input impedance Zll(Fig. 11.55).
Penard [25] has studied the mutual coupling between the hybrid-microstrip
antennas (HMA). In this element, the contribution of the apertures situated
near the electric walls is very small and can generally be ignored. In Fig. 11.56,
the coupling coefficient of HMA and OMA are shown for comparison. In the
H-plane, S12is smaller for HMA than for OMA, except when the dielectric is
replaced by a vacuum medium (Fig. 11S7).
Fig. 11.55 Experimental results of Z,, as a function of the distance dl& between the two
patches; H-plane; mode(0, I )
planE . o o o
plan H . x x x

\ m ) o = 4 c m

\;-- *--- -0-- -0 Cp.20 1

log ISlZ
mode ( 0.1)
frequence = 2.273 GHz

-
r L 2.55 \
5.0001
1.0.52 cm
-4 0 I I I I
0 0.1 0.2 0.3 0.4
dlAo

Fig. 115 4 Mutual-coupling coefficient as a function of the distance dl10 of the edges in the
E-plane and H-plane; mode(0, I )

Another interesting result, in view of array design, is that the mutual coupling
coefficient slowly increases with the substrate thickness (Fig. 11.58). The effect
of the dielectric-constant variations on mutual coupling of short-circuited and
open microstrip antennas is presented in Fig. 11.59 and 11.60, respectively.

11.3.2.2 Circular patch antennas: In this case, the circular geometry of the
Fig. 11.56 Comparison of coupling coefficient in H-plane, between HMA (a = 6cm.
elements allows the separation of the co-ordinate variables, disc centre spacing b = 2cm) and OMA (a = 6cm. b = 4cm) structures
RV,disc angular orientation 4Vand angular feed positions 0 , 0, (Fig. 11.61). E, = 2.17, fa = 2.45 GHz, r = 0.157cm. tanb = 10-4
 636 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 637

Fig. 11.59 Theoretical coupling coefficient between two OMA versus E, for different values
Fig. 11.57 Comparison of coupling coefficient in'H-plane, between H M A and OMA with
of dl10
air-dielectric
8, = 1 tl& = 0.017
f, = 1.773 GHz, t = 0.1 50cm

Fig. 11.60 Theoreticalcoupling coefficient between two HMA versus E for different values of
dl&
6, = 1 tl1, = 0.0093 A experiment
Fig. 11.58 Theoretical coupling coefficient as a function of the dielectric thickness t l k E, = 2.1 t / l o = 0.0064 0 theory
mode(0, I )
 638 Design a n d technology of low-cost printed antennas
Mahdjoubi [66] gives a semi-analytical formula for 2, which permits a great
reduction in computation charges:
m m
Z, = 1 Zy 1 Z;[Yemn(R,,) cos m(4ii - O,)cosn(4,
m=O n=O
- 0,)
+ Y,""(R,)~inm(4~- 0,) sinn(bV - O,)] (1 1.35)
m and n are the angular mode indices for antennas i and j, respectively.
Fig. 11.61 Geometry of e/ectromagnetically coupled microstrip disc antennas
= angular width of the feeding probe; Q = radial feed position; a = effective
radius of the disc antenna; t = substrate thickness; k = propagation constant
in the dielectric, J, and Y, = Bessel functions of first and second kind of
order p.
YY(R,,) and Y,""(R,) are the space-wave mutual admittances between mode m
of annular slot i and mode n of annular slot j for the E- and H-plane coupling
configurations, respectively. For two identical circular discs with only the
fundamental modes excited in both antennas (m = n = I), the mutual im-
Design a n d technology of low-cost printed antennas 639
pedance Z I 2can be written
where
The Y, factor corresponds to the E-plane (4 = 0, 8, = 0, = 0) while the Y,
factor corresponds to the H-plane ( 4 = 0, Or = 0, = 90') (see Fig. 11.62). The
E plane
H plane
Fig. 11.62 Magnetic-current distribution along the contour of circularpatch for two principal
coupling planes
mutual coupling coefficient S12 calculated at f = 1.4127GHz, for comparison
with experimental data (f = 1.44GHz) from Jedlicka et al., are given in Fig.
11.63. Agreement is quite good so long as the frequency shift ( x 2%) due to the
cavity model is taken into account. Whatever the mode, the exact dependence
of the coupling phenomena on 4 is a simple trigonornetrical function. Eqn.
11.36 and Fig. 11.64 describe the typical form of the coupling coefficient versus
4, normalised to E-plane coupling. The results are compared with those of
Bailey and Parks (Reference 5, p. 157).
The curves of coupling coefficient against the angular feed position 02(0, = 0)
are given in Fig. 1 1.65 for two edge spacings d = 0. I%, and d = Ao.The E-plane
 640 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 641

and H-plane of antenna 1 correspond, respectively, to I$ = O0 and I$ = 90". It

should be noted that the coupling coefficient varies rapidly near 0, = 90'
showing how critical is the precision of the feed angle at this location.

d'
. E plane

} theory
::] exper~rnent .>, H plane
reference 61

O Q O measured microstrip disk coupling

Fig. 11.63 Coupling coefficient versus distance d between edges
6, = 2.5, t = 1.575 mm, patch radius = 3.85 cm, probe radial distance = 1.1 cm
------ calculated circular waveguide fed
aperture coupling (Bailey and Parks)
Frequency: our theory
Measurement = 1.44 GHz
Theory = 1.4127 Fig. 11.64 Coupling coefficient versus orientation 4, normalised to the E-plane coupling
( f = 5.5GHz)
(a) Input impedance Z,
The input impedance Z, of the elements can be readily calculated by applying
the impedance matrix equations of network theory. Fig. 11.66 illustrates Z,
variations of the previous two-element array against the edge distance d. It can
be observed that, for d > 0.3& ( R > 0.68&), the coupling effect on Z , is
negligible. In studying 2, as a function of 4 and frequency, we have chosen a
very short distance ( d = 0.1&) in order to observe the coupling influence. Fig.
11.67 shows that in the E-plane ( 4 = 0), Z, x Z,, = 61.7 /-Y ohms, where
Z , , is the proper impedance of the disc antenna. However, in the H-plane
( 4 = 90') where the coupling is stronger, Z, is very different.

(b) Circular sub-array of three identical elements (Fig.11.68) [68]

In Fig. 11.69 one observes that the coupling effect on the input impedances Z,, ,
Z,, and Z,, becomes negligible for d > 0.31,. The dependence of the input
impedance on feed angular position 0 is presented in Fig. 11.70. Although two
of the three impedances can be equal for certain values of 0, nowhere they are
identical simultaneously.
82 I
Fig. 11.65 E- and H-plane coupling coefficients for m = I , n = I , versus angle OZ(B, = 0)
642 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 643
( c ) ConcIusion
The cavity method and reaction theorem represent very suitable ways to cal-
culate and predict free-space mutual coupling between array elements of simple
shape. According to these results, it seems that the H-plane coupling is stronger
for the disc than for the patch (see Figs. 11.54 and 11.63) and identical in the
E-plane ( d < 0.31,).Mutual coupling effects become important when the edge-
spacing distance d is lower than 0.31, for an array of few elements.

I
I

1 z,, 1 =61.7/

F i g . 11.67 2, of a two-identical-element array against @ ford = O . I I , and f = 1.3918GHz

11.3.3 Linear series array of corner-fed square patches [69]

The feed network of microstrip antenna arrays exhibits loses which lead to a
limit on the expected gain, and consequently a limited number of elements. The
Schelkunoff unit circle, Dolph-Chebyshev and Villeneuve methods are well
suited for the design of fairly small arrays; Taylor's method, sampling line
source and Fourier's series expansion are better for large arrays. However, the
previous synthesis methods consider only the array factor, while the overall
Fig. 11.66 Input impedance 2, of each array element as a function of d diagram obviously needs the element patterns to be taken into account; for e.g.
f = 1.391 8 G H z- E Plane ---- H Plane the cos 0 variation in the H-plane of most printed antennas has a strong effect
for arrays with few elements; the - 40dB Chebyshev design example previously
644 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 645

mentioned shows a high sidelobe level (- 23 dB) in Fig. 11.50. The sidelobe
disappears completely in the H-plane element pattern (Fig. 11.71), and the
initial specification is obtained. Two synthesis methods, taking the elementary
radiation pattern into account, are developed in Section 11.4.

Fig. 11.68 Geometry of a planar circular array of three microstrip disc antennas

Only one simple structure using square-shaped microstrip antennas is con-

sidered here. The corner-fed square patches are easily excited with a single
microstrip line (Fig. 11.72); a tapered distribution is readily obtained using
quarter-wavelength transformers along the line. In order to get a broadside
pattern, one wavelength spacing is necessary; half-wavelength spacing is also
possible, with alternate elements to keep the equi-phase condition.
The corner-fed square patches have been chosen because they provide a high
input impedance well suited for series array. It is also very easy to feed each
element on the corner. Two new aspects are considered here: reduction of
cross-polarisation using altenate location of the elements on the feeding line,
and reduction of sidelobe level using simple tapering for a linear series array.
 646 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 647
11.3.3.1 Radiation of corner-fed square patches

(a) Theory: When the patch is excited at one corner (Fig. 11.73a), the cavity
model [4, 251 shows that the main part of the internal field is the sum of two
degenerate modes with equal amplitudes, i.e. modes (0, 1) and (1, 0). If the
higher modes are ignored, the Ex and E, fields along the edges exhibit the

input I
I
A, 1
Fig. 11.72 Corner-fed square-patches array

Fig geometry of a
the patch

Fig. (O.l)and (1.0)

modes contributions

Fig. (1.1) mode

contribution

point

fK,(x)
-2 In1
---- z,"2 . . . . . z,
M T ( x ) ~ o
Fig. 11.70 Input impedance versus frequency: d = O.7&

MO(Y)
geometry and magnetic currents of the corner
fed patch
Fig. 11.73 Geometry and magnetic currents of the corner-fedpatch

variations shown in Fig. 11.736. The far field is linearly polarised either in the
E-plane (I(/ = 0") or in the H-plane ($ = 45"). For instance in the H-plane,
sin 2C
EQ, = -jM, -
(2C)' - n2
where M, is proportional to the amplitude of the magnetic current ( x = 0,
Y = 0)
2C = k,a - J 2 sin 8
2
deg
Fig. 11.71 Chebpshev pattern of a six-microstrip antenna linear array (H-plane, spac- a = side length
ing = 0.752,)
 648 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 649

AS long as the cross-polarised field is needed, higher modes must be considered. an anti-phase relationship which cancels the different contributions. The dia-
The next mode (1, l ) adds a contribution with a magnetic line distribution as gram is shown in Fig. 11.77. It shows a large reduction of the cross-polar level,
shown in Fig. 11.73~.The E,, far-field component is given by which is close to -28dB instead of - 17dB as previously. Obviously, as the
length of this array is half that of the first one, the 3 dB beamwidth is larger. To
2C
E,, = - 2M, cos2C (1 1.38) attain the same directivity, 20 elements spaced 112 would be necessary.
(2C)2 - n'
where MI is proportional to the amplitude of the magnetic current ( x = 0,
y = 0). It will be noticed that M I is much smaller than M,. The previous
formula shows that the cross-polarised component is null for B = 0 and in-
creases with 0.

UNIFORM LlNE ARRAY A1

1-1
****it****

I
FIG. Uniform linear array A1(FO- 21.3GHz) fig. H plane pattern of the
untform array A1 (FO=Z1.3GHz)
r10 ELEMENTS - copolar

*SPACING - 1 GUIDED WAVELENGTH

!----
cross-polar

*SUBSTRATE PTFE El- 2.17

THICKNESS
tgs
- 0.001
0.38
+FEEDING LlNE lOOn
+COAXIAL OUTPUT (AND TWO QUARTER
WAVELENGTH TRANSFORMERS)

Fig. 11-74 Uniform line array A, (F, = 21.3GHz)

( b ) Examples o f linear arravs

When the' spacing along the feeding line equals one guided wavelength, the
different elements are uniformly excited (Fig. 11.74). Here the ten-element array
is printed on the usual PTFE substrate (dielectric constant 2.17 and thickness
0.38 mm). The quarter-wavelength transformer enables good matching to the
50 i2 coaxial output. As expected, the H-plane diagram exhibits the well-known
- 13 dB first sidelobe; the cross-polarisation component is very low at @ = 0
and then quickly increases with 0 to - 17dB (Fig. 11.75). This cross-polarisa-
tion component can be reduced in a simple and efficient way. Let us consider
each patch located alternatively on each side of the feeding line (Fig. 11.76). Fig. 11.75 H-plane pattern of the uniform array A , (F, = 21.3GHz)
Considering half-wavelength spacing, the different co-polar fields (Emfor H- --- co-polar
plane) add in phase, while the different cross-polar components ( E B Imaintain
) ---- cross-polar
650 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 657
1I .3.3.2 Tapered linear series array

(a) Theory: The previous arrays were uniformly excited; high sidelobes are
the consequences of this illumination. The idea was to produce a non-uniform
amplitude distribution while keeping the simplicity of the previous series feed-
ing. Let us consider a linear array with wavelength spacing (guided wavelength);

UNIFORM LINEAR ARRAY

WITH ALTERNATE ELEMENTS A2

I
FIG.
.
Uniform array A2 alternate alaments
I

Fig. 11.76 Uniform linear array with alternate elements A,

Fig. H plane pattern of uniform array A2
(FO=21.3GHz)
as the radiating elements are identical, impedance transformers are necessary to
obtain the given amplitude current. To do this, a two-step quarter-wave trans- -
----
copolar

former can be used in each cell (Fig. 11.78). The transformed admittance & in
the IT: plane is given by
cross polar
I
Fig. 11.77 H-plane pattern of the uniform array A, (F, = 21.3GHz)
where Y,,, is the admittance of node (i +
l), and Y,, and Y,, are the charac- - co-polar
teristic admittances of each quarter-wavelength transformer. If necessary, four ---- cross-polar
quarter-wave transformers can be inserted when the spacing equals one
wavelength.
When the input voltage leads to a unit current in the first element, the current
distribution is readily obtained with the following relations:
However, the various ratio nican be deduced step by step from the known values
of I,, I,,&, etc. The input impedance at element 1 is
Z"
z,, =
1 + n: + nini + n:n:n: + . . . . . (1 1.41)
652 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 653
( b ) Results
TAPERED LINEAR SERIE'S A R R A Y A ten-element array has been constructed (Fig. 11.79). The requirement was to
***a******
get a sidelobe level lower than - 20 dB. Only eight transformers were used (four
lA~
INPUT

---7- aI I,

- -
10 YAV 1
1, - n,Y,+V = n l
i, - n,n,, ...n,
Fig. 11.78 Tapered linear series array: current distribution

TAPERED LINEAR SERIE'S

A R R A Y A3

~~~ **********

I
fig. H plane pattern of tapered array
A , (Fo = 21.3 G H z )

I
FIG. Tapered array A3

* 10 ELEMENTS
*SPACING = ONE GUIDED WAVELENGTH
Fig. 11.80 H-plane pattern of tapered amy A, (F, = 21.3GHz)
* 4 TRANSFORMERS (ON EACH SIDE)
* FEEDING LINE lOOn on each side), because Y,, was chosen equal to the characteristic admittance of
* COAXIAL OUTPUT the half-wavelength following line. Taking the characteristic impedance of the
I I
main line as about loon, the various transformers exhibit impedances between
75 and 95 0 , which are easily realised with microstrip lines. The experimental
Fig. 11.79 Tapered linear series array A,
H-plane diagram is plotted in Fig. 11.80. It shows that the sidelobe level is lower
 654 Design and technology of low-cost printed antennas
than - 20 dB while the gain and input-impedance matching remain very similar
to those of the uniform array.
A combination of alternate elements and tapered distribution yields a pattern
with low sidelobe level and low cross-polarisation over the whole space. Fig.
11.81 shows a ten-element array with quarter-wave transformers. The diagrams
deg
Fig. 11.81 H-plane pattern of tapered array with alternate elements
Fig. 11.82 Typical cross-fed array of square patches (without matching network)
Design and technology of low-cost printed antennas 655
are plotted in Fig. 11.8 1, which clearly shows that no degradation occurs to the
cross-polarisation level within the frequency band.
11.3.4 Two-dimensional cross-fed arrays [TI, 721
The combination of identical linear sub-arrays leads to a planar array. The
feeding network of such structures is developed in Section 11.4. Another simple
two-dimensional arrav named the cross-fed structure can be considered as a
combination of non-identical linear sub-arrays.
Cross-fed printed aerials have already been described by Williams [70]. The
basic radiating elements were 45' dipoles inclined along the feeding lines. No
analysis was proposed; however, the structure appears very attractive owing to
its simple feed geometry which avoids having any transformer. Corner-fed
patches were chosen because they are easily fed along a straight microstrip line.
Moreover, the discontinuities introduced near the corner of each element are
symmetrical and identical for all of them. Thus co-polar and cross-polar :om-
ponents remain symmetrical around the broadside direction. Figure 11.82
shows a typical cross-fed array. Matching networks can be added for a coaxial
feed. Inter-element spacing equals one guided wavelength. Design equations for
a uniform array are given below.
11.3.4.1 Uniform illumination and impedance matching: The overall array is
constructed from parallel sub-arrays. The number of elements is reduced from
+
N, to N, - 2, considering ith and (i I) th sub-array, respectively. When N is
the number along the diagonal, we find:
N, = (N - 2)N/4 elements for the upper or lower group of subarrays
N, = N2/2 elements for the whole array
The input impedances are different for the half main-line section (R,) and the
upper or lower group of sub-arrays (R,) (Fig. 1133). The impedance matching
needs one or two quarter-wave-section transformers to get suitable characteris-
tic impedances (Z;, Z,, Z',, 2,).The uniform illumination condition yields a
second relation (same voltage V for all the elements). Then the transformer
impedance ratios are equal on each side as follows:
where R, = resistance of a corner-fed antenna; R, = desired input impedance;
N = number of elements on the main line.
11.3.4.2 Radiationpatterns: The total array factor results from the combina-
tion of sub-array factors. The expression is
+ 1
2
P
cosicp,
sin(N - 2i)(py/2
sin qy/2 I (1 1.43)
 656 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 657
where 4, = kd, cos 4 sin 0; q, = kd, sin 4 sin 0; d, and d, are the inter-element
spacing along Ox and Oy; P = number of sub-arrays (above or below the central
line); N = number of elements along the central line and S,, = N 2 / 2 .

Fig. 11.83 Input matching transformers of the cross-fed array

It is interesting to consider this expression for d, = dy = d i n the two main

planes. In the H-plane, ST is the sum of the usual uniform array factors Si=
sin (Micp,/2)/sin(cp,/2). However, the nulls of each Si function have different
locations because the sub-arrays have a different number of elements. Then all
S, components are added in phase in the broadside direction while a compensa-
tion occurs from the oscillating functions outside 9 = 0. The following results
are given for usual substrates (6, x 2.2), and the spacing equals one guided
wavelength (A8 2, 00.5&).
Fig. 11.84 shows the three components So, S , , S, for the case N = 6 and
P = 2 (six diagonal elements and four sub-arrays). Summation then leads to a
diagram with a large reduction in the previous oscillations of each of the
subarrays. In the E-plane, cp, = 0 and ST is the array factor of an equivalent
array exhibiting a linear tapered excitation. Fig. 11.84 Sub-array contributionsSfof the globalarray factor S, (six diagonal elements and
Fig. 11.85a, b and c show the computed patterns in the three main planes, four subarrays)
658 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 659
pattern of crass-fed array which clearly show that the sidelobe level (SLL) is greatly reduced. However, the
PHI =go 4 = 45' plane yields the - 13 dB sidelobe of the uniformly illuminated square
element NB:6 structure. On the other hand, the half-power beamwidth &dB remains very
(diagonal)
equiampl~tude similar in the three planes ( 4 = 0, 45', 90'). &dB and SLL,, versus the
equ~phase number N of elements on the diagonal feed line is given in Table 11.10.

Table 11.10 Half-power beamwidth and sidelobe level of cross-fed array

20 40 60 80

pattern of cross - fed array

N = number of elements along the diagonal feed line

elernent NB=6

equiamplitude It will be noticed that only using the impedance matching condition does not
provide a uniform illumination; for instance, if different quarter-wave transfor-
mers are used on each arm of the previous structure (N = 6 elements on the
diagonal arm), the voltages on each element of the upper and lower sub-arrays
differ from the voltages of the central line. The computed patterns are plotted
in Fig. 113 6 when the voltage equals 1 on the main line and 0.8 on all the other
elements. It appears that the E-plane is quite transformed, sidelobes reaching
- I6 dB and 0, dB = 17.8'.
-60 -40 -20 0 20 40 60 80
11.3.4.3 Results: Various arrays have been built either with coaxial or
waveguide output. Each of them was printed on a PTFE substrate ( E , = 2.17,
pattern o f cross -fed array tg6 = lo-), thickness = 0.38 mm) or on polypropylene, as described in Chap-
PHI - 0
ter 5. An 18-element cross-fed array designed for 23.5 GHz is considered first.
element NB = 6 The same quarter-wave transformer was used on each arm to obtain a SWR
(diagonal) better than 1.5 at the coaxial output (Fig. 1137). The measured gain equals
equ~amplitude
equiphase 20dB, while a uniform aperture of the same area yields a 21 dB gain. E- and
H-plane diagrams are given in Fig. 11.88; the sidelobes reach - 18dB in the
E-plane and are lower than -20dB in the H-plane, and beam widths are 15"
(H-plane) and 18" (E-plane), as expected.
Another 50-element array, printed on polypropylene of the same thickness,
has been realised for a frequency near 20 GHz. The number of elements in the
upper and lower groups of sub arrays is 40, and the impedance to be matched
equals 8 a. The transformer then needs three quarter-wave sections as shown in
Fig. 11.89 to get equal voltage on each patch and good impedance matching.
The radiation patterns are given in Fig. 11.90. No sidelobes larger than - 25 dB
Fig. 11.85 Computedpatterns of the cross-fed array (six diagonal elements, uniform distribu- appear in the H-plane while the E-plane exhibits - 18dB sidelobes. In both
tion) cases the cross-polarisation level remains acceptable. The measured gain is
( a ) H-plane 6 = 90"
(6)45"-plane 4 = 45"
( c ) E-plane = 0'
 660 Design and technology of low-cost printed antennas
Design and technology of low-cost printed antennas 667
Pattern of cross-fed array
0-
PH 1.90
element NB=6
(diagonal)
-10- variable ampl
equiphase

-20-

-30-

-"PO -fi4!-i0 i a
i l \ o ~ idoo F(GHz)

Fig. 11.87 VSWR of a 78-element cross-fed array (six diagonal elements, F, = 23.5GHz)
pattern of cross-fed array

element NB-6

Pattern of cross-fed array

PH 1.0
element NB=6
(diagonal)
-10 variable am01
/ \ equiphase '

Fig. 11.86 Computed patterns of the crass-fed array (six-diagonal elements; non-uniform
distribution) b
Fig. 11.88 E-plane and H-plane measured pattern of the 78-element cross-fed array
 662 Design and technology of low-cost printed antennas
23.2 dB; the uniform aperture of the same area would yield 25.48 dB. Losses
reach approximately 2.3dB; the VSWR is lower than 1.8 between 193 and
20.4 GHz.
11.4 Synthesis methods for linear arrays 173, 74, 751
Design and technology of low-cost printed antennas 663
Let F(8) be the directivity pattern of a linear array; then
where f (8) = array factor; g(8) = directivity pattern of the source.
As usual printed antennas have different diagrams in the two main planes
(E-plane and H-plane), so it is better to synthesize F(8).

The usual analytical synthesis methods (Fourier, Chebyshev, Woodward-Laws-

on) are not always suitable when the directivity pattern is specified by a given
outline. Thus, the mean-squared error criterion is a global criterion that does
not permit the separation of the main-beam and the sidelobes contribution. plan E
Prescribing equi-level sidelobes does not always fit the Chebyshev requirement. I FzI9.55GHz

CROSS-FED ARRAY A4
**********
+4 4 +
++4 +4+
+ 4 4 + 4 4 4 +
+ 4 4 4 4 4 + 4 4 +
SO E U M N T S W L L Y EXCITED
SILPLE CORPORATE FEED +4++?44++
SIMPLE OUTPUT MACHINC
(A THIEE SECTION TRANSFORMER ON THE
VERTICAL B W H WAS MCESSARY)
LOW SIDE LOBES I N E-PLANE
H-PLANE
s F E W Y BM9 19.5-20.4 CHz

GAIN .
VSWR < 1.8
23 dB
FIG. Cross fed array with coaxial
output A4

Fig. 11.89 50-element cross-fed array printed on polypropylene 1 1 ~ ' ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ~ ' ' ' ~
-80 60 40 20 0 20 40 60 80
deg
b
Lastly, the Woodward-Lawson sampling method needs a great number of Fig. 11.90 E-plane and H-plane patterns of a 50-element cross-fed array printed on poly-
sources, and the choice of sample is sometimes critical. propylene
Numerical methods can take into account the envelope specification, the
directivity pattern of the source and the inter-element spacing. Two numerical
methods have been studied: 11.4.1 Relaxation methods

The relaxation method which enables real excitation coefficients 11.4.1.I Method: Let us consider a symmetrical linear array of 2N elements;
The simplex method which uses the Dantzig algorithm and yields symmetri- FJ8) is the desired directivity pattern and a = (a,):-, is the unknown excitation
cal or non-symmetrical pattern synthesis (real or complex coefficient). vector:
664 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 665
N
Fd(0) = 1 a, cos
j- l

d, = symmetrical-axis relative position of source + j ; A, = free-space

wavelength.
The following linear system is obtained using discrete values of 0:

C = [c,,],.~, cv =
(: )
cos 2a 2 sinei , a T = (a,, a,. . .a,);

The functionals to optimise are those like

We denote a' and a E RNsuch as J(af) = Min J(a) in the sense of the chosen
criterion. To realise this, a series of vectors d is built, such as J ( d + 1) < Fig. 11.91 Outline of desired directivity pattern (example of sector of pattern)

J ( h ) . The search directions are the co-ordinate axes, each of them being taken
periodically. For each component aj, we realise = 4+' such as at the
(k + 1) th iteration:

For a given quality criterion and choice of convenient functionals to optimise,

the relaxation method provides fast convergence. Some portions of the outline
pattern can eventually be preferred. Two examples illustrate the method: sector
pattern and directive broadside pattern.

11.4.1.2 Sector-pattern case: Let us define the desired directivity pattern

from an outline symmetrical on 0 (Fig. 11.91). RIimis the maximum ripple value
for 0 E [0, 0, - a] and R = 1 - D(8),. S L L , is the maximum sidelobe level
for 0, + a < 0 < a/2 and SLL = D(B),,. The pattern in the transition re-
gion is related to u values. An initial value can be chosen, such as la:1 = 1 -
j3(i - l)/(N-1) 0 6 j3 < 1, or it can be equal to the excitation obtained from Fig. 11.92 (a) Sector pattern of a 10-element array (H-plane). Relaxation synthesis:
a C, criterion
classical methods. The functional J(a) is replaced by two functionals R(a) and b C, criterion
SLL(a), and min J(a) can be expressed by the following improvement criteria c C3criterion
 666 Design and technology of low-cost printed antennas
c,, c,, c,: let a + be the actual vector of the iteration,
c,: [R(a+) < R(a) and SLL(a+) 5 SLL(a)] or [R(af) 6 R(a)
and SLL(af) < SLL(a)]
c2: [R(a+) < R(a) and SLL(af) 5 SLL,]
c,: [SLL(a+) < SLL(a) and R(a+) 5 R,;,]
The c, criterion application improves the ripple and the sidelobe level. c, gives
the best ripple for a given sidelobe level SLL,, and c3the best sidelobe level for
a given ripple R,. The c,, c,, c, criteria can be applied successively, depending
on the ripple and the sidelobe level requirements.
Fig. 11.93 Directional broadside pattern of a six-element array (H-plane)
-Relaxation synthesis
---- Experimental results
11.4.1.3 Directive broadside pattern case: For a broadside array of N equi-
spaced elements with equal excitations, the beamwidth between first nulls is
equal to 21,lNd radians. Applying c3 with R , , = 1 and 0, > Ao/Nd provides a
main beam with the lowest sidelobe level, taking d/lo and g(0) into account. If
d < 0.52, and g(0) = constant, we obtain the Chebyshev excitation. When
d > 0.5A0, the Chebyshev method does not always maintain equal sidelobe
levels. The relaxation method avoids such limitations, and printed antenna
arrays with guided-wavelength spaced sources can be considered.
11.4.1.4 Results: Let us consider a linear printed antenna array on a sub-
strate, having a relative dielectric constant E, = 2.17 (PTFE). For microstrip
transmission lines a, = z 0.75&.
(a) Sector pattern
We want to design a 10-source array with a 90' sector pattern and a transition
width, 2a = 20'. If the sources are equally spaced, the array amplitude factor
will be zero with any excitation vector for sin 0 = &/2d. To avoid a null in the
sector region, the first distance to the symmetry axis can be taken as 0.252,.
With aoT= (1; - 0.75; 0.5; - 0.25; O), g(0) = cos (0) and application of the
Design and technology of low-cost printed antennas 667
c, criterion, we obtain R = - 0.8 dB and SLL = - 29 dB for a solution vector
S (Fig. 1 1.92~).
With a0 = S, SLL,, = -25dB and application of c,. we obtain
R = - 0.3 dB and SLL = - 26 dB (Fig. 11.92b).
With a0 = S, R,,,, = - 1.5dB and application of c,, we obtain
R = - 1.5 dB and SLL = - 36 dB (Fig. 11.92~).
(6) Directive broadside pattern
On the same substrate, let us design a linear array of six sources with inter-
element spacings ,$. For equal excitations, the beamwidth between the first nulls
is 25.5" and the first sidelobe is at - 13dB. For 20, z 35", 2u = 10' (outline
corresponding to a - 40 dB sidelobe level, Chebyshev excitation), R,, = 1,
a: = 1, and a," = 0 V j # I, g(0) = cos(O), the c3 application provides a
- 40 dB sidelobe level (Fig. 11.93). The experimental pattern obtained with a
linear array of square patches is plotted in Fig. 11.93. The theoretical and
experimental outlines agree well if the low-level measurement difficulties are
taken into account.
11.4.2 Simplex method
Dantzig's algorithm [76] has been developed for linear programming with a real
variable. Under symmetrical amplitude and antisymmetrical phase conditions,
it is possible to compute real or complex excitations. The desired diagram is also
defined with an envelope specification.
11.4.2.1 Symmetrical pattern: If the pattern is considered at M angular
values (without correlation with N), the following inequalities are obtained:
F ( 0 J 5 F,
F(&) L F
2 (1 1.45)
F(0M) 5 FM
The problem is to find the ensemble V:
c, has been previously defined, and a functional J(a) = ElrJaj has to be
optimised (minimum or maximum) in order to promote some part of the
diagram for instance.
11.4.2.2 Asymmetrical pattern: Choosing a, = a-j and q5] = -q5-j, the
array factor can be written in the following form:
N
f (0) = 1 a, cos (k4 sin 0 + 4,)
j= l
668 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 669
The expansion of the cosine term leads to:
N element
F(0) = f (0)g(0) = (A, cos (ko4sin0) - B,sin (k04sin0)) g(0) I 1.000
j= l
2 0667
3 0267

A, and B, are the real unknowns to be determined from the new 2N-dimensions
linear problem. The algorithm can be used again, and at the end:

The introduction of M difference variables leads to a new linear system of M

equations (instead of inequalities) with N +
M unknowns associated with the
+
functional J(a) (or 2N M unknowns for an asymmetrical patterns). The
Dantzig algorithm shows that only one solution exists (if there is a solution),
which is found in a finite number of steps [73, 761.

element
m
./-I 1.000 D
+I-2 -0.236
+I-3 0.097
4-4 -0.073
+I-5 0.054

m
'I)

deg
deg
Fig. 11.94 Sector pattern of a 10-element array (H-plane). Simplex synthesis
g(O)=g (0)
Fig. 11.95 Directional patterns of a six-element array for various source patterns. Simplex
11.4.2.3 Examples: The second pattern previously mentioned has been com- synthesis
a Isotropic source
puted using the simplex method (Fig. 11.94). It appears that the amplitudes, b E-plane pattern
sidelobe level and ripple are very similar to the relaxation solution. Directional c H-plane pattern
 670 Design and technology of low-cost printed antennas
patterns are plotted in Fig. 11.95 for an array of six elements spaced 0.75 &;
three-element patterns have been considered (isotropic, E-plane and H-plane).
The directional patterns of a linear 32-equi-spaced-element array (spac-
ing = 0491,) in the H-plane are plotted in Fig. 11.96. The sidelobe levels were
Fig. 11.96 Computed directional broadside pattern of a 32-linear-element array (H-plane
and d l l = 0.87). Simplex synthesis
a Equi-amplitude
b CCIR-TVRO constraints and -20dB first sidelobe
Design and technology of low-cost printed antennas 671
constrained by the CCIR-TVRO conditions with an extra limitation of - 20dB
for the highest one. A cosecant-squared pattern (with a 30' window) was
achieved for a 30-element array with half-wavelength spacing. The E-plane-
degrees
Fig. 11 3 7 Computed cosecant squared pattern of 30-element linear array (E-plane and
d l l = 0.5). Simplex synthesis
degrees
Fig. 11.98 Computed 30' steered-beam pattern of 20-element linear array (H-plane and
d / L = 0.5). Simplex synthesis
element pattern is considered in Fig. 11.97. A 30" steered beam of a 20-element
array (with half-wavelength spacing) in the H-plane, with a sidelobe level lower
than - 25 dB, is shown in Fig. 11.98.
 672 Design and technology of low-cost printed antennas
Design and technology of low-cost printed antennas 673
11.4.3 Experimental results [73]
Both methods have been used to design sector and directional patterns in the
X-bands and K-bands, using corner-fed square patches.

11.4.3.1 Sector pattern: The specification is as follows:

ripple = f l dB
SLL < -25dB
To keep the feed line as simple as possible, the linear series array was chosen.
The constraint of straight lines between each element leads to a one-guided-
wavelength spacing (or a 0.75 free-space wavelength). However, a sector pattern
of 90" beamwidth cannot be obtained with this spacing, because a null occurs
in the B direction (where 0 = sin-' 1,/(21g)) whatever the amplitudes. T o avoid
this effect, one solution is to change the spacing. Fig. 11.99 shows the structure
which has been used. The two half arrays have been located closer together, then
the synthesis methods can perform the amplitude excitations, taking into acc-
ount the nonidentical spacing and the H-plane pattern of each patch. Theoreti-
cal results are presented in Table 11.I 1.

table 11.12 Amplitude distribution along near direction array (in E- and
H-plane)
Element nb H-plane Element nb E-plane
(6 elements) (8 elements)
I , , . . . , , . ,
90 60 30 0 30 60 90 +/- 1 1 +I- 1 1
Fig.
+/-2 0.627 +/-2 0.735
Measured sector pattern of a 10-element linear array and
F, = 10.8GHz) +/-3 0.2 12 +/-3 0.377
+/-4 0.110

The 180" phase shift between two neighbouring elements is easily obtained
using alternate positions along the feeding line. Quarter-wave transformers (one
Table 11.11 Amplitude and phase distribution along sector pattern array or two sections) are used to obtain the amplitude taper. The measured patterns
(co- and cross-polarised components) are plotted in Fig. 11.99.
Element Amplitude Phase
+I- I 1 0 11.4.3.2 Low side-lobe directive array (Fig. 11.100): Two steps were necess-
2 0.218 180 ary: just to synthesise the H-plane pattern (six elements) and secondly to
3 0.117 0 synthesise the E-plane pattern (eight elements). The amplitude taper was real-
4 0.082 180 ised in the usual way, using quarter-wave transformers. Table 11.12 presents the
5 0.075 0 taper values.
Figs. 1l.lOla and b show the measured patterns in the E and H-planes. The
674 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 675
sidelobe levels are higher than the expected - 10 dB. However, mutual coupling
has not been taken into account, and the reflectivity of the anechoid chamber
reaches - 40 dB.

I ,point d'alimentation

Fig. 11.100 48-element planar directionelarray (6 x 8 ) F, = 10.8GHz

11.5 New low-cost low-loss substrates 177,791

11S.1 Substrate choice

A large market demand for low-cost printed antennas has emerged from the
development of new types of civil communication such as direct-broadcasting
satellite reception, data transmission, communications between satellites and
mobiles (trucks, ships, etc.) and intruder detectors.
The price of a mass-produced printed antenna is directly related to substrate
and connector costs. The choice of the appropriate substrate depends on its
 676 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 677
temperature (LC) pressure

thn)
Fig. 11 .I02 Pressure and temperature cycles of polypropylene

Raw material

Cu-Polypropy lene-Cu

I Etching
wet process

If necessary goto step 2

Fig. 11 .I03 Steps in manufacture of double-sided printed-circuit board (step 7)

 678 Design and technology of low-cost printed antennas
application, nevertheless, many substrate properties may be involved: dielectric
constant, loss, and their variations with temperature and frequency ranges, and
mechanical and climatic stresses.
Conventionally, printed antennas require the use of a low-loss low-dielectric-
constant substrate. Unfortunately most printed-circuit boards currently used
are quite expensive; some are listed in Table 11.13. CNET (Centre National
d'Etudes des Telecommunications France) has developed a polypropylene sub-
strate whose characteristics are very similar to commercial substrates, while
remaining inexpensive.
Fabrication procedures and evaluation of this new substrate will be discussed
in the next Section.
11 S.2 Fabrication procedures
The fabrication of polypropylene printed-circuit boards is a very simple
procedure. Two types of board are considered:
Double-sided printed circuit: Cu-polypropylene-Cu
0 Thick metal-backing substrate: thick metal base (A1 or Cu) polypropylene-
Cu
Polypropylene of different thicknesses (ranging from 0.25mm to 1.6mm or
more) is manufactured by heating polypropylene granules to the melting point
(170C) and pressing them (Fig. 11.103). Pressure and temperature cycles are
Raw material
I Double sided ~ r i n t e dcircuit
Copolymere of polypropylene ethylene
I C : RT Duroid 5880
Copper foil (ZOpm or 35pm)
V after tests, and are compared with the RT Duroid 5880 substrate. The poly-
Multilayer printed cicuit
Fig. 11 .I04 Steps in manufacture of multi-layer printed-circuit board (step 2)
detailed in Fig. 1 1.102. Copper-foil or thick-metal backing is first chemically
processed and then laminated to the polypropylene-based dielectric by the same
procedure.
Laminated copper is usually selected in preference to electrical-grade copper,
which is more lossy. Thicknesses are commonly 20pm or 35pm. .
Design and technology of low-cost printed antennas 679
Thick-backing construction offers significant advantages over conventional
designs [78]:
0 It provides high reliability for connector mounting
For high-power applications heat generated by devices can be dissipated.
Because the substrate is not loaded with glass fibre, it tends to warp when the
internal stress between copper and polypropylene is too high; a thick metal
cladding can ensure flatness.
The same process can be extended to multilayer structures (Fig. 11.104), but
in order not to damage the first layer, a copolymer or polypropylene-ethylene
is used.
The electrical characteristics are almost similar, but the melting point is 20"
lower (150C instead of 170C); applications to triplate feeding network and
stacked patches are straightforward as there is no need for bonding the film
between different layers. Examples of practical applications are discussed below.
11.5.3 Electrical characteristics
Dispersion measurements were carried out by the ring-resonator technique in
triplate technology (in order to avoid unwanted radiation).
Overall losses are plotted against frequency in Fig. 11.105 for different types
of printed circuits and are compared with a PTFE substrate (RT Duroid 5880):
A: Cu (20 pm)-polypropylene-Cu (20 pm)
B: Aluminium (Al) (4 mm)-polypropylene-Cu (20 pm)
The dielectric constant remains constant with frequency. Fig. 11.106 shows that
polypropylene and the copolymer of polypropylene-ethylene have the same
electrical performance.
11 3 . 4 Environmental tests
11.5.4.1 Damp heat: 95% relative humidity at 40" and for 22 days: Again,
overall losses are plotted as a function of frequency in Fig. 11.107 before and
propylene does not seem to be very affected by the test conditions, while the
PTFE substrate losses are slightly increased. This is probably due to the fact that
this substrate is loaded with glass fibre.
11.5.4.2 Thermal shocks: Tests have been performed on a 0 4 m m printed
circuit with 4 mm aluminium (Al) backing. The board survived the following
tests:
 680 Design and technology of low-cost printed antennas 681
Design and technology of low-cost printed antennas
damp heat (9570hr 40C 21days)

PTFE af!er test (4

PTFE before test (-1

2 polypropylene before test
polypropylene after !est

0L
35 55 73 90 10.8 12.7 14 5

- frequency GHz

Fig. 11.107 Losses before and after test of damp heat

frequency GHz

Fig. 11 .lo5 Losses (dBlm) versus frequency for various printed circuits
( A ) Cu 20pm-polypropylene-Cu 2 0 p m
(8)A l (4 mm)-polypropylene-Cu 2 0 p m
(C) RT-Duroid 58-80 (copper on each side)

01
3.3 6.7 10.1 13.5 16.9
frequency,GHz
Fig. 11.108 50-element cross-fed array printed on a 0 . 4 m m polypropylene substrate. Fre-
Fig. 11.106 Comparison of polypropylene (A) and copolymer of polypropylene ethylene (6)
quency = 2 0 G H z
 682 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 683
-40C +85OC a 50-element array is shown in Fig. 11.108. This array operates at 19.6 GHz, and
- 55OC + 10o0c its measured gain is 23 dB. The antenna has already been discussed in a previous
Section.
but was destroyed at - 65OC + 125OC. The three-element array in Fig. 11.109 was designed for data transmission
between mobiles at 23.5 GHz. The required performance was:
E-plane: 3 dB beamwidth = 40'
H-plane: 3 dB beamwidth = 60'
In order to meet the 60" requirement in the H-plane, the distance between
patches has to be very small, which is the reason why this three-element geo-
metry was chosen. The measured gain is I1 dB because the patches are very
close. It is not possible to match the antenna with 0.25l microstrip line transfor-
mer. The problem has been solved by using a coaxial-line transformer machined
in the 4 mm aluminium ground plane which provides a VSWR of 1.5. Radiation
patterns are given in Fig. 11.110.

11.5.5.2 Multi-layer printed antennas: A printed-slot array was designed in

Fig. 11.109 Three-element array printed on a 0.4mrn polypropylene substrate. Fre-
quency = 23.5 GHz
113 . 5 Examples of printed antennas on polypropylene substrates
The two examples which are described here give a good insight into the possibili-
ties of this new technology.
11.5.5.1 Single-sided printed antennas: A few types of microstrip antenna
arrays have been designed for the 20GHz band with gain from I 1 dB to 23 dB
[77]. These antennas are printed on a 0.4mm polypropylene substrate with 4mm
aluminium backing, which operates as the ground plane of the patches, ensures
the flatness of the antenna and eases the connector mounting. A photograph of
order to test the possibilities of using polypropylene technology for large dimen-
sions. This array was also designed with the aim of replacing a Yagi antenna,
whose sidelobes level are very high, by a flat antenna which should have better
characteristics and be less expensive. The specifications were:
1.38-1.525 GHz
Gain: 16dB to 17dB
3 dB beamwidth = 20' to 25O
Linear polarisation
Sidelobe level < - 20 dB
The gain specification implies dimensions of 60 x 60cm. The printed-slot
element was selected because it can provide wider bandwidth than microstrip
patches, and also because it is very easy to feed. The fabrication of the antenna
requires few steps. The two copper layers are separated by a 0.8 mm polypropy-
lene layer. The slots are printed on the upper face of the substrate while the feed
network is printed on the rear side. The printed-slot plane is covered by a second
0.8 mm polypropylene layer which ensures flatness of the antenna. Photographs
of the antenna are given in Fig. 11.111.
The printed slots are electromagnetically coupled to the microstrip lines and
the array is suspended over a reflector plane in order to get unidirectional
radiation patterns. Since the dielectric thickness is very small compared with the
wavelength, the dimensions of the slot are such that 2nr = A. Input impedance
is measured in the plane of the outer edge of the slot and then matched to 50 0
by a matching network. The slot width is W/1 = 0.017. The VSWR of a single
element is plotted against frequency in Figs. 11.112a and b for two spacings
h/l = 0.12 and h / l = 0.25 of the reflector plane. In the first case, the matching
is provided by a quarter-wavelength transformer, while in the second case a
 684 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 685
quarter-wavelength transformer and two stubs are required to fit the bandwidth
plan E requirement. It appears that the bandwidth decreases as the spacing increases.
\ 23.5GHz The 4 x 4-slots array is designed for a 25 dB Chebyshev taper. The results are

Fig. 11 .I11 Printed slot array. Frequency = 1.5 GHz

a Top view of the printed-slot array beside a Yagi antenna under a radome
b Rear view of the printed-slot array and its feed network

as follows:
Slot width = 0.0171
Reflector height = 0.251
Gain: 17.2 dB
Efficiency: 67%
Radiation patterns are given in Fig. 11.113. Although the radiation patterns and
gain are good, a good match over the whole frequency band could not be
obtained. The VSWR remains below 2.5 over the whole-frequency band. This
is due to the very high coupling between slots, which is increased by the
proximity of the reflector plane.

11.6 Concluding remarks

Fig. 11.1 10 Radiation patterns of the three-element array (Fig. 1 1.107) The design of low-cost printed antenna arrays needs accurate analysis of par-
( a ) E-plane
( b ) H-plane ameters such as resonant frequency, input impedance, mutual coupling and
686 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 687
688 Design and technology of low-cost printed antennas
sidelobe level with respect to the dimensions, dielectric constant, thickness of
protective layer and inter-element spacing. The transmission-line model, the
cavity model and the spectral-domain approach are complementary tools which
can be used together. In addition, specific synthesis methods have been de-
veloped to maintain the flexibility offered by printed structures; namely non-
identical element spacing and non-identical E- and H-plane element patterns.
Finally, the polypropylene substrate, whose characteristics are very close to
commercial substrates, is very cost competitive in most applications. On the
other hand, thick-backing construction using either metal or metallised dielec-
tric, and multi-layer structures without bonding film have been experimented
with in order to reduce overall cost.
11.7 References
DERNERYD, A. G., and LIND, A. G.: 'Extended analysis of rectangular microstrip resona-
tor antennas', IEEE Trnns., 1979, AP-27, pp. 846-849
VANDESANDE, J., PUES. H., and VAN DE CAPELLE, A.: 'Calculation of the bandwidth
of microstrip resonator antennas'. Proc. 9th European Microwave Conference, Brighton,
1979, pp. 116-119
DUBOST. G.: 'Linear transmission-line model analysis of arbitrary shape patch antennas',
Electron. Lett., 1986. 22, pp. 798-799
LO, Y. T., SOLOMON, D., and RICHARDS, W. F.: 'Theory and experiment on microstrip
antennas', IEEE Trans., 1979, AP-27, pp. 137-145
JAMES, J. C., HALL, P. S., and WOOD, C.: 'Microstrip antenna theory and design' (Peter
Peregrinus, 1981)
BAHL, I. J., and BHARTHIA, P.: 'Microstrip antennas (Artech House', USA, 1980)
WOOD, C.: 'Improved bandwidth of microstrip antennas using parasite elements', IEE Proc.,
1980, 127, pp. 231-234
PENARD. E., and DANIEL, J. P.: 'Open and hybrid microstrip antennas'. IEE Proc., 1984,
131H, pp. 3 8 4 4
MAHDJOUBI, K., DANIEL, J. P., and TERRET, C.: 'Etude d'antennes imprimCes a accis
multiples', Ann. de Telecom, 1985, 40, pp. 190-203
MAHDJOUBI. K.. DANIEL, J. P., and TERRET, C.: 'Dual frequency disc antenna studied
by cavity method', Electron. Lett., 1986, 22, pp. 379-381
MAHDJOUBI, K.. and TERRET, C.: 'Exact expression for stored energies in the cavity
volume of microstrip antennas', Electron. Letl., 1985, 21, pp. 1221-1222
PENARD. E., and DANIEL, J. P.: 'Electric and magnetic currents in microstrip antenna
theory'. Int. IEEEIURSI, Albulquerque, New Mexico, 1982
PENARD, E.. and DANIEL, J. P.: 'Calcul du rayonnement d'antennes microstrip: deux
exemples. Journies Nationales Micro-ondes de Toulouse, 1981
NEWMAN, E. D.. and TULYATHAN, P.: 'Analysis of microstrip antennas using moment
method', IEEE Trans., 1981, AP-29, pp. 47-53
LEWIN. L.: 'Radiation from discontinuities in strip-lines', IEE Monograph 358, 1960
HARRINGTON, R. F.: 'Time-harmonic electromagnetic fields' (McGraw-Hill Book, NY,
1961)
AAS, J. A., and JAKOBSEN, K.: 'Radiation patterns of rectangular microstrip antennas on
finite ground plane'. 12th European Microwave Conference, Helsinki, 1982
POZAR. D. M.: 'Finite phase arrays of rectangular microstr~ppatches', IEEE Trans., 1986,
AP-34, pp. 658-665
Design and technology of low-cost printed antennas 689
19 POZAR, D. M.: 'Input impedance and mutual coupling of rectangular microstrip antennas',
IEEE Trans., 1982, AP-30, pp. 1191-1 196
20 ITOH. T., and MENZEL, N.: 'A full wave analysis method for open microstrip structures',
IEEE Trans., 1981, AP-29, pp. 63-68
21 DESHPANDE, M. D., and BAILEY, M. C.: 'Input impedance of microstrip antennas', IEEE
Trans., 1982, AP-30, pp. 645-650
22 MOSIG, J.: 'Les structures microrubans, analyse au moyen des Cquations intigrales', D.Sc
Thesis, EPFUL, Lausanne, Switzerland, 1984
23 KATEHI, P. B., and ALEXOPOULOS, N. G.: 'On the modelling of electromagnetically
coupled microstrip antennas. The printed strip dipoles', IEEE Trans., 1984, AP-32, pp.
1179-1 I86
24 ROUDOT, B.: 'Analyse d'antennes imprimees par une approache dans le domaine spectral',
D.Sc Thesis, University of Rennes, France, 1985
25 PENARD, E.: 'Etude d'antennes imprimies par la mdthode de la caviti applications au
couplage'. D.Sc Thesis, University of Rennes, France, 1982
26 BAHL, 1. J., STUCHLY, S. S., and BHARTIA, P.: 'Design of microstrip antennas covered
with a dielectric layer', IEEE Trans., 1982, AP-30, pp. 314-318
27 DE ASSIS FONSECA, S. B., and GIAROLA, A. J.: 'Microstrip disk antennas. pt. 2: The
problem of surface wave radiated by dielectric truncation', IEEE Trans., 1984, AP-32, pp.
561-573
28 ROUDOT, B., TERRET, C., DANIEL, J. P., PRIBETICH, P., and KENNIS, P.: 'Funda-
mental surface-wave effects on microstrip antenna radiation'. Electron. Lett., 1985, 21, pp.
1112-1114
29 DE ASSIS FONSECA, S. B., and GIAROLA, A. J.: 'Influence of surface-wave excitation
efficiency of space-wave launching in microstrip disc antennas', Electron. Lett., 1982, 18, pp.
406407
30 GRIFFIN, J. M., and FORREST, J. R.: 'Broadband circular disc microstrip antennas',
Electron. Lett., 1982, 18, p. 266-269
31 DAS, N., and CHATTERJEE, J. S.: 'Conically depressed microstrip patch antenna', IEE
Proc., 1983, 130H. pp. 193-196
32 JEDDARI, L., MAHDJOUBI, K., TERRET, C., and DANIEL, J. P., 'Broadband conical
microstrip antenna', Electron. Letl., 1985, 21, pp. 896-898
33 MAILLOUX, R. J.: 'Printed slot arrays with dielectric substrates'. IEEE Symposium on
Antenna and Propagation, June 1985
34 OLINER, A. A.: 'The radiation conductance of a series slot in strip transmission line'. IRE
Conv. Rec. No. 2, 1954, Pt. 8, pp. 89-90
35 BREITHAUPT, R. W.: 'Conductance data for offset series slot in stripline', IEEE Trans.,
1968, MMT-16, pp. 969-970
36 TOBARIAS, 3.: 'Fente rayonnante bidirectionnelle a la rhonance et a I'antiresonance alimen-
tie par une ligne microruban'. Gme Journees Nationales Micro-ondes, Lannion, June 1984,
pp. 236-237
37 TOBARIAS, J., and TERRET, C.: 'Fente rayonnante a la resonance et a l'antiresonance
alimentie par une ligne coplanaire'. Gme Journees Nationales Micro-ondes, Lannion, June
1984, pp. 242-243
38 DUBOST, G., and ZISLER, S.: 'Antennes a large bande', (Masson Editors, France, 1976)
39 WEEKS, W. L.: 'Antennas engineering' (McGraw-Hill, 1968)
40 MARCANO, D., SAILLARD, J., TERRET, C., and DANIEL, J. P.: 'Reseau de fentes
balayage electronique'. 5eme Journees Nationales Micro-ondes, Nice, pp. 265-267
41 YOSHIMURA, Y.: 'A microstrip line slot antenna', IEEE Trans., 1972, MTT-20, pp. 760-762
42 DAS, B. N., and JOSHI, K. K.: 'Impedance of a radiating slot in the ground plane of a
mirostripline', IEEE Trans., 1982, AP-30, pp. 922-926
43 BOOKER, G. G.: 'Slot aerials and their relation to complementary wire aerials', J. IEE, 1946,
93, Pt. IIIA, pp. 620-626
 690 Design and technology of low-cost printed antennas Design and technology of low-cost printed antennas 691

44 NESIC, A.: 'Slotted antenna array excited by a coplanar waveguide', Electron. Lett., 1982,13, 71 DANIEL, J. P., MUTZIG, J. P., NEDELEC, M., and PENARD, E.: 'Reseaux d'antennes
pp. 404-406 imprimees dans la bande 20/30GHz'. 4eme Journees Nationales Microondes, Lannion,
45 NESIC, A,: 'A printed antenna array with slots as primary radiators for phase scanned France, June 1984, pp. 246-247
antenna'. JINA, Nice. Nov. 1986, pp. 281-283 72 DANIEL, J. P., MUTZIG, J. P., NEDELEC, M., and PENARD, E.: 'Reseaux d'antennes
46 DUSSEUX, T.: 'Etude d'antennes fentes annulaires imprimees applications antennas melan- imprimees dans la bande 20/30GHz', Lbnde ~ I e c t r i ~ u e1985,
, 65, pp. 35-41
geuses, reseaux'. D S c Ing. Thesis, University of Rennes, May 1987 73 BOGUAIS, M.: 'Contribution a la synthese de reseaux d'antennes, realisation en technologic
47 COHN, S. B.: 'Slot-line on a dielectric substrate, IEEE Trans., 1969, MTT-17, pp. 768-778 imprimbe. D.Sc Thesis, University of Rennes, France
48 DUSSEUX, T., DANIEL, J . P., and TERRET, C.: 'Theoretical and experimental results of 74 BOGUAIS, M., DANIEL, 1. P., and TERRET, C.: 'Antenna pattern synthesis using a
guided wavelength of a slot on a low permittivity substrate', Electron. Lett., 1986, 22, pp. relaxation method: application to printed antennas', Electron. Lett., 1986, 22, (7)
589-590 75 BOGUAIS, M., DANIEL, I. P., and TERRET, C.: 'Deux methodes de synthese de reseaux
49 JANASWANY, R., and SCHAUBERT, D. H.: 'Dispersion characteristics for wide slot lines d'antennes, application aux antennes imprimees', JINA, Nice, 1986, pp. 3 10-31 1
on low permittivity substrates', IEEE Trans., 1985, MTT-33, pp. 723-726 76 DANTZIG, G. B., and ORCHARD-HAYS, W.: 'The product form for the inverse in the
50 KAWANO, K., and TONIMORO, H.: 'Slot ring resonator and dispersion measurement on simplex method', Math. Comp., 1959
slot lines'. Electron. Lett., 1981, 17, pp. 916-917 77 DEMEURE, L.: 'New low cost and low loss substrate: Application to printed antenna', JINA,
51 GARG, R., and GUPTA, K. C.: 'Expressions for wavelength and impedance of a slot line', Nice, France, Nov. 1986
IEEE Trans., 1976, M'IT-24, p. 532 78 BONFIELD, R.: 'Thick metal backing adds value to substrate', Microwaves and RF, Feb. 1987
52 JAMES, J. R., and HENDERSON, A,: 'High-frequency behaviour of microstrip open-circuit 79 Patent 84 402 7078: 'Support metallis6 a base de polypropylene et procede de fabrication de
terminations', IEE J. Microwaves. Optics and Acoustics, 1979, 3, pp. 205-218 ce support'
53 ELLIOTT, R. S.: 'The theory of antenna arrays', in HANSEN, R. C. (Ed.): Microwave
scanning antennas: Vol. 11, (Academic Press, 1986), chap. 1
54 HANSEN, R. C.: 'Linear arrays' and 'Planar arrays' in RUDGE, A. W. el al. (Eds): 'The
handbook of antenna design: Vol 11' (Peter Peregrinus, 1983). chaps. 9 and 10
55 BACH, H.: 'Directivity of basic linear arrays', IEEE Trans., 1970, AP-18, pp. 107-110
56 BACH, H., and HANSEN, J. E.: 'Uniformly spaced arrays' in COLLIN, R. E., and ZUCK-
ER, F. J. (Eds.): 'Antenna theory. Pt. 1' (McGraw-Hill, NY, 1969)
57 HANSEN, R. C.: 'Aperture efficiency of Villeneuve n arrays', IEEE Trans., 1985, AP-33, pp.
666-669
58 VILLENEUVE, A. T.: 'Taylor patterns for discrete arrays', IEEE Trans., 1984, AP-32, pp.
1089-1093
59 HANSEN, R. C.: 'Comparison of square array directivity formulas', IEEE Trans., 1972,
AP-20, pp. 100-102
60 WHITTAKER E. T., and WATSON, G. N.: 'A course of modern analysis', Cambridge,
London, 1962, p. 170
61 JEDLICKA, R. P., and CARVER, K. R.: 'Mutual coupling between microstrip patch anten-
nas'. Proc. Workshop on printed circuit antenna technology, Oct. 1979
62 JEDLICKA, R. P., POE, M. T., and CARVER, K. R.: 'Measured mutual coupling between
microstrip antennas', IEEE Trans., 1981, AP-29, (1)
63 MALKOMES, M.: 'Mutual coupling between microstrip patch antennas', Electron. Lett.,
1982, 18, pp. 520-522
64 VAN LIL, E. H.. and VAN DE CAPELLE, A. R.: 'Transmission line model for mutual
coupling between microstrip antennas', IEEE Trans., 1984, AP-32, pp. 816-821
65 PENARD, E., and DANIEL, J. P.: 'Mutual coupling between microstrip antennas', Electron.
Lett.. 1982, 18, pp. 605-607
66 MAHDJOUBI, K., PENARD, E., DANIEL, J. P., and TERRET, C.: 'Mutual coupling
between circular disc microstrip antennas', Electron. Lett., 1987, 23, pp. 27-28
67 BHAlTACHARYYA, A. K., and SHAFAI, L.: 'Surface wave coupling between circular
patch antennas', Electron. Let(., 1986, 22, pp. 1198-1200
68 MAHDJOUBI, K., PENARD, E., TERRET, C., and DANIEL, J. P.: 'Mutual coupling
between microstrip disk antennas'. ICAP '87, late papers, University of York, 1987
69 DANIEL, J. P., PENARD, E., NEDELEC, M., and MUTZIG, J. P.: 'Design of low cost
printed antenna arrays'. Proceedings of ISAP'85, Kyoto, Aug. 1985, pp. 121-124
70 WILLIAMS, J. C.: 'Cross fed printed aerial'. Proc. 7th European Microwave Conf. Copen-
hagen, 1977, p. 292
 Chapter 12

Analysis and design considerations

for printed phased-array antennas
D. M. Pozar

12.1 Introduction

Until the last decade or so, phased-array technology generally employed dipole
or waveguide radiating elements, with waveguide or coaxial lines for feed
networks [l-41. In more recent years, however, printed or microstrip arrays and
feedlines have become quite popular [5-71 owing to features including light
weight, conformability, ease of manufacture and, probably most important,
potentially low cost. Economics is generally the most critical factor affecting the
deployment of phased arrays into more systems, as a variety of applications
would benefit from the advantages of a phased-array antenna, which include
rapid and selective beam steering, adaptive nulling, and other controlled array-
illumination functions. The printed phased array, with its fabrication simplified
through the use of photolithographic techniques, offers the promise of lower-
cost electronically scanned arrays. This is in spite of some inherent disadvan-
tages of printed antennas, such as low bandwidth and power capacity.
Printed arrays can take many different forms. Radiating elements may be
printed dipoles, printed (microstrip) patches or slot elements. Feed circuitry
may be in microstripline, or in stripline form. Several combinations then exist
for the interconnection of feed lines to radiating elements. One approach is to
etch the radiating elements and feed lines in microstrip form on the same
substrate, while other approaches use two or more layers to separate the
radiating elements from feed circuitry. Phase shifting and other active circuitry
functions can be incorporated in hybrid form. At millimetre-wave frequencies
the physical size of the array may be small enough so that circuit integration can
be carried one step further, resulting in the 'monolithic phased array'. This
concept, discussed in more detail in Section 12.3, involves the integration of all
active circuitry required for a sub-array module of a millimetre-wave phased
array.
This Chapter first considers the rigorous analysis of several canonical printed-
array geometries (Section 12.2), and then discusses some design considerations
for printed arrays (Section 12.3). During the 1960s, a large analysis effort was
694 Analysis and design considerations for phased-array antennas
carried out for waveguide and dipole phased arrays [I-41; a corresponding effort
for printed phased arrays is still needed, but the present Chapter consolidates
some of the solutions which have been completed to date. Section 12.2.2 treats
infinite planar arrays of various printed elements. While most of these cases are
idealised in some way, they represent a starting point for the analysis and/or
design of more practical arrays. In addition, as discussed in more detail in
Section 12.2.2.5, the scan performance of phased arrays is often more dependent
on substrate parameters and element spacing than on the particular details of
the feeding method. In Section 12.3 dealing with design considerations, the
monolithic phased-array concept is emphasised, but much of this material is
relevant to non-monolithic printed arrays as well.
12.2 Analysis of some canonical printed phased-array geometries
In this Section we present analyses for several types of printed phased-array
geometries. These problems are canonical in that the geometries are idealised in
some sense, usually in terms of simplifying assumptions about the feed. The
solutions here all have a high degree of commonality, being based on the work
of the author and his colleagues at the University of Massachusetts. The
relevant work of other researchers, however, is noted and discussed in relation
to the analyses presented here.
The basic procedure for the analysis of each of the printed-array geometries
in this Section is as follows. First, the Green's function for the relevant dielec-
tric-slab geometry is derived in spectral (transform)-domain form, for a single
infinitesimal source (electric or magnetic dipole). This result is then extended to
an infinite periodic planar array of such sources, with a progressive phase shift
for scanning at the desired angle. A moment-method solution is formulated for
the unknown current distribution on the antenna element, and an appropriate
set of expansion weighting functions is chosen. An impedance matrix results,
which can then be used to determine the unknown coefficients of the expansion
modes. Because of the periodic nature of the array, the current distributions on
all of the elements are the same, except for the imposed progressive phase shift.
Thus, formulating the moment-method solution for one 'unit cell' is equivalent
to imposing the solution across the entire array. Mutual coupling is implicitly
included in the solution. This method has variously been referred to as a
'full-wave solution', or the 'Galerkin method in the spectral domain', and has
been applied to a variety of antenna and microwave circuit problems, in both
single-element and array form.
After the currents have been determined, other quantities of interest can easily
be found. The variation of input impedance with scan angle can be calculated;
this result is quite important for matching the array over the desired scan range.
A related quantity is the active-element pattern, which also gives information
about the scan performance of the array. Other quantities of interest include the
Analysis and design considerations for phased-array antennas 695
cross-polarisation level, and possibly the efficiency of the array. Pattern quan-
tities such as directivity and sidelobe level depend on the size of the array, and
so are not very meaningful for infinite arrays (since an infinite array radiates a
plane wave, its directivity is infinite, while its sidelobe level is zero).
In the following Subsections, we first present some material that is common
to most of the solutions which follow, including a brief derivation of a typical
Green's function and a discussion of the scan-blindness effect. We then treat
several canonical infinite planar printed arrays, and present results for the scan
performance of such arrays. Next, solutions are described for finite arrays of
dipoles and rectangular microstrip patches. The finite-array problem is con-
siderably more difficult than the corresponding infinite-array problem, but may
be of more practical utility since it includes edge effects.
12.2.1 Some preliminaries
12.2.1.1 Derivation of the Green's function of a grounded dielectric slab: -
Central to the solutions that follow is the exact Green's function of the dielec-
tric-slab geometry in spectral, or transform, domain form. Such Green's fun-
ctions have appeared in a number of recent papers on printed-antenna analysis
[8-121, but generally without derivation. Thus it may be useful to present a short
derivation of a Green's function, in case the reader is not familiar with the basic
procedure. We will derive the Green's function for a grounded dielectric slab,
with an infinitesimal electric-current source on its surface. This is one of the
most useful Green's function results, being applicable to all of the array geome-
tries below (some additional Green's functions are needed for the slot elements
of Section 12.2.2.4). The same procedure, however, can be used to obtain the
Green's functions for a number of more general cases, including the following:
Two (or more) dielectric-layer geometry
Dielectric layer with a lossy (surface-impedance) ground plane
Substrate with magnetic properties
Anisotropic substrates
Fig. 12.1 shows the geometry of a grounded isotropic dielectric slab of thickness
d and relative permittivity E , . The source is an infinitesimal f-directed electric
dipole, of unit strength and located on the surface of the dielectric slab at (x,,
yo, d). We desire to find the E,, Ev and E, fields generated by this source.
While it is possible, and quite common, to introduce vector potentials, it is
actually simpler to work directly with wave equations for El and Hz, and find
the transverse fields from these field components. Thus, Maxwell's equations,
V x E = -jopoH (12.14
V x H = jo~E (12.lb)
can be solved simultaneously for the usual Helmholtz wave equations in a
source-free region:
 696 Analysis and design considerations for phased-array antennas Analysis and design considerations for phased-array antennas 697
These results are for the dielectric region 0 < z < d, but can be used for the air
region z < d by setting E, = 1. Similarly, a2/az2= - k: or - k:, depending on
whether 0 < z < d, or z z d, respectively.
where k2 = m2p0c,and E = %er for 0 < z < d, and E = E~ for z > d. Anti- From the wave equations of 1 2 . 2 and~ 12.26, the general solutions for &and
cipating a plane-wave form of solution, with a propagation factor egk\" e*IkyY H~are:
etJk:', and substituting this form into eqn. 12.2a,b gives the propagation con-
stants in the z-direction as
fi -
-I
Be-jk~z
for z z d (12.6b)
Ez = Ccosk,z + Dsinklz for 0 < z < d (12.6~)
with B2 = k: +
g . In the above, the branch of the square-root function should I?, for~ Z0 < z < d
= Esink,z+ F C O S ~ (12.6d)
be chosen so that Im(k,) < 0 and Im(k2) < 0. where outgoing waves have been assumed in the region z > d. With these forms
the transverse field components of eqns. 12.5~-dcan be rewritten as

Fig. 12.1 Geometry of an infinitesimal f -directed current element on a grounded dielectric

slab

We now define a Fourier-transform pair as which also apply to the region z > dafter E, is set to unity. Applying eqns. 1 2 . 7 ~
and b to eqns. 1 2 . 6 ~and d to enforce the boundary condition that E, = Ey = 0
at z = 0 yields D = F = 0. There then remains four constants (A, B, C, E ) to
be evaluated by the continuity of Ex, Ey and H, at z = d (the dielectric-air
interface), and a jump condition in H,at z = d (due to the current_source).After
some straightforward algebraic manipulation, the results for E, and $ for
Then, in-the transform domain, the transverse fields can be written in terms of 0 < z < dare
zzand H,as,
(erg + $) Ex = jk.
a -
a, E, + oh kyfix
-
-
Hz =
-jky sink,
-
e-jk.r"o e - j k y ~ ~
(12.8b)
Te
where
T, = +
~,k,cosk,d jk,sink,d (12.9~)
T, = k, cosk,d + jk,sin k,d (12.9b)
and Zo = a. The zeros of the T,,, and T, functions correspond to the TM
and TE surface-wave poles of the grounded dielectric slab.
698 Analysis and design considerations for phased-array antennas
Using eqns. 12.7~-dand eqns. 1 2 . 8 ~and b and taking the inverse transform
in eqn. 1 2 . 4 ~allows the transverse electric fields at z = d to be evaluated as
Analysis and design considerations for phased-array antennas 699
are the direction cosines. Then by superposition, eqns. 1 2 . 1 0 and
~ b can be used
to find the total field from this infinite array (after replacing rn, n, k, and ky with
- rn, - n, - k , and - k,, , respectively):

where the following quantities have been defined:

G.:; =
-jZo (E, ki
-
- kS)k2cos k, d + jk, (g - k:) sink, d .
sin k, d
ko T, T,

, =
- kxk, sin k, d[k2cos k, d
jZo + jk, sink, d l
G'" k, T, T ,
(12.11b)
In eqns. 12.10a and b the notation E0 has been used to denote the field due to
a single source. The above field expressions are directly applicable to the analysis
of isolated antennas printed on the surface of a dielectric slab. Note that the
results for Ex and E, of eqns. 12.10~and b satisfy reciprocity, as an interchange
of x , x, and y, yo does not change the result.

12.2.1.2 Extension to an infinite array: We now show how the Green's fun-
ction of the previous Section for a single infinitesimal electric dipole can be
generalised to an infinite phased array of such sources. Fig. 12.2 shows the
geometry of an infinite periodic array of infinitesimal sources, with spacing a in
the E-plane (x) direction, and spacing b in the H-plane (y) direction. Them, nth
s a m e is thus located at
where rn, n are integer indices with - co < m, n < co. Now for scanning at the
angle 0, q5 the currents on the rn, nth source must be phased as
-jko(muu +nbv)
(12.13)
where
u = sinOcos4
Fig. 12.2 Geometry of an infinite periodic array of R-directed infinitesimal current elements
on a grounded dielectric slab
Eqns. 12.15a and b give the transverse fields at (x, y, d) due to the infinite array
of dipoles and may be thought of as the Green's function of the infinite array.
Observe, however, that E, and Ey of eqns. 12.15~ and b do not satisfy reciprocity
upon interchange of x , xo and y, yo. This is due to the asymmetry introduced by
the phasing of eqn. 12.13.
Eqns. 12.15~and b are rigorous expressions, but clearly not in a very usable
form from a computational viewpoint. The Poisson-sum formula can be applied
to greatly simplify the result. Consider an expression of the form
The Poisson sum formula can be written as

v = sin Osin q5 m
ejm""' F(rnwo) = T
m
f( t + mT)
 700 Analysis and design considerations for phased-array antennas Analysis and design considerations for phased-array antennas 707
where T = 2n/w,, and f (t) and F(w) form a Fourier transform pair: in a number of different types of arrays [2, 4, 8, 131, and is generally related to
the resonance of some type of trapped or guided mode of the array structure.
~ ( w )= J-
mm f(t) edw' dl (12.18a) For example, waveguide arrays with dielectric plugs or dielectric cover layers

f(t) = -In -"

1
2i7
F(w)eJW'dt (12.186)
have exhibited the scan-blindness phenomenon [4]. Printed phased arrays,
because of the presence of a dielectric slab, also show scan-blindness effects.
Scan blindness is total only in infinite arrays (or waveguide simulators), but in
Now let f (I) = h(r) el"". Then F(w) = H(w - w, ), and eqn. 12.17 becomes large arrays the effect can be severe enough to seriously degrade performance.
Thus it is important to both understand the scan-blindness effect and to be able
to predict the occurrence of this effect in printed phased arrays. Unless otherwise
stated, the discussion below refers to an infinite array.
One way of observing scan blindness is to look at the reflection-coefficient
magnitude at one element of an infinite phased array. Assuming a reasonable
impedance match at broadside scan (0 = 0') the reflection-coefficient mag-
Now compare eqns. 12.19 and 12.16, and let t = k,u, w, = a, t' = k,, w, = x, nitude will be small there. As the scan angle increases towards endfire (0 = 90),
T = 2nla and h = q, to get the reflection-coefficient magnitude must increase to unity, since an infinite
phased array does not transmit any real power away from the face of the array
at endfire scan. A scan blindness, however, will show its presence by unity (or
near unity) reflection-coefficient magnitude at some scan angle before endfire.
with This means that each element of the transmitting array is reflecting all the power
incident on it, and so the array is 'blind' at this scan angle.
The active-element pattern of the array provides another way of looking at
the scan-blindness effect. The active-element pattern is defined as the radiation
pattern of an array obtained when one element is driven and all other elements
So we see that the Poisson-sum formula can be used to eliminate the infinite are terminated in matched loads [3]. If no grating lobes are present, it can be
integration of eqn. 12.16. This result can be applied twice (fork, and k,,) to eqns. shown [3] that the active-element (power) pattern F(0, 4 ) of an array is related
12.15~and b), to give the following results: to the active reflection coefficient R(0, 4 ) by
w , 4 ) = (1 - I R ( ~4)1*)cose
, (12.24)
The active-element pattern is significant because it is relatively easy to measure
(no power-dividerlphase-shifter network is required), and it provides informa-
tion about the scan performance of the array. A scan-blindness condition will
show up as a null in the active-element pattern, owing to the unity reflection
where the variables k, and k, take on the discrete values
coefficient at the blind spot. Thus, some workers refer to scan blindnesses as
2nm nulls in the active-element pattern.
k, = -+k
a 'O (12.23~) One can discuss the scan-blindness effect from several viewpoints. If a specific
array geometry is being considered, a rigorous analysis of the scan performance
of the array will allow the prediction of the blindness effect. This will be done
in the following Sections for several printed arrays of practical interest. In
Note the similarity between eqns. 12.22 and 12.10; eqn. 12.22 can be con- addition, certain canonical problems can be posed which are complete enough
sidered as a discretised version of the continuous integration in eqn. 12.10. to yield data on the scan performance of printed arrays, including blindness
angles, in a more general sense. This has been done with the current sheet model
12.2.1.3 The scan-blindness effect: Scan blindness refers to a condition discussed in Section 12.2.2.5, and with the infinite array of infinitesimal dipoles
where, for certain scan angles, no real power can be transmitted (or received) by discussed below.
a phased array. This effect has been experimentally and theoretically observed
702 Analysis and design considerations for phased-array antennas
If no details of the array geometry are available, or if one wishes to look at
the blindness effect from a different point of view, an analysis based on mutual
coupling (measured or calculated) can be used. Thus, consider an infinite planar
array, with the elements indexed as in eqn. 12.12. Let the reference element be
the m = n = 0 element, and assume each element is fed with a unit amplitude
voltage source having a phase given by eqn. 12.13. If the scattering matrix
coefficient between them, n t h element and the 0,0 reference element is S,,, then
the reflection coefficient is
This result shows that the reflected wave at any given element is due to the
mismatch of the isolated element (So,,),plus contributions from all the neigh-
bouring elements. The effect of the coupling from the neighbouring elements
depends on the strength of the coupling, and on the scan angle. For coupling
coefficients of a certain magnitude and phase, it is possible for an in-phase
accumulation of coupled power to lead to total reflection at certain scan angles.
This discussion shows how scan blindness can occur from a mutual-coupling
point of view, but drawing more specific conclusions is difficult unless data on
the coupling coefficients is available, or can be assumed [14].
We now look at a specific printed-array geometry - an infinite planar array
of infinitesimal dipoles - to obtain more information about the scan-blindness
effect. This is an idealised case, of course, but is complete enough to show some
of the essential blindness mechanisms that occur in this array and in other
infinite printed arrays. The preliminary derivations of the Green's function and
the extension to an infinite array of infinitesimal %directed currents of Sections
12.2.1.1 and 12.2.1.2 can be applied directly.
We assume that the rn = n = 0 element of the array is located at x = y = 0,
so that x, = yo = 0 in eqn. 12.22~.We now compute the complex power
leaving the unit cell centered around the 0, 0 element from
where Ex is given by eqn. 12.22a, and J, is the electric-surface-current density
of the sources at z = d
m m
1 1
m--m n=-m
e
-jko(mou+nbv)
6(x - ma)b(y - nb) (12.27)
In eqn. 12.26, the power is evaluated at z = d + , and is the same for each cell
in the array. Substituting eqns. 12.27 and 12.22~
into eqn. 12.26 and performing
703
Analysis and design considerations for phased-array antennas
the integration gives
where G C ( k , , k,) is given by eqn. 12.1la. Recall that k , and k, are functions of
m, n and the scan angle 9, 4, as given by eqn. 12.23.
The result in eqn. 12.28 shows that the complex power leaving the face of the
array is a superposition of the powers contained in each of the Floquet modes
forming the field solution. Now if we assume that the array spacing is such that
a < 112 and b < 112, so that no grating lobes are present for any scan angle,
then from eqn. 12.23 it is easy to see that
lk,l < ko only for m = n = 0
From eqn. 12.3b, these conditions imply that k, is purely imaginary except for
the m = n = 0 Floquet mode, and a study of the G,:! function then leads to the
conclusion that only the rn = n = 0 term of eqn. 12.28 contributes a real part
to P; all the other terms are purely imaginary. Thus the m = n = 0 Floquet
mode is the only mode carrying power away from the array face, and all the
other Floquet modes are evanescent, storing energy near the array surface (or
carrying power across the surface of the array).
Fig. 12.3 shows the complex power P,,, for three Floquet modes (m = - 1,
n = 0; m = n = 0; m = 1, n = 0) versus E-plane scan angle for an infinite
array of infinitesimal dipoles, with a = b = do/2and a substrate with E, = 12.8
and d = 0.061,. The m = n = 0 mode is the only Floquet mode with a
non-zero real part as discussed above. All the modes have imaginary contribu-
tions, which are generally well behaved with scan angle. The exception is the
rn = - 1, n = 0 Floquet mode, which can be seen from Fig. 12.3 to have a
singularity in its imaginary part near a scan angle of about 45'. This results in
a sEan blindness at this angle.
The singularity in Im (P-,,) can be traced to a zero of the Tmfunction in the
denominator of G.2.The zeros of this function correspond to TM surface waves
of the unloaded dielectric slab. As the scan angle approaches the blindness
angle, the propagation constant of the m = - 1, n = 0 Floquet mode ap-
proaches that of the TM surface-wave mode of the dielectric slab, and resonates
this mode. The surface wave propagates along the surface of the array, and so
does not carry power away from the array. In this situation k, is real, meaning
that waves are propagating (up and down) inside the dielectric slab, while k, is
imaginary, meaning that the field above the surface of the array is evanescent.
Thus, such a wave is sometimes called a 'trapped mode'.
This is not unlike the case of total reflection of a plane wave, in a region
having a low dielectric constant, incident on a region having a higher dielectric
constant. For incidence angles greater than the critical angle, all power is
reflected but a surface wave field is excited in the higher dielectric-constant
 704 Analysis and design considerations for phased-array antennas Analysis and design considerations for phased-array antennas
region. This surface wave propagates along the interface, and exponentially pressed as:
decays away from the interface. Since this surface wave field cannot exist in the
absence of the incident plane wave, some authors refer to it as a forced surface
wave, while other authors feel that it is not a 'true' surface wave field at all, but
is only 'surface-wave-like' [13].

where I, is the free-space wavelength.

As an example, for the E, = 12.8, d = 0.061, dielectric slab of Fig. 12.3,
P,,,,/ko = 1.28582. In the E-plane, u = sin0 and v = 0, so for half-wave spac-
ing eqn. 12.29 reduces to
(1.28582)~ = (2m + +
Clearly the only solution occurs for m = - 1 and n = 0; so
O,,,, = sin-'11.28582 - 21 = 45.6'
A useful graphical technique, referred to as a surface-wave circle diagram [8,
161, can be arrived at by noting that eqn. 12.29 describes a set of circles in the
u-v plane. Fig. 12.4 shows such a diagram. The solid-line circles represent the
usual grating-lobe circles, with centres at u = - ml,/a, v = - nI,/b, and unit
I scan angle
/
/-

Fig. 12.3 Complexpowersradiatedby them = - 7 . n = O;m = n = 0;andthem = 7.n = 0

Floquet modes for an infinite array of infinitesimal dipoles, versus E-plane scan
angle
E, = 12.8, d = 0.06&, a = b = & / 2 wave circles

Another point to be noted from Fig. 12.3 is that the real part of the power
(in the m = n = 0 mode) is well-behaved with scan angle, and is finite and
non-zero at the blindness angle. Thus blindness occurs because the imaginary
part of the input impedance becomes extremely large, leading to severe im-
pedance mismatch, even though the real part may be non-zero. Some authors
[IS] have been led to erroneous conclusions in this regard. I \v~sible
space
Scan blindnesses of this type can be predicted by comparing the propagation
constants of the surface wave of the dielectric slab and the various Floquet F i g . 12.4 Surface-wave circle diagram for an infinite phased array with a = b = 0.51,.
modes. Let p,, be the propagation constant of the first (TM) surface-wave mode E, = 12.8, d = 0.061.,

of the unloaded dielectric slab where k,, < P,, < &ko. (In practice, substrates
are usually thin enough so that only the lowest-order TM mode propagates.)
Then a surface wave resonance will occur when /I,matches
, a particular Flo-
quet-mode propagation constant. Mathematically, this condition can be ex-
radius. For half-wave spacing, the edges of these circles just touch, indicating
that no grating lobes are visible. The grating-lobe circle centred at the origin
represents visible space, since, within this region, u and v are such that 0 and 4
706 Analysis and design considerations for phased-array antennas Analysis and design considerations for phased-array antennas 707
correspond to real scan angles. The dotted circles in Fig. 12.4 are solutions to Using the results of Section 12.2.1.2, the moment method can be applied to
eqn. 12.29, and so are called surface-wave circles. When a surface-wave circle a dipole in a single unit cell; by periodicity, all dipoles in the infinite array and
intersects visible space, a scan blindness can occur at those scan angles, unless their mutual interactions are then accounted for. The 2-electric-surface-current
the scan angle is such that a polarisation mismatch occurs between the fields of density on the dipole is expanded in a set of piecewise-sinusoidal (PWS) modes:
the relevant Floquet mode and the surface wave of the slab. This occurs, for
N
example, in the H-plane scan of the infinitesimal dipole array because the
polarisation of the xldirected dipoles cannot couple to ;surface wave propagat-
4x(x, YO) = 4<(%, YO) (12.31)
ing along the y-axis. Mathematically, an inspection of G.:: in eqn. 12.1l a shows
that the T,, function in the denominator is cancelled by an identical term in the
numerator when k, = 0 (H-plane scan). Thus, there is no blind spot at 45.6' in
the H-plane.
The surface-wave circle diagram is a very convenient way to study the effect
of grid spacing and substrate parameters ( E , and d ) on the potential blindness
angle. For example, the diagram shows that it is possible to completely eliminate
a scan blindness by decreasing the element spacing, since this has the effect of
moving the grating-lobe and surface-wave circles further apart. The diagram
also shows that, for half-wave spacings or greater, there will always be a scan
blindness, and that it will occur closest to broadside for E-plane scanning. In
practice, however, for electrically thin substrates 3/, will be close to k,, so the
scan blindness will occur close to endfire. Decreasing the substrate dielectric
constant will also move the blindness angle towards endfire; as E, -+ 1, the
blindness angle approaches 90".
The surface-wave-type blindnesses discussed above seem to occur in any type
of printed array. There are other types of resonances that are also possible,
however, depending on the type of array element being used. Patch elements, for plane
example, may in certain circumstances load the dielectric slab enough to support Fig. 12.5 Geometry of an infinite array of centre-fedprinted dipoles on a grounded dielectric
'leaky wave' modes [15]. substrate

12.2.2 Injnite-planar-array solutions where

12.2.2.1 Printeddipoles: In this Section we consider the analysis of an infinite

planar array of printed dipoles. The dipoles are assumed to be thin, and to be is the i th expansion mode, and
fed with idealised delta-gap generators at their mid-points. Isolated printed sin k,(h - (x, - xil)
dipoles and mutual coupling between pairs of such dipoles have been studied by fp(xo.x,) ='
sin k,h for Ix, - xi/ < h (12.33)
several workers [9,17,18]. Infinite arrays of printed dipoles have been analysed in
References 8, 19 and 20. The solution presented here follows that of Reference 8. is a piecewise-sinusoidal expansion mode with terminals at xi, and a half-length
Fig. 12.5 shows the geometry; the dipoles are of length L and width W. A of h, and
rectangular grid is shown, but a triangular grid can be treated by replacing k,
in the following solution by

is a uniform distribution representing the current variation across the width of

the dipole. An edge condition could be incorporated here, but past experience
has shown that this is generally not worth the trouble. In eqn. 12.32, the wave
where cc is the skew angle of the vertical columns measured from the x-axis
number k, can be arbitrarily chosen; here it is set to k, = k , . / m , which
(cc = 90" then reduces to the rectangular grid).
has been found to give good results.
 708 Analysis and design considerations for phased-array antennas
After the expansion modes have been selected, a Galerkin moment-method
procedure can be applied to the electric-field integral equation that enforces
E,, = 0 over the surface of a dipole in one unit cell. The other boundary
conditions at the ground plane and dielectric-air interface are guaranteed to be
satisfied through the use of the Green's function of eqn. 12.22~.A general
impedance matrix element, representing the coupling of expansion mode j to
weight mode i, is defined as
where E,(x, y, d ) is the field at ( x , y, d) due to a periodic array of infinitesimal
dipoles, as given by eqn. 12.22~.The space integratons in eqn. 12.35 can be
evaluated as the Fourier transforms of the expansion and weighting modes to
give the following:
where F,, and F, are the Fourier transforms of the& and f, functions, as defined
by
Sm
~ ( k , ) = - m f(x)ed'" dx (12.37)
Evaluating eqn. 12.37 for& of eqn. 12.33 a n d f , or eqn. 12.34 gives
2kJcos k,h - cos k,h] -jkx,l
F,.(kx) = sin keh(G - k:)
If there are N (odd) PWS expansion modes used on each dipole, the voltage-
vector elements can be defined as
=
1 for i = (N + 1)/2
0 otherwise
Then in matrix form the expansion coefficients can be found from
The input impedance at any dipole in the array is then
where k = (N + 1)/2 is the index of the mode at the dipole terminals. The
active reflection coefficient is calculated as
so that the array is conjugate-matched to its broadside scan impedance.
Analysis and design considerations for phased-array antennas 709
For typical grid spacings a, b of the order of 112, the series in eqn. 12.36
converges for upper limits of about rn, n = _f 60. This makes computational
efficiency an issue that requires some attention. One thing that helps is to
recognise the Toeplitz-like symmetry in the [Z] matrix. Since the Green's
function of eqn. 12.22~is not reciprocal in terms of interchanging xo,yo with x,
y the [Z] matrix of eqn. 12.36 is not symmetric. But if the N PWS expansion
modes are laid out uniformly on the dipole, other symmetries of the form
Z, = Z ,-,,-, , for 1 < i < j, allow the entire matrix to be filled by computing
only the first row and the first column of the matrix (2N - 1 elements), rather
than all N 2 elements. Additional time savings can be obtained by careful writing
of the computer code. In particular, it is relatively easy to write the program so
that the G,E;/function of eqn. 12.36is computed only once for a new set of k, and
k , values, and not recalculated for each impedance-matrix element. The Fou-
rier-transform functions in eqn. 12.36 can similarly be handled. Such techniques
result in an efficient computer program: a given matrix element takes less than
4 s of CPU time, and the input impedance using three PWS modes can be
calculated for one scan angle in less than 30 s. (These CPU times are for a Micro
VAX/II computer.) The infinite-array solution is considerably faster than a
full-wave solution for a single dipole.
It is interesting to note that the solution for an isolated dipole can be
recovered from the infinite-array solution by an integration over all scan angles
defined by a unit square on the grating-lobe diagram (lul < 1, Ivl < I), for
a = b = 1,/2.
We now present some results for the scan performance of infinite dipole
arrays. Fig. 12.6 shows the magnitude of the reflection coefficient of a dipole
array with grid spacings a = b = 0.51,, on a substrate with d = 0.191, and
E, = 2.55. This substrate is thicker than is usually used in practice, but serves
to clearly illustrate some of the important scan effects. The dipole is matched at
+
broadside scan (0 = 0') with an input impedance of 75 jOR, and curves are
shown for E-plane scan (4 = 0), H-plane scan (4 = 90') and a diagonal
(D-plane) scan (4 = 45'). Note that all curves tend to unity as 0 -* 90, and
that a scan blindness exists in the E-plane at 0 = 45.8'. The substrate supports
a surface wave with fl,,/k, = 1.283. Two solutions to eqn. 12.29 are possible in
the principal planes:
(b) m = 0, n = - 1; u = 0, v = 0.717
Solution (a) leads to the blindness seen in the E-plane at 0 = 45.8'. Solution (b)
would lead to a blindspot in the H-plane at 0 = 454'; but k, = 0, so the TM
surface-wave pole is cancelled, as discussed in Section 12.2.1.3. Other solutions
to eqn. 12.29, however, are possible off the principal planes. Fig. 12.7 shows a
contour plot of the reflection-coefficientmagnitude in the u/v scan plane for this
same ariay. Note the two semi-circular loci of the unity reflection-coefficient
magnitude, with one starting in the E-plane at 0 = 45.8' and leaving visible
space at 4 = 32.7'; and the other entering visible space at $ = 45.8'. The
 710 Analysis and design considerations for phased-array antennas Analysis and design considerations for phased-array antennas 71 1
region of large reflection gets smaller and smaller as the H-plane is approached,
however, until the unity-magnitude region vanishes entirely at the H-plane.
Another way of viewing this effect is with the surface-wave circle diagram shown
in Fig. 12.8. This simply-obtained diagram predicts quite accurately the scan
blindnesses seen in Fig. 12.6 and 12.7. The reader is referred to Reference 8 for
further examples, including the effects of grid spacing, substrate thickness and
dielectric constant, and a triangular grid.

..
scan angle e
a=b=0.5A -
--- E-plane
L=0.39h H-plane
w=o.o02x --- D-plane
d =O. 19X
c,=2 55 1

Fig. 1 2 6 Reflection-coefficient magnitude versus scan angle for an infinite printed dipole
array

I visible
space

Fig. 12.8 Surface-wave circle diagram for the array of Fig. 72.6

The above theory has been experimentally verified by using a waveguide

simulator. Waveguide simulators provide a convenient way to test an element
in an infinite-array environment, and have been used extensively for waveguide .
arrays [4,21]. Here, two printed dipole elements were used in a simulator to test
the above theory, and to provide experimental evidence of scan blindness in
infinite arrays of printed dipoles.
The simulator was made from a piece of C-band waveguide, with inner
dimensions of 2.22cm x 4.75 cm. Two printed monopoles of length 1.02cm
were laid on a glass-fibre epoxy substrate of thickness d = 0.95cm, with a
permittivity 6, = 4.35. The monopoles were fed with SMA coaxial connectors
through the broad wall of the waveguide. The element spacings were
a = 4.43cm and b = 2.38cm. By varying the frequency from about 4.0 to
6.3 GHz, the simulator scanned in the H-plane to an angle given by
u - s i n 8 cos 4 sin 0 = 1/46
Fig. 12.7 Contour plot of the reflection-coefficientmagnitude in the u/v plane for the array
of Fig. 12.6 Over this frequency range, only the TE,,-mode was propagating in the
 772 Analysis and design considerations for phased-array antennas
waveguide. A ground plane backed the glass-fibre substrate, and a standard
waveguide-matched load was used on the array end. The two printed-antenna
elements are fed in phase, with a Wilkinson-type power divider, even though the
array is effectively scanning off broadside. This is because the waveguide mode
corresponds to two propagating plane waves, one at angle 0 and the other at
- 8. For proper simulator operation, the array must generate both of these
waves which, by superposition, results in in-phase feed voltages.
Analysis and design considerations for phased-array antennas 713
the TM, surface wave to the rn = _f 1, n = 0 Floquet mode in the H-plane, as
can be seen from the surface-wave circle diagram for the simulator, shown in
Fig. 12.10. The diagram shows that, because of the large element spacing in the
E-plane and the large diameter of the surface-wave circle, an intersection of the
TM, surface-wave circle occurs for H-plane scan, and does not result in can-
cellation because k, is not zero (since u = 0, but m = I). +

0.4 calculated
x x x measured

........
0
20 c
8
0
4 5 6
frequency, GHz
Fig. 12.10 Surface- wave circle diagram for the dipole array simulator of Fig. 12.9
a = 0.0443 rn b = 0.0238m
L =O.O204m w=0.0013m
d =0.0095rn 6,=4.35 A slightly different solution for the infinite array of printed dipoles has been
H-plane scan
reported in Reference 20. Results from that work, which include several
Fig. 12.9 Measured and calculated active-reflection-coefficient magnitude for a waveguide waveguide simulator measurements, have been favourably compared with those
simulation of an infinite array of printed dipoles
from the present solution. The analysis in Reference 20 uses a singularity
subtraction technique in the space domain to improve computational efficiency,
Fig. 12.9 shows the measured reflection-coefficient magnitudes compared and may be limited to substrates that are not too thin.
with the calculated values. This reflection coefficient is based on the input In Reference 19, a solution is described for a planar array of printed dipoles
impedance of a printed monopole and a 50R system. The agreement is generally with a superstrate (cover layer). The method of analysis is similar to that
quite good; the measured results at 4.0 and 4.2 GHz are somewhat higher than presented above. Also treated in Reference 19 is an infinite layer array of dipoles
calculated owing to some residual mismatch at the power divider. Since the proximity-coupled to microstrip feed lines.
original measurements were made [8],these results have been improved.
The blind spot is clearly seen at 4.76 GHz, with a measured reflection-coef- 12.2.2.2 Rectangular probe-fed patches: We now consider the analysis of an
ficient magnitude of 0.96; it is presumed that a non-unity value resulted because infinite array of probe-fed rectangular microstrip patches; the geometry is
of copper and dielectric loss. The blindness in this case is caused by coupling of shown in Fig. 12.11. This solution is based on the work reported in Reference
714 Analysis and design considerations for phased-array antennas Analysis and design considerations for phased-array antennas 775
22, and assumes an idealised probe feed model. This feed model uses a constant- For the patch problem it is necessary to use both x and y expansion currents;
current filament to mode the probe, and does not attempt to model the rapid so in the interest of clarity and conciseness it becomes useful to define dyads
variation of surface current near the probe-patch junction. Such a model has representing x and y field components as
been found to work quite well for single patches on thin substrates [lo, 18,23,
241, since the patch Q in this case is relatively large, so that the resonant mode
current dominates the total current. Reference 25 shows a plot of the currents
on a probe-fed patch which graphically illustrates this effect.

Thus, consistent with the notation of Section 12.2.1.1, GE represents the$-com-

ponent of the electric field due to a $directed electric-current source. Gfi and
Gf: are given by eqns. 12.1l a and b and G E and G g can be found from eqns.
12.1l a and b by interchanging kx and k,. The results are

G$ =
-jZo ( ~ , k i- k;)k2cos k , d
-
+ jk, (ki - ki) sin k, d sin. k, d
ko T, Tm
(1 2.46~)
jZo k,k, sin k, d[k, cos k, d + jk, sink, d ]
G$' = - (12.46b)
ko Te Tm
Note that G:, = G z . We will also require the E, field due to 2 and 9 currents;
from Section 12.2.1.1 these can be derived as
Z,,k,rk, sin k, d
GLJ =
kok, T,

Fig. 12.11 Geometry of an infinite array of probe-fed rectangular microstrip patches

By reciprocity, G,EJ = - G:, and G$ = - G$. The E, fields represented by

The use of this idealised feed model considerably simplifies the analysis, yet eqns. 12.47~and b have been integrated over z for 0 6 z < d, since this is the
provides useful information on the scanning performance of the array and the form in which they will be used later.
effects of parameters such as substrate thickness and permittivity, and grid The electric-fieldintegral equation, representing the boundary condition that
spacings. While the absolute values of active impedance obtained from this the total tangential electric field must vanish on the patch conductor, can be
solution may only be valid for substrate thicknesses on the order of 0.0212, or written as
less, the active reflection coefficient, being a normalised quantity, has a greater
range of validity. In addition, comparisons of patch arrays and printed-dipole
arrays with the same substrate parameters and grid spacings show significant
similarity in terms of reflection coefficient versus scan angle, suggesting that the where E;:; is the tangential ( & j ) component of the incident electric field due to
element type or feed arrangement is not a dominant factor in the scan perfor- the probe source evaluated at z = d, J, is the total vector electric surface-current
mance or active-element patterns. This idea can be pursued further with the density on the patch (the sum of the currents on the top and bottom patch
current-sheet model discussed in Section 12.2.2.5. surfaces), S is the patch surface, and E:;' is the tangential field scattered by the
The moment-method theory of the analysis of the probe-fed patch array will patch. The surface-current density J i s now expanded in a set of basis functions
be developed below, as an extension of the dipole-array solution of Section
12.2.2.1. Then several calculated results will be presented for the scan perfor-
mance of infinite patch arrays, followed by some measurements from a
waveguide simulator.
716 Analysis and design considerations for phased-array antennas Analysis and design considerations for phased-array antennas 777
where J, is an expansion mode representing current flow in either the 2 or $ where a voltage-vector element based on expansion mode j has been defined as
direction, and I, is the unknown coefficient. Substitution of eqn. 12.49 into eqn.
12.48, multiplication by a weighting function J, and integration overs yields, for
i = I, 2, 3 . . . N ,

In words, V,' represents a voltage based on the field from the probe integrated
We can now define an impedance matrix element as over a surface-test mode, while represents a voltage based on the field from
a surface-expansion mode integrated over the probe. Thus # VV,: in general.
The above solution may be described as a Galerkin solution in the spectral
and a voltage-vector element as domain, since the matrix elements are expressed in terms of the Fourier trans-
forms of the fields and currents. The probe self-reactance has been ignored here.
The next step is to choose the expansion/weighting functions. Because of their
correspondence with the cavity model, entire domain modes of the following
where the superscript t indicates that this voltage-vector term is based on a test, form were used:
or weighting, mode. These matrix elements can be written, using eqn. 12.45, as
kx Ix
4(x, y) = f sin - ( x
L
+ L/2) cos -
W (y + W/2) (12.60~)

for f currents, and

kx In
J(x, y) = j c o s - (x
L
+ L/2) sin -
W ( y + W/2) (12.606)
x $+P &P (12.54)
where x,, y, are the co-ordinates of the feed probe, and Fi represents the Fourier for 9 currents, where k and I are integer indices accounting for the number of
transform of the Ji expansion mode, defined as variations in the x and y directions, respectively. The Fourier transforms of
F, (k, , k,) = IsJi (xo, yo) e-jkx* e-jkyyodxOdyO (12.55)
these modes can be easily calculated analytically from eqn. 12.55.
Through numerical convergence checks it was found that, for E-plane scan-
(Note: Because this definition of the Fourier transform differs from that of ning, the- (k, I) = (I, O), (3, O), (5, O), (7, 0) 2-directed currents and the (k,
Reference 22 in the signs of the exponential terms, some of the above results I) = (0,2) $-directed current gave a fairly stable solution, while for off E-plane
differ from those of Reference 22 by conjugation.) scan the (0, 1) $-directed current mode should also be included. The edge
The unknown expansion coefficients 4 can then be found as solutions to the condition could be incorporated into the above currents, but it has been found
following set of linear equations: that this is an unnecessary expense in a Galerkin-type solution for patches 110,
231.
Z,$ = V:, for all i (12.56) Fig. 12.12 shows the reflection-coefficient magnitude versus scan angle for an
J
infinite array of probe-fed microstrip patches on a substrate with E, = 2.55 and
Note that [ Z ] is not a symmetric matrix. The input impedance at the probe can d = 0.061,, but the E-plane spacing is a = 0.511, which leads to a grating lobe
be calculated as at 0 = 73". The blindspot position in the E-plane then occurs at 68.8". Because
of the presence of 9-directed currents, the H-plane also shows a blind angle, at
0 = 76.4"; this blindness has a much higher Q than the E-plane blindness since
these J-directed currents are highly reactive and radiate little power. In practice,
where I, is the current on the probe which, if d < &, is assumed to be uniform any loss, probe radiation, or random-error effects would probably 'wash-out'
along the probe, and will be chosen as 1 A. Then from eqn. 12.57, 12.49, 12.48 this spurious H-plane blindness.
and 12.45 we can write Fig. 12.13 shows the behaviour of an infinite patch array on a thin high-dielec-
tric-constant substrate having d = 0.021, and E, = 12.8. Because of the thin-
ness of the substrate, the blind spots in the E- and H-planes now occur at
 778 Analysis and design considerations for phased-array antennas
Analysis and design considerations for phased-array antennas 779
9 = 82.9". This result shows that, even though a scan blindness may be present
in the visible range of an array, it may have little effect on the scan range if it
is sufficiently close to endfire. The reader is referred to Reference 22 for further
examples.

x x x measured
-
0.6

scan angle
6,=2.55 - E-plane
d--0.06ho ---
H-plane
a -0.51h0 ---
D-plane
b = 0.50h0
L=0.28X0
W=0.30ho
Xp= L/2 Zb=86n frequency, GHz
4, = 2.33 L = 1.8cm
Fig. 12.12 Reflection-coefficientmagnitude versus scan angle for an infinite array of rectan- d=0.159cm W = 1.8cm
gular microstrip patches. d = 0.06A0, 8, = 2.55 a= 2.22cm Xp= 0.48cm

Fig. 12.14 Measured and calculated reflection-coefficientmagnitude of a microstrip array in

a waveguide simulator. The centre patch is fed

The above theory has been verified with several waveguide simulator meas-
urements, and a typical example is shown in Fig. 12.14. Unlike the dipole
simulator described in Section 12.2.2.1, this simulator uses only one fed element
centered in the guide, with two half-patches at the sides. These half-patches
carry zero feed current. Good agreement with theory is obtained, although it
should be pointed out that the data of Fig. 12.14 is dominated by the high-Q
resonance of the patch, and differs only slightly from the frequency-dependent
behaviour of the isolated patch element. It should also be noted that, although
scan angle the patches image correctly across the waveguide simulator walls, the feed
c,=12.8 -E-plane probes do not. This is felt to be a negligible factor for electrically thin substrates.
d= 0.02Xo ---
H-plane
---
D-plane Another solution for the probe-fed patch array has been reported in Re-
a=b= 0.5A0
L= 0.131h0 ference 15. This work uses a singularity subtraction technique for the patch
W= 0.15ho current near the feed point, and so overcomes the main drawback of the solution
xp= L/2 Zb=403n
in Reference 22, which is limited to thin substrates because of the idealised feed
Fig. 12.13 Reflection-coefficientmagnitude versus scan angle for an infinite array of rectan-
gular microstrip patches. d = 0.02A0,E, = 12.8
model. Reference 15 presents some useful design data for patch arrays, and
interprets the behaviour of such arrays in terms of surface-wave and leaky-wave
 720 Analysis and design considerations for phased-array antennas
effects. It erroneously presumes, however, that the solution of Reference 22 is
missing leaky-wave effects, which come about because of the patches loading the
dielectric slab. The solution of Reference 22 is full-wave and accounts for the
presence of the patches on the dielectric slab through the moment-method
procedure, and comparisons with some of the results in Reference 15 show the
same effects which are labeled as 'leaky waves' in Reference 15. A recent paper
[47] describes experimental confirmation of surface-wave scan blindnesses in
large arrays of microstrip patches, in agreement with the theory of Reference 22.
12.2.2.3 Circular probe-fedpatches: Like the rectangular patch, the circular
microstrip antenna is often used as an array element, and so we describe here
the extension of the previous analysis to an infinite array of circular patches.
This case clearly illustrates the versatility of the Green's-function/moment
method (or 'Galerkin's method in the spectral domain,' as some authors have
referred to it) which has been presented in the preceding Sections. As we will see,
the only major change needed to treat circular patches is to use the appropriate
expansion modes and Fourier transforms of those modes [26].
""""""/yg;-/"/nT "
plane
Fig. 12.15 Geometry of an infinite array of circular probe-fed microstrb patches
The geometry of the infinite circular patch array is shown in Fig. 12.15. A
rectangular grid is assumed, but a triangular grid can be easily treated. As in the
rectangular-patch case, the probe-feed model here is also idealised in that it does
not attempt to mode the singularity in patch current near the probe, thus
limiting the solution to thin substrates. The arguments presented in Section
12.2.2.2, concerning the utility and justification for such an approximation, then
apply here, as well.
Analysis and design considerations for phased-array antennas 721
The analysis for the infinite array of circular patches follows identically the
solution for the rectangular-patch case of Section 12.2.2.2, until the expansion
modes are chosen. Analogous to the rectangular-patch case, we look to the
cavity model and select the TM,, circular waveguide modes as expansion
functions. If we assume the feed point of the reference patch lies on the 4 = 0
line, the i th expansion mode can be written in cylindrical co-ordinates as
R= 0.079h0 - E-plone
c=0.077ho ....... D-plane
a=b=h0/2 H-plane
linear polarized
--
scan ongle
Fig. 12.16 Reflection-coefficientmagnitude versus scan angle for an infinite array of probe-
fed c~rcularmicrostrip patches
d = 0.021,. E, = 12.8. Single-probe feed.
where PpqRis the q th zero of Jb(x), where Jp(x) is the Bessel function of order
p, and R is the radius of the circular patch elements. In eqn. 12.61, the single
index i is used to form a one-dimensional sequence of the TM, modes. To apply
eqns. 12.51-12.59, the Fourier transform of the above expansion modes is
needed. These expressions can be derived in closed-form [26], but are too lengthy
to list here.
Fig. 12.16 shows a typical result for the scan performance of an infinite
circular-patch array. The grid spacing (a = b = &/2) and substrate parameters
(d = 0.02&, 6, = 12.8) are the same as the rectangular-patch case shown in
722 Analysis and design considerations for phased-array antennas
Fig. 12.13, and it is interesting to observe that the results are practically
identical. Only in the diagonal plane is there much difference between the
rectangular- and circular-patch results.
Multiple-probe-fed patches are also of practical interest. Fig. 12.17 shows
three common feeding circuits for circular patches, and these techniques are also
relevant for rectangular (or square, for circular polarisation) microstrip patches.
Fig. 12.17a shows the single probe-fed patch which has already been treated.
The two-probe case of Fig. 12.176 is fed with a 180' hybrid, which reduces the
amount of cross-polarised radiation. In Fig. 12.17~the two feed probes are in
orthogonal planes and are fed with a quadrature hybrid to generate circular
polarisation. The theory which has been presented above can easily be extended
to handle the two-probe feed cases of Figs. 12.176 and c.
Fig. 12.17 Three common feed circuits for patch antennas
a Single probe feed
b Balanced 180' hybrid feed for reduced cross-polarisation
c Quadrature hybrid feed for circular polarisation
A given element in the infinite-array environment can be treated as a two-port
network, with an open-circuit 'port' impedance matrix of the form
Analysis and design considerations for phased-array antennas 723
where the matrix elements can be determined from
which gives the open-circuit voltage induced on the k th feed probe due to the
fields excited by a unit current on the Ith feed probe. After the matrix of eqn.
12.62 is found (which is a generalisation of the single-port input impedance of
eqn. 12.57), an equivalent circuit which models the feed network of either Fig.
12.176 or c can be used to find the reflection coefficients R, and R,,seen looking
into the antenna element ports. If the hybrids have isolation, then, in general,
some of the reflected power from the antenna element will be dissipated in the
hybrid and some will pass back through the hybrid. Thus the reflection coef-
ficient R at the input of the hybrid does not account for this lost (non-radiated)
power. A better indication is to plot the active-element gain pattern, including
the efficiency of the feed network, defined as
where q is the feed-network efficiency:
and cc is the phase angle between the two feed ports (either 90" or 180"). Since
R,and R2in eqn. 12.65 vary with scan angle, q also varies with scan angle.
Fig. 12.18 shows such an active-element gain pattern for a circularly polarised
circular patch array. The element and substrate geometry is the same as that of
Fig. 12.16. The resulting axial ratio is shown in Fig. 12.19. Observe that, while
the single-probe-fed array of Fig. 12.16 shows a scan blindness of about 6 = 83"
in the E-plane, the corresponding circularly polarised array with the two feed
probes per element of Fig. 12.18 does not show a blindness at this angle. This
is because the reflection coefficient at the feed probe which drives E-plane
currents (probe at 4 = 0) may have a unity reflection-coefficient magnitude at
0 = 83" in the E-plane, but power can still be delivered to the cross-polarised
currents fed by the other feed probe (at 4 = 90"). This polarisation, being
H-plane directed, is decoupled from the E-plane surface wave. The axial ratio,
however, becomes infinite at 6 = 83O, as shown in Fig. 12.18. Because of
symmetry, this argument applies to both E- and H-plane scan for the circularly
polarised circular-patch array.
The above results for circular-patch arrays are preliminary; theoretical work
on this problem and experimental verifications are continuing.
12.2.2.4 Aperture-coupled patches: The next type of printed array to be
considered is one using aperture-coupled rectangular microstrip patches. The
aperture-coupled patch element [27] consists of two substrates, with a ground
plane in between. As shown in the geometry for a single aperture-coupled patch
in Fig. 12.20, a microstrip feed line is printed on the bottom (feed) substrate,
 724 Analysis and design considerations for phased-array antennas Analysis and design considerations for phased-array antennas 725
while the patch element is printed on the top (antenna) substrate. Coupling
between the feed line and the radiating element is through a small slot in the
ground plane below the patch. As will be discussed further in Section 12.3.2, this
type of element has a number of attractive features when used in a phased-array
configuration.

antenna
patch substrate
antenna

circularly
polarized

0.21 , I , I

OO 20 40 60 80
scan angle
Fig. 12.18 Active elementgain for an infinite array ofprobe-fed circularpatches. d = 0,02L,,
E, = 72.8
Two-probe feed with a quadrature hybrid for circular polarisation

Fig. 12.20 Geometry of an aperture-coupled microstrip patch antenna element

The single aperture-coupled patch has been theoretically analysed in Referen-

scan angle (deg)
Fig. 12.19 Axial ratio versus scan angle for the circularly polarised array of Fig. 12.18
ces 28 and 29, using Green's-function/moment-methodprocedures. In Re-
ference 28, the currents on the feed line were expanded in terms of travelling-
wave and piecewise sinusoidal expansion modes, following the method of
Reference 30. Since this method requires modelling of the feed line over at least
several wavelengths, it cannot be applied directly to an infinite phased array
without the feed lines running into neighbouring patches or other feed lines. In
contrast, the method of Reference 29 first treats the coupling of the aperture
fields to the patch, then relates this to an equivalent series impedance seen by
the microstrip feed line. This makes the method of Reference 29 a viable
approach for treating the infinite array of aperture-coupled patches, since the
feed-line interaction is avoided. The solution thus accounts for mutual coupling
between the patches, and between the slots, but assumes the feed lines do not
 726 Analysis and design considerations for phased-array antennas Analysis and design considerations for phased-array antennas 727
couple to each other. This should be a safe assumption for practical microstrip z = 0. The individual components are [29]:
networks.
Aperture coupling is a more complicated way of feeding the patch, as com-
pared to probe feeds or edge feeds (via microstrip line), but it is interesting to
note that it is easier to do the analysis of the former case in a more rigorous -jk,k,(~, - I) sink, d
manner than the latter. This is because the aperture-coupled patch is proximity Gg' =
T, T m
coupled, without a direct contact, eliminating the difficulty of the patch-current
singularity that occurs with probe or edge feeds. In eqn. 12.69 s represents the patch surface.
The present analysis follows that of Reference 29, as extended to the infinite Now an aperture admittance Yp can be defined as the reaction of the aperture
array. The f and $ surface currents J, are expanded in a set of entire domain- field and the field scattered by the patch:
basis functions, as in eqns. 12.49 and 1 2 . 6 0 ~and b. The electric-field integral
equation of eqn. 12.48 is then applied to the patch surface, where E"' is the field
radiated by the equivalent magnetic current where H, is the magnetic field at the aperture due to the currents on the patch,
S, represents the aperture surface, and the elements of the voltage vector [VJ]
defined as
where e",xo, yo) is the unknown aperture electric field. Since the coupling
aperture is electrically small, a good approximation is to model the field distri-
bution with a single PWS mode:

where G y = GEB + G r $ is a Green's function representing the Hy field at

where S, denotes a uniform distribution of field across the width (narrow z = 0 due to an 9 or $-directed infinitesimal electric dipole at z = d. The
dimension) of the coupling aperture, as defined in eqn. 12.34, and& denotes a individual components are [29]:
PWS distribution across the length (long dimension) of the aperture, as defined
ineqn. 12.33. If thereis some compelling reason to do so, additional PWS modes
could be included, as discussed in Reference 29.
The electric-field integral equation 12.48 then can be reduced to matrix form
as,

The coupling aperture also has a self-admittance Y caused by the direct

where [I] is the column vector of unknown patch expansion mode coefficients, radiation of the aperture on either side of the ground plane. For an assumed
[Z] is the impedance matrix of the patch with elements given by eqn. 12.53, and aperture field of the form of eqn. 12.67, we have
[ V M ] is the voltage vector due to an excitation of magnetic current in the
aperture. The elements of [VM]are

where ~i~~ refers to the Green's function representing the Hy field radiated on
either the top (+) side or the bottom (-) side of the ground plane, due to a
where &, F,, and F, are previously defined Fourier transforms, and $-directed magnetic current in the aperture. These two terms are similar in form,
+
G,EM= 2 ~ ; +G&' ~ +
is a Green's function representing the iand electric but may have different values if different substrates are used on the two sides of
field at z = d due to a $-directed infinitesimal magnetic current element at the ground plane. From Reference 29, the contribution from one side of the
 728 Analysis and design considerations for phased-array antennas Analysis and design considerations for phased-array antennas 729
ground plane is given as The former has a resonance with an infinite susceptance at 6 = 86", while the
(E,% - $)kt cos k l d + jk2e,sink1d) - latter has a resonance with an infinite susceptance at about B = 62". Thus,
G y = - surface-wave resonances are possible on both the feed and the antenna sub-
kt T m strate, but these blind spots can be moved closer to endfire by making the
substrates thinner.
Then, as derived in Reference 29, the slot-coupled patch antenna appears to the equivalent
microstrip feed line as a series impedance Z, where Z is given by feed series
line/ =
impedance =

=c =c
where Yp and Ys are given by eqns. 12.71 and 12.74, 2,is the characteristic
impedance of the feed line, and Av is a modal voltage due to the discontinuity
of the slot. From Reference 29, Av is given by
Fig. 12.21 Equivalent circuit of a stub-tuned aperture-coupled microstrip patch element

where hy(x,y) is the normalised magnetic field of the quasi-TEM microstrip-line

mode [29]:

where pm is the propagation constant of the microstrip line. The equivalent

circuit is shown in Fig. 12.21. A tuning stub is generally used to terminate the
feed line and to adjust the impedance match of the antenna; this is easily treated
via the equivalent circuit of Fig. 12.21.
Figs. 12.22 and 12.23 show results for an aperture-coupled array geometry
with an antenna substrate having E, = 2.55 and d = 0.022,, and a feed sub- 0
strate having E, = 12.8 and d = 0.0520. The feed substrate was intentionally 0 30 60 90
scan angle
made thicker than usual to show surface-wave resonances; a thinner substrate fr,=2.55 frf=12.8
would move the feed-substrate resonance closer to endfire. Fig. 12.22 shows the d ,=0.02h0 df =0.05h0
PL =0.279ho SL = 0.115Xo
reflection-coefficient magnitude versus scan angle, where surface wave blind PW=0.279ho SW= 0.01ho
spots are seen to occur at B = 62" and at 86" in the E-plane. The former is due a=b=0.5Xo Wf =O.OS
to surface-wave excitation on the feed substrate (from the coupling slots), while L, = 0.075h0
the latter is due to surface-wave excitation on the antenna substrate (from the Fig. 12.22 Reflection-coefficient magnitude of an infinite array of aperture-coupledpatches
slots and the patches). The individual contributions of the patch and slot to this
phenomenon are shown in Fig. 12.23, where the real and imaginary components
of the patch (YP) and slot (YS)admittances as seen by the microstrip feed line The results from this solution, which are seen to be qualitatively similar to
at the coupling slot are plotted against E-plane scan angle. The patch admit- those of the probe-fed patch solutions of Sections 12.2.2.2 and 12.2.2.3, lend
+
tance, YP = GP jBP, is seen to have a resonance with a near-zero real part a t credibility to the feed-model approximations used in those solutions.
about 9 = 85". The slot looks into both the antenna substrate and the feed
substrate, and so the slot admittance Y* has been separated into an antenna 12.2.2.5 Other geometries: We have analysed several of the most popular
substrate component, C" + jB", and a feed substrate component, Gsf + jBSf. types of printed phased arrays in the preceding Sections, but some other results
 730 Analysis and design considerations for phased-array antennas
could not be presented here because of space limitations. We will briefly discuss
these results, and refer the interested reader to the literature for more details.
After studying a number of different phased-array geometries of dipoles and
patches, it becomes evident that many of the dominant characteristics of printed
phased arrays are controlled by the element spacing and substrate parameters,
1 a study was the possible integration of array elements and circuitry at the
1
I consideration for general applications.
scan angle
Fig. 12.23 Active slot admittances of the patch and slot elements for the array of Fig. 12.22
as opposed to the specific type of radiating element or feeding technique. This
observation led to a current-sheet model of a printed phased array [31], based
on an extension of some early work by Wheeler [32]. The purpose of this model
was not to generate a solution or detailed data that was specific for any printed
Analysis and design considerations for phased-array antennas 731
phased array, but to develop a simple analysis that could predict the major
trends in the scanning performance of a general printed phased array. Such a
model can enhance our understanding of the operation of such antennas. A
variety of results, given in Reference 31, show how the relatively simple current-
sheet model can be used to predict many of the characteristics, such a reflection-
coefficient variation with scan angle and scan blindnesses, of various types of
phased arrays.
Another type of phased-array geometry is analysed in Reference 33, where
elements are printed on a semi-infinite substrate (half-space). Both dipole and
slot elements were considered, and the analysis employed a Green's-function/
moment-method procedure similar to the above work. The motivation for such
surface of an electrically large dielectric lens. Several problems with this con-
figuration, including poor scan performance, have precluded it from further
12.2.3 Finite-array solutions
Of necessity, all practical phased arrays are finite in size, and so it is important
to determine the efficacy of the infinite-array assumption. If it is very large, the
central elements of a finite array are generally modelled quite well by the
infinite-array approximation. The infinite-array solution, however, does not
account for edge effects in a finite array, and it is generally not known a priori
how 'big' a finite array has to be before it can be reasonably modelled as infinite.
Thus, the analysis of finite arrays may be of more practical utility than either
the analysis of isolated elements (no mutual coupling), or infinite arrays (no
edge effects).
Another reason for considering finite arrays of printed antennas concerns the
role of surface waves. It has been shown [6, 9, 12, 181 that a single printed
antenna element can convert a significant fraction of its input power into surface
waves, as opposed to radiated power. On the other hand, surface waves cannot
exist on infinite phased arrays except at blindness angles, where all input power
is converted to surface-wave power, and no radiation leaves the surface of the,
array. The question then arises as to the effect of array size on the generation
of surface-wave power. Does a finite array of printed dipoles, for example, excite
more or less surface-wave power than a single printed dipole on the same
substrate? And how does this power vary with array size, and scan angle? The
analysis of finite printed arrays provides answers to these questions, and clearly
shows the relation of surface-wave excitation to the scan-blindness
phenomenon.
Finite arrays, however, are considerably more difficult to analyse than either
single elements or infinite arrays. Using the 'element-by-element' approach [3],
the mutual coupling between each pair of elements in the array must be cal-
culated, and matrices of order equal to the number of elements in the array (or
larger, if there is more than one expansion mode per element) must be inverted.
 732 Analysis and design considerations for phased-array antennas
The size of arrays that can be handled by this method is thus quite limited. It
is important to realise that, even though mutual coupling is calculated between
pairs of open-circuited elements, the complete solution includes the effect of
terminations, and is completely rigorous in the moment-method sense.
In the following Section we present the solution for a finite array of printed
dipoles [35]. The key step in this analysis is the efficient and accurate calculation
of mutual coupling between pairs of dipoles, which is carried out with a
moment-method procedure using the Green's function results of Section
12.2.1.1. Quantities such as the active input impedance, reflection coefficient and
element patterns can then be calculated. Section 12.2.3.2gives a brief discussion
of the analysis of finite probe-fed rectangular patch arrays [36], and presents
calculated and measured results for mutual coupling between microstrip patch-
es, and some active element patterns.
Besides the element-by-element method used here, a techique called the 'finite
periodic structure approach' [34] has recently been develbped, and appears
capable of treating large arrays. It is based on a modification of the infinite-
array solution, and is similar to a technique that has been applied to finite
waveguide arrays [4].
Analysis and design considerations for phased-array antennas 733
given by eqn. 12.39. Note that reciprocity is satisfied, so that Z,, = Z,,,. The
efficient numerical evaluation of eqn. 12.80 is discussed in Reference 23.
Now consider an N x N planar array of printed dipoles with M PWS
expansion modes on each dipole. Then the order of the linear system of equa-
tions in eqn. 12.79 is N x N x M. Thus, for example, an I1 x l l square
dipole array with three expansion modes per dipole requires an impedance
matrix of size 363 x 363, and the order increases as the square of N. It is
therefore very important to minimize the number of basis functions used. The
PWS mode of eqn. 12.32, with a wave number given by k, = k , , / m ,
was found to give quite good results for resonant dipoles, even when only one
mode was used [8, 18, 341. So, for the majority of calculations in Reference 35,
and the results presented here, one PWS mode was used per dipole ( M = 1).

12.2.3.1 Printed dipoles: Fig. 12.24 shows the geometry of a finite array of
printed dipoles. Each dipole is assumed to have a length L, a width W, and to
be uniformly spaced from its neighbours by distances a in the x-direction and
b in the y-direction. The solution can treat rectangular arrays of arbitrary size,
but in the interest of simplicity only square arrays are considered here. The
dipoles are assumed to be thin, so only ?-directed currents are used. The
appropriate Green's function is then given by eqn. 12.10~.The current on the
dipoles is expanded in a set of piecewise sinusoidal (PWS) modes, as defined in
eqns. 12.32 and 12.33. The dipoles are assumed to be centre-fed with idealised
Fig. 12.24 Geometry of an N x Nplanar array of dipolesprinted on a grounded dielectric slab
delta-gap generators with series impedance 2,. The equivalent circuit of the fed
is shown in Fig. 12.25. Using a Galerkin procedure the electric-field integral
equation reduces to

where [Z] is the impedance matrix representing the mutual coupling between all
the PWS modes on the dipoles, [Z,] is the generator terminating impedance
matrix (a diagonal matrix), [ I ] is the unknown vector of expansion-mode
Fig. 12.25 Equivalent circuit of each dipole in the finite array
coefficients, and [V] is the excitation vector of generator voltages. The rnn th
element of the impedance matrix is given by
The issue of a complete modal expansion is an important one. Discussions of
current expansions for free-space dipole arrays can be found in Reference 3 and
37. In Reference 37 it was found that for arrays of thin dipoles near resonance,
not spaced too closely together, the dipole currents were practically identical.
where Fpmis the Fourier transform of the rn th PWS expansion mode given by For printed dipoles, the situation should then be even better, because the
eqn. 12.38, and F, is the Fourier transform of the uniform y-variation of current printed-dipole resonance has a much higher Q than the dipole in free space. A
734 Analysis and design considerations for phased-array antennas
judicious choice of a single basis function can thus give a very good approxima-
tion to the true current.
The best justification for a single-mode approximation, however, is a com-
parison with results computed using more than one expansion mode per dipole.
Thus, for the example shown in Figs. 12.26-12.28, the input impedance, reflec-
tion coefficient and radiation efficiency were computed against scan angle using
one and three PWS modes per dipole, for array sizes up to 9 x 9. The input
impedance differed by about lo%, and the reflection-coefficient magnitude and
radiation efficiency (which are normalised quantities) differed by less than 5%.
It should be emphasised that the solution presented above is capable of
handling any number of PWS expansion modes per dipole, and that it is desired
to use only one mode per dipole in order to analyse larger arrays. The presence
of all dipoles in the array and their mutual coupling is accounted for in the
solution. In addition, the solution can handle both the 'forced excitation'
(Z, = 0) case, as well as the 'free excitation' (Z, # 0) case.
If one expansion mode per dipole is used, the voltage-vector elements can be
written as
where x,, y,,, are the co-ordinates of the centre of the m th dipole and u, v are
direction cosines for scanning, as given by eqn. 12.14. Then after the matrix eqn.
12.79 is solved for the currents, the input impedance at the nth dipole can be
computed as
Note that the input impedance at a dipole of the finite array is dependent on the
location of that dipole, as opposed to the infinite-array case where the input
impedance would be the same for all dipoles. The active reflection coefficients
at the n t h dipole can then be calculated according to eqn. 12.43.
A quantity of interest for the finite-array case is the radiation efficiency e
based on the power lost to surface waves:
where e,,is the total input power to the array, and P,,is the surface-wave power
excited by the array. These quantities can be calculated as [I81
where Z,, is given by eqn. 12.80, and Z z is the surface-wave contribution (from
the residue of the surface-wave pole or poles) to the impedance Z,,. The
Analysis and design considerations for phased-array antennas 735
active-element pattern can also be calculated, as discussed in Reference 35; this
Reference also discusses some useful techniques for improving the computation-
al efficiency of the finite-array solution.
E-plane
- Inf. array
DO* 19x19
H - plane
---- Inf.array
a 0 0 19x19
I
0 w
theto
Fig. 12.26 Reflection-coefficient magnitude versus scan angle (E- and la lane) for a finite
(19 x 19 centre element) printed dipole array and an infinite array
E, = 2.55,d = 0.19, a = b = 0.51,. L = 0.39&, W = 0.01 A,
Fig. 12.26 shows the reflection coefficient magnitude of a 19 x 19 printed
dipole array on an 6, = 2.55 substrate, compared with the result for an infinite
array of similar dipoles. The reflection coefficient of the finite array is computed
at the centre element of the array, and is matched at broadside scan. Note that
the 19 x 19 array is sufficiently large that its reflection-coefficient magnitude
versus scan angle follows that of the infinite array relatively closely. This array
shows a scan blindness a t 0 = 45.8" in the E-plane. At this scan angle, the
reflection-coefficient magnitude of the infinite array is unity, but that of the
centre element of the finite array is actually greater than unity. This means that
the centre dipole is delivering power back to its generator and load. This power,
of course, is being transferred from other ports, and does not violate any
conservation laws. The input impedance across the finite array is thus non-
uniform. Fig. 12.27 illustrates this variation, showing the reflection-coefficient
magnitude as a function of element position across the E-plane (x-direction) of
the 19 x 19 array of Fig. 12.26, for various scan angles. The 0 = 0 (broadside)
case is symmetrical about the centre of the array, and the data shows that the
centre element (no. 10) is perfectly matched, but that other elements are slightly
mismatched. For 0 = 30" (scanning to the right of the Figure), the mismatch
is greater and is asymmetrical. The 0 = 45" data shows that a number of dipole
ports on the right-hand side of the array have reflection-coefficient magnitudes
greater than unity; it appears from the data that the left-hand elements are
736 Analysis and design considerations for phased-array antennas
absorbing power from the generators and delivering it to the right-hand ele-
ments.
Fig. 12.28 shows the efficiency of this array, based on power lost to surface
waves as defined in eqn. 12.83, versus E-plane scan angle, for various array sizes.
This is a particularly interesting result because it shows the role of surface waves
in the transition from a single-element printed antenna, to a finite array, and to
an infinite array.
1 5 10 15
element pos~tion
Fig. 12.27 Reflection-coefficient magnitude versus element position across the E-plane of
the 19 x 19 finite array of Fig. 12.26
For a single dipole (1 x l), about 22% of the input power is converted to
surface-wave power (with the remainder going into space-wave radiation); and
this ratio, of course, does not vary with scan angle. For arrays, however, a
significant variation of efficiency occurs with scanning. The general trend is that
the efficiency improves rapidly for even modest-sized arrays, and increases with
array size at all scan angles except those near 4 5 P , at which angle the efficiency
decreases (more surface-wave power) with increasing array size. This is precisely
the angle at which the infinite array has a scan blindness. If the efficiency of the
infinite array were plotted in Fig. 12.28, it would be unity at all scan angles
except 45.8', where it would be zero. Since there is no scan blindness in the
H-plane of the infinite array, the efficiency of finite arrays for H-plane scan is
near unity.
This effect can be explained as the destructive or constructive interference of
the surface wave of the unloaded dielectric slab with the radiation of the array.
As the array becomes larger, the periodicity and phasing of the array tend to
cancel the surface wave at all scan angles except at the scan-blindness angle. At
Analysis and design considerations for phased-array antennas 737
this angle, the periodicity and phasing of the array are such as to reinforce, or
resonate, the surface wave.
As a practical matter, the data of Fig. 12.28 show that the scan-blindness
phenomenon can be a problem for even relatively small arrays, and that a
prudent array design should probably limit the maximum scan range to about
10" less than the blindness angle.
Reference 35 shows the active-element patterns for the array of Fig. 12.26, as
well as examples of other arrays.
Fig. 12.28 Radiation efficiency (power loss to surface waves) of the finite dipole array of Fig.
12.26, versus E-plane scan angle for various array sizes
12.2.3.2 Microstrip patches: The above analysis for finite arrays of printed
dipoles can be readily extended to finite arrays of rectangular microstrip patch-
es, as reported in Reference 36. This solution uses the idealised probe-feed
model discussed in Section 12.2.2.2 and has been verified by mutual coupling
and active-element patterns for patches on thin substrates [23, 24, 361.
The geometry of the finite patch array is shown in Fig. 12.29. As in Section
12.2.2.2, PWS expansion modes are also used here, and impedance matrix
elements can be defined as in eqn. 12.80. It must be realised, however, that the
mutual impedance defined by eqn. 12.80 are not the same as those seen at the
inputs to the probe feeds of the patches. That is, unlike the dipole case, we must
make a distinction between the moment-method impedance matrix [Z] and the
'port' impedance matrix [ZP]defined at the probe terminals. The current flow
on the patch is related to the voltage excited at the probe terminals by the modal
voltage V, [23]:
where G$ is given by eqn. 12.47~.
738 Analysis and design considerations for phased-array antennas
The computation of Zm,,and V, as given in eqns. 12'30 and 12.85 constitutes the
bulk of the computational effort for the finite patch-array solution, and so it is
important that these terms be evaluated in an efficient manner. References 23
and 36 discuss this issue. In addition, it is possible and desirable to use only one
x-directed expansion mode on each patch. This allows a smaller matrix size to
be used for a given array, and the arguments presented in Section 12.2.2.1 for
the one-mode approximation can be used here as well. Reference 36 shows a
result for the reflection-coefficient magnitude versus scan angle for a 7 x 7
patch array, computed using one and three expansion modes. The results are in
good agreement, except for about 10% error in the H-plane scan near endfire.
Thus, although the solution can accommodate more than one expansion mode
per patch, it appears that in many cases this is not necessary, which then allows
the treatment of larger arrays.
Analysis and design considerations for phased-array antennas 739
The active reflection coefficient can then be calculated from eqn. 12.43, and a
radiation efficiency calculated as in the dipole-array case.
To talk about mutual coupling between the probe feed 'ports' of the array,
we must define a port impedance matrix [ZP] as
[V" = [ZP][Ip] (12.89)
where [ZP] is found from
Scattering matrix elements can then be calculated directly from [ZP]:
where [Z,] is a diagonal matrix with elements Z,, the characteristic impedance
of the connecting transmission lines.
The active-element pattern can be calculated as follows. From eqns. 12.87,
12.89 and 12.90, the patch currents [I] due to a set of port voltages [VP]can be
calculated as
-1
[I] = &[Y][YP][VP] = -[VP] (12.92)
&
Now define [Is] as the driving-current source vector for the active-element
pattern of the j t h element. Then all elements of [I"] are zero except for the j t h
element, which may be set to unity. The port voltages due to [I"] are found from

where [Y'] is a square diagonal matrix with elements l/Z,, and where Z, is the
0 20 40 60 80 termination impedance at each patch port. Then from eqns. 12.92 and 12.93, the
theta patch currents for the active-element pattern are
Fig. 12.29 Geometry of a finite array of rectangular microstrip patches -1
v, {[Yp]
[I] = - + [YT]} -I [I"] (12.94)

For scanning at the angle 8 , 4 , the probe (port) currents should be driven as The active-element pattern of the j th element is then computed as
El(& 4) = EO(e, 6) 1 1, e-jk~(u-rm +M (12.95)
m
where x,, ymare the co-ordinates of the probe feed on the rn th patch, and u, v
are direction cosines given by eqn. 12.14. The patch-current amplitudes are then where EO(O,$)is the pattern of a single PWS mode [35]. The active-element gain
given by the column vector [I] and [23] is then

where [Y] = [Z]-' is the inverse of the moment-method impedance matrix. The
active input impedance at the rn th patch is then [23]
This definition does not include power lost in the terminating impedance of the
fed element.
An intermediate result that can easily be obtained from this analysis is the
mutual coupling between two microstrip patches. Mutual coupling has been
 740 Analysis and design considerations for phased-array antennas Analysis and design considerations for phased-array antennas 741
calculated or measured by several authors [18,23,24,38-411, with a wide variety active-element patterns can be calculated for finite patch arrays. The patch-
of analytical methods. Fig. 12.30 shows data for the E- and H-plane mutual array. Scan range is also constrained by the scan-blindness effect, which is losely
coupling between two rectangular patches using the above formulation. Observe related to element spacing and substrate parameters, as discussed in Section
that the magnitude of the H-plane coupling decays much faster than the E-plane 12.2.1.3.
coupling. It can be shown that the H-plane coupling decays as I/?, while the
Table 12.1 Measured and calculated S-parameters
i, J S, (measured) S, (calculated)
25, 24 - 12.5dB / - 147' - 13.4dB / - 140"
25, 26 - 12'5dB / - 145" - 13.4dB / - 140'

25, 23 - 21.0dB 144' -21.5dB 159"

25, 27 - 21.0dB 1490 - 2 l . 5 d ~1590
24, 26 - 21.5dB -/ - 2 1 4 d ~1570
25, 18 - 24.5 dB 1113" - 26.0 dB /m2"
25, 32 - 25.0 dB 1
- 1120 - 26.0dB - /m
25, 11 -29.5dB /-113" - 29.8 dB / - 105"
25, 39 - 30.0dB / - 128' - 29.8 dB / - 105"

18, 32 - 30'0dB / - 115" - 29.8 dB / - 105"

7 x 7 element array; 8, = 2.55; d = 0.16cm; a = b = 3.45cm; L = 2.0cm, W = 3.0cm;
X, = 0.55cm. Y, = 0, f = 4.35GHz

Measurements were made on a 7 x 7 patch array on a thin substrate to verify

the theory. With all ports terminated in 50R, S-parameters were measured and
calculated for various element pairs at different locations in the array; typical
data are shown in Table 12.1. The elements are numbered across the H-plane,
as in Fig. 12.29. Because of the difficulty in obtaining an accurate phase
reference, it is estimated that the measured phase data in Table 12.1 may be in
error by about 10". Also note that the mutual-coupling data in Table 12.1 are
between two patch elements in the presence of all the other (terminated) ele-
ments, as opposed to the data of Fig. 12.30, which is for two isolated elements.
Element patterns were also measured for the above array, by terminating all
but the centre element. Figs. 1 2 . 3 1 ~and b show the measured and calculated
patterns. Agreement is generally within 1 dB or so, although it is clear that the
evident in the measured patterns. Asymmetry in the E-plane could possibly be
due to feed-probe radiation, which was neglected but could easily be included
Fig. 12.30 Calculated mutual coupling (magnitude and phase) between two rectangular in the solution.
microstrip patches

E-plane coupling decays much slower owing to surface-wave interaction. It is
also interesting to note from the phase data of Fig. 12.30 that there is essentially
an e - ~ k ~ phase
r dependence with distance for both E- and H-plane coupling.
Even though the E-plane coupling is dominated by a surface-wave field, the
surface-wave propagation constant of the thin substrate is close to k,.
As in the case of printed dipoles, the reflection coefficient, efficiency and
12.3 Design considerations for printed phased arrays
In this Section we will discuss a variety of considerations for the design and
development of printed phased arrays. Much of this material has appeared in
the literature [42-441, in relation to the monolithic phased-array concept, but is
also relevant for a broader class of printed phased-array antennas. Since printed
 742 Analysis and design considerations for phased-array antennas
Analysis and design considerations for phased-array antennas 743
and integrated phased arrays are still very much in the development stages, we
unfortunately cannot be completely thorough in this discussion. The integrated phased array is a general concept that refers to an antenna that
A phased-array antenna offers a number of desirable features to the systems takes advantage of photolithographic techniqes and microwave integrated cir-
designer, such as rapid beam scanning, pattern control and compatibility with cuitry (MIC) for the radiating elements, feed network and active (phase-shifterl
amplifierlswitching) circuitry. The logical extension of this concept is the mon-
olithic phased array, where the radiating elements, active circuitry, and feed
networks are all integrated on one substrate, (or in sub-array form on one
substrate). Such a purely monolithic phased array is far from realisation at the
present time and, for reasons discussed below, may not even be desirable from
technical viewpoint. Thus it has become more common to speak of an
integrated phased array that is as monolithic as possible.
The following Section will discuss some general factors affecting the design of
integrated arrays. Section 12.3.2 will then describe and discuss the relative
merits of a variety of array geometries, or architectures, that may be suitable for
- rolrulated I
I various levels of phased-array integration.
I/ o o o measured I
12.3.1 Design considerations: Design criteria for integrated phased arrays
may be categorised according to electrical or mechanical considerations:

( a ) Electrical considerations

Type of substrate: T o achieve a high level of integration, a semiconductor

(high E,) substrate (e.g. GaAs) is desirable for active devices and circuitry, but
a LOW-E, substrate is preferable for the antenna elements, to enhance bandwidth
and scan range.

I
Maximum scan range: The maximum scan range and the desire to avoid
I grating lobes controls the element spacing, and hence packing density, of the
-
H-plane
calculated
I
! array. Scan range is also constrained by the scan-blindness effect, which is
closely related to element spacing and substrate parameters, as discussed in
II 0 0 o measured I
I I Section 12.2.1.3.

Bandwidth: The substrate permittivity and thickness, and the element type,
all affect the bandwidth of the array. Thick substrates with low permittivity are
generally preferred for improved bandwidth.

Fig. 12.31 (a) E-plane, (b) H-plane measured and calculatedactive element patterns (centre Type ofpolarisation: This basically affects the complexity of the array. Linear
element) of a 7 x 7 rectangular microstrip patch array polarisation is the easiest to obtain; circular polarisation usually requires a
-
E, = 2.55. d = 0.159cm. a = b = 3.45cn-1, L = 2 0 c m .
X, 0+5crn. Y, = 0 , f = 4.35GHz
W = 34cm.
quadrature hybrid, and switched polarisation requires a switching network.
Dual polarisation is probably the most complicated, as it requires two separate
adaptive and beam-forming systems. The limiting factor in the deployment of sets of circuitry for each element.
phased-array systems, however, is cost, and the cost of such systems seems to
be increasing. There exists, then, a strong interest in the integrated phased array, Spurious radiation: Radiation from the feed network and/or active circuitry
as such a design would use the technology of integration to (hopefully) lower the may degrade the sidelobe level, polarisation or gain of the array.
cost of phased arrays.
 744 Analysis and design considerations for phased-array antennas
( b ) Mechanical considerations
Number of elements: A typical phased array may require from lo3 to loS
elements. The array architecture must be able to accommodate this number of
elements and the requisite feed and control circuitry.
Substrate area: Substrate 'real estate' must exist for radiating elements, feed
networks, active circuitry and bias control lines.
Heat transfer: The efficiency of most active devices (particularly FETs) is
low. Thus heat removal is often a necessity, especially at millimetre-wave
frequencies.
Modularity: To facilitate the reliable fabrication and repair of an integrated
phased array, some type of modularity is needed.
A number of the above electrical problems arise from the apparent requirement
of using a high-dielectric-constant substrate for both the radiating elements and
the active circuitry. For example, microstrip antennas have better bandwidth
and less surface-wave excitation for low-dielectric-constant substrates, but the
likely semiconductor substrates have a relatively high dielectric constant. In a
sense, then, it is a conflicting requirement to have a single substrate for the
distinct functions of radiation (loosely bound fields) and circuitry (tightly bound
fields). As will be seen in the next Section, a number of new printed-antenna feed
methods have been developed to resolve this basic problem by using separate
substrates for the radiating elements and the active circuitry.
Substrate space is another prime concern, since a scanning array requires R F
power-distribution networks, control and bias circuits, phase-shifter circuits,
and possibly amplifier circuits, in addition to radiating elements. The amplifiers
may be needed to compensate for increased circuit losses at millimetre-wave
frequencies. As will be discussed below, a number of array configurations use
more than a single substrate to provide more space, as well as some other
advantages. In such cases, a method is needed to couple from one substrate to
another. Via holes (plated-through holes) can sometimes be used, but in general
it is desirable to avoid such direct connections because of very low yields, and
because such connections are usually very inductive at high frequencies. As an
alternative, some proximity coupling schemes are discussed below.
A large integrated phased array will probably consist of a number of sub-
arrays. Such sub-arrays, for example, might be fabricated on a single 'chip',
perhaps with one phase-shifterlamplifier circuit feeding all the antenna elements
associated with that chip. All the subarray 'chips' could then be mounted on a
'mother board' to supply RF, bias and control lines. Interconnections here also
pose a problem.
Circular or switchable (dual) polarisation is required for a number of applica-
tions, and, of course, such requirements complicate the design. Dual polarisa-
Analysis and design considerations for phased-array antennas 745
tion is probably the most difficult case to accommodate, as this essentially
requires two separate orthogonally polarised elements, or at least a single
element (such as a square microstrip patch) that can be switched between two
polarisation states. Circular polarisation is somewhat easier to obtain, by using
a circularly polarised element or a polariser to convert linear polarisation to
circular.
I blindness '
Fig. 12.32 Scan-blindness angle and bandwidth for a patch array with 1,/2 spacing on a
GaAs substrate versus substrate thickness
As discussed in Section 12.2.1.3, the scan-blindness effect can limit the scan
range of printed phased arrays. As the substrate is made electrically thicker (as
a result of higher frequency, dielectric constant or thickness), the angle at which
scan blindness occurs moves closer to broadside. This blindness angle thus
effectively limits the scanning range of the array. Fig. 12.32 shows the blindness
angle of an infinite microstrip patch array on a GaAs substrate versus substrate
thickness. Such a substrate 0.04& thick, for example, would have a blindness
angle of about 60, which would probably limit the useful scan range of the
array to less than 50" owing to the rapid increase of the reflection coefficient near
the blindness angle. The data of Fig. 12.32 assumes an element spacing of 1,/2
- the blindness angle moves closer to broadside for larger spacings. Also shown
in Fig. 12.32 is the approximate bandwidth of the patch element, which shows
that a trade-off exists between the bandwidth of the array and its maximum scan
range.
12.3.2 Array architectures
In this Section we will discuss several types of printed-phased-array geometries,
and their relative merits. Several of these configurations correspond to specific
canonical arrays analysed in Section 12.2.
 746 Analysis and design considerations for phased-array antennas
12.3.2.1 Single-layer substrate: The type of geometry that probably first
comes to mind when considering an integrated phased array is the single-layer
substrate shown in Fig. 12.33, where radiating patches, active circuitry and the
necessary feed networks are all contained on the same substrate. A major
problem with this approach is that there may not be enough room on the
substrate for all of the components. To avoid grating lobes, antenna elements
must be spaced no more than about 1,/2 apart; so if the phase-shifter circuitry,
R F feed network and bias lines can be fitted in at all, the spurious coupling
between these components may be severe. Another problem with this geometry
is the scan-blindnesslbandwidth-trade-offwhich was discussed above. Scan
blindness will always occur at some scan angle for a printed array, but, for thin
substrates, the blind angle will be closer to endfire. Fig. 12.13, for example,
shows a calculated result for the reflection-coefficient magnitude versus scan
.angle in the three planes, for an infinite array of microstrip patches on a
0.021,-thick GaAs substrate. The blindness angle is seen to occur at about 82"
inthe E-plane (unity reflection-coefficient magnitude), although the reflection-
coefficient magnitude is still about 0.5 at 60 scan in the principal planes. If the
substrate thickness is increased, because of higher-frequency operation or a
desire for more bandwidth, the blindness angle will move closer to broadside,
as indicated in Fig. 12.32.
active microstrip
circu,itry pat~h
I n d n 1-5 /I
ground
plane
Fig. 12.33 Geometry of an array of microstrip patches on a single-layer substrate
This geometry is also susceptible to spurious radiation from the active cir-
cuitry and/or the feed network, which can degrade sidelobe levels or polarisa-
tion.
12.3.2.2 Two-sided geometry: Fig. 12.34 shows a two-sided substrate design
that eliminated many of the problems encountered with the single-layer case by
Analysis and design considerations for phased-array antennas 747
going to the root cause of those problems, and using two separate substrates for
the distinct functions of radiation and circuitry. A substrate with a low dielectric
constant holds the radiating microstrip patches, while a parallel semiconductor
substrate contains active circuitry and feed networks. The two substrates are
separated by a ground plane, and apertures in this ground plane are used to
couple R F power from the feeds to the radiating elements.
radiating
elements
coupling
aperture\
active circuitry/ ground1
and feed network plane
Fig. 12.34 Cross-sectional view of a two-sided integrated-array geometry with aperture-
coupled patch radiators
This design thus matches the substrate to the electrical function, resulting in
improved blindness/bandwidth performance. For example, with an E, = 2.55
antenna substrate, the thickness would have to be about 0.051, for a blindness
at 80, and the situation would be even better for a lower-dielectric-constant
substrate. Since we have two substrates, much more space is available than in
the single-layer case. In addition, the ground plane effectively isolates the active
circuitry and feed network from the radiating elements to reduce spurious
coupling and radiation.
This array configuration is dependent on the aperture-coupled microstrip
antenna, which has been described in detail in Reference 27 and theoretically
analysed as a single element in Reference 28 and 29, and as an array in Section
12.2.2.4. Fig. 12.20 shows the geometry of a single aperture-coupled patch
antenna, fed by a microstripline on the feed substrate. The feed line is usually
terminated in an open-circuited stub for tuning. The aperture is smaller than
resonant size, so very little radiation occurs in the back region. Models have
been successfully fabricated and tested at frequencies from 2 to 20GHz.
A final feature of the two-sided array, and the array configurations to follow,
is the fact that it offers better radiation 'hardening' from lightning or EMP
effects compared with the single-layer design, owing to the shielding effect of the
ground plane. The coupling of the sensitive active circuitry to the outside world
must take place through the microstrip antennas and coupling apertures, which
present a two-pole (or more) filter response to signals outside their bandwidth.
12.3.2.3 Perpendicular feed substrates: Another design that uses separate
substrates for the radiating elements and active circuitry is shown in Fig. 12.35.
In this case, a vertical substrate holding the radiating elements is fed by a
748 Analysis and design considerations for phased-array antennas
number of parallel-feed substrates. Coupling is again through apertures in the
ground plane of the antenna substrate. This design also allows the use of a
low-dielectric-constant substrate for the radiating elements and a separate
semiconductor substrate for the active circuitry, similar to the two-sided geo-
metry, and so has the same advantages in relation to scan-blindnessjbandwidth
performance and shielding of spurious radiation or coupling. In addition, this
architecture has a number of other advantages.
First is the fact that the feed substrate can be of virtually unlimited size, since
there is no immediate restriction on the 'depth' dimension away from the
vertical-antenna substrate. Waveguide phased arrays usually use this depth
dimension to a similar advantage. The geometry also permits a modular con-
struction, where feed modules could conceivably be plugged into receptacles on
the antenna substrate.
A feed network
antenna
substrates
I substrates
Fig. 12.35 An integratedphased-arrayconfiguration using a feed substrate perpendicular to
the radiating-element substrate
This design also allows efficient heat transfer from the ground plane of the
feed substrate. At millimetre-wave frequencies, low device efficiency requires
efficient heat transfer from active circuitry. The unobstructed ground plane of
the feed substrates allows much heat removal to take place, while the embedded
ground plane of the two-sided design makes heat removal more difficult.
Finally, such a geometry would lend itself well to space-fed phased-array lens
designs, which may be of interest for some applications. This could be im-
plemented by having antenna substrates a t both ends of the feed substrates. It
does not appear, however, that this geometry would be useful if dual polarisa-
tion were required.
Analysis and design considerations for phased-array antennas 749
The array with perpendicular-feed substrates depends on the feasibility of
feeding a single patch through an aperture with a microstrip line on a perpen-
dicularly oriented substrate. Such a geometry is shown in Fig. 12.36, and has
been discussed in more detail in Reference 45. This design has been verified
experimentally, but no theory has been developed beyond the simple arguments
presented in Reference 45.
microstrip A .,zEz:
resonant 1( . ..
-
direction r o n t n r t s tnn
of oatch of aperture
microstrip/
radiation '1
I ,
, feed line
ground
plane
Fig. 12.38 Geometry of a rnicrostrip antenna fed through an aperture with a microstrip feed
line on a perpendicularly oriented feed substrate
The geometry in Fig. 12.36 shows a direct connection from the feed line to the
top of the aperture; the two ground planes are also in electrical contact. Another
version of the perpendicularly fed antenna excites the aperture by proximity
coupling, eliminating the need for a direct connection of the feed line, as shown
in Fig. 12.37. Other variations, including the use of a co-planar waveguide feed,
are also possible.
12.3.2.4 Endfie elements: The previously discussed integrated-phased-array
designs all used microstrip patches or printed dipoles which radiate normal to
the substrate on which they are printed. An alternative to this approach is to use
elements which radiate endfire to the substrate, as shown in Fig. 12.38. This
example shows the use of tapered-slot antennas, but other elements capable of
endfire (to the substrate) radiation, such as dipoles, could be used as well.
This type of geometry then uses a single substrate for both active circuitry and
radiating elements, but in a rather different manner from the single-layer design
discussed earlier. A lot of substrate space is available for feed networks and
circuitry, and the design can readily be used for space-fed lens arrays. The
individual substrates can be made in modular form, and heat transfer should not
be a problem.
750 Analysis and design considerations for phased-array antennas Analysis and design considerations for phased-array antennas 751
rnicrostrip patch Although surface waves can still be excited on the substrates, an additional
on front side
problem is the possibility of scan blindness caused by surface waves on the
protruding grid of dielectric slabs; such effects have been observed in similar
arrays with protruding dielectrics. In addition, this configuration would prob-
ably not be useful if circular polarisation is desired.
The tapered-slot element has been discussed in Reference 7, and may be
constructed with either a linear taper or a curved slot. The slot antenna can be
proximity fed with a microstrip-linelslot-linetransition, in Reference 7, or the
slot line could be directly fed from the active circuitry. In this regard, it is
interesting to note that slot line has a number of distinct advantages over
microstrip in such millimetre-wave integrated-circuit applications [46].

12.4 Conclusion

antenna s u b s t r a t e This Chapter has discussed the analysis and design of printed phased arrays.
Analytical techniques were outlined, and applied to several canonical infinite
Fig. 12.37 Geometry of a microstrip antenna fed through an aperture which is proximity and finite printed arrays. General considerations for the design of integrated
coupled to a microstrip feed line on a perpendiculady oriented feed substrate
arrays were also discussed.
This Chapter has summarised most of the work to date on the anaysis of
printed arrays, but there is much yet to be done. Some topics include proximity-
coupled elements, the use of wide-angle impedance-matching layers, the de-
velopment of improved probe-feed models, and the effect of substrate aniso-
f e e d network tropy.
a n d active

12.5 Acknowledgment

The author would like to thank his graduate students, James Aberle and Fran
Harackiewicz, for reviewing the manuscript and making valuable suggestions
regarding the consistency of notation.

12.6 References

1 OLINER, A. A,, and KNITTEL, G. H.: 'Phased array antennas'. in Proc. Phased Array
Antenna Symposium', (Artech House, 1972)
2 STARK, L.: 'Microwave theory of phased array antennas - A review', Proc. IEEE, 1974,62,
pp. 1661-1701
3 HANSEN, R. C. (Ed.): 'Microwave scanning antennas', (Academic Press, NY, 1966)
radiators
4 AMITAY, N., GALINDO, V., and WU, C. P.: 'Theory and analysis of phased array anten-
nas', (Wiley Interscience, NY, 1972)
5 MAILLOUX, R. J., McILVENNA, J. G., and KERNWEIS, N. P.: 'Microstrip array technol-
ogy', IEEE Trans., 1981, AP-29,pp. 25-37
Fig. 12.38 An integrated phased-array geometry using tapered-slot elements that radiate in
6 JAMES, J. R., HALL, P. S.,and WOOD,C.: 'Microstrip antenna theory and design', (Peter
the endfire direction
Peregrinus, 1982)
752 Analysis and design considerations for phased-array antennas Analysis and design considerations for phased-array antennas 753
7 BAHL, I. J., and BHARTIA, P.: 'Microstrip antennas (Artect House, 1980) 33 KOMINAMI, M., POZAR, D. M., and SCHAUBERT, D. H.: 'Dipole and slot elements and
8 POZAR, D. M., and SCHAUBERT, D. H.: 'Scan blindness in infinite phased arrays of printed arrays on semi-infinite substrates', IEEE Trans., 1985, AP-33, pp. 600-607
dipoles, IEEE Trans., 1984, AP-32, pp. 602-610 34 ISHIMARU, A,, COE, R. J., MILLER, G. E., and GEREN, W. P.: 'Finite periodic structure
9 RANA, I. E., and ALEXOPOULOS, N. G.: 'Current distribution and input impedance of approach to large scanning array problems', IEEE Trans., 1985, AP-33
printed dipoles', IEEE Trans., 1981, AP-29, pp. 99-105 35 POZAR, D. M.: 'Analysis of finite phased arrays of printed dipoles', IEEE Trans., 1985,
10 DESHPANDE, M. D., and BAILEY, M. C.: 'Input impedance of microstrip antennas', IEEE AP-33, pp. 1045-1053
Trans., 1983, AP-31, pp. 740-747 36 POZAR, D. M.: 'Finite phased arrays of rectangular microstrip antennas', IEEE Trans., 1986,
11 MOSIG, R., and GARDIOL, F. E.: 'A dynamical radiation model for microstrip structures AP-34, pp. 658-665
in Advances in electronic and electron physics: Vol. 59' (Academic Press, 1982) pp. 139-237 37 KING, R. W. P., MACK, R. B., and SANDLER, S. S.: 'Arrays of cylindrical dipoles'
12 PERLMUTTER, P., SHTRIKMAN, S., and TREVES, D.: 'Electric surface current model for (Cambridge University Press, 1968)
the analysis of microstrip antennas with application to rectangular elements', IEEE Trans., 38 JEDLICKA, R. P., POE, M. T., and CARVER, K. R.: 'Measured mutual coupling between
1985, AP-33, pp. 301-31 1 microstrip antennas', IEEE Trans., 1981, AP-29, pp. 147-149
13 KNITTEL, G. H., HESSEL, A,, and OLINER, A. A,: 'Element pattern nulls in phased arrays 39 PENARD, E., and DANIEL, J. P.: 'Mutual coupling between microstrip antennas', Electron.
and their relation to guided waves', Proc. IEEE, 1968, 56, pp. 1822-1836 Letr., 1982, 18, pp. 605-607
14 LECHTRECK, L. W.: 'Effects of coupling accumulation in antenna arrays', IEEE Trans., 40 MALKOMES, M.: 'Mutual coupling between microstrip patch antennas', Electron. Lett.,
1968, AP-16 1982, 18, pp. 520-522
15 LIU, C. C., HESSEL, A,, and SHMOYS, J.: 'Performance of probe-fed microstrip-patch 41 VAN LIL, E., and VAN DECAPELLE, A,: 'Comparison of models for calculating mutual
element phased arrays'. Phased Arrays Symposium, Bedford, MA, 1985 coupling in microstrip arrays', IEEE AP-S Symposium Digest, Boston, 1984, pp. 745-748
16 FRAZITA, R. F.: 'Surface-wave behavior of a phased array analyzed by the grating-lobe 42 POZAR, D. M., and SCHAUBERT, D. H.: 'Comparison of architectures for monolithic
series', IEEE Trans., 1967, AP-15, pp. 823-824 phased array antennas', Microwave J., 1986, 29, pp. 93-104
17 ALEXOPOULOS, N. G., and RANA, I. E.: 'Mutual impedance computation between printed 43 POZAR, D. M.: 'Phased arrays of printed antennas', ISAP Symposium, Kyoto, Japan, 1985
dipoles', IEEE Trans., 1981, AP-29, pp. 106-111 44 POZAR, D. M.: 'New architectures for millimeter wave phased array antennas'. JINA
18 POZAR, D. M.: 'Considerations for millimeter wave printed antennas', IEEE Trans., 1983, International Symposium on Antennas, Nice, France, 1986
AP-31, pp. 740-747 45 BUCK, A. C., and POZAR, D. M.: 'An aperture coupled microstrip antenna with a perpen-
19 CASTANEDA, J., and ALEXOPOULOS, N. G.: 'Infinite arrays of microstrip dipoles with dicular feed', Electron. Letr., 1986, 22, pp. 125-126
a superstrate (cover) layer'. IEEE AP-S International Symposium Digest, Vancouver, Canada, 46 JACKSON, R. W.: 'Coplanar versus microstrip for millimeter wave integrated circuits'.
1985, pp. 713-717 Microwave Theory and Techniques Symposium, Baltimore, 1986
20 WRIGHT, S. M., and LO, Y. T.: 'Efficient analysis for infinite microstrip dipole arrays', 47 SCHUSS, J. J., HANFLING, J. D., and MORROW, R. E.: 'Observation of Scan Blindness
Electron. Lett., 1983, 19, pp. 1043-1045 Due to Surface Wave Resonance in an Array of Printed Circuit Patch Radiators'. 1987
21 WHEELER, H. A,: 'A survey of the simulator technique for designing a radiating element', International IEEE Antennas and Propagation Symposium, Blacksburg, VA, 1987
in OLINER, A. A,, AND KNITTEL, G. H., (Eds.): 'Phased array antennas' (Artech House,
1972)
22 POZAR, D. M., and SCHAUBERT, D. H.: 'Analysis of an infinite array of rectangular
microstrip patches with idealized probe feeds', IEEE Trans., 1984. AP-32, pp. 1101-1107
23 POZAR, D. M.: 'Input impedance and mutual coupling of rectangular microstrip antennas',
IEEE Trans., 1982, AP-30, pp. 1191-1 196
24 NEWMAN, E. H., RICHMOND, J. H., and KWAN, B. W.: 'Mutualimpedance computation
between microstrip antennas', IEEE Trans., 1983, MlT-31, pp. 941-945
25 MOSIG, J. R., and GARDIOL, F. E.: 'General integral equation formulation for microstrip
antennas and scatterers', Proc. IEE, 1985, 132H. pp. 424432
26 ABERLE, J. T., and POZAR, D. M.: 'Analysis of infinite arrays of one- and two-probe-fed
circular patches', IEEE Trans., AP. (Accepted for publication)
27 POZAR, D. M.: 'A microstrip antenna aperture coupled to a microstrip line', Electron. Lett,
1985, 21, pp. 49-50
28 SULLIVAN, P. L., and SCHAUBERT, D. H.: 'Analysis of an aperture coupled microstrip
antenna', IEEE Trans., 1986, AP-34, pp. 977-984
29 POZAR, D. M.: 'A reciprocity method of analysis for printed slot and slot-coupled microstrip
antennas', IEEE Trans., 1986, AP-34, pp. 1439-1446
30 JACKSON, R. W., and POZAR, D. M.: 'Full-wave analysis of microstrip open-end and gap
discontinuities', IEEE Trans., 1985, MTT-33, pp. 1036-1042
31 POZAR, D. M.: 'General relations for a phased array of printed antennas derived from infinite
current sheets', IEEE Trans., 1985, AP-33, pp. 498-504
32 WHEELER, H. A.: 'Simple relations derived from a phased-array antenna made of an infinite
current sheet', IEEE Trans., 1965, AP-13, pp. 506-514
 Chapter 13

Circularly polarised antenna arrays

K. Ito, T. Teshirogi and S. Nishimura

Introduction

This chapter is concerned with various techniques for circularly polari3ed

microstrip arrays. Circular polarisation is effective for many radio systems, such
as communications, remote sensing, navigation and radar systems. In par-
ticular, at present, mobile-satellite-communication and direct-broadcasting-
satellite systems use circular polarisation, because they do not need polarisation
tracking.
In these systems, it is desirable for each ground terminal to have a low-profile
and lightweight antenna. Also, it is a requirement to achieve a specific gain
which cannot be obtained by a single radiating element. In addition to high gain,
multiple functions, such as electronic beam scanning, beam shaping and low
sidelobe-radiation patterns, are often required. For these reasons, interest in
circularly polarised microstrip-array antennas is increasing rapidly.
Circularly polarised microstrip arrays are classified into three major cate-
gories. The first group includes arrays which are composed of circularly pol-
arised (sometimes linearly or elliptically polarised) microstrip patches. This type
of array is the most common and widely used, and it includes many variations.
The second type of array is composed of composite elements, which consist of
electric- and magnetic-current-source elements. The third type are travelling-
wave arrays which utilise radiation due to suitable discontinuities in travelling-
wave transmission lines. These techniques will be discussed in this chapter.

13.1 Various types of circularly polarised arrays

13.1.1 Arrays of patch radiators

The conventional method of obtaining a circularly polarised array is to arrange
circularly polarised microstrip patches with appropriate feeding.
( a ) Circularly polarised radiating elements: There are various types of cir-
cularly polarised patches or resonators, and these are described in Chapter 4. In
 756 Circularly polarised antenna arrays Circularly polarised antenna arrays 757
Fig. 13.1 typical circularly polarised patches which can be used as array ele- (b) Feedmethods andarray conjiguration: There are several feed methods for
ments are shown. a linear or planar circularly polarised array, and the detail and the correspond-
The most direct approach for obtaining circular polarisation is to excite a ing array cbnfiguration will-be described in the next Section.
square or circular patch with two orthogonal modes of equal amplitude and a
+
differential phase shift of 90, by using a 90' hybrid as shown in Fig. 1 3 . 1 ~
and b. The antenna can be excited from a single feed point by use of a dual-feed
device, such as a 90' hybrid or power splitter with the necessary phase shift [I,
21. For wider bandwidth applications, four-probe feeds with 0, 90, 180' and
270" phase differentials are used which can suppress higher-order modes formed
in the thick substrate [3], as shown in Fig. 13.1~.
QQQ

0 b c b
dual feed 4- robe feed

Fig. 13.2 Example of co-planar amys [671

feeder substrate
/I&= 2.55)

h
single teed
Fig. 13.1 Circularly polarised microstrip patch antennas

Several methods have been proposed to provide circular polarisation without
the complexities inherent in dual-feed devices. One approach is to attach a single
feed point at a location so as to excite two equal-amplitude degenerate or-
thogonal modes, and then to introduce some asymmetry into the cavity so that
the degeneracy of the modes is removed. Examples of this technique are the
square microstrip patch with a tilted slot [4], the corner-fed rectangular patch
[5, 61, the slightly elliptical patch [7, 81, the pentagon-shaped patch [9], and the
circular disc with perturbation element [lo], as shown in Fig. 13.ld-h, respec-
tively. However, these perturbation techniques for generating circular polarisa-
tion have very limited axial-ratio bandwidth - generally of the order of 1%.
...
antennas substrate
Fig. 13.3 Rear-feed microstrip array [471
The most simple is the corporate (or parallel) feed system which splits the
power between n output ports with a prescribed distribution while maintaining
equal path length from the input to output ports. The bandwidth of this type
of array is essentially wide. In practice, it will primarily be limited by the match
of the radiating elements. Figs. 13.2 and 13.3 show examples of corporate feeds.
 758 Circularly polarised antenna arrays
Fig. 13.2 is a co-planar array in which the array and the corporate feed system
are formed on the same plane. Although this array configuration is simpler to
manufacture, the radiation from the microstrip feed line deteriorates the overall
radiation characteristic of the array.
The rear feed system, in which the feed network is located behind the array,
is effective for shielding the spurious radiation from the feed lines and devices.
One example, as shown in Fig. 13.3, is a circularly polarised array, composed
of circular discs with perturbation notches, each of which is driven from the
corporate feed circuit in the rear side through a feed probe.
Another simple form of feed system is a series feed in which the circularly
polarised radiating patches are attached periodically to a transmission line. In
this configuration, the phase of the radiating elements is determined by their
spacing along the transmission line; therefore, as the frequency is altered, a
progressive phase shift results down the array, which causes the main beam
direction to change and the beam to squint.
Generally, beam squint with frequency is a particular disadvantage of
travelling-wave arrays. However, this beam squint can be eliminated by equalis-
ing the path lengths between the input and each element. A compact form of
such squintless array has been developed by Rodgers [l I].
(c) Circularly polarised array composed of linearly or elliptically polarised
elements: A circularly polarised array can also be realised even by using linearly
or elliptically polarised elements. In general, since microstrip antennas, and
particularly single-feed-type antennas, have a narrow-ellipticity bandwidth,
techniques for obtaining an array which is composed of linearly or elliptically
polarised elements, but radiates circular polarisation over a wide frequency
band, are useful. Details of these wideband techniques will be described in
Section 13.4.
Fig. 13.4 Arrangement and excitation of a pair of single-feed-type microstrip antenna [12]
(@ 1982 IEEE)
In this Section, these techniques are just introduced as a type included in the
category of circularly polarised arrays of patch radiators.
There are three kinds of array. The first is when a n array is composed of sets
of pairs, in each of which the differential orientation angle and phase shift
between two antennas is 90, as shown in Fig. 13.4 [12]. Although each element
has a narrow axial-ratio bandwidth, and radiates a heavily elliptical or almost
Circularly polarised antenna arrays 759
linear polarisation at frequencies off resonance, the cross-polarisations of the
paired elements cancel each other out and the array can maintain good polarisa-
tion characteristics over a wide bandwidth ( 3 10%).
In the second case, circular polarisation is achieved by having a basic 2 x 2
sub-array composed of single-feed linearly polarised elements with unique
angular and phase arrangements of the elements, as shown in Fig. 13.50 and b
[13]. Both the angular orientation and feed phase of the element are arranged
in a 0". 90, O0, 90, or 0, 90, 180, 270" fashion.
Fig. 13.5 Circularly polarised 2 x 2 sub-array of linearly polarised elements [13] (@ 1986
IEEE)
a 0'. 90', 0'.90' arrangement
b 0'. 90'. 180'. 270' arrangement
The third array, which is called a sequential array, is a more generalised
configuration [14]. A differential orientation angle and phase shift are provided
sequentially for each element of the array. A schematic arrangement of an
N-element sequential array is shown in Fig. 13.6. The sequential array provides
not only excellent circular polarisation in the boresight, but also low VSWR
over the wide frequency band.
13.1.2 Arrays of composite elements
In general, a proper combination of electric and magnetic radiating elements
will be able to produce circular polarisation when the two radiation fields from
the elements are perpendicular to each other and have 90" phase difference. Such
a combination of elements is sometimes referred to as a composite element.
Several kinds of composite elements for circular polarisation have been
reported, including a combination of a slot and two parasitic dipoles [I 51, as
shown in Fig. 13.7a, and a combination of a strip and a slot [16], as shown in
Fig. 13-76.
Fig. 13.7a shows the configuration of an array element composed of an
 760 Circularly polarised antenna arrays
Fig. 13.6 Configuration of a sequential array [I41
Fig. 13.7 Two types of composite elements for circular polarisation
a Slot and two parasitic dipoles [I 53
b Basic elements and working principle of CP-PASS [I61 (@ IEE)
Circularly polarised antenna arrays 761
excited slot and two parasitic flat dipoles [15]. The slot can be excited by a
microstrip line. The dipoles are placed above the slot at a distance d and angle
a. The two parameters, d and a, are determined so as to produce circular
polarisation. The array element was analysed using the technique of reaction
matching. A planar array consisting of four identical elements, as shown in Fig.
13.8, was designed, and its various characteristics were simulated taking into
account the inter-element coupling effects. The simulation showed that the
planar array could produce a gain of more than 13dB and the 1.5 dB bandwidth
of the axial ratio was about 6%. An actual array can be constructed by using
a proper substrate and by modifying the design method.
Fig. 13.76 shows the fundamental structure of a circularly polarised printed
array composed of strips and slots (CP-PASS). It consists of a strip on a thin
substrate, a slot in the ground plane and a microstrip feed line. The strip and
the slot - basic radiating elements - are almost half a wavelength long and
the spacing between them is a quarter of a guide wavelength ,I8 along the line.
The strip and the slot are excited by the electric and magnetic fields propagating
along the microstrip line, and radiate electric fields E, and Ed, respectively,
shown in Fig. 13.7b in the broadside direction.
Fig. 13.8 Configuration of array (four-element sets) [15]
Fig. 13.76 also shows the voltage and current distributions V1and I, along the
microstrip line when the line is short-circuited at I = 0. If the strip and the slot
are located at the maximum points of V, and I,, respectively, the elements can
produce circular polarisation efficiently in the broadside direction.
Fig. 13.9 shows a typical configuration of a linear-array-type CP-PASS,
which consists of three sets (i.e. six pairs) of elements. Each element pair has a
strip element (the term 'strip element' means a combination of a strip dipole and
a window) and a slot. The window, a kind of wide and long slot, is located in
the ground plane in order to effectively increase the gain and bandwidth of the
strip dipole [17].
762 Circularly polarised antenna arrays
Tapered window edges are also excited slightly by the magnetic field along the
line, and radiate unwanted waves. However, as shown in Fig. 13.9, the strip
dipoles are placed at the voltage maxima along the line with a half-wavelength
spacing, so that the unwanted radiation from the windows can be effectively
suppressed. The details of unwanted radiation will be discussed in Section
13.3.2.
Additionally, it is quite easy to control aperture distribution along the feed
Circularly polarised antenna arrays 763
assumes that electromagnetic waves radiate in the main from the bent parts of
the microstrip line. The line electric-current source method [25] assumes that the
source of the radiation is the line electric current (of uniform amplitude) along
the central line. These two radiation mechanisms, though based on different
principles, give the results that are in good agreement.
fundamental element -:&---

line by adjusting the coupling gaps between the elements and the feed line. A
design procedure for CP-PASS and a design example are given in Section
13.2.2.
(a)
r--z-------7
.
,{ , matched load

b
, ~ine a
7 '.04
.kGb,t
on 1

-
3
I dz LO;:!
- ..*
+U
1g/$ !:
I
I

B I._,
II Strip

rdipok
c
-- - - - (- -b-)- - - - matched load
t inout
3

Window

Fig. 13.9 Configuration of CP-PASS (three-element sets) [38] (@ IEE)

13.1.3 Travelling-wave arrays

It is a well known fact that discontinuities in a microstrip line produce radiation
[18]. The microstrip-line antenna utilises this phenomenon: the strip conductor
Fig. 13.10 Circularly polarised microstrip-line antennas
of a microstrip line, which is bent periodically like a meander, forms a circularly a Rampart-line antenna
polarised travelling-wave array. b Chain antenna
Fig. 13.10 shows some circularly polarised microstrip-line antennas. These c Square-loop-type microstrip-line antenna
are: d Crank-type microstrip-line antenna

(a) Rampart-line antenna [19, 201

(b) Chain antenna [21]
(c) Square-loop-type microstrip-line antenna [22]
(d) Crank-type microstrip-line antenna [23]
These antennas will radiate right-handed circularly polarised waves when the
power is fed from the left-hand end, and the right-hand end is terminated in a
matched load. On the other hand, when the power comes from the right-hand
end and the matched load terminates the left-hand end, each antenna radiates
left-handed circularly polarised waves.
The circularly polarised radiation of the four types of antennas (a)-(d) can be
explained by the magnetic-current source method [24] for (a), and the line
electric-current source method for (b)-(4. The magnetic-current source method
The rampart-line antenna shown in Fig. 13.10~is built from sets of four
right-angled bends. The right-angled bends have to be chamfered in order to
reduce the right-angled-discontinuity susceptance [26]. The radiation arises
from the bends, and the polarity has the direction shown in Fig. 13.10~.Let us
consider the fundamental element surrounded by a dotted line in Fig. 13.10~.
When the lengths a, b and c are appropriately chosen, the radiated field from the
four bends will produce circular polarisation. For example, when a = 3Ag/8,
b = Ag/2 and c = Ag/4 (Ag is the guide wavelength of the microstrip line), the
antenna will become a circularl'y polarised travelling-wave-array antenna [20].
The chain antenna and one of its fundamental elements are shown in Fig.
13.106. Each fundamental element is built from a V-shaped circularly polarised
radiating element and a U-shaped phase shifter. A U-shaped phase shifter is
 764 Circularly polarised antenna arrays
inserted between two V-shaped teeth of the zigzag line. Each arm of the
V-shaped radiating element has the length a, and the angle between the two
arms is a. For practical purposes, the optimum ranges of the parameters are:
a = 0.251g-0.51g, a = 90'-150'. To suppress reflection from the angled parts,
a capacitive ear is added to each V-shaped radiating element
The square-loop-type microstrip-line antenna, shown in Fig. 13.10c, is
formed from a series of fundamental elements. Each fundamental element is
made of a square loop, which radiates circular polarisation, and a straight
feeder. When the perimeter of the square loop is ,I8, the loop will radiate circular
polarisation. In this case, the linear or ti on will oDerate as a feeder, and its
effective length will be lgl
Fig. 13.11 Instantaneous current distribution and polarity
Fig. 13.10d shows a crank-type microstrip-line antenna which gives better
frequency characteristics for the axial ratio and radiation pattern, compared
with a single rampart-line antenna. The antenna is made of two parallel mean-
der (rampart) lines of the same dimensions. One of the meander lines is shifted
for one-half of its period. In order to reduce the susceptance, every bend is
chamfered, but in special cases the bends need not be chamfered [23]. When the
lengths of each segment of the fundamental element are selected properly, the
fundamental element will radiate circular polarisation. The method of selection
for the lengths is described in detail in Section 13.2.3.
Circularly polarised antenna arrays 765
Using the line electric-current source method, we will explain now how these
four types of antenna radiate circular polarisation. Suppose that the power is fed
from the left-hand end. The instantaneous current distribution is shown by the
arrows in Fig. 13.11. The arrows point in the opposite direction every half guide
wavelength 4. Fig. 1 3 . 1 1 ~shows the fundamental element of a rampart-line
antenna. At time t = 0, the total radiation field from each segment consists only
of the radiation from the vertical segments, because the radiated fields of the
horizontal segments cancel each other out. When t = 1/(4fi, where f is the
frequency, the total radiation field will be of horizontal left-oriented polarity.
Similarly, when t = 1/(2fi or t = 3/(4f), the polarity of the total radiation field
will be oriented downwards or to the right, respectively. As shown in Fig.
13.1la-d, the polarity of the total field of the radiated electromagnetic waves
(normal to the surface of the paper), rotates counterclockwise, and completes
one cycle in time I/$ Thus the fundamental element operates as a right-handed
circularly polarised antenna. Similarly, the three other types (b), (c) and (d) also
operate as right-handed circularly polarised antennas, as can be seen from Figs.
13.11b-d.
13.1.4 Others types of arrays
Several other types of circularly polarised microstrip or printed arrays have been
developed and reported. In the following, some typical arrays will be in-
troduced.
-
Radiating elements Feed network
Fig. 13.12 Broadband flat radiating element [27]
Dubost [27] has proposed a broadband circularly polarised flat antenna, as
shown in Fig. 13.12, which consists of two flat radiating elements that are
linearly polarised and placed orthogonally. Each radiating element is a sym-
metrically fed flat folded dipole, and it is separated from the reflector by a
dielectric sheet. Dubost et al. [28] have described a cylindrical array composed
of four such flat antennas producing circular polarisation and omnidirectional
radiation.
Ito et al. [29] have proposed a travelling-wave-type circularly polarised array
as shown in Fig. 13.13~.Inclined half-wavelength printed dipoles are arranged
along both sides of a microstrip feed line terminated in a matched load. The
spacing D, between adjacent dipoles is a quarter of a guide wavelength, and the
766 Circularly polarised antenna arrays I Circularly polarised antenna arrays 767
spacing D, and the angle a are determined from the desired main-beam direc- 13.2 Design of circularly pdarised arrays
tion. The frequency bandwidth was as narrow as those of other simple micro-
strip antennas. 13.2.1 Arrays of patch radiators
The generalised design method for circularly polarised microstrip patch arrays
, can be divided into two steps: the design of circularly polarised patch radiators
themselves and the design of the appropriate feed network. The former is
elaborated in Chapter 4.

+
Crossed printed
dioote

.....-...
;!j j
__^)

i! /j
A!
2
.
.
.i ,
.: .i
j j
/ j

.
.,0 a
3

Fig. 13.1 3 Travelling-wave printed arrays

a Dipole array [29]
b Slot array [30]

Nakaoka et al. [30] have proposed another type of circularly polarised

travelling-wave array composed of inclined slots, as shown in Fig. 13.136. The
slot arrangement can be determined in a similar way to the printed dipole array
in Fig. 13.13~.Although the slot array requires a reflector under the substrate
in practical use, its frequency bandwidth could be much wider than that of the
printed-dipole array.
The crossed printed-dipole array [3 I], shown in Fig. 13.14a, or the crossed slot
array [32], shown in Fig. 13.146, both fed from microstrip lines, can be used as
circularly polarised printed arrays. Compared with microstrip patch arrays,
such arrays have relatively wide frequency bandwidths. Fig. 13.14 Crossed printed-element arrays
a Dipole array [31]
An alternative method of constructing a circularly polarised array is to place b Slot array [32]
a circular polariser over a linearly polarised microstrip array. As a typical
example, Henderson and James [33] have proposed a DBS reception array As is well known, typical feed systems for microstrip patch arrays are series
shown in Fig. 13.15. A parallel-plate polariser is overlaid on a comb-line array feeding and parallel or corporate feeding [35]. Such feed systems will be des-
antenna [34] with a suitable spacer. cribed more fully in Chapter 14.
 768 Circularly polarised antenna arrays
Circularly polarised antenna arrays 769
A typical series-fed linear array, as shown in Fig. 13.16a, consists of identical
patch radiators producing circular polarisation, a microstrip feed line and its
terminal load. Various shapes of patch radiators, as shown in Fig. 13.1, could
be candidates for the array. Fig. 13.16~illustrates circular patch radiators with
a single feed point. These radiators should be designed to produce circular
polarisation and to meet the required conditions mentioned in Chapter 4.
For the design of such a series-fed linear array, an equivalent circuit model,
as shown in Fig. 13.16b, is employed using transmission-line theory [36]. Nor-
mally, the patch radiators are attached to the feed line with equal spacing D. If
the load admittance Y L is equal to the line characteristic admittance Y,, a
/ travelling-wave array is formed [35]. On the other hand, if YLis equal to zero
or infinity, a resonant array is formed. All the radiating-element admittances Y,
(1 < n < N) can be determined from the conditions of input-impedance
matching and aperture distribution required. A more generalised design meth-
Comb-line array Polariser od, using such an equivalent circuit, will be elaborated in the next Section.
In the array configuration, the excitation phase of the patch radiators is
Fig. 13.15 Linear polarised array and polariser [33] (0 IEE) determined by their element spacing along the feed line. Therefore, the main
beam direction will change with frequency variation. A series-fed linear array
Circular polarised can minimise the feed-line losses at the sacrifice of frequency bandwidth.
r a d i a t i n g elements
Circular polarised
radiating elements
Load

i D i i feed line i

\
Feed
I npul n e lwor k
Fig. 13.17 Example of corporate- fed linear array
Y r n : R a d i a t i n g element a d m i t t a n c e
Yc : Characteristic admittance A typical example of the other feed system, parallel or corporate feeding, is
shown in Fig. 13.17 feeding a circularly polarised linear array. In general, the
Y L : Load a d m i t t a n c e
patch radiators will be designed independently of the feed system in order to
Fig. 13.16 Series-fed patch array and its equivalent circuit produce circular polarisation and to meet the required conditions mentioned
a Circularly polarised series-fed4inear array above. The feed system, in this case, splits the input power between the output
b Equivalent circuit
ports with a prescribed distribution, while maintaining equal electrical path
770 Circularly polarised antenna arrays
lengths from the input to output ports. Therefore, the frequency bandwidth of
such an array could be essentially wide. In other words, the frequency band-
width will be limited by the patch radiators. However, the feed-line losses would
be quite large. Detailed design methods for the feed system will be described in
Chapter 14.
In a practical patch array, the antenna performance with regard to such
matters as radiation pattern and axial ratio could be deteriorated by mutual
coupling between the tadiating elements or unwanted radiation from, for exam-
ple, its feed network. These practical design problems will be disussed in Section
13.3.
In practice, the frequency bandwidth of a microstrip patch antenna is relative-
ly narrow. Section 13.4 will describe several design techniques for wideband
circularly polarised patch arrays.
13.2.2 Arrays of composite elements
This Section will concentrate on the design of a circularly polarised printed
array composed of strip dipoles and slots (CP-PASS) as a typical design method
for series-fed circularly polarised arrays.
Fig. 13.9 shows a configuration of CP-PASS consisting of M radiating
element sets which are fed in series from a microstrip feed line. For the design
of a series-fed linear array, in general, an equivalent-circuit model and trans-
mission-line theory are employed to determine the element spacing and input
immittances [36].
-k 4rnth .section
1 -
Fig. 13.18 Equivalent circuit of mth element set of circularly polarised printed array of strips
and slots (CP-PASS) [37] (@ IEE)
In an equivalent-circuit model, a strip dipole and a slot can be approximately
represented by a shunt and a series element, respectively, to a transmission line
[37]. The effect of the windows placed in the ground plane can be neglected at
a design frequency of about f;,. In a practical design, inter-element coupling
effects should be included in the determiantion of the element input immit-
tances. Therefore, an equivalent circuit of the mth element set at f;. can be
expressed as shown in Fig. 13.18, where z,,, y;,, and d,, are the normalised
element input immittances and; rIn,
etc. are the travelling-wave voltages (sub-
I Circularly polarised antenna arrays 771
scripts f and r indicate forward and reflected waves, respectively, and super-
scripts + and - indicate the load and generator sides of each element, respec-
tively).
For simplicity in the design, we will set
and also
The relationship between z,, and y,,, may be derived from the design conditions
of both circular polarisation and input-impedance matching [37]. As a result, the
design conditions for the mth element set will be written as
z,,,,, = 2y:", (* is complex conjugate) (13.3a)
where h and E , are the thickness and the relative dielectric constant of the
substrate and 1, is the free-space wavelength a t f , .
Fig. 13.1 9 Equivalent circuit of M-set CP-PASS [38] (@ IEE)
Fig. 13.19 shows an equivalent circuit of an M-set CP-PASS atf,, where Y,,,
and Z,, are the nth input admittance of a strip element and the mth input
impedance of a pair of slots, respectively, and Z , is a terminal load (normally
it is a short or open circuit). Y,,, and Z,, can be determined from the design
conditions mentioned above and the aperture distribution required.
A generalised design procedure for linear-array-type CP-PASS is briefly
described as follows [38]:
(a) A t f , evaluate the propagation constant y,(= a, + j&) of the microstrip
line on a specific substrate. The substrate used throughout this Subsection is
mentioned in Table 13.1.
(b) Calculate all the input immittances of the elements required for a specific
CP-PASS design by applying the design procedure previously described [37].
(c) As a preliminary experiment, make a linear array composed of identical
strip elements (no slots) arranged along a feed line. Derive the frequency
dependence of the average input admittance from its measured input reflec-
tion coefficient I-,, where the superscript st denotes a strip-element array. Then,
by testing some arrays which have different coupling gaps, obtain the depen-
 772 Circularly polarised antenna arrays Circularly polarised antenna arrays 773
dence of resonant conductance ei
and resonant length 4: on the coupling gap greater than a specified value AR,,, the element lengths should be adjusted
S,, atJ;. Fig. 13.20 shows measured values of typical dependences of c::and 4: until the axial ratio becomes less than AR,,,. Then obtain the frequency
on S,, at resonance. dependences of average input immittances Ef and of the elements, where the
superscript cp denotes a circularly polarised array [38].
Table 13.1 Design data for five-set Chebqchev-array CP-PASS
m 1 2 3 4 5 Measured f = 3.0GHz

Am 1.OOO 1.609 1,932 1.609 1.OOO 6.0

G m (mS) 0.420 1.116 1.641 1.148 0.445
(a) 2.10 5.58 8.21 5.74 2.23
S,rm (mm) 1.67 1.32 1.21 1.31 1.64
~~~ (mm)
a m (mm)
2.80
35.38
1.62
34.93
1.28
34.73
1.59
34.9 1
2.72
35.35
a h (mm) 38.39 38.02 37.84 38.01 38.37
Substrate: DICLAD 522 (h = 0.8 mm, E, = 2.6) w = 2.0 mm ( Z , = 50 R), a, = 2 dB/m,
4 = 0.68 j, = 3.0 GHz, D, = D, = D, = 68.0 mm, 2, = 0 R, b,, = 6 , = 2.0, a,, = 55.0,
b, = 20.0, e, = 0.3 mm O
t ' l b 2.0
' ' ' 3.0'
b,, (mm)
L

41)
l T

j Measured f =3DGHz Fig. 13.21 Measured typical dependences of average resonant resistance and resonant
length on the coupling gap for slot arrays [38] (@ IEE)

j Meosured f =3.OGHz
36.0

Fig. 13.20 Measured typical dependences of average resonant conductance and resonant
length of the coupling gap for strip-element arrays [38] (@ IEE)

(4 In a similar way, obtain the dependence of the average resonant resistance
RI and resonant length 4;on the coupling gap a,/ for some slot arrays. Typical
measured dependences are shown in Fig. 13.21.
( e ) By combining the strip elements with the slots, construct a small array as
shown in Fig. 13.9. To facilitate the procedure, the coupling gaps are chosen
from Figs. 13.20 and 13.21 as the relationship for circular polarisation
R = 2Zze: (13.4)
is satisfied, where Z, is the characteristic impedance of the line. Then, measure
the axial ratio of the array in the broadside direction. If the axial ratio at f;. is
Fig. 13.22 Measured typical dependences of average resonant conductance and resonant
length on the coupling gap for circularly polarised arrays [38] (@ IEE)
(f) By testing some circularly polarised arrays with different coupling gaps,
obtain the dependences of resonant conductance @and length aff on S,, and the
dependences of resonant resistance @ and length aff on 6 , atf,. Some typical
measured dependences are shown in Figs. 13.22 and 13.23, where AR,, was
chosen to be 3 dB. It was found that the difference between, for example, Fig.
13.20 and Fig. 13.22arose from mutual coupling between the strip elements and
the slots.
774 Circularly polarised antenna arrays I
Circularly polarised antenna arrays 775

( g ) Determine all the element lengths and coupling gaps required for the design
of CP-PASS from the experimental curves obtained in (f). There will be almost
no need to correct the element dimensions because the curves will involve the
inter-element coupling effects.

Fig. 13.23 Measured typical dependences of average resonant resistance and resonant
length on the coupling gap for circularly polarised arrays (381 (@ IEE)
Fig. 13.24 Front view of designed CP-PASS [38](@ IEE)
Slots and windows are indicated by broken lines
In addition, when a reflector is placed under the ground plane, the same design
procedure will also be available.
To demonstrate the validity of this design method, a Chebyshev-array CP-
PASS consisting of five element sets was designed and measured at S-band [38].
The sidelobe level of -20 dB was specified. Using this design procedure and
employing Figs. 13.22 and 13.23, design data for the array were obtained as
shown in Table 13.1. A,, is the amplitude ratio of each element set. For
simplicity in the experiment, a reflector was not used.
Fig. 13.24 shows the front view of the array producing RHCP in the broad-
side direction. Slots and windows are indicated by broken lines.
Fig. 13.25 shows the measured radiation patterns in the yz-plane atf,. The
sidelobe levels of the co-polar radiation were less than -20 dB. The maximum
cross-polar radiation was about -20 dB because circular polarisation was not
achieved at f,, as shown next.
Fig. 13.26 shows the measured axial ratio versus frequency. Circular polarisa-
tion was obtained at 2.95 GHz, which is 1.7% below A . The 3 dB bandwidth
of the axial ratio was 9.3%.
Fig. 13.27 shows the measured input impedance of the array versus frequency.
Good impedance matching was achieved as predicted, and the bandwidth for
VSWR < 2.0 was 8.5%.
For stricter designs of CP-PASS, e.g. a requirement for lower sidelobe levels, Fig. 13.25 Measured radiation patterns in the yz-plane 1381 (@ IEE)
the cross-polarisation behaviour of the array has to be taken into account. In f = 3,0GHz (Fig. 13.24)
 776 Circularly polarised antenna arrays
I
this case, the strip element will have to be represented instead by a T-type
equivalent circuit in the design procedure [40].
For the case of a linear-array-type CP-PASS, the arrangement of the radiat-
ing elements shown in Fig. 13.9 seems to be the most suitable and efficient.
Circularly polarised antenna arrays 777
13.2.3 Design of travelling-wave arrays
Section 13.1.3 gave an outline of the four types of circularly polarised travelling-
wave-array antennas. In this Section the crank-type microstrip-line antenna is
chosen as representative and the method for its design is described.

fundamental e l e m e n t
m i t e r e d bend
/ / d i e l e c t r i c substrate
/

Frequency ( GHz) ground p l a t e s t r i p conductor

Fig. 13.26 Frequency dependence of axial ratio measured at broadside [38](@ IEE)

< -: s t r i p current
__--.
. image current

Fig. 13.28 Crank-type microstrip-line antenna

a Antenna configuration (four elements)
b Fundamental element
c ldealised fundamental element

( a ) Equations for the circularly polarised radiation: Fig. 13.28~shows a

-1.0
Fig. 13.27 Measured input impedance versus frequency [38] (@ IEE)
However, if a planar array is formed using this arrangement, the spacing
between elements on adjacent feed lines would be so small that the design would
become more complicated because of strong inter-element coupling effects. To
avoid this, radiating elements on one side of a microstrip line should be used for
a planar array [39].
crank-type microstrip-line antenna with the fundamental element surrounded
by a dotted line. Fig. 13.286 shows the shape of the fundamental element;
however, we will study the idealised form shown in Fig. 13.28~.We make the
following assumptions:
(i) The neighbourhood of the strip conductor is the medium with the effective
relative permittivitiy.
(ii) There are two currents: the strip current which is concentrated along the
central line of the strip conductor, and the image current formed by the ground
plate.
778 Circularly polarised antenna arrays
(iii) The amplitude of the strip current is uniform, and the reflection of the
electric current on the bent parts can be neglected. Thus, mutual coupling and
substrate surface wave effects can also be neglected.
To obtain the conditions for the radiation of circularly polarised waves in
0 = Om, 4 = 0" direction, the following equation must be satisfied:
E4 = +_jER (13.5)
i.e. both components E, and E4 of the electromagnetic wave radiated from the
fundamental element must have the same absolute value, and if the phase
difference is 90" the resulting wave will be circularly polarised.
When a linear-array antenna is constructed from these fundamental elements,
the condition for the formation of the main beam in the 0, direction, i.e. the
condition that electromagnetic waves radiated in the 6, direction from both end
points of the fundamental element are in phase, is given by
kLcos 0, - P(I - 46) = 2nn (13.6)
+ + +
where k = 2n/lo,/? = 2a/Ag,L = 2a c, 1 = 2a 26 c , 1, is the free space
wavelength, ,Ig is the guide wavelength of the microstrip line, L is the length of
the fundamental element, 1 is the strip-conductor length of the fundamental
element along the central line, a, b, c are the lengths of the crank segments, 6
is the correction length for the crank segments and n is an integer. The physical
meaning of 6 is the difference between the physical and electrical lengths of the
crank segments.
From the components Eo and E, of the electromagnetic wave radiated from
the fundamental element, and from eqns. 13.5 and 13.6, we can obtain a
condition which ensures that the circularly polarised wave is radiated in the 0,
direction when n = - 2. The condition is
where
U = (1 - 5 cos em), = l J A o and W is the width of the crank segment. The
upper sign represents left-handed circular polarisation and the lower sign re-
presents right-handed circular polarisation. When the values for A,, [, 0, and b
Circularly polarised antenna arrays 779
are given, the values for a and c can be obtained from eqns. 13.7 and 13.8. The
choices of b is free; however, the optimal value for b is near 442.
Fig. 13.286 shows the distance d between the two strip conductors. The value
of d should be chosen according to
which is an experimentally derived inequality.
The conditions for circularly polarised radiation of the three remaining types
of travelling-wave arrays are shown in Table 13.2.
( b ) Details of the design: We will give the important factors in the construc-
tion of the four types of array antennas. However, a full explanation will be
given only for the crank-type antenna. The length L of the fundamental element
and the electrical length I' of the other three types can be obtained from Table
13.2.
(i) Correction of bent parts [41]
As shown in Fig. 13.29 the phase constant of the straight-line portion is P, and
the phase constant of the bent parts is the effective phase constant P'. In this
case, when the width of the straight line is W, the correction length for the crank
segments will be
W
Fig. 13.29 Right-angle bend
The value of 6 must just be determined by the following method: (i) By setting
0, = 90" and 6 = 0 , then a, b and c are obtained from eqns. 13.7 and 13.8. On
the basis of these calculated values an experimental antenna is constructed. The
value of the frequency, fm is then obtained experimentally by setting the main
i beam in the broadside direction, 0, = 90'. (ii) From the frequency f, and the
design frequency f, the value for 6 is obtained from
1 where V is the velocity of light. In this manner, a crank-type fundamental
780 Circularly polarised antenna arrays Circularly polarised antenna arrays 787
element which radiates circular polarisation in the 0, direction can be deter-
mined.
(ii) Main beam direction 8, and frequency f
The crank-type microstrip-line antenna is a series-fed linear-array antenna, and
when the frequency changes, the main beam direction will also change. The
relationship between these two factors is given by

cos 0, =
1 1-46
- ---
L( c + y)
where n represents a negative even number.
(iiij Length of the fundamental element
Normally we make n = -2 in eqn. 13.13 for the construction of a linear-array
antenna. However, depending on the choice of the main beam direction 8,, a
grating lobe can appear simultaneously. Because of this, there are limitations on
the values of 8, when only the main beam, corresponding to n = -2, is present
in the visible region. In other words,
L V
-<
Lo cos 0, + 1
v = 1 for rampart, loop
v = 2 for chain, crank
According to this equation, when the lengths of the crank segments a, b and c
are calculated from eqns. 13.7 and 13.8 for a given value for Om, the value
L = 2a + c must satisfy eqn. 13.14. The direction 8, is usually larger than 60".
(iv) Return loss
A microstrip-line antenna has its microstrip lines periodically bent. If there are
any reflected waves from the bent parts, such an antenna will exhibit a high
return loss, no matter how small the reflection. High return loss is a consequence
of the total sum of the reflected waves which come from every fundamental
element, and are in phase when the electrical length of the fundamental element,
I' = I - 46, is equal to a multiple of L,/2. Therefore, the condition I' = mLg/2
(m is an integer) must be avoided.
When 0, = 90, then I' = 21,; therefore, there is always high return loss. In
this case the following counter-measures can be applied: (i) The bent parts must
be matched very carefully; and (ii) the length of each crank segment must be
chosen carefully, so that the sum of the reflected waves from all the four bent
parts in the fundamental element equals zero.
( v ) Transmission loss and efficiency
In the case of a microstrip-line antenna, the strip current fed through the feeding
end decreases exponentially on its way from the feeding end to the matched
load, because the strip current is radiated from each fundamental element.
Therefore, amplitude distribution, as shown in Fig. 13.30, can be assumed. In
this case, the efficiency q of the microstrip-line antenna can be represented by
782 Circularly polarised antenna arrays
where q, is the aperture efficiency, q, is the feeding efficiency (1 - R2), qc is the
radiating efficiency and R2 represents the power dissipated at the matched load.
Here, q, is determined by both the conductor loss and dielectric loss.
Circularly polarised antenna arrays 783
voltages. In addition, in the case of circularly polarised arrays, it causes de-
terioration of the polarisation characteristics. Theoretical studies of mutual
coupling in microstrip arrays have been presented by Pozar [42] and Malkomes
[43], and experimental work has been done by Jedlicka et al. [44] and Haneishi
et al. [45]. All these works, however, dealt with linear polarisation.
From a knowledge of mutual coupling in a linearly polarised array, the
mutual coupling effects on polarisation characteristics of a circularly polarised
array can be derived.

Fig. 13.31 Geometry of circularly polarised linear array

0 5 I0 15 Let us consider a circularly polarised linear array composed of microstrip

TL(dB)
Fig. Efficiency of travelling-wave antenna (7, = 0.9)
Fig. 13.30 shows the results when r], = 0.9. As can be seen, there is an
optimal range for the transmission loss of between 10 and 12 dB, in which case
the antenna efficiency is about 72%.
An example of the crank-type microstrip-line antenna is given in Chapter 19.
patches with feed hybrids as shown in Fig. 13.31. The array can be represented
in terms of a scattering matrix [46], which expresses the complex coupling
coefficients between the incident + and reflected - voltage at each feed part.
Each element of the scattering matrix is given by E-plane and H-plane coupling
coefficients for the linearly polarised array. Thus, for a circularly polarised
array, if we assume the mutual coupling between orthogonal linearly polarised
V- and H-components to be neglected, we have

13.3 Practical design problems

13.3.1 Mutual coupling

Mutual coupling between radiating elements in microstrip arrays results in both
distortion of the element radiation pattern and also errors in the element feed
784 Circularly polarised antenna arrays Circularly polarised antenna arrays 785
and normal to the array, as shown in Fig. 13.31, the co- and cross-polarised fields
radiating from the array are expressed by

and

The excitation voltages Vv,,and VH, at the v- and H-ports of the nth element are r , , and rHndepend on the scan angle B,, and thus the polarisation characteris-
given by tics vary with the scan angle due to mutual coupling.
VV" = (vi, + v&) = G n ( l + rv") (13.18)
where

where

If each hybrid is ideal, V;, and V;, can be expressed by the input voltage V,f to
the hybrid as
Fig. 13.32 Measured and calculated mutual coupling IS,212
for circular patches [47]

Next, let us consider the element spacing of circularly polarised arrays. In

and order to suppress grating lobes and scan the beam widely, the spacing between
adjacent element should be small. Small spacing, however, causes large mutual
coupling, and deteriorates the radiation characteristics. Thus the spacing for a
circularly polarised array should be chosen from practical trade-offs. The cal-
+
The sign corresponds to the sense of polarisation rotation. Thus, the co- and culated and measured mutual coupling of a two-element circular-patch array
versus element spacing for both the E-plane and H-plane are shown in Fig. 13.32
cross-polarised components radiated from the nth element become
[47]. From this Figure, it can be seen that, as the spacing increases, the E-plane
coupling becomes larger than the H-plane coupling owing to the stronger
surface wave, and the magnitude of coupling for both planes is almost equal at
d = 0.681,. Fig. 13.33 shows the effect of orientation on coupling between
and circular patches for several element spacings. From this Figure, d = 0.681, may
be suggested for circularly polarised arrays, because, at this spacing, the mutual
couplings for the E- and the H-planes are nearly equal and less than - 22 dB;
furthermore, they are almost independent of the orientation angle of the array.
If the main beam is scanned in the direction B,, where 0, is the angle from the This fact almost holds good for rectangular microstrip patches.
 786 Circularly polarised antenna arrays Circularly polarised antenna arrays 787
13.3.2 Unwanted radiation
For practical design of microstrip-array antennas, and particularly for circularly
polarised antennas, so-called 'unwanted radiation' should be taken into acc-
ount. Unwanted radiation may result from the generation of higher-order
modes in microstrip patches (see Chapter 4 for details), co-planar microstrip
feed lines and discontinuities [IS], microstrip feed transitions such as connectors
[48], surface waves excited on the substrate [49], secondary current sources on
the substrate edges [50], and so forth.
Unwanted radiation, in general, sometimes causes high cross-polarisation
level, degradation of antenna gain, reduction of frequency bandwidth, and
alteration to radiation patterns such as sidelobes and nulls.
Recently, cross-polarisation effects in linearly polarised microstrip antennas
have been studied by, for example, Hall and James [51] and Hansen [52]. For
circularly poalrised antennas, the axial ratio.will readily be degraded by cross-
Fig. 13.33 Mutual coupling between microstrip circular patches as a function of orientation
polarisation. In the following, the relationship between axial ratio and cross-
[471
polarisation level will be discussed.
Fig. 13.34 shows a typical circularly polarised array composed of identical
radiating elements with a normal microstrip feed network. The array will radiate
nearly circular polarised waves in the broadside direction, which can be decom-
posed into the two orthogonal fields E, and E,. The circularly polarised radia-
tion fields can be written as

Ey t Radiation f i e l d s
where the subscripts R and L represent right-hand and left-hand circular pol-
arisation, respectively. Then the axial ratio AR is given by

Suppose that the total radiation fields E,, and Ey, are simply expressed as
Radiating elements

E,, = E, (13.32)
where IEr,l expup) represents the unwanted radiation field caused by, say,
higher-order modes in the elements or feed network as mentioned above.
If right-hand circular polarisation is desired, the relationship E, = jEy must
be satisfied for the case of no unwanted radiation. However, we then obtain

Next U R is defined as
6 Microstrip feed
Input network in order to examine the influence of unwanted radiation on the axial ratio.
Fig. 13.34 Concept of circularly polarised array Putting E, = IE,I and substituting eqn. 13.34 into eqns. 13.28 and 13.29, we
788 Circularly polarised antenna arrays
obtain
Then the axial ratio can be written as
Fig. 13.35 shows the degradation of the axial ratio caused by the unwanted
radiation defined in eqn. 13.34. The solid and the broken lines indicate the
estimates for the cases p = - n (the worst case) and p = f4 2 , respectively.
mr Calculated
Fig. 13.36 Influence of unwanted radiation on axial ratio
-p = -n (worst case)
---- p = n/2 *
The above discussion illustrates that the influence of unwanted radiation on
the axial ratio can be quite large. Therefore, it is quite important to suppress
- -
such unwanted radiation in the design of circularly polarised antennas.
Various methods for suppressing unwanted radiation in circularly polarised
microstrip antennas have been proposed. Chapter 4 elaborates on the suppres-
sion of higher-order modes in microstrip patches and describes an effective
suppression method using 'paired elements'. Another effective method is to
form a 'sequential array', which is a generalised method of paried elements. The
sequential array is described in detail in Section 13.4.3.
13.3.3 Limitations and trade-offs
In a large array, the achievable gain is limited owing to the conduction loss and
the dielectric loss in the microstrip feed line, and the radiation loss generated
from discontinuities in hybrids, impedance transformers and right-angled cor-
ners of the microstrip line.
Circularly polarised antenna arrays 789
As the array size increases, the directional gain will increase proportionately.
However, as array size increases, the feed-line length become longer, and the
feeder loss will eventually increase faster than the directional gain; the power
gain will therefore decrease.
I- \ slot
Fig. 13.36 Microstrip ground-plane slot array losses [35]
a Array corporate feed arrangement
b Power gain of array against number of elements
- theoretical gain of array with no feeder losses
---- theoretical gain of array with feeder losses
feeder attenuation
x x x measured values of gain
Corporate feed arrays in particular, have longer feed lines and larger feeder
loss than series-fed arrays. Reference 53 has quantified the limit for corporate
feed arrays. The feed geometry is shown in Fig. 13.36~;each line feeds a
ground-plane slot and there are 2L/& and L/& elements in the E-and H-planes,
respectively, where L~ is the area of the square array. Although this array is
790 Circularly polarised antenna arrays Circularly polarised antenna arrays 791
designed for linear polarisation, circularly polarised array can be treated in a I where W is the width of the strip conductor, 2, is the characteristic impedance,
similar manner. For the geometry shown, the length of feeder line from the input
point to any element is 3L/2. thus the power gain of the array is given by ! and a, is the conductivity of strip.

2L2 3 ( c ) Radiation loss (571: All the expressions for radiated power due to dis-
G = G,, + 10 log -T - -LF (13.37) continuities take the general form
n, 2
where G,, is the element gain in decibels and F i s the feeder loss in dB/unit-length
[351. where t is the dielectrics thickness, k is the free-space wave number and F(E)
The optimum value of L which gives the maximum power gain is depends only on the substrate permittivity, but is different for each microstrip
13.3 log e 5.776 configuration. This equation is for a unit incident current wave, the reflection
L = = - at the open circuit being assumed complete. The F(E)s for fundamental circuits
F F
were derived by Lewin [57]. .
Fig. 13.36b shows the power gain versus number of elements for a 12 GHz array Open circuit
with a feeder loss F = 0.075 dB/cm. The maximum gain can be seen from the
Figure to be about 30dB.
An effective method of reducing feed losses in the corporately fed arrays is to
replace part of the microstrip feed lines by a low-loss medium such as a coaxial I
Short circuit
line or waveguide. This approach is used in a synthetic-aperture radar antenna
for the SEASAT satellite [54], and in a circularly polarised microstrip array for
reception in a 12 GHz direct broadcasting satellite (DBS) [55]. The latter array
consists of four 256-disc-element sub-arrays which are connected by a
waveguide power combiner mounted on the rear side of the array. A measured Matched termination
gain of more than 33 dB was achieved for this 1024element array.
In co-planar arrays, not only conductor loss and dielectric loss of feed lines,
but also spurious radiations from power dividers, impedance transformers and
corners become significant loss factors. These losses in microstrip lines represent Right-angle corner
the major limitation of microstrip antennas. Estimates of these losses are
summarised as follows: & + I
- In-
28 G
( a ) Dielectric loss [56]:
F4(&) =
& JE-'i - IJm
n -
JE
(1 3.46)

The dielectric loss is almost constant for the substrate thickness. Conductor loss
and radiation loss, however, depend on the thickness, and furthermore the
variations of losses with thickness are quite different. Thus, a study of the
optimum thickness of the substrate is necessary.
Fig. 13.37 shows a calculated example of conductor losses and the radiation
,losses against thickness of the microstrip substrate [58]. The antenna consists of
+
where E is the dielectric constant of the substrate, E, = 1 q(&- l), q is the a 256-element circularly polarised planar array at 12 GHz. It was assumed that
dielectric filling factor and a is the conductivity of the substrate. each feed line from the input port to each element has a length of 40 cm, eight
( b ) Conductor loss [56]: 2-way dividers, and seven right-angle comers.
From the Figure, it can be seen that the conductor losses increase rapidly as
the substrate thickness decreases; on the other hand, the radiation losses in-
crease as the substrate becomes thicker. The losses also depend strongly on the
characteristic impedance of the microstrip lines. Since an optimum thickness
exists at which the sum of conductor loss and radiation loss becomes a mini-
mum, we can determine the substrate with the optimum thickness.
 792 Circular/y polarised antenna arrays
13.3.4 Non-planar scanning arrays
User terminals in aeronautical-satellite and inter-satellite communication links
require high-gain beams, capable of being steered to wide angles over a full
hemisphere. Scanning losses in planar phased arrays increase rapidly beyond
about 60, necessitating employment of non-planar scanning arrays.
L0ssdZo=50fl line
R A D I A T I O N LOSS
Loss
[dB]
Fig. 13.37 Conductor loss and radiation loss versus substrate thickness 1581
f = 12GHz,&,=2.17
Electronically switched spherical arrays are simple wide-angle scanning
arrays; an example using microstrip antennas has been described in a low-orbit-
ing satellite pointing its beam at at geostationary data-relay satellite [59].
Another example is a dome-shaped switching array which consists of 120
circularly polarised microstrip disc elements, and which provides 14dB of gain
with good uniformity over almost 300' of total angle [60]. The array produces
a beam by exciting a 12-element sub-array with non-phase compensation, and
changes the beam-pointing direction by selecting another sub-array which may
or may not contain some of the elements in the first sub-array. A disadvantage
of such electronically switched spherical arrays is that they require a large
number of elements to cover the hemisphere with high-gain beams; consequent-
ly, they are bulky and heavy.
A phase-compensated switched-element spherical array for mobile earth
stations for satellite communication was proposed by Hori et al. [61]; thus
enables one to achieve a high-gain beam with a small number of elements, by
providing the same number of variable phase shifters as excited elements.
! Circularly polarised antenna arrays 793
! Let us consider a general spherical array, whose co-ordinate system is shown
I
in Fig. 13.38. The array is composed of N elements which are located on the
limited sphere tilted at an angle a from the vertical axis (z-axis). If M elements
are excited and phase-shifted at a time, the radiation pattern is calculated from
the following equation:
Fig. 13.38 Co-ordinate system for spherical array [61] (Q IEE)
where, k is the wave number and Y,, is the phase of the nth element. e, and e,,
are the vectors of the radiation and element position, respectively, and are given
by
e, = (sin 9 cos 4, sin 9 sin 4, cos 9) (13.48)
e,, = a(sin a cos p,, sin a sin p,,, cos cc) (13.49)
where a is the radius of the sphere and g ( l , p) is the radiation pattern of the array
element.
The minimum coverage gain of a simple switched-element spherical array in
which the elements are switched one at a time is represented by the crossover
level of the radiation patterns of the adjacent array elements within the coverage
area (9, - A9 and 9, + A0 from the vertical axis). Since the minimum
coverage gain of the antenna is limited, in order to improve it without altering
the number of elements, the switched-element array proposed employs the same
number of phase shifters as excited elements, as shown in Fig. 13.39. This
antenna has both switched-element-array and phased-array functions. It uses
the first function to achieve wide scanning and the seond to increase crossover
level between adjacent beams.
The L-band array developed is shown in Fig. 13.40. It is 40 cm in diameter
and 20 cm in height. The antenna is composed of the radiator section, the
 794 Circularly polarised antenna arrays Circularly polarised antenna arrays 795
switching circuit and the controller. The radiator section consists of six micro- In the design of conformal arrays, the mutual coupling between elements
strip discs with parasitic elements for broadening the bandwidth, which will be arranged on a curved or folded plane has to be investigated. Hori [62] studied
described in the next Section. experimentally the mutual coupling between microstrip discs with parasitic
elements arranged on a roof-shaped plane as shown in Fig. 13.41. In the same

6 "ice
+Jel+ shifters

Na : number of array elements 7"

Ne : number of excited elements .4 .5 .6 .7 .8 .9 1

ELEMENT SPACING D/h
Fig. 13.39 Switching spherical array with phase shifters [61] (@ IEE)
Fig. 13.41 Measured mutual coupling of the non-planar two-element array with parasitic
elements [62]

Fig. 13.40 Inner construction of the spherical switching array (Courtesy: NTT, Japan) ARRANGEMENT RAD l US R/ A
Fig. 13.42 Measured mutual coupling of the non-planar array versus radius of the arrange-
The two-element-excitation method is applied in this test antenna. T o imple- ment [62]
ment two-element excitation, only one phase shifter consisting of two bits, one
of 45" and the other of 90, is required. Consequently, two elements can generate figure, the mutual coupling versus element spacing D/1 ( 1 is the wavelength) is
seven beams, and a six-element array can radiate 42 beams. also shown. Fig. 13.42, however, shows the mutual coupling versus the arrange-
The minimum coverage gain of a six-element array can be improved by 2.3 dB ment radius Rl1. From the Figure, it can be seen that at R > 3L, the mutual
by using the proposed method, when the coverage area is within 20'-60' from coupling becomes almost constant, and at R i21, the mutual coupling in the
the vertical axis. non-planar arrangement is 5 -
lOdB less than in planar arrangement.
 796 Circularly polarised antenna arrays Circularly polarised antenna arrays 797
13.4 Wideband circularly polarised arrays higher-order modes are excited because of the asymmetrical feed structure, and
they generate cross-polarisation. However, it is known that notches in circular
In most communication systems, transmitting and receiving frequencies are patch provide arbitrary elliptical polarisation, and therefore properly designed
separated by several percent (typically 7-10%). In general, a microstrip antenna notches can cancel the cross-polarisatin caused by the asymmetical feed struc-
has a narrow frequency bandwidth; therefore, in the practical design of circular- ture.
ly polarised arrays, techniques for achieving wideband polarisation characteris-
tics, as well as wideband impedance characteristics, are important. There are the
following techniques for achieving wideband circularly polarised arrays:
(i) Employment of wideband circularly polarised radiating elements
(ii) Stacked elements for dual-frequency resonance
(iii) Special configuration for wideband circularly polarised array.
Details of these techniques are described below.

13.4.1 Arrays of wideband elements

U
( a ) Radiating element with substrate of low dielectric constant: Since the
bandwidth of a microstrip antenna is given by [63] Fig. 13.43 Four-probe feed for higher-order mode suppression [3] (@ 7982 IEEE)

VSWR - 1
BW =
Q JVSWR
the bandwidth can be increased by reducing the Q-factor. It is well known that
the Q-factor is proportional to dielectric constant of the substrate, and inversely
proportional to its thickness. Therefore, in order to broaden the bandwidth, the
utilisation of a thicker substrate with lower dielectric constant is effective.
One such substrate consists of two thin layers of PTFE (polytetrafluoroethyl-
ene) bonded on each side of honeycomb material. This method, however,
frequently generates higher-order modes in a microstrip antenna. In circularly
polarised antennas, the higher-order modes become one of the sources of
cross-polarised waves, and therefore they must be suppressed. In a microstrip
circular patch, the dominant mode is TM,,, and the first higher-order mode is
TM,,,. One method of suppressing the undesired TM,,, is to excite a microstrip
radiator by four feeds with 0, 90, 180' and 270' phase differentials [3], as Fig. 13.44 Comparison of coupling between feed ports [3] (@ 1982 IEEE)
shown in Fig. 13.43. The TM,,,, mode is found by measuring the cross-coupling
between orthogonal ports. Fig. 13.44 gives a measured example of the coupling
between the orthogonal ports; the solid curve shows the case of the four-probe Fig. 13.45 shows a seven-element array, in which each element has two
feed in Fig. 13.43, while the broken curve shows the case of a conventional notches and is excited from orthogonally located feed points with a phase
two-probe feed. By using a thicker substrate with four-probe feeds, relativley differential of 90". Fig. 13.46 shows the measured axial ratio of this array, and
wide impedance and axial-ratio bandwidths ( 210%) can be achieved. it can be seen that the effect of the notches on the suppression of higher-order
For a large array, a multiple-probe feed system would become complicated, modes is obvious.
more expensive, and more prone to R F loss.
Another technique for suppressing the effect of higher-order modes generated ( b ) Application of parasitic elements: A two-layer microstrip antenna cap-
in a thick substrate of low dielectricconstant is to cut two notches on the patches able of broadband performance with excellent circular polarisation was
[64]. In a normal microstrip circularly polarised antenna using a thick substrate, proposed [65]. Such antennas are also referred to as electromagnetically coupled
798 Circularly polarised antenna arrays Circularly polarised antenna arrays 799
patches (EMCP), which have been shown to be broadband radiators for linear
polarisation. Fig. 13.47 illustrates the structure of the EMCP. The antenna
element consists of two circular patches of diameter Df and D, separated by a

R a d i a t l n q Patch
Foam F e e d l n P a t c h

Ground I

Fig. 13.45 Wideband microstrip array composed of notched elements [64]

Fig. 13.47 Electromagnetic coupled patch antenna [65] (@ 1984 IEEE)

3.5 3.7 3.9 4.1 4.3 4.5 4.7 4

frequency (GHz)
Fig. 13.48 Measured return loss [65] (@ 1984 IEEE)

0.01 .
2.0 2.1 2.2 2.3 2.4 distance S. The top patch is excited by the bottom patch (the feeding patch)
Frequency ( G H r ) which in turn, is fed by a coaxial line from underneath, or by a microstrip line
on the same plane as the feeding patch. The return loss of the EMCP, shown
Fig. 13.46 Improvement of axial ratio of the 7-element array by use of notches [64] in Fig. 13.48, is characterised by two resonant frequencies which vary with
800 Circularly polarised antenna arrays
separation. In general, the upper resonant frequency shifts downward and the
lower shifts upward when the separation increases.
By using EM-coupled patches, a broadband circularly polarised array can be
produced. However, when fed at two points (A and B in Fig. 13.47), the EMCP
generates highly elliptical polarisation because of its asymmetrical feed struc-
ture. One technique of achieving a good axial ratio is shown in Fig. 13.49. This
Fig. 13.49 Circularly polarised EMCP array [65] (@ 1984 IEEE)
Fig. 13.50 Microstrip disc antenna with parasitic element [67]
array employs symmetrical deployment of the radiating elements, which are
equally excited at two points. The configuration can cancel the radiation dif-
ference as a result of the symmetrical arrangement of array elements.
The EM-coupled patches are also capable of having only one resonant
frequency, and they have broadband characteristics. The antenna was first'
described by Taga et al. [66], and more recently it has been applied to a circularly
polarised shipborne antenna array for mobile satellite communication [67]. The
structure of the broadband microstrip antenna with a parasitic element is
Circularly polarised antenna arrays 807
illustrated in Fig. 13.50, where a parasitic element of radius b is mounted over
a microstrip antenna of radius a at height h. The parameters a, d and the
dielectric constant E, are determined from the substrate and the operating
frequency, while b and h are related to the bandwidth. An example of the relative
I I I I
i 0.07 0.08 0.09 0.1 t
Swclng h / Wavelength
Fig. 13.51 Relative bandwidth variation against element spacing and radius ratio of circular
disc elements [67]
Fig. 13.52 Frequency dependence of measuredgain of the 4 x 4-element array withparasit-
ic elements [67]
bandwidth variation is shown in Fig. 13.51 when hll, (3,is the wavelength) and
bla are varied; the relative bandwidth is defined as the ratio of the frequency
bandwidth over which the VSWR remains below 1.5 to the centre frequency.
A wideband circularly polarised array can be obtained by combining a
radiating and a parasitic array. Fig. 13.52 shows the frequency dependence of
the gain of an S-band array. Both the radiating array and the parsitic array are
composed of 4 x 4 elements. The radiating array has co-planar feeding circuits,
and each element is excited by a branch-line coupler. The element spacing is 0.78
802 Circularly polarised antenna arrays Circularly polarised antenna arrays 803
wavelength, and the height between the radiating array and the parasitic array parts play a role in transmission and reception, respectively. And each layer of
is 2 cm. The measured gain shows that the aperture efficiency exceeds 62% over the antenna is individually fed at two points with 90' phase shift, in order to
a frequency range of 2.6-2.8 GHz. obtain circular polarisation. Each upper-layer element is a conventional micro-
strip antenna, while the lower one is a circular microstrip antenna with an
13.4.2 Arrays of dual-frequency stacked elements
For many uses, the increased bandwidth is actually needed at only two distinct
frequencies, for transmission and reception, which may be too far apart for a
single antenna to operate efficiently at both frequencies. The behaviour of the
antenna characteristic at the range of intermediate frequencies may be of no
concern. In such cases, an antenna operating in dual-frequency bands is useful.

Fig. 13.54 3 x 3 stacked-element array for dual-frequency operation [69] (@ IEE)

A, B : Feed Point
(1.545 GHz)
2.83 GHz 3.10 GHz C. D : Feed Point
(1.6465 GHz)
In
x
0

40 30 20 10 0 -Io0 2.6 28 3.0 3.2

FREQUENCY (GHz)

Fig. 13.53 Dual-frequency stacked-patch antenna [63] (@ 1981 IEEE) Dielectric

a Cross section of typical stacked circular disc antenna Substrate
b Measured E, patterns at 2.83and 3.10 GHz
c Measured input impedance showing resonance at 2.83 and 3.10 GHz

One technique is to stack one circular patch on top of another in a sandwich
construction as shown in Fig. 13.53 [68, 631. This antenna can also be applied
as an element of an array, and one can obtain a dual-frequency circularly I
polarised array. An example is shown in Fig. 13.54. This antenna is nine-element I electrical shielding ring that provides enough space for the upper antenna to be
airborne phased array for aeronautical satellite communication, the spacing of I
I
which is 94mm (about half a wavelength at 1.611.5 GHz), and the dimensions
are about 300 x 300 x lOmm [69]. This phased array can scan its beam by
k 45' at least, for gain coverage greater than 12dBi. The element is a newly
developed stacked patch antenna as shown in Fig. 13.55. The upper and lower
I
Receive Port ' ~ r s n m i tPort
Fig. 13.55 Configurat~onof the stacked elements [69] (0IEE)
easily fed [70]. Owing to the shielding ring, this antenna has good isolation
characteristics between the transmit and receive ports. The measured coupling
between two ports is less than -30dB. Consequently, the antenna can be
designed optimally for both transmission and reception frequency bands in-
dependently.
804 Circularly polarised antenna arrays Circularly polarised antenna arrays 805
13.4.3 Wideband-array techniques ( b ) A r i a ! of 4-elerlzenr subnrrays: A circularly polarised array can also be
For a large array, the use of conventional dual-feed circularly polarised elements formed by using 4-element sub-arrays. each of which is composed of 2 x 2
has the disadvantage of complicated structure, R F losses due to many feed linearly polarised elements with unique angular and phase arrangements as
cables, hybrids and power dividers, and high cost. The application of microstrip shown in Fig. 13.5.
elements with single-point feeds is attractive for large circularly polarised arrays,
but, in general, these antennas have narrow ellipticity bandwidth. Therefore, the
techniques of configuring a circularly polarised arrays with elliptically or linear-
ly polarised elements become important in practice.
Three techniques were outlined briefly in Section 13.1.1. In this Section more
detailed explanations are provided.

( a ) Array of microstrip-patch pairs: This method is to construct a circularly

polarised array with pairs of microstrip patches [71]. In each pair, two elements
are arranged with an angular orientation of 90' to each other and fed with 90'
phase difference, as shown in Fig. 13.4. Using the pairs (or two-element sub-
arrays), a circularly polarised array can be constructed.

Fig. 13.57 64-element array composed of the sub-array units [77]

It is well known that circular polarisation can be achieved in the broadside

Fig. 13.56 Four-element coplanar sub-array unit of CP pairs [77]
Fig. 13.56 shows a four-element co-planar sub-array unit composed of the
pairs. In the Figure, T, is a quarter-wavelength impedance transformer which
provides both 90' differential phase shift and impedance matching. Fig. 13.57
shows a 64-element circularly polarised array composed of the sub-array units.
Typical measured data show that the 3 dB ellipticity bandwidth of the array with
paired elements is about 6% and it is greatly improved as compared with the
conventional array.
direction of an array composed of two linearly polarised elements with angle
and phase arranged in a 0, 90' fashion. This circular polarisation, however,
becomes very poor at angles greater than 5" off broadside owing to the spatial
phase delay between the two elements, as shown in Fig. 13.58. This spatial phase
delay no longer exists in the 2 x 2 subarray shown in Fig. 13.5, since the spatial
phase delay in one row or column is opposed to that of the other row or column,
and consequently they cancel each other out. The calculated radiation pattern
shown in Fig. 13.59 indicates that a 2 x 2 sub-array has excellent quality of
circular polarisation over a wide angular region of f 40' from boresight [72].
( c ) Sequential arrays: A more generalised circularly polarised array which
consists of an arbitrary number of identical elements with arbitrary polarisation
has been proposed by Teshirogi et a / . [14]. In this array, the incremental angular
806 Circularly polarised antenna arrays
orientation and the excitation phase difference are provided sequentially to each
element; therefore, the array is called a sequential array. The configuration of
an N-element sequential array is illustrated in Fig. 13.6. The nth element is
located at an arbitrary position on a plane, but with an orientation angle of
Circularly polarised antenna arrays 807
A sequential array has another advantage with regard to VSWR. According
to the differential path length of each feed line, the reflected wave back to the
input terminal from the nth element has a differential phase shift of 24,.

where P is an integer and 1 $ P < N - 1, with respect to the first element, say
element 1. The nth element is also fed with a differential phase shift of 4, radians
and the same orientation angle.

Fig. 13.59 Calculated radiation pattern in principal plane of 2 x 2 array shown in Fig. 13.5
[72] (0 1985 IEEE)
Element spacing = 0.9 wavelength.

THETA ( 8 1
Fig. 13.58 Calculated radiation pattern in principal plane of two-element CP array [72] (@
1985 IEEE)
Element spacing = 0.9 wavelength

We assume that the polarisation of the radiated field from element 1 is

elliptical in the boresight direction, and expressed by

where U, and V , are orthogonal unit vectors corresponding to the major and the
minor axes, respectively, of polarisation ellipse, and a and b are the amplitudes
i
of both components. The total radiated field E from the sequential array in the ( C ) P=3
boresight direction can be derived as Fig. 13.60 Possible 4-element sequential linear array

Therefore, if the reflection coefficients of all the elements are the same, the sum
of all the reflected waves V, returning to the input terminal of the array becomes
This means that the sequential array radiates perfect circularly polarised waves
in the boresight direction regardless of the polarisation of the element.
808 Circularly polarised antenna arrays Circularly polarised antenna arrays 809

Consequently, it can be seen that the sequential array provides not only perfect
circular polarisation in the boresight, but also no reflection at the input terminal.
There are several configurations of the sequential array corresponding to P
in eqn. 13.51. Fig. 13.60 shows three examples for a four-element linear sequen-
tial array. Since each phase difference 4, is usually given by adjusting the
feed-line length, these three arrays differ in ellipticity and VSWR bandwidth.
Fig. 13.61 shows the improvement factor of cross-polarisation discrimination
(XPD) and the ratio of the XPD of the sequential array to that of conventional
array. From this Figure, it is clear that XPD is improved as N increases, and the
case P = 1 is the best. It should be noticed that the examples in Figs. 1 3 . 5 and
~
b can be interpreted in terms of a sequential array, as a two-pair array of a
two-element sequential sub-array, and a four-element sequential array in the
case of P = 2. Therefore, from the point of view of ellipticity bandwidth, the
configuration in Fig. 13.60~is the widest of these three Celement arrays.

@
( I I I I I
Fig. 13.62 Arrangement of 2-dimensional sequential arrays
a Recurrence arrangment for a square array
b Generalised arrangement for a rectangular array

2 4 6 8 1 0
Number of elements
Fig. 13.61 Improvement factor of cross-polarisation discrimination [14]

Two-dimensional arrays can also be composed by combining linear sequen-

tial arrays. Fig. 1 3 . 6 2 ~shows an example of a square array. This is a recurrent
type of sequential array, in which each row and each column array is a linear
sequential array. For rectangular arrays, a more generalised configuration is
available, as shown in Fig. 13.626. In this method, the orientation angle and the
differential phase shift to be applied to the (m, n)th element 4, is defined by
4nm = Ndn + M4m (13.55)
where

Fig. 13.63 Two lest arrays 1141

Verification experiments have demonstrated wideband characteristics of the a Convent~onalarray
sequential array. The test array comprises two 2 x 4-element arrays. One is a b Sequential array
 870 Circularly polarised antenna arrays Circularly polarised antenna arrays 81 7
conventional array and the other is a sequential array, as shown in Fig. 13.63. 2 SANFORD, G.G.: 'Conformal microstrip phased array for aircraft tests with ATS-6', IEEE
All the elements are identical microstrip patches which have small notches and Trans., 1978, AP-26, pp. 642-646
are excited by single-point rear-side feeding. The material of the substrate is 3 CHIBA, T., SUZUKI, Y., and MIYANO, N.: 'Suppression of higher modes and cross
polarised component for microstrip antennas'. IEEE AP-S Int. Symposium Antennas and
glass-cloth PTFE, the dielectric constant is 2.6 and the thickness is 4mm. Propagat. Digest, 1982, pp. 285-288
4 KERR, J. L.: 'Microstrip polarisation techniques'. Proc. Antenna Applications Symposium,
1978
;:'-----
: - conventional
sequential
5 CARVER, K. R., and KOFFEY, E. L.: 'Theoretical investigation of the microstrip antenna'.
Physic, and Sci. Lab., New Mexico State Univ., Technical Report PT-00929, 1979
6 RICHARDS, W. F., LO, Y. T., SIMON, P., and HARRISON, D. D.: 'Theory and applica-
0
8.0 tions for microstrip antennas'. Proc. Workshop on Printed Circuit Antenna Technology, New
Mexico State Univ., 1979, pp. 8/1-23
7 SHEN, L. C.: 'The elliptical microstrip antenna with circular polarisation ', lEEE Trans., 1981,
AP-29, pp. 90-94
.
-
8 LONG.. S. A.. SHEN, L. C., SHAUBERT, D. H., and FARRAR, F. G.: 'An experimental
----

study of the circular-polarised, elliptical, printed circuit antenna', IEEE Trans., 1981, AP-29,
0 . 0 1 . I , , pp. 95-99
2.10 2.20 230 240 2.50 9 WEINSCHEL, H. D.: 'A cylindrical array of circularly polarised microstrip antennas'. IEEE
FREQUENCY (GHz) AP-S Int. Symposium Antennas and Propagat. Digest, 1975, pp. 177-180
10 HANEISHI, M., and YOSHIDA, S.: 'A design method of circularly polarised micrqstrip ~ ; s k

?
--.--
- conventional
sequential
FREQUENCY (GHz)
b polarized wave array antenna using slot antenna combined with parasitically excited dipoles',
Fig. 13.64 Measured bandwidth characteristics 1741
a Axial ratio
b VSWR
Fig. 13.64~and b show the axial ratio and VSWR of these arrays. From the
Figure, it is clear that the sequential array has much more wideband characteris-
tics of circular polarisation and impedance than the conventional array. For
example, 3 dB axial-ratio bandwidth of the sequential array exceeds 14%, which
is about 15 times that of the conventional array; while 1.5 VSWR bandwidth is
13.7%, which is about twice that of the conventional array.
Several applications of sequential arrays have been
for an airborne
microstrip phased array in aeronautical satellite communication [73] and for a
feed array of a contoured-beam reflector antenna [74].
13.5 References
1 HOWELL, J. Q.: 'Microstrip antennas', IEEE Trans., 1975, AP-23, pp. 90-93
antenna by one-point feed', Trans. IECE Japan, 1981, J64-B, pp. 225-231 (in Japanese)
I I RODGERS, A,: 'Wideband squintless linear arrays', Marconi Rev., 1972, 35, pp. 221-243
12 HANEISHI, M., YOSHIDA, S., and GOTO, N.: 'A broadband microstrip array composed
of single-feed type circularly polarised microstrip antennas'. IEEE AP-S Int. Symposium
Antennas and Propagat. Digest, 1982, pp. 160-163
13 HUANG, J.: 'A technique for an array to generate circular polarisation with linearly polarised
elements', IEEE Trans., 1986, AP-34, pp. 1113-1 124
14 TESHIROGI, T., TANAKA, M., and CHUJO, W.: 'Wideband circularly polarised array
antenna with sequential rotations and phase shifts of elements'. Proc. Int. Symposium on
Antennas and Propagat., Japan, 1985, pp. 117-120
15 ITOH, K., ARIGA, T., and SHINADA, H.: 'On some design examples of a circularly
Trans. IECE Japan, 1982, J65-B, pp. 1385-1392 (in Japanese)
16 ITO, K., AIZAWA, N., and GOTO, N.: 'Circularly polarised printed array antennas com-
posed of strips and slots', Electron. Lett., 1979, 15, pp. 811-812
17 ITO, K.: 'Circularly polarized printed array antenna composed of end-fed strip dipoles and
slots', Electro. & Commun. Japan, 1984, 67-B, pp. 56-69
18 LEWIN, L.: 'Radiation from discontinuities in strip-line', IEE Proc., 1960, 107C, pp. 163-170
19 HALL, P. S.: 'Rampart microstrip line antennas'. European Patent Application 79301340.0,
1979
20 WOOD, C., HALL, P. S., and JAMES, J. R.: 'Design of wideband circularly polarised
microstrip antennas and arrays'. 1st IEE Int. Conference on Antennas and Propagat., London,
1978, pp. 312-316
21 HENRIKSSON, J., MARKUS. K., and TIURI, M.: 'A circularly polarised travelling-wave
chain antenna'. Proc. 9th European Microwave Conference, Brighton, 1979, pp. 174-178
22 MAKIMOTO, T., and NISHIMURA, S.: 'Circularly polarised microstrip line antenna'. US
Patent 4 398 199, 1983
23 NISHIMURA, S., SUGIO, Y., and MAKIMOTO, T.: 'Crank-type circularly polarised
microstrip line antenna'. IEEE AP-S Int. Symposium Antennas and Propagat. Digest, 1983,
pp. 162-165
24 HALL, P. S.: 'Microstrip linear array with polarisation control', IEE Proc., 1983, 130H, pp.
215-224
25 WOLFF, E. A.: 'Antenna analysis' (John Wiley, NY 1966) Chap. 8
 872 Circularly polarised antenna arrays
26 DOUVILLE, R. J. P., and JAMES, D. S.: 'Experimental study of symmetric microstrip bends
and their compensation', IEEE Trans., 1978, MTT-26, pp. 175-182
27 DUBOST, G.: 'Broadband circularly polarized flat antenna'. Proc. Int. Symposium on Anten-
nas and Propagat., Japan, 1978, pp. 89-92
28 DUBOST, G., SAMSON, J., and FRIN, R.: 'Large-bandwidth flat cylindrical array with
circular polarisation and omnidirectional radiation', Electron. Lett., 1979, 15, pp. 102-103
29 ITO, K., AIZAWA, N., and GOT0,N.: 'Circularly polarized microstrip antennas'. Report of
Technical Group, IECE of Japan, AP78-90, 1978, pp. 21-26 (in Japanese)
30 NAKAOKA, K., ARM, K., and ITOH, K.: 'Some problems on circularly polarised micro-
strip line slot arrays'. National Conv. Records on Optical and Wave Sect., IECE of Japan,
1982, S2-15 (in Japanese)
31 OLTMAN, H. G., and HUEBNER, D. A,,: 'Electromagnetically coupled microstrip dipoles',
IEEE Trans., 1981, AP-29, pp. 151-157
32 ITOH, K., BABA, H., OGAWA, Y., WATANABE, F., and YASUNAGA, M.: 'L-band
airborne antenna using crossed slots'. Report of Technical Group, IECE of Japan, AP85-101,
1986, pp. 65-72 (in Japanese)
33 HENDERSON, A,, and JAMES, J. R.: 'Low-cost flat-plate array with squinted beam for DBS
reception'; IEE Proc., 1987, 134H. pp. 509-514
34 JAMES, J. R., and HALL, P. S.: 'Microstrip antennas and arrays. Pt. 2: New array-design
technique', IEE J. MOA, 1977, 1, pp. 175-181
35 JAMES, J. R., HALL, P. S., and WOOD, C.: 'Microstrip antenna theory and design' (Peter
Peregrinus, 1981)
36 COLLIN, R. E., and ZUCKER, F. J.: 'Antenna theory: Pt. 1' (McGraw-Hill, 1969) chap. 14
37 ITO, K., and KITAJIMA, H.: 'Design of series-fed circularly polarised printed array antenna'.
4th IEE Int. Conference on Antennas Propagat., Coventry, 1985, pp. 103-107
38 ITO, K., ITOH, K., and KOGO, H.: 'Improved design of series-fed circularly polarised printed
linear arrays', IEE Proc., 1986, 133H, pp. 462-466
39 ITO, K., ITOH, K., OHTAKE, T., and KOGO, H.: 'Circularly polarized printed planar array
composed of strip dipoles and slots'. IEEE AP-S Int. Symposium Antennas and Propagat.
Digest, 1986, pp. 561-564
40 KOBAYASHI, A., ITO, K., and BAN, M.: 'A precise measurement of input immittances of
elements coupled to a microstrip line (Continuation)'. Report of Technical Group, IECE of
Japan, MW85-106, 1985, pp. 25-30 (in Japanese)
41 NISHIMURA, S., SUGIO, Y., and MAKIMOTO, T.: 'Side-looking circularly polarised
microstrip line planar antenna'. Proc. Int. Symposium on Antennas and Propagat., Japan,
1985, pp. 129-132
42 POZER, D. M.: 'Input impedance and mutual coupling of rectangular microstrip antennas',
IEEE Trans., 1982, AP-30, pp. 1191-1 196
43 MALKOMES, M.: 'Mutual coupling between microstrip patch antennas', Electron. Lett.,
1982, 18, pp. 520-522
44 JEDLICKA, R. P., POE, M. T., and CARVER, K. R.: 'Measured mutual coupling between
microstrip antennas', IEEE Trans., 1981, AP-29, pp. 147-149
45 HANEISHI, M., YOSHIDA, S., and TABATA, M.: 'A design of back-feed type circularly
polarized microstrip disk antennas having symmetrical perturbation element by one-point
feed', Trans. IECE Japan, 1981, 564-8, pp. 612-618 (in Japanese)
46 BAILEY, M. C., and PARKS, F. G.: 'Design of microstrip disk antenna arrays'. NASA
Technical Memorandum 7863 1, 1978
47 HANEISHI, M.: 'Studies on circularly polarised microstrip antennas'. Doctoral Thesis,
Tokyo Inst. Tech., 1981 (in Japanese)
48 HENDERSON, A., and JAMES, J. R.: 'Design of microstrip antenna feeds. Pt. I: Estimation
of radiation loss and design implications', IEE Proc, 1981, 128H. pp. 19-25
49 JAMES, J. R., and WILSON, G. J.: 'Microstrip antennas and arrays. Pt. 1: Fundamental
action and limitations', IEE J. MOA, 1977, 1, pp. 165-174
Circularly polarised antenna arrays 813
50 HUANG, J.: 'The finite ground plane effect on the microstrip antenna radiation patterns',
IEEE Trans., 1983, AP-31, pp. 649-653
51 HALL, P. S., and JAMES, J. R.: 'Crosspolarisation behaviour of series-fed microstrip linear
ar ays', IEE Proc., 1984,13lH, pp. 247-257
52 HANSEN. R. C.: 'Cross polarization of microstrip patch antennas', IEEE Trans., 1987,
AP-35, pp. 731-732
53 COLLIER. M.: 'Microstrip antenna array for 12GHz TV', Microwave J., 1977,20, pp. 67-71
54 MURPHY, L. R.: 'SEASAT and SIR-A microstrip antennas'. Proc. Workshop on Printed
Antenna Technology, New Mexico State Univ., 1979, pp. 18/1-20
55 HANEISHI, M., HAKURA, Y., SAITO, S., and HASEGAWA, T.: 'A low-profile antenna
for DBS reception'. IEEE AP-S Int. Symposium Antennas and Propagat. Digest, 1987, pp.
914-917
56 PUCEL, R. A., MASSE, D. J., and HARTWIG, C. P.: 'Losses in microstrip', IEEE Trans.,
1968, MTT-16, pp. 342-350
57 LEWIN, L.: 'Spurious radiation from microstrip', IEE Proc., 1978, 125, pp. 633-642
58 MURATA, T., and OHMARU, K.: 'Characteristics of circularly polarised printed antenna
with two layer structure'. Report of Technical Group, IECE of Japan, AP86-101, 1986, pp.
83-87 (in Japanese)
59 STOCKTON, R., and HOCKENSMITH, R.: 'Application of spherical arrays - A simple
approach'. IEEE AP-S Int. Symposium Antennas and Propagat. Digest, 1977, pp. 202-205
60 MAILLOUX, R. J., MCILEVENNA, J. F., and KERNWEIS, N. P.: 'Microstrip array
technology', IEEE Trans., 1981, AP-29, pp. 25-37
61 HORI, T., TERADA, N., and KAGOSHIMA, K.: 'Electronically steerable spherical array
antenna for mobile earth station'. 5th IEE Int. Conference on Antennas and Propagat., York,
1987, pp. 55-58
62 HORI, T.: 'Mutual coupling between broadband microstrip antennas'. National Conv. Re-
cords, IECE of Japan, 1984, p. 714 (in Japanese)
63 CARVER, K. R., and MINK, J. W.: 'Microstrip antenna technology', IEEE Trans., 1981,
AP-29, pp. 2-24
64 TESHIROGI. T.. and GOTO, N.: 'Recent phased array work in Japan'. ESAICOST 204
Phased-Array Antenna Workshop, 1983, pp.-37-44
CHEN, C. H., TULINTSEFF, A., and SORBELLO, R. M.: 'Broadband two layer microstrip
antenna'. IEEE AP-S Int. Symposium Antennas and Propagat. Digest, 1984, pp. 251-254
TAGA, T., MISHIMA, H., and KANEHORI, T.: 'A broadband microstrip antenna at U H F
band'. National Conv. Records, IECE of Japan, 1979, pp. 254-255 (in Japanese)
HORI, T., and NAKAJIMA, N.: 'Broadband circularly polarised microstrip array antenna
with co-planar feed', Trans. IECE Japan, 1985, J68-B, pp. 515-522 (in Japanese)
LONG, S. A,, and WALTON, M. D.: A dual-frequency stacked circular-disk antenna', IEEE
Trans., 1979, AP-27, pp. 270-273
YASUNAGA, M., WATANABE, F., SHIOKAWA, T. and YAMADA, M.: 'Phased array
antennas for aeronautical satellite communications'. 5th IEE Int. Conference on Antennas and
Propagat., York, 1987, pp. 47-50
GOTO, N., and KANETA, K.: 'Ring patch antennas for dual frequency use'. IEEE AP-S Int.
Symposium Antennas and Propagat. Digest, 1987, pp. 944-947
HANEISHI, M., SAITO, S., YOSHIDA, S., and GOTO, N.: 'A circularly polarized planar
arrays composed of the microstrip pairs element'. Report of Technical Group, IECE of Japan,
AP 83-64, 1983, pp. 1-4 (in Japanese)
HUANG, J.: 'Circularly polarised microstrip array with wide axial ratio bandwidth and single
feed L.P. elements'. IEEE AP-S Int. Symposium Antennas and Propagat. Digest, 1985, pp.
705-708
TESHIROGI, T., TANAKA, M., and OHMORI, S.: 'Airborne phased array antenna for
mobile satellite communications'. IEEE AP-S Int. Symposium Antennas and Propagat.
Digest, 1986, pp. 735-738
BALLING, P,: 'Design and analysis of contoured-beam reflector antenna feed arrays and
contoured-beam array antennas'. JINA, 1986, Nice, pp. 315-329
 Chapter 14

Microstrip antenna feeds

R. P. Owens

14.1 Introduction

When choosing the most appropriate microstrip antenna configuration for a

particular application, the means of excitation of the radiating element is an
essential and important factor which requires careful consideration. A wide
variety of feed mechanisms is available, not just for coupling energy to
individual elements, but also for the controlled distribution of energy to a linear
or planar array of elements. The feed system may be either co-planar with the
radiating elements, or situated in a separate transmission-line layer.
Several publications have surveyed many possible types of microstrip antenna
feed [l-71. In the present discussion covering the most important types, those
aspects which are of particular practical interest to the designer wilI be covered,
emphasising available options and including references to the latest advances in
feed design. For the sake of descriptive economy and consistency, it will be
assumed, unless otherwise stated, that the antenna is radiating rather than
receiving energy.
The feed network win in general have certain undesirabIe characteristics
which must be carefully monitored in order to minimise any adverse effects on
the array performance. For example, attenuation due to conductor loss and
dielectric loss will occur in a stripline feed, and this win reduce the efficiency, and
hence the gain, of the antenna array.
The conductor loss cc,depends specificaIly on the type of striplke considered,
but it is always proportional to the length of the line and the surface resistivity,
R,, of the conductor. R, = 20n& o where f is the operating frequency and a,
is the conductivity. In general, a, also increases with increasing line impedance
and substrate permittivity, and decreases with increasing substrate thickness.
Thedielectric loss a, is proportional to the line length, the frequency, and the
dielectric loss tangent, tan 6, and it increases with increasing substrate permittiv-
ity.
816 Microstrip antenna feeds
Such features as tapered or stepped impedance transformers, bends, junc-
tions, branches, transitions and terminations will introduce electrical and physi-
cal discontinuities into the feed line. In practice it is rarely possible to eliminate
the electrical effects completely by normal matching techniques, and as a result
reflection losses will occur. In addition, such discontinuities in microstrip feeds
will cause surface-wave loss and spurious radiation. The latter will, in general,
be uncontrolled, and is likely to increase co-polar sidelobe levels in some
directions, and to increase the total energy in the cross-polar radiation pattern,
thereby reducing the antenna gain. Theoretical investigations of surface-wave
and radiation losses associated with some common discontinuities have been
carried out [3, 8-10, 391.
Direct radiation losses and surface-wave losses are eliminated in enclosed
triplate and suspended stripline feeds, but any discontinuity causing asymmetry
in the cross-section, such as a probe feed to a patch, will introduce losses due
to the transfer of energy to a parallel-plate mode propagating between the
ground planes. This energy is free to couple to adjacent probes, and may thus
ultimately result in spurious radiation. The mode can be strongly attenuated by
the use of mode-suppressing pins close to the discontinuity, or by means of
microwave-absorbent film or sheet material.
The following discussion will concentrate mainly on feeds either for conven-
tional resonant microstrip patch radiators, or for the open-end stub radiators
which are used in the comb array. However, the series-array design procedure
covered in Section 14.3.2.3 could also be applied to microstrip radiating systems
such as the serpent, chain, Franklin and rampart arrays which are discussed in
Section 14.3.4. In these arrays, radiation takes place from abrupt changes in the
direction of the feed line itself.
Two distinct types of array feed system can be distinguished. First, there is the
parallel or corporate feed, which has a single input port and multiple feed lines
in parallel constituting the output ports. Each of these feed lines terminates a t
an individual radiating element, and therefore transfers all its energy into the
element.
The second type of feed system is the series feed, which usually consists of a
continuous transmission line, from which small proportions of energy are
progressively coupled into the individual radiating elements by various means.
This constitutes a travelling-wave array if the feed line is terminated in a
matched load, or a resonant array if the termination is an open- or a short-
circuit.
The coupling mechanisms available for achievi~geither full or partial transfer
of energy to microstrip radiators are discussed in some detail in Section !4.2.
Some of the more common parallel and series feed systems are considered in
Section 14.3, together with some basic design procedures.
It will become clear from Section 14.3 that power dividers and combiners are
crucial components in the design of some series feeds and most corporate feed
systems. Section 14.4 is concerned with some basic characteristics and limita-
tions of commonly used direct-coupled stripline power dividers/combiners.
Microstrip antenna feeds 817
In Section 14.5, some alternative transmission lines for feed systems are
briefly discussed, followed by a survey of some specialised feeds for various
types of active and passive multiple-beam-forming networks.
14.2 Coupling to microstrip patches
14.2.1 Co-planar coupling to a single patch
14.2.1.1 Gap-coupledpatch: Coupling to a microstrip patch may be achieved
by means of a narrow gap between the feed line and the resonant patch, as
shown in Fig. 14.1~.The width of the gap dictates the strength of the coupling
at the resonant frequency. When the feed line and the resonant patch are
critically coupled, the latter constitutes a matched termination. An approximate
equivalent circuit is shown in Fig. 14.lb. For a wide rectangular patch, the
radiation conductance G, is given approximately [2, 3, 71 by
where Weis the width of the equivalent waveguide model for a microstrip line
of width W [3].
14.2.1.2 Direct-coupledpatch: If the feed line is directly coupled to the patch,
critical coupling a t the resonant frequency may be achieved by one of the two
configurations shown in Fig. 14.2. Fig. 1 4 . 2 ~shows a quarter-wave transformer
of impedance ZT between the feed line and a rectangular patch [7]. The input
impedance to a half-wave resonant patch is equal to R,/2, where R, = 1/G,.
Thus
Fig. 14.2b shows an inset feed arrangement which artificially moves the feed
point to a lower impedance region inside the patch [7]. For a half-wavelength
resonant rectangular patch, the input impedance Z, at a distance I from either
radiating edge is given by
where /l= 241,; 3, being the wavelength in the microstrip line.
F D a~quarter-wave short-circuited rectangular patch, the input impedance is
independent of Z,:
Z, = R,/(l + tan2(80) (14.4)
818 Microstrip antenna feeds Microstrip antenna feeds 879
Recently, the case illustrated in Fig. 1 4 . 2 has
~ been investigated, where the feed microstrip
feed line section
line enters at a point about one third of the way along a non-radiating edge [I I].

/[i
Shorter feed lines with lower loss may be possible using this configuration in a
microstrip

fz;< cOu$$g I I I
I I I

?!@$.,
f 20
feed line gap resonant patch

Fig. 14.1 Gap-coupled patch

a Patch
b Eauivalent-circuit

corporate feed network. The cross-polarised radiation is relatively high, but it

can be minimised by optimising the aspect ratio W/Lof the patch at about 1.5.
Two-port coupling at the non-radiating edges has also been tried. When applied
to series-fed linear arrays, it has the advantage that the power radiated by the
two-port patch can be controlled by adjusting the distance of the input and
output ports from a radiating edge 1121.

14.2.2 Series-array co-planar coupling

14.2.2.1 Proximity coupling: This method is used for coupling a single feed
line to a linear array of resonant patches. Although gap coupling is involved, in
some applications the patches couple over a significant distributed length of line,
so proximity coupling is a more appropriate description.
In an array configuration, the individual patches do not necessarily need to Fig. 14.2 Direct-coupled patch
be matched to the feed line, neither do they have to operate at maximum a Quarter-wave matched feed with equivalent circuit
efficiency. There are two consequences of this: the first is that the coupling gaps b Inset feed with equivalent circuit
c Non-radiating edge feed
 820 Microstrip antenna feeds
(el
Fig. 14.3 Proximity-coupled arrays
a End-coupled parasitic patches
b 45' inclined parasitic patches
c Herringbone array of circularly polarised patch pairs
d Edge-coupled parasitic patches
e Microstrip patches coupled to dielectric insular guide
Microstrip antenna feeds 821
can be varied to control the proportion of power coupled into the patches, and
the second is that the patches themselves can have characteristic impedances
rather higher than those normally associated with more conventional low-
impedance patches.
Some examples of proximity-coupled patch arrays are shown in Fig. 14.3.
Example (a) in this Figure shows discrete gap coupling to patches with a
relatively large aspect ratio, and hence fairly high impedance [13]. The gap
widths can be theoretically or empirically related to the radiation conductance
of the patches, although no design data have been published. The methods of
Section 14.3.2 can be applied to produce a tailored aperture distribution across
the array. Quite strong coupling can be achieved with narrow gaps, but there is
an ultimate limit set by the etching accuracy for the particular substrate,
metallisation thickness and line-width combinations. There is effectively no limit
to the lowest coupling available from very wide gaps.
Fig. 14.4 Direct-coupled arrays
a Comb-line array
b Cascaded patch array
Using the nomenclature of waveguide arrays, both transposed arrays (as
shown in Fig. 14.3a) and untransposed arrays of this type can be designed. The
radiated E-field polarisation is parallel to the length of the patch, i.e. transverse
to the axis of the feed line. A variant of this type is shown in fig. 14.36, where
the patches are inclined at 45" to the feed-line axis [13, 141. The polarisation is
likewise inclined at this angle. An interesting application of this variant is shown
in Fig. 14.3~.This is a transposed array of 4.5" inclined patches in which the
lower array is displaced by one quarter-wavelength relative to the top array. As
a result, the patch pairs shown in the Figure each form a self-matched circularly
polarised element [13, 141.
 822 Microstrip antenna feeds
In Fig. 14.3d, the resonant patches are capacitively coupled over their whole
length [15, 31. Again, the width of the gap dictates the strength of coupling, and
hence the radiation conductance of the patch. The polarisation in this case is
longitudinal, and the excitation is such that only untransposed arrays are
possible.
In one interesting variant of this category of feeds, the microstrip line is
replaced by a dielectric image guide, or insular guide, as shown in Fig. 14.3e.
This arrangement has been employed in millimetre-wave antennas at frequen-
cies where conventional microstrip lines would contribute excessive losses [16].
14.2.2.2 Direct coupling: Two particular examples of this type of coupling
will be considered; the comb-line array [3, 171, and the cascaded patch array [2,
181, both illustrated in Fig. 14.4. The comb-line structure of Fig. 1 4 . 4 ~is seen
to bear some resemblance to the gap-coupled array of Fig. 14.3a, but in this case
the radiation conductance of the open-circuited stub is determined by its width,
as indicated in Section 14.2.1.1 [2, 31.
The maximum conductance available for a given feed-line width is dictated by
a maximum stub width of about one half-wavelength. The use of a transposed
array overcomes this limitation to some extent by spreading the amount of
power to be radiated over twice as many elements within a given array length.
The minimum conductance available is governed by the practical limitations on
etching sufficiently narrow lines.
The direct-coupled array, an example of which is shown in Fig. 14.46, consists
of a cascaded series of low-impedance resonant patches, linked by sections of a
high-impedance feed line. The required aperture distribution is obtained by
appropriately varying the widths of the resonant patches [2].
14.2.3 Probe coupling
This method of coupling, illustrated in Fig. 14.5, has been widely analysed in the
literature, particularly for circular patches [19-241. It has the advantage that the
patch, ,pin
connector
Fig. 14.5 Probe-coupled microstrip patch
feed lies behind the radiating surface, and therefore does not itself contribute
unwanted radiation. It is a very convenient method of feeding a single patch by
means of a surface-mounted coaxial connector attached to the microstrip
ground plane, for instance for experimental purposes. The probe is positioned
at a point where the input impedance of the patch, Z,", is equal to the charac-
teristic impedance of the coaxial feed line. Eqns. 14.3 and 14.4 may be used to
determine Z , for a rectangular patch.
Microstrip antenna feeds 823
The disadvantage for an array system is that the feed network must lie in a
separate layer behind the radiating surface, so the complete antenna cannot be
etched on a single substrate. There is a consequent increase in complexity, but
this may be an acceptable penalty to pay for a resulting increase in design
flexibility. An additional disadvantage, particularly at high frequencies, is that
the necessity for inserting properly secured probes results in extra mechanical
complexity and increased manufacturing costs, particularly for large arrays. The
feed network beneath may be constructed in a variety of transmission lines, e.g.
microstrip, triplate or suspended stripline, provided that an adequate right-
angle transition into the microstrip patch can be designed.
The inductive reactance of the probe adversely affects the VSWR bandwidth
in this type of coupling, particularly if a thick, low-permittivity substrate is used
in order to increase the gain bandwidth. The VSWR bandwidth can be
broadened by adding series capacitive-reactance compensation as close as poss-
ible to the inductance, to provide a series circuit resonant at the same frequency
as the patch [25]. This may be done within the feed network behind the patch
[25], within the probe itself [26], or in the form of a small annular gap on the
surface of the patch [27]. The latter method results in high cross-polar levels due
to the asymmetrical gap, but these can be significantly reduced by introducing
a second compensated feed diametrically opposite the first.
An additional short-circuiting pin is often placed through the centre of a
circular microstrip patch, providing a DC connection between the patch and the .
ground plane [19]. This has the dual advantage of earthing the patch, and
suppressing some higher-order modes which would otherwise exist on the patch.
14.2.4 Aperture coupling
Coupling via an aperture in a common ground plane between the radiating
microstrip layer and a stripline feed layer is an attractive alternative to probe
coupling. No extra components or assembly processes are needed; the only
requirement is that the common ground plane should contain etched apertures
accurately positioned below the microstrip patch and above the feed line [28-
301.
The microstrip feed system shown in Fig. 14.6 provides stronger coupling
than a similar triplate or suspended stripline system because of the higher
concentration of fields above the feed line where the aperture is positioned.
Furthermore, a relatively high-permittivity substrate can be used if required for
the feed system, without compromising the radiating properties of the lower-
permittivity substrate carrying the microstrip patches.
Both circular and rectangular apertures have been investigated theoretically
and experimentally, to determine the relative influence on coupling factors of
the aperture dimensions and the feed line and patch overlap parameters, and
good agreement between theory and experiment has been obtained.
A dual aperture-fed patch has been reported, with 18 dB isolation between the
two ports [31]. If both apertures are excited via a quadrature feed, circular
 824 Microstrip antenna feeds Microstrip antenna feeds 825
polarisation can be obtained from a square or circular patch. Alternatively, if of the gap-coupled parasitic patch array shown in Fig. 14.3, can be designed in
a single-feed circularly polarised patch were used [20], dual circular polarisation which the coupling is varied along the array by laterally displacing the elements
would be possible with a dual-aperture feed [32]. relative to the feed line beneath them.
subst rates
substrates feed line

/ ground-plane
/

crucifor
patch
Fig. 14.6 Aperture-coupledpatch antenna (Reproduced from Pozar [28] with permission of
IEE)

A series-fed aperture-coupled array could be made in which the coupling to

each patch is controlled by varying the size of the aperture.
A further application relevant to planar arrays with stacked feed substrates
involves aperture coupling from a microstrip feed perpendicular to the array
plane. The use of a straight feed line [33] and a proximity feed line [34] have been
reported.
Fig. 14.7 Electromagnetically coupled patches
14.2.5 Electromagnetic coupling a Collinear coupled patch (Based on [35] with permission of IEEE @ 1981 IEEE)
This type of coupling, despite the somewhat broad definition implied by its title, b Circularly polarised coupled patch
is by convention taken to refer to the overlaid resonator configuration illu-
strated in Fig. 1 4 . 7 ~
[35]. The feed system is a covered microstrip network, and
the radiating elements are etched onto the covering substrate immediately above 14.3 Parallel and series feed systems
the open-ended feed lines. The elements are thus parasitically coupled to the feed
network. They may be regarded as microstrip patches on a double-thickness 14.3.1 Parallel feeds for one and two dimensions
substrate sharing a common ground plane with the feed network. Successful
feed systems have been developed using this type of coupling, but, like the 14.3.1.1 One-dimensionalpara[[e[feed: Fig. 14.8 shows a simple one-dimen-
co-planar feed systems, they can suffer from spurious radiation from the feed sional parallel feed system for a linear array, consisting of a branching network
network itself. As Fig. 14.7b shows, circular polarised radiation can be obtained of two-way power dividers. The power-divider ratios may be chosen in order to
by suitably exciting a cross-shaped patch [35]. A series-fed system, reminiscent achieve a particular aperture distribution across the array. If the distances from
 826 Microstrip antenna feeds
the input port io each of the output ports are identical, and the radiating
elements are identically coupled to thse ports, the resulting radiated beam will
be in the broadside direction. The beam position will be independent not only
Fig. 14.8 One-dimensionalparallel feed network
aperture plane
main bea
ye
direction
Fig. 14.9 Networks for producing inclined beams
a Beam squint by line extension
b Beam squint by offset power divider
of the frequency, i.e. it will be 'squintless', but also of the spacing between the
elements. This spacing can thus be chosen to meet other criteria. For example,
it may be desirable to increase the spacing, up to a limit dictated by the onset
of grating lobes, in order to minimise the complexity of the feed network by
ensuring that a particular number of elements fits within the available aperture.
Alternatively, it may be important to maximise the gain from a given aperture
by choosing one half-wavelength spacing between the elements; or maybe a
favourable spacing with regard to mutual-coupling effects will be preferred.
A parallel feed may also be used to produce a radiated beam inclined at an
angle to the broadside direction. Fig. 14.9 shows two alternative ways of
Microstrip antenna feeds 827
achieving this. In Fig. 14.9a, an incremental phase difference 6 between the
elements is obtained by progressively extending the line lengths to successive
elements by the distance A1 = (6/2n)Il, where A, is the wavelength in the
transmission line. The same result is obtained by offsetting the power-divider
T-junction by multiples of A1/2, as shown in Fig. 14.96.
The general equation relating the main-beam angle 8, to phase difference 6,
free-space wavelength 1, and element spacing d is as follows:
6 -
sin $ = - 10
2n d
It will be observed that, in the general case, the beam angle is no-longer
squintless; neither is it independent of d. It is of interest to determine the range
of parameters for which the peak of the first possible grating lobe lies outside
real space. The relevant formulae are as follows:
General phase equation for beam at f? degrees is:
kd sin 0 - 6 = 2Kn (14.6)
where k = 27r/A0,and K is an integer denoting the order of the mode.
This reduces to the main beam equation 14.5 when K = 0.
The general equation for grating-lobe angle f? = 0, is thus:
sin 8, = sin 8, + Mold (14.7)
If K = - 1 the grating lobe will not exist if d / l , < l/(sin 0, + 1).
If K = + 1 the grating lobe will not exist if d/A, i 1/(1 - sin 0,)
Another variant of the basic parallel feed pattern of Fig. 14.8 is represented
by Fig. 14.10. Here, some of the two-way power dividers are removed, but the
Fig. 14.10 Asymmetrical parallel feed network
line lengths to each element are controlled as before to provide the desired phase
front. The main advantages of this variant are twofold. First, the network can
be readily configured for even numbers of elements which are not powers of two,
and for odd numbers of elements. Secondly, it gives greater flexibility in provid-
ing tapered aperture distributions which would otherwise necessitate incon-
veniently large power-divider ratios in the conventional branching network.
 828 Microstrip antenna feeds
14.3.1.2 Two-dimensional parallelfeed: Fig. 14.1 1 shows the schematic lay-
out for a two-dimensional corporate feed with four elements per side. This form
of layout can be extended to larger arrays with, specifically, 2Nelements per side.
Any attempt to modify the feed to deal with other even numbers can result in
Microstrip antenna feeds 829
under these conditions, but the cross-polar components in the radiation pattern
are out of phase, and are suppressed over a range of frequencies. The axial-ratio
bandwidth is thus significantly increased.
Fig. 14.12 shows schematically the feed configuration required for a right-
hand-polarised group of patches. As may be deduced from the line lengths
shown, a by-product of this technique is that, around the design frequency, the
mismatches due to the coupled patches tend to cancel out at the input port.
Consequently the bandwidth over which the input VSWR is less than some
specified value, say 2: 1, should also increase. A relatively minor counteractive
effect is that, because the line lengths are unequal, the phase differential between
the four patches will be frequency-dependent.

14.3.1.3 Radiation-pattern prediction for arrays fed in paralleI: The details of

array theory may be obtained from other sources [40, 411. However, some
simple basic principles apply, and these are summarised here. Considering a
linear array first, as shown in Fig. 14.13, the far field due to an array of N

Fig. 14.11 Two-dimensional 4 x 4 corporate feed

the need for fairly large power-divider ratios. Nevertheless, there is a basic
symmetry which reduces the number of different power dividers needed. If there
is an odd number of elements per side, this symmetry is lost. The considerations
regarding element spacing, aperture distribution and beam angle are the same N
element no. 1 2 n N-l
as those discussed for the linear array.
Recently, the technique termed 'sequential rotation' has been applied specific-
am Litude of
ally to two-dimensional arrays of circularly polarised microstrip patches with r a J a t e d field A2
single feed points [36-391. As shown in Fig. 14.12, one application of the

distance dl d2 dn
from reference
point

phase on 61 62 6n
element

Fig. 14.1 3 Parameters for parallel-fed linear array

Fig. 14.12 Feedforsequentialrotation: schematic (Copyright @ Controller HMSO, London,

1986. Reproduced with permission) elements disposed along a straight line is given by
f('M8)
=
technique involves rotating each element in a 2 x 2 array of such patches
through 90" relative to its immediate neighbours. The line lengths in the feed to where the array factor is
the group of elements are adjusted to compensate for the 90" phase change N

caused by this rotation. The co-polar radiation from this array is unaffected f (8) = 1I A, exp (jk(d,, - d , ) sin 0 - 6,)
n=
830 Microstrip antenna feeds Microstrip antenna feeds 831
g(0) = element factor In many rectangular arrays, it is possible to separate the array factors in the two
A, = radiated-field amplitude of element n orthogonal principal planes [40]. In such a case, the far-field radiation pattern
d,, = distance to element n from some reference point of an M x N array with equal element spacings dm and d,, respectively, is given
6, = phase on element n by
The radiated-field amplitudes are chosen from one of the many distributions
E(0, 4 ) = g(0, 4 ) f ( 4 f Oil (14.12)
available. Apart from the uniform case, commonly used distributions are the
Taylor one- and two-parameter, and the Dolph-Chebyshev [40,41]. The choice where g(0, 4 ) is the element factor. The separate array factors in the principal
is dictated by the particular radiation-pattern characteristics required. planes are given by:
Array elements are normally uniformly spaced, but random spacing and
sparse arrays may be accounted for in this formula.
The phase components 6, are primarily dictated by the relative lengths of the
f(x) = zM

m= I
A, exp[j(mkd, sin 0 cos 4 - &,)I (14.13)

feed lines to each element, but may also include a factor due to the relative N

orientation of the elements. f ( y) = 1A, exp[j(nkd,, sin 0 sin 4 - ti,)] (14.14)

The element factor for rectangular patches is given by the following two
approximate formulas, corresponding to the two major axes of the patches [2,
31. Fig. 14.14 shows the co-ordinate system applied to the patch.
n= I
The far-field radiation pattern may therefore be derived in any plane, provided
that the element factor can be defined in that plane.
The directivity of a linear array of isotropic sources is given by the following
formula [42]:

The directivity of a two-dimensional array is separable, such that:

D(x, y) = nD(x)D(y) cos 0, (14.16)
where D(x), D(y) are the separate directivities of the orthogonal linear arrays.
In a practical microstrip array, the overall gain, measured in decibels, will be
made up of several factors:

where
Fig. 14.14 Patch co-ordinate system
q = aperture efficiency ( < 1)
G, = gain of a single element
E-plane ( 4 = n/2) a = feeder loss in dB per wavelength
L = length of the feeder in wavelengths
M represents miscellaneous discontinuity losses as discussed in Section 14.1.
The dominant first term in eqn. 14.17 increases logarithmically with array
size, but the feeder-loss term increases linearly. The gain will therefore reach a
maximum for a particular size of array.
sin (xnW .
sm 0) For example, an N x N array with half-wavelength element spacing will have
. length L will be close to N/2 wavelengths.
a directivity D(x, y) of n ~The~ feeder
= cos 0
ILW .in F,
 832 Microstrip antenna feeds
G will be a maximum when the differential of eqn. 14.17 with respect to N is
equal to zero. The differential is given by 20 log,,(e)/N - a/2, where e is the
exponential. So, for maximum gain, N will be the nearest integer to 17.371~.
14.3.2 Series feed for one dimension
Microstrip antenna feeds 833
Figs. 14.16~and b illustrate the two types in the form of two different
comb-array designs. Some microstrip arrays can be of one type only; for
example the serpent array [44] in Fig. 14.16~is a transposed array, and the
longitudinal parasitic patch array [15] (Fig. 14.16d) is an untransposed array.

14.3.2.1. Travelling-wave feed: The travelling-wave feed is commonly used I

for arrays in which the beam is required to be inclined away from the broadside
direction. In its simplest form, as shown in Fig. 14.15a, the radiating elements
are partially coupled to a continuous feed line by one of the methods described
---.
in load
-----t

in Section 14.2.2. The feed line is terminated in a matched load. This arrange-
ment has the disadvantage that the beam will squint with frequency, but, if
necessary, line-length compensation between the couplers and the elements can
be introduced to produce the 'squintless' series feed, asshown in Fig. 14.156
WI.

rodiatina elements

power divider network

Fig. 14.15 Lineartravelling-wavearrays

a Directly coupled travelling-wave array
b Squintless travelling-wave array (Reproduced from [3] with permission of IEE) Fig. 14.16 ~ransposedand untransposed arrays
a Untransposed comb array
The equations relating the element spacing to the required beam angle can be b Transposed comb array
divided into two groups. The first group concerns arrays in which there is no c Serpent array (transposed)
d Parasitic patch array (untransposed)
phase change between adjacent elements attributable specifically to the coupling
mechanism between the feed line and the elements. Using the nomenclature of Considering untransposed arrays first, the general equation is:
waveguide slot arrays, this is termed an untransposed array, or a ++
array.
The second group of equations covers the transposed array, or - type, in +
which a phase change of 180" occurs between adjacent elements due to the effect
of the coupling mechanism on the instantaneous phase of the radiated fields. where d is the element spacing.
 834 Microstrip antenna feeds Microstrip antenna feeds 835
K = 0 is not a solution in a stripline medium where /2,/1, > 1. K = - 1 is backward-firing option is free from grating lobes for all main beam angles,
the solution for the principal beam: ~rovided1,/1, > 1.
If both options are otherwise equally possible in an array required to occupy
a given aperture, the forward-firing option may be convenient if the designer
wishes to minimise the number of elements in order to simplify the feed layout.
The element spacing d required to give a beam angle 0, is given by Alternatively, if the array is to be short, it may be advantageous to increase the
d = l0/(1,/1, - sin 8,) number of elements in order to minimise the range of radiation conductances
required. In that case the backward-firing option may be chosen. It should be
If beam angles away from broadside are required, there is the choice of back- borne in mind that the array gain will depend on the spacing of the elements,
ward or forward firing. As the formulas indicate, the element spacing is less for as will the mutual coupling between elements.
backward-firing beams, so more elements are required to fill a given aperture. The amplitude distribution across an array is determined by the strength of
It is essential in either case to know whether grating lobes will exist in real space. coupling to each element. This may be represented as a coupling factor c,
The general grating-lobe equation is: defined as the ratio P,/P,, where PRis the power radiated by the element and
P, is the power incident upon it. Alternatively, the normalised radiation conduc-
sin 8, = (K + I) - - K sin 6,
10
tance g may be used. If matched conditions were to exist beyond each element
1, in the array, g would be defined by the expression g = P,/P,, where P,
K = - 2 is the solution for the first grating lobe, which occurs only if 0, is represents the power transmitted past the element. It is important to note the
positive and sin 8, 2 0.5(/2,/1, - I) distinction between c and g, as their definitions are sometimes erroneously
All of these formulas are concerned only with the effect of phase components interchanged in the literature. Section 14.3.2.3 describes a method for comput-
in the array factor on the main-beam and grating-lobe positions. If the element ing the radiation conductances for a travelling-wave array designed to radiate
factor has some directivity, then this will influence the position and relative with a specified amplitude distribution.
amplitude of any grating lobes in the final radiation pattern. The sidelobe levels The travelling-wave feed suffers from the disadvantage that the element
and positions will be closely dependent on the distribution of radiated am- spacing needed to produce a broadside beam is either one half-wavelength for
plitudes across the elements of the array, as well as their phases. a transposed array, or one wavelength for an untransposed array. In either case,
The corresponding formulas for the transposed array are as follows: the mismatches caused by each element conductance loading the transmission
The general equation for transposed arrays is: line add up in phase, resulting in a large VSWR at the input end of the array
and a consequent reduction in efficiency. This situation may be avoided by
introducing small matching stubs at each individual element. Alternatively, a
lower VSWR has been achieved in a transposed comb array by splitting each
K = 0 is the solution for the principal beam. stub into two elements spaced by one quarter-wavelength, so that each pair is
Therefore, self-matching [3, 171. The herringbone array of Fig. 1 4 . 3is
~ likewise self-match-
ing.
sin 6, = -
10 lo
--
1, 2d 14.3.2.2 Resonant feed: An alternative way of obtaining a broadside beam
or element spacing I without the mismatch problem is to use a resonant feed. In this option, the feed
line is terminated in an open-circuit one half-wavelength beyond the last ele-
d = 1,/(2(&/L, - sin 0,)) ment, or a short-circuit one quarter-wavelength beyond. The element spacings
The general grating-lobe equation is: are as indicated above. If the array is designed so that the sum of the normalised
radiation conductances is equal to unity, the input admittance of the array is
sin 0, = sin 0, (1 - 2K) + 2KA0/1, given by
K = - 1 is the solution for the first grating lobe, which occurs only if 6, is N

positive and if sin 8, 2 2(1,/1, - 0.5)/3. y = cgn+jO = 1 (14.26)

n=l
It will be seen that the transposed array can provide beams over a much wider
range of angles without producing grating lobes, and that in both cases the Within the restriction of this condition, the relative conductance values can be
 836 Microstrip antenna feeds Microstrip antenna feeds 837
freely chosen to given any desired amplitude distribution, provided the maxim- Fig. 14.17 shows the parameters which are involved in the design procedure.
um conductance achievable is not exceeded. It is convenient to choose an initial value for the fractional efficiency factor q,
From a practical point of view, it is an advantage that matched terminations such that
are not needed, particularly if the feed network consists of a microstrip system
coplanar with the radiating elements.
Open-ended lines generally suffice as open-circuits, although they are, of
course, sources of radiation in microstrip feeds. In addition, fringing fields
extend the effective electrical length of the line beyond its physical end. Short-
circuited lines are obtained by means of pins or plated-through holes providing
a conducting path from the feed line to the ground plane.
For all practical purposes, the resonant feed is only useful in the broadside
beam case. Its main disadvantage is that, because of its resonant nature, the
VSWR bandwidth is very narrow.

14.3.2.3 Series-feed design procedure: The following discussion is intended

primarily to apply to travelling-wave array design, but with certain modifica-
tions it can be used for the design of resonant arrays. Mutual coupling is not
taken into account in the analysis, which utilises conventional lossy trans-
mission-line theory available in standard textbooks [45].
The design of a series-fed array involves the calculation of the normalised
radiation conductance required at each element, in order to provide the chosen
amplitude distribution across the array. The radiation conductance is then Fig. 14.17 Power and admittance notation for lossy transmission-line array
translated into an appropriate physical variable of the array by means of a
theoretical or empirical relationship between the two. Ideally, in order to where PRnis the power delivered by the nth element normalised to C(Ai); thus
facilitate subsequent analysis of the designed array at frequencies dose to the (14.28)
design frequency, the Q-factor of the resonant element should be known. PL = (1 - d
The radiation conductance calculation should account for the complex Starting from the load end, the parameters can be calculated iteratively from the
propagation characteristics of the transmission-line feed, and for the effects of following equations, assuming that the elements are resonant at the design
mismatches introduced by the radiating elements. The termination of the feed frequency:
line is modelled as a normalised conductance of unity.
The following groups of parameters are usually defined by the design require-
ment:
(a) Design frequency f,, or wavelength &. The radiation conductance of the nth element is
Beam squint angle, 8,.
Available aperture size, or number of elements, N.
+
(b) Propagation constant of chosen transmission line, y = a: j/3, where a is
the attenuation constant and /3 is the phase constant, 2x11,. where Re(y,) is the real part of y,, and P,, = PL
(c) Parameters of particular radiating element; namely maximum achievable The power incident on the nth element is:
radiation conductance and Q-factor. Ph = p, + PT"
Secondary parameters can be calculated from these; e.g. the spacing between The power-loss ratio [45] is:
elements (normally the spacing is uniform), and the number of elements N
required to fill the aperture, if this has not previously been defined. When N is
known, the radiated-field amplitudes A, for the desired radiation pattern can be
calculated.
(Note that in Reference 45 page 94 there is an error in eqn. 3.93 and in the equation on page 95
because ITle" = =jr,le-"I,nor IrLI).
 838 Microstrip antenna feeds
where the reflection coefficient looking into element n is:
The admittance y, transformed to the load side of element n + 1 is:
The admittance looking into element n is:
Yn = Y / n + gn (14.35)
where y,, = y,.
At the input end, the input reflection coefficient is r,, and the input power
is:
The reflected power is:
The power otherwise unaccounted for is dissipated in the transmission-line:
If any of the calculated radiation conductances exceed the maximum achievable
value, the computations should be repeated with a lower efficiency factor q.
Conversely, if the conductances are relatively small, excessive power is absorbed
in the load, and q can be profitably increased.
Fig. 14.18 shows a typical radiation-conductance profile. The power entering
the feed line diminishes rapidly with distance travelled, so, in order to maintain
a symmetrical amplitude distribution, the element conductances in the second
half of the array must be consistently higher than the corresponding values in
the first half. The maximum-conductance thus occurs about two thirds of the
way along the array.
Once the radiation conductances have been calculated, then, provided the
Q-factor of the elements is known, the array can be analysed at some off-
resonant frequency, to determine the new distribution of radiated power. The
relevant formulas are similar to those used for the design, except that the
conductance is replaced by a complex admittance given by
Then, in eqn. 14.30, the new power radiated by the nth element is given by
Microstrip antenna feeds 839
and eqn. 14.35 is replaced by
Y" = Y h f Yen
(14.41)
14.3.3 Combined feeds
14.3.3.1 Parallel/series one-dimensionalfeed (centre-fed): Fig. 14.19 illustra-
tes two examples of the type of combined feed which comes under this heading.
input element no. Load
end end
Fig. 14.18 Radiation-conductance profile for 12-element array
This type is effectively a centre-fed linear array, of either the travelling-wave or
resonant-feed type. It will be noted that, in example (a), continuity of the phase
front is ensured by offsetting the feed point by one quarter-wavelength from the
centre of symmetry of the array. This is not necessary in example (b).
If a broadside beam is required, the two halves of the array are in other
respects identical; the element spacings are equal on both sides, and the radia-
tion-conductance profile is symmetrical.
To obtain a squinted beam, however, the element spacing must be different
in each half, so that one half produces a forward inclined phase front, and the
other half a backward phase front in the same direction in space.
An important property of the centre-fed array with broadside beam is that the
beam does not squint with frequency. Any change in frequency causes the phase
fronts in the two halves to rotate in opposite directions. The result is that the
beam remains fixed in direction, but becomes broader as a consequence of the
discontinuity in the overall phase front. Ultimately, for larger frequency excur-
sions, the beam will split into two.
14.3.3.2 Two-dimensionalfeeds: Two possible forms of two-dimensional net-
work involving feeds are described here. The example shown in Fig. 14.200 is
the parallel/series type, involving a one-dimensional parallel feed with the
extended output ports forming series feeds. The beam produced by this network
will squint in the plane of the series feeds only. This squint may be eliminated
840 Microstrip antenna feeds
Microstrip antenna feeds 841

by placing two such networks back to back, as shown in Fig. 14.206. Since the
space between the series feeds will be limited, the parallel feeds may have to be
located on a separate transmission-line layer.
The example shown in Fig. 1 4 . 2 0 ~is the serieslseries type, which will exhibit
beam squint in both major planes. Four such networks, fed in parallel, combine
to form a squintless array, centre-fed in both planes as shown in Fig. 14.20d. The
primary feed lines couple to the secondary lines via power dividers. The im-

radiatin
elemen2

Fig. 14.20 Two-dimensional feeds: schematic

( a ) end-fed
Parallel-series feeds
( 6 ) centre-fed
842 Microstrip antenna feeds Microstrip antenna feeds 843
pedance ratios of these may be designed to produce a tapered amplitude
distribution in the plane of the primary lines.
load f The cross-fed array shown in Fig. 14.21 [46-481 is a variant of the serieslseries
type, in which a square array of elements is series-fed from diagonal branch

---t lines coupled directly to a single centre-fed line on the opposite diagonal. All
lines are terminated by radiating elements. The input signal is split four ways at
the centre feed point. In order to produce a coherent broadside beam, the
elements are spaced by one wavelength along the diagonals, or 0.707 wavelength

Fig. 14.21 Cross-fed array (Reproduced from Williams [46] with permission of IEE)

along the sides of the square. The packing density of elements is thus increased
by this configuration. In order to achieve controlled element excitation and
good input VSWR, it is necessary to incorporate quarter-wave impedance
transformers in each feed line, and on each branch of the central cross [47, 481.

14.3.4 Discontinuity arrays

Series-series feeds ('' end-fed
(d) centre-fed
In this category, the microstrip feed line itself is the radiator. Radiation occurs
at a bend or abrupt change in direction of a microstrip line, as a consequence
of imbalance between the fringing-field densities on the outside and inside of the
bend. If the radiating discontinuities on the line occur in certain regular geome-
trical patterns, the phase, amplitude and polarisation of the radiation can be
controlled to form a travelling-wave array.
These structures may be analysed in terms of an effective magnetic-current
source representing the fringing-field imbalance [3]. Alternatively, similar results
for the radiation characteristics of a particular structure may be obtained from
an analysis of the instantaneous surface currents on the line [9]. The discussions
on particular discontinuity arrays that follow reflect the form of analysis used
in the original texts.
Fig. 14.22 is a schematic representation of part of a serpent array, shown
844 Microstrip antenna feeds Microstrip antenna feeds 845

more fully in Fig. 14.16~.The serpent is formed from a series of contiguous

half-sinusoids [44, 49, 501. Radiation is concentrated at the peaks of the sinu-
soids where the curvature is a maximum, and is polarised in a radial direction

Fig. 14.22 Serpent array: schematic

at these peaks. The serpent is therefore a transposed array. The amplitude of the
radiation is proportional to the maximum curvature, which is given by

where s is the span and a the amplitude of the half-sinusoid.

In the simplest case of a constant-amplitude constant-conductance serpent, a
plane wavefront at the desired angle 00,relative to broadside, is obtained if the
following relationship applies:

where I,, is the meander length of the sinusoid.

This is similar to eqn. 14.23, except for the extra factor I,,,/s. For a sinusoid,
this ratio is given by
lm/s = + q) .)E(k)ln (14.44)
where q = ( x a / ~ )k~ ,= ,/(q/(l +
q)),and E(k) is the complete elliptic integral
of the second kind.
Any desired amplitude distribution may be obtained by linking half-sinusoids
with varying amplitudes and spans. The design equation is then more com-
plicated, and must be solved iteratively. Theoretical and experimental studies
have been carried out with a good measure of agreement [49,50], but the design
depends on empirical measurement of radiation conductance as a function of
sinusoid amplitude and span. Two-dimensional serpent arrays have been built
using a one-dimensional corporate feed.
The rampart antenna [2, 3, 511, or, more appropriately, the crank antenna, is
perhaps the most important example of the discontinuity array. It is an extreme-
ly versatile concept which can provide a wide range of radiated polarisations by Fig. 14.23 Rampart-line amys (Reproduced from [ 3 ] with permission of IEE)
appropriate choice of array geometry. One factor dictating the geometry is the a Unit cell
beam angle, but it will be assumed here that a broadside beam is required. The b Circularly polarised array
c Longitudinally polarised array
antenna consists of a cascaded array of unit cells. The general form of the unit
d Transversely polarised array
 846 Microstrip antenna feeds Microstrip antenna feeds 847
cell is shown in Fig. 14.23~.Radiation, polarised diagonally, occurs at each of
the six matched mitre bends as shown. The line lengths, p, r, s, normalised to
the microstrip wavelength, are chosen to provide correctly phased polarisation
of the required type within a cell. To maintain the correct phasing between cells,
the normalised length t must be equal to the fractional part of 201 + r + s). In
cases where either t o r s is zero, the unit cell reduces to a four-cornered structure.
In the circularly polarised case, s = 0 is a necessary condition, and the
parametric relationship between p and r is [51]:

Successive values of n give alternate hands of circular polarisation. Specific choice

o f p or r is dictated by considerations of cross-polar levels, line losses and input
VSWR, but a good working combination is:

This solution is illustrated in Fig. 14.236.

Linear polarisation is obtained only if s = p. The parametric relationships
are as follows:

Longitudinal polarisation: either p +

r = n, or 2p +
r =n
Representative solutions quoted in the literature are: p = s = 0.25, r = 0.5,
t = 0; or p = s = r = 113, t = 0 (See Fig. 14.23~)
Transverse polarisation: either p + r = n - 0.5, or 2p + r = n - 0.5,
Representative solutions are: p = s = 0.125, r = 0.25, t = 0; or
P = s = r = 0.25, t = 0.5 (see Fig. 14.234.

If the cells are identical, the array is a constant-conductance array. A controlled

aperture distribution can be obtained by varying the geometry of successive cells
within the constraints of the parametic equations. This is a consequence of an
approximately linear relationship between the radiation conductance of a cell
and the dimension p [51]. Two-dimensional arrays using a one-dimensional
corporate feed have been manufactured [3].
The crank array may also be analysed in terms of electric surface currents
over the complete cell, rather than in terms of localised equivalent magnetic-
current sources. The circularly polarised design described above has been
analysed in this way, and a two-dimensional version reported [52, 531. In order
to improve the overall radiation pattern, the linear arrays were fed in pairs, with
a relative displacement of one half-wavelength between them.
The meander-line form of the linearly polarised chain antenna [2, 3, 541 may
also be regarded as two crank arrays fed in pairs. Fig. 1 4 . 2 4 ~shows this type,
and Fig. 14.246 shows the rectangular-loop type [2, 551. In each case the width Fig. 14.24 Travelling-wave chain antennas (Reproduced from 131 with permission of IEE)
of the loop 2p, normalised to microstrip wavelength, is about 1, and the lengths a Meander-line
r of the sections parallel to the antenna axis are about 0.4. The instantaneous b Rectangular loop
electric-current directions are such that radiation from these parallel sections c Microstrip Franklin antenna
d Circularly polarised antenna: schematic
848 Microstrip antenna feeds Microstrip antenna feeds 849
combines in phase, whilst radiation from the transverse sections tends to cancel. trol is possible by varying the line width, and hence the impedance, of the
The polarisation is therefore linear in the longitudinal direction, producing very vertical components in the grid. Good performance has been obtained at
low cross-polar levels. The beam produced can squint with frequency over a lOGHz from a two-plane monopulse array of this type, occupying a five-
wide range of angles relative to broadside, given by wavelength-diameter circular aperture.
sin 0 = (r + p - I)/r (14.46)
A two-dimensional array of this type has been reported [54] in which both ends
of the linear arrays are connected to corporate feeds, so that there is a single feed
port and a single load port.
The microstrip Franklin antenna [2, -3, 561, shown in Fig. 14.24c, is also
linearly polarised in the longitudinal direction. For broadside radiation, the line
lengths AB, BC and CD are each one half-wavelength. The instantaneous
currents are in the direction shown. Transverse currents in adjacent parallel
sections act in opposition, so that the cross-polar levels are low. Further reduc-
tion in the cross-polar levels is obtained in a two-dimensional array by making
each linear array the mirror image of its neighbour.
Another chain antenna able to generate circular polarisation is shown in Fig.
14.24d [3, 571. The basic radiating elements are v-shaped sections of line, with
an included angle a and arms of length s. The sections are linked by Schiffmann
phase shifters aranged in alternate directions so that their spurious radiation is
cancelled. The practical array consisted of eight 32-element arrays printed on
thin plastic sheets suspended above a ground plane. Analysis of the instan-
taneous current on a single element shows that the quadrature components Eg
and E, have equal magnitude, and hence generate circular polarisation, if the
following equation holds:
tan (42) cos 0 = tan (ksT/2) (14.47)
where 0 is the beam angle relative to broadside, and
T = 1 - sin (a/2) sin 0.
The Schiffmann phase shifters [58] have negligible effect on the radiation, but
introduce 90" phase shift between elements, thereby maintaining the required
phase conditions along the array for a circularly polarised beam in the 8,
Fig. 14.25 Microstrip wire grid arrays
direction. The phase-shifter length I, and width I, must be adjusted so that a 'Brick wall' wire grid array (Based on [59]with permission of IEEE @ 1981
sin 0, = (2s + I, + I, - I,)/d (14.48) IEEE)
b Hexagonal lanice array (Reproduced from Hill [60]with permission of IEE)
where d is the spacing between elements. c Square lanice array (Reproduced from Hill [60]with permission of IEE)
A further subset of discontinuity arrays worthy of mention is the wire grid o r
The lattice arrays shown in Fig. 14.256 and c utilise three- or four-line
lattice array, three examples of which are shown in Fig. 14.25 [59, 601. Each
junctions having angular symmetry [60]. Such junctions will radiate in the
example is a resonant structure radiating in the broadside direction. The 'brick
direction normal to the array, only if the line widths at the junction are unequal.
wall' configuration of Fig. 14.25~[59] consists of microstrip loops one
A non-radiating lattice of constant-impedance lines each one half-wavelength
wavelength wide and one half-wavelength high. The instantaneous currents in
long can be used, for example, to feed an array of open-circuited stubs placed
the loop are such that the vertically polarised field components combine con-
at the nodes of the lattice. However, in the examples shown, horizontally
structively, and the horizontally polarised components cancel. Amplitude con-
polarised radiation is caused by the horizontal members of the lattice having a
850 Microstrip antenna feeds Microstrip antenna feeds 851
lower impedance than the rest. Of the two alternatives shown, the hexagonal The performance of this type of power divider is limited by the relatively low
array has advantages in terms of bandwidth. isolation between the output ports. If these ports do not have perfectly matched
terminations, some power will be reflected back to the input, and some will be

14.4 Direct-coupled stripline power dividers and combiners Z:2Zo z = 2z0

output port 2 3 output port

14.4.1 Simple three-port power dividers

Most corporate feed networks use two-way power splitters for progressively
subdividing the power to the array elements. Considering equal power division
first, the simplest type is the T-junction [3, 611 shown in Fig. 14.26~.The input
port must be matched, and the output ports properly terminated. In the example
shown, if the input impedance is, say, 50R, the output impedances are each
100n. A 90" vee, cut as shown, helps to match the junction. Transformation to
50R at the output ports may be achieved by means of step transformers or
tapers.
An alternative T-junction design, shown in Fig. 14.26b, has 50 line impe-
dances at each port, and a quarter-wave matching transformer with an impe-
dance of 35.36Q. Unequal power splits at the T-junction can be obtained by the
use of lines with the required impedance ratio at the two output ports, and
appropriate impedance matching at the junction itself, as shown in Fig. 14.26~.
S-parameter analysis may be used to determine the effects of mismatched output
ports on the power-splitting characteristics of the junction.
Apart from the T-junction, the in-line power splitter is also widely used [3,61].
In this configuration, as Fig. 14.27~shows, the input line bifurcates into two
lines with the required impedance ratio. To avoid coupling between the two, the
output lines bend away from each other a short distance from the junction. It
is necessary to choose the input line width such that: (a) its impedance is
matched by the parallel combination of the output line impedances; and (b) the
output line widths, together with a small gap between them, can be accom-
modated within the width of the input line.
The triplate form of this type of splitter can be analysed by applying mode-
matching techniques to an equivalent waveguide model of the junction, as
shown in Fig. 14.276 [3, 611. For chosen reflection coefficients at the output
ports, the analysis is able to evaluate the reflection coefficient at the input port,
and the transmission coefficients to the two output ports. The S-parameters of
the junction are then simple functions of these coefficients and the line impedan-
ces. Fig. 14.26 T-junction power splitters
a Basic T-junction
When microstrip, as opposed to triplate, is used for any corporate feed, the b Quarter-wave matched T-junction
dispersive nature of the transmission line should be taken into consideration. c Unequal-split T-junction
The microstrip phase constant is dependent on both frequency and line im-
pedance. Thus, if any significant lengths of unequal impedance are used in
parallel, line-length adjustment must be made to ensure equal phase conditions coupled to the other output port. As a consequence, the aperture distribution
at the element ports. These conditions will strictly only be met at one frequency, will be distorted, and this could be a serious problem in a low-sidelobe design.
although differential dispersion between lines of unequal impedance is normally Similar internal coupling and reflection effects occur if the device is used as a
very small. power combiner.
 852 Microstrip antenna feeds
There is also a practical limit to the impedance ratio which may be obtained
using the simple three-port power divider. This limit is not much more than 2: 1,
and it is mainly dictated by the impedance of the narrowest line which it is
possible to etch reliably and accurately.
Microstrip antenna feeds 853
The essential feature of this network is that, for the correctly chosen line
impedances and with power entering port 1, the voltages at junctions a and b
are equal. A resistor of an appropriate value placed between these junctions will
therefore not absorb any power at the design frequency. If power enters port 2

view through a - a
t

view through b-

Fig. 14.27 In-line power splitter

a Unequal power split
b Triplate junction with equivalent waveguide model (Reproduced from [61]
with permission of IEE)
The following type of power divider is able to provide a larger ratio based on
the same limitation on maximum impedance. It also has the important advan-
tage that the output ports are more efficiently isolated. Fig. 14.28 Isolated in-line power dividers
a Uncompensated type
b Compensated type
14.4.2 Isolated power dividers/combiners
Parad and Moynihan 1621 first described and analysed the two-way split-tee or port 3, some energy will be dissipated in the resistor. The isolation between
power divider, or isolated in-line power divider. They based their design on the output ports will be high, and the reflection coefficient looking into any port will
N-way power divider of Wilkinson [63], after whom these devices are now be low. The device may be regarded as a T-hybrid, with the resistor acting as
named. Fig. 1 4 . 2 8 ~shows a basic unequal-split isolated power divider [62, 641. a reflectionless load on the internal series port [65]. It will operate well as a
854 Microstrip antenna feeds Microstrip antenna feeds 855
power combiner, and if designed in this mode for combining equal in-phase A six-element corporate feed using microstrip branch-line couplers has been
powers it will, in fact, be lossless. reported for an array with a - 30dB Dolph-Chebyshev aperture distribution
Two forms of isolated power divider may be distinguished; the uncompen- 1711. Difficulties in the photo-etching process limited the maximum coupling
sated type shown in Fig. 14.28a, and the compensated type shown in Fig. 14.286. factor to 9 dB, but - 28 dB sidelobes were achieved with the design.
The latter has an additional quarter-wave transformer at the input port. Over
an octave bandwidth, it has better isolation and input VSWR, but worse output
VSWR. The network impedances for an output voltage ratio of K are given in
the Figure [62, 641.
Improvements on the original design have been reported, including multi-
section-wideband, equal-power-split and unequal-power-split versions [64-671.
The original Wilkinson N-way power divider is unsuited to planar networks
because the resistors from the N output ports must meet at a common floating
star point. Recently, an improvement has been developed using a planar feed
network well suited to MIC applications [68]. It consists of a Dolph-Chebyshev
single-input tapered transmission line segmented into N strips forming the
output lines. The isolating resistors connect between the adjacent coupled
transmission lines.
Zo = a . 5 =
Despite the good performance characteristics of the isolated power divider, it normalised admittances:
- b

has the disadvantage that an additional component in the form of the resistor Za ' Zb
must be added to the printed feed network. Moreover, the value of the resistor matched condition: b2 - a2 = 1
is dependent on the power-divider ratio required. The maximum ratio available
in practice is about 4: 1. A larger ratio can be obtained from the four-port power
dividers to be described next, which are also able to provide good isolation.
power ratios: coupling,
2 (42=
14.4.3 Four-port direct-coupled power dividers
Two types of four-port direct-coupled power dividers are useful for planar feed
i n s e r t i i loss, 2 ($2

networks. These are the branch-line coupler and the hybrid-ring coupler, shown
in Figs. 14.29 and 14.30, respectively. Being four-port devices, they can both be
fully matched at the design frequency if their constituent line impedances are
correctly chosen. For both, the input power PI entering port 1 is divided
between the through port 2, P,, the coupled port 3, P,, and the decoupled port
4, P,. Since the directivity P,/P, is in practice finite over the operating frequency
band, a resistive termination is needed on port 4; this is an undesirable feature
of these couplers. The through port insertion loss is PJP,, and the coupling
factor is P,/P,. The power-divider ratio is PJP,.
The branch-line coupler [64,69,70] can provide coupling values up to about
9 dB. The phase difference between the output ports is 90, independent of
coupling and frequency, so correction for this must be made in any corporate
feed. As Fig. 14.29 shows, if the normalised admittance of the shunt arm of the
coupler is a, the power-divider ratio is given simply by a2, assuming negligible
directivity. The maximum ratio available is thus about 7:l.
Multi-section branch-line couplers can be designed for greater bandwidth,
but owing to the narrow line widths required on the outer shunt arms, only two-
and three-arm couplers are used in practice.
directivity, P4 0 at resonance
-
P
power divider ratio, 5 , a2
p2
Fig. 14.29 Branch-line coupler
The disadvantage of the branch-line coupler is that its bandwidth is limited.
It also takes up a relatively large surface area, leading to additional line losses.
The hybrid-ring directional coupler shown in Fig. 14.30, also known as the
rat race, consists of a ring of 1.5 wavelengths circumference, with the four ports
disposed as shown. The design conditions given in the Figure [64,69] show that
the power-divider ratio is the square of the admittance ratio alb. A maximum
ratio of about 9 dB, or 8:1, is available with this configuration.
A multi-layer, broadband stripline beam-forming network using hybrid rings
has recently been reported [72]. It produces both even- and odd-mode beams
from an 8 x 8-port Butler matrix (see Section 14.5.2.2). The hybrid rings are
 856 Microstrip antenna feeds Microstrip antenna feeds 857
of modified design, shown in Fig. 14.3la, to provide broader bandwidth [73]. A 14.5 Other feed systems
similar modification (Fig. 14.31b) is able to give a larger potential power-divider
ratio, with a bandwidth of about 20% [74]. The equation governing the operation 14.5.1 Alternative transmission lines
of the device is indicated in the Figure. As indicated in Section 14.2.3, if the radiating patches are coupled via a probe
An important consideration is that input at port 1 (the sum port) of the hybrid or aperture, there is complete freedom in the choice of transmission line for the
ring produces in-phase outputs at ports 2 and 3. Therefore no adjustments to
line lengths in the feed network are necessary, in contrast to the branch-line
coupler case. On the other hand, it is sometimes topographically inconvenient
to have the loaded port opposite the input port instead of adjacent to it as in
the branch-line coupler. Input at port 4 (the difference port) produces 180'
phase difference between the output signals at ports 2 and 3. This characteristic
is made use of in the monopulse phase comparators to be described in Section
14.5.2.1.

-- 4
difference
oor t

matched condition: b 2 + a 2 :1

power ratios : coupling. 3 ,a2

PI
insertion loss, P2 = b 2
'pi-
directivity,
3 I o o t resonance
p1

2=@
power divider ratio,

Fig. 14.30 Hybrid-ring coupler (Reproduced from [74] with permission of IEEE @ 1986
IEEE)
For even higher power-divider ratios, parallel-coupled lines would be necess-
ary, but these are not commonly used in microstrip-antenna feed designs.
Instead, aperture distributions with high edge-to-centre ratios have been
achieved by judicious removal of some power dividers in the corporate feed
chain, as illustrated in Fig. 14.10.
Fig. 14.31 Modified hybrid rings
a Broadband type (Reproduced from [73] with permission of IEEE @ 1982 IEEE)
b High power-divider ratio type (Reproduced from [74] with permission of IEEE
@ 1986 IEEE)
feed system. Hitherto, a microstrip or triplate medium has been assumed, but
relatively high loss is associated with both of these, particularly in the millimetre
waveband [6,75, 761. Lower loss is obtained from suspended stripline, and from
858 Microstrip antenna feeds
dielectric image guide, which has already been mentioned in the context of
co-planar coupling [16].
Air-filled waveguide has exceptionally low losses, so this must not be discoun-
ted as a possible medium. However, radiating slots are readily cut in waveguide
walls to form complete antennas [77]; so strong justification would be needed for
using waveguide to feed microstrip patches instead. Such arrays have not yet
been seriously investigated, but the low-loss advantages may become sufficiently
Fig. 14.32 Schematic layout of patch array with interlaced H feeds (Copyright @ Controller
HMSO, London, 1986. Reproduced with permission)
attractive in the future. One important feature of waveguide is that its propagat-
ing wavelength is greater than that in free space, unless the loss-inducing
complication of corrugated walls or dielectric filling is introduced. Consequent-
ly, only transposed arrays can be used to produce a broadside beam free from
grating lobes.
Low-loss feeds based on parallel-plate waveguide may also be considered for
two-dimensional-array applications, although to date they also have been used
Microstrip antenna feeds 859
only to feed radiating slots [78]. Cavity feeds and radial waveguide feeds [79]
come into this category. In view of the difficulty in obtaining circularly polarised
radiation directly from slots in parallel-plate waveguide, one possibility is to use
circularly polarised patches requiring a single feed point [20, 321 as radiating
elements coupled to a parallel-plate feed system.
14.5.2 Multiple beam-forming networks
14.5.2.1 Special-purpose two- or four-beam networks: Microstrip antennas
may be used for applications requiring the formation of two or more indepen-
dent beams, either simultaneously or sequentially. For example, with the advent
of higher-power satellites, printed antennas are becoming acceptable not only as
receive-only DBS antennas, but also as transmitlreceive antennas. The two
independent modes of operation may be separated by using two orthogonal
polarisations with the same frequency, and/or by using two separate frequency
bands. A recently developed antenna for this application utilised two interlaced,
multi-layer corporate feeds [32]. Fig. 14.32 illustrates schematically how the
feeds are coupled via probes to a two-dimensional array of square patches with
two opposing bevelled corners. The patches are capable of radiating either hand
of circular polarisation, depending on the position of the probes. The required
broad axial-ratio bandwidth was achieved by the technique of sequential rota-
tion [36-391.
In another example, one type of airborne velocity-measuring radar uses four
simultaneous independent beams, each directed into one of the four forward/
backward, leftlright quadrants. This requirement can be met by a two-dimen-
sional microstrip array with the serieslseries feed network shown in Fig. 14.33
[3]. The coupling to the radiating elements in each linear array is such that the
radiation conductance is symmetrical about the centre of the array. Each beam
is uniquely generated by a signal applied to one of the corners of the array.
A further example, applicable to a guidance-radar antenna, is a monopulse
comparator, which utilises four hybrid couplers to give sum and difference
beams in two orthogonal planes. The principle is illustrated in Fig. 14.34, as
applied to a simple 2 x 2 patch array, which could be used as the primary feed
in a reflector system [80]. However, a complete microstrip array may be design-
ed, in which each of these single patches is replaced by a two-dimensional array
occupying one quadrant of a circular aperture. Each array may be corporately
fed, with the monopulse comparator on a separate feed layer [35]. Alternatively,
the elements of the array may be fed by a co-planar serieslseries network, with
the comparator at the centre of the aperture [76, 811 (Fig. 14.35).
14.5.2.2 Multiple fan beams: Several radar applications require the genera-
tion of multiple fan beams from a single linear antenna array. Thisis done by
means of a multiple beam-forming network fed by a set of beam ports, with a
set of antenna-array output ports. A signal entering any one of the beam ports
 860 Microstrip antenna feeds Microstrip antenna feeds 861
excites all the array ports to produce a beam in a particular direction. Three [82, 841. For an equally spaced, constant-amplitude array, this condition is
alternative feed techniques are commonly used for this purpose, as shown in satisfied when the peak of each beam coincides with the first null of its neigh-
Fig. 14.36. These are: bour. The condition is only met at one frequency, but reasonable bandwidths
(a) The parallel-plate lens type originated by Gent, but known in its two more patch, or subarray
/
recent forms'as Ruze or Rotman lenses [82, 831.
(b) The series-coupled feed type known as the Maxson-Blass matrix [82, 841.
(c) The parallel-feed type known as the Butler matrix [82, 851

PORT l I 2

difference port

-- ., -
-- Fig. 14.34 Monopulse comparator using branch-line couplers

.. 7 . - can be obtained with the lens system. Recently, a Rotman lens system has been
beam l
2 direction
Fig. 14.33 Four-beam-array feed network: schematic
All are available in printed transmission-line format, and may therefore be used
in conjunction with microstrip patch arrays.
The parallel-plate Rotman lenses (shown schematically in Fig. 14.36~)are so
designed that the beam positions do not change with frequency. However,
optimum beam efficiency is only obtained if the beams are spatially orthogonal
used in an experimental flat-plate DBS antenna, to allow selection of the beam
most suitable for acquiring the satellite signal when the antenna is fixed to a
convenient wall of a house [86].
The Maxson-Blass matrix, shown schematically in Fig. 14.366,is particularly
useful when a small number M of narrow beams is required from a relatively
large number of array ports N. There is no particular restriction on the values
of M or N. A total of M x N directional couplers is necessary in the network,
and line lengths between the couplers in the series feeds to the array ports must
be designed to ensure spatial orthogonality of the beams at the design frequency.
At other frequencies, the beam directions will change, and the orthogonality
condition will not be maintained. Because the beam feed lines cross over the
array feed lines, the directional couplers must take the form of broadside-
coupled lines on either side of a substrate suspended between parallel ground
planes [87, 881.
The Butler matrix, shown schematically in Fig. 14.36c, is best suited to a
network with N beams and N array ports, where N is an integer power of 2, i.e.
N = 2". A total of N x n couplers or hybrids is required in the network, and
these can be 90" or 180" hybrids, depending on whether the beams are to be
symmetrically distributed about the broadside direction, or whether one of the
beams is to be in the broadside direction. In addition, a considerable number of
862 Microstrip antenna feeds Microstrip antenna feeds 863
phase shifters must be incorporated in the network. Furthermore. many cross- transmission -
parallel-
overs are required. and this involves the use of a multi-layer structure. and plate
possibly a combination of different transmission-line types. The overall design region

for a large number of ports is thus very complicated [85]. beam

ports

-
,antenna
elements

(a)

antenna elements

loads

. directional
- couplers

antenna elements

Fig. 14.35 Two-dimensional microstrip antenna with integrated monopulse comparator

3dB hybrids
(Photograph by courtesy of Ball Communication Systems Division, Broomfield,
Colorado, USA) phase shifters
non - intersecti*
crossovers

14.5.2.3 Active phased-array feeds: Many applications are emerging in the

areas of radar and satellite communications, for antennas with electronic-beam
steering and adaptive beam-forming capabilities. Microstrip-array antennas are
well suited to these applications, and it is relevant to consider the feed systems
required in such cases. (c)
Fig. 14.37 shows the feed network for an S H F receive-only phased array built
Fig. 14.36 Multiple beam-forming networks
entirely in microstrip, which has been developed for aircraft-to-satellite com-
a Parallel-plate Rotman lens array: schematic (Reproduced from Smith [83]with
munication [89]. The corporate feed network is on a separate layer from the permission of IERE)
radiating elements, which consist of a n 8 x 8 array of circular patches. The 64 b Maxson-Blass matrix: schematic (Reproduced from Shaw [84] with per-
arms of the feed each contain a 3-bit digital phase shifter and a branch-line mission of IERE)
c Butler matrix: schematic
864 Microstrip antenna feeds
hybrid. The latter provides the necessary phase quadrature at two orthogonal
feed points to each patch, to satisfy the requirement for left-hand circular
polarisation. Each phase shifter contains ten PIN diodes providing a combina-
tion of 180' and 90" switched-line and 45" loaded-line phase shifters, including
the necessary RF chokes and DC bias tracks. The beam is steered in the required
direction by means of a microprocessor-based controller.
Fig. 14.37 Microstrip corporate feed for SHF receive-only phased array (Photograph by
courtesy of Ball Communication Systems Division. Broomfield, Colorado, USA)
A prototype C-band phased array transmit/receive microstrip antenna inten-
ded for earth imaging from space has recently been reported [90]. It consists of
12 linear arrays of 18 square patch elements. Each array has two independent
co-planar centre-fed series feeds for orthogonal linear polarisation, as shown in
Fig. 14.38. Behind the feed layer, in a mixture of microstrip and packaged
components, are two sets of 12 T / R modules utilising two-stage high-power and
low-noise FET amplifiers with associated PIN switch and PIN limiter diodes,
Microstrip antenna feeds 865
and 4-bit PIN diode phase shifters. The T / R capability is thus available for both
vertically- and horizontally-polarised signals. The circuit is completed by two
12-way corporate feeds, one for each polarisation.
Looking to the future, much effort is currently being devoted to GaAs MMIC
realisations of T/R modules incorporating F E T amplifiers, switched-filter and
switched-delay-line phase shifters, and both F E T and PIN diode T / R switches
[9 1-94]. These are ultimately intended for high-power broad-bandwidth phased-
Fig. 14.38 Coplanar microstrip antenna and feed for C-band transmitlreceivephased array
(Photograph by courtesy of Ball Communication Systems Division, Broomfield.
Colorado, USA)
array radars with full electronic-scanning capabilities. Again, microstrip patches
are potentially good candidates for the radiating elements. provided that their
limited gain-bandwidth can be accepted. Several potential problems in imple-
menting monolithic phased-array antennas have been identified, and some
possible solutions proposed [95]. There is little doubt that many advanced
phased-array systems will incorporate microstrip-array antennas, and that
ingenious feed networks will continue to be important features in them all.
866 Microstrip antenna feeds
14.6 Acknowledgments
The author is grateful to THORN EM1 Electronics Ltd. for permission to
publish this work, and to his colleagues, particularly Mr. J. Thraves and Mr.
G. R. Selby, for their help and advice on the manuscript.
14.7 References
HALL, P. S., and JAMES, J. R.: 'Survey of design techniques for flat profile microwave
antennas and arrays.' Radio & Electron. Engr., Nov. 1978, pp. 549-565
BAHL, I. J., and BHARTIA, P.: 'Microstrip antennas (Artech House, 1980)
JAMES, J. R., HALL, P. S., and WOOD, C.: 'Microstrip antenna theory and design (Peter
Peregrinus, 1981)
MAILLOUX, R. J., MCILVENNA, J. F., and KERNWEIS, N. P.: 'Microstrip array technol-
ogy' IEEE Trans., Jan. 1981, AP, pp. 25-37
JAMES, J. R., HALL, P. S., WOOD, C., and HENDERSON, A.: 'Some recent developments
in microstrlp antenna design.' ibid. pp. 124-128
HENDERSON, A., and JAMES, J. R.: 'A survey of millimetre-wavelength planar antenna
arrays for military applications.' Radio & Electron. Engr., Nov/Dec. 1982, pp. 543-550
JOHNSON, R. C., and JASIK, H. (Eds.): 'Antenna engineering handbook.' (McGraw-Hill,
1984) 2nd edn. chap. 7
LEWIN, L.: 'Radiation from discontinuities in stripline, Proc. IEE, Feb. 1960, pp. 163-170
LEWIN, L.: 'Spurious radiation from microstrip' Proc. IEE, July 1978, pp. 633-642
HENDERSON, A., and JAMES, J. R.: 'Design of microstrip antenna feeds. Pt. I: Estimation
of radiation loss and design implications' IEE Proc. H, Feb. 1981, pp. 19-25
OBERHART, M. L., LO, Y. T., and LEE, R. Q. H.: 'New simple feed network for an array
module of four microstrip elements' Electron. Letts., 23 Apr. 1987, pp. 436-437
BENALLA, A., and GUPTA, K. C.: 'Transmission-line model for two-port rectangular
microstrip patches with ports at the nonradiating edges' Eleclron. Letts., 13 Aug. 1987, pp.
882-884
CASHEN, E. R.: British Patent Specification No. 1572273, 1977
OWENS, R. P., and THRAVES, J.: 'Microstrip antenna with dual polarisation capability'
Proceedings, Military Microwaves Conf. Oct. 1984, pp. 250-254
CARTER, M. C., and CASHEN, E. R.: 'Linear arrays for centimetric and millimetric
wavelengths' Proceedings, Military Microwaves Conf. Oct. 1980, pp. 315-320
JAMES, J. R., JOHN, G., and HALL, C. M.: 'Millimetre-wave hybrid dielectric-microstrip
antenna array' IEE Proc. H, Dec. 1984, pp. 341-350
JAMES, J. R., and HALL, P. S.: 'Microstrip antennas and arrays. Pt. 2 New array design
technique' IEE J. M O A , Sept. 1977, pp. 175-181
METZLER, T.: 'Microstrip series arrays' IEEE Trans. Jan. 1981, AP, pp. 174-178
CARVER, K. R., and MINK, J. W.: 'Microstrip antenna technology' ibid. pp. 2-24
RICHARDS, W. F., LO, Y. T., and HARRISON, D. D.: 'An improved theory for microstrip
antennas and applications' ibid, pp. 38-46
CHEW, W. C., and KONG, J. A.: 'Analysis of a circular microstrip disk antenna with a thick
dielectric substrate' ibid. pp. 68-76
YANO, S., and ISHIMARU, A,: 'A theoretical study of the input impedance of a circular
microstrip disk antenna, ibid. pp. 77-83
DAS, A., and DAS, S. K.: 'Input impedance of a probe excited circular microstrip ring
antenna' IEE Proc. H., Oct. 1985, pp. 384-390
DAVIDOVITZ. M., and LO, Y. T.: 'Input impedance of a probe-fed circular microstrip
antenna with thick substrate' IEEE Trans. July 1986, AP, pp. 905-91 1
Microstrip antenna feeds 867
25 GRIFFIN, J. M., and FORREST, J. R.: 'Broadband circular disc microstrip antenna' Elec-
tron. Letts., 18 Mar. 1982, pp. 266-269
26 FONG, K. S., PUES, H. F., and WITHERS, M. J.: 'Wideband multilayer coaxial-fed
microstrip antenna element' ibid. 23 May 1985, pp. 497499
27 HALL, P. S.: 'Probe compensation in thick microstrip patches' ibid. 21 May 1987, pp. 606-607
28 POZAR, D. M.: 'Microstrip antenna aperture coupled to a microstrip line' ibid. 17 Jan. 1985,
pp. 49-50
29 SULLIVAN, P. L., and SCHAUBERT, D. H.: 'Analysis of an aperture coupled microstrip
antenna' IEEE Trans. Aug. 1986, AP, pp. 977-984
30 GRONAU, G., and WOLFF, I.: 'Aperture coupling of a rectangular microstrip resonator'
Electron. Lett., 8 May 1986, pp. 554-556
31 ADRIAN, A,, and SCHAUBERT, D. H.: 'Dual aperturecoupled microstrip antenna for dual
or circular polarisation' Electron Lett. 5 Nov. 1987, pp. 1226-1228
32 OWENS, R. P., and SMITH, A. C.: 'Dual band, dual polarisation microstrip~antennafor
X-band satellite communications' Proceedings, Military Microwaves Conf. June 1986, pp.
323-328
33 BUCK, A. C., and POZAR, D. M.: 'Aperture-coupled microstrip antenna with a perpen-
dicular feed' Electron. Lett. 30 Jan. 1986, pp. 125-126
34 POZAR, D. M., and JACKSON, R. W.: 'An aperture coupled microstrip antenna with a
proximity feed on a perpendicular substrate' IEEE Trans., June 1987, AP, pp. 728-731
35 OLTMAN, H. G., and HUEBNER, D. A.: 'Electromagnetically coupled microstrip dipoles'
IEEE Trans., Jan. 1981, AP, pp. 151-157
36 HANEISHI, M., YOSHIDA, S., and GOTO, N.: 'A broadband microstrip array composed
of single-feed type circularly polarised microstrip antennas' IEEE AP-S Digest, 1982, pp.
160-163
37 HANEISHI, M., and TAKAZAWA, H.: 'Broadband circularly polarised planar array com-
posed of a pair of dielectric resonator antennas' Electron. Lett. 9 May 1985, pp. 437-438
38 TESHIROGI, T., TANAKA, M., and CHUJO, W.: 'Wideband circularly polarised array
antenna with sequential rotations and phase shift of elements' Proc. ISAP 85, Aug. 1985, pp.
117-120
39 HALL, P. S., and HALL, C.M.: 'Coplanar corporate feed effects in microstrip patch array
design' IEE Proc. H., June 1988, pp. 180-186
40 HANSEN, R. C. (Ed.): 'Microwave scanning antennas: Vol. 2' (Academic Press, 1966),chap.
1
41 RUDGE, A. W., MILNE, K., OLVER, A. D., and KNIGHT, P. (Eds.): 'The handbook of
antenna design: Vol. 2' (Peter Peregrinus, 1983) chap. 9
42 STUTZMAN, W. A,, and THIELE, G. A,: 'Antenna theory and design' (John Wiley, 1981)
chap. 3
43 ROGERS, A,: 'Wideband squintless linear arrays' Marconi Rev. 4th quarter 1972, pp. 221-243
44 SKIDMORE, D. J., and MORRIS, G.: 'The design and performance of covered microstrip
serpent antennas' IEE Conf. Publ. 219. Proceedings ICAP 83, Pt. I pp. 456458
45 COLLIN, R. E.: 'Foundations for microwave engineering' (McGraw-Hill, 1966)
46 WILLIAMS, J. C.: 'A 36 GHz printed planar array' Electron. Lett. 2 Mar. 1978, pp. 136-137
47 DANIEL, J-P., MUTZIG, J-P., NEDELEC, M., and PENARD, E.: 'RCseaux d'antennes
imprimkes dans la bande des 20 GHz/30 GHz' L'Onde Elecrrique, Jan./Feb. 1985, pp. 35-41
48 DANIEL, J-P., PENARD, E., NEDELEC, M., and MUTZIG, J-P.: 'Design of low cost
printed antenna arrays' Proc. ISAP 85, Aug. 1985, pp. 121-124
49 COSSLETT, M. F., FROST, R., and ROSSITER, K. 0.: British Patent Specification No.
1269633, 1968
50 SHAFAI, L., and SEBAK, A. A.: 'Radiation chracteristics and polarisation of undulated
microstrip line antennas' IEE Proc. H, Dec. 1985, pp. 433439
51 HALL, P. S.: 'Microstrip linear array with polarisation control' IEE Proc. H., Apr. 1983,
pp.215-224
 868 Microstrip antenna feeds
52 NISHIMURA, S., SUGIO, Y., and MAKIMOTO, T.: 'Crank-type circularly polarised
microstrip line antenna' IEEE AP-S, Digest, 1983, pp. 162-165
53 NISHIMURA, S., SUGIO, Y., and MAKIMOTO, T.: 'Side-looking circularly polarised
microstrip line planar antenna' Proc. ISAP, 1985, pp. 129-132
54 TIURI, M., HENRIKSSON, J., and TALLQVIST, S.: Printed circuit radio link antenna'
Proc. 6th EMC, 1976, pp. 280-282
55 TIURI, M., TALLQVIST, S., and URPO, S.: 'The chain antenna' IEEE AP-S Digest, 1974,
pp. 274-277
56 NISHIMURA, S., NAKANO, K., and MAKIMOTO, T.: 'Franklin-type microstrip line
antenna' IEEE AP-S Digest, 1979, pp. 134-137
57 HENRIKSSON, J., MARKUS, K., and TIURI, M.: 'A circularly polarised travelling-wave
chain antenna' Proc. 9th EMC, 1979, pp. 174-178
58 SCHIFFMANN, B. M.: 'A new class of broadband microwave 90 degree phase shifted IRE
Trans., Apr. 1958, M'IT, pp. 232-237
59 CONTI, R., TOTH, J., DOWLING, T., and WEISS, J.: 'The wire grid microstrip antenna'
IEEE Trans. Jan. 1981, AP, pp. 157-166
60 HILL, R.: 'Printed planar resonant arrays' IEE Conf. Publ. 274. Proceedings ICAP 87, Pt. 1,
pp. 473-476
61 HALL, P. S., and JAMES, J. R.: 'Design of microstrip antenna feeds. Pt. 2: Design and
performance limitations of triplate corporate feeds' IEE Proc. H, Feb. 1981, pp. 26-34
62 PARAD, L. I., and MOYNIHAN, R. L.: 'Split-tee power divider' lEEE Trans., Jan. 1965,
MTT, pp. 91-95
63 WILKINSON, E. J.: 'An N-way hybrid power divider' IRE Tram., Jan. 1960, MTT, pp.
11C118
64 HARLAN HOWE: 'Stripline circuit design (Artech House, 1974) chap. 3
65 COHN, S. B.: 'A class of broadband three-port TEM mode hybrids' IEE Trans, Feb. 1968,
MTT, pp. 110-116
66 LI, C. Q., LI, S. H., and BOSISIO, R. G.: 'CADICAE design of an improved, wideband
Wilkinson power divider' Microwave J., Nov. 1984, pp. 125-135
67 WAHI, P. K.: 'Wideband, unequal split ratio Wilkinson power divider' Microwave J. Sept,
1985, pp. 205-209
68 YAU, W., and SCHELLENBERG, J. M.: 'An N-way broadband planar power combiner/
divider' Microwave J. Nov. 1986, pp. 147-151
69 REED, J., and WHEELER, G. J.: 'A method of analysis of symmetrical four-port networks'
IRE Trans, Oct. 1956, MTT, pp. 246-252
70 MATTHAEI, G. L., YOUNG, L., and JONES, E. M. T.: 'Microwave filters, impedance
matching networks, and coupling structures' (McGraw-Hill, 1964) Section 13.09
71 GUPTA, C. D., and DELOGNE, P.: 'Build an integrated Dolph-Chebyshev array' Microw-
aves, Nov. 1976, pp. 54-58
72 ABOUZAHRA, M. D.: 'Design and performance of a wideband multilayer feed network'
IEEE MTT-S Digest, 1986, pp. 143-146
73 KIM, D. I., and YOSHIYUKI, N.: 'Broad-band design of improved hybrid ring 3 dB
directional couplers' IEEE Trans., Nov. 1982, MTT, pp. 2040-2046
74 AGRAWAL, A. K., and MIKUCKI, G. F.: 'An improved hybrid-ring directional coupler for
higher power split ratios' Microwave J., Nov. 1986, pp. 87-98 (see also, IEEE Trans., Dec.
1986, MlT, pp. 140-1407)
75 POZAR, D. M.: 'Considerations for millimetre wave printed antennas' IEEE Trans., Sept.
1983, AP, pp. 740-747
76 LALEZARI, F., and MASSEY, C. D.: 'MM-wave microstrip antennas' Microwave J. Apr.
1987.. OD.
.= 87-96
77 JOHNSON, R. C., and JASIK, H.: 'Antenna engineering handbook' (McGraw-Hill, 1984)
2nd edn. chap. 9
78 RAHMAN, F., SHAFAI, L., BRIDGES, E., and ITITPIBOON, A.: 'A simple coaxial fed
planar microstrip slot array' IEEE AP-S Digest, 1981, pp. 207-208
Microstrip antenna feeds 869
79 ANDO, M., SAKURAI, K., GOTO, N., ARIMURA, K., and ITO, Y.: 'A radial line slot
antenna for 12 GHz satellite TV reception' IEEE Trans., Dec. 1985, AP, pp. 1347-1352
80 JACKSON, C. M., and NEWMAN, J.: 'Low cost Ka band microstrip patch monopulse
antenna' Microwave J. July 1987, pp. 125-131
81 ANDREWS, B. J., MOORE, T. S., and NIAZI, A. Y.: 'Millimetre-wave microstrip antennas
for dual polar and monopulse applications' IEE Conf. Publ. 219, Proc. ICAP 83, Pt. 1, pp.
529-533
82 HANSEN, R. C. (Ed.): 'Microwave scanning antennas. Vol. 3' (Academic Press, 1966) chap.
3
- SMITH. M. S.: 'Design
83
- - considerations for Ruze and Rotman lenses' Radio & Electron. Eng.,
Apr. 1982, pp. 181-187
84 SHAW. E.: 'The Maxson multi-beam antenna: Theory and design for non-interacting beams'
Radio & Electron. Eng., Feb. 1969, pp. 117-129
85 MACNAMARA, T.: 'Simplified design procedures for Butler matrices incorporating 90
degree hybrids or 180 degree hybrids' IEE Proc. H, Feb. 1987, pp. 50-54
86 MADDOCKS, M. C. D.: 'Low-cost approach for steerable flat-plate antenna design with
application to reception of broadcasting from satellite' Electron. Lett., 4 Feb. 1988, pp.
173-174
87 SHELTON, J. P.: 'Impedances of offset parallel-coupled strip transmission lines', IEEE
Trans., Jan. 1966, MTT, pp. 7-15
88 MOSKO. J. A,: 'Cougling- curves for offset parallel-coupled strip transmission lines' Microw-
A
ave J. Apr. 1967, pp. 35-37
89 CIPPOLA, F. W.: 'A 7.5 GHz microstrip phased array for aircraft-to-satellite communication'
Microwave J., Aug. 1981, pp. 75-78
90 HADDAD, H., FITHIAN, M., and COOMBS, D.: 'Heading for space: C-band phased array'
Microwaves & RF, Apr. 1986, pp. 103-108
91 ARNOLD, J., and SMITH, D. C.: 'Commercial availability of GaAs MMICs challenges
system designers' MSN & CT, Sept. 1986, pp. 119-131
92 TENEDORIO, J. G.: 'MMICs reshape EW system design' MSN & C T , Nov. 1986, pp. 95-104
93 NASTER, R. J.: 'Affordable MMIC designs for phased arrays' Microwave J.,Mar. 1987, pp.
141-150
94 CHILTON, R. H.: 'MMIC T/R modules and applications' Microwave J., Sept. 1987, pp.
131-146
95 POZAR, D. M., and SCHAUBERT, D. H.: 'Comparison of architectures for monolithic
phased array antennas' Microwave J., Mar. 1986, pp. 93-104
 Chapter 15

Advances in substrate technology

G.R. Traut

Substrate materials play an essential role in microstrip antenna design, produc-

tion and finished-product performance. Several aspects of materials must be
considered in the design stage when substrates are selected. What may seem
ideal from a design viewpoint must be balanced against production and final
product requirements. Ability to measure and control critical properties, especi-
ally relative permittivity and dissipation factor, cannot be ignored. The possible
adverse effects of necessary processing steps or environment in the final appli-
cation must be taken into account. Successful antenna production will depend
on the use of appropriate processing techniques. New substrate types and
special substrate features are becoming increasingly available and often ca'n
offer significant advantages for designers and producers.
This Chapter deals with these issues in five sections:
(i) What to consider when selecting materials
(ii) Methods for measuring relative permittivity and dissipation factor
(iii) Processing techniques for antenna fabrication
(iv) Design considerations related to materials characteristics
(v) Opportunities available in special features and new materials

15.1 Considerations for substrate selection

This discussion will be limited to five Subsections concerned with properties

most important to performance, a list of available choices, discussion of clad-
ding, details about thermal behaviour of PTFE, the polymer base for most
microstrip antennas, and some information on anisotropy related to composite
structure.

15.1.1 Impact of properties of various substrate systems on microstrip antenna

performance
Selection of the most suitable substrate for a microstrip antenna needs to be
 872 Advances in substrate technology
made early in the development of a system. Substrate characteristics must be
compatible with design objectives, the processing needed to produce the design
and the environmental conditions to which the system will be exposed during its
lifetime. Generally this selection process is a compromise to get the best balance
of desirable features for a given application. While many properties could be
discussed, we can focus on a critical three: complex permittivity, water absorp-
tion and adhesion of the metal-foil cladding.
15.1.1.1 Complex permittivity Complex permittivity has two components,
which we will call relative permittivity (dielectric constant) and dissipation
factor (loss tangent). Both are critical for microstrip antenna performance.
Neither can be regarded as a fixed characteristic. Rather, they are functions of
several factors in substrate composition, processing and application of an
antenna.
In most applications low relative permittivity is desirable to the designer for
radiation efficiency. Foam materials can have very low relative permittivity, but
this must be balanced against processability and resistance to operating stresses.
If the radiating elements are to be fed by microstrip transmission lines on the
same plane, then increasing its value is usually desirable to get good perfor-
mance.
Close tolerance of relative permittivity, K', from panel to panel, and within a
panel, is important for the final antenna system to perform as designed. With
production runs of many units, it is possible to accommodate known variations
from one panel to another with adjustments in artwork, but then accurate values
of K' for each panel are needed.
Dissipation factor is a measure of the amount of electrical energy converted
to heat in the dielectric, and partially accounts for power losses in a passive
device such as a transmission line or microstrip antenna.
The dielectric behaves as a distributed capacitor. An idealised lossless capaci-
tor shows reactively a charging rate or current cycle that exactly leads the
electric field or voltage cycle by 90". The amount this phase relationship differs
from 90' relates to the amount of capacitively stored energy dissipated as heat
during charging and discharging. Dissipation factor is also called the loss
tangent; i.e., the tangent of the deviation in phase from the 90" ideal.
Consideration of the properties of relative permittivity and dissipation factor
at an atomic or molecular level may help one to relate these to frequency and
temperature.
The incidence of an alternating electric field on a dielectric material causes a
dipole response that has the effect of increasing the amount of outside charge
needed to attain a given potential gradient. This dipole response is usually a
combination of two effects: electron shifts creating induced dipoles and move-
ment of groups of one or more atoms comprising pre-existing dipoles. The
magnitude of the dipole response is directly related to relative permittivity. The
fraction of energy converted to random thermal motion in the material (heat)
Advances in substrate technology 873
as a fraction of that stored by the dipole and space-related response is the
dissipation factor.
The electron shift response is common to all dielectric substances. Electrons
in a dielectric are bound in orbits with particular atoms or particular molecular-
bond systems. They shift elastically in response to an external electric field to an
I extent dependent on molecular structure. Elastic response means the stored
I potential energy is returned with little or no loss.
The other response, movement of atoms, arises from differing electro-negativ-
ity among atom types. Such dipoles exist without an external field, but they
I respond to a field by tending to align with it. Pendent polar groups on polymer
chains will rotate, twist or stretch toward alignment. Ionically bonded inorganic
structures are usually more rigid with more limited movement of groups,
compared with covalently bonded organic polymers.
As can be imagined, in molecular structures where such movement is per-
mitted, some of the energy associated with the movement is dissipated in
collisions with other groups in the system. The mechanical analogy of this is the
visco-elastic response characteristic of most polymers.
Bonds in polymeric molecules that give rise to little or no dipole moment
include the common carbon-carbon bond and the carbon-hydrogen bond.
Bonds that tend toward dipole formation include carbon to oxygen, nitrogen or
halogen atoms, including fluorine, chlorine and bromine.
Polyethylene, a polymer with molecules consisting of a chain of carbon atoms
with two hydrogen atoms bonded to each carbon atom along the chain, has a
low dissipation factor and relative permittivity as expected from the low dipole
moment of groups within the molecule. Similar chains with chlorine (-CI) or
hydroxy (-OH) substitutions on every other carbon atom have higher dissipa-
tion factors and relative permittivity, as expected.
Electrical properties of polymers relate not only to the presence of polarised
structures but also to the degree they can respond to the external electrical field.
A polymer in a highly crystalline state, with polar groups closely packed in a
rigid structure, will respond less than one in an amorphous state with more
mobility of groups.
Polytetrafluoroethylene (PTFE), widely used in microwave devices, is of
special interest. Highly polar fluorine atoms occupy the available bond positions
along the carbon chain. Both dissipation factor and relative permittivity are as
low as for polyethylene. This seeming contradiction is explained by the large size
of the fluorine groups. The polymer chain is actually stiffened and immobilised
by the crowded sheath of fluorine groups forming a helical pattern along the
chain. The high dipole moments of the many individual carbon-fluorine bonds
have a near-zero vector sum.
The following tabulation indicates qualitatively the influence of various
factors on relative permittivity and dissipation factor:
 874 Advances in substrate technology Advances in substrate technology 875
-

Factor Relative permittivity Dissipation factor In the case of PTFE-based laminates the surfactant for stabilising aqueous
Lamination colloidal PTFE dispersion must be fully removed from saturated glass fabric
Mechanical Density change from Mechanical before clad panels are laminated to avoid lossiness.
history microvoids causes Absorbed solvents, moisture or reagents from circuit-board processes can
degradation can degrade performance. Substrates based on saturated glass-fibre fabrics can
non-uniformity, lower increase value
exhibit wicking of moisture along the glass-resin interface. Manufacturers apply
values
coupling agents to reduce this effect.
Thermal Crystallinity raises Too much heat forms The frequency at which a molecular group will respond resonantly to an
history density for higher lossy groups by electric field is a function of its mass, dipole moment, proximity to other groups
values. Degradation degradation; too and the stiffness of its chemical bonds. For smaller groups in polymer systems
changes value little with the fundamental resonance and several harmonics fall in the infra-red and
thermosets leaves far-infra-red spectrum. Harmonics of these and of larger groups become app-
unreacted lossy arent in the microwave region.
polar groups In systems where dissipation factor peaks at a given temperature for a given
Orientation - Aligned fibres give Small effect frequency owing to a phase transition, the temperature is shifted upward for
higher frequencies. A plot of the logarithm of the peak loss frequency versus the
anisotropy
inverse absolute temperature will tend to be linear with a negative slope,
Composition More fibre or filler Fibres or filler proportional to the activation energy of the transition causing the peak.
raises the value usually raise the Interestingly, results with mechanical oscillation, such as the torsional pen-
value dulum, correlate well with electrical measurements.
Board fabrication Mobility of polar groups and internal friction influence dissipation factor.
Contamination Added dipoles raise Mobility increases with temperature. At the glass transition temperature T,, the
Solvents or moisture range of transition between glassy and rubbery states of a polymer amorphous
value slightly increase the value
phase, the internal friction and mobility are both high, and, if polarised groups
Operating are present, dissipation factor shows a peak against temperature.
temperature
Expansion Reduced dipole Little effect 15.1.1.2 Moisture absorption: As indicated previously, absorbed moisture is
density reduces of concern because of the adverse effects it has on electrical properties. Environ-
value mental conditions where cycling of humidity and temperature is encountered
Group Transitions increase Increased value, can lead to degradation of resistance to moisture absorption. Moisture penetra-
mobility dipole response especially in tion can also lead to corrosion of conductor traces and degradation of the bond
transition region between conductor and substrate.
Absorption can arise through the presence of pores or microvoids in the
CrystalIinity Raises density and May lower the value
substrate. Many polymers with polar groups have an affinity for moisture
thus value involving chemical bonding. Molecular features such as ester linkages, amide
Frequencv linkages, amine linkages, chloride groups or bromide groups are subject to
~hkrmai Increased frequency Peak loss temperatures hydrolysis; absorbed moisture reacts chemically with the polymer to change its
interrelation raises temperature shift upward with characteristics.
for transitions frequency
Harmonic of a Increase value 15.1.1.3 Foiladhesion: Foil adhesion is usually tested by measuring the force
Peak loss value needed to peel an etched strip of clad foil perpendicularly from the substrate.
molecular group
vibration The amount of force required is related to the thickness and stiffness of the foil
and to the modulus of the underlying material. For thicker or stiffer foil the
radius at the region of peeling will be larger, distributing peeling force over a
Table 15.1 Materials available as clad composites
%
Material Test Typical Tol. tan d Peel Water Temp. cn
description freq., K' of strength abs. Ti Tm
(see abbr. Hz k'% N/mm mg "C "C $
list below 3
Non-woven glass-PTFE 1M 2.15-2.35 c1 0.001 2.10 1 - 327 $,
10 G 2.15-2.35 <1 0,001 2.10 1 - 327 3
Woven-glass-PTFE 1M 2.5 2-5 0.002 1.40 1 - 327 In
10 G 2.4-2.6 1-2 0.002 1.40 1 - 327 Eb
Woven-glass-high-PTFE 10 G 2.15-2.35 <1 0.001 1.40 1 - 327 3
PTFE 10G 2.1 <1 0.0005 - 0.2 - 327 0
Ceramic-PTFE max 6.5 10 G 6.0 2-5 0.002 1.05 - - 327 0
Ceramic-PTFE max 11.0
XPS
10G
10 G
9.8-1 1.0
2.5
2-5
<1
0.002
0.002
1.40
-
-
-
-
115
327
-
5
o
Glass-XPS 1M 2.6 1-2 0.002 - - 115 - 0'
PES 1G 3.4 - 0.008 0.87 24 - 238 2
PSO 10G 3.0 - 0.006 0.87 15 - 185
Glass (10%)-PSO 10G 3.3 - 0.006 0.87 22 - 185
Mineral (10%)-PSO 10G 3.2 - 0.006 0.87 13 - 185
PEI 10G 3.0 - 0.004 1.05 13 - 215
Glass-PEI 10G 3.4 - 0.006 1.05 13 - 215
Woven-glass-epoxy 1M 4.7 >5 0.025 0.70 10 130 -
Woven-glass-PI 1M 4.5 >5 0,010 0.70 25 260 -
Woven-glass-T 1M 4.3 >5 0.015 0.87 15 200 -
Woven-glass-BTE IM 4.2 >5 0.015 0.87 15 180 -
WePTFE-epoxy 10G 2.8 2-5 0.012 1.05 10 125 -
WePTFE-PI 1M 2.8 2-5 0.010 0.70 25 260 -

BTE = bismaleimide-triazine-epoxy
PC = polycyanate resin
PEI = polyetherimide resin
PES = polyethersulfone resin
PI = polyimide resin
PSO = polysulfone resin
PTFE = poly(tetrafluoroethylene) resin
T = triazine resin
WePTFE = woven expanded PTFE
XPS = cross-linked polystyrene resin
 878 Advances in substrate technology
greater bond area. Lower modulus of the underlying substrate also has the effect
of distributing force over a larger area. Thus peel-test results can vary for
reasons other than quality of bond.
Some reagents for processing circuit patterns are capable of attacking inade-
quately bonded foil at the bond line to penetrate the interface, and cause costly
damage by adhesion failure of the circuit pattern during board processing.
The bond of foil to the substrate must be able to withstand soldering and
other lead-attachment processes to minimise the cost of damage during assem-
bly operations.
15.1.2 Comparative list of available substrates
A wide range of composites with various polymer systems can be considered for
microstrip antenna substrates. Table 15.1 lists materials available as clad lami-
nates. This list was compiled from supplier's information by the raw Materials
Subcommittee of the committee on ~ i ~ h - s ~ e e d / ~ i ~ h - ~Materials
r e ~ u e of
nc~
IPC.*
An explanation of the column headings in Table 15.1 follows. Documentation
of test methods may be found in IPC TM-650.
The test frequency, either 1MHz or lOGHz, is the producer's measurement
frequency for relative permittivity and dissipation factor. The typical K' value,
or range of values, indicates the nominal relative-permittivity value(s) available.
The Tolerance of K' is the lowest commercially available range for percentage
tolerance of nominal relative permittivity available. The tand column shows
typical dissipation-factor values.
The peel-strength values shown are the lowest values in Newtons per mm
width (converted from pounds force per inch width) specified for 34pm thick
(I oz/ft2) wrought copper-foil cladding for four conditions of test, including
as-received, after thermal-stress testing, a t elevated temperature and after ex-
posure to processing solutions.
Water absorption is a typical value for mass gain inmilligrammes for a 51 mm
(2in) square specimen, etched free of foil, that was conditioned for 1h at
+
105 2OC, weighed, immersed for 24 h in distilled water at 23 1C and then
reweighed.
The temperature data shown in degrees Celsius is characteristic of the pol-
ymer portion of the substrate material.
Thermoset polymer systems, including epoxy, bismaleimide-triazine-epoxy,
polycyanate resin, polyimide resin, triazine resin and cross-linked polystyrene
resin, are amorphous, thus having no crystalline melt point. They do show a
change in the amorphous phase from a glassy state of low molecular thermal
mobility to a rubbery state of high mobility and lower modulus. This change is
referred to as the glass transition T,. It is accompanied by an increase in the
thermal-expansion coefficient. With polymers having polar groups, relative
* Institute for
Interconnecting and Packaging Electronic Circuits, 7380 North Lincoln Avenue,
Lincolnwood, Illinois 60646, USA
Advances in substrate technology 879
permittivity increases and dissipation factor shows a peak at T,. Applications
for thermosets should avoid going above T,.
Crystalline thermoplastic materials include polyetherimide resin, poly(tetra-
fluoroethylene) resin (PTFE), polyethersulfone resin and polysulfone resin.
Crystalline melting is accompanied by an expansion. Above the crystalline melt
point T,, when the crystalline phase is in an amorphous melt state, there is a loss
of mechanical properties. For polymers with polar groups, the dissipation factor
is high. The useful temperature limit is usually well below T,.
15.1.3 Selection of metal cladding for performance
Copper foils used for clad laminates fall into two broad types: rolled and
electrodeposited. They differ widely in the processes used for producing them
and also show differences in how they perform in circuit-board applications.
Often the user must choose between high bond strength and low insertion loss
for a circuit in specifying the foil type.
15.1.3.1 Rolledfoil: One foil type is referred to as rolled foil, or wrought foil.
An ingot of copper is subjected to a series of passes through a rolling mill to
finally form it into a coil of rolled foil of uniformity of thickness dependent on
process factors including the condition of the rolling mill. Annealing steps in the
process tend to enlarge the crystal structure and improve ductility. Contamina-
tion of the metal with its oxides or other impurities is avoided as much as
possible. The crystal or grain structure of rolled copper foil tends toward
domains with boundaries running largely in the plane of the foil, as can be seen
in a microsection.
The rolling process results in a foil that has a polished appearance on both
sides. For many laminate substrates this smooth surface does not result in
adequate adhesion. Proprietary surface treatments are used by foil producers to
improve adhesion. These treatments usually consist of a deposit on one side of
the foil of attached nodules or dendrites of metal to give an opportunity for the
polymer of the substrate to form an interlocking mechanical bond. The de-
posited metal can vary, with zinc or nickel used in some cases. For microwave
laminates a copper deposit is preferred.
At least one rolled-foil producer is using an electro-etch process to remove
some of the copper on one side in order to leave a 'tooth' surface for bond
improvement.
Rolled foil is generally considered to be superior in ductility, conductivity,
low conductor loss when transmitting high-frequency energy, and freedom from
pinhole defects. It is generally inferior in bond strength attainable, rate of
etching and precision of etching for fine detail or close-tolerance conductor
features.
15.1.3.2 Electrodeposited foil: Electrodeposited foil is produced by con-
tinuous electrodeposition of copper into a non-reactive roll cathode. A highly
880 Advances in substrate technology
polished inert-metal forming roll is partially immersed in a plating bath. While
the forming-roll cathode is rotating, an electric current causes transport of
copper ions from the copper anode to the cathode where they are deposited as
a thin layer. This layer is continuously peeled off the inert metal surface. The
inert metal used for the cathode is most commonly a stainless-steel alloy,
although lead has been used. Copper-foil weight per unit area or the related
thickness is controlled by roll speed and current.
Electrodeposited copper has one side with shiny finish from the forming roll,
and the other side is dull in appearance. The dull side has a microscopic nodular
appearance, produced by the deposition process. The dull side is used for
bonding the foil to substrates. For many substrates the original nodular surface
is not sufficient, and a proprietary treatment is applied in a second operation to
generate smaller nodules on the original nodules to promote better adhesion.
Electrodeposited copper foil typically has a vertical or columnar grain struc-
ture with most grain boundaries perpendicular to the plane of the foil.
Electrodeposited foil is generally considered to be superior in bond strength,
etching rate and precision of fine detail in etched patterns. It is inferior in having
occasional pinholes, low conductivity, and conductor loss in transmission of
high-frequency energy.
15.1.3.3 Other foil features: Copper-foil cladding is supplied on some lami-
nates with an adherent black copper oxide on the outer surface for promoting
adhesion to 'prepreg' interlayers in multilayer board assemblies. This is of
special value for internal ground planes. For microwave applications the oxide
layer is undesirable owing to high loss of transmitted power. With the elevated
temperature used in producing laminates based on PTFE composites, the
adhesion of the oxide layer is likely to be destroyed.
Very thin copper foils, e.g., 118 oz/ft2 weight at 4 p m thickness, are too thin
to be handled practically in the lay up for laminating. Such foils are electro-
plated onto an aluminum-foil carrier. The laminator uses this foil composite as
supplied, and usually provides laminates clad with it still carrying the aluminum
foil.
The aluminum serves very well as protection until the laminate is to be
processed. The aluminum is then removed to leave a bright surface on the
underlying copper. This is done by etching the laminate in hydrochloric acid
diluted to a concentration of 5N. The laminate, after rinsing and drying, is
ready for application of photo-resist. Etching with sodium hydroxide, while
possible, is too energetic and does not always leave the copper surface as clean
as is desired.
15.1.4 Thermal characteristics of PTFE
PTFE-based substrates excel in properties needed for microwave systems, and
therefore receive most attention in this Chapter. Unfortunately PTFE is not a
perfect material for this application. The thermal characteristics of the polymer
Advances in substrate technology 881
need to be understood. This Section reviews some of the extensive studies
of thermal characteristics of PTFE reported in the literature. While PTFE
appears to be a simple chemical structure its thermal characteristics are rather
complex, and have proved interesting to many research workers, especially in
the 1950s.
Commercially available grades of PTFE are prepared in aqueous medium
either by suspension or by colloidal free-radical polymerisation processes. As
produced, the polymer is very nearly 100% crystalline. After it has been heated
above the crystalline melting point of 600 K (327"C), the maximum degree of
crystallinity is about 60%.
The crystalline phase undergoes an unusual transition in two steps at 292 K
(19C) and 303 K (30C) which may be detected as steps in the thermal-expan-
sion plot. Most of the transition occurs at the lower temperature. The amor-
phous phase follows a 213 rule [I] with corresponding transitions at 173K
(- 100C) and 400K (133C). The mechanical loss and conformance of PTFE
specimens at levels of crystallinity from 48 to 92% have been measured against
temperature by the torsional-pendulum method [2] to reveal the transitions.
Mechanical, electrical and nuclear-magnetic-resonance techniques for detecting
transition temperatures have been compared [3]. When the logarithm of fre-
quency for the test method is plotted against the inverse absolute temperature
in K units of transition, the plot is linear with a slope that corresponds to the
activation energy of the transition.
The 19-30C crystalline transition has been examined by infra-red spectro-
scopy [4], by linear thermal-expansion measurements [5], and by X-ray diffrac-
tion [6]. Studies of the conformational energy levels for PTFE indicate that the
molecules tend to be more rigid than polyethylene [7]. The crystalline phase
below 19C has been shown [8] to be triclinic with a. twist in the zigzag -chain
formation that is 180' over a distance of 13 carbon atoms distance. Above 19OC
the twist decreases to 180" over 15 carbon atoms, accompanied by some crystal-
line disorder. The crystal structure comes into a hexagonal alignment above
30C.
The thermal transitions of PTFE-based composites are evident in printed
circuit boards as changes in the relative permittivity related to density changes
of the polymer. Unfortunately the crystalline transition near room temperature
appears as a step change in density and relative permittivity. The 133C amor-
phous transition is difficult to observe electrically, although mechanical tests do
show a peak in lossiness around this temperature.
15.1.5 Anisotropy of relative permittivity
Anisotropy of the relative permittivity K' is the degree to which the property
varies in value depending on the direction of the electric field with respect to the
axes of the material. To simplify microwave circuit-design computations, one
usually assumes isotropy of K', i.e., equal values in the X, Y and Z directions,
of laminated substrates for microstrip or stripline circuitry. This can lead to
 882 Advances in substrate technology
error, particularly where fringing-capacitance values are critical for perfor-
mance.
The nominal K' value is typically obtained by a test method such as the
stripline-resonator test method, in which the electric field is predominantly in
the Z(thickness) direction. Normally this method gives no measure of degree of
anisotropy.
Advances in substrate technology 883
ceramic-PTFE composite with nominal K' of 10.5. In both cases a thick block
was prepared from thin layers of sheet stock using a laminating press cycle
typical of the production process. The card pairs cut from the blocks were
machine-finished to suitable thickness and tolerance for use with the stripline-
resonator method at about lOGHz [9] or the fluid-displacement method at
1 MHz [lo]. The electric field is essentially perpendicular to the plane of the
specimen in both methods.

Table 15.2 K' data versus electric-field direction for non-woven-glass

microfibre-PTFE substrate
Electric X Y Z
longitudinal transverse thickness
Thickness mm (in) 1.58 (0.0622) 1.53 (0.0604) 1.60 (0.063)
Relative permittivity
l MHz 2.428 2.430 2.330
10 GHz 2.452 2.432 2.347
Dissipation factor
1 MHz 0.00 14 0.0009 0.0005
10 GHz 0.0023 0.0024 0.0016

Fig. 15.1 Schematic for cutting stripline test cards from a thick panel

A technique for measuring K' of a substrate with the electric field in each of
the principal axes should ensure that the same method is applied to each axis for
directly comparable results. The stripline-resonator test method at 10 GHz has
been used with a series of specimens machined from a specially prepared thick
block of the material of interest. As illustrated in Fig. 15.1, pairs of test-
specimen cards are cut in planes perpendicular to each of the three principal
axes. Each pair is then measured.
This technique was applied to RT/duroidm 5870, a non-woven glass microbfi-
bre-PTFE composite with a nominal K' of 2.33, and to RT/duroid 6010, a
Rela t l v e Permittivity, K'
Fig. 15.2 K' anisotropy ratio plotted versus nominal K' for woven andnon- woven glass-PTFE
substrates
Glass-fibre-PTFE substrates, either woven or non-woven, have fibres lying in
the X Y plane, with K' nearly three times greater than for the polymer matrix.
The composite will be anisotropic with K;and K;.nearly equal, but both greater
than K''.
884 Advances in substrate technology Advances in substrate technology 885
Experimental results for the non-woven glass-PTFE are summarised in Table ,-RESONATOR TEST PATTERN CARD
15.2. The Z-direction results agree with accumulated quality-assurance test data
on production laminates. The degree of anisotropy may be expressed as the ratio
of the average of the X and Y values to the Z value. For this material, the degree
of anisotropy is 1.040 by the 10GHz method and 1.042 by the fluid-displace-
ment method.
Since the only source of this anisotropy is the glass-microfibre content, one
would expect the RT/duroid 5880 material at K' = 2.20, with about half the
volume fraction of fibre, to have an isotropy ratio of about 1.02.
These results are compared graphically in Fig. 15.2 with published values for
laminates based on woven glass-fibre-PTFE [I 11. They show that the random
glass-microfibre-PTFE structure is less anisotropic than a woven-fabric struc- M3 3 n n DIA. SCREW NUT
ture at the equivalent fibre content.
This difference may possibly be explained by considering the woven-fabric M2 2 n n DIA. SCREW & NUT
structure, especially at lower fibre content, to be a series of alternating polymer-
rich and fibre-rich layers. The Z-direction field in effect 'sees' a series-capacitor
network, while the X and Y fields 'see' a parallel-capacitor network. This layer L END
LAUNCH 3MM JACK - STRIPLINE
effect augments the fibre-orientation effect in the fibre-rich layers. CONNECTOR BODY
The results for the ceramic-PTFE substrate material are summarised in Table BASE COVER BOARD
SPACER BOARD WITH THICKNESS OF PATTERN CARD
15.3.
Fig. 15.3 Exploded side view of stripline fixture assembly
Table 15.3 Summary of anisotropy of K' data for RTlduroid 6070
Electric field X Y Z RAISED PRESSURE AREA AGAINST CLAMP PLATE
longitudinal transverse thickness THERMOCOUPLE WELL
/GROUND PLANE FOIL
Dielectric constant / J
Panel A 10.64 10.69 10.61 TEMPERATURE CONTROL
Panel B 10.80 10.67 10.61 /AND PRESSURE BLOCK
Panel C 10.60 10.74 10.61
Average 10.68 10.70 10.61 -M3 3 n n DIA. CAP SCREW
\TUBING FITTING POSITION
While the ceramic-PTFE substrate is a laminated-sheet product, a significant degree of anisotropy
of K' was not found .EMBEDDED STEEL BALL

/BASE STRIPLINE BOARD

15.2 Measurement of substrate properties ~ B A S COVER

E BOARD

The complex permittivity at microwave frequencies is of prime interest. Four BASE PLATE
methods are described. All of them use resonant measurement techniques that
could be implemented at minimal cost with good sensitivity to variations in the
substrate. \\ \-- SLIDE AND BLOCK ASSEMBLY

L ~ 3 n n DIA. SCREW & NUT, BRASS OR SS

They include the most widely used stripline-resonator method, the microstrip-
M2 2 n n DIA. SCREW & NUT
resonator method for high R substrates, the non-destructive full-sheet
resonance method of increasing use, and the specialised cavity-perturbation
method for characterisation of materials. Fig. 15.4 Face view of stripline fixture assembly
 886 Advances in substrate technology
The following Subsections give details of each method together with some
performance data. The relative merits are tabulated in Section 15.2.5.
15.2.1 Stripline-resonator test method
Versions of this method exist in a military specification [I21 and in private-sector
documents [9, 131. Some information is generally available supporting adap-
tation of the method to other materials or showing actual experimental deter-
mination of end-fringing correction [14]. The most recent updates of the docu-
mentation include provision for a range of nominal K' values from 2.2 to 10.5.
15.2.1.1 Brief description of the method: The method uses a fixture with
clamp plates, ground-plane foils and a permanent pattern card bearing probe
lines and resonator, as illustrated in Figs. 15.3 -15.5. The documentation [13]
provides complete drawings for the fixture hardware and electronic-measure-
ment equipment.
RESONATOR LENGTH
CHAMFER
1 RESONATOR IWIDTH
Nilnerlc dimensions In millimeters
Fig. 15.5 Generalisedresonatorpattern cardshowing dimensions of Table 75.5,andmade of
laminate matching nominal dielectric constant of material to be tested
The specimen, consisting of two unclad cards or stacks of cards to the correct
thickness, is interleaved on either side of the pattern card. A 4.448 kN (1000 Ib)
clamping force is applied before measurement. The resonant frequency f, and
bandwidthf, andf, at 3 dB down from resonance is determined experimentally
by observing power transmitted against frequency. From these data the relative
permittivity K' and dissipation factor tan d are determined by simple formulas:
K' = {Nc/[2f,(L + AL)]}'
tand = l/Q - l/Q,
where c = speed of light in a vacuum = 300mm/ns (this rounded value for c,
Advances in substrate technology 887
though inexact, is used for consistency with previous practice in the stripline
method). L = resonator length as measured. AL = length correction for end
fringing. N = number of nodes or half wavelengths in the material. I/
Q, = conductor loss.
Q = Lllh - hl (1 5.3)
Conductor loss may be calculated from the known properties of copper, fre-
quency, characteristic impedance of the section of transmission line comprising
the resonator, and cross-sectional geometry of the resonator [I 51. This is based on
the assumption that the resonator surfaces are smooth and free from lossy films
of copper oxides. Since typical resonator pattern cards are based on laminate
clad with copper foil having a surface treatment for adhesion, the estimate for
l/Q, is generally biased low, giving a high bias to dissipation-factor measure-
ments.
For a resonant peak free of unusual features, frequency readings at power
levels differing somewhat from the 3 dB level below the peak can be used to
calculate Q by a more exact formula:
l/Q = Ih - f,ll[L{ex~(Piln(10)) -
+ IL - f,I/[L{ex~(P2W o ) ) - 1)0'51 (15.4)
where f: is the power level in dB corresponding tof;.
15.2.1.2 Some factors to consider in the stripline method: The resonator is a
short length of stripline open at both ends. A signal induced by the probe at one
end will propagate in the transverse electric field mode, TEM, at a velocity
inversely proportional to the square root of the relative permittivity of the
dielectric. When it is excited at a resonant frequency by the probe, there will be
a self-reinforcing standing-wave pattern along the line with node number equal
to the number of half wavelengths and voltage maxima at the ends. End-fringing
capacitance causes the resonator to appear electrically longer than its measured
length L, by an amount AL discussed in Section 15.2.1.3.
Inverse Q is the ratio of dissipated power per cycle to stored power in the
resonator. For the TEM-mode of stripline, dissipation by radiation can essen-
tially be ignored. The other causes of dissipation are resistance losses in the
conductor, dissipation losses in the dielectric and coupling to the probe lines.
The Q with coupling losses is referred to as loaded Q, Q,, while an ideal
resonator with no coupling would have an unloaded Q, Q,.
The degree of coupling to probe lines is observed by comparing the trans-
mission insertion loss of the probe-resonator system at resonance with the same
fixture having a straight-through transmission line on the pattern card. A wide
gap between probe and resonator reduces coupling to bring insertion loss above
30 dB. When both probe gaps are equal, the following relationship between Q,
and Q, can be derived from given formulas [16]:
QLIQu = 1 - (PLIPO)~~ (15.5)
888 Advances in substrate technology Advances in substrate technology 889
where PL = transmitted power with the resonator at resonance. Po = trans- frequency of resonance. From eqn. 15.1 these 12 data values have a linear
mitted power with the straight-through line. dB insertion loss = - 10logl0(PL/ relationship of the form
Po). The loaded Q differs from the unloaded Q as follows:
FLIN = Bo + B, FIN (15.7)
dB insertion loss 1 5 10 15 20 25 30 35 40
QL, as % less than Q, 89 56 32 18 10 5.6 3.2 1.8 1.0

Losses due to radiation are assumed negligible in stripline. Losses due to

coupling to probe lines are made low enough to ignore. The total loss is then
assumed to consist of conductor resistive losses and dielectric losses as in eqn.
15.2.
The end-fringing-capacitance correction AL, used in the relative-permittivity
calculation, has been estimated from a published formula [17]:

where K = Bln(2)/n, B = ground-plane spacing, W = resonator width. With

W = 6.35 mm and B = 3.39 mm, AL is predicted to be 1.28 mm. Experimen-
tally AL was found to be about 1.42mm for substrates of non-woven glass/
PTFE at K' values of 2.20 and 2.33. This difference arises from several causes.
The test fixture with specimen is offset from centre by half the dielectric thick-
ness of the resonator pattern card. The resonator conductor has a finite thick-
ness ignored in Reference 17. Capacitance is increased by the proximity of the
probe lines. The substrate under test has anisotropic K' from the loading with
oriented glass fibre; i.e., K'' or K;, with the electric field parallel to the X or Y
axes of the substrate, is higher than K', and end-fringing capacitance is larger
than predicted.
F o r K' = 2.20, node 4

I-
F o r K' = 2.20, node 3
15.2.1.3 Experimental determination of end-fringing correction: A series of
test pattern cards with resonator lengths for I-, 2-, 3- and 4-node resonances
near the 10GHz test frequency are prepared so that they are identical in gap
size, probe width etc. Fig. 15.6 shows typical photomask artwork for this
purpose. The resonator length L is measured by optical comparator or X, Y
co-ordinograph for each of the resonator pattern cards.
Specimens will show variations of thickness from the fixture design value with
associated variation in the AL fringing correction. An experimental procedure
to determine AL as it varies with specimen thickness i s based on a series of
specimens representing the range of thickness to be encountered. The average
thickness T of each specimen pair is determined.
Fig. 15.6 Examples o f pattern-card artwork for probe-line impedance and AL determination
With each resonator pattern card in turn mounted in the test fixture, re-
sonant-frequency readings F a r e obtained on each of the selected series of test
specimens. It is good practice to replicate this series of readings with each where
pattern card and use the average F value. Bo = C/(2K'0'5) and B, = -AL
For each test specimen there will be four sets of F, L and N data, where N is
the number of nodes of the resonance, L the resonator length and F the The slope B, and Y-axis intercept Bocan be obtained by plotting, as in Fig. 15.7,
or by numerical linear regression analysis to yield AL values for each specimen.
 890 Advances in substrate technology Advances in substrate technology 891

When this is done, the average thickness T and AL data for all the selected
specimens may be plotted to correlate AL with average thickness T, as in Fig.
15.8, with good results.
Experimental data for two substrate types are summarised in Table 15.4.

Table 15.4 Summary of experimental AL values as a function of specimen

thickness
Material RT/Duroid 5880 RT/Duroid 5870
Nominal K' 2.20 2.33
Co, null 0.032 0.150
Std. dev. about fit, mm 0.005563 0.006 121
c, 0.864486 0.800966
Std. err. of est. 0.09462 1 0.09575 1
AL for 1.524mm thickness 1.349 1.371 .
(Res, f r e q . ) / (Nodes), I / n s AL for 1.626 mm thickness 1.438 1.453
Fig. 15.7 Example of plot of FLIN versus F I N for a specimen of RTJduroid 5880 with the AL = C,,+ C, (thickness)
series of pattern cards for nodes I to 4

15.2.1.4 Adaptation of the method: The latest documentation [13] provides

for a wide range of R values at convenient specimen thicknesses, as summarised
in Table 15.5. The following comments apply to the Table.
Chamfer values are based on published design curves [la].
Calculation of probe-line widths, for 50Q characteristic impedance Z,, ass-
umes that the ground-plane spacing is twice the nominal specimen thickness
plus the thickness of the pattern card and its 34pm thick (1 oz/ft2) copper-foil
pattern. The simplified case of the stripline centred between ground planes is
used [19, 201.
To use the stripline-resonator test for a new substrate type, test pattern cards
are prepared from a clad laminate of the substrate at 0.216mm (0.0085in)
dielectric thickness. A 0 node pattern is included which consists of a straight-
through line (no resonator) to be used to verify the 50Q value of Z, for the
probe-line width.
The fixture is assembled with the new substrate type used for the base cards.
Specimens of the 0.216mm-thick dielectric laminate used for the test pattern
cards are stacked in the fixture and tested with the 1-, 2-, 3- and 4-node pattern
cards to verify that its K' matches that of the thicker substrate to be routinely
tested.
The I-, 2-, 3- and 4-node pattern cards are used with a series of substrate
specimens covering the expected range of thicknesses to determine AL as a
1.52 1.54 1.56 1.58 1.60 1.62 function of specimen thickness, as outlined previously.
Specimen thickness, mm
Fig. 15.8 Plot of experimentally determined AL versus specimen thickness for a series of 15.2.1.5 Considerationsfor precision and accuracy: Two sources of possible
specimens of differing thickness error were studied in the laboratory. The test fixture with resonator pattern card
 892 Advances in substrate technology Advances in substrate technology 893

could develop a bias with use over a period of time; and the resonator pattern electric field parallel to the plane of the substrate. Thus a value for K' obtained
card does wear with use and requires replacement. by the stripiine method must be understood to include both Z and X-Y-plane
A series of 12 test specimens were used for each of two non-woven glass- components.
PTFE substrates at R values of 2.20 (RTIduroid 5880) and 2.33 (RTIduroid There are some other sources of error in the test method. A fixed-frequency
5870). For each of these materials a test fixture was prepared with a 4-node test resonant cavity traceable to an acceptable standard should be used to verify
pattern card in place. The specimens were tested in rotation in ten test sessions performance of the frequency counter. Resonator distortion can be detected by
looking for drift in reference-standard specimens and verifying it by actual
Table 15.5 Dimensions for stripline test-pattern cards in mm referring to re-measurement. Erratic performance or multiple resonant peaks may be ev-
Fig. 15.5 idence of fatigue fracture of the probe lines where they leave the base-card
region.
Nom. Nom. Patter- Probe Chamfer Probe Resonator Conductor
K' thk. card thk. width X, Y gap width length loss as Table 15.6 Summary of K' readings for specimens and patterns
4 node l/Q, 2.20 2.33
Nominal K'
MIL-P13949F [I 21 Number of specimens 12 12
2.55 1.588 0.216 2.18 2.56 2 3 4 6.35 38.1 0.0006 Number of readings per specimen 10 10
Mean of standard deviations of K' readings 0.00131 0.00 164
on specimens with a single pattern card
Std. deviation of the specimen std. deviations 0.00032 0.00022
Number of 4-node pattern cards 18 10
Mean of standard deviations of R readings 0.00313 0.002 12
on specimens when changing pattern cards
Std. deviation of pattern card std. deviations 0.000 19 0.000 13

Earlier adaptation [I41

10.5 1.27 0.254
Note: See comments in 15.2.1.4.
0.42 radius 2.03 2.54 17.3 0-0009
. .
RESONATOR

. . $---I
TRIMMED SIZE
-

over a period of more than five working days. For each specimen the standard
deviation for K' readings with a single pattern card was determined. For each
type, a series of 18 or 10 test pattern cards were used for a measurement session
with specimens in rotation. The standard deviation of K' was determined for
each specimen as pattern cards were changed.
Table 15.6 summarises the results of the study. The stripline test applied to
laminates in the K' = 2.0-2.5 range should have a precision of better than
0.5%, considering the drift of the fixture and pattern-card changes.
Accuracy is related to certain features of the test method. The K' value
obtained is based largely on the case of electric field perpendicular to the plane
of the substrate, i.e., Z direction. However, part of the electric field in the
fringing region along the lengthwise edges of the resonator has a component of
1.27 mm THICKNESS 0.635 mm THICKNESS
Fig. 15.9 Plan views of the microstrip test specimens for 1.27 and 0.635mm dielectric
thickness
15.2.2 Microstrip-resonator test method
The microstrip-resonator test method has been used for several years with
ceramic-PTFE soft substrates for K in the 9.5-1 1.0 range, where the appli-
cation is for microstrip devices not necessarily limited to antennas. Variations
of the method have been specified by at least one user, and it has been in routine
use by at least two suppliers. It is most readily used for substrates clad on one
side with thick metal.
 894 Advances in substrate technology
15.2.2.1 Brief description of the microstrip-resonator method: Rectangular
specimen cards 30.5 x 19.0mm (1.20 x 0.75in) with thick metal backing are
prepared by photomasking and etching a pattern consisting of two probe lines
and a resonator element, as shown in Fig. 15.9. The probe-line width is designed
to match the 50R characteristic impedance of the test system.
BRASS COVER BLOCK
r.----7/
Advances in substrate technology 895
fittings and contact the probe lines to the pins. The movable end plate is applied
with force to ensure ground contact of the coaxial fittings to the thick metal
backing on the specimen. A metal cover is placed over the mounted specimen.
The resonant frequency f, and bandwidth, f, and f , , at 3 dB down from the
resonance are determined experimentally by observing power transmitted
through the fixture against frequency. From these data the relative permittivity
K' and dissipation factor tan d are determined from the simplified formula (eqn.

-
-
p- MDVABLE END PLATE
FIXED END PLATE

RIGHT ANGLE COAX

BULKHEAD FEEDTHROUGH
FITTING WITH PIN I N
15.8) with constants A-F specified as in Table 15.7.

where L = length of the resonator, in.

I BASE BLOCK
SIDE VIEW
The l / Q value may be determined from eqn. 15.3 or 15.4 for the stripline
method. For the microstrip method, the observed dissipation of the resonator
CONTACT WITH PRUBE
element is the sum of losses from radiation, conductor resistive loss, coupling
LINE ON SPECIMEN loss and loss in the dielectric. Normally no attempt is made to separate these
effects, and the material under test is rated by the Q value of the resonator.
The Q measurement is made after the specimen has been exposed to con-
\ \ I ditions representative of a circuit-board processing environment, as a way of
SPECIMEN, 0.635 nn DIELECTRIC
predicting the sensitivity of the substrate to impairment of electrical perfor-

\ WITH RESONATOR PATTERN ON TOP

AND THICK METAL CLADDING
ON BOTTOM

SPRING LOADED L I F T PLATE

mance.

END 'VIEW
Fig. 15.10 Side and end views of the microstrip test fixture with specimen in place
To perform a measurement, the specimen is mounted in the test fixture shown
in Fig. 15.10 to provide connection of the specimen ground plane and probe
lines into 3mm coaxial connectors. The spring-loaded floor of the fixture is
pushed down so that the specimen can fit under the centre pins of the coaxial
15.2.2.2 Some factors to consider in the microstrip method: This method may
be adapted to other values of substrate thickness and nominal relative permit-
tivity. The resonator pattern would require redesign. More exact formulas
relating effective K' to substrate K' [21, 221 would need to be used to derive a
simplified formula for computing K' from the resonant frequency.
The length of the resonator on each specimen should be read accurately by
an optical comparator or X, Y co-ordinograph to avoid test-result bias by
pattern variation from the photomask dimension.
The metal cover specified for the test is subject to variation in the quality of
contact it makes to ground. The presence of the cover affects the resonant
frequency. For comparability of measurements from one specimen to another,
they should always be made with the cover in place. The intention of the metal
cover is to remove the effect of radiation losses from the Q measurement. Based
on experience, there is some question as to whether this cover is effective for the
purpose.
The method is primarily intended for use with laminates clad on one side with
thick metal, typically 3.2mm (0.125in) thick or more. When specimens made
from laminates clad on both sides with thin copper foil are to be used a thick
metal back-up plate is also used.
15.2.2.3 Considerationsforprecision andaccuracy: This test method does not
use a test-pattern card that is part of the fixture, as is used for the stripline-
resonator method. The possible source of bias imposed by the pattern card is
 896 Advances in substrate technology Advances in substrate technology 897
thus eliminated. This method seems rather more relevant to microstrip appli- is based on the assumed values of dielectric thickness and resonator width.
cations, although the section of line used for the resonator is wider than a Ignoring these will bias the calculated results.
microstrip transmission line at the typical 50R characteristic impedance. More exact formulas [21, 221 were used to calculate the bias associated with
The metal cover has been observed to change the resonant frequency, and errors in the assumed resonator width and dielectric thickness, as shown in
hence the computed K' value. Constant F for eqn. 15.18, listed in Table 15.7, Table 15.8.
includes a correction for this test bias. The quality of fit of metal cover over the
specimen and end plates is expected to influence accuracy.
Table 15.8 Predicted bias in calculated bulk K' from ignoring specimen
dimension variations in the microstrip resonator test method
Table 15.7 Microstrip resonator test: pattern design data and constants for
calculation of results Bulk Dielectric Resonator Bias in calculated bulk K'
Nominal K' 10.2 10.2
.. -
R thickness, mm width, mm +
ignoring a 25 pm variation in:
Diele. thk. Resonator width
Dielectric thickness 0.635 (0.025) 1.27 (0.050)
Card length 30.5 (1.20) 30.5 (1.20) 10.2 0.635 3.18 -0.39 % +0.10 %
Card width 19.0 (0.75) 19.0 (0.75) 10.2 1.27 3.18 -0.14 % +0.10 %
Probe-line width 0.53 (0.021) 1.14 (0.045) 2.2 0.79 5.08 -0.18 % +0.031 %
Probe-resonator gap 0.46 (0.018) 1.78 (0.070) 2.2 1.58 5.08 - 0.084 % f0.031 %
Resonator length 17.2 (0.676) 17.8 (0.700)
Resonator width 0.318 (0.125) 0.318 (0.125)
Number of nodes 2 2
Constants for calculation 15.2.3 Full-sheet-resonance test method
of K by eqn. 15.8 using The full-sheet-resonance (FSR) test method, also called the parallel-plate-
length in inches: waveguide resonator method, is a valuable non-destructive method for measur-
AL 0.0205 0.0375 ing the dielectric constant at microwave frequencies of clad laminate panels.
A 69.73 34.869 While the FSR method is beginning to find use in individual cases with users and
B 0.0 0.3403 producers of clad laminates for microwave applications, a standard procedure
C 0,175 - 0.2992 or apparatus has not yet been established in the industry.
D 0.0089 0.0098
E 1.533 0.7008 15.2.3.1 Brief description of the full-sheet-resonator method: In the full-sheet-
F 0.14 0.33 resonator (FSR) test method a metal-clad laminate panel, trimmed to the
rectangular size to be supplied or used, is treated as a parallel-plate-waveguide
D~mensionsare in mm (in)
resonator in which the electric field is in the Z direction and the open edges form
The end-fringing effect AL in the calculation of K' is based on consideration an electrical boundary. Probe connections are made at two edge or corner
of the ratio resonator-widthlheight, but does not include the effect of resonator- positions, as shown in example of Fig. 15.11. The frequencies of selected
conductor thickness, proximity of the probe line and its width, or the anisotropy unambiguous resonant modes are measured experimentally by observing power
of the dielectric. Accuracy of the method would be improved by an empirical transmission between probes against frequency.
derivation of this constant for the conditions of the test method. The approach Two probe and specimen fixtures, shown in side and face views in Fig. 15.12,
used for stripline in Section 15.2.1.3 would be applicable. provide a convenient and non-destructive way to couple probes to the specimen.
The etched pattern can also vary in the gap between the probe lines and They can accommodate a wide range of panel sizes and thicknesses, with o r
resonator. This variation will have an, as yet undetermined, effect on the value without thick metal backing. They allow precise air-gap adjustment to reduce
to be used for the AL end-fringing correction. If the gap is larger due to the coupling of probes.
over-etching, the AL should be smaller. A simple formula is used to calculate K' for an observed resonant frequency
The simplified formula provides for a fixed resonator width and substrate of known mode number [M:N], ignoring the fringing-capacitance effect of the
thickness with no provision for actual variations between specimens. The open edges forming the electrical boundary.
relationship of effective K', obtained by the test, to the bulk K' of the substrate
898 Advances in substrate technology Advances in substrate technology 899
he ref,,^,.^ = resonant frequency, GHz, c = speed of light, 299.792 mm/ns,
M = integer number of nodes along length, L = length, mm, N = integer
number of nodes along width, W = width, mm, K' = dielectric constant of the
substrate.

MARKER FOR NODE SHOWING VOLTAGE POLARITY

AT A G I V E N INSTANT FOR MODE C3t21

COAXIAL PROBE CONNECTION TO CENTER OF EDGE WITH

GROUND TO TOP PLATE AND CENTER P I N TO BOTTOM

Fig. 15.1 1 Schematic of coaxial-probe connection to parallel-plate resonator with voltage

nodes for a typical resonant mode indicated by arrows

15.2.3.2 Some factors to consider in the FSR method: The FSR method
applied to microwave integrated-circuit substrates used the plane of an APC-7
coaxial connector for coupling to the metallised substrate [23, 241. The fixture
of Fig. 15.12 combines specimen support with better adjustability. Firm electri-
cal contact of the coaxial ground of the probe is made to the top metal edge,
while either contact or air capacitive coupling is made between the centre pin
and the lower metal edge. The ground contact assembly is lifted against spring
tension to permit placing the specimen.
Measurements at controlled temperatures may be carried out by clamping a
specimen between two aluminum blocks, as in Fig. 15.13, through which fluid
is circulated from a constant-temperature bath. Specimen size exceeds block
dimensions by about 1 mm.
For a given K' the limited useful frequency range for testing increases as the
panel size decreases.
The calculation is based on two assumptions: the cladding on the top and
 900 Advances in substrate technology
bottom of the specimen is parallel, i.e., thickness of the substrate is uniform; K'
of the substrate is uniform over the area of the specimen. If these assumptions
are false, inconsistent measurements can arise.
-----------
DRILL HELE AND TAP BOTH ENDS
FUR 3 nn PIPE FITTINGS
Fig. 15.13 Aluminum block for temperature control of FSR specimens: two required
The three fundamental resonant modes are of two kinds. First, the [I :0] and
[O: 11 modes have voltage maxima oppositely polarised along two opposite
edges. Secondly, the [1:1] mode has voltage maxima of alternating polarity
around the four corners. Higher-order resonant modes simply involve multiple
patterns of a fundamental mode. The node number M or N is the pattern count
lengthwise or widthwise. A specimen at resonance [M:N ] has ( N +
1)(M 1) +
voltage maxima with 2(M + N ) on the edges. Resonant modes with odd-node
numbers along an edge are suppressed when a probe is centered on the edge. The
f; for a given mode is associated with an average K' for the panel's area weighted
for regions of voltage maxima.
Observed values off, differ somewhat from the prediction of eqn. 15.9 (lower
apparent K') than [O:N] or [M:O]modes. This deviation agrees in sign with, but
Advances in substrate technology 907
is at less magnitude than, earlier work on metallised integrated-circuit substrates
1241.
The influence on apparent K' of extra fringing capacitance of open edges is
not well explained. Intuitively it seems that some modes have fringing field in
air along the edge, acting as a heterogeneous medium for lower KiPp.Other
modes 'see' the fringing as increasing electrical length - in effect, increasing K:pp.
Some modes will predominate in one effect or the other, while others may tend
to cancel the two effects. For lack of an accepted model the effect is ignored,
which appears adequate for specimens of small thickness versus length and
width.
. .
I .
0.0 1.0 2.0 3.0 4.0 5.0
CHz
Fig. 15.14 Diagram of expectedresonancesfor a 254 x 102mm (10 x 4in) cladpanel with
nominal K' of 70.2.
--- Selected modes above 0.5GHz
15.2.3.3 Ambiguous mode limitation of the FSR method: The spectrum of
resonant modes for FSR rapidly becomes too dense for practical use as fre-
quency increases for a given specimen size. For the example in Fig. 15.14
suppose that the substrate K' is known to be always within 5% tolerance of
nominal K', but measurement has to be done to detect boards that fall outside
a 2% tolerance. The 5% possible tolerance of K' will be observed as a possible
2.5% tolerance of the expected resonant frequency for a given mode.
A mode is ambiguous if the expected nominal frequency is in a region of the
spectrum where adjacent frequencies on either side are within the 2.5%
tolerance; i.e., there is no assurance as to whether a frequency observed in the
expected region is actually the expected mode or not. In the crowded part of the
spectrum one can surely find a resonance close enough to indicate whatever
value of K' is desired, whether it is correctly assigned a mode or not.
Thus selection of unambiguous modes for a particular specimen length/
width ratio is an essential part of the method. It is desirable to select unam-
biguous modes of higher order, if possible, in order to obtain a larger number
of voltage maxima better representing the area of the laminate specimen.
In the example of Fig. 15.14 all the modes except four have nominal frequen-
cies with neighbours closer than the 2.5% test, and thus are ambiguous. The five
 902 Advances in substrate technology Advances in substrate technology 903
frequencies most widely spaced from their nearest neighbours are: (I) Set L = nominal L minus largest deviation expected. Set W = nominal W
plus largest deviation expected. Repeat steps (d)-(k). Retain as unambiguous
Mode Nominal Closeness of and usable only those modes that were found before and also this time.
f, GHz nearest other (m) Set L = nominal L plus largest deviation expected. Set W = nominal W
f,, % minus largest deviation expected. Repeat steps (d)-(k). Retain as unambiguous
[3:01 0554 16.66 accent and usable only those modes that were found before and also this time.
[4:0] 0.739 12.50 accept
[4:2] 1.183 3.15 accept 15.2.3.4 Test procedure for clad laminates: Test specimens are clad laminate
[2:2) 0.995 2.66 accept panels trimmed accurately to a rectangular shape. All four edges should be
[7:21 1.589 2.31 reject straight and smooth without edge contamination such as metallic chips that
might occur with cutting. All four corners should be right angles. Failure to
achieve smooth, straight, clean, perpendicular edges will impair the accuracy of
The FSR method will usually be applied to large numbers of panels of the the test. Experimental data are not yet available on the effect of these parameters
same nominal length, width and K'. Practical application of FSR requires not on accuracy.
only consideration of possible variations of K' from the nominal but also of the Measure specimens for length and width as the distance between centres of
length and width. Variation of the aspect ratio of length to width alters the opposite edges. Use either a vernier caliper or an X , Y co-ordinograph. Record
relative position of resonances in the spectrum, and could make previously dimensions to the degree of resolution of the measuring equipment. The accu-
selected modes ambiguous. Thus the possible range of aspect ratio one could racy of dimensional readings influence the accuracy of the test. An error of 1%
encounter must also be considered in the selection of modes for test. in a dimensional reading can cause a calculated relative-permittivity error of up
The following stepwise procedure, best implemented by a computer program, to 2%, depending on the resonant mode used.
may be used for selection of unambiguous resonances:
15.2.3.5 Precision and accuracy, supporting data: Almost all data on preci-
(a) Determine the nominal and largest expected deviation from nominal for the
sion and accuracy of the FSR method obtained so far have been with high-K'
parameters of length L, width Wand relative permittivity K'; Set L = nominal
length; set W = nominal width. ceramic-PTFE substrates, such as RT/duroid 6010 laminate. The method is
(b) Decide whether probes will be on corners or centred on opposite edges. If
expected to be as useful with low-K' substrates such as glass-fibre-PTFE sub-
centred on edges, use the distance between those edges as the length, and exclude strates.
all modes with odd-width node values N from this selection procedure. For
15.2.3.5.1 Effect of temperature: Aluminum blocks, 254 x 96mm, as men-
probes at corners the increment for N = 1; otherwise it is 2.
(c) Decide on an upper frequency limit for consideration, f,.
tioned in Section 15.2.3.2, were used to control the temperature of two spe-
cimens of nominal 254 x 108 mm size, 0.64mm dielectric thickness, 34pm
( d ) Set N = Oand M = 1.
(e) Use eqn. 15.9 to predict a value off,. Record the values for M, N and S,.
copper-foil clad both sides. Mean K' values from modes [3:0], [4:0] and [32]
( f ) Increment M by 1, and repeat step (e) until f, exceeds f;,. were obtained in replicate for several temperatures, as summarised in Table
15.9. Where corresponding data were available, no difference in K' was observed
(g) Set M = 0 and increment N by the value I or 2 determined by probe
position. Perform steps (e) and Cf). with or without the aluminum blocks at 23OC ambient temperature.
(h) Repeat (g) until the firstf, for a given value of N is found to exceedf;,,. Temperature effect is similar to that observed with stripline-resonator meas-
( i ) Rearrange the sets of M, N andf, values in order of increasing value off,.
urements, and indicates that varying the laboratory temperature over a 20-26OC
( j ) For each set, determine which adjacent set has the nearestf, value to be used
range will account for less than 0.2% error in K'.
asf,,,,. Determine percentage proximity value P.
15.2.3.5.2 Effect of decoupling
Opening capacitive gaps of probes to increase the insertion loss at resonance did
( k ) Rearrange the sets of M, N andf, values in order of decreasing P value, and increase Q as expected, but has not shiftedf,, as predicted in earlier work with
note as unambiguous those modes with P greater than half the percentage value integrated-circuit substrates [24]. It does, however, yield 'cleaner' resonant
for the largest expected deviation of K' from nominal. peaks; i.e., the plot of dB versus frequency became smoother and showed less
low-level noise above and below resonance.
 904 Advances in substrate technology Advances in substrate technology 905
Work on metallised integrated-circuit substrates used coupling at a corner tions of replicate readings for specimens averaged 0.008 for K' z 10.0 in the
with the face of an APC7 coaxial connector [25]. Perturbation errors were laboratory with better temperature control.
introduced by the method of coupling into either closed or open cavities. The
conclusion was that, in the [O:N] mode, the electric- and magnetic-field effects 15.2.3.5.4 Consistency among specimen sizes and mode selections
of the probe cancel. The recommendation was to ignore all but the zero-mode A study to evalute the FSR method for consistency of test results with respect
resonances. Experience with the method applied to laminates of larger area with to specimen size was performed with RT/duroid 6010.5 high-K' ceramic-PTFE
the fixture of Fig. 15.12 seems to indicate less error for any mode. composite substrate. The study started with six panels of nominal
508 x 254 mm size, 0.64 mm dielectric thickness with 34 pm copper clad both
Table 15.9 K' values by FSR norrnalised as percentage difference from 23C sides. These were measured at full size and successively cut into smaller portions
value which were again measured and averaged. Table 15.11 summarises the sizes,
number of portions and average change in K' reading from that of the original
Specimen A Specimen B
Length, mm 254.7 254.4 size.
Width, mm 108.7 108.4
Table 15.10 Average K' value summary of FSR repeatability study on seven
Temperature Mean K' % diff. Mean K' % diff. specimens at nominal 2 5 4 x 108rnm size with modes [3:0],
setting, OC - [4:0] and [3:2] -
13 10.828 0.637 10.674 0.386 Specimen A B C D G H I
14 10.817 0.534 10.678 0.423 Supplier R R M M R R R
23 10.754 - 0,051 10.630 - 0.028
Thick metal alum. alum. alum. alum. brass brass brass
23 10.765 0.05 1 10.636 0.028 thk.. mm 3.18 3.18 3.18 3.18 6.35 6.35 6.35
32 10.702 - 0.534 10.587 - 0.433 Group X
Mean value 10.201 10.069 9.733 9.968 10.165 10.376 10.077
32 10.727 - 0.302 10.593 - 0.376 Std. dev. 0.005 0.008 0.008 0.010 0.001 0.013 0.012
40 10.660 - 0.925 10.562 - 0.668 Number 4 4 4 4 3 4 4
40 10.669 - 0.841 10.564 - 0.649 I
Group Y
45 10.641 - 1.101 10343 - 0.846 Meanvalue 10.189 10.072 9.739 9.962 10.172 10.388 10.080
45 10.642 - 1.092 10.540 - 0.875 Std. dev. 0.010 0.010 0.010 0.002 0.013 0.001 0.003
The average K' value from resonances of modes [3:0],[4:0]and [3:2]are shown for each temperature Number 4 4 4 4 4 4 4
Group Y
15.2.3.5.3 Repeatability Mean value 10.166 10.035 9.722 9.939 10.134 10.371 10,082
With three selected unambiguous modes, the FSR method has proved to be Std. dev. 0.019 0.014 0.014 0.020 0.004 0.004 0.018
reproducible in replicate test runs, comparing laboratories and test fixtures for Number 5 4 3 4 2 2 3
a series of specimens. Group X Rotated replicate runs with fixture I , laboratory E
Seven specimens from two suppliers of high-K ceramic-PTFE substrate with Group Y:Rotated replicate runs with fixture 2, laboratory E
Group Z: Rotated replicate Nns with fixture 2, laboratory W
nominal 254 x 104mm size were measured repeatedly in rotation at selected Laboratory W had higher and less controlled ambient temperature
modes [3:0], [4:0] and [3:2]. The dielectric thickness was 0.64mm with thick
metal cladding one side and 34pm-thick copper on the other. For each run an The results are consistent until the final 76 x 76mm sizes, which showed
average of the three modes was used as the K'-value. Summarised data are given lower K' values.
in Table 15.10.
Replicate runs comparing two fixtures and two laboratories showed insignifi- 15.2.3.5.5 CorreLation with the stripline-resonator test
cant fixture difference. The difference between laboratories is attributed to Among the 72 specimens of 76 x 76 mm size resulting from the study discussed
poorer temperature control in one, which was later improved. Standard devia- in Section 15.2.3.5.4, seven pairs were selected to have nearly identical average
 906 Advances in substrate technology
K' values by the FSR method and to cover a range of K' values. Specimens were
prepared by cutting the two 76 x 76 mm cards into four 38 x 76 mm cards to
be stacked to the nominal 1.27 mm thickness on both sides of the stripline-
resonator pattern card for testing. K' was measured by the stripline-resonator
method.
Advances in substrate technology 907
3 dB down from resonance is determined experimentally by observing the power
transmitted versus frequency. Electronic equipment for this may be the same as
is required for the stripline-resonator method [13], except that a higher degree
of frequency resolution is needed. The Q value is determined from eqn. 15.3 or
15.4 in Section 15.2.1.1.

Table 15.11 Summary of average K' values by FSR for panels successively
divided andlor trimmed
Nominal size, mm 508 x 254 254 x 254 254 x 108 254 x 96 76 x 76
FSR modes [5 :21 [3:21 [3 :21 [3:01 [I :0]
selected [4:21 [2:21 [3:01 [4 :01 [3 :21
[6:01 [3:01 [4:0] [4:21 [I :2]
No. of specimens 1 2 4 4 12
averaged per panel
Mean change of average K' for panel from initial size
As change in K' 0.000 0.003 - 0.020 - 0.015 - 0.200
AS % of R 0.000 0.028 -0.19 -0.014 -1.90

The FSR K' values for each mode and their means are plotted versus stripline
R values in Fig. 15.15. Correlation coefficients of the [I :0], [3:2], [1:2] and mean
FSR values versus the stripline values were 0.97,049,0.95 and 0.95, respective-
ly. As can be seen from the plots, the population of specimens seems to fall in
a similar K' region from either method, but the slope shows the FSR method
to be about twice as sensitive to variations in specimen K' as the stripline
method. This is consistent with the use of an unchanging pattern card adjacent
to the resonator in the stripline method.
We conclude that the FSR method is more sensitive to K' variations than the
stripline method, but correlates strongly with it. Specifications calling for the
method should be wider in tolerance than they would be for the stripline-reso-
nator method.
10.5 10.6 10.7 10.8
15.2.4 Perturbation cavity method K' by strlpline method
The resonant-cavity perturbation method of test for complex permittivity is
F i g . 15.15 Plot of K' values by modes and mean FSR versus stripline K' values
discussed here because it has some unique features useful for characterising
substrate materials. A standard method for measuring the dielectric constant
and dissipation factor (complex permittivity) is published [26]. It is primarily To be useful for microstrip-antenna substrate materials, the means for load-
intended for testing ceramic materials at about 10 GHz. ing the specimen into the cavity must have minimal effect on the emptyf, and
Q values. Fig. 15.16 shows a sketch of an effective perturbation cavity. The
15.2.4.1 Brief description of the method: A section of rectangular waveguide flanged joint at the centre of the cavity length is fitted with dowel pins and
is provided with flanges at the ends. At each end an iris plate is clamped between alignment holes. Clamping is done in the same way each time with lever-type
the waveguide flange and the transition into coaxial cable for connection to the clamps applied to two specific locations on the flanges.
electronic equipment. The resonant frequency, f, and bandwidth, f,and f2, at Specimen types described in the standard [26] include vertical bar, vertical
 908 Advances in substrate technology Advances in substrate technology 909
I
rod, horizontal rod, thin strip and sphere, all with dimensions qualitatively where
described as much smaller than the cavity. This discussion will focus on the
vertical-bar and sphere types most useful for microstrip antenna needs.
A cavity size for transverse electrical resonance mode TE,,, at the desired
frequency is selected with inside dimensions such that the length is seven times
= volume of waveguide cavity, = volume of specimen, f , = f, of empty
the width and the height is half the width. The TE-mode numbers designate
number of nodes in the width, height and length directions, respectively. The cavity, f, = f, of cavity with specimen, Q, = unloaded Q of empty cavity,
electric field is in the height direction. Q, = unloaded Q of cavity with specimen.
15.2.4.2 Perturbation test fixture for microstrip substrates: For the TE,,,
ENDPLATE FOR CAVITY WITH I R I S HOLE mode at about 10 GHz the cavity should be 109 x 22.9 x 10.2 mm. A recom-
mended rod-specimen diameter is 1.0mm. For a 2.0mm-diameter rod with K'
of about 8-10 an error of approximately - 2% in K' is reported [26].
Ceramic materials that lend themselves to precision grinding can be used for
CENTER FLANGE FOR OPENING CAVITY
small rod or sphere specimens. Soft microstrip-antenna substrate materials,
SPECIMEN POSITION such as non-woven glass-PTFE or ceramic-PTFE, are difficult to fabricate with
quality. woven-glass fabric-based substrates pose greater difficulties. These
problems are reduced by a larger cavity for lower frequency such as 3 GHz with
recommended rod diameter of 3.2mm.
Vertical bars from most substrate materials can be shear cut or sawn to a
width of 6-12mm at the original thickness. For spheres, bars cut from a
specially made thick laminate are lathe-turned with special fixtures.
A cavity with TE,,,-mode resonance at 3.0GHz is made from extruded
copper waveguide with inside dimensions of 486 x 72 x 34mm and is
WAVEGUIDE-COAXIAL L I N E TRANSITION provided with a 15.5 mm diameter iris opening centred at the ends. Insertion loss
of the empty cavity is about 40dB. With proper polishing of flange surfaces the
Fig. 15.16 Sketch of perturbation cavity Q is about 10000. The flanged joint for loading is within 0.13 mm of the centre
of the length.
Frequencyf, and Q of an odd-node (N-) resonance of the waveguide cavity Spherical specimens are of particular interest, since they can be variously
is measured with and without a dielectric specimen in the cavity. The method oriented with respect to the cavity field in order to observe the degree of
specifies symmetrical positioning of the specimen. The odd-node number is anisotropy of substrate materials. The vertical bar is convenient to prepare but
required to ensure a voltage maximum at the centre where the specimen is has the electric field along the length of the bar, which is not typical of micro-
located. Bothf, and Q decrease when a specimen is present to a degree related strip-antenna applications.
to the specimen shape, size, orientation, R and dissipation factor D. Collected
data on the empty and loaded cavity is reduced to K' and D values by simple 15.2.4.3 Considerations for precision and accuracy: Sources of error in the
formulas as follows: perturbation method can involve measurement errors of frequency or specimen
For the vertical bar size, with predictable consequences. Less obvious are errors related to incorrect-
ly positioning the specimen. The discussion of error is confined to the 3 GHz
cavity.
D = LIP It is important to have a frequency source that is stable and resolvable to less
For the sphere than one part in lo6 since the perturbation method is concerned with small
differences in frequencies.

15.2.4.3.1 Errors in dimension or frequency data: Previously discussed meth-

ods use a resonator based on the material under test. In the perturbation method
 910 Advances in substrate technology Advances in substrate technology 91 1

small changes in frequency readings are much more significant. Small errors in Table 15.12 shows the computed bias of K' and D measurements resulting
specimen dimension o r f , values affect the K:,, of the material significantly. A from singular errors in specimen dimension or frequency.
small error in one of the four frequency readings for bandwidth of the empty Specimen features, such as non-uniform width or thickness of bars or dia-
and loaded cavity affects D,,. The sensitivity to frequency errors is related to meter of rods or spheres, give rise to measurement errors than can be reduced
by averaging multiple readings of each dimension. Surface roughness of a
Table 15.12 Predicted bias of the perturbation method from dimensional specimen can also bias dimension readings away from the effective dimension.
and frequency errors
15.2.4.3.2 Changes in cavity performance: Cavity characteristics drift with
Vertical bar: Area = thickness x width use. This is not a problem if it is gradual and the practice of alternating empty
and loaded cavity readings is followed. Use the average of the empty cavity
Parameter with error Area f, fi readings before and after a loaded reading for calculations of K' and D.
Amount of error 1% 100 kHz l kHz 1 kHz Abrupt changes in cavity performance can arise from irregular laboratory
3.2 x 12.7mm techniques, varying flange-clamp force, changing position of clamps and in-
K' = 3.0 and D = 0.002 clusion of contamination in the cavity or on the flange face. Data showing
Resulting % bias of K' - 0.66 - 0.49 - 0.005 - variable empty-cavity performance is suspect.
Resulting % bias of D - - - - 1.2
15.2.4.3.3 Position of specimen in the cavity: Little comment is made on how
3.2 x 12.7mm
critical this is [26]. Work was done with both the vertical bar and the sphere.
K' = 10 and D = 0.002
The orientation of the width dimension of vertical-bar specimens and mis-
Resulting % bias of K' - 0.89 - 0.59 - 0.005 -
location from the centre widthwise, and to lesser extent lengthwise, were inves-
Resulting % bias of D - - - - 1.5
tigated. The data-collection programme served the dual purpose of showing
reproducibility on repeatedly loading a specimen at the same position, and the
Sphere: Diameter effect of deliberate changes in placement. A single vertical-bar specimen mach-
Parameter with error Dia f, fi ined from a dimensionally stable and uniform thermoset moulding composition,
Amount of error 0.22% 0.022% lOOkHz l kHz 1 kHz 0.1 kHz with good microwave properties but somewhat high dissipation factor, was
used.
11.4 mm diameter The 3 GHz waveguide cavity with flange joint at the centre was used. Read-
K' = 3.0 and D = 0.002 ings alternated between cavity empty and cavity loaded. The vertical bar size
Resulting % bias of K' - 0.73 - - was 34 x 12.7 x 3.2mm and fit snugly between top and bottom walls of the
Resulting % bias of D - - - cavity.
6.35 mm diameter A series of 20 specimen positions were used involving all combinations of five
K' = 3.0 and D = 0.002 degrees of widthwise offset with four orientations and lengthwise offset.
Resulting % bias of K' - 1.31 - - The five widthwise-offset values from the centre of the cavity were:
Resulting % bias of D - - -
0.00, 1.52, 3.05, 4.57 and 6.10mm
6.35 mm diameter
K' = 10 and D = 0.002
Resulting % bias of K' - 4.14 - 0.43 - 19.8 0.0, 2.1, 4.2, 6.3 and 8.4% of the cavity width
~ e s u l t i n e% bias of D - - -
The four orientation variations were:
the stability of the signal, not to the accuracy with which the frequency is (a) Specimen width crosswise and centred over the joint
known. If there is a bias in frequency readings that applies proportionally to all (b) Crosswise and flush with the joint (1.6 mm lengthwise offset)
the readings, the error cancels out in the calculation. However, if frequency is (c) Crosswise but rotated 180" and flush with the joint
unsteady and bias is variable, measurement accuracy will suffer. (d) Specimen width lengthwise in the cavity, centred over the joint
912 Advances in substrate technology Advances in substrate technology 973
A vernier caliper was used to verify each location. Each of the 20 positions was The data for K' values showed a pattern of decreasing value as the widthwise
used for a measurement. This test sequence was repeated six times to provide six offset increased. However, for the four orientations there did not appear to be
replicated data values for each position taken over a period of time. With each any pattern. At each offset the standard deviation among the four sets of means
loading, data were collected for the designed 7-node resonance at 3.0 GHz and for the orientations were remarkably small, more so for 7-node than for 9-node.
also for 9-node resonance at 3.47 GHz. This is summarised in Table 15.14. The effect of widthwise offset is shown
graphically in Fig. 15.17.
Table 15.13 Condensed statistics on reproducibility and sensitivity to posi-
tion and orientation of a vertical bar specimen in the pertur-
bation-cavity method for K' and D
Attribute for Std. devs. of Means of six Std. devs. of six -
- Values wtth 9 node resonance
which mean and six replicate replicate replicate
std. dev. are readings of K' readings of D readings of D -
shown at each position at each position at each position -
Node number of -
resonance 7 9 7 9 7 9 --
A

Mean 0.0028 0.0028 0.0066 0.0067 0.00010 0.00005 -

- Values wtth 7 node resonance
Std. dev. 0.0013 0.0012 0.0002 0.0002 0.00004 0~00002
as % of mean 46 43 3 45 39
Number of data 20 20 20 20 20 20

Table 15.14 Effect of widthwise offset of vertical-bar specimen in pertur- 3.10 I I I

0.00 1.52 3.05 4.57 6.10
bation-cavity: averaging mean K:, values for four variations in OFFSET I N nn UF VERTICAL BAR SPECIMEN FROM CENTER OF 72 m n VIDTH
orientation
Fig. 15.17 Plot showing effect of widthwise offset from centre of waveguide perturbation
Widthwise offset, mm 0.00 1.52 3.05 4.57 6.10 cavity for vertical-bar specimen
With 7-node
resonance at 3.0 GHz
Table 15.15 Comparison off, and I / Q for an empty waveguide cavity with
Mean 3.2503 3.2424 3.2189 3.1779 3.1213
and without a foam polystyrene specimen support
Std. dev. 0.0033 0.0038 0.0036 0.0050 0.0065
as % of mean 0.10 0.12 0.1 l 0.16 0.2 1 Nodes 5 7 9 11
Number of data 4 4 4 4 4 Resonant freq., GHz
With 9-node Empty freq., GHz 2386485 2.995648 3.466802 3.977994
resonance at 3.47 GHz With spacer 2.586387 2.995539 3.46680 3.977860
Mean 3.2690 3.2623 3.2420 3.2068 3.1573
1061Q
Std. dev. 0.0170 0.0180 0.0175 0.0191 0.0191 Empty cavity
as % of mean 052 0.55 0.54 0.59 0.60
Number of data 4 4 4 4 4

Similar widthwise behaviour is seen with spherical specimens. Another series

For all 20 positions there was no significant difference in D values. The of measurements was carried out with spheres of four diameters centred length-
standard deviations for the six K' or D values at each position did not show any wise and widthwise. Vertical offset was varied. Fig. 15.18 illustrates how K',,
trends. Overall averaging of these, shown in Table 15.13, indicates generally increases with increasing vertical offset from the centre of the cavity height.
very satisfactory reproducibility for the perturbation method. A block of very low-density polystyrene foam may be mounted in the cavity
 914 Advances in substrate technology Advances in substrate technology 915
to support the specimen. When this is done the foam should be left in the cavity Adapted for Full details Full details Not needed Not
for the empty readings as well. Empty-cavity resonant frequency a n d Q readings detailed
were found to change very slightly when a foam support is in place, as shown material tvves
by data in Table 15.15. Convenience
Skills for Minimal Photomasking, Minimal Machining
preparation etching,
cutting
SPECIMEN DIAMETER, nn
6.31 Equip. for To cut cards Photomasking, Trim panels Special
8 8 6 ................. preparation etch off foil etching, fixtures
11 39 - . - - . - . -
sawing size for spheres
1385 - - - - - - - - - - -
Potential for Very little High - bad None Moderate
specimen loss mask or etch
Skills for Minimal Loading High-mod., High- moderate
fixture automate automate
loadine
--

Cost
Destruction Enough cards Less than None Small,
of product 51 x 76 mm for 645 mm2, special type
1.27 or 1.5 mm locate
W E
CLW thick conveniently
aC
z I I
a w
ao 0 5 10 15 Labour for Under half Up to 1h Under quarter- Over 1h
OFFSET I N nn FROM CENTER UF VERTICAL 3 4 nn DIMENSION preparation, hour hour or less -
I N 486 x 72 x 3 4 nm WAVEGUIDE CAVITY measurement automation
Fig. 15.1 8 Plots of KLppversus vertical offsetposition in perturbation cavity for four spherical
specimens of differingdiameter Time from About one day Over one day About 1h. Over one day.
selection In-process Machine-shop
to results control help needed
15.2.5 Tabulated evaluation of methods for measuring relative permittivity and Error
dissipation factor Likelihood Unlikely Likely and Unlikely. Moderate
Feature Stripline Microstrip Full-sheet Perturbation specimen critical Rectangular
evaluated resonator resonator resonator cavity preparation

Standardisation Detectability Monitor of Monitor with Monitor with Dimensionally

Routine use. Standard for Spec. 1 user Few cases of Not routine of error fixture reference reference verifiable
K' = 2.15-2.55 K' near 10 user-supplier deterioration specimens specimens
Higher K' Apparatus Fatigue Less than Littlelnone Self-checking
starting deterioration fracture of stripline versus cavity
Reproducibility High degree Fair Excellent Good for pattern card
Stretching of
sphere
resonator
Correlation Good Acceptable Very good No data
among labs. Usefuhess
Correlation Fairly good Good vs. K' sensitivity Limited
Citations MIL-P-13949 Two user 1 or more None with microstrip; exceeds
in specs. (12) and specs. user specs. application vs. process, stripline
users cladding good scale-up
91 6 Advances in substrate technology Advances in substrate technology 977
Orientation Similar Similar Similar Vertical bar 15.3.1 Handling incoming copper-clad laminates
of field no X, Y is poor; Attention to incoming packaging, storage and pre-process handling of clad
VS.antenna fringing Sphere allows laminates for microstrip antennas will pay off in improved manufacturing yield.
control
Sensitivity- Good but Variable Radiation at Separates 15.3.1.1 Packaging: Suppliers of clad laminates for microwave applications
dissipation includes experience edges obscures dielectric generally take special precautions to ensure minimal damage from shipping or
factor conductor Includes losses storage. In addition to a rigid container and compressible spacers in the con-
loss cladding tainer, the individual panels are usually specially protected by either thick
Limitations plastic film with lightly adherent pressure-sensitive adhesive or polyethylene
Range of Limited thickness Limited thickness Wide-range Wide-range bag enclosure. Tarnish inhibiting paper is often used against the copper-foil
thicknesses and K' range 0.63 & 1.27mm; thickness, K' machinable surface.
and types for a fixture K' near 10 Such care is especially needed for PTFE-based substrates which are relatively
Field Limited to Z LImted to Z Limited to Z X, Y, Z for soft.
orientations sphere spec. While shipping containers must be opened on receipt to verify the contents
measurable and detect any damage, it is usually most efficient to take advantage of the
10 GHz/4-node, 3 GHz Limited by Practical existing protection for storage until processing is started.
Frequency
range could use specimen waveguide Most producers of microwave substrates provide individual panel identifica-
2.5, 5.0, 7.5 dimensions sizes, 3-10 GHz tion, and should include a serial number for traceability. The user should
provide practical means for maintaining panel identity through the process to
Questionable Clear result, Positioning Critical that Clear result. the final microstrip antenna unit.
results monitor cover, ground modes be K' separated
fixture contacts unambiguous from magnetic
permeability 15.3.1.2 Storage: Metal cladding of laminated substrates is the feature
having greatest susceptibility to chemical damage in storage. Individual packag-
Substrate Single or 0.6311.27 mrn Wide range Machined ing of panels provides some degree of protection, but the storage area should
thickness stacked to under 2.5 mm sphere is have a room-temperature atmosphere free of high humidity, sulphur oxides and
range 1,2711.57mm to over 3 mm small versus
cavity
other corrosive materials. Small amounts of corrosion can be removed through -
the cleaning steps to be discussed.
Problems of Fringing No account Best when Not applicable Physical-storage hazards to the quality of laminates can be avoided by
thickness correction for fringing thin versus attention to a few rules. These are particularly applicable to soft substrates such
extremes varies error length, width as those based on PTFE.
Material in original containers should be stored flat but not stacked high
enough to crush the containers.
15.3 Processing laminates into antennas Material removed from containers should be kept in the original bags or
protective film to prevent scratches, pits or dents from handling during storage.
The volume of information on processing of printed wiring boards from copper- Panels can be stored on edge to provide easy access to a variety of panels in the
foil-clad laminates is expanding rapidly. Contributors include technical socie- inventory. If they are to be stacked horizontally, the shelf should be flat and
ties, individual workers and producers of equipment and supplies used in this larger than the panels. Panels of different sizes should not be combined in a
growing industry. While producers of microstrip antenna boards can use much single stack. The area should be free of chips and particles that can get between
of this information, they also have special concerns. These include the high cost panels causing pits and dents.
of substrate materials, safety with PTFE, etch-strain relief, machining, bending For PTFE-based laminates, the timeltemperatwe cycle of the manufacturer's
boards, bonding assemblies, plating holes and edges. laminating process leaves aluminum plate or copper-foil cladding in a fully
This section is not intended to be a complete guide to processing printed annealed state that is particularly susceptible to scratches and dents. These
wiring boards, but will address selected topics of value to producers of micro- materials require greater care in storage and subsequent handling than other
strip antennas. substrate types.
 978 Advances in substrate technology
15.3.2 Handling prior to processing
Quality problems in processing can be reduced by attention to some details in
handling before the processing begins.
15.3.2.1 Avoid scratches, dents andfinger prints: The initial process steps of
shearing, sawing or blanking to size and of forming tooling holes should be done
with surface protection present. This can be the plastic film or protective paper
used in packaging. Avoid sliding or stacking unprotected surfaces against each
other.
Use gloves of knit nylon or other non-absorbent material for handling
unprotected panels to avoid finger prints, which contain acidic skin oils that
corrode copper surfaces. If a dilute-acid bright dip is used to remove visible
corrosion, the corrosive oil remaining on the copper surface causes the finger
print to reappear hours or days later.
Finger prints can be removed by a bright dip in dilute hydrochloric acid with
rinse and drying, followed by degreasing in ketone solvent or chlorinated
solvent vapours. A second bright dip, water rinse and drying is then used.
15.3.2.2 Remove any adhesive residue: If panels are protected by the lightly
adherent thick plastic film, it will need to be peeled off and the surface cleaned
just before etch masking is applied. Residual adhesive on the copper surface may
be removed by wiping the surface with a soft, lint-free cloth soaked in an alcohol
solvent. Isopropanol at 70-100% concentration is preferred for its low toxicity
and flammability. Panels that have had long-term or high-temperature storage
may require a soak in isopropanol and scrubbing with a soft bristled brush to
remove all the adhesive residue.
15.3.2.3 Ensure adhesion of the photomask: Poor adhesion of the photomask
through the etching process can occur with no apparent cause. The following
procedure has been found effective for avoiding this costly problem:
(a) Immerse the board for 5 min in Neutraclean 68 solution*
(b) Hold for at least 1 min in each of two successive water-rinse tanks
(c) Immerse for 1 min in a 10% by volume solution of sulphuric acid
( d ) Hold for at least 1 min in each of two successive water-rinse tanks
(e) Drain the board and allow it to dry in air
(f) If dry film is to be applied, preheat the board for at least one hour in an
oven set for 125C. Allow enough time for the rolls on the laminator to stabilise
at the required laminating temperature before using it to bond the film to the
board.
(g) If wet film is to be applied, use a constant speed for withdrawal of the board
from the dip tank
* Neutraclean 68 is an aqueous formulation of sodium bisulphite at neutral pH with proprietary
additives for removing oxides and contaminants from the copper surface. The supplier, Shipley
Company, 2300 Washington Street, Newton, MA 02162, USA, phone 6171969-5500, provides
instructins for preparation of the solution.
Advances in substrate technology 979
15.3.3 Safety considerations for PTFE-based substrates
PTFE is exceptional among plastics materials for its chemical inertness, in-
solubility, non-flammability, non-toxicity and long-term resistance to thermal
degradation. It is such a stable polymer that the tendency may be to assume that
no safety precautions at all are needed.
At temperatures of 380C and above, thermal decomposition of PTFE be-
comes measurable. The decomposition and toxicity of the decomposition
products have been studied and reported in detail [27-471. Good summary
bulletins are available from the polymer producers [48, 491. There are no
reported incidents of which we are aware of permanent human injury from
fumes in normal processing. Particulate decomposition products, if inhaled, are
believed to be the cause of a syndrome of influenza-like features described as
'polymer fume fever'. Symptoms arise after a latent period of a few hours. The
symptoms subside within 24-48 h with no apparent known after effects.
PTFE has a limiting oxygen index for flammability of 95% and resists
auto-ignition to very high temperatures: 575OC in air, 512C in oxygen. The
toxicity of products from very high-temperature decomposition are more dan-
gerous. The fumes and smoke from a fire containing PTFE can be hig~.,, toxic
and corrosive.
Hazards peculiar to PTFE-based substrate materials can be encountered in
machining processes and in bonding operations at elevated temperature. They
are avoided entirely by simple precautions. While PTFE is probably the most
inert of all available polymeric materials, it is capable of producing toxic fumes
at very high temperatures.
15.3.3.1 Machining hazard: Chips and particles from machining can take an
electrostatic charge and be transferred by hands or clothing onto cigarettes and
tobacco. Particles in tobacco can encounter temperatures in excess of 1600C,
where they decompose to impart 'fume fever' symptoms to the smoker. Smoking
should be prohibited in machining areas and washing of hands and changing of
clothing are advised for smokers. Non-smokers do not have this hazard.
Dispose of scrap by landfill, putting scrap into bags or boxes clearly marked
Do not incinerate.
If high temperatures are generated by machining conditions, forced venti-
lation is advisable. Enough heating to generate fumes or smoke indicates poor
machining practice, leading to sub-standard workpiece quality.
15.3.3.2 Fumes from bonding boards: While temperatures normally used for
bonding multilayer boards are below 300C, there is the possibility of a control
malfunction or set temperature error. It is good practice to ventilate the press
area as a precaution in anticipation of such an event.
15.3.4 Reducing the effects of etch strain relief
The lamination of copper-foil-clad panels involves clamping substrate material
920 Advances in substrate technology
and metal foil between steel plates, followed by heating the assembly to effect
either a thermoset reaction, in the case of thermosetting-resin systems such as
polyimides, or a fusing together of components, in the case of thermoplastic
polymer systems such as those based on PTFE. In either case the clamped
assembly is cooled to room temperature before the clamping force is removed
and the clad panels are removed from the press package.
During cooling the difference in coefficient of thermal expansion between the
substrate and the metal-foil cladding gives rise to internal strain. In addition,
there is internal strain associated with the change of state of the substrate
material. Thermoset materials tend to shrink when converted from their fusible
initial stage to the final fully reacted state. Thermoplastics may undergo a
change in degree of crystallinity with associated specific volume change.
Thus metal and substrate are strained to dimensions different from what they
would be if they were not attached. When some or all of the metal cladding is
removed by etching for production of a circuit board, some or all of the strain
on the substrate is relieved by a change in dimensions.
This characteristic strain-relief behaviour is common to some degree in all
clad laminates. PTFE substrates preferred for microstrip antenna boards are
laminated at temperatures above the 327OC crystalline melt point of the poly-
mer. This relatively large temperature excursion makes etch strain relief of
important concern to designs with tight tolerances on dimensions between
features.
The process sequence described here is specifically intended for overcoming
strain-relief characteristics of glass-PTFE substrates, though the principles may
apply to other materials.
15.3.4.1 Metal-foil removal without fiwing feature position: Photomask and
etch to a preliminary pattern with oversize features designed to remove at least
90% of the metal foil to be finally removed without precisely defining the
location of features. Provide narrow X- and Y-axis centre lines in this prelim-
inary pattern for later alignment.
When the copper has been removed, stresses related to the internal strains
become unbalanced. If the substrate material were perfectly elastic, the substrate
would instantly change dimension to relieve the internal-strain imbalance.
However, the visco-elastic nature of the PTFE makes this response time-depen-
dent. The time required for strain relief is reduced at elevated temperature.
The mask material is stripped from the board so that subsequent heating does
not make it difficult to remove later.
15.3.4.2 Complete strain relief before,fixingfeature positions: Rate of strain
relief varies approximately in proportion to the remaining unbalanced internal
stress, and is a logarithmic function. Suppose, for lack of precise data, it takes
48 h at room temperature for one half of the strain to be relieved. Allow 96 h for
three-quarters relief, 144 h for seven-eighths relief, etc. This hypothetical 48 h
Advances in substrate technology 921
time constant can be regarded as the half life for residual strain. At some later
time, further strain relief is no longer practically detectable.
About half of the PTFE in the substrate exists as the amorphous phase with
visco-elastic control of strain relief. The transition temperature from glassy to
rubbery state, T,, of the amorphous phase takes place at about 130C.
Time for strain relief decreases gradually with increasing temperature up to
T,, where a greater decrease occurs. Increased temperature does little to change
the amount of strain to be relieved, but it does much to reduce the time needed
for most of it to occur. Thus oven-heating boards well above T,, e.g., 150 or
170C, will reduce the time required to 16h or less. Higher temperatures will
cause excessive oxidation of the copper foil.
The time needed at a particular temperature for strain relief of a given
laminate construction and antenna-pattern may be verified by measuring the
change between marked locations on a test board versus time.
15.3.4.3 Process to fix feature positions: The X, Y centre lines from the
preliminary pattern are used for optical alignment for drilling or punching the
registration holes used for the rest of the process. Very little dimensional change
is to be expected from here on.
Various sequences of steps can be devised to etch the conductor pattern to its
final dimensions and to provide other required features such as holes or routed
edges. A feasible outline of steps is given as an example. This starts with a board
pre-etched as in Section 15.3.4.1 and strain relieved as in Section 15.3.4.2.
(a) Drill holes to be plated through.
(b) Treat PTFE surfaces with a sodium etch solution for adhesion of electroless
plating.
(c) Electroless plate all surfaces. This thin deposit is a base for plating in holes,
and it electrically connects isolated elements.
( d ) Apply a second photomask and develop to expose holes and other
regions to be plated. Then electroplate these areas to a desired thickness. Strip
the mask.
(e) Apply a third photomask over the final conductor pattern shape to be .
retained. If a dry film mask is used it will need to bridge over plated-through
holes. If wet film is applied by dipping, the viscosity must be low enough to
ensure that holes are coated and free of bubbles. Etch away exposed metal,
which will include electroless deposit over previously conductor-free areas, as
well as extra conductor area left from the first etching operation.
An alternative to steps ( d ) and (e) that avoids a third photomask and the
concern for protecting plated-through holes is as follows:
( d ) Apply a second photomask exposing the final conductor pattern including
holes to be plated through. Electroplate the exposed areas with copper to the
desired thickness. Electroplate a thin layer of tin. Strip the mask.
(e) Use an etchant specific to copper to remove all copper from areas not
 922 Advances in substrate technology
protected by the tin plate. Then strip the tin-plate mask using a commercially
available tin-stripping solution.
15.3.5 Machining of PTFE-based boards
While typical microstrip antennas require little machining, some discussion of
the topic with respect to substrates based on PTFE seems worthwhile. Holes are
often employed for feeds to microstrip radiation elements. Often microstrip
antenna boards are trimmed to size by routing.
15.3.5.1 Generalprinciples: Substrates based on PTFE for microstrip anten-
nas usually contain one of the following in a soft PTFE matrix:
For non-woven glass-PTFE: Discontinuous glass fibres of wide diameter distri-
bution, averaging about 1 pm, uniformly distributed in the matrix and individu-
ally oriented in all directions in the X, Y plane of the substrate.
For woven glass-PTFE: A fabric of glass-fibre strands, where each strand
consists of 30 or more continuous glass fibres of narrow fibre-diameter distribu-
tion, typically about IOpm.
For ceramic-PTFE: Particulate ceramic filler with hardness similar to chrome,
uniformly distributed in the matrix and occupying a significant fraction of the
volume.
These composites are all capable of rapidly destroying the cutting edge of a steel
tool even though the matrix is relatively soft. Carbide cutting tools are recom-
mended, and in the case of ceramic filler are mandatory.
Woven glass tends to produce rough edges when routed or drilled. Small-
diameter drills tend to be deflected, reducing hole-location accuracy.
Cutting edges of tools must be sharp as viewed under a 30-60 power stereo-
microscope. Incidents of rejection as high as 25% in new tool lots have been
mentioned. As a radius forms from wear, the cutting mechanism of fracture
propagation ahead of the edge degrades to a sliding-tearing action. This rapidly
increases edge wear and generates frictional heat which softens the matrix and
further reduces the cutting quality.
Solid-carbide drills and routers are regularly used in large quantities by the
printed-wiring-board industry, and are available from various suppliers.
15.3.5.2 Drilling burr-free holes: PTFE tends to form tough burrs at mach-
ined edges that are hard to remove once they are formed. A few precautions
assure smooth, burr-free holes:
(a) Use entry and back-up boards. Hard phenolic-paper laminates have been
found effective.
(b) If holes are to be located in areas free of copper-foil cladding, do the drilling
prior to etching away the copper.
Advances in substrate technology 923
(c) Maintain adequate clamping force on the stack during drilling to prevent
burr formation.
( d ) The exact drill-bit diameter needed for a specified hole size is based on trials.
The low modulus of the substrate matrix can result in the hole diameter differing
from the bit, depending on drilling parameters such as diameter, board type,
drill speed and feed rate.
(e) Limit the number of hits per drill bit to a number found by trial to produce
good-quality holes.
15.3.5.3 Avoiding 'smear' on hole walls: Smear on drilled hole walls is well
known with epoxy-glass laminates, where over-heated tool bits can partially
decompose the epoxy resin and redeposit it on the hole wall. Smear interferes
with adhesion of the copper deposit to the edge of the substrate and the clad foil
in plated-through holes.
With PTFE-based laminates smear arises somewhat differently. Excessive
heat from a worn tool edge, high drill speed or a plugged flute softens polymer
particles so that they can be shear-deformed to a poorly adherent very thin
transfer film.
The following principles minimise concern about smear.
(a) Do not exceed a tool surface speed of 0.76m/s (150ft/min), significantly
slower than that recommended for most substrates.
(b) Use drill bits with a relatively shallow included lip angle and with a relief
along the flutes. An included angle of 130" and a 127 pm (0.005 in) reduction in
flute diameter after the first 0.64 mm (0.025 in) length have been found to work
well with carbide drills for 0.64mm holes.
(c) Use a feed rate for a 50 pm (0.002 in) chip load.
15.3.5.4 Routing edges of PTFE-based substrates: Non-woven glass-PTFE
and ceramic-PTFE substrates can be shear cut as in punch and die tooling for
remarkably smooth edge finish. For woven-glass-PTFE or for quantities too
small to justify special tooling, edges are cut with a routing bit.
The following suggestions should be considered in setting up the operation:
(a) Select a carbide router bit with single cutting edge that is either straight and
parallel to the axis, or a reverse helix so that it tends to push the work against
the back-up board.
(b) Rout with clad foil forming the edge if possible.
(c) Use a rough cut followed by a fine cut to dimension. Tool-axis movement
relative to the work follows the cutting edge direction. i.e., for clockwise tool
rotation inside edges are cut moving clockwise, while outside edges are cut
moving counter-clockwise.
( d ) Keep tool surface speed down to 0.76 m/s (150 ft/min) as for drilling.
 924 Advances in substrate technology Advances in substrate technology 925
15.3.6 Bending etched antenna boards
A significant number of microstrip antennas based on non-woven glass-PTFE 15.3.6.2 Heat forming on a mandrel: A board clamped around a mandrel
are formed to a cylindrical shape for missile telemetry applications. The non- somewhat smaller than the desired diameter of curvature is held at elevated
woven glass-PTFE substrate is particularly suited for this application. It can be temperature to allow strain relief to occur. This process is slower than roll
bent without measurable damage to the microstructure or the mechanical and forming, but, as should be obvious from the discussion in Section 15.3.6.1, it
electrical properties. does not require as great a degree of over-bending. For a given substrate
Three methods are being used to bend antenna boards to the desired radius: thickness, smaller-diameter bends can be more practically obtained by heat
use of forming rolls, heating a board clamped on a mandrel and bonding thinner forming than by roll forming. The two processes differ in time and temperature
boards together around a mandrel. of conformation imposed on the material.
A laboratory experiment on heat forming was performed with a 114mm
15.3.6.1 Forming rolls: Forming rolls are commonly used in sheet-metal ( 4 5 in)-diameter mandrel and 1.57 mm (0.062 in)-thick non-woven glass-PTFE
work for forming bends with a radius much larger than the metal thickness. The substrate, RT/duroid 5870 microwave-circuit-board material from Rogers Cor-
three-roll former has two metal rollers working against a third base roller. The poration. The substrate was clad on both sides with copper foil of 34pm
system forms two nips along the circumference of the base roller. When material thickness (1 oz/ft2 weight). The thickness/radius ratio selected was just less than
is fed into one nip so that it is next caught in the second, it is forced to conform sufficient to cause immediate board damage when a specimen was clamped to
to the radius of the base roller. The nips can be opened to increase the radius a mandrel.
of curvature producing less conformance. Specimens were etched to leave a narrow conductor trace runing circum-
A second type of forming-roll apparatus consists of a metal base roll working ferentially on one side, leaving the other side fully clad. Experiments were run
against a rubber roll. The rolls are brought together so that the rubber roll with specimens having the trace on the concave and on the convex side. The
I specimen clamped on the mandrel was held at elevated temperature, cooled,
deforms along the circumference of the base roll. Material fed into this nip is
conformed to the radius of the base roll to an extent which is controlled by how then released from the mandrel and allowed to stand in standard laboratory
tightly the rolls are pressed together. conditions for 16 or more hours before its curvature was measured.
In either type of apparatus the principle is the same: the material is forced to
conform to a radius. If this radius causes the material near its surfaces to exceed Table 15.16 Summary of mandrel heat- forming investigation
the elastic limit of the sheet material, permanent deformation results in a smooth Temperature, OC 23 177 260 302
curved shape. When the conforming force is removed, a certain amount of ~ i m dh, 24 1 9 1 9 1 9
elastic spring back occurs. Thus, to bend a given radius, the material is overbent % Retention of curvature
to a smaller radius, held there for perhaps 0.25 s and then released. The smaller Ground-plane concave side 43 77 76 86 85 88 96
radius could be as small as one-third or one-quarter of the desired final radius. Ground-vlane convex side 38 73 76 86 87 94 99
With metals the permanent deformation and spring back are rapid. Defects in copper* 0 1 1 3 2 5 4
When the method is applied to a copper-foil-clad PTFE substrate with one * 0 to 5 represent subjective ratings of none, slight, few, much and severe
side etched to form an antenna pattern, care is needed to prevent deformation
damage in the roll-forming operation. The copper on the inner and outer
surfaces is forced to accommodate to the opposed changes in length of these Curvature retention was calculated as the ratio of the mandrel radius to the
surfaces. It is either elastically and permanently deformed by the imposed final specimen radius of curvature. Copper-foil damage was also noted. Table
change or it fractures and buckles. 15.16 is an abbreviated summary of the work.
When typical sheet metal is deformed, the deformation almost instantly For less severe bends, no defects in the copper cladding are expected at
resolves into permanent and elastic portions. The elastic portion accounts for forming temperatures up to 170C for periods of nine hours or more. With
spring back. For a visco-elastic material as discussed in Section 15.3.4.2, the protection from oxygen, higher temperatures could probably be used to im-
small instantaneous ratio of permanent to elastic portions of the deformation prove curvature retention without copper-foil damage, but this is not likely to
will increase with time. The instantaneous ratio and the rate of increase are both be worth the effort since mandrel size can be adjusted to get a desired curvature
greater at higher temperatures. Thus the amount of overbending needed in a at lower temperatures.
roll-forming operation may be reduced by pre-heating the board and by lower-
ing the speed of the forming rolls. Heating above the 130C amorphous-phase 15.3.6.3 Bonding layers on a mandrel: The process includes heat forming on
glass transition temperature of the PTFE is advised. a mandrel since heat is used to activate the bond.
 926 Advances in substrate technology
Advances in substrate technology 927
A microstrip antenna is assembled from two or more boards that are inter-
leaved with bonding film and clamped around a mandrel for a thermal cycle to perature capability. The bonding system must be serviceable at least to the
activate the bonding film. A direct bond procedure omitting bonding film has maximum application or storage temperature expected.
also been demonstrated and will be discussed in a later Section. Bonding Bonding material should match the substrate's relative permittivity. Low
processes follow the principles outlined for flat antennas in Section 15.3.7. dissipation factor is usually important as well. Compatibility of the temperature
A higher thickness/radius ratio can be achieved for the assembly than would needed to activate the bond with the application environment must be con-
be practical with a single board. Curvature retention is very good since the sidered. A comparison of advantages and disadvantages of several types of
degree of strain applied to the individual layers is low. Any spring back arising bonding systems may help to put them in perspective. The bonding systems
from individual layers is offset by the fact that the difference in length between discussed here fall into three categories: thermoplastic films, thermoset systems
the inner and outer layers is retained by the bond line. and direct bonding.
This process is often used for microstrip antennas requiring a protective Thermoplastic films of practical value for bonding include, among others,
radome. One board carries the ground plane and antenna pattern. A second low-density polyethylene (LDPE), irradiated polyolefin (IPO), chloro-fluoro-
board is etched free of copper foil and bonded to the first in the curved shape. copolymer (CFP), and fluoropolymers such as poly(tetrafluoroethene-co-hexa-
Where very high thicknesslradius ratios of curved antennas are needed, fluoropropene) (FEP). These bonding films will fail in service if the melt point
several boards can be combined in an assembly so that the innermost board is exceeded.
includes the ground plane and the outermost board, not counting an optional Low dissipation factor and permittivity values nearly matching the glass-
radome cover, carries the anntena pattern. PTFE substrates make them useful materials for bonding. If ultrasonic scan
testing of an assembly indicates bond defects, thermoplastic film offers the
15.3.7 Bonded-board assemblies advantage of repairability; i.e., a second bonding cycle can be effectively run to
The use of bonded assemblies for stripline applications has almost completely pass inspection before processing the board further.
displaced clamped assemblies for the obvious advantages of size, weight, imper- Generally, for thermoplastic films a dwell time at bonding temperature of
viousness and long-term performance stability. The topic is discussed here 15-20 min is sufficient. The clamp stress needed for a suitable bond depends on
because it affords the only practical means for environmental protection in the board and copper pattern thickness, the fraction of area occupied by
many microstrip antenna applications. conductor pattern and the rigidity of the boards being bonded. Adequate clamp
Bonded assemblies are also finding use as the means of integrating the stress levels can range from 170 to 1380 MPa, and need to be determined by trial
antenna into a compact system, including devices such as feed networks built in for a given part. Characteristics of the films mentioned are summarised in Table
stripline. 15.17.
Properly built bonded-circuit-board assemblies resist moisture or other
contaminants that degrade long-term performance. Table 15.17 Thermoplastic bonding films
Film type LDPE IPO* CFP FEP'
15.3.7.1 Selection of a bonding system: A variety of possibilities exist for Melt point, "C 110 - 200 260
producing a bonded assembly. With PTFE-based substrates, boards can be Relative permittivity 2.3 2.3 2.35 2.1
clamped together and direct bonded by heating the assembly above the crystal- Dissipation factor 0.0005 1 04)02 0.0028 0.0003
line melt point of the PTFE, so that the substrate material itself acts as the Bond temperature, OC 120 121-149 220 280
adhesive. For most substrates, the boards to be bonded can be interleaved with
bonding film or 'prepreg' and heated while clamped to activate the bonding * Irradiated polyolefin co-polymer from MPC, Inc., Lowell, MA 01854, USA
't Teflon FEP fluoropolymer film from the Du Pont Co., Wilmington, DE 19898, USA
action. The options for clamping and heating include the use of a platen press
or the use of vacuum-bag moulding techniques in a pressure vessel.
In selecting the bonding system many factors need consideration. If the Thermoset resin systems may be used where higher dissipation factors can be
boards to be bonded already have mounted components or soldered joints, low tolerated. Microstrip antenna assemblies have been bonded with two-part
bonding temperatures must be used. If substrates to be bonded are based on liquid-epoxy adhesive systems at modest clamp stress and with unsupported
PTFE then provision is usually made for the surface treatment to enhance the films of thermosettable resin. In some cases, a prepreg of woven glass fabric
bond. The bonding options may be limited by the availability of a suitable saturated with an epoxy resin or a polyimide precursor resin system could be
platen press or heated pressure vessel for vacuum-bag moulding, and its tem- used, especially for attaching the antenna to underlying circuit boards.
Thermosets have the advantage of developing heat resistance to above the
 928 Advances in substrate technology Advances in substrate technology 929
bond temperature used. In contrast, thermoplastics are bonded at a temperature promoting treatment. It is important to preserve this finish for good bond
well above the melt-point limitation to heat resistance in service. In general, the results. Avoid any mechanical scrubbing or rubbing. If possible, boards should
parts to be bonded are interleaved with the thermoset system, clamped, then be stored in racks and used in the bonding operation as soon as possible. The
heated to a set point and held long enough to accomplish the cross-linking precautions are not needed when direct bonding is used.
reaction in the resin system. The clamped assembly is then cooled. Oven baking For PTFE boards, surface treatment is performed to enhance the bondability.
after assembly may be desirable to further advance the degree of cross-linking. Some users are treating the boards with commerically available sodium etch
A few systems are summarised in Table 15.18. solutions. The alternative is the use of plasma etching with oxygen and fluoro-
carbon gas. Plasma etching is becoming a popular process for de-smearing
Table 15.18 Thermoset bonding materials drilled holes in mutlilayer circuit boards of woven glass-epoxy resin laminates.
Form Film Film Prepreg Prepreg Several companies are supplying plasma de-smearing equipment for the printed-
Resin system Epoxy* Not givent ~ ~ o x y : Polyimides wiring-board industry. Either process should leave the surface to be bonded
Relative permittivity 3.76 1 MHz 4.0 1 MHz 4.35 4.25 water wettable; i.e., water poured on the surface and drained off will form a film
Dissipation factor 0.064 0,027 0.027 0.014 rather than draw up into droplets. Neither treatment is needed for PTFE
Temperature, OC 171 171-191 171-177 215-221 substrates if direct bonding is to be used.
Pressure, MPa 0.7-1.4 1.4-2.8 1.7-2.8 2.8-4.2 A word of caution concerning plasma etching for bonding is in order. While
Time. min 45-60 30-45 45 240 plasma etching has been successful in plated-through holes where wettability
and not bond strength is the major issue, it may be less successful for bonding.
* Poly-Cast EP from Fortin Laminating Corp., San Fernando, CA 91340, USA There is a suspicion that some plasma-processing conditions not only produce
' Pyralux WA/A from the Du Pont Co., Wilmington, DE 19898, USA
:Information and prepregs from Norplex Div. of Allied-Signal Inc., La Crosse, WI 54602-1448, wettable polar groups on the I T F E surface but also attack carbon-carbon
USA bonds. This chain-scission effect forms a weak surface layer of reduced-mole-
cular-weight polymer, yielding a poor bond.
Direct bonding, also referred to as fusion bonding, is feasible with substrates Surfaces of PTFE boards etched free of copper and awaiting bonding should
based on PTFE. It takes advantage of the very high melt viscosity characteristic be protected from ultra-violet-light exposure during storage. Even fluorescent
of PTFE; i.e., when PTFE is heated above its 327OC crystalline melt point, it overhead lighting is sufficient to render the treatment non-wettable by water
continues to behave as a solid and will resist flow at low stress values. Contact- after a few days. Bonding should be carried out soon after treatment-8 h for
ing surfaces of PTFE in this melt state will intermingle at a molecular scale to plasma-etched surfaces and within 48 h for sodium-etched ones.
form a fused joint. A melted PTFE surface contacting a clean copper surface will
wet the surface to form a bond. 15.3.7.3 Bonding in a heatedplaten press: Platen-press equipment is usually
Direct bonding affords several advantages. The low dissipation factor or the more commonly available and a less costly investment than equipment for
dielectric constant of the substrate are not compromised by a layer of bonding vacuum-bag processing in a pressure vessel. The boards to be bonded are
material in the assembly. The PTFE surfaces to be joined do not need any positioned in a tooling-plate fixture with alignment pins and then subjected to
special treatment to ensure adhesion. The heat and environmental resistance of clamping force during the heating-cooling cycle in the press. Aluminum tooling
the assembly is fully equivalent to the substrate material. Direct-bonded plates are usually preferred for the fixture because the high thermal conductivity
assemblies of PTFE can be functional above 3 15C. greatly improves temperature uniformity across the bonding area, and makes up
Direct bonding has the disadvantage of requiring equipment operable at for some degree of non-uniformity in heating of the plates as well as unevenness
much higher temperature, up to 388OC. The assembly must be clamped at a well of heat transfer from platens to fixture.
controlled low-level stress in a way that will not increase stress from thermal The work is constrained between the planar surfaces of the fixture, so that the
expansion. Air must be excluded from the assembly during the temperature adhesive layer is required to flow in the plane of the boards away from areas
cycle to avoid oxygen-accelerated thermal degradation of PTFE. with conductor pattern to areas without conductor pattern where there is more
space between boards to be filled. If an inadequate quantity of bonding material
15.3.7.2 Preparation for bonding: Board surfaces to be bonded should be free is used there will be void areas. When the thin layer of bonding material melts
of contamination, including complete removal of the etching mask. it becomes a lubricating layer capable of allowing slippage between boards if the
For PTFE-based substrates the area exposed by etching away the copper foil clamp pressure is uneven.
has a textured surface formed by the nodular finish of the copper foil's adhesion In the case of direct bonding where no bonding layer is used, the board
 930 Advances in substrate technology
material itself will be required to move. Since the board thickness relative to the
total thickness difference from the presence of conductor patterns is large, this
movement is very slight compared to that required of a bonding film. The
method avoids any slippage between the layers to be bonded.
Use a press with flat parallel platen surfaces that has adequate heating and
cooling capability. For higher temperatures the edges of the press platens should
be provided with high-temperature thermal insulation to prevent uneven platen
heating due to excessive radiation losses from the edges.
When boards for bonding are laid up with bonding material in the tooling-
plate fixture, it is good practice to provide a layer of thin aluminum foil as'a
disposable separator sheet between the fixture surfaces and the boards.
In the case of direct bonding, air must be excluded from the work. This is done
by enveloping the work in 50-100pm-thick aluminum foil. The edges of the foil
envelope are closed by folding them over twice and rolling the folds into tight
creases. A nitrogen purge in the envelope is desirable. A whitened appearance
of the PTFE substrate or corrosion of the copper surface are evidence of failure
to protect from air.
A series of board assemblies may be bonded in the same lay up by providing
interleaving plates between assemblies. Too many assemblies in a stack may
impair temperature uniformity, so that a compromise is needed for the optimum
productivity.
It is critical to provide for frictionless relief of expansion of the work in the
press during heating to avoid damage from build up of clamp stress.
15.3.7.4 Bonding with vacuum-bag moulding equipment: The boards and
bonding material are assembled for a single assembly on a single-plate tooling
fixture placed on a flat metal plate of larger area. The plate is fitted with a
vacuum line. Usually grooves are provided on the plate in the fixture area to aid
evacuation. A barrier film is draped over the work and attached to the plate with
a high-temperature pressure-sensitive adhesive strip.
Barrier films and adhesive strip are commercially available as high-
temperature bag-moulding supplies. The major application is in construction of
composite airframe parts.
When a vacuum is drawn, the barrier film will draw down to the plate. This
assembly, including vacuum connection, is enclosed in a pressure vessel
provided with a heater and circulation system. Inert gas such as nitrogen or
carbon dioxide is used to provide both clamping stress and heat transfer to, and
later from, the plate and bonded boards.
With suitable fixtures, bonded boards of non-woven glass-PTFE can be
prepared with a curved shape.
15.3.7.5 Bonding of curved shapes clamped on a mandrel: The reader is refer-
red to the discussion of bending in Section 15.3.6.3. Microstrip antenna boards
based on non-woven glass-PTFE substrate are combined with a cover board or
Advances in substrate technology 931
I
I radome of the same substrate material to form antennas conforming to the
I
1 fuselage of a missile or piloted craft. Many of these units serve for telemetry on
missiles.
1 A preferred method uses a cylindrical mandrel that is hollow in order to keep
I the heat capacity down to reduce the time and energy needed for the heating and
cooling cycle. Aluminum is a desirable mandrel material. Its low density facili-
I
tates handling and the high thermal conductivity tends to ensure uniform
I
heating. The higher thermal coefficient of expansion of aluminum compared
with the clamping hardware ensures a clamping force during the heating of a
I
bonding unit.
The microstrip antenna pattern is masked and etched on the board while it
i
I is still flat. In some cases the board layers to be bonded are cold-formed to the
curved shape as discussed in Section 15.3.6.1 in order to ease layup on the
I
mandrel. The layers and interleaving bonding material are positioned around
the mandrel. Usually a disposable thin sheet of aluminum foil is placed between
~ the mandrel and the first layer to ensure release after the bonding cycle.
Any of the bonding materials mentioned in Section 15.3.7.1 may be used. In
I at least one instance, a two-part heat-resistant epoxy-resin system was mixed
I
and applied to the first board with a hand-held spreading tool. The excess epoxy
resin was allowed to flow out of the edges when the assembly was clamped and
heated.
Over the outer board an outside release foil is applied. Over this is placed a
curved sheet of stiff sheet metal, such as a 0.5 mm (0.020 in) thick stainless-steel
strap, with a screw mechanism for applying tension circumferentially. The stiff
sheet metal distributes the clamping force evenly to the work.
In another arrangement a series of screw and nut connections between ends
of the wide strap provide circumferential tension on the strap. This tension,
provided either by a series of narrow stainless-steel straps or by take up on the
wide strap itself, resolves itself into a stress normal to the mandrel circumferen-
tial surface. If the screw and nut connections are replaced by an extended
threaded rod and a steel compression spring mounted on the rod, tension
applied by tightening the nut will also compress the spring and provide a
mechanism for adjusting to thermal-expansion differences.
The assembly of curved boards and clamp straps around the mandrel is
placed in a forced-circulation air oven. The oven is programmed through a
heating, dwell and cooling cycle to activate the bonding layer.
A technique has been demonstrated for directly bonding non-woven glass-
PTFE substrate boards into a curved microstrip antenna-radome assembly
capable of remaining functional as a telemetry antenna for a short time at or
above the PTFE crystalline melt point of 327C under frictional heating of
hyper-velocity flight. The apparatus and procedure are described here as an
example of a feasible process.
The material used was RTtduroid 5870 non-woven glass-PTFE microwave
laminate from Rogers Corporation. A microstrip patch array with a distribu-
 934 Advances in substrate technology
The last assembly was retained and sectioned for analysis of thickness unifor-
mity, substrate-specific-gravity uniformity and quality of bond of the cover
board to the top surface of the copper-foil patch areas. Seven locations evenly
spaced around the curved assembly were examined at seven positions across the
width for a 49-point analysis.
No failures of the bond were found between the cover board and base or
between the cover board and copper-foil surfaces. Thickness uniformity over
the points measured had a standard deviation of 1.3% of the mean. Specific-
gravity standard deviation was 1.0% of the mean.
The oven used lacked adequate radiation shielding between the electric-heater
elements and the mandrel assembly, which probably created considerable tem-
perature non-uniformities. The rolled-steel mandrel was not finished to a true
cylinder. A thicker and somewhat narrower steel strap would have better
resisted stretching to provide more even stress across the width of the boards.
It is believed that improvements of these features would greatly improve unifor-
mity of thickness and specific gravity, although the initial efforts yielded quite
good results.
15.3.8 Plating through holes in microstrip antenna boards
Very often microstrip antenna designs require an array of microstrip radiating
elements to be connected electrically to feed lines at a lower level, e.g., a stripline
distribution network on a level below the ground plane. A plated-through hole
(PTH) is usually the preferred way to accomplish this in a production run.
Plated edges may also be used in some cases. If done properly, this provides
consistent performance with minimal labour.
I
15.3.8.1 Quality of edges and hole walls for plating: Attention is needed for
several types of hole-wall and edge defects that contribute to faulty PTHs and
edges. Some of these can be avoided by precautions in drilling and other
machining steps discussed in Section 15.3.5.
Smear, a non-adherent layer deposited during drilling, can interfere with the
attachment of plated copper to the edges of clad copper, a defect that acts as a
stress riser to induce metal failure with thermal shock or exposure to stress.
Protrusions from the edge, such as burrs, can have a similar effect.
A less obvious defect causing failure in thermal stress and shock tests is a
stepped hole or stepped edge. The machined edge of the clad copper does not
align with the machined substrate edge. Heating and excess sideways force from
a cutting tool deflects the substrate, which recovers after the tool is removed.
When plating is deposited over a stepped edge it tends to form a smooth radius.
The thickness of metal drops to a minimum at the protruding corner of sub-
strate. This thinned region is a stress riser during thermal shock or stress, which
leads to a fracture.
A similar effect results from a hole or edge where the thickness of the clad
copper is reduced at the edge. In a microsection it appears to taper to a knife
Advances in substrate technology 935
edge. Such a defect can arise when abrasive scrubbing or other metal-removal
methods are used on a drilled board, presumably to remove metal burrs.
Substrates based on woven-glass fabric with concentrated fibre bundles
produce edges and hole walls that characteristically have fibre bundles extend-
ing out of the matrix. The remedy for this has been to etch the extended fibres
back with hydrofluoric acid. Such a treatment corrects the rough surface, but
tends to penetrate into the exposed fibre-bundle ends, making them vulnerable
to absorption of corrosive reagents from later board processing.
15.3.8.2 Wettability of surfaces to be plated: If copper deposition is to be
made on hole walls and edges of a board by the usual wet-processing techniques,
it is essential that the surfaces be wettable. For most substrates other than PTFE
this wettability is attained with the process steps and reagents conventionally
used in the printed wiring industry. PTFE substrates require special surface
preparation to promote wettability either by sodium etching or by plasma
etching. These were discussed briefly in Section 15.3.7.2.
When edges or holes are plated on boards that have been bonded with a
thermoplastic fluoropolymer film, it is particularly important to ensure that the
exposed edges of the bonding film are also wettable.
15.3.8.3 Electroless copper deposition: Through-hole and edge plating of
adherent copper requires a layer of electroless copper, either to the required
thickness or thick enough to provide the conductive base for electroplating more
copper to the final thickness. The deposition of electroless copper from a
copper-formaldehyde solution of carefully controlled stability is usually ini-
tiated on a surface by a layer of colloidal palladium particles previously de-
posited from a commercially available aqueous 'catalyst' solution.
One common problem encountered in through-hole plating is the occurrence
of voids or open spots in the hole wall. They can arise from several causes. If
the process is not started with suitable reagents for cleaning and conditioning
the surface, subsequent depositions will fail to attach. Air bubbles can be
trapped in the hole during some process steps, particularly the catalyst treat-
ment. An excessively heavy deposit of the catalyst, by overheating, long immer-
sion time or high concentration, results in a non-cohesive deposit that can later
break away after the initial electroless copper is deposited. Inadequate use of the
accelerator can result in tin contamination that will inhibit electroless copper
deposition.
Several electroless systems are available from various manufacturers that
should provide adequate plating if the instructions are followed. The rinses
between reagent baths cannot be over-emphasized. They are needed to prevent
carrying the reagent from one bath to subsequernt baths. If such contamination
does not destroy the performance of the process it certainly impairs its quality.
De-ionised water for rinses is recommended as a way of maximising the life of
reagent baths.
936 Advances in substrate technology Advances in substrate technology 937

15.3.8.4 Electroplating: A large body of information is available in the litera-
ture and from plating-system suppliers on the electroplating process for apply-
ing copper to edges or hole walls in printed-circuit boards.
Ductility is of special concern with glass-PTFE substrates. All microwave
printed-circuit-board materials based on PTFE with woven or non-woven
reinforcement share the characteristic of a high Z-direction (thickness) thermal-
expansion coefficient compared with copper. The high thermal expansion of
PTFE is enhanced in the Z-direction when constrained in the X- and Y-axes by
reinforcing fibre. This puts strain on the plated copper during soldering or other
assembly operations. Further cyclic strain can occur on the finished printed-
circuit board in service. If copper ductility is low, the strain of thermal mismatch
leads to fracture failure of plated-through holes.
Special care is needed in electroplating to maintain ductility. Excessive bright-
ener additives and contamination, carried in with the boards being plated or
from degradation of the organic additives in the plating bath, induce low
ductility. Frequent testing of the bath for composition, and of ductility of test
deposits on polished stainless-steel plates, is required.
Copper-plating baths fall into four general classifications: copper cyanide,
acid copper, copper pyrophosphate and copper fluoborate. While the cyanide
bath offers the best ductility, it is deficient in throwing power and in its tendency
to attack most resist films, which discourages its use in printed-circuit manufac-
turing. Copper pyrophosphate has been used primarily for circuits, but requires
careful control. Acid copper seems to be gaining popularity for circuit boards..
Copper-fluoborate systems are seldom specified because of poor throwing
'
Cold peel tests at 90" angle were run at standard laboratory conditions after
specimens were floated on solder for 30 s at a series of temperatures. Hot peel
tests at 90" angle were done with the specimen immersed in the solder. In both
cases the test was run at a speed or 045ms-' (2inImin).
For the shear test the peel strip was separated from the substrate at its narrow
end by exposure to a flame to destroy bond. It was then peeled back to the
6.35 mm-square pad which remained bonded. The pad end of the specimen was
immersed for at least one minute in the solder at temperature. The force on the
peeled strip required in a direction parallel to the specimen surface for shear fail-
ure of the bond of the pad was measured. This test was intended to simulate the
effect of sideways force applied to circuit elements on a board during component
attachment.
The data are summarised in Table 15.19. In all cases the solder provided
shielding from ambient oxygen during heat exposure. Exposure to temperatures
approaching the 327C PTFE melt point does not damage the peel strength if
the specimen is allowed to cool without applied stress. Peel strength at ttm-
perature, while decreasing with temperature, is measurable up to at least 260cC.
Shear strength at temperature decreases with heating, but persists to near the
PTFE melting point. This indicates that, with protection from ambient oxygen
in air and care to avoid shear or pulling forces, one can use high temperatures
to attach components.
Table 15.19 Heat effects on conductor adhesion to PTFE substrate

power and the tendency towards non-uniform plating thickness. Solder temp. OC182 204 232 260 288 316 343 371
OF360 400 450 500 550 600 650 700
15.3.9 Device attachment to microstrip antenna substrates Peel, N/mm width
Integration of active devices onto the same board as the microstrip antenna cold/30 s float 2.59 2.59 2.59 2.59 2.22 2.08 1.10 1.12
offers advantages in space, weight and performance for many applications. Immersed in solder 0.94 0.56 0.28 0.14 0.02 0.00 0.00 0.00
Many of the methods for making the electrical connections from the device to Shear N16.35 mm square
the substrate were developed for use with ceramic substrates. Substrates based immersed in solder 71+ 71+ 71+ 57 36 28 28 7
on other materials either become soft at the temperatures used and/or can be
(1) Cold-peel failure up to 288C was into the substrate. All other peel failures were between foil
damaged by deformation, loss of adhesion of conductive circuit elements, or and substrate.
thermal degradation. (2) The cold-peel value for material not floated on solder averaged 240N/mm width with failure
into the substrate.
(3) The 71 + N hot shear values represent breakage of the 118 in copper-foil strip used to apply force
15.3.9.1 Effect of heating on conductor bond with PTFE: A study has been to the pad.
carried out on the effect of heating on conductor adhesion for PTFE-based
substrates. The work was performed with RT/duroid 5870 non-woven glass-
PTFE substrate clad with 34pm-thick (1 oz/ft2 weight) electrodeposited-type
copper foil. Similar results should be expected for other PTFE substrates. A 15.3.9.2 Various lead bonding methods: Several lead bonding methods are
series of 76 x 76 mm (3 x 3 in) specimens were taken from a single sample of briefly described here to provide some definition of terms.
1.57 mm (0.062 in) thick laminate. Each was masked and etched to provide The methods employed to achieve low-resistance electrical connections and
several 3.18 mm (0.125 in) wide peel-test conductor lines having a good mechanical integrity fall into one of the two categories of welding or
6.35 x 6.35 mm (0.25 x 0.25 in)-square pad at one end. diffusion bonding.
 938 Advances in substrate technology
Fusion-welded bonds are formed when metals of the lead and circuit board
are melted to resolidify upon cooling. A variety of techniques are available.
Resistance welding is done by pressing the lead and circuit conductor together,
and passing a high electric-current pulse between them to momentarily melt
metal at the contact points. Parallel gap welding is a variation of resistance
welding where parallel electrodes with a gap press the lead to the conductor so
that heat from a high-current pulse is conducted to the contact. Percussive arc
welding uses R F energy to initiate a pulsed DC arc between lead and pad, which
are then forced together to weld. Electron-beam welding focuses high-energy
electrons in a vacuum at the clamped lead and conductor to heat and weld them
together. Laser welding uses the energy of a laser beam to do the heating without
the need for vacuum. Solder reflow uses either localised or general heating by
infra-red or air oven to melt layers of low-melting alloy which were previously
plated onto both the lead and the circuit conductor.
Diffusion bonding depends on solid-state metal-diffusion phenomena to join
the lead and circuit conductor without the temperature ever reaching the melt
point. Ultrasonic welding uses ultrasonic energy to rub the surfaces together.
Friction from this cleans and heats the surfaces so that diffusion below the melt
point is induced. Thermal-compression bonding uses heat and pressure to force
cleaned surface of the lead and circuit-board conductor together for diffusion to
occur. Thermosonic bonding combines ultrasonic and thermal-compression
bonding by pre-heating the parts before use of ultrasonic energy.
15.3.9.3 Lead bonding to PTFE-based substrates: Ultrasonic wire bonding,
thermosonic bonding, parallel-gap welding, thermal-compression bonding and
solder re-flow are effective with PTFE-based microstrip-antenna substrates.
Ultrasonic wire bonding using a 25 pm (0.001 in)-diameter wire of aluminum
alloyed with 1 % silica can produce bonds of no less than 39 mN (4 gf) and
typically 54 mN against copper conductors with or without pumice scrubbing
beforehand. The abrasive oxide on the wire seems to aid in the rubbing action
to clean the surfaces. This can be done without distortion of conductors or the
dielectric substrate.
Thermosonic ball bonding uses a gold wire in a capillary stylus. The protrud-
ing tip of wire is formed into a ball by flame heating. The ball forms a nail-head
shape when pressed and ultrasonically rubbed against the preheated gold-plated
conductor surface to form the diffusion bond. Preheating is to 200C with
25pm-diameter gold wire. The stylus moves to the next connect point to make
a joint, after which the wire is cut. Gold plating is needed because gold is not
able to break through the thin oxide layer on copper that exists at the preheat
temperature. If the gold is plated directly onto the copper the copper can bloom
to the surface and interfere with bond. A nickel strike coat before 2-5 pm of gold
is plated prevents this. Some deformation of conductor and substrate may
occur.
Parallel-gap welding of 13 x 51 1 pm (0.0005 x 0.002 in) gold ribbon leads
Advances in substrate technology 939
with typically IOOA current for 04-1.3 s produced bonds averaging 343 mN
(35 gf) when the copper conductor is cleaned first. Machine settings are possible
to avoid distortion of the substrate or conductor. Plating on the circuit-board
conductor would improve the bond.
Thermal-compression bonding uses a resistance-heated stylus and just
enough pressure to get good contact without distortion. Very clean surfaces
with gold ribbon and gold-plated circuit conductors are necessary for suitable
results.
Solder reflow is effective with the tin-lead plated alloys widely used in the
printed-wiring-board industry. Low-melting-point solders based on indium are
effective where severe process-temperature limitations exist. The device lead and
circuit-board conductor are both plated. A resistance-heater head can be used
in a parallel-gap welding machine to form bonds and hold them until they cool
to get good-quality bonds.
15.4 Design considerations with selected materials
This Section offers a few selected topics for consideration in microstrip antenna
design which particularly pertain to substrate materials. There is potential for
degradation of performance from substrate changes in the application environ-
ment. Conductor losses become increasingly important at higher frequencies.
The use of multilayer circuit-board technology offers potential benefits in space,
weight and cost.
15.4.1 Environmental effects on antenna substrates
Environmental effects of space, temperature change and weather exposure can
impair performance by changing critical substrate properties.
15.4.1.1 Effects of space environment: Two environmental factors peculiar to
space and affecting design are vacuum outgassing and radiation exposure.
15.4.1.1.1 Outgassing: Many materials lose mass in the form of gases or
volatile condensable matter when subjected to a vacuum, especially when they
are heated, as is likely for an antenna exposed to sunlight in space. Excessive
mass loss degrades the ability to control the space vehicle. If the material is
condensable on a cooler surface, then, in a space environment, critical areas of
a vehicle can become contaminated as well as experiencing a change in the mass
of various parts.
Non-woven glass-PTFE and ceramic-PTFE substrates have outstanding
resistance to outgassing in space applications according to compiled test data
[50] (see Table 15.20). This would be expected of PTFE-based substrates gener-
ally, since the PTFE polymer is highly stable thermally and any volatiles would
be driven off during the high temperatures encountered during manufacture.
 940 Advances in substrate technology Advances in substrate technology 941
Test data shown in Table 15.20 were obtained on specimens etched free of The degree to which PTFE is affected by radiation is essentially a function of
copper foil. the amount of energy absorbed, regardless of the identity of the radiation; i.e.,
The test procedure [51] consists of vacuum heating 100-300mg specimens in p, y, X-rays, etc. have about an equivalent effect. The radiation dose unit usually
a copper enclosure, with the exit port at 125OC, for 24 h while a chrome-plated employed in radiation studies is the rad. I rad equals IOOergs/gm.
collector, located 12.7mm from the exit port, is maintained at 25OC. The total A dose rate of 10rads/h is often given for the Van Allen radiation belt. At this
mass loss (TML) and the collected volatile condensable materials (CVCM) are rate PTFE could operate for 5-50 years before a threshold-level damage would
expressed as a percentage of the original specimen mass. In general, materials be detectable mechanically.
with TML over 1.0 or CVCM over 0.10 should be avoided in spacecraft
applications.
Table 15.21 Summary of radiation dosage versus damage to PTFE
Rads in air Rads in vacuum
Table 15.20 Outgassing data
Threshold damage level (2 to 7) x lo4 (2 to 7) x 10' or more
Substrate type Non-woven Non-woven Ceramic-PTFE
50% tensile strength 1O6 lo7 or more
glass-PTFE glass-PTFE lo7 or more 8 x 10' or more
40% tensile strength
Product name RT/Duroid 5870 RT/Duroid 5880 RT/Duroid 6010 (2 to 5) x lo6
Retain 100% elongation (2 to 5) x lo5
Nominal K' 2.33 2.20 10.5
ASTM E 595 test
% TML 0.05 0.03 0.03 Since the primary function of microstrip antenna substrates is electrical, with
% CVCM 0.00 0.00 0.00 mechanical support usually provided by metallic components, the exposures
cited above can be expected to be well below the point where electrical perfor-
15.4.1.1.2 Radiation: Exposure to high-energy radiation is an important mance is impaired. The resistance of PTFE to radiation damage is generally
factor in space applications. Cosmic radiation is similar to nuclear radiation in better than that of solid-state electronic devices such as transistors and diodes.
many respects. It can damage materials after the prolonged exposure typical of
a space-vehicle mission. 15.4.1.1.3 Weathering: The features of low profile and conformability that
For PTFE-based substrates the component most susceptible to nuclear-radia- distinguish microstrip antennas also make them candidates for outdoor appli-
tion damage is the PTFE itself. Several investigations of the resistance of PTFE cations. Thus they are exposed to varying degrees of moisture, temperature and
to nuclear radiation have been reported [52-581. ultra-violet radiation that constitutes weathering.
Primarily, radiation of PTFE reduces molecular weight by chain scission. Many polymer systems are badly affected by weathering. Polymer-chain
Low cohesive forces between PTFE molecular chains require that the polymer scission occurs from ultra-violet-activated free radical mechanisms, from hyd-
have very high molecular weight to exhibit polymer-like mechanical properties. rolysis in hot and wet conditions, or from a synergistic interaction of both.
Thus PTFE ranks poorly in ability to withstand nuclear radiation. Thermoplastic materials generally become brittle, develop porosity, and exhibit
While this sounds very unfavourable at first glance, the other feature of space surface crazing that serves to increase vulnerability. Thermoset materials
environment, vaccum, tends to reduce the problem. Oxygen is essential to many usually show an increased tendency to absorb moisture as the quality of the
of the radiation-induced scission reactions. Its absence in vacuum reduces the interface between resin and filler particle breaks down. Woven-glass-fabric
damage or delays the effect of the damage. reinforced substrates show wicking effects as moisture follows polymer-glass
Molecular-weight reduction of PTFE principally affects mechanical proper- interfaces along continuous fibre multi-filament strands. The wet/dry cycling
ties, increasing brittleness and reducing tensile strength, modulus and extensi- tends to promote dissolving of components from glass or mineral fillers.
bility. Mechanical changes in PTFE appear to depend on the total radiation Compared with the rather mediocre rating PTFE has earned for resistance to
dose and to be independent of dose rate (see Table 15.21). nuclear types of radiation, PTFE-based substrates show very good resistance to
Radiation affects dielectric properties by embedding an electrical charge in weathering. The extremely hydrophobic nature of PTFE, combined with the
the PTFE, which decays with time. Thus dose rate, not total dose, is important absence of low-energy bonds or hydrolysable groups in the polymer itself,
to electrical performance. During irradiation the permittivity and loss factor at account for this. PTFE-impregnated glass fabric is well known for its long-term
lower frequencies will be temporarily increased. This effect is lower at the weathering resistance as the roof material in inflatable buildings, and as a
elevated frequencies of interest in microstrip antenna applications. protective radome for steerable microwave antenna systems. Non-woven glass-
 942 Advances in substrate technology Advances in substrate technology 943
PTFE is especially resistant to long-term moisture penetration owing to the between 10 GHz in X-band and the 35 GHz start of the millimetre-wave fre-
discontinuous nature of the glass-fibre reinforcement. quency spectrum, the losses associated with the conductor become the major
As manufactured, PTFE-based substrates have colour ranging from light tan contributor to transmission-line attentuation.
to dark grey. This is attributed to trace quantities of organic materials which
have been partially decomposed at the high processing temperatures used in
PTFE-substrate manufacture. Weathering tends to bleach the substrate. This Table 15.22 Predicted thermal change in thickness, Ki,, and Z, of stripline
visible effect of weathering is not accompanied by any measurable degradation (SIL) and microstrip on non- woven glass-PTFE with nominal
dielectric thickness of 7.57mm clad with 341tm foil
of properties, however.
Z, nom. 25 50 75 25 50 75
15.4.1.2 Temperature exposure: The effect of temperature on electrical f, GHz 1 1 1 18 18 18
properties of the substrate must be taken into account. Some microstrip designs Width 12.29 4.80 2.44 13.94 5.79 3.07
are more sensitive than others to a change in K' of the substrate. With PTFE, Temp. % Change from 20C value:
the effect unfortunately includes a change in the R value at the 19OC second-
order crystalline phase transition discussed earlier. Various schemes have been
devised to handle this effect.
The stripline-resonator method of Section 2.1 and a modified version of this - 100 - 1.31
I I
OC of thk. of K' of K& in microstrip
SIL
1.37 1.31 1.23 1.16 1.40 1.36 1.32
method have been used to collect data on the variation of K' with temperature -60 -0.89 1.04 0.96 0.91 0.89 1.07 1.02 1.01
for PTFE-based substrate materials. The stripline-resonator test is a precise -20 -0.48 0.71 0.65 0.64 0.61 0.74 0.68 0.71
observation of propagation velocity along a stripline, which is then converted 20 0.00 0.00 0.00 0.00 0.00 0.00 0.00 0.00
to a K' value. 30 0.26 -0.27 - 0.30 - 0.27 -0.28 -0.28 -0.29 -0.20
Since propagation of signals in stripline is TEM, the thickness change in the 70 1.32 -0.55 -0.60 - 0.53 -0.50 -0.56 -0.53 -0.51
substrate under test does not directly affect the test result. For PTFE-based 110 2.37 -0.91 -0.96 -0.91 -0.89 -0.88 -0.87 -0.81
substrates the thermal change in the observed value of K' appears to be ex- 150 3.42 - 1.53 - 1.61 - 1.49 - 1.44 - 1.53 - 1.50 - 1.42
plained primarily by the thermal change in density.
The situation is somewhat different for microstrip. The effective K' of micro-
strip is related not only to the K' of the substrate but also to the frequency and % Change of Z,,er from 20C value
cross-sectional geometry of the transmission line. Thus thermal variation with - 1.60 - 1.40 - 1.33 - 1.60
temperature of microstrip characteristic impedance and propagation velocity is - 1.20 - 1.00 -0.93 - 1.20
not only related to thermal variation of K' but also to thermal variation of -0.80 - 0.60 -0.53 -0.80
thickness. 0.00 0.00 0.00 0.00
Table 15.22 for non-woven glass-PTFE and Table 15.23 for ceramic-PTFE 0.40 0.20 0.27 0.40
were generated as an example of the expected response of microstrip trans- 1.20 1.20 0.93 1.60
mission line to temperature change. The microstrip data are derived from 2.40 2.00 1.60 2.40
known test data on the Z-direction (thickness) expansion and stripline dielec- 3.60 3.00 2.53 4.00
tric-constant/temperature relationship. The known thickness and stripline
dielectric-constant data at various temperatures were interpolated for the tem-
perature increments in the Table and then used for the microstrip-transmission- There are two causes for this. First the skin effect increases with frequency,
line designs shown. The analytical formulas of Hammerstad and Jensen [59] reducing the effective available cross-sectional area of conductor at a given line
were used for the microstrip calculations. width. Secondly, the maximum allowable dielectric thickness to avoid mode
problems decreases with increasing frequency. This requires a narrower line
15.4.2 Conductor losses at millimetre-wave frequencies width for a given impedance, further reducing the available conductor cross-
Even with low-loss PTFE-based substrates, the dissipation factor of the dielec- section.
tric becomes an important contributor to attenuation of signals on transmission The problem can be offset somewhat by selecting a substrate with the lowest
lines in either stripline or microstrip from about 0.5 GHz upwards. Somewhere possible K', maximising line width for a given impedance and substrate thick-
ness.
 944 Advances in substrate technology
The problem of conductor loss sets an upper frequency limit to practical
microstrip antenna designs. As clad laminates combining adequate copper-foil
adhesion with a smoother dielectric material-copper interface become available,
the limit will be extended.
Table 15.23 Predicted thermal change in thickness and KA,, of stripline
(SIL) and microstrip on ceramic-PTFE with nominal dielectric
thickness of 1.57mm clad with 34um foil
- --
Advances in substrate technology 945
accomplished by a divider network on the same conductor level. In other cases
there are constraints that require more elaborate circuitry. The available area
may be limited. Beam-width requirements may call for limited interaction
between radiator elements and distribution lines. Broad-band ability to control
phase relationships among radiators may be required. The power budget may
be tight. Space constraints may call for combining the antenna and RF process-
or in the same unit. The interconnecting circuitry for the microstrip array to
serve multiple antenna functions may be complicated. These are some of the

Zo nom. 25 50 75 25 50 75 reasons for multilayer circuit-board technology in microstrip antenna boards.

In'many cases protection of the microstrip radiator pattern by a radome is
f, GHz 1 1 1 12 12 12
Width 4.67 1.40 0.48 5.61 1.88 0.74 needed. This is provided for by bonding a superstrate layer similar to the
substrate over the exposed microstrip pattern.

7
Temp % Change from 20C value: Successful systems are being built as multilayer units. Some features offered

I I
"C of thk. of K' of K:, in microstrip
S/L
T -
% Change of Zo,e,from 20C value
by the technology include:
(a) Bonding boards into a unitised assembly
(b) Combining stripline and microstrip layers in a single board
(c) Combining boards of dissimilar relative permittivity, as needed
(d) Copper-plated holes to provide vias between layers
( e ) Assembling multilayer boards with buried and blind vias designed to mini-
mise reflection coefficients at the transition from one signal layer to another
(f) Alternating ground plane and signal lines
(g) Complicated interconnections with crossovers
(h) New materials with good microwave properties, combined with low Z-
--
direction thermal-expansion coefficient for minimal thermal stress on plated-
through holes
(i) Practical tight registration tolerances among layers
( j ) Combining stripline, where its features of low radiative losses and low
dispersion are needed, with the features of microstrip
The limits of what can be done with multilayer techniques for microstrip
antennas is being extended by ingenious designers and by the emergence of new
materials.

15.5 Special features and new materials developments

Alternative ways to interface with microstrip radiator elements, such as
conductor-free dielectric waveguides, have been proposed. Suitable substrate
materials, and the technology to form such structures cost-effectively, will also
push the limit upward.
15.43 Multilayer circuit-board technology in microstrip antennas
For many applications an adequate microstrip antenna can be a single micro-
strip board. Distribution of the signal to or from the radiating elements is
The number of options in substrate materials for microstrip antennas is expand-
ing. Some of these are special features that offer value exceeding added cost;
others are new substrate materials that may meet a specific need in certain
microstrip antenna applications.
Selected topics are discussed in this Section. Thick metal-clad substrates offer
built-in mechanical support for microstrip antennas. Co-polymers of PTFE
improve antenna performance in changing temperature. Resistors can be incor-
porated by printed-circuit processing for better designs. Microwave-quality
 946 Advances in substrate technology Advances in substrate technology 947
thermoset composites introduce a new approach to antenna needs. Developing thermal cycling in the application is avoided. The antenna can be self-support-
technology for very low-K' substrates shows promise of improved antenna ing without added hardware.
performance. If the metal cladding is not sufficiently thick, the unbalanced construction of
the laminate with stresses induced from the laminating cycle will tend to produce
Table 15.24 Characteristics of thick metals for clad substrates a bowed shape.
Metal Aluminum Copper Brass Special care is needed when etching antenna patterns and when plating to
Alloy 606 1 110 cartridge ensure that the thick metal ground plane is masked to prevent either damage to
Composition, % the cladding or contamination of the etching or plating baths.
Al 97.5 0 0
Cr 0.25 0 0 15.5.2 Low thermal coeficient of K' inJuoropolymer laminates
CU 0.28 99.8 70 PTFE is the preferred polymer matrix for microwave-circuit composites in spite
Mg 1.o 0 0 of the undesirable characteristic of a step change in thickness and K' during the
Si 0.6 0 0 19C crystalline transition. This undesirable feature is particularly troublesome
Zn 0 0 30 for microstrip antennas, where changing ambient temperatures can result in
some areas being below the transition while others are above.
Machinability indext 20 20 30
-- Co-polymers of tetrafluoroethylene with other perfluorinated olefin mon-
Hardness, ~ i i n e l l 30 25 45 omers are available that either do not exhibit this crystalline transition beha-
Tensile strength, MPa 124 226 314 viour, or show it to a much lesser degree at a much lower temperature. However,
Specific gravity 2.7 8.9 8.5 this improvement is a trade off. Co-polymers melt at lower temperatures, have
Specific heat, mJ/g/K 960 385 375 a lower melt viscosity and have a slightly increased dissipation factor.
Thermal conductivity, W/m/K 180 390 120 As an example of a special-feature material, RT/duroid 5500 non-woven
Thermal expansivity, 10-6/K 24 17 20 glass-fluoropolymer composite from Rogers Corporation is based on such a
Resistivity, R/m 47 30 110 co-polymer and has proved an effective solution to the microstrip-antenna
Relative-cost/unit-volume 1.O 1.5 1.4 of changing ambient temperature in various microstrip-antenna appli-
t On a scale where clock brass is 100. cations. Typical properties include R = 2 5 0 f 0.04, dissipation fac-
tor = 0.0025, and a linear thermal coefficient of permittivity of - 110 parts in
lo6 perdegK at 0-80C. Other properties are similar to those of PTFE-based
15.5.1 Substrates clad on one side with thick metal substrates, but the temperatures in processing and service must be kept below
Producers of glass-PTFE and ceramic-PTFE substrates are supplying clad 260C to avoid the lower crystalline melt point.
laminates having one side clad with heavy-thickness metal plate. The thickness
can be in the range from 0.5-13 mm (0.020-0.500 in). Table 15.24 outlines the 15.5.3 Microwave laminates with a resistive layer
characteristics of three frequently used metal-cladding materials. Other possi- Divider networks for feeding arrays of microstrip radiator elements usually
bilities include stainless-steel alloys and copper-Invar-copper laminates for require resistors for suppressing unwanted signal propagations, especially in
matching thermal-expansion coefficients with other components. Wilkinson power-divider designs. The high cost of mounting resistors onto
After the high-temperature exposure required for laminating PTFE sub- circuit boards often discourages wider use of this design approach.
strates, both aluminum and copper are fully annealed and too soft for easy Ceramic-PTFE and glass-fibre-PTFE substrates are now becoming available
machining. Where machining is critical, brass is preferred. Usually aluminum is clad with Ohmega-Ply* foil. Ohmega-Ply foil consists of either 17pm
preferred for its combination of thermal conductivity, low cost and low specific (0.5 oz ft2)or 35 pm (1 oz ft2) electrodeposited copper foil with an added 0.4pm
gravity. layer of resistive metal alloy on the side against the substrate.
For microstrip antennas the thick metal-backing offers several useful features:
Connectors can be mounted directly onto the board. The higher modulus of the 15.5.3.1 Characteristics: When microstrip or stripline transmission lines at a
thick metal cladding controls dimensional change of the softer substrate characteristic impedance of 50 p are formed with the Ohmega-Ply resistive layer
material, so that precise location of antenna-pattern features is attainable.
Fracturing of features in the thin conductor layer from cyclic strain induced by * Ohmega-Ply is a trademark of Ohmega Technolgies, Inc., Culver City, CA, USA
 948 Advances in substrate technology
between the line and the substrate, comparisons with conventional electro-
deposited foil cladding in various laboratories have shown no detectable in-
crease in insertion loss over the frequency range, at least up to 18 GHz.
Selective etching techniques provide areas where the copper has been removed
to leave a trace of the thin resistive alloy on the substrate surface to serve as an
integral resistor element. When properly processed and provided with a pro-
tective coating, tight tolerances of resistance can be produced and maintained
through exposure to high humidity, solder dip, thermal cycling and immersion
in boiling water, as detailed in Table 15.25.
Table 15.25 Characteristics of Ohmega-Ply resistive layer
Surface resistivity, 0 25
Resistivity tolerance, % 6.0
DC power-dissipation capability, kW/m2 UD to 620
Substrate non-woven ceramic-
glass-PTFE PTFE
Percent change in resistance maximum after:
I00 h/95% RH/35'C
20 s dip in 280C solder
1.O
1.O
100 cycles between - 55C and 125C 3.0 1.O
15 min in boiling water I .O
in -;>, .-,,,sion,
,I:-- the resistive layer is characterised for simplicity in terms of
surface resisitivity, ignoring its constant and small thickness. This is essentially
the same as the term 'sheet resisitivity'. Some refer to the units as Q/square to
distinguish the resistivity from the resistance.
The presently available cladding with 2552 surface resisitivity value is ade-
quate for designing most of the DC resistors needed in microwave boards. For
resistors handling R F it is desirable to keep the resistor length as short as
possible, and higher resisitivities are desirable so that the width of resistive
elements may be as wide as possible for better control of value. At present, for
Ohmega-Ply, resistivity values higher than 2 5 0 have not been adequately
reliable for consideration in antenna applications on PTFE-based substrates.
Development effort is in progress to produce a reliable material of higher
resistivity.
15.5.3.2 Designprinciples: If the resistor element is of rectangular shape with
copper conductors connected to opposite edges of the rectangle, the resistance
is simply the product of the surface resistivity and the ratio of length to width.
Advances in substrate technology 949
gular-shaped resistor element between edges attached to conductors,
W = width of the rectangular-shaped resistor element between unconnected
edges. For a square resistive element connected on two adjacent edges the
formula for resistance becomes
and its effective length is
A meander line with resistive layer can be designed for high values. If the line
is of uniform width and simple right-angle corners without radius or chamfer
are used, the line can be considered as a collection of rectangular elements in
series, where the length of each element is the orthogonal distance between
inside corners and the corners are considered to be square elements connected
at adjacent edges.
Thus a meander line of uniform width W, having five straight sections of
lengths L,, L2, L,, L, and L, with four square corners would have a resistance
of
0.5
1.O
R = 4(0.441)W +
(L, + L2 L,+ + +
L4 L5)/W [15.21]
The short segments of a meander line and the corners may be eliminated by
0.5 retaining conductive-copper-foil rectangles as jumper connections. The copper-
foil areas are made slightly oversize to ease registration requirements in process-
ing.
15.5.3.3 Processing boards with a resistive layer: A circuit pattern with inte-
gral resistors is generated in a process sequence that uses two photomasks and
three etching steps. The first mask for the composite pattern protects areas that
will finally be either copper or resistive conductor. The second mask protects
copper areas while copper is being removed over areas that will become resis-
tors.
With the composite-pattern mask in place, copper is removed with one of the
conventional copper etchants - ferric chloride, acid cupric chloride, alkaline
cupric chloride, or persulfate.
Etch rates for the resistive layer in copper etchant are slower. Attempting to
remove the grey-coloured exposed resistive layer with the same etchant will
result in undercutting of the copper foil and loss of pattern resolution. Etching
is stopped as soon as all the exposed copper is removed.
A second etchant specific for the exposed resistive layer is used next to remove
exposed resistive layer without undercutting the copper. The etchant is made up
as follows:
Water to which sulfuric acid is added: 800 ml

where R = resistance, 1, P = surface resistivity, 250, L = length of rectan- Concentrated sulphuric acid (100% H2S0,): 2 ml
 950 Advances in substrate technology
Copper-sulphate pentahydrate (CuS0,-5H20): 250g
Water for a final volume of: I000 ml
Note: Never add water to concentrated sulphuric acid as it may boil and
spatter with possible personal injury.
Etching conditions of 3-5min at 80-100C with agitation are sufficient to
remove the resistive layer. Since this etchant does not attack copper, it can be
used on boards from which the photomask has been stripped.
A second photomask is applied to the board, which is designed to leave
copper exposed only in areas where resistors are to be formed. This mask should
be designed so that its coverage extends beyond the copper to be protected. This
will minimise the potential for damage to the copper conductors by under-
cutting or slight misalignment of the photo-tool.
Exposed copper is etched away with a chromic-acid solution, which also
serves to passivate the newly exposed resistive layer. It is prepared as follows:
Water to which sulphuric acid is added: 800 ml
Concentrated sulphuric acid (100% H,SO,): 30 ml
Chromium trioxide, anhydrous (CrO,): 300 g
Water for a final volume of: 1000 ml
Notes
(i) Never add water to concentrated sulphuric acid as it may boil and spatter
with possible personal injury.
(ii) Be sure to keep the chromic-acid etchant free of surfactants and other
additives which can inhibit passivation.
Etching conditions are 5-7 min at 45-55OC with agitation. Rinse in deionised
water followed by a spray rinse with acidic sodium-bisulphite solution. Water
rinse again and dry in an air oven at 100C.
The acid sodium-bisulphite solution is made up by first dissolving 50g of
sodium bisulphite in 1 litre of water. Add dilute sulphuric acid to pH 3.
Use care to avoid abrading resistor elements. Strip the photomask. If possible
use an aqueous resist stripper. Solvent-based resist strippers may attack and
degrade the exposed resistors.
15.5.3.4 Resistor protection: In addition to passivation of resistors to
enhance thermal stability, protection against thermal shock, moisture and
mechanical damage is required. This can be accomplished by a conformal
coating on the resistor area.
The considerations in selection of a conformal coating for Ohmega-Ply
resistors include ease of application, electrical properties, heat resistance and
moisture resistance. The coating may be hand-brushed or silk-screened only
Advances in substrate technology 951
onto resistor areas. Types of coatings can include cross-linkable epoxy-resin
systems, solvent or aqueous dispersion forms of polyester-resin systems, acrylic-
resin systems, and various silicone-resin systems.
Hysol PC-17, a two-component epoxy coating from the Hysol Division of the
Dexter Corporation, 15051 East Don Julian Road, Industry, California 91749,
USA, has been effective. HumiSeal lB3 1, a one-component acrylic resin from
the HumiSeal Division of Columbia Chase Corporation, 26-60 Brooklyn-
Queens Expressway, West Woodside, New York 11377 USA, has proven to be
a reliable protective coating.
15.5.4 Thermoset microwave materials
A completely new class of microwave substrate materials is being introduced
with strong prospects for major benefits to the microwave industry. Some types
of microstrip antennas should benefit. The substrate materials are referred to
here as thermoset microwave materials, or TMM.
Thermoset resin systems, in general, seem to be characterised by K' values
well above 3, and with dissipation factor values too high for many microwave
applications. This is evident in the materials listed in Table 15.1, where cross-
linked systems include polyimides, triazine systems, bismaleimide resin and
epoxy systems. These materials are characterised by highly polar organic groups
for the cross-linking or curing reaction, which contribute to high dissipation
factors. The nature of the cross-linking reactions used is such that highly polar
and lossy groups are required. Typically the cross-linking is a condensation-type
reaction with volatile by-products that also contribute to dissipation. With
extended heating to complete the cure, low-molecular-weight fractions or by-
products volatilise and the voids left are susceptible to moisture penetration,
making the thermoset sensitive to humidity and moisture with respect to elec-
tronic properties. The thermoset resins are an unexpected quarter from which
to find good microwave materials.
Polar groups and volatile by-products in TMMs are at low levels, if present
at all. The polymer portion of the composition is almost entirely hydrocarbon
polymer chains with a very high cross-link density. TMMs are highly cross-
linked hydrocarbons. Before the cross-linking reaction TMM resin systems have
low melt viscosity, so that much higher levels of filler content can be accom-
modated with intimate blending compared with the higher-molecular-weight
and higher-melt-viscosity thermoplastic polymers, with PTFE as the extreme.
This ability to accept fillers at high levels allows TMM composites to be
designed for low thermal-expansion coefficient as well as other desirable attri-
butes.
Copper-clad panels of TMM are being evaluated for several microwave
applications including microstrip antennas. These are based on a series of
designed TMM substrate composites, as summarked in Table 15.26. Dielectric
thicknesses presently practical for manufacture start at 0.38 mm (0.015 in).
Clad laminates of TMM exhibit linear thermal change of K'. There are no
 952 Advances i n substrate technology Advances i n substrate technology 953
thermal transitions detectable up to 400C. Some composites have extremely comparable to that done on metallised ceramic substrates, with the important
low thermal coefficients of K' values. Weight loss with thermal aging is low, difference that large flat areas with precise thickness control can be processed.
namely 1.5% after lOOh at 270C. Weight loss exceeding 1% in air or nitrogen This capability should make more complex microstrip antennas an economic
does not occur until well above 400C with a heating rate of 10K/min for option.
20-40 mg specimens. Complex moulded shapes of TMM composites are feasible and could serve
Table 15.26 Summary of characteristics of various TMM composites as subsirates for special types of microstrip antennas.
TMM composites are readily drilled and routed with numerically controlled
TMM tvue machine tools as commonly used in the printed-wiring-board industry. The
K' at 3 GHZ (1) carbide-tool bits required are typical of those used for other types of laminates.
D at 3 GHz (2)
Therm. coeff./K', 10-6K-1(3) TMM composites differ in not having any tendency to form smears in drilled
CTE, 10-6K-1 holes. The chip produced in machining is in the form of granular particles easily
from 0 to 140C X, Y dir. cleared by a vacuum system. Machined surfaces can be very smooth, with
Z dir. precise control and no distortion during machining. Hard entry and backer
from 20 to 280C X, Y dir. boards are needed to prevent chipping away edges.
Z dir. Plated-through holes have been produced with standard procedures for elec-
Ins. res., TR/23OC troless and electroplating deposition of copper. Boards with plated-through
after 96 h/95 RH holes have not shown any hole wall damage after several minutes immersion in
solder at 288OC.
Water absorption, % 48 h/50C
0.125 in thick 15.5.5 Low-permittivity ceramic-PTFE laminates
0,050in thick A special ceramic-PTFE composite combines low K' with low Z-direction
Flexural strength, MPa thermal expansivity close to that of copper. This material* has demonstrated its
Flexural modulus, GPa value for multilayer printed wiring boards with high interconnection density for
Tensile strength, MPa high-speed digital-electronics applications. The thermal-expansion match to
Rockwell hardness (E scale) copper gives reliability to the plated-through hole vias widely used in such
Specific gravity boards.
Dielectric strength, kV/mm (4) In addition to the reliability of vias, the low modulus of the composite has
Water permeability, mg/m/h proved to be of value for ceramic-surface-mount chip-carrier devices. X, Y shear
(5) strains are induced either from thermal mismatch of the ceramic with the
substrate, or from actual differences in temperature during start up of equip-
( I ) K' values by perturbation cavity at 3 GHz for types 3,3E and 20, by FSR method for types IOT
and I ? ment. The low modulus of the low-K' ceramic-PTFE results in strain being
(2) D values were by perturbation cavity at 3GHz on a vertical-bar specimen: absorbed by the substrate rather than by the solder used to attach the surface-
(3) The FSR method at 4.3, 2.4, and 2.3GHz, respectively, was used for determining thermal mount chip package to the board. The substrate does not harden, fatigue and
coefficient of K'.
(4) The short-timemethod [60]in air with type 1 electrodes was used with 0.3-0.4mm-thick x 102- fracture, as does solder.
mm-d~ametermoulded disc specimens. A third special feature has become apparent - the combination of PTFE and
(5) The equilibrium rate was measured for 0.012-0,021 in thick specimens from 95RH/21.5C. ceramic filler shows essentially a zero thermal coefficient of K' over the 0-100C
Compare with published [61]values: 51-75 for polyvinyl chloride, 13 for high-density polyethylene,
6.4 for PTFE film, 15 for Mylar film control. temperature range. This makes it of special interest to microstrip antenna users
and producers, who have had performance problems with less thermal stability.
TMM composites are typically inert with respect to solvents and most strong K' at 2.94 f 0.04 is somewhat higher than glass-PTFE substrates.
reagents, and are also resistant to penetration by moisture.
Dissipation factor D is similar to PTFE-based composites at microwave 15.5.6 Very-low-dielectric-constant substrates
frequencies. Techniques have been found for producing a uniform fine-structured foam
The high degree of rigidity, bordering on brittleness, of TMM composites * Low-K' ceramic-PTFE substrate is designated RO 2800TMPTFE composite by the Microwave
rules out any bending of substrates, but it allows fast automated lead bonding Materials Division of Rogers Corportion, Chandler, AZ 85226, USA.
 954 Advances in substrate technology Advances in substrate technology 955
based on the TMM polymer system, and referred to as thermoset microwave -12. 'X-band effective stripline dielectric constant and dissipation factor for copper clad glass
foam (TMF). The foam is unusual in its combination of attributes, including woven fabric GR and GX laminates,' US Military Specification, MIL-P-13949F
high degree of heat resistance, resilience under compressive loading, low affinity 13. Institute for Interconnections and Packaging Electronic Circuits, 'Stripline test for permittivity
and loss tangent (dielectric constant and dissipation factor) at X-band', ibid., 1988, IPC-TM-
for moisture and low dissipation factor. 650, Method 2.5.5.5
Clad panels in 0.76-6.35 mm (0.030-0.25 in) thickness with precise thickness 14. 'Modified ASTM D 3380 stripline test method for X-band measurements of dielectric constant
tolerance of less than 51 pm (0.002in) make the foam of particular interest to and dissipation factor of RT/duroid 6010 ceramic-PTFE laminates.' Rogers Corporation,
designers of microstrip antennas where a K' value of 1.4-1.6 in a clad laminate RT6.1.2, I983
can offer efficient radiation performance over a wide bandwidth. 15. COHN, S. B.: IRE Trans., 1955, M T T J , pp. 119-126
16. SUCHER, M.: 'Measurement of Q' in' Handbook of microwave measurements' (Polytechnic
Although TMF is in the early stages of investigation and availability is Press, 1963). p. 456
limited, the characteristics summarised in Table 15.27 indicate its value for 17. ALTSCHULER, H. M., and OLINER, A. A.: 'Discontinuities in the center conductor strip
future microstrip-antenna applications. transmission line,' IRE Trans., 1960, MT-8, p. 328
18. MATI'HAEI, G. L., YOUNG, L., and JONES, E. M. T.: 'Microwave filters, impedance-
Table 15.27 Summary of characteristics of TMF matching networks and coupling structures' (McGraw Hill, 1964) p. 206
19. COHN, S. B.: 'Characteristic impedance of the shielded-strip transmission line,' IRE Trans.,
IS by waveguide perturbation at 3 GHz 1.4 or more July 1954, M l T , pp. 52-57
Tolerance of K' + 0.02 20. COHN, S. B.: 'Problems in strip transmission lines,' IRE Trans., March 1955, MTT
21. EDWARDS, T. C.: 'Foundations for microstrip circuitry' (Wiley, 1981) pp. 45, 58, 73, 74, 104
D at 3GHz < = 0.0006 22. HAMMERSTAD, E. O., and JENSEN, 0.: 'Accurate models for microstrip computer aided
Specific gravity 0.4 or more design.' IEEE MBTT-S International Microwave Symposium Digest, May 1980, pp. 407409
Cell size, pm 100 23. NAPOLI, L. S., and HUGHES, J. J.: 'A simple technique for the accurate determination of
CTE, 10-6K-' 50-60 the microwave dielectric constant for microwave integrated circuit substrates,' IEEE Trans.,
Flexural modulus, MPa 9.7 1971, MlT-19, pp. 664-665
Failure strain, % 5 24. HOWELL, J. Q.: 'A quick accurate method to measure the dielectric constant of microwave
integrated circuit substrates,' IEEE Trans., 1973, MlT-21, pp. 142-143
Retention of modulus at 100C. % > 90 25. LADBROOKE, P. H., POTOK, M. H. N., and ENGLAND, E. H.: 'Coupling errors in
cavity-resonance measurements on MIC dielectrics,' IEEE Trans., 1973, MTT-21, pp. 560-562
26. 'Test methods for complex permittivity (dielectric constant) of solid electrical insulating
15.6 References materials at microwave frequencies and temperatures to 1650C.' Annual Book of ASTM
Standards 10.02, 1985, D 2520-81
I. KOO, G. P.: 'Structural and mechanical properties of fluoropolymers' in L. A. WALL, (Ed.): 27. HARRIS, D. K.: Lancet, 1951, 2, p. 1008
'Fluoropolymers' (Wiley Interscience, 1972) pp. 516-521 28. American Industrial Hygiene Associate Quarterly, 1956, 17, p. 98
2. MCCRUM, N. G.: 'An internal friction study of PTFE,' J. Polymer Sci. 1959, 34, p. 355 29. HARRIS, D. K.: British J . Industrial Medicine, 1959, 16, p. 221
3. BUR, A. J.: 'Dielectric properties of fluorine containing polymers' in WALL, L. A. (Ed.): 30. WAGNER, W. D.: Letter Report to Research & Technical Service Br., August, 1961
'Fluoropolymers' (Wiley Interscience, 1972) pp. 475-503 31. CLAYTON, I. W.: J . Occupational Medicine, 1962, 4, p. 262
4. BROWN, R. G.: 'Vibrational spectra of PTFE: Effects of temperature and pressure,' J . Chem. 32. Federal Register Title 21, 12125555, 13 October, 1962
Phys., 1969,40, p. 2900 33. LEHMAN, A. J.: Association of Food and Drug Oficials US Quarterly Bull., 1962, 26, p. 109
5. KIRBY, R. K.: 'Thermal expansion of PTFE Teflon from - 190" to 300C.' J. Research NBS, 34. ZAPP, J. A,: 'Toxicity of plastics and resins,' Arch. Environmental Health, 1962, 4, p. 335
1965, 57, pp. 91-94 35. Hygienic Guide Series (American Industrial Hygienic Association, 1963), p. 198
6. BUNN, C. W., and HOWELLS, E. R.: 'Structure of molecules and crystals of fluorocarbons,' 36. CLAYTON, J. W.: Fluorine Chemistry Rev., 1967, 1, pp. 197-252
Nature, 1954, 174, p. 549 37. LEWIS, E. E., and NAYLOR, M. A. J.: American Chem. Soc. J., 1967, 69, p. 1968
7. DESANTIS, P., GIGLIO, E., LIGUORI, A. M., and RIPIMONTI, A,: 'Stability of helical 38. WARITZ, R. S., and KWON, B. K.: 'The inhalation toxicity of pyrolysis products of
conformations of simple linear polymers,' J. Polymer Sci., 1963, A-I, p. 1383 polytetrafluoroethylene heated below 500 degrees centigrade.' American Industrial Hygiene
8. MCCRUM, READ, and WILLIAMS: 'Anelastic and Dielectric Effects in Polymeric Solids' Assoc. J., 1968, 29, pp. 19-26
(Wiley) 39. COLEMAN, W. E., SCHEEL, L. D., KUPEL, R. E., and LARKIN, R. L.: 'The identification
9. 'Standard test method for permittivity (dielectric constant) and dissipation factor of plastic- of toxic compounds in the pyrolysis products of polytetrafluoroethylene (PTFE),' American
based microwave circuit substrates* Annual Book of ASTM Standards, 10.02, 1985, D Industrial Hygiene Assoc. J., 1968, 29, pp. 33-40
3380-82 40. COLEMAN, W. E., SCHEEL, L. D., and GORSKI, C. H.: 'The particles resulting from
10. 'Standard test method for dielectric constant and dissipation factor of polyethylene by liquid polytetrafluoroethylene (PTFE) pyrolysis in air,' American Industrial Hygiene Assoc. J., 1968,
displacement procedure' Annual Book of ASTM Standards, 10.02, 1985, D 1531-81 29, pp. 54-60
11. NOWICKI, T. E.: 'Microwave substrates present and future,' New Electronics, 1980, 13, pp. 41. KUPEL, R. E. and SCHEEL, L. D.: American Industrial Hygiene Assoc. J., 1968, 29, p. 27
85-86. 88
 956 Advances in substrate technology
42. SCHEEL, L. D., LANE, W. C., and COLEMAN, W. E.: American Industrial Hygiene Assoc.
J., 1968, 29, p. 41
43. SCHEEL, L. D., MCMILLAND, L., and PHIPPS, F. C.: American Industrial Hygiene Assoc.
J., 1968, 29, p. 49
44. 'Listing of Plastic Materials'. National Sanitation Foundation, March 1968, p. LO
45. DUPONT DE NEMOURS & CO.: J. Teflon, 1976, 1 1 , p. 8 Chapter I 6
46. GASKILL, J. R.: 'Smoke development in polymers during pyrolysis or combustion,' Smoke
and Products of Combustion, 1973, 2, pp. 1-34
47. PACIOREK, K. L., KRATZER, R. H., and KAUFMAN, J.: 'Oxidative thermal degradation
of polytetrafluoroethylene,' J. Polymer Sci.: Polymer Chemistry Edition, 1973, 2, 1465-1473
Special measurement techniques
48. 'Teflon fluorocarbon resins-Safety in handling and use' (DuPont de Nemours & Co., 1970)
p. 19898
for printed antennas
49. 'Health and safety aspects of fluoro-polytetrafluoroethylene' (ICI Technical Service, 1978) 2nd
edn
E. Levine
50. CAMPBELL, W. A., JR., and MARRIOTT, R. S.: 'Outgassing data for selecting spacecraft
materials.' NASA Reference Publication 1124 (revised), August 1987
51. 'ANSIIASTM E 595-77 standard test method for total mass loss (TML) and collected volatile
condensable materials (CVCM) from outgassing in a vacuum environment' Annual Book of 16.1 Introduction
ASTM Standards, 1980, pp. 595-77
52. MORRIS, P. 0.. JR.: 'The effects of combined environments on PTFE.' AIEE CP62-1284,
1962
The measurements of printed antennas are essentially the same as the measure-
53. FLORIN, R. E., and WALL, L. A.: J. Appl. Polymer Sci., 1959, 2, p. 251 ments of other antennas. The majority of these measurements lie within two
54. BOPP, C. D., and SISMAN, 0.: 'Physical properties of irradiated plastics.' ORNL-928, 1951 basic categories: impedance measurements and radiation pattern measurements.
55. 'Radiation resistance of Teflon in a simulated space environment.' Components & Materials The first category deals with the complex reflection coefficient or equivalently
Laboratory, Hughes Aircraft Co., TM-687, August 1961 the input impedance at the antenna terminals. The second category is broad in
56. LINNENBOM, V. J.: 'The radiation challenge,' Instrlation, Feb. 1962, p. 80
57. FRISCO, L. J.: 'Dielectrics for satellites and space vehicles.' John Hopkins University,
scope and includes various radiation properties such as beamwidth, sidelobe
Dieletric Lab., ASTIA No. AD276-867, 1962 levels, gain, polarisation etc. Measurements of the noise figure and the efficiency,
58. DUPONT Co.: 'Radiation tolerance of Teflon resins.' J. Teflon, Jan.-Feb. 1969, 10 both of which are a combination of the two categories, are sometimes also
59. HAMMERSTAD, and JENSEN: IEEE MTT-S Int. Microwave Symposium Digest, May required. In this Chapter we will not describe standard experimental methods
1980, pp. 28-30 in antennas. Readers interested in such methods should refer to textbooks such
60. 'Standard test method for dielectric breakdown voltage and dielectric strength of solid electri-
cal insulating materials at commercial power frequencies.' Annual Book of ASTM Standards,
as [I-41. Furthermore, experimental techniques for microstrip circuits in general
10.02, 1985, D 149-81 can also be found in the textbooks [5-61.
61. 'Encyclopedia of Polymer Science & Technology. Supplement Vol. 1.' 1976, p. 65 The purpose here is to reveiw some specific experimental techniques which
can be useful in the design and manufacturing processes of microstrip and other
printed antennas. These techniques are suggested, of course, only to supplement
measurements of the far fields and the input impedance. These techniques are
motivated by several reasons: first, the use of dielectric materials whose physical .
properties are not always known accurately or the use of multilayer substrates
made of different materials. Secondly, the transition from a coaxial transmission
line or a waveguide into the printed network is a difficult analytical problem.
The electrical properties of such transitions cannot be neglected in many cases,
and moreover the electrical properties can be used in the antenna design.
Thirdly, the use of complicated feed networks in large arrays suggests that their
properties should be checked experimentally, by resonant techniques, by time-
domain reflectometry (TDR) or by probing the near field. Direct efficiency
measurement of such arrays may also be of great help. The characterisation of
printed antennas can therefore be divided into three levels, each of which will
be described in subsequent Sections of this Chapter:
958 Special measurement techniques for printed antennas Special measurement techniques for printed antennas 959

(a) Physical constants of the substrate (Section 16.2) and properties of the where m, n are the mode numbers of the resonance. Certain improvements can
connector (Section 16.3). be achieved by preparing an almost totally enclosed resonator whose narrow
(b) Measurements of the electrical properties of printed lines and networks edges are metallised as well [9]. The resonator is excited by input/output printed
(Section 16.4). strips via small apertures, as shown in Fig. 16.2.
(c) The characterisation of the entire antenna by probing the near field (Section
16.5) and efficiency measurement (Section 16.6). parallel plate
input resonator output
coaxial A
coaxial
connector connector

16.2 Substrate properties

Commercial substrates are supplied with measured data on the two main
physical properties: the dielectric constant (or permittivity) E, and the loss
tangent (or dissipation factor) tan 6. The measured values are usually given at
a low frequency, e.g. 1 MHz, and at some operating microwave frequency like
10 GHz. Another physical property of interest is the surface resistivity of the
metallic cladding R,,which is related to the conductivity a, by

Fig. 16.1 A parallel plate resonator with a coaxial coupling (Source: Reference 10)

circular wholly
where f is the frequency and p, = 4n x (in SI units) is the permeability
of the vacuum. Conductivities of bulk metals are well known in the literature.
For example, copper has a conductivity u, = 5.88 x lo7 mho/m [7].
Although E, and tan 6 are usually known, there are instances in which their
precise characterisation is required. One typical case is when the substrate is
used at higher frequencies than those tested by the manufacturer. Another case
is when high accuracy and reproducibility are needed in mass production and
one wishes to cross-check the variations among different production series. In
other cases, multilayer substrates made of several materials, such as combina-
tions of spacer foam plates with dielectric layers, have to be characterised. Since
the tested materials used for antenna applications are shaped as flat plates, with Fig. 16.2 A parallelplate resonator with aperture coupling. The resonator is wholly metallised
one o r more metallised walls, it is natural to choose techniques based on (Source: Reference 10)
parallel-plate resonators. This simple geometry is preferred over various techni-
ques in which a small dielectric sample is inserted into a waveguide or other The main source of measurement error in these resonators is the shift in the
cavity, although the achieved accuracy is limited. resonant frequency due to losses. If the quality factor Q is also measured, the
A measurement procedure which utilises a flat dielectric plate, metallised on following correction to the frequency can be made
both top and bottom sides [8], is described in Fig. 16.1. The rectangular cavity
whose dimensions are a x b x h (h is much smaller than one wavelength) is
excited in its corner by a coaxial connector. The resonant frequency f,, is
measured either by the reflection from the connector or by the transmission into wheref, is the measured frequency and& is the corrected frequency that should
another connector. The dielectric constant of the substrate is given by be used in eqn. 16.2. Another source of error, caused by the coupling of the
probe into the resonator, is discussed in Reference 10. It is shown there that the
coupling errors in the two cases of totally and partly metallised substrates are
of opposite sense. Thus the two methods may be averaged to improve the overall
960 Special measurement techniques for printed antennas
measurement precision, which may be of the order of 0.5%. A re-examination
of the results obtained and a comparison with other methods is given in
Reference 11. The conclusion drawn there is that it is difficult to account for the
fringing errors; hence the overall precision in the measured dielectric constant
is of the order of 1 %. An interesting procedure for measurement of tan 6 and
u,,of substrates is outlined in Reference 12. The procedure consists of cutting the
double-clad substrate to some arbitrary-shaped closed resonator and measuring
its Q. A second cavity is constructed of twice the thickness of the first cavity.
Since the ohmic losses are not changed much in the two cases, one can separate
the dielectric and the conductive loss parameters.
resonator tested sample
strip \ /(non-metallised)
input
strip
strip
( standard
T V
substrate
Fig. 16.3 A microstrip resonator coupled by a gap. The resonator is loaded by the tested
dielectric sample (Source: Reference 73)
A different method suggested in Reference 13 is shown in Fig. 16.3. Here, a
non-metallised flat sample is placed above a microstrip section, creating a
cavity. The microstrip section, which is built on a standard and known substrate
(with EL, tan 6') is coupled to input and output terminals through small gaps. In
addition to the measurements of the resonant frequencyf, and the quality factor
Q, of this cavity, a second measurement (giving f , and Q,) is done while
replacing the tested sample by a reference material with known properties. The
four measured valuesf, ,f,, Q, , Q, and the knowledge of the reference properties
E,, , tan 6, enables one to compute the unknown properties E,,, tan d2, using the
equations given in the appendix of Reference 13. Although the calculations
involved in this technique are quite complicated, and the reported accuracy of
the dielectric constant is only 3% (but often as good as 0.5%) it has the
significant advantage that the loss tangent is also obtained.
Another test configuration, described in Reference 14, is suitable especially
for high E, materials like ceramics and semiconductors. The dielectric layer to be
measured is placed between two films of polymide (such as Dupont Kapton) as
shown in Fig. 16.4. The bottom side of the lower film is metallised while the
upper film contains a half-wavelength circular resonator and a coupling strip fed
Special measurement techniques for printed antennas 967
by a coaxial connector. The fixture is closed by air pressure in order to prevent
air gaps. By comparison with a known reference, checked in the same fixture,
the dielectric constant of the unknown substrate is obtained from the relation-
ship
connects
connects t o
outer cond ucror
of coax.
n.*-F- measured
'0' ring
air
pressure
Fig. 16.4 A test fixture for precise measurement of thin dielectric substrate (Source: Re-
ference 14)
where AE,/E,is the relative change in the dielectric constant and Afm/fmis tlie
relative change in the measured resonant frequency. K is a constant determined
by the shape of the resonant circuit and the material used. The value for the
specific structure in Reference 14 was K = 2.15. This relatively complicated test
configuration gave repeatable results of the dielectric constant within 0.1% +
relative to the known standard, and the deviations from the described techni-
ques can provide the measurement of the dielectric constant with accuracies of
the order of 1%. The measurement of the loss tangent is more complicated and
less accurate. A possible way to measure the loss tangent is to build a non-
radiating transmission line (like stripline) and measure the attenuation along the
line.
962 Special measurement techniques for printed antennas
The limited accuracies mentioned so far are inherently affected by the use of
medium-Q cavities in the described geometries. In cases where higher accuracies
are needed, one should use either small samples inserted into closed cavities, or
employ open interferometry techniques. Various such methods are summarised
in Reference 15 and reviewed in Reference 16 with a comprehensive list of
references. More recent contributions on dielectric-rod measurements and qua-
si-optical spectroscopy in millimetre waves can be found in References 17 and
18, respectively.
Special measurement techniques for printed antennas 963
experimental techniques for the characterisation and improvement of tran-
sitions from the printed antenna to the 'outer world' of the microwave circuitry.
Furthermore, it is strongly recommended that any design process of printed
antenna will include independent characterisation of the connector in use, by
one of the techniques described in this Section.

input output
16.3 Connector characterisation connector connector

In most cases there is a transition from the printed antenna to transmitlreceive

modules or to some measurement equipment having coaxial or waveguide
terminals. The inherent discontinuity of the currents in the vicinity of the
transition is a source of radiation and surface-wave losses. Also, its reactive Fig. 16.5 Testing a coaxial connector by a mated pair of connectors
nature may cause unwanted reflections and sometimes even changes the re-
sonant frequency of the antenna. These effects influence the antenna perfor- flange
mance, since all the radiated power of the antenna passes through this tran-
sition. The typical and most frequently used transition, from a coaxial line to the
microstrip line, is accomplished by extending the inner conductor of the coaxial
line and soldering it to the printed board. This connection is especially useful for
thin substrates, where the discontinuity is small in size and many commercial
connectors can do the job properly. Other transitions of interest are between
microstrip lines and waveguides, striplines and slotlines.
There are many cases in which the transition effects should be studied or taken
into account:
(a) In applications where very high level of impedance matching is necessary.
(b) In thick substrates, which are chosen in order to achieve increased band-
width, the connector effects are severe and cannot be neglected. In such cases,
several simple geometrical changes can often improve the transition. I \
(c) In the design of printed antennas, especially when large bandwidth for Fig. 16.6 A coaxial/microstrip transition test fixture (Source: Reference 30)
matching is needed [19-211, the parasitic impedance of the transition may be Typical dimensions are: dielectric-substrate thickness ranges between 0.25 mm
used. and 3.2mm and the central pin extension should be less than 1 mm
(4 In research and development of new printed radiators, especially those
combined with modern integrated and monolithic circuits, unconventional Commercial connectors are usually tested by the manufacturer in a coaxial
feedings are often used. structure consisting of two identical connectors in contact [32] as shown in Fig.
The analytical modelling of the coaxial/microstrip and the other transitions is 16.5, or through a high-precision air line. This configuration is tested in reflec-
a complicated field-theory problem. Remarkable progress in this subject has tion or in transmission by a network analyser or by a TDR technique [33]. A
been achieved in recent years [22-261, but still the works published seem to be basic test fixture of a coaxial/microstrip transition, mounted on a jig, is presen-
impracticable for design use. On the other hand, simple approximations which ted in Fig. 16.6 [34, 351. The microstrip line here is open ended, but a shorted
model the transition as short transmission lines or as lumped element networks end may be used as well. The reflection coefficient (or VSWR) of the transition
[27-311 are not accurate enough in many cases. It is therefore important to use is found from the overall measured reflection at the connector input. An
interesting suggestion [36] is to repeat the measurement for several lengths of the
 964 Special measurement techniques for printed antennas 965
Special measurement techniques for printed antennas
printed line. Averaging the measured data reduces the error caused by the connector is terminated by a high-quality coaxial matched load PI(/). The total
radiation loss of the open end. Naturally one would like also to measure the reflection coefficient is a superposition of the two reflections coming from the
reflection while the end of the printed line is matched by a termination. How- two connectors. This coefficient is given by
ever, the realisation of high-quality microstrip terminations is not easy. Several
types of terminations are described in the literature [6, 37, 381 and in technical
catalogues such as References 32, 39 and 40. Three examples are shown in Fig.
16.7. The first example is a 50 iZ thick-film resistor inserted into the substrate
and shorted at one end to the ground plane [32]. The quality of such resistors microstrip ground
line \ plone
in terms of VSWR is not high, but they are easy to build. A second termination
is made of a precision attenuator pad made in thin-film technology [39]. The
significant advantage here is its small and compact size. The third example is a
tapered resistive layer mounted on the top of the printed line [40]. This con-
figuration gives good matching over a broad band of frequencies, but it requires
a long length of at least half a wavelength. A comparison of various printed
terminations and experimental testing of connectors by such terminations are
given in Reference 37.

thin film resistive

pod box

w
Fig. 16.8 Svmmetrical pair of coaxial/microstrip transitions makes a sensitive test set
a .!'? metallic short
I to ground
b L-- plane Since the summing- of the two waves is coherent,
(16.6)
~ ~ r o l ~=2 ll-I + r 2 I 2
tapered
absorbing where l-, is the reflection from connector 1, and l-, is the reflection from
film connector 2, the individual reflection can be written as
I Er V I-, = jr, le'+ (16.7)
ground plane
. -.
Fig. 16.7 Typicalmatched terminations for use in microstrip lines (Source: References 32.39.
401
where I r,1 and 1 Fzlare the respective amplitudes, 4 is an arbitrary phase and
a shorted thick-film resistor [32], 4nlKf/1,is the phase added by the round-trip propagation in the microstrip
b Precision attenuator pad [39] frame. The phase 4 can be selected to be zero and the reflected amplitudes of
c Tapered resistance film [40] the two connectors are assumed to be identical. Hence, the total reflection is

A simple, sensitive and accurate test set can be made from a pair of identical
transitions [3O, 31, 411 where no shorts, open ends or terminations in printed . . .
form are needed. The schematic structure, shown in Fig. 16.8, enables not only
which is the well known expression for interference effects. The measured
characterisation of the connectors, but also gives valuable information about
reflection coefficient will be a periodic function of the frequency, with maxima
the printed line itself. The test procedure is as follows: two measurements are and minima
made over as wide a frequency range as possible. The first measurement is the
reflected power from a coaxial short (without any transitions) PI0).The second Il-torliax = 41I-l2
measurement is the reflected power from the test fixture while the second
 967
966 Special measurement techniques for printed antennas Special measurement techniques for printed antennas

and with a period of i centre-conductor diameter [42,43], insertion of the connector into the dielectric
[42,44], addition of a metallic ring around the central conductor 121,421,drilling
l a hole in the ground plane [34], an off-centre shift of the central pin [41] and the

where c is the speed of light in vacuum. This measurement is very sensitive 1 commercial x connector [45] Measured graphs of the last two improved connec-

lOOn microstrip iOOn microstrip

because the coherent summing multiplies the reflection from the single connec-
tor by a factor of 4, as shown in eqn. 16.9. This technique may also be used to
measure the effective dielectric constant of the printed line using eqn. 16.10, and
additional information about the losses can be gained by measuring the trans-
mission to the second connector. 0 b

lOOn microstrip L
I 50, rnicrostrip

\ hole in base plate

d

Fig. 16.9 The microstrip test frame consists of a closed microstrip line printed on a dielectric
substrate and two identical connectors (Source: Reference 42)

e
A modification of this method, introduced in Reference 42, is useful for
vertical connectors where the outer conductor is soldered to the ground plane
of the antenna. Since usual coaxial connectors have an impedance of 50 Q, it is
convenient to attach the connector to the middle of a 100 R microstrip line. In alternate attachment
techniques
this way, two lines of 100 R are seen in parallel, providing an impedance of
50 R. Such 100 R lines are used, of course, later on in the feed network of the f
antenna. The measurement structure in this case is made up by a closed micro-
strip line and two identical connectors which are soldered to the line so that the
line is divided into two equal lengths (Fig. 16.9). Smooth corners are preferable Fig. 16.10 improved coaxial/microstrip transitions (Sources: References 21. 34. 47-45)
for the frame, but chamfered corners are also acceptable. a Narrow pin 142. 431
The measured reflection coefficient of the coaxial transition is the result of the b Inserted connector [42. 441
discontinuity between different types of transmission lines. The transition area c Metallic ring [21, 421
is modelled sometimes by some parasitic reactance in an equivalent electrical d Compensating hole 1341
e Off-centered pin [41 I
network [27-311. In cases where information about the reactive nature of the
f K connector [451
connector is needed, the complex reflection coefficient should be studied; e.g.
both amplitude and the phase are to be measured. Several geometrical changes tors, which were performed using the two-identical-connectors approach (via a
in the transition can help compensate for the reflections, either by a 'cut and try' straight microstrip line) are shown in Fig. 16.11. A close look at these graphs
procedure or by numerical modelling of an equivalent circuit. A few examples reveals the periodic behaviour of the reflection and the remarkable VSWR
of improved connectors are shown in Fig. 16.10, both for horizontally and values of the single connector. In Fig. 16.11a [41] the period is Av = 0.9 GHz,
vertically mounted connectors. These improvements include: reduction of the the length of the microstrip section is 2.0 in and thus Ld
= 3.28. The maximal
 Special measurement techniques for printed antennas Special measurement techniques for printed antennas 969
968
measured VSWR is 1.1 and hence the VSWR of the single connector is 1.05. In
Fig. 16.11b [45] the period is Av = 3.9 GHz, the microstrip length is 0.5 in, and
I,
wafer
thus Kf = 3.03. The maximal return loss is - 15 dB, which is equivalent to 1 under
test
vlew
a VSWR of 1.43, and thus the VSWR of each connector is 1.2, most impressive
for this band of frequencies. 1 contacts

coplanar

max VSWR=1.1
VSWR of each wafer bottom
connector 1.05 under view
test

frequency, GHz

contact

m'0 wafer view

under
rnax RL=-15dB $
20 test
VSWR of each " Fig. 16.12 Details of the 'on wafer'probe in top, bottom andside views (Source: Reference
connector 1:l.Z $ 49)
30

40
0 10 20 30 40
frequency, GHz

Fig. 16.11 Measured results of VSWR and return loss of two improved connectors. The
measurements were done by a connector pair (Sources: References 41, 45)
a Two off-centered pin connectors of Fig. 16.10e [41],
b Two K connectors of Fig. 16.1Of [45] balanced
probe heads

The idea of two identical connectors can be applied to all kinds of transitions.
For example, the integrated waveguide/microstrip transition described in Re-
ference 46 or the microslot/microstrip transition shown in Reference 47 were
tested in this way. In another recent work [48], a versatile mounting fixture has
been proposed for Gallium-Arsenide substrates. This fixture has an adjustable
ground plane which is capable of acommodating substrates of arbitrary thick-
nesses. Good electrical contacts are achieved by a conductive rubber layer.
A novel tool for integrated and monolithic circuits, known as 'wafer probing'
[49, 501, is applied to the bonding pads of a semiconductor chip. Two probes NETWORK
ANALYZER
described in Fig. 16.12 are made of coplanar lines and small metallic contacts.
They are connected to a network analyser through wideband baluns, as shown
in Fig. 16.13. This microwave-wafer probing has been shown to be an accurate
and convenient tool for the detailed network analysis of monolithic elements, Fig. 16.13 'On wafer' measurement by two balanced probes and a corrected network
and any new feeding configurations of printed antennas can be tested as well. analyzer (Source: Reference 49)
 970 Special measurement techniques for printed antennas
16.4 Measurements of printed lines and networks
After presenting the characterisation of substrate properties and measurement
of connector effects, we come to the printed line itself and the building blocks
of feed networks, to be discussed. First, one is interested in the three fundament-
al parameters of the basic printed line:
(a) The effective dielectric constant ce8, related to the propagation constant P by
where k,,is the free-space wave number.
(6) The attenuation factor a, consisting of an ohmic part a, and a dielectric part
ad.
(c) The characteristic impedance of the line Z,.
It should be noted that these are not physical properties of the substrate, but
rather the electrical properties of a specific given line.
Secondly, one should investigate the properties of certain printed structures
used in feed networks: bends, width changes, T junctions, cross-junctions etc.
Each one of these structures is actually some kind of a discontinuity that causes
reflections and losses. One is interested either in quantitative measurements of
these effects or in modelling them by equivalent electrical circuits. In antenna
applications, the power splitters and the delay lines are the most important
building- blocks to be considered.
General methods for measurement of microwave transmission lines are well
known in the literature and in practice. The great majority of the measurements
are performed in the frequency domain using network analysers, with or with-
out automatic error correction. The main problem with the network analysis of
printed circuits is the fact that the test ports and the standard calibration units
(short, open and matched terminations) are available in coaxial or waveguide
forms. Any measurements of printed lines have to be made, therefore, through
transitions. The inherent ,reflections from the transition and the difficult deter-
mination of the reference planes limit the obtainable precision. One simple
solution is, of course, to use excellent connectors (VSWR of about 1.01) for the
network analysis. Another solution is to use a pair of the tested devices,
separated sufficiently from the connectors, and to get different resonances for
the connectors and for the devices under test. A third solution, which is often
the most practical one, is to use resonators which contain the devices under test.
These resonators, which are coupled by air gaps or by other means of weak
coupling, have high sensitivity and high accuracy, as will be described later on
in this Section. A different approach is to use time-domain reflectometry in
which discontinuities, including the transitions, are separated on the time axis.
Direct probing along printed lines and printed structures may also be considered
as a quantitative tool, and this technique will be discussed in the next Section.
The purpose of this section is therefore to review several practical methods for
1i Special measurement techniques for printed antennas 971
I the characterisation of printed-line parameters and networks. Most of the
examples deal with microstrip lines, but they are general enough to be applied
to all kinds of printed or dielectric transmission lines.
16.4.1 Measurements of printed-line parameters
16.4.1.1 E ~ The
~ . three fundamental parameters to be measured are the effec-
tive dielectric constant, the attenuation factor and the characteristic impedance.
An example of the measurement of cef has been shown in eqn. 16.10. This
4
method, based on a straight printed line between two connectors, is not very
accurate, mainly because the distance between the two connectors is subjected
I to different definitions. High-sensitivity and well-defined lengths are achieved in
I printed resonators. The idea is to create an exact printed structure and bring it
into resonance by weak coupling via small air gaps. The ring resonator is an
attractive candidate for this purpose [S, 6, 51-53]. The ring shown in Fig. 16.14
is excited and measured by an 'input' strip and an 'output' strip. The effective
dielectric constant of the tested line is related to the resonant frequency of the
ringf, by
where n is the order of resonance and 1 is the mean circumference of the ring.
input
resonator
Fig. 16.14 A ring resonator for measuring the effective dielectric constant of a microstrip line
(Source: Reference 51)
The main advantage of a ring resonator is that it is free of end effects. However,
the curvature effect reduces the accuracy and a corrected procedure for deter-
mining the value of E~ is described in Reference 53 as follows: The ring is
characterised now by three geometrical parameters: ri, r, (inner and outer radii
of the ring) and w (width of the line). Now define two effective radii Ri and R,
by
 972 Special measurement techniques for printed antennas
The effective width at frequency f is defined by
where Wcff(0)is the effective width at zero frequency, and
Here h is the thickness of the substrate, q, is the free-space impedance (377 R)
and Z, is the characteristic impedance of the line. ~ ~ ~is( the 0 )electrostatic
approximation for E@ and f , is defined by
The k number of the resonance is found by solving the equation
J"(~R,)Y"(~R.)- J"(~R,)Y"(~R,)= o (16.17)
where Ji, are derivatives of the Bessel function of the first and second kinds
of order n. Finally one gets
This procedure is shown to provide improved accuracy over eqn. 16.12, and it
is estimated to be better than 1%. The accuracy is better for large radii and for
thin substrates.
Another configuration for measuring the effective dielectric constant is a
straight resonator of length I, coupled at its edges or through its sides [6]. This
resonator has two abruptly open ends which are best accounted for by consider-
ing the line to be longer by I, on each edge. The effective dielectric constant is
therefore
The value of I, should be taken from theoretical calculations or from known
data.
A most practical arrangement, made of two straight resonators with lengths
of I, and 1, [6, 54-56] is shown in Fig. 16.15. The two resonators are measured
independently and two resonant frequencies f , and f, are found. These two
measured values are then used to find both E,, and 1, [54]
nc(2fi - f,)
f f = 2f,t;h(12 - I,)
Special measurement techniques for printed antennas 973
For successful operation of this technique it is recommended to make
I, = 21, + 1,; thus an approximate a priori knowledge of I, is required. The
obvious advantages of this 'two-resonators' technique over the ring resonator
are that the accuracy does not depend on the size of the resonator and that the
production tolerances for straight section are better than those for curved lines.
11
input )7 1 t 2 = 211
Fig. 16.15 Two open resonators for measuring the effective dielectric constant (Sources:
References 6 and 54)
input (7
11
Fig. 16.16 Two open resonators for measuring the effective dielectric constant and the
end-effect of a microstrip line (Sources: References 6 and 57)
Another version of the straight-line resonator is described in Fig. 16.16 [6,57].
Here, the first resonator consists of two sections 1, and 1, while I, equals 1,/4.The
second resonator is made of the section I, alone, or practically it is the first
resonator after the I, section has been removed. This configuration distinguishes
between the effective length of the gap and the effective length of the open end.
The procedure is as follows: first, the effective dielectric constant of the tested
line is measured by one of the techniques described earlier. Then the two
resonant frequenciesf, andf, are found and the two equations
nc
-J- = li+lq+L0 (16.23)
2f 8.8
are used, to find the effective length of the gap, I,, and the effective length of
the open end, .I,,. Now the final value of the effective dielectric constant is
obtained iteratively.
 974 Special measurement techniques for printed antennas
16.4.1.2 a: The attenuation factor a is another parameter of interest, and
detailed theoretical and experimental data on microstrip lines are available in
References 58-60. Direct measurement of the power loss of a straight line is not
practicable unless the losses are high as occurs with millimetre waves [61]. Usual
values of a for commercial substrates are between 0.05 and 0.1 dB/&. In the
presence of radiation and surface-wave losses, such values are difficult to meas-
ure without careful calibration. It may be useful to use a stripline made of the
substrate under test and measure the attenuation.
An alternative way is to measure the quality factor Q of a resonator and then
solve for the attenuation factor using the basic expression [62]
General reviews of Q-factor measurements can be found in References 63-65.
To determine a from the Q measurement one has to separate the radiation and
the surface-wave losses. The situation is summarised by the equation
where Q is the measured quality factor defined by f/AA where f is the resonant
frequency and Af is the width between half-powc- points. Q, is the unloaded
quality factor and Q, is the unknown, to be found.
A suggested procedure for reduction of the error [6] is the following:
(a) Measure the quality factor of a ring resonator where radiation losses are
smaller and find an approximate value of Q = Q,.
(b) Measure the quality factor in a straight-line resonator having the same
parameters, and find the value of Q.
(c) Use eqn. 16.25 to calculate Q,.
(4 Use eqn. 16.24 to find a.
More information on the printed-line attenuation can be achieved by measuring
another line with very low dielectric constant, where the dielectric and the
surface-wave losses are close to zero. Also, the procedure described in Section
16.2 of measuring two resonators of two thicknesses may be used for the same
purpose.
Another technique requires only a straight-line resonator. A relationship
between Q, and the reflection coefficient p at the resonator input is proved in
Reference 6 to be
n is the order of resbnance. Thus the measured end-reflection magnitude p
enables the estimate of Q,, and hence of a.
16.4.1.3 2,: 'Characteristic impedance' is a fundamental concept in micro-
wave circuits; however, its exact meaning in quasi-TEM (transverse electric
Special measurement techniques for printed antennas 975
magnetic) lines is not always very clear. Getsinger [66] uses the term 'apparent
characteristic impedance' to denote the parameter which describes how printed
line (microstrip in this case) exchanges power with a TEM line. The term
'characteristic impedance' defines how one TEM line exchanges power with
another TEM line. Experimentally, there are at least three accurate and reliable
methods for impedance measurements [66, 671, denoted as: the slotted line, the
real-axis intercept and the group-delay method. A brief description of each
technique is given here for the microstrip case.
The slotted line procedure for finding Z, is as follows:
(a) Prepare a microstrip line short-circuited at both ends and find the effective
dielectric constant by a resonant technique.
(b) Replace one short-circuit with a high-quality transition to the coaxial line
and predict the frequencies for which the line presents short-circuits, open-
circuits or +jZ, at the junction of the connectors
(c) Use a slotted line to measure the reactance at each of these frequencies. Use
a shorted coaxial reference to define the terminal plane at the connector end of
the microstrip. This technique eliminates equipment calibrations because only
distances and frequencies are measured, and the overall accuracy is of the order
of 0.1%.
The real-axis intercept procedure for determining Z, is done as follows:
(a) A long uniform microstrip line is placed between two connectors with very
low reflections. One of the connectors is matched by a coaxial load with an
impedance Z , = 50 a.
(b) The input impedance at the other connector is measured by a network
analyser as a function of frequency, and a Smith chart is recorded.
(c) The input impedance makes nearly circular spirals on the Smith chart and
XI and X, are the extreme values of the reactance. The algebraic average Xis the
residual reactance of the transition, i.e.
+
x = (XI X2)/2 (16.27)
The intercept of the input impedance with the real axis of the Smith chart is
found experimentally at some value R and the characteristic impedance of the
tested line is
(16.28)
Z, = (RZ, - y ) l i 2
The real-axis intercept is both a simple and a fast procedure, but it has a
significant disadvantage: If R is close to Z , , the impedance locus is nearly
parallel to the real axis and a small reflection from the transition causes a large
error in the real-axis crossing. Thus poor accuracy would be expected for
printed lines with Z, close to 50 8.
The group-delay method is important because it is almost independent of the
coaxial-connector parasitics and it is fairly simple to perform. First, prepare a
 976 Special measurement techniques for printed antennas
microstrip line terminated with a flat-plate short. Second, measure the reflection
delayT about frequencies for which the shorted microstrip presents zero reac-
tance at the transition. The equation for T is
which represents the total phase shift of the reflected wave (i.e. the angle of the
reflection coefficient). Z, is then given by
Special measurement techniques for printed antennas 977
analysers. Typical examples of such resonant techniques will be shown in this
Subsection, followed by a short description of time-domain reflectometry. A
classical example of the characterisation of a bend in microstrip line [68] is
shown in Fig. 16.17, In this method, two bends are joined by curved lines of
lengths I, to form a continuous 'ring' resonator. This resonator keeps two
resonances related to 1, and to 21,, and thus the properties of the bends can be
found. Additional comparison is made with a circular ring built on the same
substrate.

where n, an integer, is the length of the microstrip in half wavelengths atf, and
D is given by

E, is the relative dielectric constant of the substrate and ~ ~ ~is (the

0 effective
)
dielectric constant at zero frequency (see eqn. 16.15 and 16.16).
The last method needs specific theoretical adaptation for each type of printed
line in use. In summary it can be seen that the slotted-line method is accurate
but difficult to perform. The real-axis-intercept method is the easiest to make
but requires excellent connectors and loads. The group-delay method is simple
and accurate, and does not require good transitions. But the interpretation of
the result is specific to microstrip lines.

Fig. 16.18 An arrangementfor measuring the equivalent electricalparametersof a T-junction

(Source: Reference 69)
The components L,, L,, C and n are measured in ( a ) - ( d ) , respectively. The
output equivalent circuit is given in ( e )
Fig. 16.17 A closed 'ring' resonator with bends for characterisation of the bend properties
(Source: Reference 68)
Another common discontinuity is the T-junction, often found in power split-
ters. A detailed arrangement for measuring the equivalent electrical parameters
16.4.2 Measurements of printed networks
of the T-junction is suggested in Reference 69. Four different test junctions are
shown in Figs. 16.18a-d. Each of these junctions gives the best modelling for a
16.4.2.1 Resonant techniques
different component in the equivalent electrical circuit, shown in Fig. 16.18e.
Networks and line discontinuities can be characterised by the techniques des-
More theoretical and experimental information about discontinuities can be
cribed heretofore, namely, the investigation of resonant structures by network
found in References 69-73.
 978 Special measurement techniques for printed antennas
Further demonstrations of practical measurements of T-junctions are shown
in Figs. 16.19 and 16.20. Fig. 16.19 shows a straightforward arrangement in
which a 50 Q line is connected to two 50 Q lines, using a quarter-wave transfor-
mer. The overall losses and phase shifts of this splitter can be easily found, but
the connectors have a significant effect on the results. A modified version, shown
in Fig. 16.20, is intended to check the T-junction between 50 R line and two
oppositely directed 100 R lines. Here, two symmetrical junctions are built with
a distance I2 between them, while the distance between the connector and the
T-junction is I,. Recalling the interferometric relation in eqn. 16.10, one can
easily find that the periodic ripple due to the connectors is
Fig. 16.19 Practical testing of a T-junction
, 79-81 by the use of combined reflection and transmission measurements, or by
Fig. 16.20 Improved test of a T-junction by two junctions
Different resonances are measured due to the connectors and due to the T-
junctions
while the periodic ripple due to the T-junctions is
A sufficiently large difference between I2 and (21, +
1,) is required in order to
achieve high accuracy. An additional test of two connectors alone will give a
good reference for this measurement.
A few words should be injected here about an alternative network analysis
Special measurement techniques for printed antennas 979
known as the 'six-port reflectometer' [74-781. The main feature of this technique
is that a six-port network with square-law detectors on four of the 'output'
ports, can be calibrated by a computer, to determine the amplitude and the
phase of signals at the other two 'input' ports. The simplicity of the detection
reduces the cost of precise hardware, although the accuracies are of the same
order as in error-corrected vector network analysers. One important attraction
of the six-port technique is the relative ease with which it can be employed in
the millimetre-wave range. So far as the characterisation of printed lines is
concerned, the six-port technique, being based on coaxial components, has the
same drawbacks as the vector network analysers.
16.4.2.2 Time-domain rejectometry: Time-domain measurements are per-
formed by transmitting a known waveform out into the test network o r into a
discontinuity and measuring the waveform returned as a function of time. The
delay of the returned waveform aids in finding the distance to the network, and
the shape of the returned waveform gives information about the impedance of
the network. Since commercial pulse generators have a typical rise time of about
t, = 30 ps, the resolution of the distance is of the order of
for air lines, and up to A[,& z 1.5 mm for printed lines on alumina. The
time-domain technique eliminates the problems involved with coaxial connec-
tors and coaxial standard references. It gives an immediate recognition of the
impedance of various networks, while the accuracies are limited by the rise time
of the source. Improved accuracies and sensitivities are reported in References
statistical signal processing.
An example of TDR results is shown in Fig. 16.21. A 16-element microstrip
antenna, designed to operate at 10 GHz is shown in Fig. 1 6 . 2 1in
~ top view. The
central 100 R line is fed at its centre by a coaxial SMA connector. The other feed
lines have impedance of 200 R,which is also the calculated radiation impedance
of each radiating element at resonance. The TDR oscillogram of this antenna
was made with Tektronix equipment (7603 mainframe, S-52 pulse-generator
head and S-6 sampling head) having a typical rise time of 30 ps. Three points
of interest are noted as A, B and c, both on the antenna layout and on the TDR
oscillogram. Point A is the connector area which is found to have a small
inductance. Point B is the first T-junction which has a small capacitance, and
point c is the second junction which is almost free of any parasitic reactance.
The time delays between A-B and B-c are equal and measured from the oscillo-
gram to be
AT = (1.1 f 0.1)200 picoseconds
This time delay is equivalent to a distance of
A&
! = c(A~/2) = 3.3 $. 0.3 centimetres
 980 Special measurement techniques for printed antennas
The physical distances between the noted points, A1 = 2.4 cm, and the electrical
distances AIJE~~z 3.4 cm, are in good agreement with the TDR results. The
dominant slope, which comes after c, is the result of the capacitive nature of the
radiating elements.
Fig. 16.21 A TDR oscillogram of a 16-element rnicroslrip array (Courtesy: Dr. P. Pertmutter,
Bellcore NJ, USA)
Points A. B and C are shown both on the antenna layout and on the TDR trace.
Special measurement techniques for printed antennas 981
An indirect TDR can also be made by measuring the frequency response and
using a fast fourier transform (FFT) to convert the data into the time domain.
Such measurements made by wideband network analysers have gained much
interest in recent years [82, 831, and several commercial products like the
HEWLETT PACKARD 8510 network analyzer 1841or the WILTRON 360 [82]
are adequate for these applications.
16.5 Near-field probing
Near-field measurements are usually considered as a special case of pattern
measurements, which are done in the vicinity of the radiating aperture. They are
performed in cases where the far-field pattern measurement is very difficult,
expensive or even impossible, e.g. high-gain or low-frequency antennas, anten-
nas mounted on large bodies, or if there is a need in secure and controlled test
environments or with all-weather capability. The near field is detected by a
scanning probe over a chosen surface (planar, cylindrical or spherical), and the
measured results are processed by analytical or numerical methods to give the
required far field. General reviews on near-field techniques, including scanning
procedures, data analysis and error corrections - especially errors caused by
the probes - can be found in References, 2 and 85-89. Near-field scanning is
also required in cases where the knowledge of the exact current or field distribu-
tion on the aperture is important; e.g. in arrays with tapered aperture distribu-
tion, in phased arrays where the relative phases of excitation should be charac-
terised, or in R F applicators which operate on biological tissues at close distan-
ces. In most of these cases, precise quantitative information is needed, which
means that the measurement of the near field must be very accurate and the
effects of the probe on the measured fields must be taken into account. There
are, however, a variety of applications in which a qualitative mapping of
radiating aperture is very useful. One such application is the near-field probing
of microstrip antennas. This near-field probing, which is done at distances of
millimetres from the printed board, can serve as a diagnostic tool in the design
and the production stages of the antenna. It can be used to find local defects,
to reveal asymmetries in the feed networks, to improve excitation details of the
radiating elements, and even to provide quantitative profiles of the aperture field
distribution.
The discussion in this Section is restricted to practical and immediate techni-
ques which can be used by any microstrip antenna designer. The most important
consideration for these applications is the choice of a miniature probe. The
probe should be sensitive enough to the tested fields on one hand, while having
minimal effect on the tested aperture. Furthermore, the probe should have high
spatial resolution. A second consideration is the use of efficient yet inexpensive
scanners. Both of these considerations are discussed here, several schemes are
presented and some experimental mappings are demonstrated.
 982 Special measurement techniques for printed antennas
The majority of the miniature probes can be divided into three types: (a)Jield
probes which are small antennas that are sensitive to electric or magnetic fields,
(b) intensity probes which are basically square-law detectors connected to
miniature dipoles, and (c) thermal probes which are flat layers of liquid crystals
or thermo-electric junctions that are sensitive to the R F power in their vicinity.
The specific choice between the three types depends mainly on the required
information. A comprehensive review of electric-field probes of the first and the
second types is given in Reference 90. A detailed analysis of miniature electric-
field probes with resistive tranmission lines is given in Reference 91. Examples
of a sub-millimetre probe and a three-dimensional probe are given in References
92 and 93, respectively.
Special measurement techniques for printed antennas 983
sensitive. Their major disadvantage is the undesired change in the phase of the
transmission line which connects them to the receiver, while the probes are
moving. These probes are usually sensitive with respect to the frequency. In
addition, special baluns (balanced to unbalanced transitions) are needed to
reduce their influence on the tested apertures. Four practical field probes are
shown in Fig. 16.22. The first one (a) is a very simple monopole which is created
microwave A

Schotky diode
( M A 402149)

a / Af3soREEFi \ n
whiskers I
b

'*guided or
SEMI- RIGID free-air beam

@
CABLE
b electro-optical
modulator

I
photo-detector

electronic
detector

Fig. 16.22 Practical field probes for near-field mapping (Sources: References 2, 85,95,96)
a A shon monopole with vertical polarisation [95]
b A split-coaxial balun with horizontal polarisation [85, 961
c A small loop for magnetic-field probing (perpendicular to the plane of the paper)
d A small loop with balun [2]
The field probes are generally used when both the amplitude and the phase
of the tested aperture are needed. They are small in size and their sensitivity
depends on the quality of the receiver in use. Field probes are polarisation
Fig. 16.23 Electric-field probes with square-law detectors (Sources: References 90,98)
a A diode mounted in a plastic holder [98]. The dipole length is 3 mm.
b A printed version [SO] where the dipole length is 1.5 mm
c An electric probe coupled to optical modulator [go]
by extending the inner conductor of a coaxial line. This sensitive and reliable
probe has been used for precise measurements of printed-line parameters [94,
951. A second field probe (b) is a combined balun, made of a split coaxial section
and an absorber [85,96], both intended to reduce its effect on the tested aperture
to a minimum. A simple loop for magnetic-field detection is shown in Fig.
16.22c, and another balun arrangement for the magnetic field [2] is shown in
16.22 (4.
 984 Special measurement techniques for printed antennas
The second type - miniature square-law detectors - are very attractive for
automatic scanning over large surfaces and for diagnostic purposes. Like the
field probes they are polarisation sensitive, but, unlike the field probes, mini-
ature square-law detectors cannot measure phase. These detectors have a weak
effect on the field at the antenna aperture and perform best when they are
connected to the recording system by highly resistive transmission lines. Three
inlensiry probes used to measure the electric field with square-law detectors are
shown in Fig. 16.23. A first example (a) is a sensitive diode mounted in a plastic
holder and connected to a coaxial amplifier through highly resistive coatings [97,
981. The mean size of such a probe can be between 1 and 4 mm and its low
sensitivity requires high-quality amplifiers for the DC output. Fig. 16.236
presents a printed version of the intensity probe [90] and Fig. 1 6 . 2 3 illustrates
~
a novel concept of an electric probe that modulates a laser beam [90]. Any R F
voltage developed by the dipole causes a direct, instantaneous change in the
amplitude of the beam of light passing through the modulator. This probe
ensures high isolation of the probe's antenna and a very quick response which
can be used for phase measurements.
antenna transparent liquid
under test\ cover \ crystal paint
plane
Fig. 16.24 Liquid-crystal diagnostics of a microstrip radiator (Source: Reference 100)
The third group - thermal detectors - are the least sensitive and least
accurate, but they can be used in the most direct and simple way. A thin layer
of liquid crystal, placed on the antenna as shown in Fig. 16.24, can give
immediately a good visual impression of the standing waves existing in the feed
network, the relative magnitudes of the element excitations and the field profile
on each radiator. Further details and examples can be found in References
99-102.
The second consideration in preparing a near-field system is the scanning
instrumentation. Although the liquid-crystal diagnostic is extremely simple
since no scanning is needed, it seems that the requirements of high spatial
Special measurement techniques for printed antennas 985
resolution, and the possibility of extracting quantitative information, make the
scanning probes more attractive for near-field measurements. What specific
applications can be done with scanning probes? The first is checking the para-
meters of a printed transmission line, especially the effective dielectric constant.
Several examples can be found in References 94, 95 and 103-105. Of special
interest is the description given by Ladbrooke [I041 of how a microstrip stand-
ing-wave indicator can be used to study the line parameters and the effects of
discontinuities. Such probing can be most useful for unconventional printed
lines like the suspended air line shown in Fig. 16.25 or dielectric lines [94], where
analytical expressions or design curves are not available.
to recorder
arm
ground
Fig. 16.25 Near-fieldprobing along a printedline can be used for the measurement of the line
parameters
A second application for scanning probing is to locate defects and asymme-
tries in printed feed networks. Any mechanical scanning system can be used to
move the probe in a planar, cylindrical or on any desired surface. An example
of a low-cost mechanical system, based on two general-purpose XY recorders is
shown in Fig. 16.26 [98]. Here, one of the recorders serves as a scanner ('mas-
ter'). Another recorder, referred to as the 'slave', receives the X and Y signals
from the 'master', and, in addition, its Y input also receives the detected output
from the probe. In this Figure a lock-in amplifier in used in order to improve
the sensitivity. Another example of standard laboratory equipment used for this
purpose appears in Reference 106. A modified version uses a moving-field probe
connected through a semi-rigid line to a high-quality detector. Such an arrange-
ment can give accurate mapping of the near field, since the detector is far away
from the aperture, as well as high sensitivity. For example, antennas with area
of 0.1 m2, fed by a source of 20 mW, can easily be detected with a conventional
detector (like the HP 8473) and XY recorders without any external amplifiers.
The phase distribution on the aperture is also often needed. For instance, in
beam-shifted arrays one is interested in the relative phase changes between the
elements. Also in circular-polarisation designs one is interested in checking if an
 986 Special measurement techniques for printed antennas Special measurement techniques for printed antennas 987
appropriate 90" phase shift is achieved. The scanning with a phase-sensitive with an aperture size of 47 mm. The deviations from the theoretical profile were
probe is quite complicated. The main difficulty is that it is necessary to convey about 0.5 dB in amplitude and 5" in phase.
the signal back to a phase-sensitive receiver from the mobile probe without Special phase mapping, which is done by a low-frequency modulation of the
.robe., is described in Reference 107. Small dipoles which are coupled to optical
-
signal
generator Pin
switch antenna

-------
mech.
modulators are expected to play an important role in future developments. As
mentioned earlier, they are both sensitive and 'transparent' to the tested fields.
Another novel idea is the use of an active radiating probe [I081 to test printed
1master and even integrated circuits. This is done by very thin radiators which are placed
A recorder on the printed circuit, and the change in their radiation pattern is detected by
another receiving antenna.
switch
driver
(25 Hz)
lock-in
amplifier
ref in
slave
recorder

Fig. 16.26 Block diagram of a scanning system for near- fieldprobing (Source: Reference 98)
The probe is a square law detector connected to a small dipole

n recorder 2
(plotter)

recorder 1
(scanner)
amplitudelphase
output

probe Fig. 16.28 Near-field mapping of a four-element array at 10.1 GHz

a in original state
b After two cuts were made near the left elements

We conclude this Section with several demonstrations of near-field mappings

a
RF cable
Fig. 16.27 Block diagram of amplitude andphase mapping by a network analyser. The probe
is a small antenna
changing the electrical path length of the transmission line. A possible solution
is to keep the probe in one place and move the tested antenna instead. The block
diagram of Fig. 16.27 shows how both amplitude and phase information can be
achieved using a network analyser (HP 8410) and two recorders. The detector
in use here is the split coaxial line and the R F cables are Gore flexible cables.
This scheme has been checked on the aperture of a standard horn at 10 GHz,
of microstrip antennas [98]. All these mappings were done by miniature electric
probes (Fig. 16.226 or 16.23 a) and two XY recorders, as described in Fig. 16.26.
Fig. 16.28 demonstrates the influence of a defect on the near field of a four-
element array. Fig. 1 6 . 2 8 is
~ the layout of the antenna, Fig. 16.286 is an original
field pattern, which was taken at 10.1 GHz at a distance of 3 mm from the
antenna, and Fig. 16.28~shows the near-field pattern after two cuts were made
near the two left elements. Such a diagnostic test can reveal local defects even
in cases where ohmic contacts do exist. A second example, shown in Fig. 16.29,
shows the layout of a 16-element array, built at 10.6 GHz (Fig. 16.29a), a
near-field pattern which was produced at distance of 1 mm from the antenna
(Fig. 16.29b), and another pattern which was made at distance of 30 mm. It can
be seen here that the asymmetry between the left and the right sides of the
 988 Special measurement techniques for printed antennas
------'-
Fig. 16.29 Near-field mapping of a 76-element array at 10.6 GHz
a At distance of 1 mm
b At distance of 30 mm from the aperture
Fig. 16.30 Near-field mapping of a four-element array with a corporate feed network, at
6.3 GHz
The scanning was performed at a distance of 3 mm from the antenna.
989
Special measurement techniques for printed antennas
antenna almost disappears at a distance of one wavelength, which is much closer
than the far-field region. The local asymmetries in this specific antenna are
caused by different excitation angles of the elements. In corporate-feed designs,
all the elements are fed from the same angle and the local asymmetries are
expected to be reduced. The third example, shown in Fig. 16.30, presents the
near-field pattern of a four-element array at 6.3 GHz, with corporate-feed
network. In this case there are standing waves in the central feed line that have
some influence on the elements. Such near-field mapping can be used in this case
to check how relative translations of the central line change the antenna excita-
tion.
Fig. 16.31 Near-field mapping of a 16-element array of double-sided printed dipoles at
4.7 GHz
The dark area in the antenna layout is an upper print and the dashed area is a lower
print. The two-sided board is placed above a metallic ground plane. Two or-
thogonal mappings are presented
Fig. 16.31 shows scanning results on the aperture of a 16-element array of
printed dipoles at 4.7 GHz. The array is printed on both sides of a thin dielectric
substrte. The black area is the upper print and the dashed area is the lower print.
The printed substrate is placed above a metallic ground plane with a foam
spacer of 0.15 A, thickness in between. The two scans in Fig. 16.31 were carried
out in the same polarisation of the probe. The standing waves in the feed
network and some local asymmetries are shown in the two orthogonal views.
 990 Special measurement techniques for printed antennas
These standing waves, recorded at a distance of 2 mm from the printed board
(0.03 Lo), are related to the overall feed radiation, although a quantitative
interpretation is difficult. The last example, shown in Fig. 16.32, is the near-field
mapping of a single disc, fed at its centre by a coaxial connector, at 7.1 GHz.
I-\--
ground
\+plane
/'
/ / " - - ' , \
\ antenna, but it is quite difficult to identify the contribution of each loss. Experi-
center-fed
I disk in
I top view
/ placed in front of an ideal antenna; thus they reduce directly the overall noise
/ figure of the system. On the other hand, radiation losses change the radiation
/
Fig. 16.32 Near-field mapping at 7.1 GHz of a centre-fed disc, with foam spacer as a
dielectric substrate. The disc is placed 3 mm above the ground plane and the
scanning was done 6 mrn above the ground plane. The mode number is n = 0
fJbfk,rJ = 0)
The size of the probe (length of the dipole) is 4 mm. The disc is mounted 3 mm
above a ground plane with a foam spacer in between, and the scanning was done
6 mm above the ground. This map of the element excitation shows the existence
of the first symmetrical mode on the patch. Important information can be
gained from this map as to the decay of the field along the ground plane. In this
case the fields are bounded within a circle with a diametre of 70mm. Near-field
probing can be thus used for combined theoretical and empirical modelling of
printed radiators, testing the influence of a finite ground plane or testing the
contribution of the vertical currents in the feed.
Special measurement techniques for printed antennas 997
16.6 Efficiency measurement
Printed array antennas show low efficiency owing to inherent dielectric and
ohmic losses in their feeding network and owing to the excitation of surface
waves in the substrate. The efficiency limitation is most severe in large arrays
where the feed network is long and complicated, and in arrays operated at high
frequencies. In addition to the dissipation losses, there are usually also 'radia-
tion losses' caused by unwanted radiation from the feed lines or from the
connectors. All these mechanisms reduce the available gain of the printed
mental characterisation of the different losses is frequently desired. One major
motivation is the fact that in low-noise receivers there is a difference between
dissipation losses and radiation losses. Dissipation losses act as an attenuator
pattern, but they do not affect the noise figure unless the sidelobes are directed
towards a noisy environment. A similar separation between losses is also re-
quired when a plastic cover (radome) is put in front of the antenna.
The ohmic and dielectric losses of microstrip lines are well known in the
literature both from theoretical and experimental aspects. The losses of any
specific materials or feed lines can be measured independently by standard
methods mentioned in Sections 16.2 and 16.4. We would like to concentrate here
on the overall power efficiency of the printed antenna, and, in particular, to
describe a simple technique for its measurement.
The power efficiency of an antenna is the ratio of the total power radiated by
the antenna to the input power accepted by the antenna at its terminals. The
efficiency can also be expressed in terms of the gain and the directivity
gain
directivity
Here, the radiated power PRis given by integrating the radiation intensity P(0,
4) over the far-field surface s completely enclosing the antenna. Thus an
accurate evaluation of requires the measurement of P(B, 4) for a sufficiently
large number of angles, and a numerical integration of the results. This
is often slow and complciated and requires a well-designed antenna
range [109-1101. A simpler, but less accurate method, requires the measurement
of the two principal E- and H-plane patterns of the tested antenna, and the use
of the following approximation for the directivity D (for a uniform illuminated
aperture) [Reference 3, p. 331
 992 Special measurement techniques for printed antennas Special measurement techniques for printed antennas 993
where 8, and 8, are, respectively, the half-power beamwidths in degrees. Com- Table 16.1 List of components used in radiometric systems
bining this directivity with the measurement of gain gives the efficiency (eqn.
10 GHz 20 GHz 35 GHz
16.35).
A quick method for measuring the efficiency of an antenna, which directly RF Miteq none none
integrates the antenna pattern, is described in Reference I I 1. In this method, the Amplifier AFD4040 130
antenna to be tested is at the input of a radiometer that is directed towards NF = 4dB
extended 'warm' and 'cold' targets. A convenient cold target is a clear sky, and Mixer Anaren Alpha Honeywell
a warm one is an extended absorber, usually at room temperature. The use of 7G0118 ATD9606K09 F35UP
extended sources automatically integrates over the whole radiation pattern. The NF = 5dB NF = 5dB
relationship between the desired efficiency q and the measured voltages at the
radiometer output is given by Local Varian Varian Varian
oscillator VSX900 1M G VSK9004FS VSA90 10JC
IF Avantek Avantek Avantek
amplifier SD30632M SD30632M SD30632M
where E is the ratio of the voltage when the antenna is directed towards a warm
target to the voltage when the antenna is directed towards a cold target. 6 is the LP filter 200 MHz 200 MHz 200 MHz
same measured ratio for a high-efficiency antenna such as a horn, for which we Reference FXR self-made TRG
assume q = 1. A block diagram of the system is shown in Fig. 16.33. The R F horn X638A A861
amplifier is the most important component because it has the dominant effect
on the overall noise figure of the radiometer. In many radiometers an integrated
mixer/amplifier is used as an input stage, and in these cases its noise figure is
dominant. The validity and the accuracy of this measurement technique were
tested by connecting a waveguide calibrated attenuator after the horn antenna,
and measuring the radiometer output. For example, a test set in the Ku-band
was built with a low-noise amplifier Miteq AMF-3D-1218 with 20 dB gain and
a 5.7 dB noise figure. The I F bandwidth of 200 MHz was determined by a Table 16.2 Measured and calculated results (in dB) of the losses in micro-
low-pass filter that was introduced between the I F amplifier and the detector. strip arrays
The difference between the attenuator readings and the losses computed from Number of elements
the measured radiometric results was less than 0.3 dB.
16 64 256 1024
reference 10 GHz
horn
D(ca1culated) - Gtmeasured) 0.7 1.O 1.5 -
Calculated loss 0.8 1.1 1.6 2.3
I mixer detector Measured loss 0.5 1.2 1.7 -
1 RF
amplifier
20 GHz
tested D-G 0.3 1.5 2.0 3.2
antenna Calculated loss 0.9 0.2 2.1 3.0
Measured loss 0.7 1.3 2.2 2.7
Fig. 16.33 Block diagram of a radiometer for antenna-efficiency measurement (Source:
Reference 1 7 7)
35 GHz
A set of 16-, 64-, 256- and 1024-element microstrip antennas [I 121 in frequen- D-G 1.1 1 .5 2.7 4.8
Calculated loss 1.1 1.5 2.5 4.1
cies of 10, 20 and 35 GHz were tested by this technique, and their measured
efficiencies were found to be in a good agreement with theoretical estimates. A Measured loss 1.4 1.8 2.8 4.0
 994 Special measurement techniques for printed antennas
list of the components used in these measurements is given in Table 16.1 and
measured results are given in Table 16.2.
A few remarks should be noted concerning the accuracy and the repeatability
achieved by this method:
(a) The expression used for the efficiency is valid under the assumption that the
antenna is perfectly matched. However, standing waves in the measurement
system may cause significant errors in the interpretation of the measured data.
It is therefore recommended the measurements be repeated at several frequen-
cies near the operating frequency of the antenna.
(b) The transition from the printed antenna into a waveguide input of a mixer
or an amplifier is also a source of systematic errors. These errors can be reduced
by performing an additional measurement of 6 for the reference horn, while
connecting it through two waveguide/coaxial-line adapters, and averaging the
results with the original value of 6.
(c) Special care should be taken to ensure that the extended cold target will
cover the upper half-sphere around the tested antennas. Any surface-wave
launching in near-endfire angles can change the entire effective temperature of
the antenna; thus the area in the vicinity of the ground plane should be clear of
obstacles.
(4 The radiometric measurement is not sensitive to the antenna polarisation;
and hence it can handle both linear and circular polarisations with the same
accuracy.
16.7 Concluding remarks
The experimental techniques described in this Chapter can be most helpful in the
design and the manufacturing processes of microstrip and other printed anten-
nas. They are suggested as supplements to the conventional far-field measure-
ments and network analysis. The frequency range discussed here covers micro-
waves up to 40 GHz, and in some cases perhaps a little higher. It should be
pointed out that all these measurements are limited by calibration errors and by
wrong interpretation, and not by signal/noise or dynamic range considerations.
This is, of course, due to the passive nature of the antennas and their feed
network.
The topics covered in this Chapter include: the substrate properties, the
connector, the printed network and two specific techniques of near-field probing
and radiometric measurement. The properties of the substrate are usually given
by the manufacturer. However, in some cases one wishes to characterise the
substrate by oneself. Examples are: combination of different substrates, the use
of substrates at higher frequencies than reported, or when an exact characterisa-
tion is needed as a function of a parameter such as the temperature. The
accuracy achieved for the dielectric constant in the described techniques is of the
order of 1%. The measurement of the loss tangent is difficult for common
Special measurement techniques for printed antennas 995
low-loss materials. This information can be achieved by a stripline made of the
materials under test.
The characterisation of connectors is very important since the manufacturers
do not give specific data about transitions into printed lines. Two typical
examples in which the connector measurement is recommended are: thin
(i 0.4 mm) and thick ( > 3.2 mm) substrates, and at higher frequencies than
recommended by the manufacturer. The characterisation of connectors is sens-
itive and accurate while using the interferometric methods. The reactance of the
connector may even by used for broadband matching of antennas.
Printed lines and structures can be tested accurately by the resonant method
given in this Chapter. Of special interest is the TDR technique, which can
identify local defects with accuracy of several millimetres.
Near-field probing is a qualitative tool which can help in understanding the
behaviour of printed antennas and finding local defects. The main limitation is
the interpretation of these mappings. The radiometric technique is useful,
especially for large printed arrays with high losses. One can expect to find the
dissipation losses of medium- and high-gain arrays with accuracies better than
1 dB.
Future developments in the instrumentation for printed antennas will prob-
ably include the following: better substrates with accurate data sheets; novel
calibration sets for accurate measurement of connectors and printed structures;
new feed lines and new feed techniques for the radiators. Another promising
area is the combination of radiators with active devices and MMIC chips. Such
combinations will also lead to better measurement techniques and instrumenta-
tion. In the future there will be also combinations of printed antennas with
electro-optical devices and optical fibres. Such combinations will, of course,
require special measurement techniques. Finally, one may expect in the future
to find some miniature probes, fed by optical means, for near-field probing.
16.8 References
1 HOLLIS, J. S., LYON, T. J., and CLAYTON, L. (Eds.): 'Microwave antenna measurements'
(Scientific Atlanta, 1970)
2 APPEL-HANSEN, J., DYSON, J. D., GILLESPIE, E. S. and HICKMAN, T. G.: 'Antenna
measurements' (in RUDGE, A. W., MILNE, K., OLVER, A. D., and KNIGHT, P. (Eds):
'The Handbook of antenna design' (Peter Peregrinus, 1982) Chap 8
3 BALANIS, C. A,: 'Antenna theory - analysis and design' (Harper and Row, NY, 1982)
Chap. 15.
4 BLAKE, L. V.: 'Antennas' (Artech House, Mass., 1984) chap 8. chap 9-9
5 GUPTA, K. C., GARG, R., and BAHL, I. J.: 'Microstrip lines and slotlines' (Artech House,
Mass., 1979) pp. 28-38, 184-193, 331-336
6 EDWARDS, T. C.: 'Foundations for microstrip circuit design' (John Wiley, 1981) pp. 172-207
7 LAVERGHETTA, T. S.: 'Microwave materials and fabrication techniques' (Artech House,
Mass., 1984) p. 56
8 NAPOLI, L. S. and HUGHES J. J.: 'A simple technique for the accurate determination of the
microwave dielectric constant for microwave integrated circuit substrates', IEEE Trans., 1971,
MTT-19,pp. 664-665
 996 Special measurement techniques for printed antennas Special measurement techniques for printed antennas 997
9 HOWELL, J. Q.: 'Quick accurate method to measure the dielectric constant of microwave
microstrip connector suitable for evaluation of microstrip Zports', Electron. Lett., 1976, 12,
integrated circuit substrates', IEEE Trans., 1973, MTT-21, pp. 142-143
pp. 430-43 1
10 LADBROOKE, P. H., POTOK, M. H. N., and ENGLAND, E. H.: 'Coupling errors in cavity
1 36 GOURLEY, S. E. and CHAPMAN, A. G.: 'Broadband characterization ofcoaxial to micro-
resonance measurements on MIC dielectrics', IEEE Trans., 1973, MTT-21, pp. 560-562
strip transitions'. Proc. 12th European Microwave Conference, 1982, pp. 622-627
11 OWENS, R. P., AITKEN, J. E. and EDWARDS, T. C.: 'Quasi static characteristics of
microstrip on an anisotropic sapphire substrate', IEEE Trans., 1976, MTT-24, pp. 499-505 i 37 RICKARD, D. C.: 'Thick film mic components in the range 10-20 GHz. Proc. 6th European
Microwave Conference, 1976, pp. 687-691
I2 RICHARDS, W. F., LO, Y. T. and BREWER, J.: 'A simple experimental method for
38 FINLAY, H. J., HOPKINS, L. G. T. and OZAMIZ, J. M.: 'Design and applications of
separatmg loss parameters of a microstrip antenna', IEEE Trans., 1981, AP-29, pp. 150-151
13 ITOH, T.: 'A new method for measuring properties of dielectric materials using a microstrip
cavity', IEEE Trans., 1974, MTT-22 pp. 572-576
14 GERHARD, A. R.: 'Measuring dielectric constant of substrates for microstrip applications',
i
I
precision microstrip multioctave attenuators and loads'. Proc. 6th European Microwave
Conference, 1976, pp. 692696
39 KDI-Pyrofilm: RF resistive components catalog, PSA 1/80, 1980
40 EMI-Varian: Microstrip circuits data sheets, mc2m/10/70
lEEE Trans., 1976, MTT-24, pp. .. 485437 41 EISENHART, R. L.: 'A better microstrip connector'. IEEE Int. Symp. Digest MTT-S, 1978,
IS BUSSEY, H. E.: 'Measurement of RF properties of materials - survey', proc. IEEE, 1967,
pp. 318-320
55, pp. 1046-1053
16 LYNCH, A. C.: 'Precise measurements on dielectric and magnetic materials', IEEE Trans.,
1 42 LEVINE, E., and TREVES, D.: 'Test technique improves coax to microstrip transitions',
Microwaves and RF, July 1986, pp. 99-102
1974, IM-23, pp. 425-431 i 43 GUPTA, C. D., and TOMAR, R. S.: 'Resonance method of measurement of input impedance
17 KOBAYASHI, Y. and KATOH, M.: 'Microwave measurement of dielectric properties of low
of any broadwall launched discontinuity in microstrip transmission lines', IEEE Trans., 1986,
loss materials by the dielectric rod resonator method', IEEE Trans., 1985, MTT-33, pp.
IM-35, pp. 126-129
586-592
44 SEALECTRO Co.: SMA catalog, SMA-9, 1980, pp. 25-26
18 AFSAR, M. N.: 'Dielectric measurements of millimeter wave materials', IEEE Trans., 1984,
45 WILTRON Co.: K Connector and semirigid cable, data sheet, 35K, Oct. 1984; K l I0 Series
MTT-32, pp. 1598-1 609
sliding contacts data sheet, DS 110-1, April 1985
19 PUES, H. F. and VAN D E CAPELLE, A. R.: 'Wideband impedance matched microstrip
46 HEUVEN, VAN 1. H. C.: 'A new integrated waveguide microstrip transition', IEEE Trans.,
resonator antenna'. Proc. 2nd ICAP, 1981, pp. 402-405 1976, MlT-24, pp. 144-147
20 GRIFFIN, J. M. and FORREST, J. R.: 'Broadband circular disc microstrip antenna', Elec-
47 SCHIEK, B. and KOHLER, J.: 'An improved microstrip to microslot transition', IEEE
tron. Lett., 1982, 18, pp. 266-269 Trans., 1976, MTT-24, pp. 231-233
21 FONG, K. S., PUES, H. F., and WITHERS, M. J.: 'Wideband multilayer coaxial fed
48 SHEPHERD, P. R. and POLLARD, R. D.: 'Direct calibration and measurement of micro-
microstrip antenna element', Electron. Lett., 1985, 21, pp. 497-499 strip structures on Gallium Arsenide'. IEEE Int. Symp. Digest MTT-S, 1986, pp. 629-632 ,
22 CHEW, W. C. and KONG, J. A.: 'Analysis of a circular disk antenna with a thick substrate',
49 CARLTON, D. E., GLEASON, K. R. and STRID, E. W.: 'Microwave wafer probing',
IEEE Trans., 1981, AP-29, pp. 68-76
Microwave J . Jan. 1985, pp. 121-129
23 YANO, S. and ISHIMARU, A,: 'A theoretical study of the input impedance of a circular
50 HEWLETT PACKARD COMPANY: 'On wafer measurements using the HP 8510 network
microstrip disk', IEEE Trans., 1981, AP-29, pp. 77-83 I
analyzer and Cascade Microtech wafer probes'. Product note 8510-6, May 1986
24 HENDERSON, A. and JAMES, J. R.: 'Design of microstrip antenna feeds. Part I: Estimation
51 TROUGHTON. P.: 'Measurement techniques in microstrip', Electron. Lett., 1969, 5, pp.
of radiation loss and design implications', IEE Proc., 1981, 128H. pp. 19-25
25-26
25 HENDERSON, A. and JAMES, 1. R.: 'Estimation of radiation from microstrip antenna feed
52 WOLFF, I. and KNOPPIK, N.: 'Microstrip ring resonator and dispersion measurement on
transition', Radio Sci., 1981, 16, pp. 1119-1 123
microstrip lines', Electron. Lett., 1971, 7 , pp.779-781
26 PINHAS, S., and SHTRIKMAN, s.: 'Vertical currents in microstrip antennas', IEEE Trans.,
53 OWENS, R. P.: 'Curvature effect in microstrip ring resonators', Electron. Lett., 1976, 12, pp.
1987, AP-35, pp. 1285-1289
356-357
27 WIGHT, 1. S., JAIN, 0. P., CHUDABIAK. W. J., and MAKIAS, V.: 'Equivalent circuits of
54 DEUTCH, J. and JUNG, H. J.: 'Measurement of the effective dielectricconstant of microstrip
microstrip impedance discontinuities and launchers*, IEEE Trans., 1974, MTT-22, pp. 48-52
lines in the frequency range from 2 GHz to 12 GHz', Nachrichtenlech Z., 1970,12, pp. 620-624
28 AJOSE, S. 0 . and MATHEWS, N. A,: 'Equivalent circuit of coaxial to microstrip connector
55 EDWARDS, T. C. and OWENS, R. P.: '2-18 GHz dispersion measurements on 10-100 ohm
over the 8-12 GHz range, Electron. Lett., 1977, 13, pp. 465-466
microstrip lines on sapphire', IEEE Trans., 1976, MTT-a, pp. 506-513
29. HALL, P. S. and JAMES, J. R. 'Design of microstrip antenna feeds. Part 2: Design and
56 ROMANOFSKY, R. R., BHASIN, K. B., PONCHAK, G. E., DOWNEY, A. N. and
performance limitations of triplate feeds' IEE Proc., 1981, 128H, pp. 26-34
CONNOLLY, D. J.: 'An experimental investigation of microstrip properties on soft substrates
30 MAJEWSKI, M. L., ROSE, R. W. and S C O T , J. S.: 'Modeling and characterization of
from 2 to 40 GHz' IEEE Int. Symp. digest MTT-S, 1985, pp. 675-678
microstrip to coaxial transitions', IEEE Trans., 1981, MTT-29, pp. 799-805
57 RICHINGS, J. G.: 'An accurate experimental method for determ~ningthe important par-
31 PUES, H. F. and VAN DE CAPELLE, A. R.: 'Computer aided experimental charaterization
ameters of microstrip transmission lines', Marconi Rev. 1974, pp. 209-216
of microstrip to coaxial transitions'. Proc. 14th European Microwave Conference, 1984, pp.
58 BELOHOUBEK, E. and DENLINGER, E.: 'Loss considerations for microstrip resonators',
137-141
IEEE Trans., 1975, MTT-23, pp. 522-526
32 EMC Technology: Microwave components catalog 882-15, 1982, p. 38
59 PUCEL, R. A,, MASSE, D. J. and HARTWIG, C. P.: 'Losses in microstrip', IEEE Trans.,
33 MIA-COM Omni Spectra Inc: Microwave coaxial connectors catalog, 1983, p. I I
1968, MTT-16, pp. 342-350, 1064
34 ENGLAND, E. H.: 'A coaxial to microstrip transition', lEEE Trans., 1976, MTT-24, pp. 60 DENLINGER, E. J.: 'Losses in microstrip lines', IEEE Trans., 1980, MTT-28, pp. 513-522
4748 61 JAMES. J. R. and HENDERSON, A,: 'Planar millimeter wave antenna arrays in Infrared and
35 AJOSE, S. O., MATHEWS, N. A. and AITCHISON, C. S.:'Characterisation of coaxial to millimeter waves'; Vol. 14 (Academic Press, 1985) pp. 189-247
 998 Special measurement techniques for printed antennas
62 TROUGHTON, P.: 'High Q factor resonators in microstrip', Electron. Lett., 1968, 4, pp.
520-522
63 AITKEN, J. E.: 'Swept frequency microwave Q factor measurement', Proc. IEE, 1976, 123,
pp.855-862
64 KAJFEZ, D. and HWAN, E.: 'Q factor measurement with network analyzer', IEEE Trans.,
1984, MTT-32, pp. 666-669
65 LADBROOKE, P. H.: 'Some effects of field perturbation upon cavity resonance and disper-
sion measurements on MIC dielectrics', IEEE Trans., 1977, MTT-25, pp. 892-903
66 GETSINGER, W. J., 'Measurement and modeling of the apparent characteristic impedance
of microstrip', IEEE Trans., 1983, M'IT-31, pp. 624-632
67 SHEPHERD, P. R., and DALY, P.: 'Modeling and measurement of microstrip transmission-
line structures', IEEE Trans., 1985, MTT-33, pp. 1501-1506
68 STEPHENSON, I. M. and EASTER, B.: 'Resonant techniques for establishing the equivalent
circuits of small discontinuities in microstrip', Electron. Lett., 1971, 7, pp. 582-584
69 EASTER, B.: 'The equivalent circuit of some microstrip discontinuities'. IEEE Trans., 1975,
MTT-23, pp. 655-660
70 MENZEL, W. and WOLFF, I.: 'A method for calculating the frequency dependent properties
of microstrip discontinuities', IEEE Trans., 1977, MIT-25, pp. 107-1 12
71 EASTER, B., GOPINATH, A. and STEPHENSON, I. M.: 'Theoretical and experimental
methods for evaluating discontinuities in microstrip', Radio & Electron. Eng., 1978, 48, pp.
73-84
72 RIZZOLI, V.: 'A general approach to the resonance measurement of asymmetric microstrip
discontinuities'. IEEE Int. Symp. Digest MTT-S, 1980, pp.422424
73 GARG, R. and BAHL, I. J.: 'Microstrip discontinuities', Int. J. Electron., 1978,45, pp. 81-87
74 ENGEN, G. F.: 'The six-port reflectometer: an alternative network analyzer', IEEE Trans.,
..
1977. M'IT-25. DD. 1075-1080
75 CRONSON, H. M. and SUSMAN, L. A.: 'Six-port automatic netowrk analyzer' IEEE Trans.,
1977, MTT-25, -DD.
- 1086-1091
76 CULLEN, A. L.: 'The six-port and the microprocessor in microwave measurements'. Proc. 9th
European Microwave Conference, 1979, pp. 74-82
77 SOMLO, P. I. and HUNTER, J. D.: 'A six-port reflectometer and its complete characteriza-
tion by convenient calibration procedure' IEEE Trans., 1982, MTT-30, pp. 186-192
78 EL-DEEB, N. A,: 'The calibration and performance of a microstrip six-port reflectometer',
ZEEE Trans., 1983, MlT-31, pp. 509-514
79 ELLIOT, B. J.: 'High sensitivity picosecond time domain reflectometry', IEEE Trans., 1976,
IM-25, pp. 376-379
80 PILLER, U.: 'Time domain immittance measurements'. Proc. 4th European Microwave
Conference, 1974, pp. 61-65
81 CASPERS, F.: 'Precision time domain measurement system', Electron. Lett., 1980, 16, pp.
29-30
82 'Microprocessor based fault finder pinpoints transmission line faults within inches', Microwave
- 49-57
J., Nov. 1981. *DD.
83 'Real time measurements in a wideband network analyzer', Microwave J., Jan. 1984, pp.
138-145
84 HEWLETT PACKARD COMPANY: H P 8510 network analyzer, operating and service
manual, 1984, pp. 127-149
85 DYSON, J. D.: '~easurement of near fields of antennas and scatterers', IEEE Trans., 1973,
AP-21, pp. 445-460
86 JOHNSON, R. C., ECKER, H. A. and HOLLIS, J. S.: 'Determination of far-field antenna
patterns from near-field measurements', Proc. IEEE, 1973, 61, pp. 1668-1694
87 PARIS, D. T., LEACH, W. M. and JOY, E. B.: 'Basic theory ofprobe-compensated near field
measurements', IEEE Trans., 1978, AP-26, pp. 373-379
88 JORY, V. V., JOY, E. B. and LEACH, W. M.: 'Current antenna near field measurement
999
Special measurement techniques for printed antennas
research at the Georgia Institute of Technology'. Proc. 13th European Microwave Conference,
1983, pp. 823-828
89 YAGHJIAN, A. D.: 'An overview of near field antenna measurements', IEEE Trans., 1986,
AP-34, pp. 30-45
90 BASSEN, H. I. and SMITH, G. S.: 'Electric field probes - a review', IEEE Trans., 1983,
AP-31, pp. 710-718
91 SMITH, G. S.: 'Analysis of miniature electric field probes with resistive transmission lines',
IEEE Trans., 1981, MTT-29, pp. 1213-1224
92 BATCHMAN, T. E. and GIMPELSON, G.: 'An implantable electric field probe of sub-
millimeter dimensions', IEEE Trans., 1983, MTT-31, pp. 745-751
93 KANDA, M. and RIES, F. X.: 'Dipole based EM probe grabs complex fields', Microwaves
& RF, Jan. 1981, pp. 63-66
94 SOLBACH, K.: 'Electric probe measurements on dielectric image lines in the frequency range
of 26-90 GHz', IEEE Trans., 1978, MTT-26, pp. 755-758
95 DAHELE, J. S. and CULLEN, A. L.: 'Electric probe measurements on microstrip', IEEE
Trans., 1980, MTT-28, pp. 752-755
96 WEEKS, W. L.: 'Antenna engineering' (McGraw Hill, 1968) pp. 179-180
97 CHUNG, I., ANDREWS, C. L. and LIBELO, L. F.: 'Near field diffraction on the axes of
disks'. J. Opt. Soc. Am., 1977, 67, pp. 1561-1566
98 LEVINE, E., SHTRIKMAN, S. and TREVES, D.: 'Near field mapping of microstrip anten-
nas'. Proc. 12th European Microwave Conference, 1982, pp. 337-342
99 -KERNWEIS.
> >--- .. -- - . N. P. and McILVENNA, J. F.: 'Liquid crystal diagnostic techniques -An
-
antenna design aid', Microwave J., Oct. 1977, pp. 41-44
100 GIANNINI, F., MALTESE, P. and SORRENTINO, R.: 'Liquid crystal technique for field
- - .- ..... in microwave integrated circuitry' AIta Frequenza, 1977, XLVI pp. 170-178
detection
101 GIANNINI, F., MALTESE; P. and SORRENTINO, R.: 'Liquid crystals improved technique
for thermal field measurements', Applied Optics, 1979, 17, pp. 3048-3052
102 NEWHAM, P.: 'Monolithic patch array antenna for small missile applications'. Military
Microwaves Conf. MM-86, 1986, pp. 335-340
103 HASEGAWA, H., FURUKAWA, M. and YANAI, H.: 'Measurements on high frequency
transmission characteristics of metallization patterns in monolithic ICs', Electron & Commun.
in Japan, 1971.54-B, pp. 52-60
104 LADBROOKE, P. H.: 'A novel standing-wave indicator in microstrip', Radio & Electron.
Eng., 1974,44, pp. 273-280
105 HUBBELL. S. and ANGELAKOS, D. J.: 'A technique for measuring the effective dielectric
constant of microstrip line', IEEE Trans., 1983, MlT-31, pp. 687-688
106 GAJDA, G., STUCHLY, M. A. and STUCHLY, S. S.: 'Mapping of the near field pattern in
simulated biological tissues', Electron. Lett., 1979, 15, pp. 120-121
107 DAVIES, D. E. N. and VAKIL, S. M.: 'Field probe for measuring both amplitude and phase
of antenna radiation patterns', Electron. Lett., 1980, 16, pp. 873-875
108 HELIER, M. and BOLOMEY, J. C.: 'Test of monolithic microwave integrated circuits with
radiating probes', Proc. 13th European Microwave Conference, 1983, pp. 621-645
109 NEWMAN, E. H., BOHLEY, P. and WALTER, C. H.: 'Two methods for the measurement
of antenna efficiency', IEEE Trans., 1975, AP-23, pp. 457461
110 KUMMER, W. H. and GILLESPIE, E. S.: 'Antenna measurements-l978', Proc. IEEE, 1978,
66, pp. 483-507
111 ASHKENAZY, J., LEVINE, E. and TREVES, D.: 'Radiometric measurement of antenna
efficiency', Electron. Lett., 1985, 21, pp. 111-112
112 LEVINE, E., MALAMUD, G. and TREVES, D.: 'High gain modular microstrip antennas',
Proc. 16th European Microwave Conference, 1986, pp. 655-660
 Chapter 17

Computer-aided design of
microstrip and triplate
circuits
J-F. Zurcher and F.E. Gardiol

17.1 Introduction, definition of the structure

17.1.1 Outline
The basic purpose of this Chapter is to provide general background information
about circuits in microstrip and balanced stripline (triplate) technologies, that
are currently used to interconnect elements and realise antenna feed networks.
It will describe the general appearance of the circuits, the techniques utilised to
fabricate them and interconnect them, the materials most currently used and,
finally, the very powerful computer programs presently available to analyse,
design and actually draw the pattern and cut the masks required for the
manufacturing process. The Chapter will be completed with worked examples
and an extended Bibliography.

17.1.2 Microwaves
The field of microwaves extends over the frequency range 300 MHz -300 GHz
or, in terms of wavelengths, from 1 mm up to 1 m. The sizes of instruments
required to generate them and to measure them are thus of the same order of
magnitude as a wavelength. One cannot assume that circuits are much smaller
than a wavelength, as one generally does in circuit theory. One cannot either
assume that they are much larger, as is the case in optics. This means, in fact,
that the finite velocity of light must be taken into account.
The traditional applications of microwave antennas, in radar and communi-
cations, cover the frequency bands below 12-15 GHz, while heating and medi-
cal applications are restricted to a narrow band around 2450 MHz and, in
certain countries, 915 MHz [14]. Recently, frequency bands in the millimetre-
wave range (up to 100 GHz) are also being considered.

17.1.3 Transmission lines for microwaves

Metallic waveguides, i.e. hollow metallic tubes of rectangular, circular or ellipti-
 7002 Computer-aided design of microstrip and triplate circuits
'cal cross-sections, have generally been used in the past to carry microwave
signals, in particular in antenna feed systems. They are most often rigid and
bulky, difficult to interconnect with discrete components (diodes, transistors),
for which special fixtures must be designed and machined. They are presently
used only when specifically required, for high power levels or at very high
frequencies. In all other situations, they are replaced by a variety of planar lines
(Fig. 17.1).
Strlpllne Mlcrostrlp
Computer-aided design of microstrip and triplate circuits 7003
critical). A mask of the circuit is made, and the conductor outline $then
imprinted on a suitable photoresistive layer deposited on the printed-circuit
material. Metal is then either etched away from the exposed region, or deposited
onto it (Section 17.4). The main difference between low-frequency printed
circuits and microwave planar structures is that, at microwaves, all dimensions
are significant.
Microstrip antennas are also planar structures, similarly realised by the
photolithographic process, but adapted for their larger size (in particular for
arrays). When the antenna patches are fed by microstrip lines or striplines, one
obtains a fully integrated system, in which both the radiators and the circuit
elements are realised using the same technology. Solid-state components like
transistors can be inserted to realise active antennas, while whole monolithic
antenna structures are deposited on a semiconductor substrate. In contrast,
waveguide or fin-line feeds (at millimetre wavelengths) are better suited for use
with horn radiators.

17.1.4 Balanced stripline or triplate

A thin centre conductor, of width w, is located between two flat metal plates, or
Slotllne Coplanar llne ground planes (Fig. 17.2). The material between the two plates is a low-loss

Suspended stripline lnuerted stripllne

Dielectric

Flnllne Metal Fig. 17.2 Ideal homogeneous balanced stripline

Fig. 17.1 Planar transmission lines
Planar lines make use of the photolithographic technique developed to
realise electronic circuits, permitting miniaturisation and series production. A
sheet of insulating material (ceramic or plastic), the substrate, provides mech-
anical rigidity and permits the accurate positioning of components, while metal
strips desposited on the substrate provide the necessary connections (at low
frequencies, only a low-resistance path is required; i.e. strip dimensions are not
dielectric, which may be air (in which case some provision is needed for mech-
anical support). The structure is homogeneous; i.e. the electromagnetic fields
extend over one single propagation medium of uniform dielectric properties
- at least in theory. In practice, however, the centre conductor is deposited on
a dielectric substrate, and a dielectric plate of the same thickness is then placed
on top. Since the metal layer cannot be infinitely thin, an air-filled gap remains
between the two dielectric layers (Fig. 17.3). When very high performance is
required, this gap may in turn be filled with a special dielectric glue, of the same
permittivity as that of the two plates.
 7004 Computer-aided design of microstrip and triplate circuits
Basically, the balanced stripline is a homogeneous transmission line, whose
dominant mode is purely transverse electromagnetic (both the electric and the
magnetic field are perpendicular to the longitudinal direction of the trans-
mission line). While it is an open structure, since the ground planes do not
Fig. 17.3 Balanced stripline with air gap between the two dielectric layers
extend to infinity, the fields decay quite rapidly in the transverse direction and
there is practically no radiation.
17.1.5 Microstrip
A microstrip line may be considered to be one-half of a balanced stripline (Fig.
17.4), in which one of the ground planes and half of the dielectric have been
Fig. 17.4 Microstrip line
removed. This means that the line is inherently open and inhomogeneous, with
fields extending all the way to infinity over both the dielectric substrate and the
air. Radiation and surface waves (see Chapter 8) can to somt: extent be avoided,
by using thin substrates of high-permittivity material. The fields are then mostly
concentrated within the dielectric, up to a certain frequency limit. The velocity
Computer-aided design of microstrip and triplate circuits 7005
of propagation for a plane wave is different in the two media, so that the
dominant mode of the structure cannot be tranverse electromagnetic: longi-
tudinal field components are required to satisfy the boundary conditions on the
air-dielectric interface. The longitudinal components, however, remain much
smaller than the transverse ones and can be safely neglected in most pratical
situations. The dominant mode of the transmission line is then called quasi-
TEM.
17.1.6 Adjustments
'Classical' microwave designs, based on waveguide technology, always permit
final adjustments: one may always reduce a mismatch by placing an inductive
post or a dielectric plate in a waveguide. Frequency-sensitive waveguide com-
ponents such as filters generally require a 'final tuning' step to meet their
specifications. Printed structures, on the other hand, have little or no capability
for adjustments, because they result from a lengthy fabrication procedure. One
must first prepare a layout, analyse its theoretical response, optimise it to meet
the desired performance, draw the circuit's outline, cut the mask, reduce it
photographically, expose it, etch it, dry it, and then mount the finished circuit
and measure its actual performance (Section 17.8). If the circuit does not meet
expectations, the entire procedure must be carried out again.
The design of printed circuits should therefore be right the first time. Accurate
descriptions of the components must be available, resulting from a thorough
theoretical analysis. The performance predicted should coincide with the meas-
ured data (since Maxwell, electromagnetics is an exact science).
The realisation of feeds for microstrip antennas, however, is not as critical as
that of filters: microstrip or striplines provide fairly broadband operation,
whereas the frequency band of microstrip patches is notoriously narrow. As a
result, feeds may be easier to realise than the antenna itself.
A waveguide design is assembled by bolting components together. In micro-
wave printed-circuit technology, on the other hand, connectors introduce mis-
matches that can badly damage the performance of the system. These are
difficult to avoid or to compensate for. Wherever feasible, one should realise and
assemble all the components on the same substrate. In the case of microstrip
antennas, the feed system can be deposited on the same substrate, or a multiple-
layer system can be used.
17.1.7 Multiple inhomogeneity
In terms of the electromagnetic field distribution, printed structures (circuits or
antennas) are quite complex, since most of them possess three different types of
inhomogeneity:
(a) With the exception of the balanced stripline (Fig. 17.2), the fields on printed
microwave structures extend over an inhomogeneous region, formed partly of
dielectric (one or several layers) and partly of air. Waves propagating along the
structure cannot be transverse electromagnetic, and therefore exhibit dispersion
(fortunately, this effect is small for microstrip) [14].
 1006 Computer-aided design of microstrip and triplate circuits Computer-aided design of microstrip and triplate circuits 1007

(b) A metal layer extends only partially across the structure (center or upper The longitudinal dependence of the fields on the line is expressed in terms of
conductor): the boundary conditions for the fields are not the same at all points the line voltage U ( z )and of the line current I(z). Complex phasors simplify the
on a plane, e.g. the air-dielectric interface of inhomogeneous structures. notation when the time dependence is sinusoidal. The actual voltage and current
(c) The whole structure has finite transverse dimensions. Circuits are enclosed
in a box, while antennas are open. Radiated waves and surface waves on the
air-dielectric substrate bounce back and forth, scattered by the edges, pro-
ducing spurious coupling between elements.
An accurate analysis of the electromagnetic fields on printed structures becomes
almost prohibitively difficult, owing to the presence of inhomogeneities. Circuits
in stripline and microstrip can, however, be designed quite satisfactorily using
electrostatic and quasi-static approximations, which are quite adequate for most
practical applications (radiation and surface waves can be neglected at suf-
ficiently low frequencies).

17.1.8 Measurement problems

In all technical developments, measurements are used to check the validity of
theoretical derivations and the accuracy of numerical calculations. When
analysing stripline or microstrip structures, the measurements are never made
on the structure itself. Instruments are always connected to the coaxial line or
waveguide, so that the structure is 'seeen' across transitions and connectors (like
through a glass, darkly?). It is then difficult to analyse its behaviour, since
transitions and connectors add reflections, attenuation and phase shifts that
combine in a complex manner with the characteristics of the structure itself (Fig.
17.5). Several ways of 'de-embedding' the circuit were proposed. One may
compare it with straight sections of line (assuming connections to be identical).
Also, one may insert the circuit within a resonant structure, and deduct its
properties from the measured changes of the resonance parameters (Fig. 17.6).
The 'time gating' function of modem vector network analysers permits one to
separate connector and circuit parameters. No reliable data can, however, be
obtained across a transition that is very lossy or mismatched.
The difficulties encountered in the measurement process mean that it may be
difficult to check the validity of a model and to determine the accuracy of the
computation process. This, in turn, affects the capability of CAD procedures. Fig. 17.5 Connections from microstrip to coaxial line
a Soldered strip
b Pressed beam
c Direct soldering
17.2 Basic relationships for uniform lines d Sliding sleeve

17.2.1 Uniform lines are given by the real part of the corresponding complex quantities multiplied by
We first consider straight transmission lines, aligned with the longitudinal J2 e x p m t )
co-ordinate z. Sliding the whole structure along this axis does not modify it, the
cross-section and the material parameters being independent of z (Fig. 17.7). U(z, t) = Re [$ U ( z ) exp (jot)] IV] (17.1)
The transmission line is then called uniform or translation-invariant. It is then I(z, t) = Re [$ I(z) exp (jot)] [A] (17.2)
possible to separate the transverse and longitudinal dependences of the fields
along the line [16]. In complex phasor notation, all time derivatives are replaced by the factor jw.
1008 Computer-aided design of microstrip and triplate circuits
The time dependence is thus altogether suppressed. Solving the line equations
(derived from Maxwell's equations) yields the longitudinal dependences of the
voltage and the current along the line:
U(z) = U+ exp ( - yz) + U - exp (+ yz) [vl
I(z) = Y, [U+ exp ( - yz) - U- exp (+ yz)] [A]
where y is the propagation constant (metre-') and Y, = I/Z, is the character-
Fig. 17.6 Resonant ring with slot
[used to determine the equivalent circuit of the slot (Section 17.3.8)]
istic admittance of the line [Siemens]. Both quantities depend on the transverse
distribution of the fields across the line. The quantities U+ and U- are, respec-
tively, the amplitudes of the forward wave (travelling towards increasing values
of z) and the reverse wave (travelling towards decreasing values of z) (Fig. 17.8).
These two terms are determined from boundary conditions at the ends of the
line (generator and load).
17.2.2 Conformal mapping
In stripline and microstrip lines, the conductor boundaries within the transverse
plane are located along rectangular co-ordinate lines, but cover only over part
Computer-aided design of microstrip and triplate circuits 1009
of them (inhomogeneous conditions). In a homogeneous transmission line, the
transverse-wave equation becomes Laplace's equation in the two-dimensional
transverse plane [16]. Since the boundary conditions are inhomogeneous, La-
(17.3)
(17.4)
Fig. 17.7 Uniform infinite straight transmission lines
place's equations cannot be solved directly in the rectangular system of co-
ordinates. Particular cylindrical co-ordinate systems can, however, be defined
by conformal mapping, a technique based on transformation properties within
the complex plane [45]. One may then obtain exact solutions within the transfor-
med system. A complex number z is assigned to every point in the transverse
plane of the transmission line, referenced by its rectangular co-ordinates .u and
y (the complex number z should not be confused with the direction of propaga-
tion z)
In another complex plane, there exists another complex number w:
that is connected to z by a complex function, which defines the conformal
mapping (Fig. 17.9)
1010 Computer-aided design of microstrip and triplate circuits Computer-aided design of microstrip and triplate circuits 701 1

The function f(z) must be analytical, meaning that its derivative is continuous
and single-valued. The basic idea is to map the transverse plane of the trans-
mission line (z-plane) onto the complex plane of the transformed function w in
*
forward wave
17.2.3 Schwartz-Christoflel transforms
When boundaries are located along straight lines, the Schwartz-Christoffel
transform provides the conformal mapping for the problem. It allows one to
'straighten up' the angles. A polygon is thus transformed into a straight line,
which is most often taken as the real axis. This provides an integral equation,
and the function w = f(z) is then obtained by integration. A similar transform
is applied to a section of parallel-plate capacitor. The desired conformal map-
ping is the combination of the two transforms.
The integration is the most crucial part of the whole process. When this
integral cannot be evaluated analytically, z cannot be expressed as an explicit
function of w. In the case of striplines and microstrip, the integration can be
performed analytically, but yields rather exotic functions.
17.2.4 Zero-thickness balanced stripline
Outside the centre conductor strip, the two transverse co-ordinate axes (Fig.
17.2) form electric-field lines, or perfect magnetic conductors. Assuming that the
two ground planes extend sideways to infinity, and that the centre conductor is
infinitely thin (b = O), the Schwartz-Christoffel transform maps one quarter of
the stripline's cross-section to a section of parallel-plate capacitor [ I 31. Carrying
out the calculations eventually yields the characteristic impedance of the bal-
anced stripline [9, 211
z, = ( w & ) R(k)/K(k) [a] (17.8)

- reverse wave
where K(k) is the complete elliptic integral of first kind [l] and
k = [cosh (rrwl4h)l-' [l] (17.9)
Fig. 17.8 Forward and reverse travelling waves The symbol [I] indicates that a quantity is dimensionless.

17.2.5 Finite-thickness balanced stripline

Y 2 plane v qy plane An exact analysis by means of conformal mapping is also feasible when b 0, +
i.e. when the thickness of the centre strip is taken into account [59]. However,
the resulting developments become quite involved, yielding an implicit ex-
pression for the characteristic impedance. Calculated values have been publish-
ed in Tables [21]. A simplified approximate formula for the impedance, derived
from the exact values, is

where
Fig. 17.9 Principle o f conformal mapping F(x) = (x + I)'" + "/(x - 1)'" - ') [I] (17.11)
x = l/(l - b/2h) [l] (17.12)
such a way that the transforms of the conductor boundaries become straight The characteristic impedance of a balanced stripline is displayed in Fig. 17.10
lines. The mapping process replaces the complex geometry of the transmission as a function of geometrical dimensions.
line by a simple two-plane geometry.
 7072 Computer-aided design of microstrip and triplate circuits Computer-aided design of microstrip and triplate circuits 7073
17.2.6 Equivalent homogeneous microstrip line
The inhomogeneous microstrip line is replaced by an 'equivalent' homogeneous
line (Fig. 17.1 I) with conductors having exactly the same geometry (w, h, b), but
The relative error included in these approximations is smaller than 0.2% for
0.01 ,< w/h < 100 and 1 < E, < 128.
The phase velocity v4 and the line wavelength ig are related to the effective
permittivity by
v4 = GI/& [m/d (17.16)
Ag = &I& [ml (17.17)
Both the velocity and the wavelength are functions of the transverse geometry
of the transmission line.

17.2.7 Characteristic impedance of microstrip

h Oo 1 2
wlh
3 4 5
T h e homogeneous microstrip (Fig. 17.1 1) structure was analysed by means of
the Schwartz-Christoffel transform by Schneider [55]. The mapping is carried
out by means of the logarithmic derivative of the theta function 0, (t, k):
Fig. 17.10 Characteristic impedance of balanced stripline z ( t ) = - (2hK/7c) 8,In [8,(t, K)] [I] (17.18)
where
K = K'(m)/K(m) [I] (17.19)
with K(m) the complete elliptic integral of the first kind with modulus m.
The characteristic impedance 2, of the microstrip having width w, substrate
height h and thickness b = 0 is obtained by solving the set of simultaneous
equations

all the other parameters remain unchanged

Fig. 17.11 Definition of the equivalent homogeneous microstrip

where E(m) is the complete elliptic integral of the second kind, and dn is the
Jacobian elliptic function. The characteristic impedance cannot be expressed
surrounded by a single homogeneous dielectric of effective permittivity E,. This explicitly from these equations, so that approximate design formulas have been
permittivity is determined by calculating the DC capacitance of the inhomoge- derived by Hammerstad and Jensen [23], which provide an accuracy better than
neous structure [23] 0.01% for w/h < 1 and 0.03% for w/h ,< 1000 with respect to the Schwartz-
Christoffel transform
E, E (E, + 1)/2 + [(E, - 1)/2][1 + IO~/W]-'~[I] (17.13)
with

where
F, = 6 + (2a - 6) exp [ - (30.666 h / ~ ) ~ "[l]~ ~ ~ ] (17.24)
 1014 Computer-aided design of microstrip and triplate circuits
and where Zo 2. 12077 = 376.6 ohms
The characteristic impedance of microstrip is displayed in Fig. 17.12.
17.2.8 Finite-thickness homogeneous microstrip
The thickness b of the upper conductor can be approximately taken into account
Computer-aided design of microstrip and triplate circuits 1075
17.2.9 Microstrip line synthesis for b = 0
The equations given in the previous Sections yield the line's electrical charac-
teristics E,, Z, and IZ, in terms of the geometrical and material parameters
(analysis of a specified structure). Most often, in practice, one wishes to deter-
mine the wlh ratio that yields a specified impedance 2,. This reverse operation
of synthesis is carried out by means of approximate expressions (within 1%
accuracy) derived by Wheeler [60].
For wlh < 2:
wlh r 4[(1/2) exp (A) - exp (- A)]-' [I] (1 7.26)
with

while for wlh 2 2:

with
Fig. 17.12 Characteristic impedance of microstrip

17.2.10 Dispersion in microstrip

For high frequencies, the fields tend to concentrate within the dielectric sub-
strate, so that the effective permittivity E, increases. This may be taken into
account by means of the following approximation [17]:

all the other parameters remain unchanged

where E, is the low-frequency permittivity defined in eqn. 17.13,f is the frequency
Fig. 17.13 Microstrip with conductor of finite thickness of the signal and the remaining parameters are given by

by defining an effective width we to be used instead of the actual width in the

calculations [2 11 (Fig. 17.13):

17.2.11 Effect of an enclosure

with x = h if w > h/2n The relations presented above were all obtained for structures assumed to be
and x = ~ X W if h/2n > w > 26. 'uncovered', in which the air-filled half space above the microstrip extends to
 10 7 6 Computer-aided design of microstrip and triplate circuits Computer-aided design of microstrip and triplate circuits 1017

infinity, the structure also being infinite in the tranverse directions. In practice,
a circuit is always placed within a box, the walls and covers of which are
relatively close to the circuit. The characteristic impedance and the effective
permittivity may then be affected, in ways difficult to determine accurately [27].
Rules of thumb determine roughly when the uncovered line expressions are
valid. For aluminia (E, = 9.8), this is the case when the height to cover is more
than eight times the substrate thickness and the distance to the walls is more
than five times the conductor thickness.
The characteristics of lossy microstrip can be evaluated using a program
developed by Kajfez and Tew [37].
17.2.13 Higher-order modes and radiation
Just like any other transmission line, striplines and microstrips cannot be
utilised above a certain frequency limit, since other modes of propagation,

17.2.12 Attenuation
Three kinds of losses are encountered:
Ohmic losses within the conductors, owing to the finite conductivity of the metal.
The losses may be increased by the presence of an adhesion layer between the
substrate (in the case of ceramics) and the conductor. Also, the surface rough-
ness increases the attenuation. For striplines, the ohmic-loss contribution to the
attenuation is given by 1301

x
{x wz
+ + "'(I b'h' In *}
x - 1
for w/2h 2 0.35 (17.33)

where x is defined in eqn. 17.12.

In microstrip lines, an approximate value is given by Janssen [36]

Fig. 17.14 Circuit with bends and junctions

with the metal-wall resistance
Dielectric losses are produced by the energy dissipated in the substrate, propor-
tional to its dielectric loss factor tan 6. In stripline, the resulting attenuation is
given by [30]
called higher-order modes, would then start to propagate. In the homogeneous
stripline, these modes are either transverse electric (TE) or transverse magnetic
(TM) modes, similar to the ones encountered in hollow metallic waveguides. In
the inhomogeneous microstrip, higher-order modes are hybrid modes (like
modes on optical fibres).
Since both structures are open, i.e. not completely enclosed within a metal
envelope, they may also start radiating when the frequency increases, then
behaving like antennas. For a microstrip line, radiation becomes significant for
frequencies larger than [22]:

ci, -
while for microstrip lines, they become [22]:

27.3
E, -
--
E,
1 E, tan 6
- 1 E,
- [dB/ml
A8
Radiation losses: An infinite straight transmission line propagating in the domi-
17.3 Discontinuities: bends, junctions

nant mode does not radiate. However, at every discontinuity, higher-order 17.3.1 Dejnition
modes are excited, some of which radiate part of the signal (Section 17.2.13). In actual circuits, transmission lines are neither straight nor infinite; they start,
and then stop, at some definite points, bend, change width, branch out, etc (Fig.
1018 Computer-aided design of microstrip and triplate circuits
17.14). The discontinuities most often encountered in microstrip are sketched in
Fig. 17.15, while the main mathematical techniques used for their study are
listed in Table 17.1. Discontinuities produce reflections of the propagating mode
Open line
Double step
Fig. 17.15 Common discontinuities
(TEM or quasi-TEM), with accumulation of reactive energy in evanescent
higher-order modes, and also radiation. The latter effect only becomes signifi-
cant for very large signal frequencies; i.e. when the structure cannot be used for
signal transmission anyway (Section 17.2.13). Microstrip discontinuities are
difficult to analyse, due to their inhomogeneous nature (Section 17.1.7) and the
presence of radiation together with propagation phenomena.
17.3.2 Models
The boundary conditions for the fields in the presence of discontinuities can
only be matched by the complete set of modes in the structure. Higher-order
modes are excited, but if the dimensions of the lines are selected properly, they
cannot propagate along the transmission line and remain evanescent (Section
17.2.12: the same situation is encountered in metallic waveguides [14]). The
fields of the higher-order modes decay as one moves away from the disconti-
Computer-aided design of microstrip and triplate circuits 1019
nuity; these modes store electric and magnetic energy locally. An increase in
charge density on a conductor can be represented by an additional capacitance,
while changes in current distribution produce inductances.
Discontinuities can therefore be represented by LC equivalent circuits, in
which the components are frequency-independent at low frequencies (static/
quasi-static models), becoming generally frequency-dependent as frequency
increases (dispersive models). Reactive elements can further be combined to
form sections of transmission lines with particular characteristics. An alter-
native description provides the scattering parameters of the discontinuity (more
generally used for high-frequency studies). The relationships between different
representations of a discontinuity were considered by Mehran 1471.
Discontinuity with
infinite impedance - I 7-
Discontinuity with
zero impedance
-r 7-
Fig. 17.16 Zero- and infinite-impedance discontinuities
While one might expect results from dispersive models to be more accurate,
this is not always true. The DC extrapolation of results provided by a dispersive
model was found to yield incorrect values in many situations [lo].
Discontinuities can generally be separated into two main families (Fig. 17.16):
With zero DC resistance, with a continuous upper conductor extending com-
pletely across the discontinuity. The equivalent circuit then contains both
capacitances and inductances.
With injnite DC resistance, when the upper conductor is interrupted between
 1020 Computer-aided design of microstrip and triplate circuits Computer-aided design of microstrip and triplate circuits 1021

the ends of the discontinuity. No direct current can flow across the structure,
which is represented by a purely capacitive equivalent circuit.

17.3.3 TEM-line models

The shunt capacitance of an open microstrip line is given by a perturbation of
the uniform line (Section 17.2), that takes into account the fringing fields at the
open edge [33, 321. Fairly approximate results were obtained, and the approach
could not really be extended to consider other discontinuities.

17.3.4 Variational techniques

A variational (stationary) mathematical expression for the shunt capacitance of
an open circuit was established by Maeda [42]. The unknown potentials within
a closed structure are expanded over the infinite set of solutions for Laplace's
equation. The approach is rather involved and would be difficult to generalise
in order to analyse more complex discontinuities.

17.3.5 Fourier transform

The approach used for an uniform transmission line [62] was extended to cover
also three-dimensional structures [31]. The Green's function is solved in the
spectral domain by a Galerkin technique [34]. The calculated capacitance of an
open microstrip line agrees quite well with measured data. This approach was
used by Koster and Jansen [39] to analyse step discontinuities.

17.3.6 Dielectric Green'sfunction

This flexible approach to determine capacitance remains accurate even in the
DC limit. It was applied to most geometries (Table 17.1) by Farrar and Adams
[ l l , 121, and by Silvester and Benedek [6, 56, 5 7 . A three-dimensional Green's
function is defined by generalising the one introduced for the uniform microstrip
line. The integral equation is solved with the moment's method, either with
functions defined over the whole discontinuity [56], or with step functions that
are constant over a rectangular sub-domain and vanish everywhere else [I I]. A
similar approach was developed to solve the much more complex problem of
radiation from microstrip antennas (Chapter 8).

17.3.7 Integral equations for induttances

The inductance of a bend was determined from the magnetic energy, replacing
the surface current by a wire net [18]. Since inductance becomes infinite for
infinitely thin wires, some approximations were introduced, but they proved to
be rather inaccurate.
These drawbacks were avoided by the use of an integral equation based on
the skin effect. Results were obtained for the microstrip bend [58] and for the
impedance jump in Xand Yjunctions [19]. Since this technique considers actual
surface currents, the magnetic energy can be determined without mathematical
difficulties. However, the computation process becomes quite involved, and one
7022 Computer-aided design of rnicrostrip and triplate circuits 1023
Computer-aided design of rnicrostrip and triplate circuits
must first compute a scalar-potential solution of Laplace's equation at the
discontinuity, which becomes the kernel of the integral equation.

17.3.8 Green's function and integral equation

The two techniques presented in previous Sections were improved upon, sim-
plified and combined to determine the complete equivalent circuit of a slot [52],
of a double step and a mitred bend [5].

17.3.9 Green's function and electrostatic-inductance computation

Another combined technique determines the capacitance by means of a dielec-
tric Green's function. Inductances are then assumed to be proportional to the
square of the difference of the electrostatic potentials with and without the
discontinuity, producing a virtual lengthening of the microstrip line. The im-
pedance step [28] and the bend [29] were analyzed by this approach. Calculated
results did not compare well with measured data.

17.3.10 TLM (transmission-line-matrix) method

A technique developed for uniform transmission lines was extended to study
discontinuities like impedance steps [3] in completely enclosed structures. The
technique provides resonant frequencies for discontinuities of variable length,
but the equivalent circuit is not uniquely defined by these resonances.

17.3.11 Waveguide model

The first development of the waveguide model for stripline discontinuities is due
to Altschuler and Oliner [4]. Babinet's principle allows one to replace the
original problem by an equivalent rectangular waveguide, for which solutions
are known [44]. This approach was applied directly to microstrip T-junctions
1411.
The width and the internal permittivity of the waveguide are 'effective' values
obtained from the static analysis of the uniform line (Section 17.2). The
frequency-dependent scattering parameters were determined for the impedance
step [61, 381, the simple bend [46], the mitred bend [48], the T-junction [61,46],
the Y-junction [49] and the X-junction [50]. The LC equivalent circuit of a slot
was determined using a simplified version of this model [24, 251.

17.4 Technological realization: materials and manufacturing process

17.4.1 Introduction
Physically, any microstrip structure - circuit or antenna - is made of two
parts: the substrate, a dielectric material with losses as small as possible, and a
metallisation (partial or total) on the substrate's faces. Various processes can
provide the desired metallisation pattern.
 7024 Computer-aided design of microstrip and triplate circuits
17.4.2 Dielec?ric substrate
The substrate fulfils two different functions: it is the mechanical support for the
structure, and it is also an integral part of the transmission lines, determining
the electrical characteristics of the circuit or antenna (Section 17.2, Chap. 8).
Mechanically, the following properties of the substrate must be considered:
0 Mechanical strength (e.g. breaking point), which determines the impact and
vibration resistance
Shape stability, in particular for encasing
0 Dilatation factor, which should be small, as close as possible to that of the
metal used for the conductors and the enclosure
Long-term behaviour in the presence of difficult environmental conditions
(moisture, temperature cycling).
The electrical parameters that must be considered are:
0 Relative permittivity E , , which determines the miniaturisation factor. When
all other parameters are kept equal, the size of a circuit is proportional to ~ / J E , .
By choosing a large permittivity one may reduce the circuit dimensions
0 Uniformity of the permittivity 8, over the whole circuit
0 Low dispersion of the permittivity E, and the thickness among different
batches of a given material (circuit reproducibility)
0 Small dielectric losses (one should have tan 6 < 0.001) in order to have
high-performance circuits and acceptable quality factors for resonant circuits,
filters and radiating elements
0 No absorption of water (water exhibits a high permittivity and high losses).
The following physico-chemical parameters are significant:
0 Mechanical stability up to high temperatures (soldering, deposition of com-
ponents in the thick-film technique)
Resistance to chemicals, in particular during the different stages of the
photolithography process
0 Surface flatness (bending tends to render the encasing procedure difficult)
Smooth surface, to reduce losses and ensure good adhesion of conductors,
0 Easy machining, for the cutting and drilling of holes.
Production requirements are:
0 Low cost
0 Guaranteed availability of the material
Availability of adequate sizes
0 Non-hazardous machining.
17.4.3 Comment
Considering all the above requirements, some being conflicting, it is fairly
obvious that no actual substrate would simultaneously meet them all. For every
application, one must carefully consider the different requirements and select
the material providing the best compromise. There are, in fact, many kinds of
Computer-aided design of microstrip and triplate circuits 1025
substrates, and they can be grouped in four main categories: inorganic, plastic,
semiconductor and ferrite. Table 17.2 provides the relevant data on some
common substrate materials.
17.4.4 Inorganic substrates
This category contains mainly ceramics.
Alumina (AI,03) is one of the most commonly used substrate materials. It is
characterised by good surface quality, very low losses and very little dispersion
between batches. However, it is slightly anisotropic. Common thicknesses are
0.254, 0.635 and 1.27 mm, while dimensions are generally stated in inches
(1 x 1 in to 4 x 4 in). Alumina is a very hard and brittle material; hence it is
quite difficult to machine and its permittivity depends on its porosity. Since
adhesion of copper and gold to alumina is poor, an intermediate layer of
chromium (or of some other lossy conductor) is required.
Sapphire is the monocrystalline form of alumina. It is used in particular appli-
cations at very high frequencies, when a very smooth surface is required.
Sapphire exhibits crystalline anisotropy.
Beryllia (BeO), with a lower permittivity than alumina, presents a very large
thermal conductivity, which makes its use particularly appropriate for high-
power applications (removal of heat produced by semiconductors). However,
Be0 powder is highly toxic, so that particular precautions must be taken while
machining.
Rutile (TiO,) has a very high permittivity, and is unfortunately temperature-
sensitive.
17.4.5 Plastic substrates
Pure synthetic materials may be used, such as PTFE (Teflon) or polyolefin.
Their permittivity is generally low (E, = 2-3) and the mechanical properties are
rather poor (mechanical distortion, poor temperature behaviour). By adding
ceramic powders or glass fibre, the mechanical stability can be improved and the
permittivity increased, but losses also become larger.
PTFE has a low permittivity and very low losses. It is quite poor mechanically.
(e.g. dilatation).
Glass-Jibre reinforced plastics (such as RT-Duroid 5870) are much better sub-
strates mechanically, but their permittivity is slightly higher, as well as their
losses, and they present some anisotropy (in particular, for woven fibres).
Ceramic-loaded plastics (such as RT-Duroid 6006 and 6010.5) can have a
permittivity approaching that of alumina, with very good mechanical proper-
ties. Machining and drilling is easy, and they can adequately replace alumina
during prototype development, even though their losses are somewhat higher.
Synthetic substrates are available in large sizes, which is very convenient in
practice. A slight anisotropy can be tolerated in most usual applications.
1028 Computer-aided design of microstrip and triplate circuits Computer-aided design of microstrip and triplate circuits 1029
17.4.9 Circuit realisation 17.4.12 Removal of photoresist
A particular metallisation pattern has to be realised on the substrate: centre Once the previous steps have been completely carried out, the remaining
conductor in stripline, upper conductor in microstrip. In all cases, this is done photoresist is removed with a solvant or a concentrated alkaline solution. The
by means of a mask, first designed and cut to a larger scale (on a co-ordinato- metallic layer is sometimes thickened by electrolytic deposition of metal, or a
graph or plotter, Section 17.6), then photographically reduced to the proper protective layer (e.g. gold) is deposited to prevent oxidation.
size.
A photosensitive lacquer is deposited on the structure; several processes may metal
be used to do this: printed circuit
dielectric material
Dipping: The entire structure is dipped into the lacquer, and then pulled out at
-metal
constant speed. The thickness of the layer depends on the withdrawal speed and
the viscosity of the photoresist. This process generally yields rather thick layers,
with a bulge on the lower edge.
Spraying: The photoresist is sprayed on the structure through a nozzle. It is
difficult to obtain a constant thickness in this manner.
Centrifuge: The structure rotates rapidly, and the photoresist is deposited at the
centre of rotation, being swept towards the outside by the centrifugal force. Thin
and uniform layers are obtained in this manner, but the technique can only be
used for small substrate sizes.
After deposition, the photoresist must be cured at a high temperature, becoming
tougher and more adhesive.
The photographic mask is then vacuum-pressed on the structure, with the
emulsion side next to the photoresist layer. The photoresist is exposed to
ultraviolet rays through the mask. The UV radiation must be parallel and the
photoresist layer thin, to ensure accurate reproduction of the pattern. development
By developing in a suitable chemical, either the UV-exposed part of the
photoresist layer (positive photoresist) or the non-exposed one (negative
photoresist) is removed. The circuit is then ready for the next step, which is
either the deposition or the removal of metal.

17.4.10 Etching
When starting with an entirely metallised substrate, part of the metal must be etching

removed by etching (Fig. 17.17). The structure is exposed to an acid that

dissolves the metal but does not affect the remaining photoresist. The process proter tiue lacquer
can become complex when several metallic layers are involved, since specific
etching solutions are required to remove each metal layer. Careful rinsing is
necessary between the different baths. photo1 esirt
remoudl
17.4.11 Metal deposition
This is the opposite process, in which one deposits metal on the substrate
Fig. 17.17 Realisation af a circuit by the etching process
throught the holes in the photoresist (Fig. 17.18).~everalprocesses may be used:
vacuum evaporation, sputtering, electroless plating. The layers obtained are
generally thin (2-1 5 pm). All the steps in the procedure must be carefully separated by rinsing and
cleaning operations, sometimes followed by drying in an oven. Drying is par-
ticularly critical for some plastic substrates, which tend to absorb water.
 7030 Computer-aided design of microstrip and triplate circuits 1
I Computer-aided design of microsfrip and triplate circuits 1031
I
17.4.13 Under-etching
so long as the electrical conduction requirements are satisfied. Under-etching is
Of course, one wishes to realise a circuit having a pattern as similar as possible
to the one of the photographic mask. However, in both the etching and the
, a predictable process, which can be taken into account when designing the
circuit and cutting the mask (Section 17.6).
metal-deposition processes, the most annoying phenomenon of under-etching
takes place. Considering the cross-section of a transmission line (Fig. 17.19), one

cerarnlc
substrate

chromium

gold

electrolytic gold
enporure t o uu /
mask

deuelopment

Fig. 17.19 Effect of under-etching

m e t a l deposition
17.4.14 Thin and thick film .
The techniques described so far realise 'thin-film' circuits, used, in practice, for
most microwave circuits. The 'thick-film' approach deposits a paste through a
silk screen. It is not usually accurate enough for microwave circuits, but is used
to realise components like resistors or capacitors (Section 17.7.5).
pholorerisl
removal
17.5 Analysis and synthesis programs

Fig. 17.18 Realisation of a circuit by metal deposition 17.5.1 Introduction

notices that the conductors do not have rectangular, but trapezoidal cross-
sections. The acid does not remove the metal uniformly, while metal-deposi-
tion processes do not yield constant growth. Under-etching is more important
when the metal layer is thick: for this reason, thin conductor strips are preferred,
While CAD packages for the design of low-frequency electronics systems have
been available for some time, the choice of microwave programs used to be
limited and expensive. As a result, traditional 'cut and try' methods still remain
a commonly used way to realise circuits, despite their many drawbacks and
inadequacies.
 7032 Computer-aided design o f microstrip and triplate circuits
The situation has changed radically since the arrival of personal computers:
considerable computing power has suddenly become available, while increasing-
ly sophisticated microwave CAD packages have appeared on the market. The
designer, faced with a wide variety of choice, must now determine which
software is best suited to his particular requirement.
Software comes in many kinds, and it is therefore difficult to make com-
parisons. Most packages are based on models and equivalent circuits, while
some make use of electromagnetic-field analysis. Programs can be classified into
two main categories: those devoted to analysis, in which the user describes a
structure and wishes to know its electrical response; and those devoted to
synthesis, where the program defines a physical structure meeting a specified
electrical behaviour. The latter can be done either by circuit-synthesis techni-
ques, or by an analysis software inserted into an optimisation loop (many
packages then require an approximate solution as starting point). The models
used must be rigorous enough to provide an accurate analysis, but not too
complex, as computations would become too lengthy [26]. Some software
packages were developed specifically to design particular devices (couplers,
filters, amplifiers), while others have more general scope. The next Sections
describe some of the present programs for microstrip design (the list is not
exhaustive), on the basis of available literature and opinions received from
users. Surveys of available CAD software appear occasionally in the technical
literature [51, 151.
17.5.2 EEsof: Touchstone (IBM-PC, HP 9000, Apollo)
Touchstone is a very fast general-purpose RF and microwave circuit-design,
analysis and optimisation program, specifically developed for personal com-
puters. More than 80 elements, including microstrip, stripline, waveguide and
co-planar lines and discontinuities, lumped elements and electronic-device
models are contained in its catalogue. In addition, the user can define his own
elements, specifying their scattering matrix. It offers possibilities of adjustment,
and two optimisation algorithms (random perturbation and gradient method)
for up to 15 variables. The companion program Monte Carlo determines the
sensitivity of the circuit. Touchstone is the CAD software most often used
nowadays in microstrip circuit design.
EEsof also provides several device-synthesis programs compatible with
Touchstone: waveguide, microstrip and stripline bandpass, low-pass and band-
stop filters, and coaxial low-pass and band-stop filters. The program E-Syn
synthesises matching circuits. Microwave SPICE provides nonlinear circuit
analysis and synthesis. The program Anacat is geared towards measurements,
particularly with vector network analysers, and provides embedding and de-
embedding functions (Section 17.1.8). A graphics program MICAD, running on
the same software, draws the artwork and can generate masks on co-ordinato-
graphs and photo-plotters [43] and on some regular plotters as well, using a
diamond stylus (Section 17.6.3).
Computer-aided design o f microstrip and triplate circuits 703.3
17.5.3 CCC: The Supercompact Family (VAX, IBM, IBM-PC, Apollo, PC,
HP 9000)
Supercompact is a general analysis and optimisation design tool for microwaves
and RF. Four-port circuit analysis and optimisation can be achieved, as well as
transistor impedance modelling (its library includes specifications of commer-
cially available transistors) and matching network synthesis. Single, coupled
and inter-digitated microstrip designs can be created. Microstrip and other
planar lines can be specified, both in terms of their electrical and physical
dimensions; approximate expressions are included, with accuracies of I % or
better. The effects of dispersion, radiation, discontinuities, multi-layers, metal-
lisation, surface roughness, and dielectric and conductor losses are taken into
account. An FFT time-domain option is available. Circuits are optimised by the
random perturbation and gradient techniques, their parameter sensitivity is
determined with a Monte Carlo algorithm. The companion program Autoart,
that uses the same hardware, was the first microwave CAD layout package
offered commercially (Section 17.6.2). Super-compact and Autoart were f rst
developed on main-frame computers; they are also available now for perso1.al
computers. Microcompact is a simpler program, that runs on HP computers of
the 200 series.
17.5.4 CCC: CADECf (Computer-Aided Design of Electronic Circuits, on
HP 200/500, Tektronix and IBM-PC XTIAT)
CADEC, a circuit analysis and optimisation package, is one of the oldest
programs (1973), and became commercially available in 1980. Covering the
range VLF-40 GHz, it performs frequency- and time-domain analysis, utilising
matrix representation and nodal analysis for two-ports, including lumped ele-
ments, stripline, coplanar lines and discontinuities, and couplers [53]. Disper-
sion is taken into account, and the domain of validity determined. Amplifier
analysis also evaluates the noise parameters. The optimisation is based on a
search technique followed by a gradient approach, and the sensitivity is deter-
mined. Design kits run on the same hardware: Microwave Design Kit, (am-
plifiers), Filter Design Kit, (filters) and Sonata (oscillators).
17.5.5 Acline (VAX, Apollo and HP 9000)
A c h e is a high-level program, somewhat similar to Supercompact, developed
by Professor C. Vidallon of the Universite Paul Sabatier in Toulouse, France.
It provides impressive optimisation facilities to its users, and a number of
microstrip elements are available in an interactive way. A new version, called
AC-LINE, offers considerably increased performances. Unfortunately, little
information about Acline has appeared so far in the technical literature.
17.5.6 Thom'6: Esope (Vax, IBM, Apollo)
Esope is an interactive software for the analysis and optimisation of linear
microwave circuits. Most components (lumped RLC, transmission lines,
 1034 Computer-aided design of microstrip and triplate circuits
waveguides, coupled lines, microstrip) are implemented. Active circuits are
described by their scattering parameters. Optimisation can be carried out with
one or several objectives: three algorithms are available (min-max, least
squares, fixed tolerance), the sensitivity and worst case are determined, and
results are represented graphically. The program is in Fortran 77. Drafting of
the circuit's outline is carried out by software Hyper'6-D.
17.5.7 RCA: Midas (UNIX and SUN on most 32 bit processors)
This program analyses and optimises circuits with up to 20 nodes and five ports,
utilising algebraic expressions to define the parameters required to analyse the
circuit. It is compatible with the PIana software for two-port measurements on
the HP 8409 network analyser. All the usual components (lumped RLC, trans-
mission lines, coupled lines, discontinuities, gyrators, ideal transformers and
controlled sources) are implemented. Particular emphasis is placed on striplines
an? nicrostrip. Inter-element coupling is determined for an arbitrary net of
h e s , in order to analyse comb and inter-digitated structures. The companion
program N-Fet is specifically devoted to the design of FET amplifiers.
17.5.8 LINMIC (HP 9000 series 500 and 300, Microvax 11)
The CAD package LINMIC introduces a significant new approach to the
layout-oriented design of single or multi-layered planar and monolithic struc-
tures: it combines, apparently for the first time, a rigorous field analysis - based
on an enhanced spectral-domain technique [34] - with the more usual models
and equivalent circuits (for simple and coupled lines, discontinuities, T-junc-
tions, capacitances, resistors, transistors and other lumped elements) [35]. The
LINMIC package can actually describe complex structures for which no analy-
tical models are available: interdigitated capacitors (Fig. 17.20), multi-turn
Fig. 17.20 Interdigitaced capacitor
square inductors (Fig. 17.21) and coupled meander lines. Up to four dielectric
layers can be considered, including passivation and metallisation layers, ohmic
and dielectric losses, coupling and higher-order-mode effects. Up to 40 connect-
ing points to the outside can be considered, and the structure can be divided into
20 sub-lattices.
An interactive optimisation procedure is provided, based on a conjugate
Computer-aided design of microstrip and triplate circuits
1035
direction algorithm, in terms of reflections, gain, insertion losses, ripple, noise
and stability, as defined by 60 or more parameters. The sensitivity with respecr
to those parameters is analyzed. Very high accuracy, wide validity range and a
high computation speed drastically expand the range of applications. It is highly
portable, since 98% of its code is written in Fortran 77.
/
metal strip
j i e l e c t r i c bridge
substrate
dielectric br
1
Fig. 17.21 Multi-turn square inductor
MCAD comes from the same source as LINMIC, for considering various
simple and coupled planar lines (up to 20 coupled lines) on multi-layer sub-
strates (up to 4 substrates). It is used to analyse discontinuities, junctions,
couplers and filters, as well as lumped elements. Three optimisation algorithms
are provided.
17.5.9 High Tech. Tournesol: Micpatch ( V A X , PC)
The package Micpatch was developed by Dr. Mosig at LEMA. Lausanne,
Switzerland, primarily to analyse microstrip patch antennas. It can just as well
be applied to the study of microstrip circuits of arbitrary shape. The scattering
matrix of multiple-port components is determined, taking into account radi-
ation and surface waves, metal and dielectric losses. It is written in Fortran 77,
and a more detailed description of the procedure used is given in Chapter 8.
17.5.10 Spefco Software: CiAO (IBM PC-XT or AT)
CiAO is a general analysis and optimisation program for circuits composed of
RLC lumped elements, controlled sources, gyrators, lossy or lossless trans-
mission lines, and one- and two-ports described by their scattering matrix.
The companion program Design synthesizes wideband matching circuits for
linear-gain amplifiers.
 7036 Computer-aided design of microstrip and triplate circuits
17.5.11 Made-it-associates: Mama (Measurement And Microwave Analysis,
HP 9836 or HP 9000 series 300)
This program analyses and designs: quarter-wave transformers, Lange couplers,
directional couplers, microstrip lines and discontinuities, hybrid circles, power
dividers [in fact, most of these elements can be realised by Micros (Section
17.6.4), which additionally draws them and cuts the masks]. It synthesizes filters
and rectangular printed antennas. It can be used for de-embedding components
(Section 17.1.8) and interfaces with Microcompact (Section 17.5.3).
17.5.12 Ampsa: Multimatch ( I B M PC)
Multimatch is a matching package for microwave loads and amplifiers de-
veloped by Abrie [2]. The potential performance of a transistor is determined
from its scattering matrix, and matching circuits are directly synthesized for
stable operation, either with lumped components or with transmission-line
sections. Discontinuities are taken into account and corrected for. The approach
provides the user with several possible designs to choose from, together with
their electrical performances. If needed, the circuits developed can be used as
starting points for an optimisation program. In many practical applications, the
solutions proposed by Multimatch were found to be close enough to optimal not
to require further refinements.
17.5.13 Radar systems technology: Analop (IB-PC and CP/M-80)
The program analyses and optimises, on 15 variables at 15 frequencies, two-
ports having up to 200 elements: lumped RLC components, series and parallel
resonators, transmission lines, transformers, impedance inverters, microstripli-
nes and discontinuities, rectangular waveguides (dominant mode), controlled
sources, filters and couplers.
17.5.14 MicrokoplSuspend (IBM PCJXT)
An iteration algorithm allows both analysis and synthesis of coupled lines for
microstrip and suspended substrates [8].
17.5.15 Microwave software applications (MS-DOS or PC-DOS)
A series of small programs specifically dealing with couplers is proposed:
single-section, symmetrical or asymmetrical multiple-section, Lange in various
technologies: balanced stripline, microstrip, rectangular and ridged waveguides.
17.5.16 Planim
The planar circuit approach to the characterisation of microstrip circuits is
combined with the image-parameter method to give a new powerful technique
for integrated-circuit design of microstrip filters [54]. The inclusion of two-
dimensional effects in the synthesis procedure makes the Planim approach
particularly suited for monolithic microwave circuits.
Computer-aided design of microstrip and triplate circuits 1037
17.5.1 7 DGS Associates: SIFilsyn (PC, HP 9000, V A X )
A broad series of programs for the synthesis of filters, active and passive, analog
and digital, lumped component design or microstrip, that were developed in
1967 by Dr. Georges Szentirmai.
17.5.18 Webb Laboratories: Transcad (IBM PC-XTIAT)
This specialized program is devoted to the study of transitions between trans-
mission lines (coaxial, bifilar, planar structures) and waveguides.
17.6 Layout of circuits and cutting of masks
17.6.1 Description
When a design has been analysed with the most accurate decription available,
and then optimised by sophisticated CAD techniques (Section 17.5), a most
critical part of the procedure still remains to be carried out: its physical realisa-
tion. Before moving on to the photolithographic process (Section 17.4.9), the
pattern of the upper conductor must be drawn, and then the scaled mask must
be cut. Co-ordinatographs or photoplotters can generate very accurate masks,
but these machines are quite expensive, and well beyond the reach of many small
research laboratories or academic institutions. The layout and cutting of masks
can now be done accurately enough, but at a much lower cost, directly on a
standard plotter connected to a desktop computer.
17.6.2 CCC: Autoart
The Autoart program converts to artwork microwave circuit models described
by their physical dimensions (single and coupled transmission lines, circular
radial stubs, tapered lines, Lange couplers, circuit discontinuities). It is presently
restricted to planar geometries with a single path from input to output (this
restriction may, however, be overcome by declaring other ports to be open or
short-circuited). Circuit data provided by the Supercompact package (Section
17.5.3) can thus be used to prepare the conductor pattern, and the geometrical
data is then transferred to a co-ordinatograph or a photoplotter. Autoart
interfaces directly with Wild Heerbrugg precision flat-bed plotting tables, with
Aristo Graphics automatic co-ordinatographs, with Gerber Scientific photo-
plotters and, through an Initial Graphics Exchange Specification, with pattern
generators and mechanical drafting equipment (these instruments can all be
used to generate masks) [40].
17.6.3 EESOF: Micad
In a similar manner, the data provided by the general-purpose program Touch-
stone can be transferred through the Micad software to generate a mask on a
co-ordinatograph or a photoplotter [43]. Using a specially configured diamond
stylus (MICknife), masks can also be cut on standard plotters.
 1038 Computer-aided design o f microstrip and triplate circuits
17.6.4 High Tech. Tournesol: Micros (HP 9000 series 200 and 300)
In contrast with the two previous packages, which are additions to general
analysis/synthesis software packages, Micros is a complete design tool, that
generates the drawings of selected components, interconnects them and cuts the
mask on a standard plotter [63]. The program is 'user-friendly' and completely
Fig. 17.22 Microstrip components realised by the program Micros
self-contained (it does not require extensive reading of manuals before use). The
operator specifies circuit dimensions, operating frequency, substrate permit-
tivity and width, and characteristic line impedance. A number of elements then
become available to design the circuit: couplers, hybrids, dividers, step transfor-
mers, bandpass and low-pass filters (Fig. 17.22). He can position and intercon-
Computer-aided design o f microstrip and triplate circuits 1039
nect them at will, with compensated bends and sections of line, right on his
computer screen, in order to prepare his layout. He may also define his own
elements, e.g. patch antennas. The mask is then cut on a Rubylith sheet with a
specifically designed sharp cutting tool, all in a matter of minutes to computer
accuracy. Circuits realised with Micros were measured on a vector network
analyser, and showed very good agreement between specifications and measured
data, even for highly frequency-sensitive elements.
The design of frequency-sensitive components such as filters requires a precise
knowledge of the substrate characteristics (permittivity and thickness). Some
manufacturer's specifications were found to be much too loose for design
purposes, particularly for composite high-permittivity substrates. A companion
program, Epsilon, was developed to measure the substrate's permittivity under
actual operating conditions. A circular resonant ring is designed to resonate
close to the operating frequency, and a mask is generated. When the resonant
frequency has been measured, the program determines the permittivity [64].
17.6.5 British Telecom: Temcad ( H P 9000 series 200 and 300)
A general program for the layout of planar circuits was recently developed by
British Telecom, combining theoretical models with experimental data, and
including a wide range of flexible facilities to improve designer productivity [20].
17.7 Insertion of components
17.7.1 Definitions
Lumped elements, such as capacitors, resistors and inductors, can either be
manufactured by metal or dielectric deposition, right on the microstrip or in the
balanced stripline, or alternatively discrete manufactured miniature com-
ponents can be connected on the microstrip, or within the stripline (the last
procedure is somewhat more complex, since a cavity must be carved within the
dielectric materials). A number of semiconductor devices, diodes and transis-
tors, as well as nonreciprocal ferrite devices (isolators and circulators), can also
be inserted into planar circuits to realise active antenna feeds and interconnect-
ing networks. These two categories are distinguished in the forthcoming Sec-
tions, and labelled, respectively, 'deposited' and 'discrete'.
17.7.2 Discrete components
Discrete components are in general commercially available, and ready to be
inserted into a circuit. Most of the common electronic components - active
and passive - are presently available at microwave frequencies in a package
suitable for insertion within a microstrip circuit. Their main characteristic is
their small size: a lumped element is small with respect to the wavelength. In
addition, parasitic effects must be kept as small as possible, so that wire connec-
tions are extremely short or altogether absent, being replaced by metallised
1040 Computer-aided design of microstrip and triplate circuits
connecting surfaces. Special device cases were developed, to reduce the discon-
tinuities as much as possible.
Resistors: are small 'blocks' with sides from a few tenths of a millimetre up
to a few millimetres (Fig. 17.23). They are generally made of a ceramic block,
on which a resistive layer is deposited, between two metallised regions for
connection purposes. The geometry of the resistive layer is adjusted with a laser
during the fabrication process. It is possible in this way to meet tight tolerances.
resistive
laser trimming
for connections
Fig. 17.23 Miniature resisto~
Sometimes, two connecting ribbons are attached, to facilitate the insertion
procedure. Resistive loads and attenuators (Fig. 17.24) are manufactured by the
same technique.
connecting
strips
Fig. 17.24 Miniature attenuator
Seen from the outside, capacitors look very much like resistors, they consist
of a sandwich of ceramic and metal layers. Available capacitance values range
from fractions of a picofarad up to several nanofarads.
Adjustable capacitors are available for insertion into microstrip circuits, and
Computer-aided design of microstrip and triplate circuits 104 1
can be either series or shunt connected (in which case a hole must be drilled
through the substrate). These capacitors are a few millimetres in size, so that
their use is limited up to a few gigahertz, while resonances appear at higher
frequencies. These adjustable components are most useful for fine tuning, or to
compensate for component non-uniformity.
Miniature inductors for microstrip are fairly recent additions. They consist of
helically deposited metal strips on very thin multi-layer ceramic substrates,
connecting from a layer to the next at each turn. A multi-turn 'coil' of very small
dimensions is obtained in this manner. Large inductance values with small loss
can be realised in a small volume. In practice, however, inductances are directly
deposited on the microstrip (Section 17.7.5).
Junction isolators and circulators can be mounted on microstrip. They are
small ferrite discs, a few tens of millimetres in diameter, with small ribbons for
connection within the circuit (Fig. 17.25). Holes must be drilled into the sub-
strate to permit insertion. These components are relatively heavy and large,
owing to the presence of permanent magnets.
connecting
strips
Fig. 17.25 Circulator for insertion in microstrip circuit
Most kinds of diodes suitable for mounting on microstrip are presently
available: Schottky, PIN, varactors, Gunn, Impatt, Trapatt, etc. The diodes can
be either inserted in chip form, or encased in various packages: LID ( Leadless ,
inverted device, Fig. 17.26a), beam lead (connection with thin strips, Fig.
17.266) or cylindrical cases, with or without a mounting screw (Fig. 17.26~).The
last package is mounted through the substrate, with the lower part in close
contact with the ground plane to ensure heat removal.
The increasing significance of microstrip circuits is closely related to the
availability of transistors, both bipolar and MESFET, that have steadily im-
proving characteristics. Cases that degrade as little as possible the performances
of the chip have been designed. The usual cases (Fig. 17.27) permit easy
insertion, with minimal discontinuities. A hole must be drilled through the
substrate for power devices. However, the best performances are achieved when
using chip-mounted transistors.
 7042 Computer-aided design of rnicrostrip and triplate circuits Computer-aided design of microstrip and triplate circuits 1043
ceramic support Precautions are required to avoid overheating of delicate components during
the mounting procedure
/ I Connections may deteriorate owing to aging (oxidation, etc).
Since components are generally quite small, their mechanical stability is most
often ensured by the electrical connections themselves.

\
metallization metal bond
ceramic case(Be0)

a. Leadless I n v e r t e d Oeuice ILIOI

connecting
strips

b. "beam-lead"

Fig. 17.27 Transistor for microstrip circuit

The classical soldering, currently used in electronics, can be used for mounting
components on a microstrip substrate, but particular care and equipment are
required. The soldering iron must have a very narrow tip, and a number of
solders based on indium, tin and lead can be used, with melting points in the
range 143OC to 280C [40].
Thermocompression bonding, applying heat and pressure at the same time,
c. t h r e e cyllndrlcal ceramic cares
produces an inter-atomic diffusion, and thus a high-quality weld. This process
Fig. 17.26 Various diode packages usually utilises gold as the welding material. Several bonding methods are
a Leadless inverted device (LID) shown in Fig. 17.28~-c.The duration of the bonding operation (typically in the
b Beam lead
c Three cylindrical cases
1-3 s range) is a critical parameter in the bonding process.
Ultrasonic welding uses the same physical principle as bonding: heat and
17.7.3 Mounting procedure pressure. In this case, the heat is produced by mechanical rubbing by a point
Particular care must be taken when mounting lumped elements on microstrip vibrating at ultrasonic frequencies (Fig. 17.29). This technique is particularly
substrates: well suited to very sensitive components.
Welding can also be realised by Joule heating, by circulating a current be-
The electrical connections must be made very carefully tween two electrodes (Fig. 17.30). The interest of this process lies in the fact that
Discontinuities between component and circuit must be as small as possible heat can be applied very locally. On the other hand, semiconductor devices may
The mounting must be mechanically rigid become damaged.
 1044 Computer-aided design of microstrip and triplate circuits Computer-aided design of microstrip and triplate circuits 7045
Finally, devices can also be glued using conductive epoxy, loaded with fine
metal particles, generally gold or silver. This process is extremely useful for
mounting very delicate components, and also for encasing microstrip circuits

ultrasonic
transducer

heating plate

I Fig. 17.29 Ultrasonic welding

gold thread Tungsten
\ Jube pressure lifting

!I!!
current source

electrode electrode
pressure
\
drop solQer
I conductor
torch
b

gold thread heat and

pressure

sudstrate
Fig. 17.30 Electric-current welding

17.7.4 Drilling holes in the dielectric substrate

Several components require a hole to be drilled through the substrarte. either to
provide a ground connection or a good thermal contact to remove the heat
,dissipated in the device. With plastic substrates, the process is straightforward,
and perfect holes can be drilled, so long as the drill and the speed are selected
correctly. The matter is far more complex with ceramic substrates, which are
hard and brittle. Two techniques can be used:
Drilling with an abrasive powder (carborundum) and a rotating tube (Fig.
cutting 17.3 1) or an ultrasonically driven point. This approach is rather time-consuming
of thread
A laser can realise perfect holes, even non-round ones, in a fraction of a
C
second. However, a special drilling set-up is required.
Fig. 17.28 Themocornpression bonding
a Edge bonding
In all situations one tries to drill as few holes as possible in ceramic substrates.
b Drop bonding
c Lateral-point bonding
 7046 Computer-aided design of microstrip and triplate circuits
17.7.5 Deposited components
These components are not introduced from the outside, but are built right on
the substrate, and thus are an integral part of the circuit. Resistors, capacitors
and inductors can be realised by this process. They are, in general, better suited
to very high-frequency operation, since they do not present the discontinuities
that always appear when a component is inserted. They are also generally
smaller.
Resistors are realised with either of two techniques (Fig. 17.32). A resistive
rotation
Computer-aided design of microstrip and triplate circuits 7047
can be used to realise resistors in an easy way: one just has to remove the
superior high-conductivity layers.
Capacitors can be realised by the successive deposition of dielectric and metal
layers, either by silk screening (thick-film process) or by evaporation (thin-film
process). Additionally, a single-layer capacitor can also be made by the
photolithographic process, using an interdigitated geometry (Fig. 17.20).
Inductances are also realised by the photolithographic process, in the form of
a loop or a spiral. In the latter case, the inner connection must be 'pulled out',
either by bonding a thin gold wire, or by building a dielectric bridge with a metal
strip deposited on top (Fig. 17.21).

17.8 Examples
nickel
tube pressure 17.8.I Design of a broadband amplifier
One of the most current applications of microstrip CAD is the design of
abrasiue
Substrate
amplifiers to have a constant gain over a specified frequency band. As an
example, an amplifier with a gain as flat as possible over the range 3-5 GHz
was designed around a GaAs MESFET of the type DXL-2501 (Dexcel).
The first step in the procedure is the actual measurement of the transistor

-
parameters, carried out on a specialised transistor test fixture connected to an
Fig. 17.31 Drifting of microstrip substrate with a rotating tube
automatic vector network analyser. The complete test set-up is first calibrated
to compensate for recurrent errors. The transistor is then mounted on the test
resistive paste lPd/PdOt fixture, and measurements are carried out at 101 frequency points over the range
2-8 GHz. The scattering parameters of the transistor are then obtained by
computerised 'de-embedding' of the measured data, making use of the equiv-
alent circuit of the transistor test fixture. The values obtained agreed quite well
with the manufacturer's indications. The transistor was found to be practically
unilateral (very small values of s,,). It was noted that the values of r provided
by a commercially available program were erroneous, apparently due to incon-
flu sistent definitions.
Cr The CAD program used, Touchstone, is an analysis and optimisation pro-
ri
gram, but not a synthesis program. This means that a first approximation must
be provided as starting point. This step is carried out by using the Smith chart,
substrate
and conjugate match is provided at 5 GHz at the input and output. The first
a. r e s l s t l v e p a s t e b. use o f t h e adheslue
design proved unacceptable, as a transmission-line section was too short. A
deposlted by Cr or TI l a y e r
silk screening suitable matching structure was obtained on the second try and it is sketched in
Fig. 17.33.
Fig. 17.32 Resistors deposited on circuit The corresponding structure and values are then introduced into the CAD
program Touchstone (Section 17.5.2), and optimisation was carried out, requir-
paste (palladium-oxide mixture) may be deposited by silk screening (thick-film ing a flat gain within the range 9.5 dB < G < 10.5 dB over the range 3-5 GHz.
technique), and then cured. This process can only be applied to substrates that The microstrip losses are taken into account, as well as the effects of discontinui-
can withstand high temperatures. Alternately, the low-conductivity nickel or ties at T-junctions and at steps. The predicted gain of the amplifier, prior to
chromium layers required for adhesion on ceramic substrates (Section 17.4.8)
 1048 Computer-aided design of microstrip and triplate circuits Computer-aided design of microstrip and triplate circuits 7049

optimisation, is given in the curve a of Fig. 17.34, whereas the optimised value
appears in curve b.
The two circuits then realised, the layout was drawn with the help of the
program Micros (Section 17.6.4), and the mask was cut on a plotter. The
mask is cut on a Rubylith sheet. The mask is reduced and the circuit realised by
means of the photolithographic process (Section 17.4.9). After mounting and
connecting, the filter is measured on a vector network analyser; the transmission
factor is shown in Fig. 17.36. The passband falls right in the specified range, and
the 25 dB off-band requirement is also met.

Transistor

-
Fig. 1 7 . 3 3 Transistor amplifier with matching circuit

complete photolithographic process was carried out (Fig. 17.17), and the tran-
sistor was soldered on the microstrip circuit, under a microscope. Measurements
were then carried out on an automatic network analyser, that had been
previously calibrated, and the measured results are presented in Fig. 17.34,
curve c, showing the unoptimised amplifier, and curve d the optimised one. It
will be noted that the measured amphfication curve (d),while showing the same
general behaviour as the predicted one, lies a few dB lower and is shifted
somewhat towards lower frequencies.

17.8.2 Bandpass filter design

We wish to realise a bandpass filter for the range 3.8-4.2 GHz, that has to meet
I
the following requirements: IC
' ------ measurements
Passband ripple: 0.3 dB \\\ "'
Stopband attenuation: 25.0 dB at 4.6 GHz \c mpned
values
Substrate permittivity: 2.33 (relative)
Substrate thickness: 0.51 mm frequency (GHz)

These requirements are introduced into the CAD program Micros [63], which
determines the response for a Chebjrshev design, formed of anti-resonant series
cells and resonant shunt cells. The program determines the order of the filter Fig. 17.34 Amplifier gain
that is required to meet the specification, and the operator can check the a Predicted unoptimised gain
theoretical response before proceeding further with the filter design. The micro- b Predicted optimised gain
c Measured unoptimised gain
strip structure is then realised with broadside-coupled resonant strips. The d Measured optimised gain
program carries out the complete synthesis process, taking into account the
effect of open ends (Section 17.3). The dimensions of the strips and the spacings
17.8.3 Design of a miniature Doppler radar
are determined, and the layout is drawn (Fig. 17.35). It is possible to combine circuit and antenna functions on the same microstrip
The filter structure is in turn connected to the sides of the circuit, and a scale
substrate, and this was used to develop a small proximity detector, sensitive to
 1050 Computer-aided design of microstrip and triplate circuits Computer-aided design of microstrip and triplate circuits 1051

the presence of moving objects. A rectangular patch antenna resonates at
2.45 GHz and is also used as the frequency-sensitive element (filter) in the
feedback loop. A MESFET amplifier is connected across the resonator-
antenna. The low-frequency Doppler signal is detected directly at the MESFET
connections, taking advantage of its nonlinearity to provide signal mixing. The
mask of the circuit is shown in Fig. 17.37. While this circuit is quite simple, its
performance is limited: the dielectric permittivity and thickness that are ade-
quate for the design of a circuit are far from optimal as far as antenna radiation
is involved, and vice versa. A better approach would be to separate the two
functions by making use of a multi-layer structure (Section 8.12).

-
3.6-6.2 GHZ

Fig. 17.35 Bandpass-filter design

52 1 l o g MAG
REF 0 . 0 dB
A s.ade/

Fig. 17.37 Mask for miniatwised Doppler radar

17.9 Conclusion

Computer-aided design requires suitable mathematical descriptions of the ele-

ments, which should be accurate enough, but not too complex mathematically
[26]. The particularities of microwave printed circuits, outlined in Sections
i7.1.6-1-711.8, make the design particularly demanding in terms of accuracy. It
is seldom possible to define equivalent circuits that are at the same time general,
simple and easy to use, and whenever electromagnetic-field calculations are
required within an optimisation loop, the resulting computation time may well
prove prohibitive.
Highly accurate approximate expressions are now available for straight trans-
mission lines and for many discontinuities, sometimes with errors smaller than
START 3.000000000 M x fractions of a percent (Sections 17.2 and 17.3). However, these approximations
STCP 5.000000000 GHz are only valid over specified ranges of parameters, because they neglect radiated
Fig. 17.36 Measured bandpass filter response and surface waves: this is quite acceptable for microstrip lines, but obviously not
 7052 Computer-aided design of microstrip and triplate circuits
for radiating patches. Furthermore, the number of structures analysed until now
is limited, and proximity effects cannot be evaluated as accurately. The interac-
tions of a line with a nearby wall, with the edge of a microstrip substrate, with
a discontinuity or with a patch have not as yet been determined to a similar
degree of accuracy, and much remains to be done in this respect. In fact, the
possible combinations of components are almost infinite in number. The pro-
blem becomes particularly stringent for microwave- and millimetre-wave mono-
lithic integrated circuits, where different lines, circuits and components are
located very close to one another.
It must also be kept in mind that the physical dimensions are generally not
accurately known: substrate thickness is specified within 5% or even 10% by
some manufacturers. This means that one should measure it before actually
drawing the circuit pattern and cutting the mask. A given design is valid only
with the substrate material for which it was prepared - or an almost identical
one. Dimensional effects like under-etching also introduce errors that are much
larger than the ones encountered in the theory. Fortunately, systematic errors
can be compensated for.
Section 17.5 shows that there is no shortage of CAD software packages for
microstrip design. Some are rather broadband general-purpose analysis/
synthesis packages, while others have been specifically elaborated to design a
single kind of component. The task of the designer, who is supposed to select
one particular software is clearly not obvious, since many of the programs
apparently fulfil similar tasks.
To compare different software packages, one should actually use several of
them to develop the same circuit, and compare them in terms of the results
actually measured, their ease of operation, the computer time and memory
requirements, compatibility and portability, service, bugs etc. Many different
computer systems are presently available and in constant evolution, so that the
task is by no means simple. Cost, availability and service are non-technical
parameters that become most significant in actual operation.
The accuracy provided by a program is a most significant parameter for
selection; in particular, when optimisation processes are involved. Besser [7]
noted that the use of low-accuracy models in the initial microwave CAD
packages considerably slowed down their actual implementation within the
industrial market, but this problem has been corrected to a large extent in more
recent designs. It was also noted that many optimisation algorithms did not
perform as well as was claimed, and they did not always converge towards the
best possible solution.
In conclusion, one must remember that the CAD is a most valuable tool,
useful to a designer who understands its operation and is aware of its limi-
tations, but useless and wasteful when put in the wrong hands. It can significant-
ly increase the capabilities of a skilled operator, both in terms of time and
quality. On the other hand, the GA+A principle applies (garbage in, garbage
out): for instance, if the substrate properties are not accurately introduced, the
I
!
7053
Computer-aided design of microstrip and triplate circuits
results cannot be expected to be accurate (even though miracles sometimes
I happen - but generally at random!). An expert microwave designer will im-
mediately detect possible trouble areas, where lines are too close to one another,
or too close to an edge etc., but it would be practically impossible to program
a computer to detect and take care of all such possible combinations. In
microstrip and stripline circuit design, like in many other technological areas,
the computer does not replace the skilled designer.
1
17.10 Acknowledgments
The authors wish to thank Miss Anja Skrivervik, Mr. Lionel Barlatey and Dr.
Juan Mosig, who contributed to the realisation of circuits and to the elaboration
of parts of the text. The continuous support of the Swiss National Research
Foundation is gratefully acknowledged. Thanks are also extended to the firm
High Tech Tournesol, that granted permission to reproduce Figures originally
prepared for the intensive short course 'Initiation to Microstrip'.
I
17.11 References
I ABRAMOWITZ, M., and STEGUN, I. A. (1972): 'Handbook of mathematical functions'
(New York, Dover)
2 ABRIE, P. L. (1986): 'The design of impedance-matching networks for radio-frequency and
microwave amplifiers' (Norwood, MA, Artech House)
3 AKHTARZAD, S., and JOHNS, P. B. (1975): 'Dispersion characteristics of a microstrip line
with a step discontinuity', Electron. Letts., 11, pp. 31&311
4 ALTSCHULER, H. M., and OLINER, A. A. (1960): 'Discontinuities in the center conductor
of symetric strip transmission line' IEEE Trans. MTT-8, pp. 328-338
5 ANDERS, P., and ARNDT, F. (1980): 'Microstrip discontinuity capacitances and inductan-
ces for double steps, mitered bends with arbitrary angle, and asymetric right-angle bends'
IEEE Trans. MTT-28, pp. 1213-1217
6 BENEDEK, P., and SILVESTER, P. (1972): 'Equivalent capacitances for microstrip gaps and
steps' IEEE Trans. MTT-20, pp. 729-733
BESSER, L. (1985): 'What is the direction of commercial microwave technology?, Microwave
7
Syst. News, 15, pp. 65-67
R- - -.
ROCHTLER.
- - - - - U.. and ENDRESS, F. (1986): 'CAD program designs stripline couplers',
Microwaves and RF, 25, pp. 91-95
9 COHN, S. B. (1955): 'Problems in strip transmission lines', IEEE Trans. MTT-3, pp. 119-126
10 EASTER, B. (1975): 'The equivalent circuit of some microstrip discontinuities', IEEE Trans.,
MTT-23, pp. 655-660
II FARRAR, A,, and ADAMS, A. T. (1971): 'Computation of lumped microstrip capacities by
matrix-methods: rectangular sections and end effect', IEEE Trans, MTT-19, pp. 495-497
12 FARRAR, A,, and ADAMS, A. T. (1972): 'Matrix method for microstrip three-dimensional
problems', IEEE Trans.. MTT-20. pp. 497-504
13 FRANKEL, S. (1977): 'Multiconductor transmission line analysis' (Artech House, Dedham,
MA)
14 GARDIOL, F. E. (1984): 'Introduction to microwaves' (Artech House, Dedham, MA)
 7054 Computer-aided design of microstrip and triplate circuits
15 GARDIOL, F. E. (1986): 'Microstrip computer-aided design in Europe', IEEE Trans. M m -
34, pp. 1271-1275
16 GARDIOL, F. E. (1987): 'Lossy transmission lines' (Artech House, Norwood, MA)
17 GETSINGER, W. J. (1973): 'Microstrip dispersion model', IEEE Trans., MTT-21, pp. 34-39
18 GOPINATH, A., and EASTER, B. (1974): 'Moment method of calculating discontinuity
inductance of microstrip right-angle bends', IEEE Trans. MTT-22, pp. 880-883
19 GOPINATH, A., THOMSON, A. F., and STEPHENSON, I. M. (1976): 'Equivalent circuit
parameter of microstrip step change in width and cross junctions', IEEE Trans, MlT-24, pp.
142- 144
20 GOSLING, I. G. (1985): 'A new microwave CAD layout program'. IEE/EI2 Colloquium on
Computer Aided Design of Microwave Circuits, London, Nov. 1985
21 GUNSTON, M. A. R. (1972): 'Microwave transmission line impedance data' (Van Nostrand
Reinhold, NY)
22 HAMMERSTAD, E. O., and BEKKADAL, F. (1975): 'Microstrip handbook'. Norwegian
Institute of Technology, Trondheim, report ELAB STF44 A74169.
23 HAMMERSTAD, E. O., and JENSEN, 0. (1980): 'Accuratemodels for microstrip computer-
aided design', IEEE MTT-S International Microwave Symposium Digest, pp. 407-409
24 HOEFER, W. J. R. (1977): 'Theoretical and experimental characterization of narrow trans-
verse slits in microstrip', Nachrichrentechni. Zfr, 30, pp. 582-585
25 HOEFER, W. J. R. (1977): 'Equivalent series inductivity of a narrow transverse slit in
microstrip', IEEE Trans., MlT-25, pp. 822-824 -
26 HOFFMAN, G. R. (1984): 'Introduction to computer aided design of microwave circuits'.
Proceedings of the 14th European Microwave Conference, Likge, Belgium, pp. 731-737
27 HOFFMANN, R. K. (1983): 'Integrierte Mikrowellenschaltungen' (Springer, Berlin)
28 HORTON, R. (1973): 'The electrical characterization of a right-angle bend in microstrip line'
IEEE Trans., MTT-21, pp. 427-429
29 HORTON, R. (1973): 'Equivalent representation of an abrupt impedance step in microstrip
line', IEEE Trans., MlT-21, pp. 562-564
30 HOWE, H. (1974): 'Stripline circuit design' (Artech House Dedham, MA)
ITOH, T., MITTRA, R., and WARD, R. D. (1972): 'A new method for solving discontinuity
problems in microstrip lines'. IEEE-GMTT International Symposium Digest, pp. 68-70
JAIN, 0. P., MAKIOS, V., and CHUDOBIAK, W. J. (1971): 'Coupled-mode model of
-. .
- . Electron Letts. 7. DO.
dispersion in microstrio'. rr
405-406
--
JAMES, D. S. and TSE, H. S. (1972): 'Microstrip end effects', Electron. Letts., 8, pp. 46-47
JANSEN, R. H. (1985): 'The spectral domain approach for microwave integrated circuits',
IEEE Trans., Mm-33, pp. 1043-1056
JANSEN, R. H. (1986): 'LINMIC, a CAD package for the layout-oriented design of single-
and multi-layer MICs/MMICs up to mm. wave frequencies', Microwave J., 29, (2)
JANSSEN, W. (1977): 'Hohlleiter und Streifenleiter' (Hiithig, Heidelberg)
KAJFEZ, D., and TEW, M. D. (1980): 'Pocket calculator program for analysis of lossy
microstrip:, Microwave J., 23, pp. 39-48
KOMPA, G. (1976): 'S-matrix computations of microstrip discontinuities with a planar
waveguide model', Archiv Elekr. Uber~ragun~stech., 30, pp. 58-64
KOSTER, N. H. L., and JANSEN, R. H. (1986): 'The microstrip step discontinuity, a revised
description', IEEE Trans., MTT-34, pp. 213-223
LAVERGHETTA, T. S. (1984): 'Microwave materials and fabrication techniques', (Artech
House, Dedham, MA)
LEIGHTON, W. H., and MILNES, A. G. (1971): 'Junction reactance and dimensional
tolerance effects on X-band 3-dB directional couplers', IEEE Trans., MlT-19, pp. 814-824
MAEDA, M. (1972): 'An analysis of gap in microstrip transmission line', IEEE Trans.,
MTT-20, pp. 390-396
MARCH, S. L. (1984): 'Microwave circuit layout: a dynamic plot emerges', Microwaves and
RF, 23, pp. 59-161
Computer-aided design of microstrip and triplate circuits 7055
44 MARCUVITZ, N. (1951): 'Waveguide handbook'. MIT Rad. Lab. Series, 10, (McGraw-Hill,
New York)
45 MARSDEN, J. E. (1973): 'Basic complex analysis' (W. H. Freeman, San Francisco)
46 MEHRAN, R. (1975): 'The frequency-dependent scattering matrix of microstrip right-angle
bends, T-junctions and crossings', Arch. Elek. Uberrragungstech., 29, pp. 454-460
47 MEHRAN, R. (1976): 'Frequency dependent equivalent circuits for microstrip right-angle
bends, T-junctions and crossings', Arch. Elek. Ubcrrragungstech., 30, pp. 80-82
48 MENZEL, W. (1976): 'Frequency-dependent transmission properties of truncated microstrip
right-angle bends', Elecrron. Lett, 12, pp. 641
49 MENZEL, W. (1978): 'Frequency-dependent transmission properties of microstrip Y-
junctions and 120' bends', IEE J . Microwaves. Optics and Acoustics, 2, pp. 55-59
50 MENZEL, W., and WOLFF, 1. (1977): 'A method for calculating the frequency dependent
properties of microstrip discontinuities', IEEE Trans., MTT-25, pp. 107-1 12
51 Microwaves and RF (1984): 'The software selector', 23, pp. 70-95
52 MOSIG, J. R., and GARDIOL, F. E. (1977): 'Equivalent inductance and capacitance of a
microstrip slot'. Proceedings 7th European Microwave Conference, Copenhagen, pp. 455-459
53 ROHDE. U. L. (1985): 'Models and nonlinearities: major factors in microwave CAD soft-
ware', Microwave Syst. News, 15, pp. 123-143
54 SALERNO, M., and SORRENTINO, R. (1986): 'Planim: a new concept in the design of MIC
filters'. Electron. Letts, 22, pp. 1054-1056
55 SCHNEIDER, M. V. (1969): 'Microstrip lines for microwave integrated circuits', BeN Syst.
Techn J.. 48. pp. 1421-1444
56 SILVESTER, P., and BENEDEK, P. (1972): 'Equivalent capacitances of microstrip open-
circuits' IEEE Trans., MTl-20, pp. 51 1-516
57 SILVESTER, P., and BENEDEK, P. (1973): 'Microstrip discontinuity capacitances for right-
angle bends, T-junctions and crossings', IEEE Trans., MlT-21, pp. 341-346
58 THOMSON, A. F., and GOPINATH, A. (1975): 'Calculation of microstrip discontinuity
inductances', IEEE Trans., MTT-23, pp. 648-655
59 WALDRON, R. A. (1970): 'Theory of guided electromagnetic waves' (Van Nostrand Rein-
hold, London)
60 WHEELER, H. A. (1965): 'Transmission line properties of parallel strips separated by a .
dielectric sheet', IEEE Trans., MTT-13, pp. 172-185
61 WOLFF, I., KOMPA, G., and MEHRAN, R. (1972): 'Calculation method for microstrip
discontinuities and T-junctions', Electron. Letts., 8, pp. 177-179
62 YAMASHITA, E., and MITTRA, R. (1968): 'Variational method for the analysis of micro-
strip lines', IEEE Trans., MTT-16, pp. 251-256
63 ZURCHER, J. F. (1985): 'MICROS3 - A CAD/CAM program for fast realisation of
microstrip masks'. Proceedings IEEE MTT-S International Microwave Symposium, Saint
Louis, Missouri
64 ZtiRCHER, J. F., BARLATEY, L., and GARDIOL, F. E. (1986): 'Computer-aided method
to measure the permittivity of microstrip substrates'. Proceedings MIOP Symposium, Wies-
baden, Germany
 Resonant microstrip antenna
elements and arrays for aerospace
applications
A.G. Derneryd

18.1 Introduction

Several advantages associated with microstrip antennas, namely light weight,

low profile and structural conformity, make them ideally suited to aerospace
applications. A number of single microstrip-patch antenna elements and micro-
strip arrays for communication and radar systems are described in this Chapter.
The objective is to demonstrate practical designs and results, together with the
engineering tools used. Antennas for dual frequency bands, dual beams and
dual polarisations are considered. The frequency range covered is from L-band
to c-band.
Patches that are resonant in the dominant mode are usually employed.
However, in Section 18.2 different mode structures actually excited on a circular
disc are displayed using a liquid-crystal detector. This is an effective tool for
visualising the RF field excited on a microstrip antenna.
In Section 18.3 a dual-band circularly polarised patch antenna element is
presented. The antenna is a low-gain antenna used for data communication. It
was successfully flown on the Giotto space probe that encountered the comet
Halley at its recent appearance.
A monopulse array antenna for secondary surveillance radar applications is
discussed in Section 18.4. The array aperture consists of two rows of microstrip
antenna elements. A corporate feed network in stripline is mounted as an
integrated part behind the antenna aperture. A low-density foam is used as a
dielectric to achieve a lightweight design.
An array-antenna concept suitable for a space-borne imaging radar system is
described in Section 18.5. It is an integral feed-array structure with two separate
feeding networks to enable horizontal and vertical polarisations to be obtained
using a single antenna aperture. This antenna was designed and tested by Dr
Lars Pettersson of the National Defence Research Institute, Sweden.
 7058 Resonant microstrip antenna elements
18.2 Circular antenna element
Various design techniques for resonant microstrip antenna elements of simple
geometries are available. The simplest method, useful for predicting radiation
characteristics, is the cavity model. This model has been used to predict the
resonant frequencies of a circular disc. An antenna element was manufactured
and the different mode structures were visualised using a liquid-crystal detector.
The basic circular antenna element comprises a conducting circular patch on
a thin dielectric substrate backed by a ground plane, as shown in Fig. 18.1. The
antenna can be viewed as a cylindrical cavity bounded at its top by the patch
and at its bottom by the ground plane.
Fig. 18.1 Circular microstrip antenna element and liquid-crystal detector in wooden mount
(Courtesy: N. P. Kernweis and J. F. Mcllvenna. RADC)
The resonant frequencies are functions of the radius a of the patch, the
thickness h of the dielectric and the dielectric constant E,. However, an effective
radius a,, slightly larger than the physical one, is introduced to account for the
fringe field along the edge of the resonator cavity. The relation between the
effective and physical radii is given by [I]:
This expression is derived assuming a quasi-static field distribution. However,
it can be used to estimate the higher-order resonant frequencies as well.
Resonant microstrip antenna elements 7059
Thus, for each mode configuration, a resonant frequency is evaluated from
[I]:
where Xis a zero of the derivative of the Bessel function of order n, and c is the
velocity of light in free space.
In practice, only the first few zeros are of interest. These are listed in Table
18.1 in ascending magnitude of X for convenience, since no closed-form ex-
pression exists for the roots X of J;(X). For any given radius, the mode corre-
sponding to n = 1 has the lowest resonant frequency, and is known as the
dominant mode.
Table 18.1 Roots of J i ( X ) = 0
An antenna element was manufactured from a 1/16 in (1.58 mm) doubly-clad
Teflon glass-fibre board with a dielectric constant of 2.55. The actual radius of
the disc is 18.3mm but the effective radius is 5% larger, as calculated using eqn.
18.1. The 5 0 n feeding point is located at a radius 0.3 times the total radius of
the patch. There is a shorting pin at the centre of the patch in order to reject the
static mode (n = 0).
The calculated resonant frequency for the dominant mode is 2.8 GHz using
eqns. 18.1 and 18.2. The next two higher-order resonant frequencies are 4.7 GHz
and 6.5 GHz, similarly calculated. The theoretical E-field structures for these
three modes are shown in Fig. 18.2 [2].
It is possible to monitor mode changes as antenna patch dimensions or
frequency are varied, by using a liquid-crystal detector. This consists of an
encapsulated liquid-crystal solution that changes colour with applied heat.
When a resistive coating is sprayed onto the back of the crystal, RF field
components produce localised heating. Thus, a colour display results that varies
with the intensity of the field.
The liquid-crystal detector shown in Fig. 18.1 was placed on top of the
circular patch [3].Visual colour displays of the E-field, corresponding to the first
three modes, were observed using a liquid crystal with a resistive backing of
1100 Rlsquare. Black-and-white reproductions of the different mode structures
actually excited on the circular patch are shown in Fig. 18.3. To effectively excite
the second higher-order mode, the feed had to be repositioned closer to the
7060 Resonant microstrip antenna elements
Resonant microstrip antenna elements 7067
centre of the patch. The new feed position produces changes both in the
dominant mode and in the first higher-order mode behaviour, which can be
observed with the liquid-crystal detector.

Fig. 18.2 Calculated E-field structures on a circular microstrip antenna element at resonance
a Dominant mode, TM,
b First higher order mode, TM,
c Second higher order mode, TM,

The photographs were taken with about 1 W of power applied to the antenna
input. The liquid-crystal detector was directly on top of the patch. The loading
effect on the antenna is thus at its maximum, but the displays are always bright
and well defined. There is about 1% decrease in resonant frequency with the
detector at this maximum loading position. However, the liquid-crystal detector
is an effective aid in experimentally determining the relation between the feed
position and the mode structure actually excited on a patch antenna.

18.3 Dual-band circularly polarised antenna element

On 13 March 1986 the European Space Agency (ESA) spacecraft named Giotto
encountered the comet Halley at a distance less than 500 km from the nucleus.
 762 Resonant microstrip antenna elements
Resonant microstrip antenna elements 7063
Among the antennas on board was an S-band microstrip antenna for data
communication between Earth and the Giofto space probe [4]. The antenna is
a low-gain antenna providing an omnidirectional radiation-pattern coverage,
together with a second low-gain antenna of the helix type.

The principal requirement of these two antennas was to provide real-time

receiveltransmit of telemetry and telecommand signals during the geostationary
transfer orbit. The antennas operate on a right-hand circular-polarisation mode
in both frequency bands.
The microstrip antenna is located on the outer surface of the dust-protection
shield of the spacecraft. The gain requirement, including a 1 dB allowance for
cable loss, is given in Table 18.2. This provides the necessary radiation-pattern
overlap in the overall coverage pattern.

Table 18.2 Gain requirement including cable loss

Angular interval, deg Minimum gain, dBi

Fig. 18.3 Observed mode structures on a circular microstrip antenna element (Courtesy: N.
P. Kernweis and J. F Mcllvenna, RADC)
a Dominant mode (2.8 GHz) The microstrip antenna is a square radiating element resonant in its dominant
b First higher-order mode (4.7 GHz) mode. The element is etched on one side of a doubly clad PTFE board. The
c Second higher-order mode (6.5 GHz)
thickness of the laminate is 1/16in (1.58 mm) and the dielectric constant is 2.32.
 1064 Resonant microstrip antenna elements
The element is fed at two adjacent sides via a 90" hybrid to generate right-hand
circular polarisation.
Each feeding line between the patch and the hybrid contains a two-section
impedance transformer for double tuning [5]. The geometry of the transformer
and the radiating patch is shown in Fig. 18.4. The input impedance, as seen from
the 50R line, can be expressed in terms of the transformer parameters and the
input impedance ZAof the radiator alone as
Resonant microstrip antenna elements 1065
The impedance matching transformers and the hybrid are etched on a thin
doubly clad board which is contained in a stripline network bonded to the rear
side of the radiating element board. A layout of the different boards is shown
in Fig. 18.5. The transmission lines on the top board and in the stripline board
are connected via copper strips surrounded by mode-suppressing pins. The
fourth arm of the hybrid is terminated in a matched load placed inside the
stripline board.

The explicit expression is

Zin = Z* 2, + jZz tan (k21z)

Z2 + jZL tan (k21z)
where

z, = z, Z, + jZ,
tan (k, I,)
+
Z , jZA tan (k, I, )
and k, and k, are the propagation constants of the two transformer sections,
respectively.

Fig. 18.5 Layout of the complete microstrip antenna boards

The overall dimensions of the complete antenna, including a mounting frame,

rodlalor 2-section transformer 5011
Fig. 18.4 Dual-frequency microstrip antenna using an impedance-matching transformer
The characteristic impedances Z, and Z, of the transformer and the corre-
sponding lengths I, and I2 are the parameters to be chosen to fulfil the matching
conditions
The single and double prime signs refer to the two frequencies to be matched.
In this case the input impedance is set to 50R, i.e. the characteristic impedance
of the feed line. The input impedance of the patch alone is found from measure-
ments at the two frequencies to be matched.
are 139mm x 139 mm and the weight is 0.2 kg. The front surface is covered
with a thin layer of black thermal paint with low electrical conductivity to
prevent electrostatic charging in space. A picture of a prototype antenna is
shown in Fig. 18.6.
The measured return loss of the antenna including the cable, is presented in
Fig. 18.7. The voltage standing-wave ratio is better than 1.2:1 at both the
telecommand frequency (2116.7MHz) and at the telemetry frequency
(2298.7 MHz).
Recorded antenna patterns at three different cuts through the broadside
direction are plotted in Fig. 18.8. The antenna is mounted on a ground plane
to simulate the dust-protection shield. The specified gain as a function of the
angle is shown as a broken line.
The dual-band circularly polarised microstrip antenna was successfully flown
on the Giotto spacecraft, and it was proved to fulfil all the specified data.
 1066 Resonant microstrip antenna elements Resonant microstrip antenna elements 1067

Fig. 18.6 A prototype microstrip antenna for the Giotto space probe

-50 -25 0 25 50 75
angle from broadside, degrees

Fig. 18.8 Recorded antenna patterns at 2298.7 MHz for three different cuts (0". 4 5 , 90")

frequency, MHz

Fig. 18.7 Measured return loss of the antenna including the cable Fig. 18.9 Monopulse microstrip-array antenna
 7068 Resonant microstrip antenna elements
18.4 Monopulse-array antenna
In this Section an L-band planar-array antenna of the monopulse type intended
for a secondary surveillance radar is described. Besides the usual sum radiation
pattern, the antenna also generates a difference radiation pattern in one plane.
In the current application the latter is used to separate signals received in the
main beam from those received in the sum sidelobe region. This requires that
the difference sidelobe level exceeds the sum sidelobe level at all angles.
The antenna array consists of an aperture with 2 x 6 rectangular patches, as
shown in Fig. 18.9. A stripline feed network is placed on the rear side of the
aperture. The size of the antenna is 650 mm x 1300mm and the weight is 10 kg.
Mechanically, the antenna is of sandwich design with foam as a spacer to
achieve low weight and low losses. A cross-section of the aperture and the
stripline feed network is shown in Fig. 18.10. The radiating patches, the trans-
mission lines and the ground planes are supported by glass-fibre-reinforced
plastic (GFRP) skins. Standard techniques are used to etch the radiating ele-
ments and the feed lines. The final assembly is done by bonding the different
layers together in a step-by-step procedure.
radiating elements
GFRP skin
foam
GFRP skin
ground Plane -.
foam
feed network - GFRP skin
foam
GFRP sktn
ground plane ---,
Fig. 18.1 0 Cross-section of the monopulse-array antenna
The basic radiating element used in the array is the rectangular patch with
dimensions 90mm x 121 mm, resonant in the dominant mode. Two such
elements, spaced 258 mm and fed in series from a single feed point, are employed
to form the E-plane radiation pattern. The element pair is fed at the lOOQ
input-impedance point of one of the patches. The second patch is connected to
the first one by a half-wavelength microstrip line. The resulting input impedance
of the pair is thus 50Q since the elements act in parallel when transformed to
the common input port. The H-plane pattern is generated using six such pair of
elements. The centre-to-centre spacing is 175mm.
Resonant microstrip antenna elements 1069
A stripline corporate feed network is connected to the microstrip elements
from underneath. Unequal power dividers of the Wilkinson type are used to get
a 25 dB Chebyschev amplitude taper in the H-plane. A layout of the corporate
feed network is given in Fig. 18.11. The first power divider, however, is a rat-race
hybrid with equal power split to feed the two antenna halves. The sum signal
of the two identical antenna halves is formed at one port of the hybrid, while
the difference signal is formed at an opposite port.
The far fields are calculated assuming a two-slot model for each patch [6].
However, this model does not take into account the effect of the finite ground
plane. The edge eiYects are included by adding the field components diffracted
from the edges to those radiated directly from the slots [7]. This effect is
particularly noticeable at angles close to endfire and in the backward direction.
Fig. 18.11 Layout of the corporate feednetwork (Courtesy:J. P.Starski, Chalmers University
of Technology)
The radiation mechanism in the E-plane of a single slot in a finite ground
plane is illustrated graphically in Fig. 18.12. The pattern is calculated by
summing three rays; i.e. the direct geometrical-optics field, the singly edge-
diffracted field and the doubly edge-diffracted field. The direct field is obtained
by assuming the ground plane to be infinite in extent. Using the co-ordinate
system of Fig. 18.12, the normalised direct field in the E-plane from each slot
can be written as
sin (nh& sin 8) - e-jk"
Eoo = e
~ h &sin 8 s
where k is the propagation constant in free space, h is the slot width normalised
to the free-space wavelength, E, is the relative dielectric constant of the substrate
and s is the distance from the slot centre to the observation point. The slot width
is usually assumed to be equal to the substrate thickness.
The singly diffracted field from each edge, generated from the same slot, is
given as the incident field at the edge times a hard-boundary diffraction coef-
ficient neglecting the dielectric effect [a]:
 1070 Resonant microstrip antenna elements Resonant microstrip antenna elements 1077

The far field diffracted from the edges is thus expressed as ( + sign refers to edge where
1 and - sign to edge 2)
n + P
C + ( p , n) = cot -
E: = + e sinah&
(xh&) e-jkdt
-Dr(e
Jz
= d,, Pi, n = 2) 2n
a - P
C - ( p , n) = cot -
2n

g+ (P) = 1 + cos (/I - 2 n n N + )

g-(P) = 1 + cos ( p - 2 n n N - )

Fig. 18.12 Co-ordinate system and geometry in the E-plane of a slot in a finite groundplane

with

. . . . . . b . o . l l l l l . a I

-90 -60 -30 0 30 60 90

angle from broads~de~degrees

Fig. 18.13 Recorded and calculated E-plane radiation patterns of an element pair at
1060 MHz
and D' is the reflected diffraction coefficient given in eqn. 11.716 of Reference
8 and reproduced here:
in which N + and N - are the integers that most closely satisfy the equations
2nnNf -P = n (18.15)
2nnN- -P = - n.
1072 Resonant microstrip antenna elements
Since the edge is relatively close to the slot, cylindrical-wave propagation is
assumed. The doubly diffracted field is expressed as in eqn. 18.9. However, the
incident field in this case is the singly diffracted field from the other edge
assuming cylindrical-wave propagation.
A calculated radiation pattern along the E-plane of an element pair is plotted
in Fig. 18.13. The far field is computed using eqns. 18.8 and 18.10 referred to
the same phase centre and assuming four radiating slots. The double edge-
diffracted fields are also included; e.g. fields diffracted from a first edge and
incident on a seond edge. Diffracted rays on both sides of the aperture are
considered. The corresponding measured pattern at 1060 MHz is included in
Fig. 18.13 as well. The half-power beamwidth is 27'.
-90 -60 -30 0 30 60 90
angle from broads~de.degrees
Fig. 18.14 Recorded and calculated sum-and-difference radiation patterns along the H-
plane at 7060MHz
Recorded H-plane sum and difference patterns of the antenna are presented
in Fig. 18.14. Calculated patterns using the method outlined in Reference 7 are
also included as broken lines. The sum sidelobe level is below - 22 dB and the
half-power beamwidth is 16'. The gain measured at the sum input port is 17 dBi
and the losses are estimated to 1.3 dB. The difference sidelobe level exceeds the
sum sidelobe level as required.
Resonant microstrip antenna elements 7073
18.5 Dual-polarised-array antenna
Part of the active microwave instrumentation on board the European Remote
Sensing Satellite 1 (ERS-I) developed for the European Space Agency is a
scatterometer. It is an incoherent radar system at 5.3 GHz which measures the
back scattering of the ocean. The values are used to calculate wind speed and
wind direction.
Fig. 18.1 5 Series-parallel fed array of square microstrip patches
The scatterometer antenna is a narrow-band (2 MHz) planar array generating
a fan beam. As an alternative to the current slotted-waveguide antenna solution
a microstrip array is presented. The advantage of such a design is that the same
aperture can be utilised for two orthogonal polarisations simultaneously [9].
A section of the microstrip array is shown in Fig. 18.15. It consists of 6 x 6
square microstrip patches arranged in 12 identical linear sub-arrays with three
elements each. Every element is excited in two orthogonal modes from two
separate feed networks.
The feed point of the network for vertical polarisation is marked with an A
on the central feeding line in Fig. 18.15. The feed network for horizontal
polarisation is split into two symmetrical parts. These are fed 180 out of phase
7074 Resonant microstrip antenna elements
at points B and C on the outer feeding lines in order to generate a broadside
beam.
Each square element is fed at the mid-point of the edge of the element. This
is the low-impedance point of the orthogonal mode in each case, thus maximis-
ing the isolation between the two polarisations. This cross-coupling is further
surpressed by the symmetric feeding of the array.
The three elements in each sub-array are fed by resonant feed networks. In
the vertical-polarisation case the elements are fed from a common microstrip
line of constant-impedance level at positions spaced one wavelength apart. The
total input admittance of a sub-array is thus the sum of all patch admittances
first transformed through a quarter-wave section. This is matched to the rest of
the feeding structure by two quarter-wave transformers.
In the horizontal-polarisation case the elements are fed in series separated by
half-wave transmission lines. The input admittance of the three series patches is
the sum of the patch admittances transformed through a quarter-wave section.
The networks to feed the 12 identical sub-arrays are non-resonant to make
them less sensitive to internal reflections and to enable future beam shaping by
feeding the sub-arrays with different amplitudes and phases. The feed networks
for the two polarisations are basically identical since the input impedance of a
three-element sub-array for vertical polarisation is chosen to be twice the input
impedance of the sub-array for horizontal polarisation.
Table 18.3 Impedance levels and dimensions of the microstrip lines in the
feed network
Line Impedance, Length, mm Width, mm
The characteristic impedances of the feed lines are given in Table 18.3,
together with the corresponding lengths and widths. The effective dielectric
constants and the line widths have been determined using a low-frequency
Resonant microstrip antenna elements 1075
(TEM) method [lo]. The radiation resistance at each end of the patches is
assumed to be 360R, a value found from measurements on test circuits.
A microstrip substrate of the type shown in Fig. 18.16 is used. The base plate
is a carbon-fibre-reinforced plastic (CFRP) skin to provide high mechanical
stiffness and low thermal expansion. The metallic coating forming the ground
plane is realised by an aluminium foil of 0.1 mm thickness. The top dielectric
skin, carrying the patches and the feed network, is made from three layers of
Kevlar-epoxy prepreg. A prefabricated 0.005 mm copper foil on an aluminium
carrier is cured to the skin in a vacuum press. The aluminium carrier is then
removed by means of an alkaline etching solution and the etching of the pattern
is carried out by the standard photo-etching technique. Spacers in CFRP are
used to support the dielectric skin, and they are positionedat locations with low
electric field.
c~rcutt spacer
dielectrtc layer
l g r o u n d plane
Fig. 18.16 Cross-section of the microstrip substrate
The dimensions of the radiating patches and of the microstrip transmission
lines are functions of the substrate thickness and the dielectric constant. The
thickness has been determined from a minimum-loss point of view. The vari-
ation of the different types of losses as a function of the substrate thickness at
a fixed impedance level can be derived, for example, from References 11-13. The
proportional factors are determined by calculating the various losses for dif-
ferent thicknesses. The total losses in this particular design and for the substrate
used, given a fixed impedance level, depend on the height h (in mm) approxi-
mately as
Conductor loss 0.2/h dB
Dielectric loss 0.3 dB
Radiation and surface wave losses 0.05h2dB
The minimum-loss substrate thickness calculated from the above is 1.3 mm. The
calculations were made on the assumption of a uniform dielectric. However, a
suspended Kevlar skin is used, and instead the dielectric loss decreases slightly
with increasing thickness. The actual optimum is therefore slightly larger than
found above. A thickness of 1.6 mm is used, i.e. 0.4 mm for the Kevlar skin and
1.2 mm for air with sparsely scattered spacers.
A photograph of a section of the antenna is shown in Fig. 18.17. The size of
 7076 Resonant rnicrostrip antenna elements Resonant rnicrostrip antenna elements 7077

the radiating patches is 23.5 mm x 23.5 mm and the element separation is

23.7mm. Two recorded radiation patterns of the array are presented. A cut
along the H-plane for the vertical-polarisation case is plotted in Fig. 18.18. The
sidelobe level is about -13dB, as expected from a uniform amplitude distri-
bution. The corresponding pattern cut for the horizontal-polarisation case
along the E-plane is shown in Fig. 18.19. The measured sidelobe levels are
somewhat unsymmetrical, mainly owing to an amplitude unbalance at the input
ports. The half-power beamwidth is 10" for the measured patterns. The isolation
between the two polarisation ports is 31 dB, and this will improve with a better
amplitude balance.

Fig. 18.17 Dual polarised rnicrostrip array antenna

a n g l e from b r o a d s ~ d e ,degrees

' Fig. 18.19 Recorded radiation pattern along the E-plane at 5.3GHz. Horizontal polarisation

18.6 Concluding remarks

The procedures employed for designing the microstrip antenna elements and
- 401 I, \ 8 I arrays make use of transmission-line and cavity models coupled with experi-
-180 620 -60 0 60 120 180 mental iterations of the initial designs. This is adequate for antennas with
a n g l e from broadside, degrees
moderate requirements in terms of bandwidth and sidelobe levels. The effect of
a finite ground plane on the radiation-pattern predictions is included by adding
fields diffracted off the edges to the direct radiated fields. By doing this, the
Fig. 18.18 Recorded radiation pattern along the H-plane at 5.3G M Vertical polarisation
1078 Resonant microstrip antenna elements
radiation patterns are more accurately estimated in the backlobe and wide-angle
regions.

Chapter 19
18.7 References
Applications in mobile and satellite
1 SHEN, L. C., LONG, S. A,, ALLERDING, M. R., and WALTON, M. D.: 'Resonant
frequency of a circular disc, printed-circuit antenna', IEEE Trans., 1977, AP-25, pp. 595-596 systems
2 WATKINS, J.: 'Circular resonant structures in microstrip', Electron. Lett., 1969, 5, pp.
524-525
3 KERNWEIS, N. P., and McILVENNA, J. F.: 'Liquid crystal diagnostic techniques, an
K. Fujimoto, T. Hori, S. Nishimura and K. Hirasawa
antenna design aid', Microwave J., Oct 1977, 20, pp. 47-51, 58
4 BENGTSSON, P., HALM, R., and CRONE, G. A,: 'The Giotto spacecraft antenna subsystem
design', IEEE Int. AP-S Symp. digest, 1986, pp. 711-714
5 DERNERYD, A. G.: 'Microstrip disc antenna covers multiple frequencies', Microwave J.,
May 1978, 21, pp. 77-79
19.1 Introduction
6 DERNERYD, A. G.: 'Linearly polarized microstrip antennas', IEEE Trans., 1976, AP-24, pp.
846-851 Mobile communications often require antennas having small size, light weight,
7 HUANG, J.: 'The finite ground plane effect on the microstrip antenna radiation patterns', low profile and low cost. Microstrip antennas (MSA) are a type of antenna
lEEE Trans., 1983, AP-31, pp. 649-653 which can meet these requirements, and various MSAs have so far been de-
8 BALANIS, C. A,: 'Antenna theory: Analysis and design', (Harper & Row, 1982), pp. 502-514
9 DERNERYD, A. G.: 'Microstrip array antenna', Proc. 6th European Microwave Conf., 1976, veloped and used for mobile communication systems. The practical applications
pp. 339-343 for mobile systems are in portable or pocket-size equipment and in vehicles.
10 SMITH, J. I.: 'The even- and odd-mode capacitance parameters for coupled lines in suspended UHF pagers, manpack radars, and car telephones are typical of those. Base
substrate', IEEE Trans., 1971, MTT-19, pp. 424-431 stations for mobile communications need antennas with sector radiation pat-
11 DENLINGER, E. J.: 'Losses of microstrip lines', IEEE Trans., 1980, MlT-28, pp. 513-522 terns. Small, simple antennas are also favoured, since the antenna tower built
12 JAMES, J. R., HALL, P. S., WOOD C., and HENDERSON, A.: 'Some recent developments
in microstrip antenna design', IEEE Trans., 1981, AP-29, pp. 124-128 for the base station can then be smaller and need less support for the weight.
13 LEWIN, L.: 'Spurious radiation from microstrip', Proc. IEE, 1978, 125, pp. 633-642 Ships and aircraft also demand small, lightweight antennas, and sometimes
conformal structures are desirable to allow antennas to be mounted flush on the
body of the moving vehicle. MSAs are considered to be suitable for such
conditions and many antennas have been developed and installed on ships
and aircraft. Examples are a marine radar antenna and a surveillance radar
antenna.
In satellite communications, circularly polarised radiation pattefns are re-
quired and MSAs of either square or circular patches with one or two feeding
points can be used for generating the circular polarisation. Beam shapes such as
a sector beam and a multi-beam can be produced by an array of MSA elements,
which can be easily fabricated to form a flat structure, even though thousands
of elements are used, by means of a photo-etch technique applied to the
copper-clad dielectric substrate. A flat structure can be a feature of an MSA
array used for receiving satellite broadcasting. Parabolic antennas are very
popular for receiving broadcasts from satellites, but replacing them by small,
flat antennas is preferable, especially for the home use. A large parabolic
antenna, with the primary feed placed in front of the reflector, needs a wide area
for installation, while a small, flat antenna can possibly be mounted flush on the
wall of the house or even placed inside the window at home, depending on the
field strength at the receiving environment. Several types of flat antennas have
 1080 Applications in mobile and satellite systems
been developed: antenna elements used are the crank type, four-element square
patch of either single- or two-point feed, etc..
In this Chapter, various types of MSAs which have been developed and
applied in mobile and satellite systems are described.
19.2 Mobile systems
19.2.1 Design considerations
MSAs used for land, maritime and aeronautical mobile systems are described
in this Section. Antennas for land-mobile base-stations are also included. In
these applications, the features of MSAs, such as flat structure, lightweight and
compactness, which are generally required for mobile systems, are fully taken
into account in the design. Antennas used for cellular mobile base-stations are
described in the first part of this Section. In the cellular mobile phone system a
service zone covered by a base-station antenna is divided into small sectors to
increase the use of radio-frequency channels. The antennas are then designed to
have a sector or a multi-beam pattern to form three to six zones in a 360'
coverage. These patterns can be synthesised by an antenna array and MSAs are
employed as being the most suitable array elements. Using antennas such as a
dipole and a paraboloidal or corner reflector, beam shaping cannot easily be
realised. In addition, the utilisation of MSA elements for the base-station
antenna is advantageous, because the array can be made in flat structures and
of light weight; and the antenna tower construction is easier and less costly from
installing the MSA array on the tower than by using conventional heavy
metallic antennas.
A base station of small power, applied, for example, to home security, may
represent an interesting problem. To overcome severe multipath fading in signal
reception during indoor operation, the system is designed to receive both electric
(E) and magnetic (H) field components at the same time, thereby obtaining
smooth signal output. For this purpose, a system with a combined slot and
rectangular patch antenna has been developed to receive both E- and H-field
components simultaneously. Those antenna elements are placed flush on the
panel of the equipment box, which can be hung on the wall of a room when used
operationally.
The second part of this Section describes the applications of MSAs to the
automobile. Two types of antennas have been introduced: one is an antenna
which can be used in the interior of an automobile and the other is an antenna
which is installed on the roof of a vehicle. Both should be a thin and compact
antennas, and MSAs can satisfy such requirements. Two antennas are located
in the interior: one on the dashboard and the other on the back of the rear seat,
for the purpose of the diversity performance. Successful operation of the diver-
sity system has been reported after an experiment was undertaken in the Tokyo
metropolitan area.
Applications in mobile and satellite systems 1087
When an antenna is placed on the roof of a car, the radiation pattern usually
tends to be directed upwards; however low-angle radiation is preferred for
urban mobile operations. An annular slot is used to obtain a relatively low-angle
pattern when installed on the roof of a car. Also its flat structure is preferable,
because a thick or bulky antenna cannot be mounted on the roof.
The next Subsection introduces antennas for pedestrians. This application
requires special consideration, especially when the equipment is used in an
operator's pocket or near a human body. The effect of adjacent materials on
antenna performance can scarcely be avoided; however, by using an antenna
excited by a magnetic current, instead of electric current, the degradation of
antenna performance due to adjacent materials can be reduced. The ground
plane of an MSA acts as a type of shield against adjacent materials such as
circuit components and other metallic materials, and yet the image of the
magnetic current, with respect to the ground plane, will provide enhanced
radiation in front of the MSA element, which results in reducing the degra-
dation of the radiation. Four types of modified quarter-wavelength MSAs used
for the pocket-size equipment are introduced.
The remainder of the Section deals with radar antennas. The first is a marine
radar antenna. Antennas used on ships are mainly either reflector antennas or
slot arrays. An MSA array, formed as a flat structure and placed on a rotating
pedestal similar to the usual radar antennas, is introduced. The MSA applica-
tion reduces the weight, and makes the antenna rotation simple, smooth and
tolerable over a long period of time. This is very important for use in small
boats, because a heavy and large antenna with bulky counterweight cannot be
mounted on the top of a mast.
19.2.2 Base stations
(a) Sector-beam array: The microstrip-antenna concept is applied to base-
station antennas for land-mobile communications to create small, simple and
low-cost antennas.
The configuration of base-station antennas which have sector beams is shown
in Fig. 19.1 [I]. Fig. 19.l a shows a 60"-120" sector-beam array [2], and Fig. 19.lb
shows a 180' one [3].
The former is composed of two or four sub-arrays, and the sub-array is
composed of 2 x 4 microstrip patch elements. The patch element is a broad-
band microstrip antenna with a parasitic element. This antenna radiates vertical
and horizontal polarisation in the 900 MHz band. It has 60"-120" sector-beam
patterns in the horizontal plane. For realising these patterns in the horizontal
plane, a two-element excitation method is used, in which two elements are
excited with different amplitudes and out of phase. The feeder section is com-
posed of two hybrids and a phase shifter. The amplitude ratio of the two
elements can be changed by controlling a phase shifter. The measured radiation
patterns are shown in Fig. 19.2. The gain deviation is less than 4 dB within the
 7082 Applications in mobile and satellite systems
60"-120" beam. The measured VSWR is less than 1.5 over a frequency band-
width of 9%.
The other array (shown in Fig. 19.lb) is composed of 8 or 16 printed dipole
elements with a corner reflector. This antenna radiates vertical polarisation in
the 900 MHz band. It has 180" sector-beam patterns in the horizontal plane.
Applications in mobile and satellite systems 7083
(b) Multi-beam array: The configuration of the multi-beam base-station
antenna for land-mobile communications is shown in Fig. 19.4 [4]. This antenna
is composed of 8 x 8 microstrip patch elements. The patch element is a broad-
band microstrip antenna with a parasitic element. The antenna operates in

Parasitic,
Element

-30 1
-180 -90 0 90 180
Angle (')

Fig. 19.3 Measured azimuthal radiation patterns of the 180' sector-beam array (@ 1988
IEICE)
--- 860 MHz
-900 MHz
- - - 940 MHz
. . . . Cal (900 MHz)
Fig. 19.1 Configuration of the sector-beam base-station antenna (Courtesy: NTT, Japan)
( a ) 60-120' sector-beam array (V/H polariation) Input/Output
( b ) 180' sector-beam array (V polarisation)

Butler matrix

Fig. 19.4 Configuration of the multi-beam base-station antenna (@ 1988 IEICE)

Fig. 19.2 Measured azimuthal radiation patterns of the 60- 12LT sector-beam array (@ 1988
IEICE)
( a ) amplitude ratio, 1 :0.27
(6)amplitude ratio, 1 : -0.33.
- V polarisation, - - - H polarisation.
The corner angle of the reflector is optimised to realise these patterns in the
H-plane. The measured radiation patterns are shown in Fig. 19.3. Here, the
corner angle is 26O0, and the reflector size is 0.62 wavelength width. The
measured VSWR is less than 1.2 over a frequency bandwidth of 9%.
vertical/horizontal polarisation, and has eight beams within 120" area. The eight
beams are switched by a Butler matrix circuit. For simplicity and low cost, the
microstrip antenna elements and the feed circuits are arranged and etched on the
same surface. The measured radiation patterns are shown in Fig. 19.5. The
measured gain is more than 22 dBi at 900 MHz, and the circuit loss is less than
2 dB.
(c) E-H antenna: Conventional or shortened dipole antennas have been
used as base-station antennas for small indoor communication systems. One of
the biggest problems is that indoor communication is sometimes interrupted
 7084 Applications in mobile and satellite systems Applications in mobile and satellite systems 7085
when a receiving antenna happens to be located near the minimum points of a by a standard dipole antenna, but the solid line shows little fluctuation in the
standing-wave distribution of electric fields, as shown in Fig. 19.6. combined signals received by the E-H antenna.
Several techniques, such as space diversity or frequency diversity, have been In addition, the present antenna can receive both vertically and horizontally
introduced to the communication system to solve the problem. However, the polarised waves using a proper signal combiner. This will be quite useful for
system may be rather bulky and expensive for most small base stations. Indoor communication.

wide strip

microstrip

Angle (')
out
Fig. 19.5 Measured azimuthal radiation patterns of the multi-beam base-station antenna (@ I
1 9 8 8 IEICE)
- measured Fig. 19.7 Configuration of the E-H antenna (@ 1988 IEEE)
--- calculated

antenna present .................................................

antenna
....................
.....-....
... ,,.........*.
standing metal 1 ic ...
waves -1 1
i I
standard !
,
: ... ,,,,'
'*
dipole
.\! .
..
t i
y
floor measured
Fig. 19.6 Idealised standing- wave distribution (@ 1 9 8 8 IEEE)

Ito et al. [5] have proposed a simple printed antenna, as shown in Fig. 19.7,
composed of a half-wavelength slot, a wide strip and a signal combiner. The
antenna can receive transmitted signals through both magnetic and electric
fields, so that there will be almost no interruption from standing waves. The
antenna is referred to as an E-H antenna.
The slot and the strip are constructed on both sides of a printed circuit board.
Received signals from the slot and the strip are combined directly and sent to
a receiver. A reflector is placed under the ground plane of the substrate to form
a unidirectional pattern.
Fig. 19.8 shows an example of received-signal fluctuations for varying anten-
na location. Each maximum value was normalised to unity. D denotes the
distance of the antenna from a metallic wall in wavelengths (the frequency was
320 MHz). The dotted line shows the standing wave of the electric field received
Fig. 19.8 Example of received-signalfluctuations (@ 1 988 IEEE)
Each maximum value was norrnalised to unity
-E-H antenna
- - - standard dipole antenna
19.2.3 Wheeled vehicles
( a ) Cabin antenna: The circular microstrip antenna is used for cabin-anten-
na applications [6].It is necessary for a cabin antenna to have 1 dB more gain
than a haif-wavelength dipole, based on the assumption that the degradation in
the average received power caused by installing an antenna inside the vehicle is
3 dB, and the improvement caused by receiving both dominant and cross-
7086 Applications in mobile and satellite systems
polarisation is 2 dB. The cabin antenna is a broadband microstrip antenna with
a parasitic element, as shown in Fig. 19.9. The measured radiation patterns are
shown in Fig. 19.10. The relative gain compared with a half-wavelength dipole
is about 1.5 -- 2 dBd for a microstrip antenna with a parasitic element.
Applications in mobile and satellite systems 7087
switching the capacitance loaded on the slot by the bias supply. This type of
pattern can be considered for a vehicular antenna system to reduce the mul-
tipath fading in urban mobile communications.
1

Fig. 19.9 Outer view of cabin antenna (Courtesy: NTT, Japan)

( b ) Rooftop (annular slot): The annular slot [7, 81 is a candidate for use in
vehicular antennas for mobile communications, since it can radiate power at low
elevation angles. In urban mobile communications, incident waves to mobile
stations come mostly from directions having low elevation angles, - about 30"
up from the horizontal plane. The antenna structure is shown in Fig. 19.1 I . The
radiation pattern depends on the radius of the slot, and its variation is shown
in Figs.19.12 - 19.14, where the antenna radius is taken as a parameter. Fig.
19.15 illustrates a way of mounting an annular-slot antenna on the roof of a
vehicle. An example of an annular-slot antenna with k, = 0.5 (k = 2n/l, where Fig. 19.10 Measuredradiation patterns of cabin antenna (Courtesy: NTT. Japan)
I is the wavelength) is shown in Fig. 19.16. (a) Vertical plane (E- lane)
The radiation pattern can be controlled by loading a capacitor on a slot as (b) Vertical plane ( ~ i p l a n e )
(c) Horizontal plane
shown in Fig. 19.17 and by varying its capacitance electronically. One method
is to make the radiation pattern asymmetric. An almost one-sided radiation
pattern is obtainable, and a typical example is shown in Fig. 19.18, where the 19.2.4 Railways
antenna parameter k, = 1.5, the loaded reactance is - 80 ohms, and the operat- A train antenna is required to have wide beamwidth, with a low profile and
ing frequency is 1.5 GHz. The radiation pattern can be rotated electronically by small structure. TOsatisfy these requirements, a short-circuit rectangular micro-
 1088 Applications in mobile and satellite systems Applications in mobile and satellite systems 1089
z

X / ' f e e d point

"
10 ABAIV. C 10 WOIV.

ielectrlc material
0 0'

10 dB/DIV. d 10 W l V .

Fig. 19.11 Annular-slot antenna structure and the co-ordinate system Fig. 19.13 Radiationpatterns
( a ) Antenna structure and co-ordinate system
( b ) Feed point

b 10 dWOIV.
d 10 WON.

Fig. 19.12 Radiation patterns Fig. 19.14 Radiation patterns

 7090 Applications in mobile and satellite systems 1091
Applications in mobile and satellite systems
strip antenna is used [9]. A train antenna is shown in Fig. 19.19; it is 2300 mm
in width, 350 mm in height and 90 mm in depth. The antenna is composed of
the transmitting antenna and the receiving antenna. Each antenna is a four-
patch array, and it operates at 413 MHz and 452 MHz. The measured radiation

Fig. 19.15 Concept of mounting an annular-slot antenna on the roof of a vehicle

Fig. 19.18 Measured radiation patterns

Fig. 19.16 Example of annular-slot antenna element

&N#M~B&Y~ rodlotor
i,i $5 t'i $5 j J-
, feeder

1 ! mn ! Dower d i v i d e r

Fig. 19.19 Configuration of train antenna (@ 1988 IEICE)

reactance
component 'feed point
patterns are shown in Fig. 19.20. The half-power beamwidth is more than 34'
in the horizontal plane, and more than 153" in the vertical plane. The measured
Fig. 19.17 Dimensions of antenna
gain is 5.8 dBd at 452 MHz, and the VSWR is less than 1.4.
 7092 Applications in mobile a n d satellite systems Applications in mobile a n d satellite systems 7093

I units in the VHF/UHF bands. A practical example of one of these antennas

mounted in a pocket-size pager at the 900 MHz band is shown in Fig. 19.21 [IO].
1 The gain required for this type of pager antenna is about -4 dB compared to a

Fig. 19.20 Measured radiation patterns of train antenna (@ 1988 IEICE)

-measured
- - - calculated

19.2.5 Pedestrian
Antennas for VHF/UHF hand-held portable equipment, such as pagers, port-
able telephones and transceivers, must naturally be small in size, light in weight,
and compact in structure. Some of this equipment, especially that used most of
the time in an operator's pocket, demands either flush-mounted or built-in
antennas. It is well known that the smaller the antenna size, the lower the
antenna efficiency. There is a growing tendency for portable equipment to be
made smaller and smaller as the demand for personal communication rapidly
increases, and the development of hand-held or hand-portable units has become
urgent. Requirements on antenna performance for such small equipment are
becoming increasingly severe, since the antenna performance should not be
significantly degraded as the size becomes smaller. The microstip antenna is one Dimensions 1 8 5 x85 x 1 9 mm
of the most preferable for small equipment, especially when a flush-mounted or Weight ( 140g
built-in antenna is required. Since the microstrip antenna can be made with a Display 12 digits numerals
very thin and compact structure, it can easily match various types of portable Sensitivity 1 5pVh
units. One possible problem to be considered when using a microstrip antenna Battery life 1 3 months
is its narrow bandwidth, which is usually only a few percent, depending on the Spurious I -40dB
thickness of the antenna and the manner of feeding. Efforts have been made to -

increase the bandwidth, but a wide bandwidth of, say, 10% would be hard to . -
Freauencv / 9OOMHz band with
frequency stobility
obtain in the present state of the art. Fortunately, some systems such as pagers 2.5 part in 1 0 ~ ( - 1 0 + 5 0 ~ )
do not need a wide bandwidth in their operation, and so the microstrip antenna
can be applied to small VHF/UHF equipment used in such systems. Fig. 19.21 Example of pocket-sized pager having an MSA element
(Courtesy: Matsushita Communication Ind. Co. Ltd., Japan)
Four types of microstrip antennas (MSA) are introduced in this Section -
QMSA, PMSA, WMSA and FVMSA: Q stands for quarter-wavelength, P for
post-loading, W for window-attached, and FV for frequency-variable. Anten-
half-wave dipole. The four types of MSA variations for 900 MHz pagers are as
nas other than QMSA are modified from QMSA. They have basically similar
follows:
radiation patterns. The difference is that PMSA has two radiation apertures in
order to increase the gain, WMSA has a reactance slit on the patch to make the
( a ) Quarter-~avelengtlmicrostrip antenna ( Q M S A ) : QMSA is a quarter-
QMSA length shorter, and FVMSA is a QMSA with its resonance frequency
wavelength rectangular patch antenna with one end of the patch shorted elec-
electronically variable. Any of these antennas can be applied to small portable
trically. (Fig. 19.22) There are slight differences from an ordinary quarter-
 7094 Applications in mobile and satellite systems
wavelength patch antenna; the sides of the patch are cut so that the ground
plane has the same width as the radiation patch, and a part of the ground plane
is extended from the radiation aperture by a length of G, as shown in Fig. 19.22.
Applications in mobile and satellite systems 1095
(6, = 2.4), Teflon ( E , = 2.5), and glass-fibre-reinforced epoxy resin ( E , = 3.7);
E, is the relative permittivity of the substrate. 0 dB in the Figure is the gain of
the standard half-wave-dipole antenna. This is also used in later Figures in this
Section.

0-

;
i
-c-
B
-0 .
-
[?$,/
5
d .
patch
radiator\

; 1
-10
polyethylene
.feed v teflon
X epoxy-fibreglass

0 50
L
Fig. 19.22 Quarter wavelength microstrip antenna (OMSA) structure
Fig. 19.24 Gain versus patch length L (QMSA)
f

teflon
x epoxy-f lbreglasr

Fig. 19.23 Gain versus length G, Fig. 19.25 Gain versus patch width W (QMSA)

The length G, plays an important role in increasing radiation. The gain versus the total length L is illustrated in Fig. 19.24. The patch width
The variation of gain of a QMSA with respect to the length G, is shown in W also affects the gain, as shown in Fig. 19.25. An example of measured
Fig. 19.23: Three types of substrate material are treated as typical: polyethylene radiation patterns of the QMSA is shown in Fig. 19.26, where patterns in the
 7096 Applications in mobile and satellite systems Applications in mobile and satellite systems 7097
three planes XY, Y Z and ZX are illustrated. Consideration of the radiation i form another radiation aperture at its end. The ground plane is further extended
patterns in these three planes is important, since their evaluation becomes from the end of the radiation aperture by a length G, as shown in Fig. 19.27. The
meaningful when the antenna is used in an urban environment for mobile dimensions of the antenna and the co-ordinate system are also shown in the
communications, where multipath fading problems exist. The antenna perfor- I Figure.
mance should be evaluated three-dimensionally: not only in the horizontal
plane, but also in the two other planes. For example, the antenna-pattern radiator
maximum may exist in a plane other than the horizontal, and incoming waves post /patch
may often come from directions not in the horizontal plane. For these reasons,
gain defined only in the XY plane is not sufficient for evaluation of the actual
antenna performance. The antenna parameters used are as follows:
L = 76.7 mm, G, = 27.9 mm, W = 30 mm, t = 1.2 mm and 8, = 2.5
(Teflon).

Fig. 19.27 PMSA antenna structure

m -
9
.c -5-
% -

- 10
o polyethylene
0 teflon
a epoxy-fibreglass

I , , , , , , ,
Fig. 19.26 Radiation pattern (G, = 27.9 mm) (QMSA) 0 '0 . 20 (mr
Gz
(b) Post-loaded microstrip antenna (PMSA): PMSA [I 1 , 121 is an antenna Fig. 19.28 Gain versus length G, (PMSA)
modified from an ordinary QMSA and designed to have two radiation apertures The gain of the PMSA with three posts versus the length G, ( = G, = G,) is
driven by a single feed, as shown in Fig. 19.27. Several reactance posts are used shown in Fig. 19.28. The operating frequency is 930 MHz. Three kinds of
to replace the shorted termination of the QMSA, and the patch is extended to substrate materials (8, = 2.4,2.5 and 3.7) are again taken into account, and the
 1098 Applications in mobile and satellite systems 7099
Applications in mobile and satellite systems
Table 19.1 Dimensions of PMSA used in the experiments (see Fig. 19.27,
antenna parameters t = 1.2 mm, W = 30 mm, and lengths a and b in Table
G m = G, = Gz) 19.1 are used. The gain versus total length L is shown in Fig. 19.29. The length
Substrate er t W a b is about a quarter wavelength, while the length a is about 24 mm, which is
G, b L
(mm) (mm) (mm) (mm) (mm) (mm) (mm) slightly altered by the change in G,. Fig. 19.30 shows the variation of gain with
Poly- 1.2 30 48.9 4.5 24.5 82.1 length a, where the parameters used are the number of posts. Two examples of
ethylene 47.9 7.0 24.2 85.1 measured radiaton patterns are shown in Figs.19.31 and 19.32. They resemble
those of QMSA, although there is a difference in two apertures driven in this
instance.
47.4 20.0 24.0 111.4
Teflon 1.2 30 46.1 3.8 23.9 77.9

apoxy- 1.2 30 39.6 4.0 20.2 66.9

glassfibre 38.8 7.5 22.0 75.3
- -

O
- e polyethylene

0 teflon
epoxy-fibreglass
Fig. 19.30 Gain versus length a (PMSA)
m
9

( c ) Window-reactance-loaded microstrip antenna (WMSA): WMSA [I31 is

designed to have a shorter length than QMSA by putting a reactance window
(slit) with a length W, on the patch, as shown in Fig. 19.33. The window acts
as a reactance component and its value is altered by the length W, of the
-10
window. The resonance frequency versus the length W, of the window is given
in Fig. 19.34, which also gives the dimensions of the antenna. The gain is
affected by the length W,, and is shown in Fig. 19.35.The gain differs depending
upon the location of the window, and the distance S-W of the window from the
end of the patch (see Fig. 19.35). For larger S-W, the gain is higher. Two slits
0 50 placed on the patch, as shown in Fig. 19.36a, can also form a reactance
100 (mn)
L component in the antenna structure, and the length of the radiation patch can
Fig. 19.29 Gain versus patch length L (PMSA)
be shortened. If the length Wsis altered, the resonant frequency alters as shown
in Fig. 19.36b. It increases with W,.
 7 700 Applications in mobile and satellite systems 7 707
Applications in mobile and satellite systems

( d ) Frequency-variable microstrip antenna (FVMSA): The resonance fre-

quency of a QMSA can be altered electronically by varying the value of a
reactance loaded on the antenna structure. For this pirpose, the shorted ter-
mination of the QMSA is replaced by a reactance component and the reactance
is varied by its bias supply. This type of antenna is called FVMSA (Frequency-
variable MSA) [14]. A diode can be used as a reactance element, and its

substrate

-
Fig. 19.31 Radiation pattern (G, = 7.0 mm) (PMSA)

Fig. 19.33 Window reactance loaded microstrip (WMSA no. 1 )

(unit rnm)

Fig. 19.34 WMSA no. 1

( a ) Dimensions
( b ) Resonant frequency versus slot width W,.,

resonance frequency, depending on the bias voltage to the diode, is shown in

Fig. 19.32 Radiation pattern (G, = 20.2 mm) (PMSA)
Fig. 19.37, which also shows the gaintfrequency as the bias voltage is altered
from 6 to 10 V. The diode used is an MA325 (Fig. 19.38) and its characteristics,
 7 702 Applications in mobile and satellite systems Applications in mobile and satellite systems 1103

0
0 - 8
0

o bias voltage - 7
0
0
0 5 10 15 20 (mm) 0

Ww I ~ ~ ~ ~ ~ ~
Fig. 19.35 Gain versus slot width W, (WMSA no. I ) 8 50 900 950
FREQ (MHz)

Fig. 19.37 Gain versus frequency and bias voltage (FVMSA)

Unit :mm

" I" Jl.

Ws (mm)
a b
__H___
Fig. 19.38 WMSA no.2 1' Cathode
( a ) Dimensions 2 Anode
( b ) Resonant frequency versus slot width W, Fig. 19.38 Dimensions of M A 325
 1704 Applications in mobile and satellite systems Applications in mobile and satellite systems 7 105
and capacitance variation with bias voltage, are shown in Fig. 19.39. The
40 antenna structure and its dimensions are given in Fig. 19.40 and Table 19.2,
respectively.

30

-a 20
-
LL

0
U

10

0
0.1 0.3050.71 35710 3050X)lOO
v, (v)
Fig. 19.39 Bias voltage V, versus capacitance C, ( M A 325)

Fig. 19.41 Microstripline planar antenna for manpack radar

Fig. 19.40 FVMSA antenna structure 19.2.6 Radar

( a ) Manpack radar: Portable manpack radars operating by the pulse Dop-

pler method can serve as detectors of moving targets such as people, vehicles etc.
They are small, light devices which utilise microstripline antennas, as shown in
Fig. 19.41. such-an antenna is constructed of 16 centre-fed Franklin-type
Table 19.2 Dimensions of FVMSA (see Figs. 19.22 and 19.40) microstripline antennas [15]. It is centrally fed through the front side hybrid
er t W F~ b Gz Diode type networks. Each network is fed through the parallel-feeding network on the rear
(mm) (mm) (mm) (mm) (mm) side. Each microstripline
. . antenna works as a standing-wave antenna terminated
by an open clrcult.
The characteristics of this antenna are as follows: frequency, X-band; antenna
 1106 Applications in mobile and satellite systems Applications in mobile and satellite systems 1107
aperture, 34cm x 45cm; half-power beamwidth, about 4.5"; gain, 30dBi; ( c ) Radar rejector: A bidirectional communication system is shown in Fig.
weight, about 2 kg. 19.45 [17, 181. The radar reflector can transmit information from the reflector
site to a radar station as well as receiving signals from a radar station. The
( 6 ) Marine radar: Microstrip arrays have been used in marine radars [16]. reflector consists of a Luneburg lens with a reflector plate on it. Fig. 19.46 shows
Fig. 19.42 shows an inside view of a radar system installed on ocean vessels. This
type of radar system has been in service since 1981. The antenna array consists
of 48 (3 x 16) circular patch microstrip elements and is mounted on a rotating
pedestal. The specification of the system is given in Table 19.3. Fig. 19.43 shows

I I
0.97
I
0.98
I
0.99
I
1.00
I
1.01
I
1.02
I
1.03
flfo
Fig. 19.43 Sidelobe level and gain versus frequency

Fig. 19.42 Marrne radar antenna (Courtesy: Japan Rad~oCo. Ltd.)

Table 19.3 S~ecificationsof marine radar

Frequency X band
Gain over 22 dB
Beamwidth 6.0" in azimuth
25" in elevation
Dimension 367 x 88.0mm

the gain and first-sidelobe characteristics against frequency in the azimuth

direction. Typical radiation patterns are shown in Fig. 19.44. By replacing the
metallic waveguide-fed antennas with microstrip elements, the antenna system
can be made smaller and lighter, so that smoother rotation of the antenna and
increased reliability is achieved. In addition, the production cost can be reduced
and mass production becomes easier.
Fig. 19.44 Radiation pattern
details of the reflector plate. The two quarter-wavelength sections A and B are
an impedance transformer and RF blocks, respectively. When the diode is off,
signals are received. When the diode is switched on/off by coded pulses to be
1708 Applications in mobile and satellite systems Applications in mobile and satellite systems 7 109

Reflecting transmitted from the reflector to a radar station, the radar cross-section is
modulated by them and is detected at the radar station. Thus no transmitter is
needed to transmit information from a reflector site to a radar station. A
Luneberg lens focuses on incoming R F waves at any angle incident on the
antipodes of the lens sphere. Then a reflecting plate placed on the antipodes
Demodulator reflects the waves back exactly in the incident direction. The same lens sphere
Encoder can be used to establish communication channels with other radar stations in
Transmitter different directions by placing reflecting plates at the corresponding antipodes.
The gain is 26.5 dB and the beamwidth is 7" at 9.375 GHz when the relecting
plate is used as a receiving antenna. At the same frequency the radar cross-
Recorder section is 33 m2 and the beamwidth is 3" when the reflecting plate is used as a
Fig. 19.45 Block diagram of bidirectional communication systems (@ 1988 IEICE) reflector. Fig. 19.47 shows the VSWR when the diode is off, and the return loss
when the diode is on, as a function of frequency. At the centre frequency, a
Switching ( MSA VSWR of less than 1.1 and a return loss of 0.5 dB are obtained. Fig. 19.48 shows
circuit n the radar cross-section of the reflector normalised to that of a metal plate of the
same area. A modulation ratio of 13 dB is obtained when the diode is switched
To Recewer
on and off. Field experiments using a marine radar and the reflector have been
carried out successfully.
Rodidor
top view
-2 1
Bias ON

Bias
side view
C

Fig. 19.48 Configuration of reflector plate (@ 1988 IEICE)

a Equivalent circuit
b Top view Fig. 19.48 Radar cross-section of reflector with lens (@ 1988 IEICE)
c Side view

( d ) Secondary surveillance radar: In order to cope with the rapidly increasing

Fig. 19.47 VSWR and return loss of reflector without lens (@ 1988 IEICE)
air traffic, replacement of the existing secondary surveillance radar (SSR) with
the next generation of SSR (mode S) is being studied at an international level.
For the purposes of improving data rate, the cylindrical electronic scanning
antenna, which can be turned instantaneously in any direction and aimed at any
target, is believed to be more promising than the existing SSR with a constant
rotation.
Fig. 19.49 shows an experimental antenna. This is a cylindrical array of
one-third arc, and it has a 90" active sector. Its radiating elements are vertically
polarised circular patch antennas, and it operates in the frequency range 1030-
1090 MHz. A paper honeycomb substrate is used. Ten radiating elements are
arranged in elevation, forming a cosecant-squared pattern. The azimuth is
111 0 Applications in mobile and satellite systems Applications in mobile and satellite systems 1 1 11

composed of 32 active sector lines and a pencil beam is formed on the azimuth
plane, which can be scanned horizontally by gradually changing the transfer
switch.

( e ) Three-faced array: The vertically polarised antenna, which has an om-

nidirectional pattern in the horizontal plane and a desired pattern in the vertical
plane, is mainly utilised for broadcasting, telecommunications and as a trans-
ponder. In addition, this antenna can be used in an aircraft system. When used
together with the MLS (Microwave Landing System) it works as a DME
(Distance Measurement Equipment) antenna.

Fig. 19.49 Cylindrical-array antenna for SSR mode S

reflector

microstrip
line
slot

feed point

3 way power

Fig. 19.51 Three-faced microstrip slot array antenna (Courtesy: Japan Radio Co., Japan)

Fig. 19.50 Three-faced array antenna

The shape of the antenna is shown in Fig. 19.50. The antenna gives an
a antenna configuration omnidirectional pattern when a suitable flared reflection board is set on each of
b Feeding network the three faces. Each face produces a cosecant-squared pattern through pattern
(After Hara, and Goto [I91,@ 1988 IEEE) synthesis.
 1 1 12 Applications in mobile and satellite systems Applications in mobile and satellite systems 1 1 13

Fig. 19.51 shows a trial antenna. The characteristics of this antenna are: antenna is not considered wise, owing to its large size and the complexity of
frequency, 3 GHz; height, about 80 cm; peak gain, 9.5 dBi; antenna efficiency, manufacture. Again, an array structure is preferred to a large-aperture antenna,
85% [19]. and MSA elements in the S- and L-bands have been employed in such array
systems.

19.3 Satellite system 19.3.2 Direct-broadcasting reception

Several planar antennas for satellite TV reception using circularly polarised
19.3.1 Design considerations printed arrays have been proposed. In Japan, with the start of satellite broad-
This Section describes MSA applied to three areas of satellite systems: direct casting in December 1986 utilising the satellite Yuri 26 (BS-2b), efforts have
broadcasting systems (DBS), the earth stations and the satellite-borne systems. been made with the R & D of the planar antenna. When the characteristics of
a planar antenna are set at 11.7-12.0 GHz and the gain is 33-34 dBi, the
( i ) DBS antennas: In the design of the DBS antennas, there are the pro- antenna will receive signals in the high-field-strength area (except during heavy
blems of reducing the size and weight, and yet it must~havea gain high enough rains), as shown in Fig. 19.52.
to be competitive with the popular parabolic reflector antennas. A number of
flat antennas using MSA elements have been developed so far with the features
of light weight and ease of installation on the walls of a house or a building, or
location inside a room. Another advantage of using a flat structure is that its
performance is less affected by the snow or wind than by using parabolic
reflector antennas.

(ii) Earth stations: The general requirements for an earth-station antenna

are: simple structure, low cost, tracking capability and fading-suppression
performance. Two types of satellite tracking are considered: the first is to use the
radiation patterns of either a wide beam or a conical beam, thereby ensuring
that the antenna beam always tracks a satellite, and the second is to use either
a spherical array or a phased array. Antenna-beam shaping in the vertical plane
is designed to achieve fading suppression.
By making best use of the MSA feature, wide-beam or the conical-beam
antennas of simple structure and low cost have been developed and are presently
in practical use. MSA elements are also considered to be suitable as L-, S- and Fig. 19.52 Operational area (Japan)
C-band antennas, and applicable to spherical (switching) arrays, phased arrays
and sector-beam arrays. We shall describe a few types of planar antennas for satellite-TV reception,
A unique transportable equipment for satellite communications has been and a parabolic-cylinder reflector. These antennas have the following charac-
developed and tested by using a link with the Japanese communication satellite teristics;
ETS-V. The equipment hardware is put in an small attachk case which the Frequency range : 11.7 - 12.0 GHz
operator carries. Two antennas, one for the transmitter and the other for the Polarisation : right-hand circular polarisation
receiver, are used instead of only one antenna operating as both a transmitter Gain : more than 33 dBi
and a receiver. By this means the heavy diplexer can be eliminated. As the Axial ratio : less than 1 dB
antennas must be located on the back of the attachk-case lid, the MSA element VSWR : less than 1.5
is chosen as the best fit antenna.
( a ) Circular-patch array: Fig. 19.53 shows a sub-array of singly fed circular-
(iii) Satellite-borne antennas: Light weight, compactness, vibration-toler- ly polarised patch elements [20]. In order to provide wide-bandwidth axial ratio
able characteristics etc. are the major requirements for satellite-borne antennas. and impedance characteristics, the paired elements, described in Chapter 4,
Although high gain in the L-, S- and C-band is desirable, the use of an aperture form the fundamental element of the antenna.
 1 1 74 Applications in mobile and satellite systems
Fig. 19.54 shows a circular-patch array. The antenna exhibits a 30' inclination
of the beam direction (towards the front side of the board), it has 1024 elements
(size of the antenna is 48 cm x 64 cm) and has about 33 dBi gain. In order to
decrease the feeder loss, the main feeding line for four 256-element panels
tY
I I I
di=0.6h0, dzr0.8ho
Fig. 19.53 Sub-array of singly fed patch elements
Fig. 19.54 Planar array of circular patch elements (Courtesy: Yagi Antenna, Japan)
utilises the rear-mounted rectangular waveguide. As shown in Fig. 19.55 the
beam-tilt type can be set almost vertically. Thus, it has advantages such as being
easily installed on a wall, and snow does not affect it.
Applications in mobile and satellite systems 1 1 15
( 6 ) Square-parch array: Fig. 1 9 . 5 6 ~shows the structure of the elements. In
order to increase the bandwidth and to decrease the feeder loss of the patch
element, the thickness of the substrate, which supports the patch and the feeder,
was altered. In this case the substrate has a complex two-layered structure which
results in increased bandwidth and efficiency of the antenna [21].
Fig. 19.55 Flat antenna (Courtesy: Yagi Antenna. Japan)
Fig. 19.57 shows a square-patch planar array which has 512 patch elements
(see Fig. 19.566). The main feeding line for the eight 64-element panels has a
rear-mounted rectangular waveguide. The size of the antenna is 32 cm x 64 cm
and its gain is 34 dBi.
( c ) Crank-rype tnicrostrip-line array: Fig. 19.58 shows the configuration of
a crank-type microstrip-line antenna. This antenna, described in Chapter 13, is
formed by crank-type undulation of two strip conductors of a microstrip line.
 11 1 6 Applications in mobile and satellite systems Applications in mobile and satellite systems 11 77

fundamental element dielectric substrate

feed

*x
g r o u n d plate s t r i p conductor

Fig. 19.58 Configuration of crank-type microstripline antenna

gap (height 1 m m )

Er = 2.17
- A1 -4
circularly polarised
microstrip patches

Fig. 19.56 Circularly polarised patch element and sub-array

a Patch element with two-layered structure
b Sub-array
(After Murata and Ohrnaru [Zl],@ 198 IEICE)

Fig. 19.59 Planar array of microstripline antennas terminated in patch element (Courtesy: DX
Fig. 19.57 Planar array of square patch elements (Courtesy: NHK, Japan) Antenna, Japan)
 I
7 118 Applications in mobile and satellite systems , Applications in mobile a n d satellite systems 7 719

The two strip conductors are shifted by half their periods with respect to one ( e ) Circular-patch-slot array: Fig. 19.64 shows the structure of a circular-
another. In the Figure, the section surrounded by a dotted line shows the patch-slot array. This antenna consists of a ground plate, a patch and feeding-
fundamental element of a travelling-wave array [22]. I line plate, a slot plate and a radome. The circularly polarised patch elements,
Fig. 19.59 shows a crank-type microstrip-line planar array, and Fig. 19.60 with a single feeding point 1201and the feeding line, are printed on a plastic-film
shows the same antenna array set in its actual position. This antenna, with 332
elements, has a gain of 33-34.2 dBi at a frequency of 11.7 - 12.0 GHz. The size
radome
of the antenna is 40cm x 60 cm, its thickness is 2.1 cm and its weight is 4.7 kg
[23]. As shown in Fig. 19.59, an open end of each microstrip-line antenna -radiation plate
terminates in a square-patch element. The power at the patch element is totally (slot and patch)
radiated in order to improve the efficiency of the antenna [24].
feeding line plate
ground plate

Fig. 19.61 Structure of rectangular-slot array

feeder slot patch

Fig. 19.60 Actual position of planar antenna (Courtesy: DX Antenna, Japan)

(d) Rectangular-slot array: Fig. 19.61 shows the structure of a rectangular-

slot array. This antenna consists of a ground plate, a feeding-line plate, a
radiation plate and a radome. Each plate is supported by a honeycomb-foam
spacer in order to reduce the feeder loss. The radiating element consists of a
rectangular slot with a patch, and it operates by electromagnetical coupling to
the feeding line. The feeding line and radiating elements are printed on a film
substrate as shown in Fig. 19.62. The fundamental element of this antenna
seems to be complementary to the square-patch element with the slot open on
its surface.
Fig. 19.63 shows a rectangular-slot planar array. This antenna with 512
elements has a gain of 35 dBi. The size of the antenna is 36 cm x 72 cm, its
thickness 1.7 cm and its weight is 6 kg.
Fig. 19.62 Mask of sub-array
a Feeding-line plate
b Rad~ationplate
substrate. The substrate is covered with two sheets of shield board. Both the
upper and lower board are at a distance of 1 mm from the substrate, with air
between. The feeding line has the form of a suspended line system having low
loss. The circular-patch element radiates through a circular slot open on the
upper board.
 1720 Applications in mobile and satellite systems Applications in mobile and satellite systems 1121
. Fig. 19.65 shows a scaled-down mask of the patch elements and the slots. Fig. ~atch feeder slot
19.66 shows a circular-patch-slot planar array. The main beam direction of this
antenna is inclined at 10' to the front board. With 476 elements it has a gain of
34.1 dBi. The size of the antenna is 42 cm x 50.4 cm, its thickness 2.2 cm and
its weight is 4.2 kg.

(b)
Fig. 19.65 Mask of sub-array
a Feeding line with patches plate
b Slot plate

Fig. 19.63 Planar array of rectangular-slot elements (Courtesy: Matsushita Electric Works,
Japan)

n radome

slot plate

p a t c h and f e e d i n g
line plate
ground p l a t e

Fig. 19.64 Structure of circular-patch-slot array

(f) Parabolic-cylinder rejector: Fig. 19.67 shows a parabolic-cylinder reflec- &,' 7--

tor antenna. It is an example of the microstrip-line antenna application. The

antenna utilises a side-looking circularly polarised microstrip-line antenna [25]
on the line-source feed of the parabolic-cylinder reflector. The reflector can be
Fig. 19.66 Planar array of orcular-patch-slot elements (Courtesy Sony, Japan)
 1122 Applications in mobile and satellite systems Applications in mobile and satellite systems 1123
parabolic-cylinder reflector
installed in the vertical position because the direction of line-source feed is
inclined. Consequently, it has advantages such as ease of installation on a wall
and ;now does not affect it.

line-source feed
(side-looking microstrip line antenna)

Fig. 19.67 Configuration of parabolic-cylinder reflector-antenna

Fig. 9.69 GPS microstrip antenna (Courtesy: Toyo Communication Equipment Co. Ltd.,
Japan)

Fig. 19.68 shows an experimental antenna. It has a reflector of :size

Fig. 19.68 Parabolic-cylinder reflector with side-looking line-source feed (Courtesy: DX
Antenna, Japan)
60 cm x 70 cm, the size of line-source feed is 4 x 42 cm, and the gain at
12 GHz is 34 dBi. The axial ratio is 0.8 dB and the VSWR is less than 1.5.
 1124 Applications in mobile and satellite systems
iI Applications in mobile and satellite systems 1125
I

19.3.3. E a r t h stations Fig. 19.72 shows the vertical radiation pattern. Although the gain near the
I horizontal direct~onis less than -5 dBi, the low-noise amplifier provides enough
( a ) Wide-heam antennas gain to receive signals in this direction.
(i) G l o b a l positioning system: Microstrip antennas are being developed for
automobile navigation using the NAVSTAR satellite global positioning system
(GPS). They are designed to be installed on the roof of an automobile. The
antenna of a GPS receiver presents characteristics little different from those of
most satellite systems. Hemispherical beamwidth and right-hand circular pol-
arisation are required to receive signals from NAVSTAR satellites anywhere in
the sky.
Fig. 19.69 shows a microstrip antenna fed by two feeds from the bottom,
where the two feed voltages are of equal magnitude and 90" out of phase. In the
frequency range 1574-1577 MHz, the axial ratio is less than 2 dB in the
boresight direction and the gain is greater than -3 dBi in the direction between
6.5' and 90" above the horizontal plane. The vertical radiation pattern is shown
in Fig. 19.70.

Fig. 19.70 Vertical radiation pattern (Courtesy: Toyo Communication Equipment Co. Ltd.,
Japan)

Fig. 19.71 shows another GPS microstrip antenna viewed from the top and
the bottom. The antenna is fed at the two points from the bottom, with feeds
of equal amplitude and 90' out of phase. The thickness of the antenna is
1.6 mm. In the frequency range 1573 - 1577 MHz, the axial ratio is less than
1 dB and the gain is 5.5 dBi, both in the boresight direction.
Fig. 19.71 GPS microstr~pantenna (Courtesy: Toyota Central Research Laboratory, Japan)
( i i ) Transportable earth station: A hand-held message-communication ter-
minal (HMCT), shown in Fig. 19.73 for a very small earth station, can transmit
and receive messages of 20 - 30 characters via geostationary Engineering Test
 1726 Applications in mobile and satellite systems
Applications in mobile and satellite systems 7 127

Satellite - Five (ETS-V) at 150' east 1261. The HMCT is contained in an attach6
case. On the lid are stuck very thin and light circular-patch antennas as shown
in Fig. 19.74, and they are left-hand circularly polarised. The lid is opened and
placed to face the satellite at a suitable angle to communicate with a fixed earth
station. Each antenna is used, respectively, for transmitting (1645 MHz) and
receiving (1543 MHz). The advantages of this configuration are that each
antenna can be designed independently to maximise its efficiency, and a di-
plexer, which is usually heavy, is unnecessary. The gain and axial ratio of each
antenna in the boresight direction are about 7 dBi and 2.5 dBi, respectively.

Fig. 19.72 Vertical radiation pattern (Courtesy: Toyota Central Research Laboratory, Japan)

Fig. 19.74 Circular microstrip antennas for the HMCT (@ 1988 IEEE)

( b ) Conical-beam antennas: T o create a low G/T and low-cost antenna for

land-mobile and maritime satellite communications services, conical-beam
antennas are required because there is no need for tracking. For this require-
ment, three types of circularly polarised microstrip antennas with a conical
baam have been developed: (i) six-element circular array, (ii) higher-mode
microstrip antenna and (iii) circular array of strips and slots.

( i ) Six-element circular array: A conical-beam microstrip array which is

fabricated in the S-band and radiates circularly polarised waves is shown in Fig.
19.75 [27]. The antenna is a six-element circular array, and each pair, located
symmetrically with respect to the array centre, is fed out of phase. Therefore the

a
''
Fig. 19.73 Schematic view of the HMCT (@ 1988 IEEE)
antenna has radiation null at the boresight, and hence a conical beam. The
radiation pattern is calculated using the following equation:
 I
7 728 Applications in mobile and satellite systems Applications in mobile and satellite systems 1129
I

where g(%,4-$,) is the element pattern, k , is the propagation constant in free

I space, and e, and e, are the vectors of the radiation and element position,
respectively, and are given as:
e, = (sin Q cos 4 , sin % sin 4 , cos 8) (19.2)
e, = (d cos $, d sin $, 0) (19.3)
Here, the variable d represents the radius of the circular array, and I//,, is the
angular position of nth element, given as:

For simplicity of feeding, elliptical patches with a single feed which can radiate
circularly polarised waves are used.
The measured gain of the fabricated antenna is 6.6 0.9 dBi at 30" from the
vertical axis. The measured axial ratio is less than 2.5 dB in the same direction.
The measured radiation patterns are shown in Fig. 19.76.

(ii) Higher-mode microstrip antenna: A higher-mode microstrip antenna

can be designed to have a conical beam. In the case of the TM,,,-mode micro-
strip antenna, circular polarisation can be obtained by exciting the patch with
two feeds which are 90" out of phase and displaced 45" and 135" apart.
For the purposes of mobile communication services, the earth-station anten-
nas must have broadband characteristics.

Fig. 19.75 Outer view of conical-beam microstrip array (@ 19 8 8 IEICE)

tic element

Excitation element

Fig. 19.77 Configuration of conical-beam broadband microstrip antenna (01 9 8 8 IEICE)

A circularly polarised broadband microstrip antenna with a conical beam is

shown in Fig. 19.77 [28]. This antenna is a TM,,,-mode microstrip antenna with
a parasitic element which is placed in front of the excitation element. The
parasitic element is used to broaden the usable frequency band.
Fig. 19.76 Measured radiation patterns of conical-beam microstrip array (circular polarisa- The frequency dependence of VSWR with and without a parasitic element is
tion) (@ 1 9 8 8 IEICE) shown in Fig. 19.78. The parasitic element increases the bandwidth more than
 7 730 Applications in mobile and satellite systems Applications in mobile and satellite systems 1131

. ,
2 5 1 p 1 a l f l l ~1 ,,
,96 1
I
1.04
, 1
Normal lzed frequency
Fig. 19.78 Frequency dependence of measured VSWR with and without parasitic element
(@ 1988 IEICE)

radiation pattern 8=3 5-

0
---- ---------------- --------------_--
___-
Fig. 19.79 Measured radiation patterns of conical-beam broadband microstrip antenna (cir- -1 - 8=45'
cular polarisation) (@ 1988 IEICE) -2 -

strip
dipole

Fig. 19.81 Calculated co-polar radiation patterns and axial ratios (Courtesy: J.P. Daniel,
University of Rennes; and Koichi Ito, Chiba University)
/ \ p I ane a Vertical plane
ref lector subst r a t e -4 = 0' plane
Fig. 19.80 Configuration of the circular array (Courtesy: J.P. Daniel, University of Rennes;
- - - 4 = Wplane
b Conical-cut plane
and Koichi lto, Chiba University)
-0 = 3 5 plane
(Windows and feed network are not shown)
- - - 0 = 4 5 plane
 1732 Applications in mobile and satellite systems Applications in mobile and satellite systems 1133

six times. The measured directive gain of the fabricated antenna is 6.5 dBi in the as shown in Fig. 19.83~.The switching circuit is composed of two SP-3T
frequency band of 8%. The measured radiation patterns are shown in Fig. switches and one DP-DT switch. Each radiator is electronically switched one by
19.79. one through a control which compares its received power level with that of the
(iii) Circular arra.v of strips and slofs: The third type of conical-beam
antenna with circular polarisation can be realised by modifying a circularly
polarised printed array composed of strip dipoles and slots [29]. Fig. 19.80
shows an example of a circular array consisting of four-pair strip dipoles and
slots. Windows, not shown in the Figure, are placed in the ground plane to
increase the gain and bandwidth of the strip dipoles [29]. The radiating elements Array Element
are arranged radially and fed by an appropriate microstrip feed network. A
substrate with high dielectric constant should be used to reduce the antenna
diameter.
Fig. 19.81 shows calculated results of co-polar radiation patterns and axial
ratios for a four-pair array with a larger reflector. The frequency was 2.5 GHz
and the diameter and height of the antenna were about 10 cm and 3 cm,
respectively. The axial ratio was less than 2 dB over most of the hemisphere. The A
(a) C o o r d i n a t e S y s t e m
ripples of both the radiation pattern and the axial ratio in the conical-cut planes
were less than I dB.

Fig. 19.82 Outer view of six-element switched-element spherical array (@ 1988 IEE)

( c ) Spherical arrays ( i ) six-element switched-element spherical array: An ( c ) 2-dimensional radiation patterns

outer view of six-element switched-element spherical array, which is fabricated
in the L-band and radiates circularly polarised waves, is shown in Fig. 19.82
[30]. It is 40 cm in diameter and 20 cm in height. This antenna comprises the
radiator section, the switching circuit and the controller. The radiator section is
composed of six circularly polarised elements. Each element is a circular micro-
strip antenna with a parasitic element for broadening the bandwidth, and is
located on the limited sphere tilted at an angle a from the vertical axis (Z-axis)
Fig. 19.83 Calculated two-dimensional radiation patterns of optimised six-element spherical
array (@ 1988 IEE)
adjacent radiators. It is postulated that the coverage area of the antenna is
between B,, - AB, and Oo+ AO, from the vertical axis and omni-directional in the
horizontal plane, as shown in Fig. 19.83~.
 1134 Applications in mobile and satellite systems Applications in mobile and satellite systems 1 135

The radiation pattern is calculated from the following equation:

4

where, ko is the propagation constant in free space, and JI, is the phase of the
nth element. e, and e, are the vectors of the radiation and element position,
respectively, and are given as
e, = (sin 0 cos 4, sin 0 sin 4, cos 0) (19.6)
ep = a (sin ct cos p,, sin a sin fin, cos a) (19.7)
Here, the variable a represents the radius of the spherical array. g(l, p) is the
radiation pattern of the array elements. Assuming the use of a broadband
microstrip antenna, g(l, p) can be approximated by the following equation:

The relation between 0, 4 and 1 is therefore given as follows:

cos l = (er .ep)
-
lerllepl
Fig. 19.84 Configuration of 86-element spherical array (Courtesy: KDD, Japan)
The calculated 2-dimensional radiation patterns of an optimised six-element
array antenna are shown in Fig. 19.83. The angle 0 from the antenna vertical
axis is indicated by the distance from the origin, and the angle 4 rotated about
the vertical axis is indicated by the angle from the horizontal axis. The six
radiators are located at a position a = 44" as shown in Fig. 19.83. The solid and
dotted curves in Fig. 19.83 indicate the contour lines in which the directive gain Antenna -
for each beam is 7 and 8 dBi, respectively. It can be seen in Fig. 19.83 that the
minimum coverage gain is 7 dBi when the coverage area is between 40' - 20'
and 40" + 20".
The measured directive gain of the fabricated array is more than 7 dBi for a
coverage area of within 20"-60" from the vertical axis and a frequency band of
8%. The peak gain is 8.5 + 0.3 dBi given the same frequency band. The
measured axial ratio is less than 3 dB and the measured VSWR is less than 1.4. 8x6
The total insertion loss for the switching circuit is less than 1.6 dB. Therefore, Switch 611
Combiner
actual coverage gain can be more than 5.4 dBi. Matrix
/
(ii) 86-element spherical array: The configuration of an 86-element spheri-
cal array is shown in Fig. 19.84 [31]. This array is composed of 86 microstrip
radiators, and the sub-array, which is composed of six or seven radiators, is Group
switched one by one. The coverage gain of this antenna is more than 12 dBi #8
within 115' from the vertical axis. The configuration of feeder section is shown
in Fig. 19.85. The active elements are used in order to provide higher G/T. The
calculated radiation patterns of an 86-element spherical array are shown in Fig. Fig. 19.85 Configuration of feeder section of 86-element spherical array (Courtesy: KDD,
19.86. Japan)
1136 Applications in mobile and satellite systems Applications in mobile and satellite systems 7 137

( d ) Sector-beam array: The sector-beam array is used for shipborne anten-

nas [32]. The outer view of the sector-beam array, which is fabricated in the
L-band and radiates circularly polarised waves, is shown in Fig. 19.87. The size
of the array is 1000 mm x 1200 mm. The array is composed of 48 circular
microstrip discs. This antenna has a 40" sector beam in the elevation plane.
The measured radiation pattern of the fabricated antenna is shown in Fig.
19.88. The half-power beamwidth is less than 39" in the vertical-plane and less
than 20" in the horizontal-plane. The sidelobe level is lower than -25 dB, the
axial ratio is less than 1.5 dB and the measured gain of the antenna is more than
16 dBi.

Fig. 19.87 Outer view of sector-beam array (@ 1988 I EICE)

Fig. 19.86 Calculated radiation patterns of 86-element spherical array (Courtesy: KDD,
Japan)

( e ) Phased arrays (i) Slot antenna: The airborne antenna shown in

Fig.19.89 is a 16-element cross-slot antenna (XSA) developed for satellite
communications at L-band [33, 341. The element spacing is 97 mm and its
volume is about 560 x 560 x 20 mm.
Fig.19.90 shows the configuration of the XSA element. It is called a cavity-
backed cross-slot antenna and is fed with equal amplitude and phase at the two
points shown in Fig.19.90 in order to get wider-band impedance matching
characteristics compared with a centre-fed slot. The two perpendicular slots are
fed with a 90' phase difference through hybrid circuits to obtain right-hand
circular polarisation. The length and width of the slots are about 112 mm and Azimuth Angle (') Elevation Angle (')

5 mm, respectively, and the volume of the cavity is about 80 x 20 x 20 mm. Fig. 19.88 Radiation patterns of sector-beam array (@
1988 IEICE)
I
 1738 Applications in mobile and satellite systems
Fig. 19.89 Cross-slot antenna (@ 1988
Applications in mobile and satellite systems 1139
Beam scanning is performed by controlling 4-bit variable phase shifters
attached to each antenna element. Fig.19.91 shows the patterns at 1.6465 and
1.545 GHt. In this Figure, there is a slight difference between the beam direction
to scan and the actual beam direction. The boresight gain is 15.7 dBi.
I
I
I slot feed Lircuit
I
I
I metal dielectric I
Fig. 19.90 Element configuration (@ 1988 IEE)
(ii) Patch antenna: The airborne antennas shown in Fig.19.92 is a 3 x 3-
element microstrip antenna (MSA) developed for satellite communications at
L-band [33, 341. The element spacing is 94 mm (about half a wavelength at
1.6113 GHz) and its antenna volume is about 300 x 300 x 10 mm.
Fig.19.93 illustrates the configuration of an MSA element. This is a newly
developed two-layer patch antenna. Glass-microfibre-reinforced PTFE with a
dielectric constant of 2.3 (thickness = 3.2 mm) is used as a dielectric substrate
because of its good temperature characteristics. The two-layer MSA is adopted
in which the upper and lower parts are used for transmissin (16465 GHz) and
reception (1.545 GHz), respectively. Each layer is independently fed at two
points with a 90' phase difference to obtain right-hand circular polarisation. The
upper antenna is a conventional circular MSA and is used for transmission,
IEE) while the lower one is a circular MSA with an electric shielding ring to separate
 1140 Applications in mobile and satellite systems
Applications in mobile and satellite systems 1141

e ('1
b
Fig. 19.91 Antenna patterns (@ 1988 IEE)
a Radiation 1$)
b Reception
X
Fig. 19.92 3 x 3-element microstrip antenna (@ 1988 IEE)
 1 142 Applications in mobile and satellite systems Applications in mobile and satellite systems 1143

the lower from the upper MSA and to feed the upper MSA easily. The diameters
of the upper and lower MSA and the shielding ring are about 66 mm, 84 and
27 mm, respectively. The distance from the centre to the feed points are about
10 mm and 20 mm for the upper and the lower MSA, respectively.
Beam scanning is performed by controlling the 4-bit variable phase shifters
attached to each antenna element. Fig.19.94 shows the antenna patterns of
radiation (1.6465 GHz) and reception (1.545 GHz). In this Figure, there is a
slight difference between the beam direction to scan and the actual beam
direction. The boresight gain is 15.2 dBi.

Feed Point
(1.545 GHz)
Feed Point
(1.6465 GHz)

Dielectric
Substrate

I \
Receive Port Transmit Port

Fig. 19.93 Element configuration (@ 1988 I E E )

(iii) Sequential-array antenna: Japanese experimental domestic mobile-

satellite communications, using Engineering Test Satellite-V (ETS-V) launched
in 1987 [35], will provide high-quality links for ships and aircraft at L-band. For
this purpose the antenna shown in Fig.19.95 has been developed [36]. It is
mounted in the fairing of a Boeing 747 Jumbo Jet. Each of the two array panels
consists of 2 x 8 microstrip circular patches, and only one of the array panels
facing the satellite is used for communication.
The beam can be scanned only in the horizontal plane. Thus each of eight
4-bit digital phase shifters is connected to two vertical elements, as shown in
Fig.19.96. Each circular patch fed by two points is sequentially rotated and
differentially phase-shifted [35]. In such a sequential array, perfect circular
1144 Applications in mobile and satellite systems Applications in mobile and satellite systems 1 745
polarisation is obtained at a boresight direction independently of the polaris- Table 19.4 Sequential phased-array characteristics
ation of elements, and the reflected waves returned to the input port cancel1 each
other. For the rectangular sequential array shown in Fig.19.95 high cross- Frequency 1545-1548 MHz (transmitting)
1647-1 650 MHz (receiving)
Polarisation left circular

995mm-4
Gain 12-14.5 dBi (beam scanning)
sequential phased array Bandwidth 8% (VSWR < 2)
(2x8)
Antenna element circular patch antenna
Array type 2 x 8 element
substrate sequential phased array
A (6 = 2.8)
Phase shifter 4 bits (digital)
Substrate Teflon ( E , = 2.6)
thickness = 4mm
Volume 15 x 40 x 90cm3
Weight 18 kg

Fig. 19.95 Airborne phased array (@ 1988 IEEE)

sequential feed in fairing

-1- in fuselage
arravs circuit I

1 ~ w rDIV.:
. Power divider

Fig. 19.96 Configuration of antenna systems (@ 1988 IEEE)

polarisation discrimination is obtained over a wide angle in the two principal

planes, which is useful for reducing fading due to reflections from the sea
surface. The characteristics of the array in Fig.19.95 are given in Table 19.4. Fig. 19.97 Outer view of synthetic aperture radar (Courtesy: MELCO, Japan)
 7 746 Applications in mobile and satellite systems Applications in mobile and satellite systems 1747

19.3.4 Satellite-borne antenna

( a ) Synthetic-aperture radar (SAR): A rectangular microstrip array is

applied to a synthetic-aperture radar (SAR) [37]. The outer view and the
configuration of this radar are shown in Figs.19.97 and 19.98, respectively. This
antenna is composed of eight panels, the size of leach being
1390 mm x 2060 mm. Each panel has eight sub-arrays and each sub-array has
16 rectangular microstrip antennas, which radiate linearly polarised waves. The
substrate is made of honeycomb foam, the dielectric constant being 1.14. The
amplitude distribution of the array is uniform in the E-plane, and it is tapered
in the H-plane so as to obtain -18 dB sidelobe level.
The measured gain is more than 26 dBi in the 1.3 GHz band. The calculated
two-dimensional radiation patterns are shown in Fig.19.99. The half-power
beamwidth is less than 8.7' in the E-plane and less than 6.3' in the H-plane; the
sidelobe level is lower than -12.7 dB in the E-plane and lower than - 17.9 dB
in the H-plane.

U b .
1.3 9m
1m -$*k !microstrip array
coax. cab

flexible cable 4 Iflcxlble joint

coax. wave gu l do

Fig. 19.98 Configuration of synthetic-aperture radar (@ 1988 IEICE)

( b ) 19-element muliibeam array: The outer view of a 19-element multibeam-

array antenna for a data-relay satellite is shown in Fig. 19.100 [38]. The antenna
operates at 2.1 and 2.3 GHz. The receiving system employs a fixed multibeam
antenna of 19 contiguous beams, while the transmitting system uses a single-
beam phased array; 12 radiating elements are shared by both transmission and
reception, and the remaining seven elements are dedicated to reception.
The sub-array of the 19-element array is composed of seven circular micro-
strip patches, as shown in Fig.19.101. To broaden the bandwidth, these patches
are printed on a Nomex honeycomb substrate of 10 mm thickness. Each patch
is excited at two points with 90' phase shift by the rear feeding circuit. The
 1148 Applications in mobile and satellite systems Applications in mobile and satellite systems 7 749
microstrip antenna is arranged so that each patch has small notches to cancel
the elliptically polarised components generated owing to the asymmetrical feed
structure. The measured gain of this sub-array is more than 15.1 dBi at
2.1 GHz. The measured radiation patterns are shown in Fig.19.102.

Fig. 19.100 Outer view of 19-element multi-beam array (@ 1 988 IEICE)

Fig. 19.102 Radiation patterns of sub-array of 19-element multi-beam array (@ 1 988 1 EICE)

honey-comb-core substrate

19.4 References

I KURAMOTO, M., and SHINJI, M. (1986): 'Second generation mobile radio telephone
system in Japan', IEEE Commun. Mag. 24, pp. 16-21
2 HORI, T., and NAKAJIMA, N. (1983): 'Sector-beam base station antenna for land mobile
communication' Natl. Conv. Rec. IECE Japan 754 (in Japanese)
3 NAKAJIMA, N., NARA, T., KAMEO, S., ABE, H., and TAKAMATSU, Y. (1985):A major
angle comer reflector antenna with 180" beam width'. Natl. Conv. Rec. IECE Japan 752 (in
Japanese)
2.2 h 4 NAKAJIMA, N., and HORI, T. (1984): '900 MHz-band multibeam antenna using butler
h:2.3GHz I feeding circuit 1 matrix', IECE Japan Technical Report, AP84-50 (in Japanese)
5 ITO, K., and SASAKI, S. (1988) 'A small printed antenna composed of slot and wide strip for
Fig. 19.101 Structure of sub-array of 19-element multi-beam array (@ 1 9 8 8 indoor communication systems', IEEE Int. Antennas and Propagation Symp. pp. 716-719
IEICE)
 1150 Applications in mobile and satellite systems Applications in mobile and satellite systems 1 151

6 MISHIMA, H., and TAGA, T . (1982): 'Antenna and duplexer for new mobile radio unit', Rev. general mobi!e satellite communications', IECE Japan Technical Report, AP84-30 (in
Elect. Commun. Labs NTT, 30, pp. 359-370 Japanese)
7 JASlK H, (1961): 'Antenna Engineering Handbook' (McGraw Hill, NY) 32 OHMORI, S., MORIKAWA, H., MIYANO, N., SUZUKI, Y., and CHIBA, T., (1980):
8 OHISHI, Y. (1983): 'Analysis of reactance-loaded annular slot antenna directivity perfor- 'Circularly polarized sector-beam shipborne antenna', Nat. Conv. Rec. IECE Japan S1-4 (in
mance'. Graduation Thesis, University of Tsukuba, Japan (in Japanese) Japanese)
9 KONDO, M., SASAKI, S., MATSUMOTO, K., ABE, H., CHATANI, Y., FURUNO, T., 33 SHIOKAWA, T., er 01. (1986): 'Cross slot array antenna for aeronautical satellite communica-
and MANO, S. (1986): 'A train antenna for a leaky coaxial cable'. Natl. Conv. Rec. IECE tions', IECE Japan Technical Report, AP86-59, pp. 17-21 (in Japanese)
Japan 618 (in Japanese) 34 YASUNAGA, M., el al. (1987): 'Phased array antennas for aeronautical satellite communica-
10 YAMAMOTO, J. (1985): '900 MHz numeric pager' Nat. Conv. Rec. IECE Japan 2411 (in tions'. IEE lnt. Antennas and Propagation Conf., pp. 47-50
Japanese) 35 OHMORI, S. el al. (1986): 'Aircraft earth station for experimental mobile satellite system'.
11 KUBOYAMA, H., el al. (1985): Post loaded microstrip antenna for pocket size equipment at IEEE Int. Conf. on Communications, pp. 1392-1395
UHF'. Int. Symp. on Antennas and Propagation, Japan, pp. 434-436 36 TESHIROGI, T., er al. (1986): 'Airborne phased array antenna for mobile satellite commun-
12 FUJIMOTO, K., er al. (1986): 'Small Antennas' (Research Studies Press, London). pp. ications:. IEEE Int. Antennas and Propagation Symp., pp. 735-738
243-249 37 HISADA, Y., ITO, Y., AKAISHI, A,, IMURA, N., and ONO, M. (1983): 'The results of
13 BABA, T. (1987): 'Analysis of reactance-loaded square microstrip antenna performance'. partial experimental manufacturing of syntheticaperture radar antenna', IECE Japan Techni-
M.S.C. Thesis, University of Tsukuba, Japan, pp. 78-84 (in Japanese) cal Report, AP83-39 (in Japanese)
14 BABA, T. (1987): 'Analysis of reactance-loaded square microstrip antenna performance'. 38 TESHIROSI, T., CHUJO, W., AKAISHI, A,, and HIROSE, M. (1986): 'Multibeam array
M.S.C. Thesis, University of Tsukuba, pp. 85-89 (in Japanese) antenna for data relay satellite'. Trans. IECE Japan, J69-B, pp. 1441-1452 (in Japanese)
15 NISHIMURA, S., NAKANO, K., and MAKIMOTO, T. (1979): 'Franklin-type microstrip
line antenna'. IEEE Int. Antennas and Propagation Symp. pp. 134-137
16 FUJII, K., and ISHIKAWA, H. (1984): 'Low sidelobe microstrip array antenna', Nat. Conv.
Rec. 49, Optical and Electronics, IECE Japan (in Japanese)
17 HASEBE, N., and ONOE, M. (1984): 'Radar reflector with bidirectional communication
capability'. IEEE Int. Antennas and Propagation Symp., pp. 788-791
18 HASEBE, N., er 01. (1985): 'Radar reflector with bidirectional communication capabaility',
Trans IECE Japan, J66-B, pp. 1177-1 184 (in Japanese)
19 HARA, Y., and GOTO, N. (1984): 'An omnidirectional vertical shaped-beam three faced
microstrip slot array antenna', IEEE Int. Antennas and Propagation Symp., pp. 527-530
20 HANEISHI, M. (1985): 'A circularly polarized SHF planar array composed of microstrip
pairs-element', Int. Symp. on Antennas and Propagation, Japan, pp. 125-128
21 MURATA, T., and OHMARU, K. (1986): 'Characteristics of circularly polarized printed
antenna with two layer structure'. IECE Japan Technical Report, AP86-101 (in Japanese)
22 NISHIMURA, S., SUGIO, Y., and MAKIMOTO, T. (1983): 'Crank-type circularly polarized
microstrip line antenna'. IEEE Int. Antennas and Propagation Symp., pp. 162-165
23 WATANABE, T., FUJITA, T., and DEGUCHI, F. (1987): 'Microwave planar array antenna
design'. ITE Japan Technical Report R E 8 7 4 (in Japanese)
24 NISHIMURA, S., NISHIGAKI, A., WATANABE, T., SUGIO, Y., and MAKIMOTO, T.
(1987): 'Circularly polarized microstrip line antenna terminated by patch antenna'. IECE
Japan Technical Report, AP86-124 (in Japanese)
25 NISHIMURA, S., SUGIO, Y., and MAKIMOTO, T. (1985): 'Side-looking circularly pol-
arized microstrip line planar antenna', Int. Symp. on Antennas and Propagation, Japan, pp.
129-132
26 HASE, Y ., et al. (1987): 'Very low speed message communication system using hand-held earth
station'. IEEE Int. Conf. on Communicatins, pp. 520-524
27 HORI, T., ITAMI, Y., and NAKAJIMA, N. (1982): 'Circularly polarized microstrip array
antenna with conical beam'. Natl. Conv. Rec. IECE Japan 655 (in Japanese)
28 HORI, T., TERADA, N., and KAGOSHIMA, K. (1986): 'Circularly polarized broadband
microstrip antenna radiating conical beam'. Natl. Conv. Rec. IECE Japan 637 (in Japanese)
29 ITO, K. (1987): 'Circularly polarized printed arrays composed of strip dipoles and slots',
Microwave Jol, 30, pp. 143-153
30 HORI, T., TERADA, N., and KAGOSHIMA, K. (1987): 'Electronically steerable spherical
array antenna for mobile earth station'. IEE Int. Antennas and Propagation Conf., pp. 55-58
31 SHIOKAWA, T., WATANABE, F., and NOMOTO, S. (1984): 'Spherical array antenna for
 Chapter 20

Conical conformal microstrip

tracking antenna
P. Newham and G. Morris

20.1 Introduction

This Chapter describes the design, construction and testing of a prototype

monopulse tracking antenna that is realised as a thin microstrip/triplate struc-
ture conformal to the surface of a cone. The main application of such an
antenna configuration is to provide gul'ded-weapon seeker antennas for high
speed ( > Mach 2.5) missiles and shells where the pointed (high fineness) front
ends make conventional antenna (reflector or flat plate) plus radome designs
impracticable owing to severe radome aberration effects. The latter is due to the
high angles of incidence with which the antenna must necessarily illuminate the
radome, as well as radome-tip scattering and blockage. Other applications exist
where conical, or near conical, geometries are found - there are obviously many
possible aircraft sites, and the antenna configuration described should find
applications here.
In the context of guided weapons, tracking antennas are required for both
narrow and broadband applications: narrow band for active and semi-active

ed is narrow band, having a bandwidth of -

seekers and broadband for passive anti-radiation seekers. The design as discuss-
10% centered at 10 GHz;
techniques to extend the bandwidth to over 50% are discussed at the end of the
Chapter.
The antenna as described consists of microstrip radiators on the conical
surface with an adjacent (underneath) triplate feed network. Emphasis is given
in the chapter to the practical engineering problems that have been encountered
and solved during the design and development programme.

20.2 Single patch element

20.2.1 Choice of array element

The choice of a sutable printed array element is governed mainly by its perfor-
mance in the endfire direction. The directivity of an antenna array in a particular
 1 154 Conical conformal microstrip tracking antenna
direction is only as good as that of its constituent elements in that direction.
Thus, for the conical array, a radiator with almost omnidirectional coverage in
the plane containing the cone axis is required. The choice is essentially limited
to either a half-wavelength rectangular patch element or its quarter-wavelength
short-circuited version. Other alternatives, such as the slot or dipole, possess a
natural null in their radiation patterns at endfire. Others, such as the V-antenna,
with good directivity at endfire, are seriously degraded by the necessary presence
of a ground-plane associated with the inner, but adjacent, conformal feeding
circuitry. The effect here is to squint the pattern considerably from endfire.
For several reasons the quarter-wavelength patch is seen to offer the best
solution. In particular:
(a) Its theoretical pattern is more omnidirectional than that of the half-
wavelength patch. For example, in the E-plane at endfire (with a substrate of
relative dielectric onstant E, = 2.45), the pattern drops to - 6 dB below peak,
as against - 11 dB for the half-wavelength patch [I]. For the case of the 20' cone
under consideration, the main interest is in the performance at 10' from endfire,
i.e. along the cone axis. Here the values are -2 dB and -8 dB, respectively;
and since the theoretical gains are 5 dBi and 7 dBi, respectively [2], the gain of
the quarter-wavelength patch along the cone axis exceeds that of the half-
wavelength patch by 4 dB.
(b) The quarter-wavelength patch occupies approximately half the area of the
half-wavelength patch.
One particular drawback of the quarter-wavelength patch, however, is the
difficulty in feeding it from microstripline, owing to the high input impedance
present at its radiating edge - this is typically 240R. It becomes necessary to
break into the patch via a notch in order to find a suitable impedance level.
The design of the single patch element is now considered in detail. ,
20.2.2 Choice of substrate
Several factors govern the choice of substrate for the microstrip layer onto
which the antenna arrays are printed. First and foremost is the requirement that
the substrate must be capable of being formed into a truncated cone without
physical damage, and preferably without changing its electrical properties.
Secondly, the material should be low loss with a low relative dielectric constant.
This is in order to maximise the available bandwidth of the patch element to
allow for small performance differences between array elements owing to man-
ufacturing tolerances. For this purpose the substrate should also be as thick as
possible, but not so thick as to generate excessive surface-wave energy. The
criterion adopted for the maximum thickness h is such that h < 0.071, for
E, = 2.32 [3]; at 10 GHz this gives h < 2.1 mm. RTIDuroid 5880, manufac-
tured by Rogers Corporation with E, = 2.2 and loss tangent tan 6 = 0.001, has
been found to be a suitable material, using a standard thickness of 1.57 mm.
Rogers provide valuable information on bending their Duroid products [4], and
1 Conical conformal microstrip tracking antenna 1 155
I
i
it is the randomness of the glass microfibres embedded in the PTFE base
material that suppresses any tendency for the material to fracture under bending
stresses.
20.2.3 Feeding the patch
A patch antenna may be fed either from a microstrip line on the same substrate
or via a probe extending through the ground plane. Probe feeding was rejected
for the conical array application for two important reasons. First it would have
entailed a combination of both the power division and hybrid tracking circuitry
on the same triplate feed network. Owing to the limited space available on the
triplate substrate, as will be evident from Section 20.4, this was not possible to
achieve in practice. Secondly the additional probe inductance, associated with
the thick antenna substrate, would also need to be matched within the limited
space on the triplate.
As mentioned in Section 20.2.1, the quarter-wavelength patch is difficult to
feed via a microstrip line owing to the high input impedance at its radiating
edge. In order to feed the patch at a suitable impedance level, it is necessary to
break into the patch via a notch in the radiating edge. At frequencies below
about 5-6 GHz, the size of a patch becomes large relative to the width of a
typical microstrip feed line, which remains almost constant with frequency for
a given impedance. In this case, the patch may be fed through its short-circuit
plane with negligible effect on its performance [5]. However, at 10 GHz this is
no longer possible unless a very narrow high-impedance feed line is used. Even
at 100 R, the width of the line is comparable to that of the patch. For reasons
which will become apparent later, the 100 R feed line was the most suitable
means of feeding the patch, and for this reason the patch was fed through its
radiating edge. Weinschel [6] has measured the impedance variation with the
notch width ( s in Fig. 20.1) for a half-wavelength patch. From his graphs it is
apparent that the measured impedance corresponds to that of a probe-fed patch
when the notch width is nearly equal to the substrate thickness. The notch width
for the patch design was thus fixed at 1.5 mm.
20.2.4 Theoretical design method
The design of the quarter-wavelength patch is based on the procedure outlined
in Reference 7, and considers the patch as a short-circuited length of resonant
transmission line.
Fig. 20.1 shows the geometry of the patch together with its equivalent circuit.
Yo is the characteristic admittance of the microstrip (patch) transmission line.
The feed point is located a distance x from the short-circuit plane and d from
the radiating edge. The patch resonant length x + d is somewhat less than a
quarter wavelength in the transmission line owing to the fringing field extension
beyond the physical edge of the patch. This field extension is incorporated into
the susceptive part of the aperture admittance Yo of the radiating edge, which
is made up of a radiation conductance G, and the susceptance B,.
 1156 Conical conformal microstrip tracking antenna Conical conformal rnicrostrip tracking antenna 1157

Looking into the right-hand side of the transmission line from the feed point, The f sign in eqn. 20.3 is chosen according to whether the denominator is
the admittance, assuming a lossless line, is given by i negative or positive. On evaluation, Podis then inserted into eqn. 20.2 to obtain
B,,. Resonance occurs, of course, when the shunt susceptance to the left of the
Gh
- = G, (Yo - B, tan pd) +
G, (B, +
Yo tan pd) tanpd
(20.1) feed point cancels that to the right; i.e. when:
yo (Yo - B, tan pd)' + Gi tan2 pd
- j Yocotp0X = - j B , , ,
Bi,, - B, Yo + (Y: - Gi - &) tan pd- B, Yo tan2 Dd
(20.2)
YO +
(Yo - B, tan fld)2 Gt tan2 j d
where is the propagation constant at the frequency in question. Note that
eqns.20.1 and 20.2 are in error in Reference 7.
The patch resonant length x + d is thus obtained, together with the feed-point
short -
circuit radiating position relative to the short circuit plane.
plane edge An expression for 2, = l/Y,, the characteristic impedance of the microstrip
(patch) transmission line, has been taken to be [8]:

quarter -
wavelength where
patch murostrtp
feed lfne

short -
and
circuit
d

transmission
line
model
I / ,$ where W = patch width
h = substrate thickness
t = conductor thickness
E, = substrate relative dielectric constant
driving
polnt
/
Fig. 20.1 Microstrip quarter-wavelengthpatch and its equivalent transmission-line model
The aperture conductance Gois taken to be [8]:

The short circuit to the left of the feed point simply presents a shunt suscep-
tance of value - j Yo cot px. To proceed, Gin is set equal to the desired 2n
driving-point admittance G,, at the centre frequency (P = Po), and eqn. 20.1 is where ko = -
20
and Lo is the free space wavelength.
then solved for d. This takes the form of a quadratic equation with solution:
It should be noted that the expression for the aperture susceptance B, differs
considerably between References 7 and 8. In Reference 7 the aperture fringing
capacitance is given by:
 1758 Conical conformal microstrip tracking antenna
where L is the patch resonant length. Then B, = jw,C', where w, is the
resonant angular frequency.
In Reference 8 B, is expressed in terms of the fringing field extension, AI:
Conical conformal microstrip tracking antenna 7 159
choice of 100 R input impedance is that a superior performance dual patch
element - described in Section 20.3 - can be fed directly and symmetrically from
a 50 Cl coaxial probe without the need for impedance transformers.

Table 20.1 Single patch design at 70 GHz

where Susceptance term Width Length Short-circuit to feed
(mm) (mm) (mm)
50R l OOR
Post and Stephenson [7] 11.9 4.17 1.35 1.94
Bahl and Bhartia [8] 11.9 - 4.59 1.40 2.02

Both expressions have been examined experimentally and the results are discuss-
ed in Section 20.2.5.
The width W of the patch is given by the following well-proven formula [8]:

Finally, it should be noted that it is, in practice, convenient to realise the short
circuit at one end of the patch by using a grid of conducting pins, as shown in
Fig. 20.4. However, this produces a residual inductance at the patch edge rather
than the desired perfect short circuit, so that the patch will need to be shortened
slightly. James et al. [9] provide an expression for the shift Alin the short-circuit
plane, based on the equivalent circuit of an array of posts across a parallel-plate
transmission line backed by an open circuit:
Fig. 20.2 Measured return loss for alternative parch designs
Aperture susceptance (from Reference 7)
-
- - Aperture susceptance (from Reference 8)

where r = radius of each post, a = distance between posts, 1, = free-space

The patch dimensions of Table 20.1 are given using both expressions for the
wavelength; A1 is positive when the short circuit plane lies away from the patch
aperture susceptance B, (eqns. 20.10 and 20.1 1) taken from References 7 and 8.
and behind the grid.
Fig. 20.2 shows the measured return loss for two patches built to the two
Eqns. 20.1 - 20.14 were incorporated into a patch design software package,
alternative designs. It is evident that the measured resonant frequency of
requiring inputs of patch width, substrate thickness and dielectric constant,
9.88 GHz for the patch designed using B, taken from Reference 7 is in better
input impedance and resonant frequency, and producing the patch resonant
agreement with theory than that designed using Reference 8, which resonates at
length and feed position as output data.
9.08 GHz. The former patch was therefore taken as the design.
However, the resonant frequency of the patch is still 120 MHz below the
20.2.5 Patch design
theoretical value. This problem was investigated further with the aid of liquid
Table 20.1 lists the computed patch dimensions for an input impedance of 100 R
crystal diagnostic techniques. Liquid crystals of the cholesteric variety possess
at a resonant frequency of 10 GHz. The choice of 100 R rather than the usual
a molecular structure which enables them to exhibit colour changes over a
50 R is twofold. First, the width of a 50 12 microstrip feed line on 1.57 mm thick
narrow temperature band. As the temperature increases, the crystals reflect the
RT/Duroid is 5.13 mm, which is close to 50% of the patch width, and hence an
spectral range of colours from red to blue over a temperature range dependent
unacceptable choice of feed. On the other hand, the width of a 100 R line is
on the concentration of a dopant additive. This technique has been used to
1.52 mm, and is thus suited to the patch dimensions. The second reason for the
observe the electric field in the vicinity of the patch. The approach taken was to

1 760 Conical conformal microstrip tracking antenna
mount a commercially available sheet of liquid crystal material, sensitive to a
temperature range around 35"C, onto a resistive card, and to place this card
over the patch, which was radiating several watts of R F power at the resonant
frequency. The resistive card is sensitive to the electric field component tangen-
tial to its surface and dissipates power as heat, which results in bands of colour
in the liquid-crystal sheet dependent on the electric field strength. Fig. 20.3
shows the results of this test applied to the patch, and indicates a concentration
Fig. 20.3 Liquid-crystal display showing coupling between patch and feed line, for patch
resonant at 9.88 GHz
of the fringing field around the centre of the radiating aperture. Under normal
conditions the dominant mode within the patch should yield a uniform electric
field strength across the aperture. This concentration is therefore indicative of
a coupling between the patch aperture and the microstrip feed line. This results
in an extension of the fringing field, equivalent to an apparent lengthening of the
patch resonant length and hence lowering the resonant frequency. By shortening
the patch in the ratio of the measured to the desired resonant frequency, the
patch was made to resonate at 10 GHz. Fig. 20.4 shows the final dimensions of
the single patch element design.
Conical conformal microstrip tracking antenna 7 161
20.3 Dual patch element
20.3.1 Choice of design
Endfire radiation may be further enhanced by constructing the basic array
radiating element out of two quarter-wavelength patches phased for construc-
tive interference in the endfire direction. Fig. 20.5 shows such an arrangement,
in which the two patches are set 'face-to-face' and joined by a straight length of
microstrip line. The feed point position has to be chosen to produce the required
excitation phase of each patch. Using the 100 R input impedance patch design
of the previous Section implies that the microstrip connecting line is of 100 R
impedance, thus making for a 50 R termination at the element feed point. This
dual patch element, with its 50 R feed point on a microstrip line, is eminently
suitable for a through-substrate probe connection to an underside feeding
network. The design of this element is now considered in detail.
U
loon
microstrip
Fig. 20.4 Final single patch design
Dimensions in rnm
20.3.2 Location of patch phase centre
In order to produce the required beam at an angle of 10" to endfire, correspond-
ing to the axis of the cone, it is necessary to know the location of the apparent
phase centre of each patch of the dual patch element. This position is some
 7 7 62 Conical conformal microstrip tracking antenna Conical conformal microstrip tracking antenna 7163

distance beyond the physical boundary of each patch, owing to the extent of the
fringing field. It was assumed that this extension dl (eqn. 20.12) defined the
postion of the phase centre, being exactly one quarter-wavelength in the dielec-
tric from the patch short-circuit plane. This position is indicated in Fig. 20.5.
Knowing the wavelength in the dielectric and the physical size of the patch from
Table 20.1 it may be shown that A1 x 1.2 mm at 10 GHz.
relate to RTIDuroid 5880 substrate with E , = 2.2 and thickness 1.57 mm.
VSWR measurements made on a prototype element, fed from below by a
coaxial surface launcher, showed a shift in the resonance frequency to 9.6 GHz
from the case of a single isolated patch. This was attributed to mutual coupling,
and the original 10 GHz resonance could only be restored by empirically
adjusting the notch depth of the patch closest to the feed probe. A reduction of

m-
approximately I mm was found to be optimum.

fi $-
forward
-- phasb centre

d~rection 19. 5 0.9A element size

elemenl
feed point

- mod~fied feed posilion

Fig. 20.5 Dual-patch element design

Dimensions in mm; other patch dimensions as for Fig.20.4

20.3.3 Design and optimisation

The distance between the patch phase centres was taken to be 0.9 wavelengths
3ci I O.45h element size

at 10 GHz. This value was chosen for several reasons. First it enables a more
directive four-patch array to be constructed, if desired, by interleaving two dual
patch elements with correct phasing, as shown in Fig. 20.6. The inter-patch
spacing of 0.451 would thus be suitable for minimising the grating lobe structure
in the endfire radiation pattern. If the size of a dual patch element were halved
(to 0.451) then interleaving would not be possible in a four-patch array, and'
more space would be required for the array. Fig. 20.6 contrasts these two
alternative constructions. The four-patch array was not used in the final con-
struction since the intention was only to investigate the performance of dual
patch elements mounted on the conical surface.
Secondly, mutual coupling between the two patches of the dual patch element
at 0.9 wavelength spacing is less than at 0.45 wavelength, thus requiring only
small changes to the original single patch design.
Finally, the two nulls in the far field pattern for the dual patch element (clearly
defined by the 0.9 wavelength spacing) enable an independent check to be made
of the true phase centre positions.
The feed-point position was chosen to phase the element at 10" to endfire with
due regard being taken of the 180" phase difference inherent in the face-to-face
array configuration. The dimensions are given in Fig. 20.5. These dimensions
Fig. 20.6 Comparison in overall width of four-patch array for different element size
w, < w2
Fig. 20.7 shows the measured far-field E-plane radiation pattern for the dual
patch element. From the angular positions of the two nulls, an estimate was
made of the actual phase centre position of each patch. This was found to be
1.4 mm from the physical boundary of each patch and is in close agreement with
the 1.2 mm prediction of Section 20.3.2. The radiation pattern displays a slope
beyond 60" from broadside and is attributable to the single patch element factor
which tends to fall off as endfire is approached. Nevertheless, at 10" from endfire
the fall off is only 3 dB below the peak level.
20.4 Hybrid feeding network
20.4.1 Overview
The most efficient means of generating a tracking signal using four circum-
 7 764 Conical conformal microstrip tracking antenna Conical conformal microstrip tracking antenna 1 165

ferentially mounted antenna elements is shown in Fig. 20.8 [lo]. Both sum and antennas positioned at 90'
mtervals around c~rcumference
difference channels are generated by a combination of three 180" hybrid cou-
plers, one 90" hybrid and phase-matched lines. It can be shown that each port
of the 90" hybrid will provide a progressive 360" phase shift around the cone -element polor~~tion
perpendicularto
circumerence, with opposite sign for each port. This corresponds to a circularly cone surface
polarised sum channel along the cone axis, with opposite hand of polarisation

1
A I
channel channel

0. I phase
(0). A phase

Fig. 20.8 Schematic of monopulse tracking feed network

hybrid
element

elevalion angle.deg antenna

Fig. 20.7 Measured E-plane radiation pattern of dual-patch element feed port

for each of the two ports, provided the other port is well matched. One port of
the first 180 hybrid will generate equi-phased signals at each antenna element. substrate
In this case a difference channel is formed which is sensitive only to horizontal 'periphery
polarisation in the horizontal plane, and vertical polarisation in the vertical
plane. The second port generates alternate 0, 180, 0, 180" phasing around the
cone circumference. This is a difference channel sensitive to vertical polarisation A
9
in the horizontal plane and horizontal polarisation in the vertical plane. With channels channels
the addition of one further 90" hybrid in front of the first 180" hybrid, these two Fig. 20.9 Schematic of triplate feeding network
 1 7 66 Conical conformal microstrip tracking antenna Conical conformal microstrip tracking antenna 7 767
modes can be generated in phase quadrature, resulting in a circularly polarised
difference channel. This last refinement has not been adopted in the current
design.
A monolithic version of the hybrid feeding network requires that the antenna
elements be situated within the confines of the circuit itself in order to avoid the
crossing of lines. This can be most easily achieved by placing the complete
feeding network on a triplate layer beneath the antenna substrate and joining
the two electrically by coaxial probes. A schematic arrangement of this circuit
is shown in Fig. 20.9, where each hybrid is represented in block form. -
The physical implementation of this arrangement is complicated by the need 4
to mount the assembly on the surface of a cone. The procedure undertaken was
to design the circuit on an opened-out section of the cone, using tracks of
0 -
in both cases
@ i @ "=soe
@ 1s loaded
variable curvature dependent on their distance from the cone apex. The triplate
substrate components have been chosen again to be RT/Duroid 5880 with
thickness 0.79 mm. The resulting triplate is therefore of the same thickness as
the antenna substrate, and hence uses the same bending procedure. The main
track impedance was chosen to be 50 0,since the corresponding track width of
1.21 mm is well suited to the limited available space, as well as being able to be
fed directly from standard SMA edge connectors. A further advantage is in the
fact that the antenna elements can be fed directly from 50 Q, as mentioned in
Section 20.3.1. The following Sections deal with the design of each circuit
component.

20.4.2 Hybrid designs Fig. 20.10 Atternative 9V hybrid configurations

The conventional narrow-band 90" hybrid coupler is essentially a four-port
device comprising four arms, each of length one quarter-wavelength at mid-
band. Two configurations are shown schematically in Fig. 20.10 for the case of
50 R output lines. The two configurations are, in effect, identical. However, the
second configuration is more suited to our purposes since it comprises, essenti-
ally, two continuous 50 R tracks joined by two parallel arms of impedance
5 0 1 4 = 35.35 Q. The operation of the coupler is to split a signal entering at
port 1 into two equal output signals at ports 2 and 4, but with a 90" difference
in phase. Port 3 is loaded with its characteristic impedance. The device is totally
symmetric in that port 3 could be used as the input port with port 1 loaded.
Implementation of the 90" hybrid in triplate is complicated by the effect of the
junctions on reference plane positions. Ifjunction effects were absent, each arm
of the hybrid would be exactly one quarter-wavelength in length and the
complete device would be square. However, each corner of the hybrid is actually
a T-junction, and may be represented as shown in Fig. 20.1 1 for the general case
of series arm impedance Z,, and shunt arm impedance Z,,, with corresponding
track widths W, and W2.The reference plane for each track is displaced from
the geometric centre by an amount dl and 4 , respectively. Consequently, the
series arms of the hybrid are extended by an amount 2d, and the shunt arms by
2d2. Reference 11 provides design curves for these displacement factors in Fig. 20.11 T-junction in triplate
1 168 Conical conformal microstrip tracking antenna Conical conformal microstrip tracking antenna 1 769

air-filled stripline, together with suggestions for the case when 8, > 1. For the The even and odd-mode characteristic impedances for coupled triplate lines
triplate configuration defined, these values are calculated to be d, = 0.23 mm are given by
and d, = 0.87 mm. Fig. 20.12 shows the layout of the final theoretical 90" 94.15
hybrid design operating at a mid-band frequency of 10 GHz. ZOe = Je. ( W/b + [ln2 + In (1 + tanh m/2b)]/n)
The 90" hybrid design may be simply converted into a 180' design by extend-
ing one of the output arms by one quarter wavelength. In this case the two 94.15
2
, = (20.16)
output signals are either equi-phased or displaced by 180, depending on which JE,{W/b + [In2 + In (1 + coth ns/2b)]/n)
input port is energised. and 2;; = Z , Z , (20.17)
Voltage coupling Cois given by

and CJb = - 20 logl0CO

where s = distance between coupled lines
W = width of each line
b = triplate thickness

Fig. 20.12 Final SO' hybrid design

Dimensions in mm

20.4.3 90" benak

Several 90" bends are required in the triplate circuitry, and consequently a
suitable design is required to minimise the bend VSWR. At the same time it is
necessary to know the effect of each bend on the reference plane in order that
the correct phasing can be applied to each antenna element. Fig. 20.13 shows a
mitred bend in which the optimum mitre dimension a for minimum VSWR has
been given in Reference 11 as a = 1.1 W for the triplate parameters under
consideration. For a 50 R track, W = 1.21 mm; hence a = 1.33 mm. The
electrical line extension associated with the bend is also given as Fig. 20.13 Mitredbend
I = 0.58 W = 0.7 mm. I= electrical length of bend between two reference planes

20.4.4 Minimum track distance

The narrow confines of the triplate substrate required that tracks be brought
close to each other. It is therefore necessary to know the minimum acceptable
distance between neighbouring tracks in order to avoid excessive cross coupling.
Howe [I21 provides formulae for the evaluation of cross-coupling, and these are
surnmarised in eqns. 20.15 - 20.19.
For 50 R coupled lines in 1.57 mm triplate with E, = 2.2, it can be shown that,
for s/b = 1, the coupling factor is -40 dB, and for s/b = 2 it is -66 dB. A
minimum distance between tracks corresponding to s/b = 2 was chosen, corres-
ponding to 5 mm between track centres.
 Conical conformal microstrip tracking antenna Conical conformal microstrip tracking antenna 1171
rn~crostrlp 20.4.5 Feed-point terminations
The triplate hybrid network is connected electrically to the antenna elements on
the upper microstrip layer via four probes soldered to the track terminations.
These probes possess a flat tab to enable a clean soldered joint to be made at
each triplate junction. A typical track termination is shown in both plan and
cross-section in Fig. 20.14. A discontinuity of this type in triplate, however,
tends to generate parallel-plate waveguide modes which emanate from the probe
probe tab and couple to adjacent tracks, degrading circuit performances. These modes
may be suppressed by a grid of conducting pins surrounding the discontinuity.
The grid design has been taken from the positions of the mode suppression
50a tnplate
screws in standard SMA surface-launch connectors. The grid is also shown in
track Fig. 20.14.

20.4.6 Track lengths

feed track 7.62 rnm
The individual track lengths of the feeding network must be accurately cal-
culated and realised in order to avoid any phase error at the antenna ports. The
situation is complicated by the need to design these tracks on an opened-out
cross-section plan (in tr~plate)
section of a cone. For convenience and to facilitate the drawing of the circuit
Fig. 20.14 Feed-point termination mask, the tracks were divided into two classes: radial and circumferential. These
are shown in Fig. 20.15 together with the shape of the substrate. The angular
width of the substate is given by the expression
cone
apex Om,, = 360 sin 0, degrees (20.20)

where 0, is the semi-angle of the cone of which the substrate is a part. For the
required 0, = lo0, Om, = 62.5'.
The circumferential track is just an arc of a circle of radius r centered on the
cone apex. Its length a, can be expressed in terms of its angular width AO:

7cr
a = AO-
180

where A0 is in degrees
The length of a radial track is simply its physical length along a radial line
from the cone apex.
This track classification is particularly useful in that all circumferential tracks
are parallel and are always perpendicular to radial tracks, thus simplifying the
design of 90" bends. The main design problem, however, is related to the hybrid
couplers. The use of radial rather than parallel tracks results in unequal circum-
ferential track lengths, as illustrated in Fig. 20.16, and thus introduces phase
radial track / errors. The typical phase error, however, is only 1.3' at each hybrid, resulting
clrcurnferent~al in a maximum of 2.6' for each antenna port. These errors are negligible in
track
comparison to that expected at the electrical connections between triplate and
Fig. 20.15 Types of triplate track on opened out conical surface antenna elements, and can thus be safely ignored.
1 1 72 Conical conformal rnicrostrip tracking antenni, Conical conformal microstrip tracking antenna 1 7 73
20.4.7 Overall design mode suppression
The hybrid network was designed for the cone (semi-angle 10' and length grid circle
300 mm) with the substrate extending to the base perimeter. The substrate width
was taken to be 55 mm, which is the minimum possible to enable all circum-
ferential tracks to be spaced apart by a minimum of 5 mm according to the
criterion given in Section 20.4.4. These dimensions are indicated in Fig. 20.15.
The circuit layout is shown in Fig. 20.17, where each track is represented by a
line. Extra quarter-wavelength sections of track have been added to create 180'
hybrids from 90" hybrids as required. All other tracks have been designed, with
due regard to their radial distance from the cone apex, to ensure correct phase
matching at each antenna port.
cone
output pwt \
apex \substrate
periphery

Fig. 20.17 Final hybrid-circuit layout

Fig. 20.16 Phase error at each hybrid

It is of interest to note the position of the first 90 hybrid in the difference

channel. This position, although resulting in exceptionally long input tracks,
gives the network an elegant symmetry and minimises the total number of bends
in the tracks. On bending the triplate around the cone, it also results in all input antenna element
ports being situated within the same region of the cone base.
radius 47.32mm

20.5 Conical antenna array

The dual patch antenna elements are printed on a microstrip layer lying imm-
ediately above the triplate hybrid network to form the conical antenna array. As
trtplate track
such, there is an inherent problem associated with the different radii of curvature termmatson
of the two layers. It is necessary to ensure that, on assembly, each antenna- Fig. 20.18 Geometry of triplatelmicrostri~interface
1 1 74 Conical conformal microstrip tracking antenna
element feed point lies immediately above its associated triplate output port to
enable the two to be joined electrically. Consider the situation as depicted in Fig.
20.18. It is assumed for the moment that both the microstrip and triplate layers
terminate at the base of the cone and that the connecting probe is in position.
From Fig. 20.17 it is known that the triplate track terminations are situated
midway along the substrate, and thus 272.5 mm from the apex of the 20' cone
of which the circuit is a part. The radius of the cone at this point is then 272.5
sin 10' = 47.32 mm. From the geometry it can then be shown that the antenna
element feed point, which is at a height of 2.38 mm above the circuit, is situated
cone
aoex
antenna micrdstrip substrate
element outline
Fig. 20.19 Layout of antenna array elements
on another coaxial 20" cone at a radius of 49.66 mm, corresponding to a
distance of 286.0 mm from the apex. Thus the feed points of all four antenna
elements on the opened-out microstrip substrate surface will lie on an arc of
radius 286.0 rnm centered on the cone apex. This is depicted in Fig. 20.19, where
the microstrip track of each element is oriented along the radius at that point
to ensure that, in the final assembly, all elements lie along the generatrices of the
cone. The base of the substrate was chosen to allow an 8 mm overlap by the
triplate on final assembly to facilitate the attachment of edge connectors. The
overall width of the microstrip substrate was taken to be 70 mm. This ensured
a one wavelength ground plane extension beyond the forward patches in order
to minimise edge diffraction effects. The antenna elements are situated at equal
angular intervals on the substrate as shown in the Figure.
Conical conformal microstrip tracking antenna 7 175
20.6 Substrate fabrication
20.6.1 Overview
There are several stages involved in the fabrication of the three substratelayers
comprising the antenna array and feeding network. The procedure commences
with the drawing of all the masks, corresponding to the two sides of each
substrate layer, on a CAD computer-graphics terminal. The output, in the form
of photoplots, are then reduced to the required scale via precision photographic
reproduction. The resulting masks are used in the substrate etching process. The
three substrates are then cut to the correct shapes. All necessary holes are then
drilled and various sections are milled out. Finally the two triplate components
are bonded together. At his point the substrates are ready for bending into the
required shape, prior to final assembly and attachment of electrical connectors.
20.6.2 Mask drawing and preparation
A standard CAD system was used to produce the masks. The main difficulty in
drawing the feeding network was due to the fact that the majority of tracks are
arcs of circles of large radius of curvature. This necessitated defining each
curved track by a set of three points which could then be used by the CAD
software to generate the curve passing through the points. By necessity, a
rectangular co-ordinate system was required, and the origin was defined at the
lowest point of the substrate periphery, denoted as '0' in Fig. 20.17. The
rectangular co-ordinates of each point are easily calculated from its radial
distance r from the cone apex and angle 0 from the central axis passing through
the origin:
y = 300 - r cos 0
x = r sin 0 (20.22)
where 300 is the radial distance in millimeters from the substrate lower periphery
to the cone apex. The substrate periphery was included in the mask to facilitate
cutting to the correct shape.
A second mask was drawn for the upper triplate ground plane comprising the
outer substrate periphery and four small circular non-metallised regions opp-
osite the antenna output ports. These regions prevent the probes from short-
circuiting to the ground plane and form a quasi-coaxial region with the probes,
of approximately 50 C2 impedance.
Outside the substrate periphery, each of the two masks was provided with
four fiducial crosses at the corners to enable the two to be lined up accurately
with each other during the etching process. A further mask was drawn, compris-
ing only the outer substrate periphery, to define the shape of the lower triplate
half.
In a similar manner the two masks for the microstrip antenna substrate were
drawn, one comprising the four antenna elements and the other containing the
four circular non-metallised regions at the feed input positions. The antenna
1 176 Conical conformal microstrip tracking antenna
mask was drawn using four identical dual patch elements oriented at the angks
shown in Fig. 20.19.
All masks were drawn to a scale of 2: 1, and the resulting photoplots were then
photographically reduced and printed in reverse on clear plastic film to form the
final masks.
20.6.3 Etching
Etching involves the removal of copper cladding in a controlled manner in order
to print the mask pattern onto the surface of a substrate. The process is standard
in a well equippedlaboratory and will not be described here. It should, however,
be emphasised that the alignment of the two masks defining both the triplate
circuit pattern and its upper ground plane (as well as those of the antenna
substrate) is crucial in order that the circular non-metallised regions in the
ground plane are accurately positioned at the antenna feed ports. For this
purpose holes were punched, centered on each of the four fiducial crosses on
each mask. Four additional holes were drilled through the substrate, and the
three layers were aligned by the use of dowel pins prior to etching. This process
gives an alignment accuracy of 0.1 mm.
Conical conformal microstrip tracking antenna 1 177
of Duroid were prepared for filling the cut-outs at the track input points. The
complete assembly is shown in Fig. 20.20.
20.6.5 Triplate bonding
Bonding of two substrates under high pressure and temperature, using an
intermediate layer of thermoplastic dielectric film, is a well established
procedure in microstrip circuit technology, and details are given in the Rogers'
literature [4]. For the case of RT/Duroid. the surfaces are pre-etched to provide
a key for the thermoplastic (Dupont FEP) film by immersion in a sodium-
naphthalene solution. Alignment of the two components was maintained during
bonding by dowel pins passing through both the bolt holes and edge connector
holes. A heated hydraulic press was used to provide the necessary pressure and
temperature.

20.6.4 Substrate preparation

After etching, the substrates were cut to shape and drilled while still in their flat
state. For the case of the upper microstrip layer, holes were drilled at marked
positions on each patch to take the short-circuit grid of conducting pins.
Additional holes, defined by the mask, were drilled through the substrate to take
the bolts through which the antenna would eventually be attached to the cone.
The position of the feed probe for each antenna element had been defined by a
small circle of etched copper on the microstrip track, and this was accurately
drilled to a diameter of 0.92 mm to take a standard OSSM probe.
In a similar manner, the various holes in the upper triplate layer were drilled.
These comprised: the antenna feed ports, the circular grid of mode-suppression
screw holes (defined by the mask), bolt holes (aligned with those of the micro-
strip substrate) and two holes on either side of the track input points to take the
triplate edge connectors. The two triplate halves were then brought together and
the various hole positions were marked through to the lower substrate, includ-
ing the positions of the track input points. The bolt and edge connector holes
were then drilled as before.
It is impossible to solder the antenna feed probes into position prior to
bonding the triplate halves and subsequent bending. Consequently, in order to Fig. 20.20 Substrate components
gain access to the track terminations, a circular hole centered on the output
probe position and of 10 mm diameter was milled out at each of the four
20.7 Forming the antenna
antenna feed points on the lower triplate half. In addition, a semicircular cut-out
was milled at each track input point to enable the edge connectors to be
20.7.1 Bending the substrates
attached. Four Duroid discs were punched out to fill the feed-probe holes, and
It is possible to bend RT/Duroid at room temperature. However, the large
were drilled to take mode-suppression screws. Finally small semicircular discs
degree of springback imposes an excessive strain on the substrate when it is
 1 1 78 Conical conformal microstrip tracking antenna Conical conformal rnicrostrip tracking antenna 7 7 79

clamped in postion and may result in a fracture. It is necessary to first stress- into place on both the patch surfaces and the ground plane. These can be seen
relieve the substrate by passing it through a temperature cycle which exceeds the in Fig.20.216, which shows the microstrip layer in its final shape, prior to
PTFE glass/rubber transition temperature of 130C; beyond this temperature, assembly.
stresses are minimised [4]. For this purpose, two conical aluminium mandrels
were fabricated for bending both the triplate and microstrip layers. Each
mandrel contained an inner depression for seating the substrate, and an outer
depression to take a stainless-steel clamping band. In each case the substrate and
band were bent slowly around the mandrel at room temperature and held in
postion with studs. The assembly is shown in Fig.20.21~.The mandrel was then
placed in an oven at 100C in an inert atmosphere. The temperature cycle
consisted of a warming-up period of one hour to 180C, one hour at that
temperature, and a one hour cooling-off period. Upon disassembly, no major
faults were seen in either the dielectric or the copper ground plane, except for
some very minor wrinkling in the copper. The smooth walls of both the mandrel
and clamping band provided excellent surfaces against which the substrate
could rest and thus avoid distortion. Any springback tendency for both sub-
strates was found to be minimal.

b the antenna substrate

a the mandrel

Fig. 20.21 Bending the substrates

20.7.2 Attachment of components c the triplate (fully assembled)

Once the substrates have been bent into shape, the various probes, pins and
mode-suppression screws can be attached. For the upper microstrip layer, the
short-circuit grids of pins were inserted into the patch elements and soldered
 1180 Conical conformal microstrip tracking antenna
For the case of the triplate layer, the antenna feed probes were inserted
through the access holes in the lower substrate and their tabs soldered into place
on each track. Small relief channels were cut into the circular Duroid discs into
which the soldered tabs could sit without creating an air gap, and the discs were
pushed in place into the four access holes, ensuring that the mode-suppression
screw holes lined up with those of the upper triplate half. The discs were held
in position by the mode-suppression screws themselves. The countersunk screw
heads were embedded into the upper ground plane to produce a uniformly
smooth surface. Both the upper and lower ground planes around the screws
were coated with a conducting silver paint to ensure good electrical continuity.
In order that the probes simulate a coaxial line, it is essential that good
electrical continuity exists between the upper triplate ground plane and that of
the microstrip in the vicinity of the probes. It is very unlikely that this can be
guaranteed when the two substrates are finally brought together because of the
large surface areas in contact with each other. For this reason thin metallic
shims, milled from the flanges of standard OSSM surface launchers, were
inserted past the probes and bonded to the ground plane with conducting epoxy.
The shims were bent slightly to conform to the shape of the triplate surface. The
final triplate assembly is shown in Fig. 20.21~.
Fig. 20.22 Final antenna assembly
On mating the triplate and microstrip substrates together, the shim surfaces
were coated with a thin layer of conducting epoxy to ensure good electrical
continuity between the substrates. The probes were then inserted through the
Conical conformal microstrip tracking antenna 7 18 7
feed holes in the microstriplines and the two units bolted together with nylon
screws. Finally the probes were soldered into place and DC electrical continuity
was checked between the exposed triplate track terminations and the patch
surfaces.
20.7.3 Final assembly
The cone itself was machined from aluminium. A de~ressionat the rear end
allowed the triplate to seat properly while the upper microstrip layer could rest
on the main body of the cone. The remaining bolt holes were used to attach the
assembly to the cone using nylon screws. The free ends of the substrates were
brought together and held in position with Duroid fishplates against the press-
ure of the residual springback.
The final procedure involved the attachment of the triplate coaxial connec-
tors. These were standard OSM in edge launchers. Curved shims were fabri-
cated to interface between the flat connector flanges and the curved substrate
surface. After soldering each centre tab to the exposed triplate track, the access
cut-out was filled with the semicircular Duroid disc and the complete connector
unit with shims was screwed into position. The final antenna assembly is shown
in Fig. 20.22.
20.8 Antenna performance
Electrical measurements have been carried out on the conical conformal anten-
na over a 1 GHz bandwidth centered at 10 GHz. Both sum and difference
radiation patterns have been obtained from all four ports of the antenna using
a linearly polarised source antenna. With reference to Fig. 20.8, the scan axis is
taken to be the x-axis with the individual antenna elements oriented at 45' to
this axis. In this way, both the difference-channel ports could be tested individu-'
ally without needing to rotate the cone. From Section 20.4.1 it will be recalled
that the two difference channels are sensitive to orthogonal linear polarisations;
thus the equi-phased channel will receive horizontally (x-directed) polarised
signals, whereas the alternate 0, 180" channel will receive vertically @-directed)
polarised signals.
Measured radiation patterns at 9.5, 10.0 and 10.5 GHz are shown in Fig.
20.23 for the two sum channel ports and in Fig. 20.24 for the two difference
ports. Fig. 20.25 shows the measured response in one sum channel to rotating
linear polarisation at 10 GHz. Finally, Fig. 20.26 shows the measured antenna
boresight gain matched to the linearly polarised source antenna for vertical
@-directed) polarisation.
The antenna-performance results of Fig. 20.23-20.26 require some detailed
explanation, because, on careful consideration, many characteristics cannot, of
course, be explained by recourse to conventional (planar) array behaviour. The
subjects of grating lobe supression, axial ratio, gain and tracking slope are now
 Conical conformal microstrip tracking antenna 7 183
1 182 Conical conformal microstrip tracking antenna
considered in more detail. This treatment is largely qualitative owing to the demand explanation. The most prominent feature is the lack of grating lobes in
electromagnetic complexity of the problem and the bounds of this book. A the sum channel for horizontal polarisation, whereas those for vertical polarisa-
detailed mathematical treatment of conformal array antennas may be found in tion are well developed. This phenomenon may be explained, most easily in the
References 13 and 14.
porl 1 - horizonlal polarisation port 2 -vertical oolarisotion
port 1 port 2

m
P
-20

-30 9.5 GHz 9.5 GHz

-301 I 9.5 GHz -301 9.5 GHz
ri
-40 -20 0 20 40 -40
u -20 0 20 40
-40 -20 0 20 40 -40 -20
Azimuth angle (deg. )
0 20 40

Azimuth anale (dea

Or Or

m
u
-2 0

-30 10.0 GHz

I I , L
I I -40 -20 0 20 40 -40 -20 0 20 40
-40 -20 0 20 40 -40 -20 0 20 40 Azimuth angle (deg.)
Azimuth angle (deg.)
0 0r 0r

-10

B -2 0
-20
I \
\ I \
\ I -30 10.5 GHz -3 0
-30 \ l 10 5 GHz
\I
-40 -20 0 20 40 -40 -20 0 20 40
-40 -20 0 20 40 -40 -20 0 20 40
Azimuth angle (deg.) Az~muthangle (deg.)

- vertical
-- horizontal 1 polarrsatlon Fig. 20.24 Difference channel radiation patterns (measured)

Fig. 20.23 Sum channel radiation patterns (measured)
20.8.1 Grating lobe suppression
Although the sum and difference patterns are well defined over the frequency
band in the angular region close to boresight, there are several features which
transmit mode, by the difference in the fields radiated towards an observation
point between a visible antenna element and one hidden by the presence of the
cone. Use of GTD surface-field methods in the analysis of radiation from
conical arrays is well documented [15] and only a summary need be given here.
With reference to Fig. 20.27, each diametrically opposed pair of array ele-
 1184 Conical conformal microstrip tracking antenna Conical conformal microstrip tracking antenna 1 185
ments are equi-phased with respect to the observation point in the sum channel.
The opposite pair, in phase quadrature, may be ignored in this explanation.
Grating lobes will appear when the path difference between diametric elements
at the observation point is an integral number of wavelengths. The cone dia-
meter at the centre of the array elements is 110 mm, and therefore the horizontal
distance between diametric pairs is 1 1 0 ~ = 4 78 mm. At 10 GHz the condition
for grating lobes is that sin 0 = &Id = 30178 = 0.384; this corresponds to
0 = ) 22.5', and indeed corresponds exactly to the measured positions at
10 GHz for vertical polarisation. At 22.5' from boresight, one array element is
in the lit region and is therefore sensitive to both vertical and horizontal
polarisations. The diametrically opposed element, however, is in shadow. In this

iurface ray
- 40 -20 0 20 LO I
Fig. 20.27 Rad~atlonfrom a h~ddenarray element on a conical surface
A z ~ m u t hangle (degrees)

Fig. 20.25 Rotating linear sum pattern at 10 GHz (measured) case a surface ray creeps around the cone on a geodesic with respect to the
observation point, radiating tangentially as it proceeds. The ray that is eventu-
ally directed towards the observation point will radiate from the cone horizon
as perceived by the observer. Now the magnetic field vector associated with the
surface ray remains parallel to the cone surface over which it travels, and
therefore the observer will perceive this vector as parallel to the cone horizon,
which itself subtends a small angle to the horizontal at the 22.5" scan angle. This
implies that the diffracted field is predominantly vertically polarised as seen at
I
the observation point, and will thus contribute to the observed grating lobes
seen in vertical polarisation, but not to horizontal polarisation. This further
i1 implies that the grating lobes in the vertical polarisation pattern are about 6 dB
higher than the mean signal level at the same scan angle in the horizontal
1
I
polarisation pattern. On normalising the co-polar peaks to each other in Fig.
20.23 this is seen to be indeed the case.

20.8.2 Axial ratio

The second important feature requiring explanation is the difference in
boresight amplitude in each sum channel between vertical and horizontal pol-
Fig. 20.26 Measured antenna boresight gain in sum channel for vertical polarisation
1 186 Conical conformal microstrip tracking antenna Conical conformal microstrip tracking antenna 1187
arisations. The effect is reversed between the two ports, so that for one port the
vertical component exceeds the horizontal by an average of 3.5 dB across the
frequency band, whereas for the other port the reverse is the case. The effect is
clearly visible in the response to rotating linear polarisation, as seen in Fig.
20.25.
~ , = c o s ( w t + 2~ * ~ )
E, ' C O S W ~
\

Thus the maximum amplitudes in the vertical and horizontal directions are
given by:

and the ratio in decibels is

X
Fig. 20.28 Effect of phase error in the sum channel 2 factor changes sign and:
For the opposite hand of circular polarisation the -

This feature may be explained if it is assumed that a constant phase error

exists at the input to each antenna array element. This may be due to differing
path lengths at the interface between the triplate and microstrip layers. Consider
the situation depicted in Fig. 20.28 for the case of two adjacent elements in phase
quadrature, each element polarised at 45' to the vertical. Let one antenna

r3
element possess a constant phase error 4 with respect to the other, both
radiating at equal amplitude. The time-dependent electric field amplitudes for
one hand of circular polarisation are then: = - I0 log,, cot2 - = -Rm,
Element 1 : El = coswt
This is equivalent to saying that the difference between vertical and horizontal
Element 2 : E2 = cos(wt + 4 + Z) X
polarisation is reversed between the two sum channels, as seen in the measured
results. The average measured difference of 3.5 dB is then equivalent to an
The resultant component in the + y direction is then (ignoring constants): inherent phase error of 22.5'. It must be remembered, however, that this phase
error is distributed around the four antenna elements in a random manner. It
Ey = coswt + cos(wt + 4 X
(20.24) is therefore clear that the inherent phase error in each element is typically 5.5'.

and in the + x direction: 20.8.3 Antenna gain

E, = coswt - cos(wt
The measured antenna boresight gain, shown in Fig. 20.26, depends to a great
+ - 4 + -)2
X
extent on the tilt angle of the polarisation ellipse, which in turn depends on the
inherent element phase error as discussed above. As such, the measured am-
7 188 Conical conformal microstrip tracking antenna Conical conformal microstrip tracking antenna 1 189
plitude is somewhat irrelevant, and it is the shape of the response as a function radius of the antenna substrate, resulting in a spacing of several wavelengths
of frequency which is of interest. The gain around 10 GHz exhibits a sinusoidal between adjacent elements. However, it is noted that the measured 12' coverage
frequency dependence which drops off rapidly above 10.3 GHz. Two mechan- could be acceptable for many purposes.
isms are attributable to this shape. In the former case the element phase error
is changing as a function of frequency, thus changing the tilt angle of the
polarisation ellipse, resulting in a fluctuating vertical polarisation component.
The latter effect is attributable to the bandwidth of the elements. With reference
to Fig. 20.2, the single patch return loss increases to - 3 dB at 10.5 GHz. This
implies that half the power is no longer radiated. The gain is therefore expected
to drop by 3 dB from that at resonance. It can be seen in Fig. 20.26 that this is
indeed the case.

20.8.4 Tracking slope

Fig. 20.29 shows the tracking characteristics at band edges and centre extracted
from the sum and difference patterns. The tracking parameter is defined as the
ratio, in amplitude, of the difference signal to the sum signal, and is plotted as
a function of scan angle for both vertical and horizontal sum channel patterns.
It is important that the tracking slope remain constant over the angular region
of interest in order that a single parameter at each frequency may be used to
locate the angular position of a target. It is evident from the Figure that, for the
conical conformal antenna, this is true over an angular width of about 12'
across the frequency band; although some degradation is seen at 10.5 GHz. This
angular width is totally dependent on the positions of the two main lobes in the
difference channel, which are in turn dependent on the cone circumference at the
antenna elements. The only means of increasing this coverage is to mount the
elements closer to the cone apex. With the mechanical construction described in
this Chapter this is not possible. The 12" angular coverage is thus a limitation
of the present design, but may be acceptable for many purposes.

20.9 Conclusions and future developments

This Chapter has described in detail the application of microstrip technology to

a particularly demanding requirement - the fabrication of a conical microstrip
tracking antenna. The antenna element design has centered on the need for
acceptable endfire performance and a suitable solution has been described. The
problem associated with the siting of the tracking circuitry has been solved by
the use of an independent triplate feed network located beneath and adjacent to -
----
X (vertical)
X (horizontal)
the printed antenna substrate. This, however, raises additional problems in the
electrical interfacing of the two units when mounted in position on the surface Fig. 20.29 Tracking characteristics of conical conformal antenna (measured)
of the cone. One possible technique has been described in detail. The antenna
electrical performance has been measured and analysed, and suggestions have
The development and fabrication of the antenna has shed light on the various
been given to explain some of the peculiarities and trends in the data. Angular
engineering problems associated with the design of a cone-mounted system. In
tracking coverage is found to be somewhat limited owing to the minimum bend
particular, the need for designing the antenna and feed circuitry on an opened-
 1190 Conical conformal microstrip tracking antenna
out conical surface has emphasised the procedures necessary for non-linear
track layouts. It is evident, however, that several enhancements to the design are
possible in order to improve further the electrical performance and to facilitate
easier manufacture; these are now briefly described.
The antenna bandwidth, being a function of both the thickness and dielectric
constant of the substrate, may be increased by increasing the former and
decreasing the latter. However, the improvement is limited by the onset of
higher order modes, which result in additional circuit inductance and degraded
VSWR. The effect has been used to advantage in a novel procedure outlined by
Griffin [16]. He uses the unwanted inductance, together with that of the probe
feed, of a microstrip disc antenna to design an impedence-matching network in
the triplate feed track. He reports impressive performance, using a thick, low
dielectric constant (8, = 1.2) substrate, with a VSWR bandwidth of 20%. The
technique is applicable to any probe-fed patch antenna.
Increased antenna directivity at endfire is desirable to reduce the high sum
channel grating lobes seen in Fig. 20.23. It can readily be shown that an eight
element linear patch phased array will suppress all sidelobes in the forward
hemisphere to below - 10 dB over the 20% bandwidth, provided the inter-
element spacing is less than 0.45 wavelengths.
antenna elements
printed on this layer
antenna substrate
(expanded foam)
upper triplate
(power divlsion and
broad bonding)
-lower tr~plate
(tracking)
Fig. 20.30 Cross-section through a broadband conformal antenna
A single triplate circuit containing the hybrid tracking network together with
power-division and impedance-matching components is not feasible within the
limited space available over the cone circumference. A more attractive solution
is to share the circuitry between two triplate layers linked together electrically.
By careful choice of track impedances, the two layers could be made extremely
thin and bonded together as one unit before being bent into shape.
The use of a thick low-permittivity foam as the antenna substrate for broad-
band performance means that the antenna elements cannot be printed onto such
a surface. A possible solution would be to print the arrays onto a thin sheet of
copper-clad Kapton or similar material, and then to bond this to the foam,
which will have been pre-formed into conical segments and bonded to the
triplate. The final antenna assembly would therefore comprise six substrate
layers as shown in Fig. 20.30.
Conical conformal microstrip tracking antenna 1191
Acknowledgment
The authors wish to thank the directors of the Marconi Company Ltd. for
permission to publish this chapter, which is an expanded version of a paper
presented at the Military Microwaves 1986 Conference held at the Metropole
Hotel, Brighton, England, 24th-26th June 1986.
20.10 References
JOHNSON, R.C., and JASIK, H. (1961): Antenna engineering handbook, (McGraw Hill,
NY, pp. 7-10)
RUDGE, A.W., et al. (1986): Handbook of antenna design, (Peter Peregrinus) Section 7.4
JAMES, J.R., HALL, P.S., and WOOD C, (1981): Microstrip antenna, theory and design,
(Peter Peregrinus) Deduced from Fig.3.16~
Mektron Circuit Systems Ltd. (1983): Product information-RTIDuroids; and private com-
munication
HALL, P.S., WOOD, C., and JAMES, J.R. (1981): Recent examples of conformal microstrip
antenna arrays for aerospace applications. 'Antennas and Propagation'. IEE Conf. Publ. 195,
Pt. 1, pp. 397-401
WEINSCHEL, H.D. (1979): Measurement of various microstrip parameters, Proc. Workshop
on Printed Circuit Antenna Technology, New Mexico State University, pp. 2-1
POST, R.E., and STEPHENSON, D.T., (1981): The design of a microstrip antenna array for
a UHF space telemetry link, IEEE Trans. AP-29, pp. 129-133
BAHL, LJ., and BHARTIA, P. (1980): Microstrip antennas (Artech House, London)
Reference 3, Section 4.6
Reference 2, Section 12.6
MATTHAEI, G.L., YOUNG L., and JONES, E.M.T. (1964): Microwave filters, impedance
matching networks and coupling structures (McGraw Hill, NY Section 5.07)
HOWE, H. (1974): Stripline circuit design (Artech House, London)
RIZK, M.S.A.S., MORRIS, G., and CLIPTON, P. (1985): Projected aperture synthesis
method for the design of conformal array antennas. 'Antennas and Propagation' IEE Conf.
Publ. 248, pp. 48-52
SHAPIRA, J., FELSEN, L.B., and HESSEL, A. (1974): Ray analysis of conformal antenna
arrays, IEEE Trans, AP-22, pp. 49-63
Reference 2, Section 11.6
GRIFFIN, J.M. (1985): Broadband microstrip disc antenna for satellite communications,
M.Phil./PhD Transfer Thesis, GEC Hirst Research Centre, Wembley, Middx.
 Chapter 21

Microstrip field diagnostics

P. G. Frayne

21.1 Introduction

In this Chapter we are concerned with the relatively undeveloped subjec of

surface-field metrology for 'open' microstrip and other related 'open' plaiiar
transmission structures. Early measurements on microstrip lines were aimed at
determining the dispersive properties of the medium using a number of different
techniques, some of which are listed below:
(a) Measurements on open-circuited and short-circuited resonant microstrip
lines
(6) Ring-resonator techniques
(c) Variation of phase shift of a line with frequency
( d ) Nodal-shift techniques
None of these methods, however, provide much insight into the detailed field
distributions that actually reside on the conductors. Liquid crystals have been
employed to render mode patterns and regions of high electric field 'visible' to
the unaided eye, but the spatial resolution and general applicability of the
technique is very limited. The growing need for reduction in the size, weight and
cost of millimetric guidance and radar equipments has stimulated the design of
circuits with a much higher degree of component integration than is possible
with the modular approach to circuit construction. With the growing complex-
ity of circuit-integration techniques, both fault location and test procedures
become increasingly difficult. The accessible R F input and output ports provide
little information about the detailed internal operation of the system, and the
conventional network analyser, much valued for the assessment of isolated
modular components, is of limited use as a diagnostic for large-scale circuits.
The use of a greater degree of component integration not only makes circuit
evaluation difficult, but can also create severe problems owing to unwanted
electrical coupling between components.
 1194 Microstrip field diagnostics Microstrip field diagnostics 1195
21.2 Surface analytical techniques the probe circuitry. A highly resistive probe circuit was also considered desirable
in order to reduce the possibility of internal probe resonances and reflection of
To the author's knowledge there have been at least two prior experimental the microwave signal from the probe metallisation.
investigations of possible surface diagnostic techniques for microstrip, namely
the work of Dahele and Cullen [I] and also that of Ladbrooke [2]. The detailed
plots of the near-field radiation pattern of a polyrod antenna obtained by
Neumann [3, 41 are also of interest in the context of field-plotting techniques.
In the work of Ladbrooke, a special jig was constructed which enabled the top
microstrip conductor, defining the circuit pattern, to be moved independently of
the ground plane and the 50R microstrip feed line. The ground plane contained
a small disc probe 0.1 mm in diameter, separated from the surrounding region
by an annular gap. The thin dielectric sheet which supported the microstrip
circuit was pressed into contact with the ground plane and feed line. The probe
disc was connected to the inner conductor of a miniature coaxial cable whilst the
outer conductor was grounded. Connection of the microstrip feed and probe
output signal to a vector network analyser enabled the circuit transmission
(amplitude and phase) to be measured over a wide frequency range. The
measurements presented in Ladbrooke's paper appear to be restricted to lon-
gitudinal distributions on open-circuited microstriplines. Evidence for the ex-
citation of surface waves beyond the edge of the open microstrip termination
was discussed in a later paper. It is interesting to note that a distortion of the
standing-wave pattern similar to that discussed in the books by Ginzton [5] and
Fig. 21.1 Scanning-nework probe-plotting table
MOI'-ornery [61. due to probe reactance, was actually observed, but accredited
by Ladbruoke to a mismatch of the feed line.
The investigation by Dahele and Cullen was more concerned with evaluating, 21.3 Scanning-network probe
the response of a small coaxial probe with an extended inner conductor to a
calculable RF electric-field distribution produced by a boxed cylindrical con- The scanning network probe was originally intended for use as a circuit diagnos-
ductor. A wire, 1.6 mm in diameter was supported along the axis of wide rebate tic of high frequencies above the normal working range of commercially avail-
(of rectangular section) cut in a metal plate. The rebate was closed off by a able network analysers. With this application in mind, a mechanically stable,
movable plate which also supported the coaxial probe. The inner probe conduc- high-precision two-axis transport mechanism shown in Fig. 21.1 was construc-
tor protruded about 2mm into the waveguide region and the probe tip was ted. A conventional leadscrew drive was employed which had a minimum incre-
6.4 mm from the surface of the wire. The experimentally determined VSWR mental step size of 0.5pm. Owing to the extreme fragility of the probe sensors
pattern agreed well with theoretical predictions, and the experiment was subse- necessary at frequencies above 100GHz the majority of the data presented in
quently repeated after replacing the cylindrical wire by a flat microstrip line. this Chapter was obtained over the frequency range 26-40 GHz with the coaxial
In a more recent paper accredited to Schwarz and Turner [7], a co-planar probe shown in Fig. 21.2. The probe is very simply constructed from a rigid
waveguide probe is discussed which uses a miniature bismuth bolometer for the single-bore alumina tube of 0.5 mm external diameter. The alumina is metallised
direct detection of the microwave signal at the probe tip. Although the bol- on its outer surface and the inner conductor extends approximately 0.2mm
ometer can be made very small, it suffers from the disadvantages of being beyond the end of the tube. The other end of the probe is terminated in an
restricted to scalar measurements and also the need to be specially constructed adjustable cylindrical post, which, in conjunction with a tunable backshort,
for each width of co-planar waveguide under investigation. effectively launches a TE,, waveguide mode. The coupled signal from the probe
Co-planar waveguide probes have been investigated in our own laboratory is fed to a dual-channel homodyne (phase and amplitude) detection system. In
where a beam-lead Schottky barrier diode was employed for detection purposes. order to eliminate the need for soldering the ground plane to a flat rigid backing
In order to avoid over-coupling the probe to the co-planar waveguide circuit, it plate, a vacuum clutch was employed, which had the additional facility for
was necessary to employ a very thin low-permittivity substrate for supporting mounting coaxial microstrip launchers anywhere around its periphery. Since the
 7 796 Microstrip field diagnostics
probe shown in Fig. 21.2 is axisymmetric, it may be used for recording the
distribution of electric field over any irregularly shaped planar circuit configured
in open microstrip, slotline, co-planar waveguide or co-planar stripline. The
completely automatic data storage and processing facility is extremely versatile
in so far as it provides a range of options such as monochrome and colour-
contouring routines or a three-dimensional representation of the measured data.
The contours may also be sectioned so as to generate a plot of the voltage-stan-
ding-wave ratio. An axis of symmetry would normally be selected for this
purpose so as to ensure (a) the probe-strip capacitance remains sensibly con-
stant along the line of measurement and (b) the magnetic coupling c a n c e ~ ~theat
VSWR maxima and minima.
Fig. 21.2 26-40 GHz monopole probe
When the tip of the probe approaches an edge discontinuity, the capacitive
coupling to the top conductor rapidly decreases. The recorded field intensity,
however, does not fall off as rapidly as might be expected, owing to a local
enhancement of the surface charge and current densities due to the skin effect.
Close to an edge, the recorded distribution can only be regarded as a qualitative
indication of the true excitation. However, this apparent failure of the technique
does not seriously detract from the physical insight into high-frequency circuit
behaviour that is revealed by the tutorial nature of the contour maps. For many
purposes it is irrelevant that the magnitude of the charge and current densities
Microstrip field diagnostics 7 197
on the two surfaces of the top microstrip conductor are different. It is significant,
however, that the edge conditions for the two surfaces are identical, which leads
to a common axial periodicity of the charge and current densities on a micro-
strip line. for, example.
z
Fig. 21.3 Probe geometry for deriving near-field pattern of a monopole from an equivalent
azimuthal magnetic current M+
21.4 Theory of the monopole probe
An exact theory for the near field coupling between a short monopole and a
narrow conducting strip does not exist at the present time. A realistic theory for
the charge and current-density distributions on the top microstrip conductor
would need to take into account the lateral skin effect and the proximity effect
due to the ground plane. Reference 8 and Fig. 21.3 show that if the probe were
considered to be a transmitter located in free space, it would generate both
radial and axial electric-field components in the near field. Expressions for these
components may be calculated in terms of an equivalent azimuthal magnetic
current density k , / ~flowing within the annular aperture of the monopole.
7 198 Microstrip field diagnostics
Expressions for the field components are given in eqns. 21.1 and 2 1.2. Converse-
ly, in reception, the monopole will respond to the net radial and axial field
components generated by the microstrip line.
When the monopole is located above the symmetry axis of the microstrip line
it will respond not only to the vertical component of the electric field but also
to the surface-field gradient taken in the direction of the symmetry axis. The
latter form of coupling alternatively may be considered magnetic in origin, as
shown in Fig. 21.5. The total coupling to the probe arises from both the
conservative and non-conservative sources of electric field, which may be ex-
pressed in terms of the vector magnetic potential A , given by
The derivation of an expression for the magnetic vector potential at a height h
above a flat conducting strip of finite width and length is straightforward for the
case of a steady axial current I;. The appropriate expression is given by
Strictly, the derivation of this expression assumes the presence of a coaxial
return circuit of infinite radius.
In the case of a high-frequency current in the form of a travelling wave, a
number of additional phenomena need to be taken into account. The most
improtant of these is the rapid variation of phase over the dimensions of the
strip and also the lateral non-uniformity of the surface current due to the skin
effect. Owing to the close proximity of the probe tip to the surface of the flat
conductor, the effect of electrical-image charges should also be considered in
any realistic calculation of the probe response. When the lateral skin effect is
neglected and only the axial phase variation associated with a uniform travelling
wave is taken into account, the resulting integral equation 21.5 for the axial
component of the magnetic vector potential cannot be solved analytically in the
near field:
cos (kh tan 8) sec 8 exp (- jkh sec 0) dB (2 1.5)
Microstrip field diagnostics 1799
In the majority of practical applications of the probe technique, the surface
fields arise from a standing-wave distribution rather than a simple travelling
wave, so it is more appropriate to calculate the contributions to A: from each
half-period zone along the axis of a narrow strip as given in eqn. 21.6 and shown
in Fig. 21.4, where 0 is confined to the Y Z plane:
Fig. 21.4 Geometry for computing the magnetic vector potential at a height h above an
isolated conducting strip
In the case of a large VSWR on the strip, the computation of A, is modified to
some extent by the fact that the phase varies discontinuously along the length
of the line, increasing rapidly by n at the current maxima and remaining almost
constant at the voltage maxima. Since the amplitude of the probe signal falls off
rapidly with increasing height h, the slow variation of phase at the voltage
maxima could be neglected. However, the rapid phase variation at the voltage
minima rather implies that the spatial resolution or radius of the probe 'foot-
print' will be a function of the standing-wave ratio and frequency, and not
purely a constant geometrical factor dependent on the probe diameter.
By developing the idea of a combined response to the vertical component of
the electric field and transverse component of the magnetic field, a simple
phenomenological theory for the probe coupling has been derived.
Reference to Fig. 21.5 shows that a quasi-static contribution to the probe
signal is visualised in terms of more electric lines of force terminating on the
central conductor than the sheath owing to its proximity to the charged micro-
strip. Increasing the length of protrusion h will cause more lines of force to
terminate on the central conductor and a reduction in the number of lines
terminating on the sheath, which results in a larger potential difference V being
induced on the open-circuited coaxial line. The voltage V can be expressed in
terms of line integrals around the contours a-b-a and a'-&-a' in a vertical
1200 Microstrip field diagnostics
section through the probe taken along the symmetry axis of the microstrip. Thus
v = r ~ . d i +J : , ' ~ . d i (21.7)
For any other probe location, the line integrals should be replaced by surface
integrals over the probe aperture and tip. The circuital EMFs (5,) from the
non-conservative sources-may be found from a similar contour integration.
Under the assumption that the probe capture area is determined by the sheath
radius b and the probe height h, the circuital E M F may be expressed in terms
of the mean phase error 4, ( 4 = b&) as given by
outer conductor
radius b
dielectric
X Inner conductor,
rad~us a
I
Fig. 21.5 Schematic of coaxial monopole probe (2 figs.)
If the electric coupling reis taken to be predominantly due to the vertical
component of the potential gradient E,h, the magnitude of the total coupled
electric field k;, and associated phase angle A are given by eqns. 21.9 and 21.10:
I ~ =, I~ , h { l- ( b ~ , ) ~ ) (21.9)
A z tan-' -
{1 2Bej
The second term in eqn. 21.9 represents the magnetic contribution to the total
measured electric field associated with a travelling wave, and may be regarded
as an error arising from unwanted coupling. The variation of the amplitude and
phase errors as defined above are plotted in Figs. 21.6 and 21.7 as a function of
increasing frequency for three different probe diameters.
If the probe is used to measure the voltage standing wave ratio of a mismatch-
ed microstrip line, the magnetic contribution cancels at both the current and
voltage maxima owing to the symmetry of the tangential magnetic field along
the axis of the microstrip. At some intermediate axial position or a some point
Microstrip field diagnostics 120 1
not on the symmetry axis, there will be a net magnetic contribution to the total
measured electric field. The analysis of lines in terms of the voltage standing
wave ratio implicitly requires that the probe sensor responds either to the
conservative electric field or the magnetic field, but not a combination of both.
Under these circumstances the vertical component of electric field close to the
surface is proportional to the local surface charge density, whereas the-local
transverse component of the magnetic field is proportional to the axial current
density. Figs. 21.6 and 21.7 give some indication of the probe size necessary for
a particular frequency range. In high-permittivity lines the magnetic coupling is
relatively large and is proportional to the effective permittivity.
10
frequency IGHz
I
I'
Fig. 21.6 Travelling-wave amplitude error against frequency (e,, = 7.7)
The extension 6 of the central probe conductor beyond the dielectric tube and
co-axial screen has a significant effect on the voltage sensitivity and spatial
resolution of the probe. Whereas the sensitivity is easily seen to be proportional
to 6, the resolution cannot be readily quantified. As shown in Fig. 21.4, the
electric coupling must be determined by summing contributions from an area of
line equivalent to the probe 'footprint'. Although this area is difficult to cal-
culate, it may be investigated experimentally by comparing data obtained with
progressively smaller-diameter probes.
A more tractable approach to the problem of obtaining a quantitative valida-
tion of probe technique is by detailed comparison of experimentally measured
and theoretically computed surface distributions. For example, resonator
geometries possessing high symmetry generally have a calculable field distribu-
7202 Microstrip field diagnostics Microstrip field diagnostics 1203

tion and mode structure. The book by Bahl and Bhartia [9] contains many The boundary condition at the magnetic wall, [aEpe]',, requires that the azi-
references to resonant discs, rings and triangular patches, all of which possess muthal component of the magnetic field should vanish at the boundary. The
high symmetry and are suitable for comparative studies. roots of the equation J:,(ka) = 0 determine the TM,,,,, resonant modes of the

Fig. 21.7 Travelling-wave phase error against frequency (E, = 1.7)

21.5 Resonant microstrip discs

The four disc resonators shown in Fig. 21.8 were constructed on a proprietary
substrate of relative permittivity 2.2. Each disc was excited from a 140Q
microstrip line which was tapered up so as to match a 50Q ridged waveguide
transformer. Provided that the substrate thickness is much less than the guide
wavelength, the system can be treated as a T M cavity bounded by perfect
magnetic walls. The solution to the wave equation in cylindrical co-ordinates
results in the following equation for the vertical component of electric field E2
inside the cavity:

where J,(ke) is the Bessel function of order n and k = wJpo~,. The components
of the magnetic field are given by
Fig. 21.8 Microstrip disc resonators

cavity. The integer n represents the order of the Bessel function describing the
electric field, and physically corresponds to the number of half wavelength
changes around the edge of the disc. The integer rn represents the m th zero of
1204 Microstrip field diagnostics
Microstrip field diagnostics 1205
J,(ka) and corresponds to the number of minima in the range 0 < Q < a. The
resonant frequency of the nm th mode is given by:

where c is the velocity of light, K,,, = m th zero of the first derivative of the
Bessel function of order n. An expression for the effective radius a, of the disc
is given in Reference 9. Eqn. 21.14 was plotted as a function of Q in order to
determine the physical radius necessary to tune each mode to a centre frequency
of 35 GHz. The comparative theoretical and experimental plots for the T M , ,,
TM,, and TM,, modes given in Figs. 21.9-21.11 generally show a striking
resemblance to each other except for obvious differences at the feed point.
It is evident from eqn. 21.11 that the cavity modes have the same intensity and
are spaced in azimuth by (360/2n) degrees.
Table 21.1 lists the experimentally determined excitations and angular loca-
tion of the maxima for each of the three modes investigated.
The 'corrugated' edges seen in the theoretical plots are due to the finite size
of the sampling grid used for the computations. Since the theoretical model
assumes 'perfect' magnetic walls, there are no fringe fields leaking out of the
sides of the cavity, which explains why the surrounding ground plane is com-
pletely free of detail. The model is incapable of giving the surface field just above
the top conductor, but since the two surfaces share the same boundary, it is
assumed that the two distributions are identical to within a constant scale factor.
In reality, the side walls are imperfect and the field lines leaving the upper
surface of the disc must eventually return to the ground plane. Phase plots have
shown that a local ground-plane feature is n out of phase with the corresponding
feature on the upper surface of the disc as expected. The lack of nodal symmetry
seen in Fig. 21.1 1, arises where a particular voltage node in the upper surfack
is locally diminished and the corresponding feature on the ground plane is
enhanced. This effect is commonly observed, and will be referred to again in the
context of microstrip lines. An enhanced ground-plane feature implies either an
enhanced leakage of electric-flux lines out of the cavity, or for some reason the
distribution within the cavity is not exactly 'mirrored' on the upper surface of
the disc. It is encouraging to note from the data given in Table 21.2 that there
is very good agreement between the return-loss figures deduced from slotted-line
measurements made in the waveguide feed and those deduced from the VSWR
in the microstrip line. The VSWR on the line was deduced from a computer-
generated axial section taken through a region of the map corresponding to the
parallel-sided section of the line which exists between the stepped ridge transfor-
mer and the microstrip taper leading to the disc feed point. The good agreement
between the two sets of data suggests that the transformer was well matched at
the frequency of operation and that the scanning-network probe yields quan-
titatively accurate data.
7206 Microstrip field diagnostics Microstrip field diagnostics 1207
7208 Microstrip field diagnostics Microstrip field diagnostics 7209
21.6 Resonant microstrip triangles

Equilateral triangular patch antennas also possess high symmetry and support
a readily calculable TM-mode structure which is discussed in a paper by Hels-
zajn and James [lo]. The electric-field distribution within the microstrip cavity
is given by eqns. 21.15 and 21.16 for a triangle centered at the origin with one
of its sides normal to the x-axis: -

where

+ cos [(g+ q) [26;~ r n ] cos '1 y ]

+ cos [(z q)+ rz] cos [ 2 x ( 1 ~m, y ] (2 1.16)

Table 21.2 Comparison of scalar network analyser and scanning-network

probe return-loss measurements for the microstrip disk resona-
tnrs

Mode Waveeuide return loss (dB) SNP return loss (dB)

A,,, is the amplitude of the mode described by the integers m, n, I which must
satisfy the conditions rn + +
n 1 = 0. Clearly, the integers cannot all be simul-
taneously zero. The patch geometry and resonant frequencies are determined by

Patches were constructed on low-permittivity substrates (e, = 2.2) with the edge
dimension scaled so that each mode was nominally resonant at a frequency of
35 GHz. The E: distribution within the microstrip cavity and the experimental
distribution for the upper surface were determined for the TM,,, TM,,, TM,,,
TM,, and TM,, modes. The input impedance of the corner feed point generally
increased with increasing complexity of the mode structure. The TM,, mode
achieved the best match to the tapered 140R feed line and exhibited a VSWR
of approximately 1.4. In order to investigate the mode intensity distribution, the
signal levels were determined from the area scan data at six equi-spaced test
Table 21.3 Test-point (TP) data for triangular patches
2:
Mode TP I TP2 TP3 TP4 TP5 TP6 m,
Q
TM,, Theory 1.O 0.0 0.25 0.0 0-25 0.0 $
Expt. 1.0 + 0.2 0.0 +0.02 0.27 + 0.03 0.0 f 0.02 0.34 +_ 0.05 0.0 $_ 0.02 2
TM,, Theory 1.O 0.25 1.O 0.25 1 .O 0.25 2
Expt. 1.0 $ 1 +
0.25 0.03 0.7 + 0.14 0.1 + 0.05 1.0 + 0.1 0.25 $_ 0.03 5
TM,, Theory 1.O 0.25 0.25 1.O 0.25 0.25
Expt. 1.0*0.1 0.14&0.02 0.14+0.02 1.1 f 0.1 0.4 + 0.05 +
0.31 0.04
TM,, Theory 1.O 0.0 0.25 0.0 0.25 0.0
Expt. 1.0 + 0.1 0.0 + 0.02 +
0.69 0.07 0.19 + 0.03 +
1.1 0.1 0.0 $_ 0.02

TM,, Theory 1.O 1.O 1.O 1.O 1.O 1.O

Expt. 1.0 f 0.1 0.37 + 0.04 1.1 f 0.01 0.48 + 0.05 0.79 & 0.07 0.48 + 0.05

scale
X,Y =2.Ommldivision Z~ZdBlievel

Fig. 21.128 Theoretical (E,I2 for TM,, triangular patch mode Fig. 21.1 2b Experimental IEJ2 for TM,, triangular patch mode
72 72 Microstrip field diagnostics Microstrip field diagnostics 72 7 3

points. The measured intensities were normalised with respect to the feed-point the outer edges of the line is less pronounced in narrow lines and is not observed
intensity and then compared with the theoretical values calculated from eqn. experimentally because of the overriding influence of the reduction of the
21.15. The results of this investigation are summarised in Table 21.3 for the five probe-coupling capacitance near an edge discontinuity. However, the depress-
different modes. ion of the axial current density by the skin effect is more pronounced in the wider
The theoretical and experimental distributions for the TM,, mode are given lines and is clearly resolved by the probe, as shown in Fig. 21.14. Furthermore,
in Fig. 21.12. The TM,, mode has particularly high symmetry owing to the
equality of the indices n, rn and also the equality of the intensities at the six test
points. However, in practice there is considerable variation between the test-
point intensities, and some asymmetry is observed between features on opposite
sides of the feed-point axis. It is of interest to note that, among the five modes
investigated, the asymmetry maxima, for example, d o not always occur on the
same side of the feed-point axis. Furthermore, there are neither observable
blemishes in the etch-back process nor irregularities in the flatness of the
substrates. Owing to the rather long recording time necessary for some of the
plots, frequency drift could possibly account for some of the asymmetries that
have been observed.
As a result of the foregoing studies, the potential use of the network probe for
mode analysis and in situ circuit diagnostics has been demonstrated. The various
test pieces have also shown beyond any reasonable doubt that the distributions
observed on the upper surface of the metallisation closely resemble those inside
the microstrip cavity.

21.7 Open-circuited microstrip lines

In order to investigate the mode structure supported by wide microstrip lines

and the frequency dependence of the open-circuit terminations, a series of lines
were constructed on 0.254mm-thick PTFE glass-fibre-reinforced substrates
with a relative permittivity E, = 2.2. The lines were linearly tapered over a
distance of 10 mm to a feedpoint impedance of 50 R. Stepped ridge transformers
were used to couple the lines to waveguide WG22, and the overall line length
of 60 mm was accommodated on a substrate of dimensions 40mm x 80 mm.
The area scans for line impedances of 50 and 22 R given in Figs. 21.13 and normal to str~p-lmelrnm
21.14 have been optimised to show the overall field distribution, which inevit- Fig. 21.13 50n open-circuited microstripline at 36GHz
ably results in an apparent loss of detail on the conducting strip. In order to
obtain the VSWR pattern, the contours can be sectioned along the symmetry
axis of the microstrip line. If the area scan is not required, it is only necessary there are complicated periodic longitudinal shifts between the VSWR maxima
to record a single axial scan in order to obtain the VSWR. on the strip and those on the ground plane. These shifts are particularly large
Although the practising engineer will be pleased to learn that high-impedance close to the open-circuit termination. Another striking observation is the lack
lines possess relatively simple field distributions, the wider lines exhibit features of radial symmetry of the features over the ground plane owing to a pronounced
which are difficult to explain mathematically. In spite of the overall complexity curvature of the contours, which periodically alternates from concave to convex
of the mode structure, the distributions have a number of features in common. along the line. This effect is associated with the waxing and waning of the
For example, both the sectioned axial plots (not shown) and the area scans intensity maxima along the microstrip conductor. It will also be observed that
exhibit a long-wavelength periodicity of the VSWR along the length of the line. a strong feature over the microstrip is generally correlated with a weak feature
The tendency for the lateral skin effect to concentrate the current density along over the ground plane. It should be noted that the VSWR maxima and minima
72 7 4 Microstrip field diagnostics
observed along a section through the symmetry axis are equi-spaced for suf-
ficiently narrow lines with characteristic impedances greater than about 4 0 0 .
This observation demonstrates that the probe susceptance causes a negligible
distortion of the measured data in these lines.
normal to str~p-llnelmm
Fig. 21.14 22R open-circuitid microstripline at 35GHz
21.8 Antenna diagnostics
One of the most important applications of the scanning-network probe techni-
que is to the analysis of the mode structure supported by planar microstrip
antennas. Antennas are necessarily open structures and are therefore generally
accessible to probing. Knowledge of the surface charge-density distribution,
both in amplitude and in phase, reveals a considerable amount of information
about the mode of operation of the antenna, which cannot in the case of
multi-element arrays be deduced simply from S-parameter measurements taken
Microstrip field diagnostics 72 75
at the input port. However, when the network probe area scan data is correlated
with measurements of the return loss, the E-plane, H-plane and cross-polar
radiation patterns, a very detailed understanding of the performance of the
system is obtained. Area scan analysis is particularly useful for studying the
mutual coupling between parasitically coupled array elements, the voltage
standing wave ratio on the feed lines, and the relative excitation of directly
coupled elements in an array.
scale
X,Y =l.Ornmldivision Z=l.OdB/level
Fig. 21.1 5 Rectangular patch antenna excited at resonance, F = 3 0 5 G H z
21.8.1 Rectangular patch
A return-loss sweep for the isolated patch antenna depicted in Figs. 21.15 and
21.16 showed that the principal resonance at 30.6GHz was associated with a
return loss of 18dB. The bandwidth for a lOdB return loss was k415MHz.
The fringing effects associated with microstrip open circuits are usually
considered to be limited to a few substrate thicknesses distant from the edge; yet
here there is evidence for an effect extending over more than 10 substrate
thicknesses beyond the edge discontinuity. Another consistent feature of these
12 16 Microstrip field diagnostics
plots is the appearance of a weak 'mushroom shaped' maximum which is
located over the ground plane almost a quarter wavelength away from the
discontinuity. This feature is particularly large at resonance. As the frequency
is raised, the VSWR maxima move along the microstrip line towards the
antenna feed point owing to overall shrinkage of the mode pattern supported
by the patch. The shift in the positions of the VSWR maxima and minima along
the feed line is particularly rapid close to the resonant frequency of the patch.
The variation of guide wavelength with increasing frequency can be accurately
measured by the probe technique, and generally is in excellent agreement with
the theoretically predicted values. Movement of the first feed-point maximum
with respect of the edge of the patch may also be used to investigate the change
of input impedance of the patch with increasing frequency.
scale
X,Y =l.Ornm/divislon Z = l OdB/level
Fig. 21.16 Rectangular patch antenna just above resonance F = 31.2GHz
As the frequency of excitation is raised above resonance there is evidence for
the excitation of the cross-polar TM,, mode, as shown in Fig. 21.16. Further
'mushrooms' appear on the ground plane and are clearly associated with the
Microstrip field diagnostics 7217
lateral splitting of the patch mode. At somewhat higher frequencies than shown
here, all four corners of the patch become relatively strongly excited.
The network probe has also been of immense value in clarifying the mode
structure supported by simple self-oscillating microstrip patch antenna systems
which incorporate a Gunn or Impatt diode. Space does not permit inclusion of
this material here, but further information can be found in References 1 1 and
12.
scale
Z. Zd Bllevel
X,Y=Z.Ommld~v~s~on
Fig. 21.17 I E , I ~distribution for linear four-element array, F = 3 3 . 8 ~ ~ 1 '
21.9 Linear four-element patch array
A striking example of the application of the technique to arrays is given in Fig.
21.17. The four-element array was constructed on a low-permittivity substrate
(E, = 2.2) and has return-loss dips at 33.8 and 37.8 GHz. The main advantage
of the T-junction splitter is its wide bandwidth and simple design, but it does
suffer from the disadvantage of being non-isolating. Since the patch feed-point
discontinuity produces a substantial reflection, each element in the array directly
interacts with every other element owing to the back-reflected waves. However,
owing to the high overall symmetry of the array and also a measure of good
fortune, each element receives approximately the same excitation in this design.
In order to avoid the problem of etching very narrow 200R lines, the output
arms of the primary splitter were tapered to 50R and then split again into two
 12 18 Microstrip field diagnostics Microstrip field diagnostics 12 19
IOOR outputs. Contrary to the normal practice of avoiding the use of sharp
microstrip bends, the corporate feed was laid out using parallel or perpendicular
sections of line in order that the position of the VSWR maxima and minima
could be determined more precisely for analytical purposes.
The array elements exhibit a mode structure similar to that found with the
isolated element discussed in the previous Section. One striking observation is
that the locus of the current maxima along the width of the array is curved in
such a way that the two outer elements appear to be excited below resonance,
whereas the two inner elements appear to be excited above resonance. The
observed curvature of the phase front suggests that there might be significant
mutual-coupling effects between the elements.

scale
X.Y = Z.Ornrn/divlsion Z=66.66pV/level

Fig. 21.1 8 IEzIZdistribution for four-element circularly polarised array, F = 34.9GHz

21.10 Circularly polarised patch antennas

Two possible methods for generating circularly polarised radiation are shown
in Figs. 21.18 and 21.19. It was not immediately apparent, at the outset, what
problems might arise with these designs, and they were originally constructed
with a view to using them as antenna test pieces purely for demonstrating the
two-dimensional field-plotting capability. However, it was clear from Reference
13 that the circularly polarised pentagonal patch mode had been successfully
employed by Weinschel in a practical antenna array operating in the UHF
frequency band.
 7220 Microstrip field diagnostics
In the case of the four-element array shown in Fig. 21.18 the corporate feed
network generates a progressive phase lag of n/2 between adjacent elements,
starting with the top right-hand patch, in a clockwise direction~aroundthe array.
The primary power splitter contains an additional 112 section in the left-hand
output arm so that each of the secondary power splitters is excited with a relative
phase of n. The secondary power splitters are of identical design and incorporate
an additional 114 section in one output arm in order to generate the required
progressive lag. This arrangement by no means represents an optimum solution
for the generation of circularly polarised radiation owing to the impedance-
transformation properties of the 114 sections of line. Nevertheless, it does
provide an opportunity for examining an array in which the mutual coupling
between adjacent elements should be quite small, and also for studying the
behaviour of a T-junction splitter whose output arms are asymmetrically
loaded.
The return-loss sweep exhibits four strong resonances at 28.9, 31.8, 34.9 and
39.8 GHz, respectively. The principal resonance occurs at 39.8 GHz and minor
resonances also appear at 30.9 and 37.4 GHz. The area scan shown in Fig. 21.18
measured at 34.9GHz corresponds to the only frequency at which any of the
patches were strongly excited. This plot was thresholded lOdB above the noise
floor of the detector in order to enhance the visibility of the regions of high
excitation. It can be seen that only the top right-hand and bottom left-hand
patches are strongly excited and also that both appear to be operating slightly
below their natural resonance frequencies owing to the shift of the voltage null
on the patch towards the feed point. From the slight asymmetry in the position
of the VSWR maxima along the vertical arms of the primary splitter, it was
found that the relative phase was close to 0,9n, and therefore 10% in error. Since
the patches are grossly mismatched to the feed lines, the return-loss dips are
probably caused by feeder resonance. This view is confirmed by Fig. 21.18,
where it can be seen that the route between the two more strongly excited
patches contains an integer number of voltage maxima and forces the primary
T-junction to become a voltage minimum. The area scan corresponding to the
resonance at 28.9GHz also showed that the route between the top left-hand
patch and bottom right-hand patch was resonant at this frequency, whilst the
primary T-junction became a voltage maximum in this case. The two patches are
more weakly excited at 29.8 GHz because the natural patch resonance is much
closer to the original frequency of 34.9 GHz than 28.9 GHz.
The area scans (not shown) measured at 31.8 and 39.8 GHz suggest that line
resonance can also occur between the pair of patches coupled by either the
left-hand or the right-hand secondary splitters. Since the effective length of a
patch is 112 at the natural resonance which occurs at approxmately 35.8 GHz,
the patch element is physically shorter than 112 at 31.8 GHz and longer than 112
at 39.8 GHz. At frequencies greater or less than the half-width of the funda-
mental resonance, the edge discontinuities are no longer strongly coupled and
can participate independently of one another in generating two additional
Microstrip field diagnostics 122 1
resonances in conjunction with some other discontinuity located elsewhere in
the network.
In the present example, the resonance at 31.8 GHz is associated with the
relatively strong excitations observed at the feed-point edge of the lower left-
hand patch and also the more distant edge of the upper left-hand patch. The
principal return-loss dip at 39.8 GHz, however, is associated with the relatively
high cross-polar T M , ,-mode excitations observed at the feed-point edge of the
lower left-hand patch and the upper right-hand patch at this frequency. Further-
more, re-examination of Fig. 21.18 suggests that the return-loss dip observed at
34.9 GHz is also closely associated with the resonant length of line connecting
the more distant edges of the lower left-hand patch and the upper right-hand
patch as well as the proximity of the natural patch resonances to this frequency.
Owing to the complexity of the observed resonance phenomena it would be
quite difficult to design an array which exhibited only a single return-loss dip.
Nevertheless, the physical insight gained by studying the mode patterns ob-
tained at the various resonances of the structure does clearly identify which line
lengths might profitably be adjusted in order to reduce the total number of
resonant frequencies present in the system.
The basic reason for the failure of the array geometry is due to the mismatch
at the feed point of the patch elements, which leads to an imbalance of power
division at the secondary splitters. The effect could, in principle, be compensated
for by making the secondary feed line an integral number of half wavelengths
long at the natural resonance frequency of the patch, and displacing the T-
junction one-eighth of a wavelength from the mid-point between the patches.
Under these conditions the two branch lines form a parallel resonance circuit
which should result in an equal division of power.
Having established the probable cause of failure, the measurement procedure
is repeated on the revised array geometry in order to confirm the original
diagnosis. The entire process of empirical optimisation is repeated until the
required performance has been achieved. In the hands of a skilled microwave
engineer, a single one-dimensional scan along the axes of the feed lines could
provide sufficient information for the optimisation of an array once the problem
areas have been physically identified by a two-dimensional scan of the complete
antenna.
The pentagonal patch was designed by scaling the dimensions of the circularly
polarised VHF antenna discussed in Reference 13 to a nominal frequency of
35 GHz. The return-loss sweep within the frequency range 26-40 GHz indicated
the presence of three resonances, one of which was very weak. The strong
resonances at 31.35 and 29.5 GHz exhibited return-loss dips of 3OdB abd 14dB,
respectively. Area scans for the I E , ~ ' and relative phase distributions corres-
ponding to these frequencies are shown in Figs. 21.19 and 21.20, where it may
also be seen that the vertical microstrip feed line is offset a short distance from
the apex of the pentagon. Examination of Fig. 21.19a indicates the presence of
three voltage maxima of equal intensity which are equidistant from a deep
 1222 Microstrip field diagnostics Microstrip field diagnostics 1223

minimum located at the centre of the patch. The large voltage gradients that
exist in directions perpendicular to the edges of the patch suggest that this mode
of excitation is not linearly polarised.
The phase plot given in Fig. 21.19b indicates that the position of the voltage
minimum also corresponds to the 'phase centre' of the antenna, which is the
point from which all phase contours appear to diverge. In this particular
example the phase levels are at 30" intervals and the software operates in the
range from - 180" through zero to + 180'. The 180" phase boundary is
represented by the broad contour that spirals out from the phase centre at the
top of the Figure, and the small zigzag irregularities on it have no physical
significance. It can be seen that the narrow basal edge of the pentagon has an
almost constant relative phase of - 180" with respect to the feed point, which
was arbitrarily set to zero phase. The intensity contour plot indicates the
presence of three voltage maxima located at the vertices of an equilateral
triangle inscribed within the pentagonal patch, and the fringe fields in the
immediate vicinity of these maxima suffer a progressive phase shift of approxim-
ately 120". The phase gradient is particularly large along the two side edges of
the pentagon. Viewed along the direction of radiation propagation, the left-
hand rotational sense of the contours clearly indicates that this mode of re-
sonance does, in fact, correspond to the left-hand circularly polarised mode
discussed in Reference 13.
The phase plot given in Fig. 21.20 for the resonance at 29.5 GHz shows that
the basal edge has a constant phase of - 180, whilst the two edges meeting at
the apex now have a constant zero phase. The maximum phase gradient occurs
approximately midway between the apex and base in a direction parallel to the
axis of symmetry of the pentagon. This mode of resonance corresponds very
closely to a simple linearly polarised rectangular patch mode.

21.11 Microstrip travelling-wave antenna

The final example of antenna analysis using the scanning-network probe is to

a rampart-line array similar in type to that discussed in References 14 and 15.
The rampart line is an interesting slow-wave structure because it radiates in the
backward direction. The particular design shown here consists of ten meander
sections which radiate a principal lobe at an angle of about 140" to the plane of
the substrate. The width of the transmission line was increased towards the
centre of the array so as to achieve a tapered aperture distribution as suggested
in Reference 14. The principal return-loss dip for this antenna occurred at a
frequency of 16.7 GHz and the area scan recorded at this frequency is shown in
Fig. 21.21. Inspection of Fig. 21.21 shows that half of the structure receives little
or no excitation, whilst the other half exhibits a significant standing-wave
pattern. Theoretically, of course, this array should support a smoothly tapered
voltage distribution which is larger at the two ends owing to the reduced width
 1224 Microstrip field diagnostics
of the transmission line. A colour-coded version of the same area scan also
clearly shows an enhanced excitation at alternate corners of the rampart line.
This rather unexpected result may be due in some way to the existence of mutual
coupling between the relatively closely spaced parallel-line sections. The ram-
part-line antenna gives a very clear indication of the value of the area-scan
technique in microstrip antenna diagnostics. A single two-dimensional scan has
provided, in this instance, a considerable degree of physical insight into the
problem areas of the design geometry. Comparable information could not
possibly have been obtained from network-analyser measurements.
Fig. 21.21 IEZl2distribution for rampart-line antenna at 16.7GHz
21.12 Acknowledgments
The author wishes to thank the SERC for the provision of a research grant
which made this work possible. The Marconi Company is also gratefully ack-
nowledged for supporting a CASE studentship award during the early stages of
the programme. Finally, the author is much indebted to his research students,
Microstrip field diagnostics 1225
Dr. J. Whitehurst, Mr. A. Leggetter, and Mr. N. Piercy, for many hours of
painstaking effort, and also to Mr. J. Taylor and Mr. L. Ellison for constructing
the probe-transport mechanism.
21.13 References
1 DAHELE, J. S., and CULLEN, A. L.: 'Electric probe measurements on microstrip', IEEE
Trans., 1980, MTT-28, p. 752
2 LADBROOKE, P. H.: 'A novel standing wave indicator in microstrip, Radio Electron. Engin.,
1974,44, p. 273
3 NEUMANN, E. G.: 'Radiation from the free end of a dielectric rod transmission line', Z.
Angew Physik., 1967, 24, p. 1
4 NEUMANN, E. G.: 'The electric field near a curved dielectric transmission line', NTZ, 1969,
3, p. 161
5 GINZTON, E. L.: 'Microwave measurements' (McGraw-Hill, 1957), pp. 249-271
6 MONTGOMERY, C. G.: 'Technique of microwave measurements' (McGraw-Hill, 1947) p.
485
7 SCHWARZ, S. E., and TURNER, C. W.: 'Measurement techniques for planar high frequer,cy
circuits', IEEE Trans., 1986, MTT-34, pp. 463-467
8 HARRINGTON, R. F.: 'Time harmonic electromagnetic fields' (McGraw-Hill, 1961)
9 BAHL, I. J., and BHARTIA, P.: 'Microstrip antennas' (Artech House, 1980)
10 HELSZAJN, J., and JAMES, D. S.: 'Planar triangular resonators with magnetic walls', IEEE
Trans., 1978, MTT-26, pp. 95-100
11 FRAYNE, P. G., and RIDDAWAY, C. J.: 'Resonance in self-oscillating antennas', Electron.
Lett., 1986, 22, pp. 1269-1270
12 FRAYNE, P. G., and RIDDAWAY, C. J.: 'Resonance in an active millimetric conformal
array antenna with quasi-optical feedback'. 5th Int. Conf. on Antennas and Propagation,
York, ICAP 1987, p. 177
13 WEINSCHEL, H. D., and CARVER, K. R.: 'A medium gain circularly polarised microstrip
UHF antenna for marine DCP communication to the GOES satellite system'. IEEE AP-S
Symp. Digest., 1976, p. 391
14 HALL, P. S., WOOD, C., and JAMES, J. R.: 'Recent examples of conformal microstrip
antenna arrays for aerospace applications'. 2nd Int. Conference on Antennas and Propaga-
tion, York, ICAP 1984, p. 397
15 HALL, P. S.: 'Microstrip linear array with polarisation control', IEE Proc., 1983,13OH, p. 215
 Chapter 22

Microstrip antennas on a
cylindrical surface
E. V. Sohtell

22.1 Introduction

Owing to their ability to conform to the underlying structure, microstrip anten-

nas have a variety of applications to objects with a curved surface. The utiliza-
tion can be, for example, on aircraft, missiles, ships, satellites etc. In many cases,
where the radius of curvature is large, a planar theoretical approach is sufficient.
However, when the radius of curvature is small, the curvature of the surface
cannot be neglected. The purpose of this Chapter is to describe how theoretical
design models, previously developed for planar structures, are extended to the
cylindrical case, and to verify the theory with experimental results.
A theoretical treatment of a microstrip patch on an infinitely long circular
cylinder is presented in Section 22.2. The theory is used in the analysis of a single
patch in Section 22.3. Measured results are shown for comparison. Section 22.4
describes the design of a complete phased array consisting of 32 patches. Input
impedance, radiation patterns and mutual-coupling coefficients are displayed.
The feed network for the array was designed and built by Dr. J.P. Starski,
Division of Network Theory, Chalmers Univ. of Technology, Gothenburg.

22.2 Theoretical models for a patch on a cylinder

This Section desciibes two theoretical models of the rectangular patch. They are
useful both in calculating the input impedance and in finding the radiation
pattern from the antenna. The description will concentrate on the radiation
pattern from a cylindrical structure. The input impedance can in most cases be
found very successfully by applying a planar theoretical approach [I]. An
investigation of the influence of the curvature on the resonant frequency and the
1228 Microstrip antennas on a cylindrical surface Microstrip antennas on a cylindrical surface
input inpedance was made by Luk et al. [2]. It was found that both the resistance where
and the susceptance of circumferentially polarised square patches vary with the
cylinder radius. For axially polarised patches, on the other hand, only the
resistance varies.
In the cavity model, the patch is considered as a cavity bounded by two
electric and four magnetic walls [3-51. From the modes set up in the cavity by
the feed, the field distribution in the four magnetic wall are deduced. These field
distributions are then used as sources for the radiation from the patch. The
cavity model is developed for a rectangular patch on a cylinder in Section 22.2.1.
*,(4, z) = aces (2) (y) COS

The effect of the substrate surrounding the patch can be taken care of by
solving the entire boundary-value problem, in which case the patch may be
represented by a surface current on the substrate. When the appropriate integral
equation is solved, with the feed probe or a microstripline as the source, the
radiation patterns, as well as the input impedance, can be found [6]. Ashkenazy
et al. used assumed surface currents in solving the boundary-value problem on
a cylinder [7]. The latter description is what will be called the surface current
model of the patch. This model will be discussed in Section 22.2.2.
The two approaches described here are suitable for cylinders with a radius of
up to 4 or 5 wavelengths. For larger cylinders the numerical evaluation becomes
very time consuming and ray-tracing techniques are preferable [8].

22.2.1 Cavity model of the patch

The rectangular patch is modelled by two axial and two circumferential slots.
The field distribution in the slots may be found by solving the boundary-value
problem inside the cavity, with the feed as the source. The solution is described
by a number of modes, which then gives rise to a set of field distributions in the
side walls. These distributions are used as sources for the radiation from the
patch. Special cases of radiation-pattern calculations are described in Referen-
ces 9 and 10.
2 C,
Internal fields
The rectangular patch described by Fig. 22.1 is considered. The antenna is fed Fig. 22.1 Axially polarised patch on a circular cylinder
via a probe, which sets up a field underneath the patch. When the substrate is
thin, we can assume that the E-field underneath the patch has only a p-com-

(2)(e)
and
ponent. Following the procedure for the planar case [3, 41, we derive the
following expressions for the E- and H-fields: G~~ = sinc sinc

p and q are the modal numbers: 0, 1, 2, - - - ;ko = 2n/Lo; E, is the relative

dielectric constant of the substrate; +/, zfis the feed location; and A,, A, are the
probe extensions in 4 and z for a rectangular probe (planar approximation).
6, = d/2R, z, d and R are given by Fig. 22.1. The effective loss tangent defl is
incorporated to take care of all the losses in the cavity. Radiation losses and
losses due to finite conductivity of the conductor, as well as losses in the
7230 Microstrip antennas on a cylindrical surface Microstrip antennas on a cylindrical surface 7231

substrate, can be estimated using the procedure described for the planar case in
Reference 11. Another way of describing the losses in the cavity is to use an
impedance-boundary condition at the surrounding wall [4].
where q is the free-space impedance.
External fields An appropriate Fourier transform is defined [I21 and the Fourier-transformed
Next, we want to find the radiation from the lossy cavity. The E-field in the aperture fields are inserted into eqns. 2 2 . 8 ~and b. The radiated far field from
cavity walls can be replaced by equivalent magnetic currents M, and MZ. When each mode in the axial slots will now be given by
the substrate is thin compared to the wavelength, the magnetic currents are
exp (- jkor) kocos O[exp (jkoz, cos 0) cos (qn) - 11
narrow and can be approximated by collapsed currents on the conducting E r e = 5 ZnZrR ( q n / ~ , ) ~- cos2e
cylinder. We may now replace the magnetic currents by flush-mounted slots in
* 4011
the cylinder. We thus end up with two axial and two circumferential slots in the
ground plane. The field distribution along the four slots are considered to follow
the distribution in the cavity (eqn. 22.1) [12].
e-,kocos8~/2

z0
" &!J"cOs[n(4
Hi2y(koR sin 9)
The theoretical calculations carried out in this Chapter are all based on the
assumption that the cylinder extends to infinity in both axial directions. A
two-dimensional Fourier transform can therefore be applied in solving the
boundary-value problem. In doing this we will have to make a suitable expan- where the upper and the lower terms within the bracket are used for slots
sion of the field outside the cylinder and match to the known aperture distribu- +
positioned at 4 = - 4, and 4 = 4,, respectively, and - z,/2 < z < zJ2.
tion of a slot in the cylinder.
The field from the aperture of a slot can be obtained from two orthogonal
components of the vector potentials [13]. These components can be chosen as,
for example, A, and F,. A denotes the magnetic and F the electric vector The circumferential slots give rise to both components of the E-field. For each
potential. The expressions for the axial components of the vector potentials are P, we get
found by expanding both the field in the slot and the radiated fields outside the
cylinder in cylindrical modes. The radiation condition at infinity indicates that
Hankel functions of the second kind are to be used for an ep' time dependence.
The tangential E-fields are matched in the slot and are set to zero on the rest of
the cylinder. The following expressions are obtained for A, = - Azsin 0 and
F, = - F. sin 0 in the far field when an asymptotic formula for the near-field/
far-field transformation is used [13]:

A, = exp(-jk0r) f ejnbjn 4 ( n . k0cose)

koV "=-a HA2'(koR sin 0) sin 0
cos enEi(n, kocos 8)
F, = exp ( -jko r)
konr ,,-, k,R sin20HA2y(koR sin 0)

- E$(n, k, cos 9)
w y ( k oR sin e) I
where g:(n, k:) and E$(n, kz) are the Fourier transforms of the slot fields Ei(q5)
and E$(z). HA2)(z)is the Hankel function of the second kind and nth order, and
HA2"(z) is the derivative with respect to the argument of the Hankel function. where the upper and the lower terms within the bracket are used for slots
The radiated electric and magnetic fields are then found from the far-field positioned at z = + zd/2 and z = - zJ2, respectively, and - 4, < 4 < &.
approximations The origin in 4 and z is located in the centre of the patch in the expressions
 7232 Microstrip antennas on a cylindrical surface
given by eqns. 22.10 and 22.1 la, b. We add the contriubtions from the four slots
and make a summation over all the cavity modes p, q to find the total field.
All the above expressions contain an infinite summation, which is a sum of
cylindrical modes in which the fields have been expanded. The number of
cylindrical modes necessary for the summation to converge is dependent on the
cylinder radius, the 0 angle and the 4 angle. In the circumferential slots there
is also a dependence on the excitation mode. The number of terms is usually less
than or around 2 kR.
The coding of the formulas is fairly straightforward. However, at certain
angles and for n = 0 care must be taken. In eqn. 22.10 special attention must
be given for 0 angles where cos 0 = q1/2zm,since the denominator will be zero.
The expression will have a limiting value, since the numerator also has a null for
the same angle 0. It is found that
! Microstrip antennas on a cylindrical surface 7233
the interior fields in the cavity. This is a very crude approximation when the
metal thickness is much larger than the skin depth of the metal. We only present
an outline of the current method, the details can be found elsewhere [7, 121.
Calculated results with an assumed current distribution will be shown in the
next Section.
The relation between the interior H-field and the surface current will be
where the H-field is given by eqns. 2 2 . 2 ~and b.
I
The current densities on the patch are then represented as follows
J:pq = ) (2)
H4,pq(4,Z) = - I,, sin - cos (22.150)

jk, cos 0[exp (jkoz, cos 0) cos (qn) - 11

( q n / ~ , ) ~- ~ C O S ~ B kOcod8-tqn/z,,,*
z, q = O
zm/2 9 Z 0
J;,pq= - H q ( z) = cos (y) (2) sin (22.156)

If z, is smaller than L/2, which is always the case for half- wavelength resonant
patches, the only zero of the denominator will be for the combination q = 0, b is the cylinder radius including the substrate.
0 = 90". A similar problem is encountered in eqns. 22.1 l a and b when both p The radiated far field from the patch is:
and n are zero. A limiting value exists for these functions also. This is given in
eqn. 22.13:

- jn[cos(pn) exp (-jn24,) -

bpl(240)12 - n2
11
-10
240 n and P = 0
n = o , p + o
(22.13)

Since the cylinder is considered infinite in the axial direction, the formulas are
and
exp(-jk,r) .
not valid for angles coinciding with the cylinder axis. The limit to as how close E4(r74 , 0) = ko nr
sin 0 1 j"ejn4 C,(n, kocos 0) (22.163)
to the axis they can be used is actually set by the accuracy of the computer. A n=-m

small-argument approximation for the Hankel function is needed for very small
0 angles, but the formulas are still valid as long as 0 is not precisely 0' or 180".
Calculated radiation patterns can be found in Sections 22.3 and 22.4, where
comparisons are also made with practical results.

22.2.2 Surface-current model

We consider an infinitely long cylinder coated with a substrate. The patch is
represented by axial and circumferential surface currents that have to be found
I
where
I:[ = M-'[:] = L[
det(M) - M,,
I:[]: :-
M22 (22.17)

by a solution of the integral equation for the problem. The fields in the substrate
and in free space outside, as well as the currents on the patch, are expanded in
cylindrical modes. The proper boundary conditions are satisfied and the result-
ing currents and fields are found.
Alternatively, the cavity modes as deduced in the previous Section may be
used to obtain the tangential H-field on the inside of the metal patch. When the
'metal is considered infinitely thin, the surface currents can be directly related to
J: and J; are the Fourier-transformed z-directed and 4-directed currents. The
ekpressions for the elements in the M-matrix are given by Ashkenazy et al. [7].
The summation of cylindrical modes is similar to the summation for the
expressions in the cavity model. However, the expressions are naturally more
complicated since the substrate effects are included in the model. For every 0
angle, the Hankel function for three different arguments is necessary, from order
zero up to as many modes as are required. It is also desirable to include a
1234 Microstrip antennas on a cylindrical surface
complex dielectric constant to take into account the dielectric losses. The
pattern will otherwise show unnaturally large ripples in a 0 cut. A complex
dielectric constant, however, leads to a complex argument in the Hankel fun-
ctions. Routines with such facilities are unfortunately not available in all stan-
dard sub-routine packages.
22.3 Single-patch application
A study of the radiation performance of a single patch mounted on a circular
dielectric-clad cylinder is described in this Section. Two different frequencies are
applied, 2.615 GHz and 5.7 GHz. The element is oriented for axial as well as for
circumferential polarisation. Radiation patterns in both 0- and &cuts are
studied and measured patterns are compared to theoretically derived curves.
Cross-polarisation levels are also of interest. Additional comparisons can be
found in Reference 12.
1 1235
Microstrip antennas on a cylindrical surface
22.3.3 Radiation-pattern comparisons
The cavity model (eqns. 22.10 and 22.1 la, b) and the surface-current model
(eqns. 22.15-22.17) have been used in the theoretical calculations. Two patch
sizes are examined here: one patch is 35 x 35 mm2and has a resonant frequency
of 2,615GHz, and the other patch is nominally 15.2 x 15.2mm2 with a re-
sonant frequency of 5.7 GHz. In the lower-frequency case, the cylinder has been
covered with substrate all around in order to investigate the pattern behaviour
in a complete q5 cut.
Axially polarised patch at 5.7 GHz
The H-plane radiation patterns for an element oriented for axial polarisation are
shown in Fig. 22.2. The E,-field is the co-polar component in this diagram and

22.3.1 Mechanical design

The cylinder radius is 0.1495m for all experiments described in this Section
(1.3 1 and 2,851 at 2.615GHz and 5 7 G H z , respectively). The length of the
cylinder is 0.635 m. The element is etched on a substrate which is easily curved
about the cylinder. The substrate has a dielectric constant of 2.32 and a thick-
ness of 3.18mm (118"). The loss tangent is given by the manufacturer, Tellite
Co., to be 0.00015. The material is sensitive to heat, and, in applications where
soldering is required, a low-temperature solder is recommended. This was tested
for the soldering of the feed probe in this study. It was found, however, that the
strength of the solder was not adequate for the kind of application where the
antennas were connected and disconnected several times. A careful soldering
with ordinary Pb/Sn solder was therefore made. An alternative feeding of the
patch, e.g. microstrip feeding, could have been applied to avoid the soldering
problem. In that case the radiation from the microstrip feed line has to be taken
into account.

22.3.2 Measurements
An anechoic chamber was used for all far field radiation-pattern measurements.
The measurement room is equipped with standard instrumentation and has a
measurement length of approximately 6 m. In this application, where the radia-
tion pattern is broad, there are problems with reflections from objects nearby.
The edge radition from the substrate caused additional difficulties. This radia-
tion was reduced as much as possible by placing microwave absorbers as caps
on the ends of the cylinder. An evaluation of the measurement room gave a
peak-to-peak variation of the amplitude of around 1.6 dB at a level of - 10 dB,
which indicates that the reflectivity level is about - 32 dB.
angle ?)
Fig. 22.2 H-plane radiation pattern for an axially polarised patch at 5.7GH.z
-measured
0 0 0 0 cavity model ( 5 modes)
- - - surface-current model ( p , q = 0 , 1 and 2 , 0 )
the E+-field is the cross-polar component. The solid lines show the measured co-
and cross-polar levels. The broken lines are the computed fields with the
surface-current model and the circles give the cavity-model prediction. The
cavity model was used with five modes included, of which the p, q = 0, 1 and
 7236 Microstrip antennas on a cylindrical surface
2, 0 modes dominate. The same two modes, with the level of the p, q = 2, 0
mode 40% of the dominant mode level, were also included in the surface-current
model calculations. Both co-polar and cross-polar predictions agree very well
with the measured curves.
In the E-plane, i.e. in a 4 = 0' cut, we do not predict any cross-polarisation.
The results of the measurements and calculations are given in Fig. 22.3. The
measured cross-polar level is very low. The measured co-polar curve shows a
Fig. 22.3 E-plane radiation pattern for an axially polarised patch at 5 . 7 GHz
-measured
0 0 0 0 cavity model (5 modes)
--- surface-current model ( p , q = 0, 1 and 2. 0 )
ripple due to substrate edge radiation. This radiation also prevents the pattern
from dropping down in the 0 = 0" and 0 = 180" directions. The same excita-
tions as in the H-plane were used. Note the behaviour of the cavity-model
prediction close to the cylinder axis. These peaks are caused by the sin 0 factor
in the denominator of eqn. 22.1 l a for the 0:th cylindrical mode.
Circumferentially polarised patch at 5.7 GHz
Fig. 22.4 shows a comparison between the measured and computed curves for
a circumferentially polarised patch in the H-plane (4 = 0'). The E6-com-
ponent, i.e. the co-polar component, agreement is good for both theoretical
Microstrip antennas on a cylindrical surface 7237
models. The computed cross-polar curve derived via the surface-current model
has used the excitation 1.0 and 0.4 for modes p, q = 1 , 0 and 0,2, respectively
(the p, q = 0, 2 mode is the main contributor). The level is correct as it cuts
through the measured edge-radiation ripple. We can observe in this measure-
ment cut that the cross-polar-component ripples much heavier than the co-
polar-component ones. This is due not only to the lower power level, but also
to the stronger E-plane field excitation. (This is the E-plane for the cross-polar
component.) The cavity model has used five modes, but only two give a signifi-
b,
cantcontribution, the q = 1, 0 and 0, 2 modes.
Fig. 22.4 H-plane radiation pattern for a circumferentially polarised patch at 5 7 G H z
-measured
0 0 0 0 cavity model (5 modes)
- - - surface-current model ( p, q = 1 , 0 and 0, 2)
The E-plane cut for the same patch is displayed in Fig. 22.5. Since the cylinder
was not entirely covered with substrate, the measurements are not relevant
outside an angle of approximately 100'. The surface-current model curve has a
small ripple caused by circling waves around the cylinder. The shapes of both
computed curves are symmetrical, whereas the probe excitation makes the
measured curve slightly asymmetrical. The cross-polar level is low, which is
expected.
 7238 Microstrip antennas on a cylindrical surface
Circumferentially and axially polarised patches at 2,615 GHz
Fig. 22.6 shows a complete 4 cut at 0 = 90, i.e. in the E-plane for an element
oriented for circumferential polarisation. It is interesting to note how well both
theoretical models predict the interference from creeping waves circulating
angle t)
Fig. 22.5 E-plane radiation pattern for a circurnferentially polarised patch at 5.7GHz
- measured
0 0 0 0 cavity model (5 modes)
- - - surface-current model ( p, q = 1 , 0 and 0 , 2)
0 Microstrip antennas are very well suited to conformal array applications, and
angle t)
Fig. 22.6 E-plane radiation pattern for a circurnferentially polarised patch at 2.61 GHz
- measured
0 0 0 0 cavity model (6 modes)
- - - surface-current model ( p, q = 1 , 0 and 0 , 2 )
I
Microstrip antennas on a cylindrical surface 7239
around the cylinder. The amplitude of the interference is much higher when the
patch is oriented for circumferential polarisation, compared to what is noted
when the patch is rotated 90" (Fig. 22.7). An explanation for this is that the
patch creates a stronger field along the ground plane in the E-plane compared
to the H-plane. The asymmetry is due to the asymmetric probe location in 4. The
cross-polar level is low, around - 35 dB, in the measured curve. This indicates
that the probe is located exactly centered in the z-direction.
angle t)
Fig. 22.7 H-plane radiation pattern for an axially polarised patch at 261 GHz
-measured
0 0 0 0 cavity model (6 modes)
- - - surface-current model( p , q = 0, 1 )
The co-polar component in the H-plane of an axially polarised patch is shown
in Fig. 22.7. Only the dominant p, q = 0, 1 mode contributes visibly. The
predicted 'back'-lobe, which is caused by the constructive interference from two
waves travelling in opposite directions around the cylinder, only reaches a level
of -47dB, as compared to - 18 dB in the circumferentially polarised patch
case.
22.4 Array application
22.4.1 General
several projects have been reported in the literature [14-251. Sanford [I41 descri-
bes a phase-steered array, while References 15-21 report omnidirectional ap-
plications. A conical beam is produced by the cylindrical array in Reference 22,
and a high-gain spherical array was studied by Dubost and Vinatier [23].
Various microstrip conformal arrays are briefly described by Munson [24] and
a conical circularly polarised array with monopulse was reported by Newham
1240 Microstrip antennas on a cylindrical surface
[25]. Conformal printed dipole arrays have been analysed theoretically in Re-
ferences 26 and 27, and dipoles above a cylinder were investigated in terms of
their active radition pattern in Reference 28.
A bibliography of the recent literature on conformal antennas in general has
been compiled by Hansen [29].
22.4.2 Theoretical treatment ofjinite and injinite arrays
A cylindrical-array antenna may be treated theoretically as an infinite array in
the axial direction and an infinite periodic array in azimuth. As an alternative,
an element-by-element approach may be used. When the number of elements is
large, the infinite model is preferable, since.all the calculations may be perfor-
med by considering a single unit cell [28]. When the array is finite, the elements
close to the edge behave differently from the centre elements, owing to the
difference in mutual coupling. These edge effects must be taken into account in
the design of small and moderately sized arrays, and also when very low
sidelobes are required from a large array. Steyskal used an element-by-element
approach for the analysis of a finite array of circular waveguides on a cylinder
[30]. The same approach was also adopted by Pozar in a study of planar arrays
of microstrip elements [31].
The radiation-pattern calculations of a finite-array antenna, using an el-
ement-by-element approach, involves the modelling of the antenna element and
the incorporation of the mutual coupling. The modelling of the patch antenna
was treated in the previous Section. A theoretical model for the mutual coupling
between microstrip elements on a cylinder is not yet available; so in the array
design described in this Section the measured S-matrix was incorporated to
account for the mutual coupling.
Since the measurement of the entire S-matrix is a very time-consuming,
procedure and the accuracy is limited, an approximation is used. The idea is to
assume that elements located similarly will couple the same way. For instance,.
the two centre elements in one band are assumed to couple the same way as the
two centre elements in another band, which might not be a bad approximation.
However, when this reasoning is extended to be valid for the corner element and
its closest neighbour in the same band, we have reason to believe that it is a poor
approximation. Still, it will be shown that the approximation is acceptable in
this application, where a sidelobe level of - 20 to - 30 dB is investigated.
22.4.3 Design of a phased array on C-band , predicts. The bandwidth is -4.7% for a VSWR of 1.5.
In this Section we will describe the design of a microstrip antenna array on a
cylindrical surface. The array is described in more detail in Reference 12 and the
entire project was briefly reported by Sohtell and Starski [32]. The frequency of
operation was set to 5.7 GHz. Beam steering f 30' was desired in azimuth, but
as an extension of the program the possibility to steer the beam in elevation was
also desirable. Another option was the incorporation of circular polarisation.
The sidelobe level was to be controlled by using variable attenuators. The
Microstrip antennas on a cylindrical surface 1241
number of phase shifters and attenuators was limited to 8 in the first configura-
tion.
As in the design of a planar array, the size of the antenna, the element pattern,
the type of element grid and the element separation will govern the characteris-
tics of the far-field radiation pattern. Since the elements are placed on a curved
surface, it is not possible to separate the element pattern from the array factor,
which complicates the synthesis. In a beam-steering situation it is also essential
to beware of the limits of the active angle of the array. The element pattern, the
element grid and the active angle will be discussed below. The design will focus
on the azimuth plane, since no restrictions were set on the radiation pattern in
the elevation plane.
Antenna element
According to the specifications, circular polarisation was required for a possible
continuation of this project. There are several types of microstrip antennas that
produce circular polarisation [5]. In this project, however, we wanted to be able
to predict the radiation patterns theoretically as accurately as possible. An
' element which could easily be described in a cylindrical geometry was therefore
wanted. In addition, we did not want any radiation from the feed network to
interfere with the desired radiation. Consequently, a square-shaped probe-fed
element was chosen. The dimensions of the element were predicted using stan-
dard methods available. The following simple formula for the resonant fre-
quency of a patch on a curved surface is very useful [2]:
where 2R4, = 2R4, + h/& is the effectivecircumferential length of the patch
and z, = z, + h/& is the effective axial length. The actual dimensions of the
patch are z, and 2R4,. R is the cylinder radius,pq is the mode number, h is the
substrate height and E, the dielectric constant of the substrate. In our case we
found that with E, = 2.32 and h = 3.18 mm, for a TM,, mode at 5.7 GHz, the .
length z, should be 15.18 mm.
The dimensions of the microstrip element designed here are: axial length
+
measured on the etched board is 15.10 0.05mm, and the width is
+
15.20 0.05mm. The measured return loss for this element is given in Fig.
22.8. The resonant frequency is 5.725GHz, which is exactly what eqn. 22.18
Active angle
The active angle of the cylinder can be seen from Fig. 22.9 to be limited by the
beamwidth of the constituting elements. Beam steering to 0' and 30' is illu-
strated. Since the elements that were chosen for this array are a little smaller
than 4 2 , a fairly broad beam was expected. To get an approximate idea of the
beamwidth, a measured radiation pattern for a single element on a cylinder was
7242 Microstrip antennas on a cylindrical surface Microstrip antennas on a cylindrical surface 1243

studied. The element at one end of the array should not be too much out of An active angle smaller than 70" thus seemed necessary if all elements were to
phase when the beam is steered 30' in the other direction. The usable angle of contribute to the main lobe for all steering angles without phase compensation.
the element studied was about 130" for 22' phase error and lOdB power loss.
Element grid
The specifications asked for beam steering _f 30' in elevation and azimuth with
eight phase shifters. Steering of the beam in elevation requires more phase-shif-
ters than beam-steering in azimuth, owing to the non-separable geometry. The
array was therefore arranged to be steered in azimuth, in a first configuration,
because of the limited number of phase shifters. The difference in beam steering
in the two planes is discussed further in Section 22.4.4.
The array was designed to allow for beam steering in both planes at the same
time. A triangular-element lattice was chosen with a total of 32 elements. The
elements were arranged in eight columns of four elements each. The power was
thus first split into eight channels, each channel containing a phase shifter and
an attenuator. The four elements in each column were uniformly fed with equal
phase.
0 I I I I I I I I I 1 The element separation was the next parameter to be found. On a planar
5.2 5.7 6.2
Frequency (GHz) surface, the discussion would proceed as follows. The largest separation between
elements can be found in adiagonal plane. To be able to steer to 30" in this plane
Fig. 22.8 Measured return loss for an isolated probe-fed square patch
without introducing grating lobes, we would need an element separation smaller
than 0.62. We do, however, know that the grating lobes are not as high and
distinct in the azimuth cut, owing to the non-uniform element spacing [29]. The
element separation in a diagonal plane was chosen to be 0.581, which gave a
column/band separation of 0.4141 in q5 and z.
The above considerations led to the choice of a cylinder diameter of ap-
proximately 0.3m. The actual radius of the cylinder used in the following
measurements was 0.1495m (2.842 at 5.7 GHz) excluding the substrate. This
resulted in an active angle of 63" when the substrate thickness was 3.18 mrn. A
photograph of the array is shown in Fig. 22.10.

22.4.4 Measured performance

Fig. 22.9 Active angle in azimuth of a cylindrical array
Input impedance
The cylinder with the array was placed in a small anechoic chamber built up by
four walls covered with absorbers. The roof and the floor were not shielded. The
room was examined for this application by moving the cylinder about in the
room to find the in-and-out-of-phase interference of the reflections. An
HP 8409B semi-automated network analyser was used to perform the measure-
ments. The total error in the return-loss measurements was around f0.4dB and
f 3" at the 25 dB level.
All the neighbouring elements were terminated in matched loads during the
measurements. The loads were all better than 20dB in return loss.
Fig. 22.1 1 shows the return loss measured at all 32 input ports. The design
centre frequency of 5.7 GHz is clearly demonstrated. A bandwidth of 100 MHz
1244 Microstrip antennas on a cylindrical surface
for a VSWR < 1.5 for all the elements was achieved. The variations in return
loss are due to the etching, the probe locations, and the mutual coupling
between the elements. Considering this, the result is very good. A tendency
towards lower return loss was noted for the elements in the centre of the array.
Fig. 22.10 32-element phased-array antenna
This is caused by the mutual coupling between the elements, since the patches
were designed for optimum matching as isolated elements.
Mutual co~~pling hetnwn rlw elements
The same provisional anechoic chamber described previously was also used for
the mutual-coupling measurements. The measurement error for these measure-
+
ments was found to be less than f 0.4dB and 3" at a 20dB level. It was noted
that the surface-wave radiation along the cylinder axis influenced the mutual
Microstrip antennas on a cylindrical surface 7245
coupling heavily. The errors caused by surface-wave reflections at the edges of
the board are not included in the estimations above.
Fig. 22.12 gives the amplitude and phase of the measured mutual coupling at
5.7 GHz from a centre element to all the other elements and Fig. 22.13 shows
the coupling from a corner element. A study of Figs. 22.12 and 22.13 gives the
following information:
I
5.5 5.6 57 58 5.9
frequency (GHz)
Fig. 22.11 Measured return loss for 32 active elements
The coupling upwards from the centre element is stronger than the coupling
downwards, except for the closest elements. The reason for this can probably be
found in the second-order mutual-coupling effects, which are different owing to
the asymmetric location of the element in the array.
The E-plane coupling is stronger than the H-plane coupling for both ele-
ments, except for the centre element's coupling to the closest elements. The
difference is very pronounced for the corner element. It is striking to see the
difference in phase between the two planes. There are quite clearly two types of
coupling involved. The surface-wave coupling is strongest in the E-plane. The
radiation coupling, on the contrary, is probably stronger in the H-plane (cf.
dipoles). This is also verified by far-field radiation-pattern computations made
for single elements oriented for circumferential polarisation and elements orien-
ted for axial polarisation (Section 22.3.3). It is shown that the surface wave
radiated off the edge of the substrate is much stronger when the element is
oriented for axial polarisation.
1246 Microstrip antennas on a cylindrical surface
A comparison between Figs. 22.12 and 22.13 shows that there are large
differences in the measured amplitude and phase values for two-element com-
binations located similarly, which means that the mutual coupling is not only
t %-vector 1 amplitude
phase
Fig. 22.12 Amplitude and phase of the measured mutual coupling from a centre element
a function of distance and direction between the elements. The second-order
coupling effects result in this difference. This should be remembered when the
mutual coupling is included in the radiation-pattern calculations.
Microstrip antennas on a cylindrical surface 1247
Feed network
A block diagram of the feed network is shown in Fig. 22.14. The first power
divider splits the power equally into eight channels, corrresponding to the eight
columns of the array. In each channel there is a phase shifter, an attenuator and
1 amplitude
phase
Fig. 22.1 3 Amplitude and phase of the measured mutual coupling from a corner element
finally a 1:4 power splitter. The phase shifter can control the phase in steps of
45' between 0" and 360". The attenuator has an attenuation range of 0-40 dB
continuously. The final power divider splits the power equally and with equal
phase to the four elements in a column.
7248 Microstrip antennas o n a cylindrical surface
The computer-controlled phase shifter and attenuator are described by Star-
ski and Albinsson in References 33 and 34, respectively.
In order to be able to predict the radiation patterns of the array, it was
4 k?'
important to know the total resulting phase and amplitude at each element. A
1:8 -.
.. 45' 90"
Computer
180'
Fig. 22.14 Block diagram of the feed network
0 -40 dB
thorough measurement of the complete feed nework was therefore made. The
phase and amplitude at each one of the 32 ouput ports were measured as a
function of phase-shifter setting and attenuation. These data were then recorded
by a HP9816 computer and could be displayed during the antenna measure-
ments. The values were used as input to the computer program for the predic-
tion of the radiation patterns (see below). However, excitation errors due to
reflections at the antenna elements were not compensated for. A discrepancy
between the displayed values and the actual output values could also be expec-
ted owing to the fact that the semi-rigid cables after the measurement had to be
bent to be attached to the elements. The total difference in phase and amplitude
in the displayed values and the output values from the feed network is roughly
f0.5dB and 7". +
Antenna gain
The antenna gain was measured in the anechoic chamber described in Section
22.3.2. The reflectivity level of the room is around - 40 dB for the entire array
measurement.
The feed network was set to give a main lobe in 4 = 0". The average power
level at the elements was -25.0dB. An ideal power division would give
- 15 dB, which means that 10 dB was lost in the feed network including the
cables. This loss is mainly due to the phase shifters, which were not in their 0"
state during the gain measurements. An increase in phase shift in the phase
shifters results in an additional attenuation, and since the attenuation was not
equal in all the channels, a compensating attenuation was applied in the least-
attenuated channels.
The measurements gave an antenna gain of 13.1 (k0.8)dB. Neglecting the
feed loss, the antenna by itself has a gain of 23.1 dB. It should be pointed out
that a constant amplitude over the aperture would give an equivalent taper which
can be described by a function I/cos (4,) x g(4,,). The first term is due to the
increased density of elements towards the edge, in a projected aperture plane.
4, is the 4 angle of the m th column, as measured from 4 = 0". g(4) is the
Microstrip antennas on a cylindrical surface 7249
element pattern, which can be approximated by cos(4) for small 4 angles.
Consequently, we wanted an excitation with equal amplitudes to the elements
during the antenna-gain measurement.
An antenna gain of 23.1 (k043)dB gives an effective aperture A, of 16.3
[+3.3/-2.7]A2, using G = 4 ~ A , / 1 ~This
printed pattern.
Radiation patterns
. can be compared to the physical
projected aperture of the array, which is 9.66A2. The actual aperture with a 112
margin around the printed array is 17.541'. The effective aperture of the array
is thus approximately equal to the area of the array, including 142 around the
The cylinder was covered with substrate to a total height of 595 mm. In azimuth
the substrate covered 620mm, corresponding to an angle of 230". Radiation-
+
pattern recordings outside the range 100" could therefore have been influen-
ced bv the substrate edge. -
The calculated radiation patterns were all computed using the surface-current
model of the patch (Section 22.2.2). The mutual coupling was incorporated
using the mesured coupling coefficients from the centre element. In patterns
where a cross-polar level is predicted, mode p, q = 2, 0, with an amplitude of
40% of the dominant mode (p, q = 0, 1) amplitude, was included.
Beam steering
Let element number nm be excited by the current I,, = I(+,, z,,), where z, is the
z-co-ordinate of then th band and 4, is the 4 angle of them th column. In order
to steer the beam in a direction O0, 4,, the element excitation should be
I - j k R ~ i n B ~ c o -&)
s ( ~ ~ -jki,casOo
Inrn = nme (22.19)
when the element phases are neglected. When we analyse this expression, we find
that, to steer the beam in elevation 0, we need to calculate and change by
individual amounts the phase of every element in the array. It is thus not
possible to use one band as an entity, a so-called 'row and column' beam
steering, the same way as in a planar array. A beam steering in azimuth only
(0, = 90), on the other hand, involves the phasing of the columns.
Active-element patterns
When a single element was excited, the surrounding elements were terminated
in matched loads. The active-element patterns of a centre element and a corner
element were recorded. Fig. 22.15 gives the predicted and measured radiation
patterns for the centre element and Fig. 22.16 shows the radiation pattern for
the corner element. The former is also compared to the radiation pattern of an
isolated patch.
Array patterns
For the entire array, beam steering was performed in azimuth only. Fig. 22.17
7250 Microstrip antennas on a cylindrical surface Microstrip antennas on a cylindrical surface 7251
0 , I shows beam steering to 0, 17O, 34", 55" and 63'. The element phases have been
compensated for in the two latter diagrams. The recordings were made with the
intention of putting as much power as possible into the main beam, but still keep
a constant amplitude taper. None of the columns was therefore attenuated,
although this would have helped to reduce the sidelobes.

I I I I I

, I I I I
30
I
60
I
90
I
120
I
-40 120 90 60 30 0
angle (7

Fig. 22.15 Measured and predicted active radiation pattern for a centre element
-measured
- - - surface-current model, mutual coupling included
. . . surface-current model, isolated element

angle

Fig. 22.17 Beam steering to O", 7 7 , 344 55 and 6 3 in azimuth

The co- and cross-polar diagrams for a beam pointing in 0" are shown in Fig.
22.18. The sidelobe level for a constant-amplitude distribution is found to be
- 13dB, which is also predicted. The cross-polar level is measured to be
- 24 dB. The calculated results lie 3 dB below this value. The discrepancy is due -
to the fact that the cross-polar coupling has not been considered in the theoreti-
cal calculations and also because of the inaccurate feeding of the patches. The
elements are sensitive to cross-polar components, since mode p, q = 1, 0 is
resonant at the same frequency as the p, q = 0, 1 mode. A displacement of the
feed probe only slightly will therefore cause cross-polarisation.
Fig. 22.19 shows the co- and cross-polar curves when the beam is steered to
36". The patterns are predicted quite well with the theoretical model. The
I I I I I I I I I columns are all fed with equal amplitude.
-401 90 60 30 0 30 80 90 120 150
angle (7
Pattern synthesis
Fig. 22.16 Measured and predicted active radiation pattern for a corner element
A good synthesis method is difficult to find for conformal arrays. The projected
-measured
- - - surface-current model, mutual coupling included aperture method is the most commonly used [29, 351. Matrix inversion
 7252 Microstrip antennas on a cylindrical surface
procedures were reported by Ziehm [36] and James [37] and an iterative method
was developed by Guy [38] (also in Reference 39).
We will show here the results of a projected aperture synthesis procedure. The
sidelobes were reduced in an azimuth cut by using a sampled continuous Taylor
distribution [40].
angle (7
Fig. 22.18 Measured and predicted rad~attonpattern with the mutual couplmg mcluded for
a beam pointmg ~n0"
- measured co- and cross-polarlsatlon
- - - surface-current model, co-polarlsatlon
x x x surface-current model, cross-polar~sation
As mentioned earlier in this Section, the increased element density towards
the edge of the array is approximately compensated for by the reduced element
gain. As long as the angle is small, this is a reasonable approximation.
A 30 dB Taylor distribution was applied, in order to reduce the sidelobes. The
recorded and predicted radiation patterns are displayed in Fig. 22.20. The
predicted pattern uses the excitation delivered by the feed network with the
desired phases approximated the best that could be done with the 3 bit phase
shifters. The mutual coupling is also included. Theoretical and measured curves
agree very well in this diagram. A considerable sidelobe reduction was gained
at the cost of lobe width and directivity.
A reduced sidelobe level of a beam steered to 10" requires a new projection.
The new projection leads to an asymmetrical element distribution. Consequent-
ly, a 'new calculation of element currents has to be done. A 20dB Taylor
distribution applied to a beam steered to 10" gives the pattern shown in Fig.
22.21. The OdB level in both tapered diagrams is the level of the main lobe in
Microstrip antennas on a cylindrical surface 7253
the case of constant excitation and 0" beam steering. It should be pointed out
I that the reduced antenna gain is mainly due to the attenuation in the feed
network, since the power in the outer channels is attenuated. A constant total
power output would theoretically result in a loss of antenna gain of 0.1 dB for
a 20 dB, and 1.4 dB for a 30 dB Taylor distribution, as compared to a constant
distribution.
Fig. 22.1 9 Measured and predicted radiation pattern with the mutual coupling included for
a beam pointing in 3 6
-measured co- and cross-polarisation
- - - surface-current model, co-polarisation
x x x surface-current model, cross-polarisation
Source: E.V. Sohtell, 'Microstrip patch phased array on a cylinder', IEEE Int. Symp.
Digest Antennas and Propagation, Vol. 111, 1988, pp 1152-1 155. @ (1 988) IEEE
22.5 Summary
Two theoretical models for the microstrip patch on a cylinder have been
investigated in this Chapter. Both the cavity model and the surface-current
model give a good description of the radiation properties of the patch. Apart
from effects occurring for infinite-length cylinders in the theoretical models, and
radiation off the substrate edge in the measurement case, the agreement between
measured and theoretical curves is good. The cavity-model calculations require
much less computer time than those performed with the surface-current model,
and the higher-order modes are found quite easily with the cavity model. The
 1254 Microstrip antennas on a cylindrical surface
Microstrip antennas on a cylindrical surface 1255
surface-current model does, however, give a better prediction of the radiation
pattern for angles near the cylinder axis.
- - - - - - rO It has been shown that the patch antenna is very well suited for application
in a cylindrical phased array. The ease of fabrication and low cost make this type
of array very attractive. Beam steering & 70' in azimuth is possible without
commuting the active region. A three-step commutation will then provide
coverage of 360" while maintaining a high antenna gain.
It has also been shown that a theoretical model where the measured mutual
coupling is included gives a good picture of the electrical performance of a patch
array. Pattern synthesis may thus be performed with reasonable accuracy.

22.6 References

I KROWNE, C. M.: 'Cylindrical-rectangular microstrip antenna', IEEE Trans., 1983, AP-31,

Fig. 22.20 Measured andpredicted radiation patterns with the mutual coupling included f o ~
a 3 0 dB Taylor distribution with the beam pointing in O'
-measured
- - - surface-current model
Source: E.V. Sohtell. 'Microstrip patch phased array on a cylinder', IEEE Int. Symp.
Digest Antennas and Propagation. Vol. 111. 1988, pp 1152-1 155. @ (1 988) IEEE
Fig. 22.21 Measured and predicted radiation patterns with the mutual coupling included for
a 20dB Taylor distribution with the beam pointing in 1O'
-measured
- - - surface-current model
pp. 194-199
2 LUK, K. M., LEE, K. F., and DAHELE, J. S.: 'Input impedance and Q factors ofcylindrical-
rectangular microstrip patch antennas'. IEE Int. Conf. on Antennas and Propagation, Part 1,
York, June 1986, pp. 95-98
3 LO, Y. T., SOLOMON, D., and RICHARDS, W. F.: 'Theory and experiment on microstrip
antennas', IEEE Trans., 1979, AP-27, pp. 137-145.
4 CARVER, K. R., and MINK, J. W.: 'Microstrip antenna technology', IEEE Trans., 1981,
AP-29, pp. 2-24
5 JAMES, J. R., HALL, P. S., and WOOD, C.: 'Microstrip antenna theory and design' (Peter
Peregrinus, 1981)
6 NEWMAN, E. H., and TULYATHAN, P.: 'Analysis of microstrip antennas using moment
methods', IEEE Trans., 1981, AP-29, pp. 47-53
7 ASHKENAZY, J., SHTRIKMAN, S., and TREVES, D.: 'Electric surface current model for
the analysis of microstrip antennas on cylindrical bodies', IEEE Trans., 1985, AP-33, pp.
295-300
8 PATHAK, P. H., and KOUYOUMJIAN, R. G.: 'An analysis of the radiation from apertures
in curved surfaces by the geometrical theory of diffraction', Proc. IEEE, 1974, 62, pp. 1438-
1447
9 WU, KUANG-YUH, and KAUFFMAN, J. F.: 'Radiation pattern computations for cylin-
drical-rectangular microstrip antennas'. IEEE Int. Symp. Digest Antennas and Propagation,
Vol. 1, 1983, pp. 39-42
10 JAKOBSEN, K. R.: 'The radiation from microstrip antennas mounted on two-dimensional
objects', IEEE Trans.,1984, AP-32, pp. 1255-1259
II PENARD, E.: ' ~ t u d ed'antennes imprimkes par la method de la caviti. Application au
couplage'. D.Sc. Thesis, University of Rennes, 1982
12 SOHTELL, E. V.: 'Microwave antennas on cylindrical structures'. Technical Report 173,
Ph.D. Dissertation, School of Electrical and Computer Engineering, Chalmers University of
Technology, Sweden, Sept. 1987
13 HARRINGTON, R. F.: 'Time-harmonic electromagnetic fields' (McGraw-Hill, 1961) chap. 3
14 SANFORD, G. G.: 'Conformal microstrip phased array for aircraft tests with ATS-6'. IEEE
Trans., 1978, AP-26, pp. 642-646
IS MUNSON, R. E.: 'Conformal microstrip antennas and microstrip phased arrays', IEEE
Trans., 1974, AP-22, pp. 74-78
16 WEINSCHEL, H. D.: 'A cylindrical array of circularly polarized microstrip antenna'. IEEE
Int. Symp. Digest Antennas and Propagation, 1975, pp. 177-180
1256 Microstrip antennas on a cylindrical surface
17 AGRAWAL, A. K.: 'Cylindrical array'. IEEE Int. Symp. Digest Antennas and Propagation,
1986, pp. 549-552
18 JAYAKUMAR, I., GARG, R., SARAP, B. K., and LAL, B.: 'A conformal cylindrical
microstrip array for producing omnidirectional radiation pattern', IEEE Trans., 1986, AP-34,
pp. 1258-1261
19 DUBOST, G., SAMSON, J., and FRIN. R.: 'Large bandwidth flat cylindrical array with Chapter 23
circular polarisation and omnidirectional radiation', Electron. Lett., 1979, 15, pp. 102-103
20 OSTWALD, L. T., and GARVIN, C. W.: 'Microstrip command and telemetry antennas for
communications technology satellite'. IEE Int. Conf. on Antennas for aircraft and spacecraft,
Extensions and variations t o the
London, 1975, pp. 217-22
21 WEINSCHEL, H. D., and WATERMAN, A.: 'Cylindrical microstrip array - C-band beacon
microstrip antenna concept
antenna array with 48 rectangular radiating elements fed in phase'. Report AFGL-TR-83-
0218, AD-A135 14414, July 1983 P. S. Hall, A. Henderson, J. R. James
22 HALL, P. S., WOOD, C., and JAMES, J. R.: 'Recent examples of conformal microstrip
antenna arays for aerospace applications', IEE Int. Conf. on Antennas and Propagation,
York, 1981, pp. 397401
23 DUBOST, G., and VINATIER, C.: 'High gain array at I2GHz for telecommunications'. Int. 23.1 Introduction
URSI Symp. on Electromagnetic Waves, Munich, 1980, pp. 213A1-4
24 MUNSON, R. E.: 'Conformal microstrip communication antenna', IEEE Military Commun-
ications Conference, Monterey, 1986, pp. 23.3.14 A large number of contributions to this handbook have described analytic or
25 NEWHAM, P.: 'Monolithic patch array antenna for small missile applications'. Military design aspects relating to single microstrip elements or their use in large micro-
Microwaves, Conf. Proc. Brighton, England, 1986, pp. 335-340 strip arrays. This reflects a response to an important need for engineering
26 LEE, K. S., and EICHMANN, G.: 'Elementary patterns for conformal dipole arrays mounted developments in the communications, radar and navigation areas, where many
on dielectrically clad conducting cylinders', IEEE Trans., 1980, AP-28, pp. 81 1-18 of the ultimate uses of microstrip will be found. The number of ways of
27 ALEXOPOULOS, N. G., USLENGHI, P. L. E., and UZUNOGLU, N. K.: 'Microstrip
dipoles on cylindrical structures', Electromagnetics, 1983, 3, pp. 31 1-326 configuring a microstrip patch antenna, let alone the different array topologies,
28 HERPER, J. C., HESSEL, A,, and TOMASIC, B.: 'Element pattern of a cylindrical phased is seemingly vast, yet the fundamental performance limitations discussed in
array. Part I: Theory', IEEE Trans., 1985, AP-33, pp. 259-272 Chapter 1 are intrinsic in any microstrip antenna design. There we have cited
29 HANSEN, R. C. (Ed.): 'Conformal antenna array: Design handbook'. AD A1 10091, Jan. bandwidth extension, radiation pattern control, efficiency and feeder architec-
1982 ture, substrate technology and manufacture etc., as major fundamental issues in
30 STEYSKAL, H.: 'Analysis of circular waveguide arrays on cylinders', IEEE Trans., 1977,
AP-25, pp. 610-616 design. Any one property can generally be optimised to a large degree, but only
31 POZAR, D. M.: 'Finite phased array of rectangular microstrip patches', IEEE Trans., 1986, at the expense of the others, and copious illustrations are given throughout the
AP-34, pp. 658-665 Handbook. It is perhaps a sign of maturity in the microstrip antenna field that
32 SOHTELL, E. V., and STARSKI, J. P.: 'Cylindrical microstrip patch phased array antenna ways of combining microstrip printed technology with other constructional
- Chalscan C'. Military Microwaves Conf. Proc., Brighton, June 1986, pp. 317-322
methods are now attracting attention. The central idea here is to sacrifice to
33 STARSKI, J. P., and ALBINSSON, B.: 'Coupled transmission line as a DC isolating phase
shifter network'. TR8407, Division of Network Theory, Chalmers Univ. of Technology, some degree the manufacturing simplicity or the low profile or the choice of
Sweden, Sept. 1984 substrate materials and so on, in order to relax one or more of the above
34 STARSKI, J. P., and ALBINSSON, B.: 'An absorptive attenuator with optimised phase limitations fundamental to established microstrip technology. The concept of
response'. TR8402, Division of Network Theory, Chalmers Univ. of Technology, Sweden, mixing technologies is well known in engineering design and is best described by
May 1984 considering specific examples. The purpose of this Chapter is to report various
35 RISZK, M. S. A. S., MORRIS, G., and CLIFTON, M. P.: 'Projected aperture synthesis
method for the design of conformal array antennas'. IEE Int. Conf. on Antennas and
examples with which we have had particular design experience.
Propagation, Coventry, 1985, pp. 48-52 Section 23.2 deals with two arrangements for exercising greater control of
36 ZIEHM, G.: 'Optimum directional pattern synthesis of circular arrays', Radio Electron. Engr., radiation patterns whilst maintaining the presence of simplistic microstrip patch
1964, 28, pp. 341-355 antennas. First, in Section 23.2.1, the microstrip radiator is used as a feed
37 JAMES, J. R.: 'Conformal antenna synthesis using spherical harmonic wavefunctions', Proc. element for a large
IEE. 1975, 122, pp. 479486
- reflector antenna, thus combining the simplicity of the
38 GUY, R. F. E.: 'A synthesis technique for array antennas of general shape with various
microstrip concept with the high efficiency and good sidelobe and cross-polaris-
aperture constraints'. IEE Int. Conf. on Antennas and Propagation, Coventry 1985, pp. 35-39 ation control properties of the reflector. Similarly the combination of dielectric
39 GUY. R. F. E.: 'Power pattern synthesis of conformal arrays'. Military Microwaves, Conf. spheres with patches (Section 23.2.2) is also seen to lead to improvements in
Proc., Brighton, 1986, pp. 341-346 pattern control with novel sparse array facilities, and a possible beam-switching
40 BALANIS, C. A,: 'Antenna theory, analysis and design' (Harper & Row, NY, 1982)
7258 Extensions.and variations to the microstrip antenna concept Extensions and variations to the microstrip antenna concept 7259
technique is also realisable. Ways of satisfying multi-octave bandwidth demands
are presented in Section 23.3, and the log-periodic concept in Section 23.3.1 is
a particular demonstration of a significant improvement being made in one
parameter, namely bandwidth, at the expense of another, in this case the
physical area of the array; from this point of view, it is compatible with the w
\O
N
w
findings elsewhere in the Handbook relating to the bandwidth-size trade off, but
is, of course, an extreme case. A very new technique (Section 23.3.2) utilises the
frequency-selective properties of a printed mesh and the fact that at low fre-
quency a microstrip antenna can be composed of such a surface without
impairing its performance. The outcome is a composite structure containing two
sandwiched microstrip arrays, one radiating through the other and offering a -
0
N
I
dual-band aperture with multi-octave frequency separation.
The problem of microstrip-line loss and its increase at millimetre wavelengths
is noted in Section 23.4, and a hybrid dielectric/microstrip antenna is described
to reduce feeder loss. The merits of this antenna structure, which combines the
best properties of microstrip technology with dielectric-line technology, is dis- G-
e
cussed in the wider context of overall losses, which includes that associated with a
launching waves into the feeder. Finally in Section 23.5 we discuss the use of
extreme values of substrate permittivity and also magnetic materials to create a f
multiplicity of engineering devices with quite different applications. In Section B
23.5.1 we note the results of an investigation into the use of expanded-foam ob l o b
N VN
polystyrene substrates to reduce the cost of a large array for direct broadcasting ' I
reception from satellites. A novel robust electronic beam-scanning element for
satellites using a magnetic overlay is described in Section 23.5.2. In contrast, the
use of very high-permittivity substrates enables a reduced-size patch antenna to
radiate into human tissue for cancer therapy, as described in Section 23.5.3.

23.2 Radiation pattern control

23.2.1 Reflector feeds

Reflectors are well known for their good radiation pattern control, and when
combined with high-precision horn-type feeds lead to low-sidelobe envelopes and
low cross-polarisation over significant bandwidths. The horn is, however, con-
sidered to be the point of greatest complexity in a reflector antenna, and the need
for simple antennas makes the implementation of the microstrip reflector feed
attractive for many reasons. Their lightweight properties make them suitable for
satellite multiple-beam applications [I] and the ease of circuit integration has
resulted in investigations into their use as monopulse guidance antennas with
integrated comparator systems [2].
Simple microstrip disc elements have been used to feed circularly symmetrical
reflectors [3]. Small ground-plane sizes have to be used to reduce feed blockage.
This results in some equalisation of the E- and H-plane feed patterns as shown
in Fig. 23.1, and a consequent improvement in efficiency and cross-polarisation
 1260 Extensions and variations to the microstrip antenna concept Extensions and variations to the microstrip antenna concept 1261
compared with that expected from patches on large ground planes. Table 23.1 where
indicates typical results obtained after optimisation of the ground-plane size.
The need for a circularly symmetric feed pattern [4] implies a balanced magnetic Em = sin + cos 8 cos 44
40 (23.2)
E, = cos 0 sin 40 + cos 44
where Em and E, are the radiation fields from the magnetic and electric sources,
respectively, and the polarisation is in the y-direction. Substituting eqn. 23.2
into eqn. 23.1 and converting to reference and cross-polarised fields Ed and
Em, respectively, gives
E,~, = 1 + cos e (23.3)
EC,, = 0

\30( ,
- A-.
/ y \.)
-90 -60 -30 0 30 60 90
8 ldegl
(01
Small magnetic dipole
\
Small electric dipole

Fig. 23.2 Geometry for calculation of fields due to magnetic and electric sources

I 1-120 , I
--.
.-.,
-90 -60 -30 0 30 60 90
8 ldegl
Ibl

Fig. 23.1 Measured radiation patterns of microstrip disc on circular ground plane
Patch diameter = 16.4 mm; frequency = 5.89 GHz; substrate diameter =
32.8 mm, thickness = 3.18 mm, (hll, = 0.06). e, = 2 . 5

--
H plane
-.-.-
----
H plane
plane
I cross-polar
Fig. 23.3 Patch and ring feed

and electric source as indicated in Fig. 23.2. The radiated field is given by which indicates zero cross-polarisation. Circular polarisation can be generated
by a further pair of sources orthogonal to the first and excited in quadrature.
E = Em + E, (23.1) However, for microstrip radiators the source is primarily magnetic. The ideal
 1262 Extensions and variations to the microstrip antenna concept
Huygens source can be approached by various practical adaptations such as the
finite ground plane noted above. A further example is to surround the disc by
a quarter-wavelength-deep short-circuited annular ring [3, 41 as shown in Fig.
Fig. 23.4 Measured peak cross-polarisation of microstrip feeds
-Disc on circular ground plane (Fig. 23.1)
-- Patch and ring feed (Fig. 23.3)
---- Co-planar fed patch array (Fig. 23.6)
- Overlaid patch array (Fig. 23.9)
f = f, is patch or array resonant frequency
3 deg
Fig. 23.5 Measured radiation patterns of patch and ring feed
Patch diameter d = 19.5 mm. gap width g = 2.25mm, ring width r = 9.0mm
frequency = 5.2 GHz, substrate thickness h = 3.18mm ( h l l , = 0.06),E, = 2.5
- H plane ---- E plane,
upper curves co-polar; lower curves cross-polar
23.3. Fig. 23.4 indicates that the peak cross-polarisation is reduced by about
7 dB. Coupling to the slightly off-tuned annular resonant section also increases
the feed bandwidth by up to 50%. A typical feed radiation pattern is shown in
Fig. 23.5.
Extensions and variations to the microstrip antenna concept 7263
Feed-pattern symmetry can also be achieved by using small patch arrays [5],
and the narrower beamwidths are more appropriate to feeding larger F/D
reflectors or offset reflectors. Fig. 23.6 shows a typical four-element feed with a
co-planar corporate feed circuit. A 4% bandwidth is achieved and typical
patterns are shown in Fig. 23.7. The sidelobe asymmetry is primarily due to
radiation from the feed lines. Reflector performance of between -20 and
- 30 dB cross-polarisation level and about 66% efficiency are indicated in Table
23.1. Fig. 23.8 shows computed reflector patterns of this feed. Use of overlaid
feed networks, although resulting in increased constructional complexity, re-
duces feed-line radiation and hence cross-polarisation, as shown in Fig. 23.9 and
Table 23.1.
Fig. 23.6 Co-planar-fed patch arrays
For F / D = 0.8reflector, patch size = 9.7 mm x 7.2mm, patch spacing = 17 mm,
a = 45 mm, b = 60 mm, substrate h = 1.59mm ( h l l , = 0.06).E, = 2.32,fre-
quency = 12.2GHz
It is clear that microstrip can be combined with reflectors to give antennas
with good radiation pattern control and high gain over restricted bandwidths.
The above examples illustrate the capabilities of this hybrid microstrip reflector
antenna using mainly conventional microstrip concepts to realise the feed. There
is no doubt much further scope for innovation and improvement, perhaps in the
1264 Extensions and variations to the microstrip antenna concept Extensions and variations to the microstrip antenna concept 7265
forming of hybrid feeds. An illustration of this is the dielectric-sphere and patch
feed [6], described in the next Section. An example of optimised use of basic
patch elements within a dielectric-filled cylindrical cavity has also been given in

Fig. 23.8 Calculated radiation patterns of prime-focus-fed axi-symmetric reflector with li-
nearly polarised patch array of Fig. 23.6
( a ) E-plane
( b ) H-plane

Progress is also likely in the use of optimised array structures using the
sequential rotation technique. Fig. 23.10 shows a 16-element array [q suitable
for use as a feed for a high FjD reflector. The patches are circularly polarised
and arranged with sequentially rotated feeding [8].A simplified feed structure
compared with the conventional corporate network (Fig. 23.6) results in lower
feed radiation. The close grouping of the patch input points means that four
bends are obviated and the 180' phasing at the T-junction results in this source
of unwanted radiation being directed into the main beam, hence reducing
spillover and increasing efficiency. The changes in the feed-network radiation
characteristics are illustrated in Fig. 23.1 1. Fig. 23.12 shows that array sidelobe
levels are reduced by up to 10 dB by the new feed network, and that reductions
can also be obtained by thinner substrates albeit at the cost of reduced band-
width. Reductions in cross-polarisation are similar to those for sidelobes. Im-
provements in bandwidth are also noted using sequential rotation, with the
higher-order modes associated with low-Q patches being suppressed, leading to
the possibility of reflector feeds with bandwidths in excess of 10%.
 1266 Extensions and variations to the rnicrostrip antenna concept 1267
Extensions and variations to the rnicrostrip antenna concept

23.2.2 Spherical dielectric overlays

A truncated dielectric sphere is attached to a microstrip patch antenna as shown
in Fig. 23.13. The composite structure is no longer of low profile and its manufac-
ture is more complicated, but an additional dimension in radiation pattern
control is achieved [6,9, 10, 1I]. The spherical-lens action occurs in the near field
of the microstrip patch and can be approximately analysed [I I] by representing
the patch radiator by Hertzian dipole elements at its perimeter and involving
equivalent current sources on the sphere surface. The effect of the sphere on the
patch resonant frequency and input impedance is of second order and readily
allowed for. An example of the theoretical and measured radiation patterns are
given in Fig. 23.14. As expected, the sphere tends to focus the broad patterns
of the patch, with the resulting gain of the structure increasing with sphere
diameter until a threshold diameter is reached. For spheres having relative
permittivities of 2.3, the threshold was reached for a sphere diameter of about
three free-space wavelengths and a gain of about 16.5 dBi. At larger diameters
the field distribution over the sphere aperture departs from the illumination
required for low sidelobes and narrow main beam. This is also confirmed by the
analysis which allows for the dissipative losses in the sphere and the degree of
truncation; progressively increasing the truncation also reduces the gain.

Fig. 23.1 0 Silhouette of disc array for circular polarisation using sequentially rotated feeding
Array details: E, = 2.32, substrate thickness = 1.59 mm = 0,05&, fre-
Fig. 23.9 Radiation patterns of overlaid patch array quency = 12.0 GHz. patch spacing = 0.7&.
FID = 0.8, patch size = 1 2 . 0 mm x 8.0 mm, patch spacing = 1 8 mm.
a = 5 6 mm, b = 5 0 rnm, patch height = 1.59 mm ( h l l = 0.06). feed
height = 0.79 mm, 8, = 2.32, frequency = 10.6 GHz
This new composite element offers greater freedom of design with regard to
-co-polar
cross-polar I measured
-- co-polar
I
---- cross-polar the'w the equality of the E- and H-plane beamwidths, and there is also some suppress-
ion of the cross-polarisation levels of the patch antenna. These aspects are
 1268 Extensions and variations t o the microstrip antenna concept Extensions and variations to the microstrip antenna concept 1269

II Fig. 23.13 Sketch of dielectric sphere overlaid on a microstrip patch antenna showing
truncation of sphere

o', degrees

Fig. 23.1 1 Computed radiation patterns of feed network of array of 4 x 4 elements with disc
radiation suppressed
a Sequentially rotated feed
b Conventional feed as Fig. 23.6
-co-polarised ---- cross-polarised 4 = 0'
Patterns are normalised to peak radiation of array and feed together

array

as Fig. 23.10

/
Fig. 23.14 H-plane pattern of a rectangular patch element with overlaid sphere at 15.4 GHz
Sphere radius a = 25 mm. 8, = 2.2. tanS, = 0.0003; microstrip substrate
,, = 2.3, rand, = 0.0001. height = 0.79 mm, patch size = 5.9 mm x 5.9 mm
e
-401 I -co-polar measured
0 0.05 0.1 -.-.- co-polar, theory
hlh, ---- cross-polar measured
Fig. 23.1 2 Computed peak sidelobe level of feed radiation of 4 x 4-element array . . . . . co-polar pattern of patch alone, theory
Array details as Fig. 23.6 and Fig. 23.10
-e,= 1.06---e,= 2.32
 1270 Extensions and variations to the microstrip antenna concept Extensions and variations to the microstrip antenna concept 1277

9,' degrees
-90 -60 -30 0 30 60 90
I I

L b
Fig. 23.1 5 Measured radiation patterns of circular microstrip patch with overlaid truncated
sphere at 1 1.98 GHz
Sphere radius a = 11.1 mm. truncated sphere height h' = 0.92 x diameter,
E,, = 2.2, tans, = 0,001; microstrip substrate E , ~= 2.3, tans2 = 0.001,
height = 0.79 mm. patch radius = 8 . 5 mm, patch probe = 3.2 mm from centre.
Element gain = 10.8 dBi.
- H -plane co-polar
---- H-plane cross-polar
E-plane co-polar
. . . . . E-plane cross-polar
a Amplitude
b Phase centred at z = 7 mm from substrate groundplane
 1272 Extensions and variations to the rnicrostrip antenna concept
Extensions and variations to the rnicrostrip antenna concept 1273
illustrated in the design [6] of a feed for a reflector antenna. The spherelpatch
feed element was required to illuminate the reflector with E- and H-plane
patterns of similar beamwidths over a bandwidth of 11.7-12.5 GHz. Cross-
polarisation levels were maintained below - 15 dB while the phase deviation
over the reflector illumination sector of 50" is small (Fig. 23.15). In another
development of this antenna, orthogonal feeds were attached to the patch to
facilitate circular polarisation. The weight of the spherejpatch exceeds that of a
small microstrip array having the same gain, although the pattern control of the
former is superior and is an attractive alternative to a horn feed.
The deployment of the spherelpatch antenna in large arrays has some signifi-
cance at millimetre wavelengths because the spheres can be constructed as a
moulded planar radome. For arrays with moderate sidelobe levels sparse array
techniques can be used, and an example is shown in Fig. 23.16. This arrange-
ment offers a simplified feeder system for the patch antennas. When smaller
spheres are overlaid on an array with conventional element spacing, there is
some improvement in the cross-polarisation levels. These properties are sum-
marised in Table 23.2.
' I , , I t , , Another application of the dielectric overlays concerns a way of beam scan-
ning by placing several microstrip patches beneath the sphere. Computations
[ll] show the effect on the radiation patterns when three isolated quarter-
wavelength patch radiators are switched on and off. A beam swing of 25" is
achieved which, without the sphere, would require many more microstrip
patches plus a phase-shifter system.
It is expected that the spherical dielectric overlay technique will find use in
special applications demanding additional pattern control, and Dubost has
subsequently described a cylindrical implementation using a dielectric-rod lens
and a linear array [12].

23.3 Wide-bandwidth techniques

23.3.1 Log-periodic structures

Fig. 23.1 6 Sparsely distributedpatch array with overlaidspheres at 90 GHz, with no trunca-
tion of spheres
a Photographs of arrays with 6 4 and 16 elements
b E-plane radiation patterns of 64-element array
Element spacing = 2a = 6.35mm. h'12a = 1 .Or E,, = 2.2. tan 6, = 0.001,
E,) = 2.3, tansr = 0.001. h = 0.127 mm, patch size = 1 .O mm x 1 .Omm.
- co-polar measured
---- co-polar theory
. . . . . cross-polar measured
Microstrip antennas and arrays have bandwidth limitations that have been
extensively documented. Owing to the inherent resonant action, patch antennas
have bandwidths typically less than 10% [13, 311. Increases on this can be
achieved by the use of thick patches with probe inductance compensation [14],
stacked patches [I51 or the use of coplanar parasitic elements [16]. Even in these
cases bandwidths are usually less than 30% but the use of log-periodic scaling
of arrays of microstrip patches allows bandwidths of up to two octaves (120%)
to be achieved. Applications of such wideband arrays include electronic warfare
and wideband radar and measuring systems [17] in those cases where the low
profile is an important system consideration.
The log-periodic principle applied to microstrip patches [la, 191 is illustrated
in Fig. 23.17, where the patch size and spacing increases along the array by a
1274 Extensions and variations to the microstrip antenna concept Extensions and variations to the microstrip antenna concept 7275

factor z, where: region and the array or element pattern should have a null in the direction of
the wave propagating along the array. These conditions can be met by most
microstrip arrays. However, to ensure wideband action the propagation charac-
teristic of the array should have no stop bands below the resonant frequency.
where I, w and dare defined in Fig. 23.17 and the subscripts refer to the nth and
(n f 1)th patches. In addition, the substrate thickness for each patch should be
similarly scaled, effectively producing an array on a tapering substrate as shown
in Fig. 23.17~.Such scaling will ensure that the array characteristics vary
periodically with the logarithm of the frequency, provided certain conditions
relating to the patch connection to the feed line are observed. This arrangement
is the microstrip analogue of the log-periodic dipole array [20]. In both cases the
elements close to resonance form an active region, giving rise to strong radi-
ation, and hence attenuation of the travelling wave on the array.

ty
- H plane

L x
input

feed ,patches

U
.z
I
0 2 -1 0 1
pd lradsl u,d lnepers)
(b)
Fig. 23.18 ( a ) Equivalent circuit of single overlaid patch as shown in Fig. 23.17
Fig. 23.17 Log-periodic electromagnetically coupled overlaid patch array
( 6 ) Propagation characteristic of uniform overlaid patch array
a With scaled feed line and substrate
-calculated with mutual coupling--- calculated without mutual coupling
b With uniform patch displacement p, substrate thicknesses h, and h, and feed-
0 measured points
line width w,
I = 1 0 m m . w = 8 m m . p = 1 . 2 5 m m , d = 9 4 2 m m , w , = 2 . 5 m m , h,=
1,586 mm, h, = 0.793 mm, 8, = 2.32

Criteria for successful log-periodic operation of series-fed arrays have been The complex propagation constant 8' = fi + ju of a uniform array is given by
deduced. To prevent excitation of higher-order mode resonances in the low-
frequency elements beyond the active region, the array should be fed from the cos kd -sin kd
+ jy,,
high-frequency end, should have high attenuation within and beyond the active cos /?'d = 2 yo
K Z kd
1 - j-sin
r,
 1276 Extensions and variations to the microstrip antenna concept
i Extensions and variations to the microstrip antenna concept 1277
where adjacent-cell mutual coupling only is assumed. d is the array period
length, Y,and k are the feed-line admittance and wave number respectively, and
Y , , and Y,, are the self and mutual admittances of the array elements. Fig. 23.18
I Fig. 23.20 shows the input return loss and gain of a 36-element overlaid patch
log-periodic array designed for a near-broadside beam. Return loss is less than

shows the equivalent circuit of the overlaid patch and the propagation charac- 1 - 10 dB and the gain is greater than 8 dB over the bandwidth 4-16 G H z The

teristic of a uniform array of such patches, deduced using eqn. 23.5. k,, = 2~f/c
is the free-space wave number. The normalised propagation constant pd is -a
at zero frequency (kod = 0) due to the alternation in patch feeding. pd rises
smoothly from - T to the resonant region at kod = 1.7, where heavy attenu-
ation takes place owing to strong radiation. On the other hand, a stop band is
noted in the propagation characteristic of a quarter-wavelength-line coupled
patch array (Fig. 23.19). Similar characteristics have also been noted 1191for the

1
8 12 16 20
Freq.(GHz)
(cl
Fig. 23.19 Calculated propagation characteristic of quarter- wavelength line coupled patch
array Fig. 23.20 36-element overlaid patch log-periodic array
Array configuration and equivalent circuit of single period shown inset at top right a Array silhouette
b Measured input return loss IS,, I and transmission loss IS,, I
and left, respectively - measured
I = 10mm. w = 8 m m t d = 9 ~ 8 2 m r n .w , = 2.5mm.h= 1.586mm.e, =2.32.
w, = 0.5 mm, I, = 7.0 mm
--- calculated
c Array gain -measured x calculated
I, = 3.67 mrn, w, = 2.92 mm, dl = 3 6 7 mm, r, = 1.05.
comb-line array and the series-connected patch array [21], and it has been h, = 0.793mm, h, = 1,586mm. p = 1.25.8, = 2.32, T, = 1 .O
concluded that, for good log-periodic action, the patches need to be electromag- Overall array size = 340 mm x 50 mm
netically coupled to the feed line in some way. This introduces a series resonance
into the element equivalent circuit which simulates that of the successful dipole calculated efficiency ranges from 85% to 70% across this band. Array analysis
log-periodic element. Such conclusions are, however, deduced from heuristic is based on an equivalent-circuit model of the array and reasonable agreement
reasoning, and it may well be that other forms without coupling gaps may be with theory is noted. Fig. 23.21 shows radiation patterns in the H-(longitudinal)
discovered which also produce good log-periodic action. plane. Significant narrowing of the beamwidth occurs at high frequencies. This
 1278 Extensions and variations to the microstrip antenna concept
is believed to be due to the increased electrical thickness of the active-region
patches at high frequencies, which reduces feed-line patch coupling, and hence
increases the active-region length. The high cross-polarisation is also believed to
be due to this effect. The lack of tapered substrate thickness is noted as the main
limitation on the overall bandwidth of this log-periodic array.
Extensions and variations to the microstrip antenna concept 1279
Figs. 23.22 and 23.23 show results for the quarter-wavelength line coupled
log-periodic array, the so-called quasi-log-periodic array [22], shown inset in
Fig. 23.19. The array is constructed on a single uniform-thickness substrate. Fig.
23.22 shows input return loss and gain, and Fig. 23.23 shows E- and H-plane
radiation patterns. A bandwidth of 22% was obtained for a practical antenna
with five elements.
frequency , GHz

2.6 2.8 3.0 3.2 3.4

frequency, GHz
Fig. 23.22 Performance of quasi log-periodic antenna; antenna shown inset in Fig. 2 3 . 1 9
a Measured and computed return loss as a function of frequency
- measured
-
- - computed
b Measured power gain in broadside direction as a function of frequency
Fig. 23.21 H-plane patterns of array of Fig. 23.22
( a ) 4 GHz; ( b ) 16 GHz
-measured co-polar The log-periodic principle can be also applied to microstrip to form endfire
--- calculated co-polar arrays [23], in contrast to the broadside-beam types described above. Appli-
-.-.- measured cross-polar
cations here include mounting on nose cones of aerospace vehicles for a range
 7280 Extensions and variations to the microstrip antenna concept Extensions and variations to the microstrip antenna concept 1281
n"
of electronic-warfare functions. The key features in design are choice of element
I spacing and phasing and choice of element. The necessary close element spacing
[24] and strong element radiation in the endfire direction are achieved by using
quarter-wavelength shorted patches as shown in Fig. 23.24. These are parasiti-
cally coupled by tabs to the feed line to avoid stop bands [19]. Fig. 23.25 shows
freq IGHzI

trans-
rniss~on
loss

frea IGHzI

return
Loss
Id01

N)

15

dB Fig. 23.25 Performance of endfire log-periodic microstrip patch array

Fig. 23.23 Measured and computed co-polar radiation patterns of quasi log-periodic anten. a Transmission loss
na ---- measured
( a ) E-plane pattern at 2.81 GHz --- computed
( b ) E-plane pattern at 2.87 GHz b Measured return loss
( c ) H-plane pattern at 2.8 GHz c Measured radiation patterns respectively, at:
-measured -4.25 GHz;
--- computed --- 4.5 GHz;
4.75 GHz,
. . . . .5.0 GHz.

the performance of an array which was designed for a 25% bandwidth. Similar
short-circuit
performance trade-offs to those for the broadside log-periodic array are expec-
Fig. 23.24 Silhouette of endfire log-periodic microstrip patch array ted. as are limitations in overall bandwidth due to the use of flat substrates.
 7282 Extensions and variations to the rnicrostrip antenna concept Extensions and variations to the rnicrostrip antenna concept 1283
23.3.2 Dichroic dual-function apertures 23.31 is a photograph showing its application. Many variations of the dichroic
There is a growing interest in apertures that can be used at different frequency microstrip concept have been exploited including a rectangular mesh with
bands, and for microstrip arrays this is most readily achieved by embodying integral-array feeders [29] and a novel window antenna [27l whereby both the
dual-frequency patch elements in the array. The resulting band separation is ground plane and patches of a microstrip array are composed of frequency
generally small, but a multi-octave separation is now possible using the new
dichroic microstrip concept [25-271 which permits two microstrip arrays to be
sandwiched together as in Fig. 23.26. The innovation rests with array B com-
posed of a printed mesh, which at f, is found to function as a conventional
microstrip antenna. At frequency f,(> f,) the frequency-selective properties of
the mesh render it 'transparent' to the radiation from array A. Typically f H f L

Fig. 23.28 Geometry of the dual-band dichroic array showing the frequency-selective sur-
face array 6 superimposed on a microstrip array A

will be about 8 : l . The measured and computed radiation patterns of a mesh

Fig. 23.27 NEC-programme computations and measurements of square-mesh microstrip
patch on a low-permittivity substrate are shown in Fig. 23.27, and apart from patch antennas. H-plane radiation pattern of a mesh patch with transmission-line
a higher input impedance the co- and cross-polar patterns are very similar to feed (inset)
those of a conventional solid patch. A key issue is the degree of transparency a = 5.6cm; c = 1 cm, t = 0.1 cm, h = 0.06cm, frequency = 2.3GHz
experienced because the mesh patch is of limited size and furthermore is in the
near field of the radiation sources. Computations using the NEC program 1281 } measured 6, = I .05
and measurements establish, however, that the resulting transparency is sur-
prisingly good, as the measurements in Fig. 23.28 show. A scattering analysis
[29] gives further insight into the transparency, and establishes that the mesh
perimeter has a significant effect, causing the ripple-interference pattern (Fig.
I
---- computed E,
(i) co-polar
= 1.0

23.28). In the practical deployment of the dichroic concept there are many
effects to consider which differ with the application, and numerous examples
have been studied. The effect of a mesh on a 4 x 4 microstrip patch array is
illustrated in Fig. 23.29, and the extension to higher-gain arrays in Fig. 23.30,
which also shows an increase in sidelobe level due to scattering from the feeders
of array B. The technique has been applied at millimetre wavelengths, and Fig.
(ii) cross-polar
selective mesh, thus allowing the arrays to be used as an electromagnetic
window at fH and as a conventional array at f,. Having introduced the dichroic
microstrip concept, it is clear that much further exploitation remains using
different geometries of mesh and substrate permittivities. The use of high-
 7286 Extensions and variations to the microstrip antenna concept Extensions and variations to the microstrip antenna concept 7287
has established several issues:
The increase in microstrip-line loss with frequency is not as excessive as
originally contemplated.
Travelling-wave dielectric antennas have lower loss than the microstrip
counterparts, but control of radiation patterns is a problem.
The lower transmission loss of dielectric antennas is negated by their incom-
patibility with waveguide feeds because of launcher radiation losses.
Microstrip technology is the most attractive option up to 140 GHz, but there
I are some merits in the hybrid dielectric/microstrip linear array that combines
the best features of microstrip and dielectric technologies.
We will outline here the main features of the hybrid antenna, but first re-
capitulate briefly on the level and source of the increased microstrip-line loss at

Fig. 23.31 Photograph of 16 x 16, 90 GHz array showing superimposed 2 x 2, 9 GHz

array removed.

waveguide Iaunc

Fig. 23.30 Measured H-plane radiationpatterns of a comb array A under a mesh-parch array,
6; mesh patch has a size a x a
a = 7.95 cm, c = 1.1 1 cm. t = 0.48 cm, frequency = 17 GHz, E,, = 2.32,
e, = 1.05, h, = 0.08 cm, h, = 0.6 cm, comb array size = 28 cm x 28 cm.
(i)co-polar
(ii) cross-polar
- without mesh-patch array
--- with mesh-patch array

Fig. 23.32 Hybrid dielectriclmicrostrip antenna array

1288 Extensions and variations to the microstrip antenna concept Extensions and variations to the microstrip antenna concept 1289
millimetre wavelengths. A factor that has previously obscured the accurate Table 23.3 Power budgets of 9 0 GHz hybrid linear microstrip patch anten-
measurement of microstrip loss at these higher frequencies is the leakage of nas
radiation from the transitions attached to the line. A key contribution in the
investigation [30] was the quantification, and hence separation, of the radiation (9 (ii)
from the dissipative loss. It was found that for RT Duroid substrates the loss Beamwidth, deg. (theory)
at 90 GHz was 0.13 dB/1,, as opposed to 0.1 dB/L, at 20 GHz where 1, is the (measured)
microstrip-line wavelength. Experiments and computations indicated that some Squint angle, deg (measured)
loss reduction could be achieved by polishing the printed conductors to decrease Sidelobes, dB (measured)
surface roughness, but the practical implementation was unrealistic. Even less I Cross-polar, dB (measured)
realistic in manufacture is the process of thickening the conductor so that its Directivity*, dB
edges can be rounded. Gain, dB (measured)
The hybrid dielectric/microstrip-antennaconcept aims at replacing the micro- Launcher loss, dB (theory)
strip line in a linear array with an insular guide feeder; the practical arrangement Mismatch loss, dB (measured)
is illustrated in Fig. 23.32. The microstrip patches electromagnetically couple to Feeder loss?, (theory)
the insular guide feeder, and analysis and measurements establish that the array Resonator lossg*, dB (measured)
distribution is controlled by the feeder-patch separation. A summary of the Load loss*, dB
array performance is given in Table 23.3 and a photograph of the structure in Element spacing, mm
Fig. 23.33. The radiation pattern control is superior to that of a planar dielectric Efficiency , % (assuming no
antenna, yet the low-transmission-loss qualities of the latter are preserved. The launcher or load loss)
hybrid antenna only retains its advantage over a microstrip array when used in Efficiency" of microstrip array
a compatible dielectric-feeder arrangement, and the test set-up in Fig. 23.32 with of same physical size, %
a waveguide/insular guide transition will lower the system efficiency owing to ' Directivity = 10 log ( A / & cos (squint angle) dB; A = aperture area
increased launcher loss. t Calculated using 0.05 dB/& over 40 wavelengths
'' From measurements on patch-resonator Q-factors
'(i) Load loss = 24.2 - (17 + 2 + 0.04 + 2 + 0.97) = 2.19 dB
23.5 Novel use of materials
(ii) Load loss = 23.6 - (18 + 2 + 0.04 + 2 +
0.97) = 0.59 dB
5 (i) Efficiency = 24.2 - (17 + 2 + 2.19) = 3.01 dB = 50%
(ii) Efficiency = 23.6 - (18 + 2 + 0.59) = 3.01 dB = 50%
23.5.1 Foam substrates for Iarge direct-broadcast-satellite domestic receiving " Microstrip feeder of same physical length (39 wavelengths), E~ = 2, Z , = 50 R, line
arrays loss = 0.1 3 dB/&, patch loss = 0.97 dB
The use of expanded polystyrene foam for microstrip substrates is well known,
and for a patch antenna it is generally beneficial in terms of the bandwidth and
efficiency [31]. The use of a foam substrate in a large array appears to offer the
outstanding advantage of a low-cost substrate, and this has recently been
investigated [32,33]. Fig. 23.34 shows the laminar construction technique of the
foam array. A distinct problem concerned the design of triplate corporate feeds
for this electrically large array because the foam did not confine the feed fields
as well as a plastic substrate and a significant radiation loss was experienced at
the transition region between the corporate feed and antenna. The array specifi-
cation was complicated by the requirement for a squinted beam and cikular
polarisation, the latter being achieved using a polariser grid. Fig. 23.35 shows
CMs
the measured radiation patterns. The pattern and bandwidth specifications were 1 1 1 1 1 1 1 1 1 1 1
difficult to meet with a good margin, and the corporate feed loss contributed
somewhat to the noise performance of the receiving system and increased the
physical size of the array. A detailed comparison with conventional dish anten-
Fig. 23.33 Photograph of 80-element 90 GHz hybrid dielectric/microstripantenna array
 1290 Extensions and variations to the microstrip antenna concept Extensions and variations to the microstrip antenna concept 7297

nas [33] is given in Table 23.4, illustrating the relative effects of noise due to
antenna loss and the receiver noise. The conclusion for this particular applica-
tion is that the use of a foam substrate leads to low costs but there is a
compatibility problem with the feed arrangement, and the specification could be

H-plane
Fig. 23.35 Measured circularly polarisedradiationpatterns o f foam array at 17.9 GHz for four
33-finger combline sub-array with metal-plate polariser

Table 23.4 Comparison of antenna types and projected carrierlnoise ratios

( C I N ) at 17.9 GHz: NF = noise figure
0.9 m 0.6 m Large microstrip
dish dish array
Fig. 23.34 Laminar construction o f foam microstrip array
1 Metal ground plane 0.5 0.5 5.2 Antenna dissipation, dB
2 Foam substrate -4.56 - 8.07 -4.50 Aperture capture area, dB m2
3 Combline array elements
Total C/N, dB:
4 Foam spacer
5 Polarising grid 14.0 10.5 9.0 8 dB NF receiver
6 Waterproof cover 17.5 14.0 12.2 5 dB N F receiver
20.7 17.2 14.9 2.5 dB NF receiver
more readily met using a higher-permittivity substrate and perhaps some lower- These figures assume:
loss waveguide feeders for the longer lengths of feed. The feasibility then rests (a) 1.4 dB rain loss
with manufacturing a substrate capable of robust stable operation at low cost. (b) 0.3 dB galactic noise
(c) 160 W transmitted power
The array size, however, is very dependent on the noise performance of the (4 Antenna is modelled as a simple attenuator situated before receiver
receiver, and recent developments in semiconductors make possible a smaller For C-MAC system, picture qualities are 'fair' for C / N > 9 and 'good' for C / N > 12
 1292 Extensions and variations to the microstrip antenna concept
more efficient array, thus endorsing the feasibility of a flat plate DBS antenna.
This research project [32, 331 has been a milestone in the development of a
low-cost DBS array and clearly demonstrates the trade-offs between perfor-
mance and cost. The additional difficulties of electronic beam scan are also
highlighted.
Extensions and variations to the microstrip antenna concept 1293
23.5.3 Use of very-high-permittivity substrates in hyperthermia applicators
Microstrip structures have also proved useful in the design of applicators to
dispense high-power electromagnetic waves into human tissue in cancer

quor ter wavelength

section X - X

Fig. 23.37 Hyperthermia applicator using quarter- wavelength microstrip patch

Fig. 23.36 Scanned microstrip patch antenna using thick ferrite slab

-
a Construction of ferrite microstrip antenna showing co-ordinates
b H-plane patterns of antennas
e, = 10. e, = 13.8.p, p2 2. 1, 2a = 6 mm, frequency = 8.6 GHz

(ii) h, = 3 mm
(iii)h2 = 3 mm magnetised
Field strength of magnet was measured as 0.04 T in free space

23.5.2 Magnetic materials and beam scanning

Magnetised substrates have been used as substrates for patch antennas [34].
More recently [35] a thick ferrite superstrate with an applied magnetic field, as
shown in Fig. 23.36a, was devised and led to the creation of a beam that could
*
be scanned between 15" (Fig. 23.366). An approximate analysis of the element
illustrates the fundamental non-reciprocal action of the ferrite, but the precise Fig. 23.38 Photograph of microstrip hyperthermia applicatol
behaviour embraces many effects including leaky-wave and surface-wave action
in the ferrite slab. This type of element is of interest in certain communication therapy. This fairly recent use of electromagnetic waves [36] necessitates the
applications demanding a robust construction, and the use of modem magnetic design of applicators that are well matched to the impedance of the human body
materials with temperature compensation will lead to a compact assembly. and are compact in size for clinical convenience. An annular microstrip-loop
 1294 Extensions and variations to the microstrip antenna concept Extensions and variations to the microstrip antenna concept 7295
radiator has been devised [38] using plastic substrates, but the more recent 23.7 Acknowledgements
compact applicators [36, 371 use a very high substrate permittivity. A sketch of
a typical compact applicator is shown in Fig. 23.37 and a photograph in Fig. The authors would like to acknowledge the contribution to the work performed
23.38. In these designs the size reduction is achieved both by the high permit- at RMCS and described in this chapter, of Dr G Andrasic, Dr C Hall, Dr C
tivity and the use of a quarter-wave as opposed to half-wave resonator. Further Prior and Mr R Johnson.
details are given in Table 23.5 for lower-frequency applicators. The penetration
depth is a measure of performance, indicating the distance over which I/e2of the
power is dispensed. Focused arrays of these applicators bring about an in- 23.8 References
creased penetration depth. The compact applicator can be designed for a good I
stable input impedance match in operation without the use of additional match- I WOO, K.: 'Multiple beam antenna feed development'. IEEE Int. Symp. on Antennas and
ing elements, and further size reductions are possible at low frequencies by Propagation, Philidelphia, USA, 8-13 June, 1986, Vol. 1, p. 409-412
inserting ferrite material in the low-impedance regions of the substrate. 2 OLTMAN, H. G., WEEMS, D. M., LINDGREN, G. M., and WALTON, F. D.: 'Microstrip
components for low cost millimetre wave seekers' In 'Millimetre and submillimetre wave
propagation and circuits'. AGARD Conf. Proc. 245, pp. 27-1 to 27-9
3 HALL, P. S., and PRIOR, C. J.: 'Wide bandwidth microstrip reflector feed element'. 15th
Table 23.5 Characteristics and performance of low-profile applicators European Microwave Conference, Paris, Sept. 1985, p. 1039
4 HALL, P. S., and PRIOR, C. J.: 'Microstrip feeds for prime focus fed reflector antennas', IEE
Applicator A B ' Proc., 1987, 134H, pp. 185-193
Material: 6, 30 30 5 HALL, P. S., and PRIOR, C. J.: 'Microstrip array for reflector feed applications'. 14th
European Microwave Conference, Liege, Sept. 1984, pp. 631-636
Pr 1 2.5 6 JAMES, J. R., HALL, C. M., HALL, P. S., and PRIOR, C. J.: 'Dielectric sphere reflector feed
Element: quarter wavelength, cm 7.5 7.5 with microstrip excitation'. Proc. ISAP 1985, Tokyo, Aug. 1985, pp. 101-104
Dimensions, cm 15 x 15 x 4.5 15 x 15 x 4.5 7 HALL, P. S., and HALL, C. M.: 'Coplanar corporate feed effects in microstrip patch array
Thickness, cm: Substrate 2.2 2.2 design', IEE Proc., 1988, 135H. pp. 180-186
Superstrate 2.2 2.2 8 TESHIROGI, T., TANAKA, M., and CHUJO, W.: 'Wideband circularly polarised array with
Frequency, MHz sequential rotation', Proc. ISAP 85, Tokyo, Aug. 1985, pp. 117-120
200 120 9 HALL, C. M., ANDRASIC, G., and JAMES, J. R.: 'Microstrip planar arrays with dielectric
Net power for 235 W kg-' 156 406 sphere overlays', Electron Lett., 1985, 21, pp. 356-357
Penetration depth, cm (phantom) 4.0 4.3 10 HALL, C. M., JAMES, J. R., and ANDRASIC, G.: 'Microstrip patch arrays with spherical
dielectric overlays', Proc. Int. Conf. on Antennas and Propagation, Warwick University, IEE,
April 1985, p. 89-93
11 JAMES, J. R., HALL, C. M., and ANDRASIC, G.: 'Microstrip elements and arrays with
spherical dielectric overlays', IEE Proc., 1986, 133H, pp. 474482
23.6 Summary comment 12 DUBOST, G., and NICOLAS, M.: 'Broad angular coverage and large bandwidth antenna.
17th European Microwave Conference, Rome, Sept. 1987
The extensions and variations to the microstrip-antenna concept described are 13 JAMES, J. R., HENDERSON, A., and HALL, P. S.: 'Microstrip antenna performance is
determined by substrate constraints', Microwave Syst. News, 1982, 2, pp. 73-84
vastly different in application, and emphasise both the ingenuity that has 14 HALL, P. S.: 'Probe compensation in thick microstrip patches', Electron Lett., 1987, 23, pp.
already been experienced and the fund of innovative ideas that will continue to 606-607
be created in future years. It was stated in Chapter 1 that the inspiration for I5 HALL, P. S., WOOD, C., and GARRETT, C.: 'Wide bandwidth microstrip antennas for
innovation lies in the demands made by new systems, and this is well illustrated circuit integration', Electron. Lett., 1979, 15, pp. 458-460
by the efforts on one hand to exert more control of radiation patterns, and on 16 KUMAR, G., and GUPTA, K. C.: 'Non-radiating edge and four edges gap coupled multiple
resonator broad band microstrip antennas', IEEE Trans, 1985, AP-33, pp. 173-177
the other to increase bandwidth by an order. In each case one parameter is 17 BENNETT, C. L., and ROSS, G. F.: 'Time domain electromagnetics and its applications',
optimised at the expense of the others in a way which is compatible with the Proc. IEEE, 1978, 66, pp. 299-318
system constraints prevailing, but above all it is the microstrip technology that 18 HALL, P. S.: 'New wideband microstrip antenna using log periodic technique', Electron Lett.,
is diluted by other construction techniques to effect the desired optimisation. As 1980, 16, pp. 299-318
already mentioned, it is a sign of maturity in a technology when a designer has 19 HALL, P. S.: 'Multi-octave bandwidth log periodic microstrip antenna array' Proc. IEE.,
1986, 133 H, pp. 127-136
learnt just how to exploit selectively its best features with confidence. 20 ISBELL, D. E.: 'Log periodic dipole arrays' IRE Trans., 1960, AP-8, pp. 260-267
21 DONG, W. R., and SENGUPTA, L. L.: 'A class of broad band patch microstrip travelling
wave antennas', IEEE Trans., AP-32, (I), pp. 98-100
 7296 Extensions and variations to the microstrip antenna concept
22 PUES, H., BOGAERS, J., PIECK, R., and VAN DE CAPELLE, A.: 'Wideband quasi log
peripdic microstrip antenna', IEE Proc.. 1981, 128H, pp. 159-1 63
23 HALL, P. S., and SPARROW, A,: 'Microstrip log periodic antenna array with end fire beam',
Electron Lett., 1987, 23, pp. 912-913
24 GREISER, J. W., and MAYS, P. E.: 'The bent backfire zigzag', IEEE Tram., 1964, AP-12,
pp. 28 1-290
25 BOND, K., and SHELLEY, M. W.: 'Dual frequency antenna integration using invisible
grating structures', IEE Proc., 1986, 133H. pp. 137-142
26 JAMES, J. R., and ANDRASIC, G,: 'Dichroic dual band microstrip array', Electron Letts., Index
1986, 22, pp. 1040-1042
27 ANDRASIC, G., and JAMES, J. R.: 'Microstrip window array', Electron. Lett., 1988,24, pp.
96-97
28 BURKE, G. J., and POGGIO, A. J.: 'Numerical electromagnetic code (NEC) method of
moments', Tech. document 116, Naval Electronic Systems Command, July 1977
29 JAMES, J. R., and ANDRASIC, G.: 'Superimposed dichroic microstrip antenna arrays', IEE Abberation, 96 reciprocity method, 278
Proc., 1988, 135H. No 5, pp. 304-312 Acid, copper, 936 Richmonds reaction equation, 597
30 JAMES, J. R., and HENDERSON, A.: 'Planar millimetre wave antenna arrays' In BUTTON, Activation energy, 875 segmentation, 456,488
K. I. (Ed.): 'Infra-red and millimetre waves'. Vol. 14: Millimetre components and techniques- Active angle, 1241 spectral domain approach, 590
Part V. (Academic Press, 1985), pp. 189-247 Active element pattern, 701,739, 1249 steepest descent method, 278
31 HENDERSON, A., JAMES, J. R., and HALL, C. M.: 'Bandwidth extension techniques in Active elements in a cylindrical array, 1249 surface current model, 1228
printed conformal antennas'. Military Microwaves Conference, Brighton, 1986, pp. 329-334 Active impedance, 732 synthesis, 361
32 HENDERSON, A,, JAMES, J. R., HALL, P. S., STOTT, J. H., and BOARDMAN, D. H.: Active patches, 34 transmission line model, 456
'Investigation of a cost constrained 12 GHz flat plate antenna for DBS'. Proc. ICAP, War- Active reflection coefficient, 708, 739 variational method, 235
wick, 1985, pp. 108-112 Additives, brightener, 936 Wiener-Hopf method, 478
33 HENDERSON, A,, and JAMES, J. R.: 'Low cost flat plate array with squinted beam for DBS Adhesion, 936 Angled slot array, 39
reception', Proc. IEE, 1987, lMH, pp. 509-514 Adjustable height patch. 30 Anisotropy, 881, 909, 917
34 DAS, N., CHOWDHURY, S. K., and CHAlTERJEE, J. S.: 'Circular microstrip antenna on Admittance, aperture, 727 Annealed, 946
ferromagnetic substrate', IEEE Trans, 1983, AP-31, pp. 188-190 Aerospace systems, 1057 Annular ring, 25,879,1262
35 HENDERSON, A., JAMES, J. R., and FRAY, A.: 'Magnetised microstrip antenna with Air gap, adjustable, 193 Annular slot, 111,611,1086
pattern control' Electron. Lett., 1988, 24, pp. 45-47 Air-filled waveguide, 858 Annular slots
36 JAMES, J. R., HENDERSON, A,, and JOHNSON, R. H.: 'Compact electromagnetic appli- Airborne antenna, 1139 attenuation coefficient, 614
cators'. In HAND, J. W., and JAMES, J. R. (Eds.): 'Physical techniques in clinical hyperther- Airborne phased arrays, 802,810 guide wavelength, 611
mia' (Research Studies Press, John Wiley, 1986), pp. 149-209 Alumina substrates, 14, 1025 impedance, 611
37 JAMES, I. R., HALL, C. M., and ANDRASIC, G.: 'Angled dual compact hyperthemic Aluminium, 93 1, 946 reflector planes, 617
applicators', Proc. IEE, 1987, lMH, pp. 315-320 Aluminium foil, 1075 Antennas
38 BAHL, I. J., BHARTIA, P., and STUCHLY, S. S.: 'Design of microstrip antennas covered Amplitude distributions, 830 conical, 601
with a dielectric layer', IEEE Trans., 1982, AP-30, pp. 314-318 Analytical methods DBS,859,861
aperture radiation model, 583 dielectric, 1287
cavity model, 458,580,1058,1228 directly coupled three resonator, 507
Cohns method, 611 hybrid microstrip, 580
contour integral method, 472 integrated, 14
desegmentation, 456,494 gap coupled multiresonator, 507
direct form of network analogue, 462 millimeter wave hybrid, 1285
EMF,249 multi-terminal, 236
Filton Integration method, 285 open microstrip, 580
integral equations, 715 window, 1283
method of moments, 364,708 wraparound, 85
moments method, 593 Antenna location, received signal
multiport network approach, 455 fluctuations, 1084
multiport network model, 462 Antennas for portable equipment, 1092
network analysis, 488 Aperture admittance, 727
plane wave spectrum method, 285 Aperture blockage, 96, 104
real-space integration method, 286 Aperture coupled patches, 330
1298 lndex lndex 1299

Aperture coupling, 35, 332, 823 linear, 345,449 Attenuators, 1241 Brass, 946
Aperture distribution, 762,821, 822, 825, linear centre fed, 839 Axial polarisation, 1239 Brick wall arrays, 848
851 linearly polarised, 846 Axial ratio, 723, 787, 1185 Broadband microstrip antenna, 1083, 1086,
Aperture distribution Dolph-Chebyshev, log periodic, 39, 387 Axial-ratio bandwidth, 758,829, 859 1129,1134
855 microstrip, 443 Azimuthal modes, 53 Broadband matching, 558
Aperture radiation analysis method, 583 millimeter wave, 1282 Built in antennas, 1092
Apertures, dichroic, 1282 monopulse, 849, 1068, 1153 Butler matrix, 855, 860, 1083
Applications, 8 non uniform, 630 Backward firing beams,834
Array factor, 829 omnidirectional, 39, 382 Bag-moulding, 930
Array lattice parasitic, 801 Balanced stripline, 1003 Cabin antenna, 1085
rectangular, 706 parasitic patch, 825, 833 Bandwidth, 5, 11, 111, 128,219, 335, 338, CAD, 14, 1031
triangular, 706, 1243 parasitically coupled, 37, 345 554,743,796,895 Acline, 1033
Array structures, 36 patch, 1263 Bandwidth, 9 Analop, 1036
Arrays, 7,35 planar, 345, 776,790 axial-ratio, 829, 859 Autoart, 1037
angled slots, 39 rampart, 816, 844 extension, 9 CADEC, 1033
architecture, 745 rampart line, 38,223, 763, 1222 gain, 823 CiAO, 1035
asymmetric step, 39 recurrent sequential, 808 VSWR, 118,823 computer graphics, 1175
bonding, 916 resonant, 769,816 Base-station antenna, 1083 Esope, 1033
brick wall, 848 scanning, 792 Basis functions, 282 LINMIC, 1034
cascaded patch, 822 sector beam, 1081 entire domain, 429, 437 Mama, 1036
chain, 37, 763, 816 sequential, 759, 788, 805 Maxwell, 282 Micad, 1037
Chebyshev, 774 sequentially rotated, 12 Maxwell, modified, 282 Micpatch, 1035
circularly polarised, 755,765 series, 816 pulse, 282 Microkop/Suspend, 1036
co-planar, 758 series fed patch, 515 subdomain, 282,437 Micros, 1038
comb line, 37, 766, 822 series fed patches, 624 subsectional, 423,440 microstrip, 1001
composite element, 759,770 series-fed, 789, 8 18 Beam squint, 758 microwave software applications,
conformal, 795, 1153, 1227 series-fed circular polarised, 770 Beam steering, 1241, 1249, 1273 1036
conical, 1154 serpent, 816,833,843 Beam-forming, adaptive, 862 Midas, 1034
constant-conductance, 846 serpent line, 38 Beams Multimatch, 1036
corner fed patches, 647 sparse, 1273 backward firing, 834 photoplots, 1175
corporate-fed, 789 spherical, 34,793, 1239 forward firing, 834 Planim, 1036
crank, 844 square patches, 1073 Beamwidth, 104 S/Filsyn, 1037
crank-line, 777 square-looptype, 764 Bend measurement, 976 Supercompact, 1033
crank-type, 764 squintless, 758,841 Bending, 916 techniques, 517
cross fed, 36, 655, 843 strip dipole and slot, 39 Bends, 1168 Temcad, 1039
cross printed-dipole, 766 sub arrays, 655 Beryllia substrates, 1025 Touchstone, 1032
cross slot, 766 synthesis, 662 Bessel function, 1058, 1204 Transcad, 1037
cylindrical, 34,765, 1227 tapered, 650 Bidirectional communications, 1107 triplate, 1001
DBS, 766,790,1288 transposed, 821,822 Bismaleimide, 95 1 Fringing, 901
design of planar, 622 travelling wave, 755, 762, 765, 769, Bismaleimide-triazineepoxx878 Capacitance, 901
discontinuity, 843 816,1222 Black copper oxide, 880 Capacitive tuning, 329
dual frequency, circular polarised, 802 triangle line, 38 Blind spots, 336, 729 Capacitors, 1040
dual polarised, 1073 two dimensional, 808 Blockage, aperture, 96 Car telephones, 1079
electronically switched, spherical, 792 two sided, 746 Boards, printed wiring, 916 Carbide, 922
finite, 301, 731 untransposed, 821 Bolometer, 1194 Carbon fibre reinforced plastic, 1075
flat, circular polarised, 765 wide bandwidth, 372 Bonding Cavity feeds, 859
four element, 1217 wideband, 804 lead, 952 ~ a v i t model,
; 112,458, 580, 884, 1058,
Franklin, 8 16, 848 wire grid, 848 thermal-compression, 938 1228
Franklin line, 38 Assemblies, 916 thermosonic, 938 Cavity-perturbation, 884
herringbone, 835 Asymmetric feeds, 797 wire, ultrasonic systems, 938 CCIR-TVRO conditions, 671
herringbone line, 38 Asymmetric step array, 39 Boundary conditions, 49, 115,395 Central shorting pin, 85
infinite, 698, 731 Attachment mode, 433 Branch-line coupler, 854 Centred fed dipole, 287
lattice, 848 Attenuation, 1016 Branching network, 825 Ceramic-PTFE, 953
 1300 lndex
lndex 1301
Chain array, 37, 763, 816, 940 Compensating hole transitions, 967
Chain scission, 940 Components, 1039 probe, 822 Diagonal slot square patch, 499
Chebyschev taper, 774, 1069 Composite element, 755, 759, 770 proximity, 8 18 Diagonally fed nearly square patch, 499
Chloride Computational efficiency, 709 Coupling gaps, 762 Dielectric
acid cupric, 949 Computer graphics, 1175 Coupling mechanisms, 816 antennas, 1287
alkaline cupric, 949 Conductance Crank arrays, 764,777,844,1115 losses, 788,815, 1016
ferric, 949 edge, 476 Cross-printed dipole arrays, 766 Dielectric constant, 790, 897
Choke, peripheral, 104 radiation, 476, 554, 822, 835, 859 Cross fed arrays, 36,655,843 Dielectric constant, effective, 481
Chokes, 97 surface wave, 476 Cross patch, 501 Dielectric filling factor, 790
Chromic-acid, 950 Conductivity, 879 Cross polarisation, 68,74,76, 100, 104, Dielectric image guide, 822, 858
Circuit, etching, 1028 Conductor, losses, 788, 790, 815, 1075 590,1234 Dielectric loss, 1075
Circular array, 1127 Conformal antenna, 1227 Cross slot arrays, 766, 1136 Dielectric rod array, 39
Circular array feed, 1132 Conformal arrays, 795, 1153 Crossed dipoles, 356 Dielectric spheres, 3 1
Circular array of slots, 1132 Conformal mapping, 1008 Crossed slot, 28 Difference radiation pattem, 1068
Circular array of strips, 1132 Conical antennas, 601 Crossover level, 793 Diffraction coefficient, 1069
Circular microstrip discs, 1136 Conical arrays, 1154 Crystalline, 873, 878, 942 Diffraction effects, 1069
Circular patches, 45, 63, 111, 720, 1058, Conical beam, 1112, 1127, 1129, 1132, Current Diodes, 1041
1202 1239 ! distributions, 112 Dipole
Circular polarisation, 5, 130, 219, 722, 744, Conically depressed patch, 29 magnetic, 726 centre fed, 287
755,787, 846,848, 1127, 1132, Connections, 332 sources, 112 flat folded, 765
1261,1273 Connector characterisation, 962 Current sheet model, 731 horizontal electric, 276,403
Circular polarised radiation, 824 Connector test fixture, 963 Current-ribbon, 151 infinitesimal, 276
Circular polariser, 766 Connectors, microstrip to coax, 1007 Cyanide, copper, 936 printed, 765
Circulardisc patches, 756 Constant-conductancearrays, 846 Cycling, wet/dry, 941 circularly polarised, 356
Circular-patch array, 1113 Contour integral method, 472 Cylindrical antenna, 1227 crossed, 356
Circular-patch-slot array, 1119 Coplanar line probes, 968 Cylindrical array, 1109 efficiency, 367
Circularly polarised, singly fed patches, 221 Coplanar stripline patch, 31 active-element, 1249 EMC, 295
Circularly polarised arrays, 755,765 Copolymer substrates, 679 array feed, 1248 flat, 353
Circularly polarised circular patches, 232 Corporate feeds, 301 design, 1240 multiple, 299
Circularly polarised composite type patches, Copper electroless, 935 gain, 1248 mutual impedance, 359
222 Copper foil, 1075 impedance, 1243 parasitic, 759
Circularly polarised dipoles, 356 Copper-Invar-copper, 946 mutual coupling, 1244 polarisation, 361
Circularly polarised elements, 12 Comer fed patches, 647 pattem synthesis, 1251 printed, 706,732
Circularly polarised line antennas, 762 Comer reflector, 1082 scanning, 1249 stacked, 299
Circularly polarised patches, 27,499,821, Corporate feeds, 13, 35, 757,789, 816, Cylindrical modes, 1230 strip, 761
1218 850, 1069,1288 Cylindrical near field scanning, 981 synthesis, 361
Circulators, 1041 Correction, end-fringing, 888 Cylindrical patches, 1227 variable directivity, 372
Circumferential polarisation, 1234 Correlation, 905 Cylindrical wave propagation, 1072 Direct coupling, 822
Co-axial feeds, 276 Corrugated ground plane, 28 Direct form of network analogue, 462
Co-planar arrays, 758 Cosecant squared pattern, 671 Direct-BroadcastSatellite, 1288
Co-planar coupling, 817 Coupled triplate lines, 1169 Dantzig algorithm, 662 Directional coupler
Co-planar feeds, 8 15 Coupled-resonator, 340 Data-relay satellite, 1146 hybrid-ring, 855
Co-polymers, 945 Coupler DBS antennas, 766,790,859, 1112 rat-race, 855
Coating, conformal, 950 branch-line, 854 De-smearing, 929 Directivity, 127, 372, 554, 831
Coaxial excitation, 432 hybrid-ring, 854 Decomposition, thermal, 919 Directivity, linear arrays, 625
Coaxial probe, 29, 433, 1194 Coupling, 817, 887, 895, 897 Deformation, 924 Directly coupled three resonator antenna
Cohns method, 61 1 Coupling Degenerate modes, 756 disc,25,507,843,1258
Collected volatile condensable materials agents, 875 Dendrites, 879 Discontinuity arrays, 843
(CVCM), 940 aperture, 823 Dents, 917 Discontinuity radiation, 791
Colloidal, 88 1 co-planar, 8 17 Desegmentation method, 456,494 Dispersion, microstrip, 1015
Comb array, 37, 766, 822, 833 direct, 822 Design procedure, series-feed, 836 Dissipation, loss, 1285
Combined feeds, 839 electromagnetic, 824 Device attachment, 936 Dissipation factor, 871, 878, 886, 895
Comparator, monopulse, 859 factor, 835 Diagnostics, 1159,1193.1214 Distribution
Diagnostics, liquid crystal, 984 amplitude, 830
1302 lndex lndex 1303

aperture, 825 Electro-etch, 879 Far-field approximations, 408 Flush mounted antennas, 1092
Dolph-Chebyshev, 830 Electromagneticcoupling, 207, 797, 824 Feed isolation, 320 Foam, 953
Taylor, 830 Electronically switched spherical arrays, Feed structures, 35 Foils
Dolph Chebyshev distributions, 830, 855 792 Feeder of a cylindrical array, 1248 adhesion, 875
Dose rate, 940 Electroplating, 935 Feeds, 32, 756,767, 1001 aluminium, 1075
Double tuning, 1064 Electrostatic charging, 1065 3 dB hybrid, 220 copper, 1075
Doubly diffracted field, 1069 Element architecture, 13 electrodeposited, 879
Drills, 922 endfire, 749 asymmetric, 797 rolled, 879
Dual aperture-fed patches, 823 tapered-slot, 751 cavity, 859 wrought, 879
Dual feeds, 219 Element factor, 831 co-axial, 276 Folded slots, 353
Dual frequency circularly polarised arrays, Element grid, 1241 co-planer, 8 15 Forming rolls, 924
802 Element pattern, 732 coaxial probe, 29 Forward firing beams, 834
Dual ijarised array, 1073 Elliptical patch, 25, 182,235,756, 1129 combined, 839 Four element arrays, 1217
Dual-fed circularly polarised patches, 220 Elliptical polarisation, 186 corporate, 13, 35, 306, 757, 816, 850, Four element sub-arrays, 805
Dual-frequency patches, 30, 188, 197,200, EM-field, far-zone, 117 1069 Four point feeding, 28
312,313,796,802 EMC dipole, 295 corporate, triplate, 1288 Four-probe feeds, 756
Dual-polarisation patches, 3 12, 3 18 EMF method, 249 dual, 219 Fourier Transform, 403,406,708, 716,
Ductility, 879, 936 End fringing, 887 four point, 28 721,726
Dust protection shield, 1063 Endfire arrays, 1279 four-probe, 756 Franklin array, 38, 816, 848, 1105
Dyadic Green's function, 284,399 Endfire elements, 749 Lecher line, 353 Free radical, 941
Energy, confonnational, 88 1 microstrip line, 29 Frequency, resonant, 895
Entiredomain basis functions, 429, 437 novel, 12 Frequency agility, 187
E-H antenna, 1083 Environmental conditions, 875 Frequency diversity, 1084
overlaid, 1263
Earth stations, 1112 Environmental effects, 679 Fringing fields, 113
parallel, 335, 757, 816, 825, 828
Edge Epoxy, 878,951 perpendicular, 747 Full sheet resonance test method, 884,897,
conductance, 476 Equitriangular patches, 111 90 1
phased array, 862,1248
effects, 1069 Equivalence Fumes, 919
radial waveguide, 859
equivalent admittance network, 463 Functions
external, 47 . rear, 758
ground plane, 12, 731 internal, 49 testing, 53
reflector, 96
non radiating, 481 Equivalent sources, 112 resonant, 835,1074 triangular, 54
Effective dielectric constant, 481
Equivalent circuit, 237, 728 sequentially rotated, 36 Future prospects, 1
Effective loss tangent, 461 Eauivalent circuit model., 769.. 776 series, 335,758,767,816,832
Effective permittivity two-layer medium,
Equivalent current sources, 114 series compensated, 36
196
Eauivalent edne admittance network. 463 single, 219 GaAs substrate, 332
Effective radius, 65, 137 ~quivalentmagnetic current, 116 GaAs superstrate, 281
spurious-radiation, 332
Effective width, 115
Equivalent slot, 534 squintless, 832 GaAs transitions, 968
Effects, environmental, 939
Equivalent surface currents, 49 stripline, 353 Gain, 127, 554, 831
Efficiency, 5,13, 117,313,346,367,413,
Equivalent waveguide model, 817 travelling-wave, 832 minimum coverage, 793
554,781,837,1258
Etch two-dimensional, 839 bandwidth, 823
Efficiency
plasma, 929 two-line, 276 factor, 104
computational, 709
sodium, 929 Ferrimagnetic substrates, 1027 Gain of a cylindrical array, 1248
measurement, 991
Etchant, 949 Ferrite superstrates, 1292 Galerkin solution, 284
radiation, 9, 598, 734, 872
Etching, 1028 Fibres, glass, 922 Gap coupled multiresonator antenna, 507
spill-over, 104
Eulers constant, 474 Field, excitation, 396 Gap-wupled patch, 8 17
Elastic, 924
Excitation, coaxial, 432 Filler, ceramic, 922 Geometric optics field, 1069
Electromagnetically coupled patch, 31
Excitation field, 396 Films, barrier, 930 Giotto spacecraft, 1061
Electric current analysis method, 583
Excitation voltage, 784 Filton integration method, 285 Glass transition temperature, 875
Electric dipole, 403 Expansion functions, 53 Finite arrays, 301, 731 Glossary, 24
Electric shielding ring, 803
External equivalence, 47 Finite ground plane, 12, 1069 Grain structure, 879
Electric source, 1260
External matching circuits, 28 Flat dipoles, 353 Grating lobes, 428, 826, 834, 1183
Electric walls, 113
Flat folded dipole, 765 Green's function, 50, 116, 398, 41 1, 421,
Electric-current source method, 762,765
Fabric, 922 Floquet modes, 703 426,462,695,726
Electric-field integral equation, 715, 726,
woven-glass, 935 Flouborate, copper, 936 dyadic, 399
732
1304 Index lndex 1305

planar configurations, 519 Incoherent radar, 1073 Linear centre fed arrays, 839 Materials, magnetic, 1292
spectral domain, 327 Indoor communications antenna, 1083 Linear polarisation, 130 Mathematical modelling, 14
Ground plane Indoor receiving antenna, 1084 Linear slots, 606 Matrix
corrugated, 28 Inductance feed probe, 329 Linearly polarised arrays, 846 Butler, 861
slot, 32 Inductors, 1041 Lines excitation, 58
edge, 12,731 Infinite arrays, 698, 731 dielectric, 1285 formulation, 50
GTD, 1183 Infinite phased arrays, 698 discontinuities, 762, 1017 impedance, 708,716
Guide Infinitesimal dipole, 276 losses, 624 Maxson-Blass, 861
dielectric image, 822 Infra-red, 875 open circuited, 1212 Maxon Blass matrix, 860
insular, 822 Input parallelaupled, 856 Maxwell basis functions, 282
Gunn diode. 34 impedance, 9, 118,238,432,439, parameters, 971 Maxwell, modified basis functions, 282
708,875 parameter measurement, 970 Measurements, 957, 1006
admittance, 770 synthesis, 1015 bends, 976
H-shaped patch, 26 conductance, 554 width, 943 efficiency, 991
Half-wave patch, 313, 817 Input resistance o f patches, 324 Liquid crystal diagnostics, 984, 1058, 1159 line parameters, 970
Hand held message communication Inserted connector transitions, 967 Loading, reactive, 204, 826 radiometric, 993
terminal, 1125 Insular guide, 822 Lobes, grating, 826, 834 resonant techniques, 976
Handbook, 17 Integral equation, 47 Log periodic arrays, 39,387,834, 1273 T-junctions, 977
Hankel Transform, 406 Integral equation, electric-field, 715, 726, Longitudinal polarisation, 822, 846 Melt point, crystalline, 879
Hankel function, 264, 473, 1230 732 Lorentz's gauge, 398 Melt viscosity, 928
Hard boundary diffraction coefficient, 1069 Integrated antennas, 14 Loss tangent, 117, 461 Metal failure, 934
Herringbone array, 38, 835 Integrated phased arrays, 742 Losses, 174,407 Metallic ring transitions, 967
Higher order modes, 356,787,796, 1058, Internal equivalence, 49 conductor, 788,815,942,1075 Method of moments, 282,295,364,401,
1129 Iris, 906 dielectric, 1016, 1075 423,708
Historical development, 1 Isolated power dividers, 852 dissipation, 1285 Metrology, 1193
History Isolators, 1041 line, 624 Microstrip
mechanical, 874 ohmic, 1016 circuit realisation, 1028
thermal, 874 radiation, 1016, 1075 impedance, 1013
Holes, burr-free, 922 Junction effects, 1166 reflection, 8 16 line, 1004
Honeycomb substrate, 796 resistance, 887 materials and manufacture, 1023
Horizontal dipole, 403 resistive, 895 Microstrip antenna
Humidity, 875 K connector transitions, 967 surface wave, 816, 1075 hybrid, 25
Huygens sources, 1262 K', thermal coefficient of, 947 Low cost substrates, 674 frequency variable, 1092, 1101
Hybrid coupler, 1064,1081,1164 Kevlar epoxy, 1075 post loaded, 1O92
Hybrid microstrip antenna, 25, 580 quarter wavelength, 1092
Hybrid sources, 353 Machining, 916 window attached, 1092
Hybrid-ring coupler, 854, 855 Land mobile satellite communications, Magnetic Microstrip dispersion, 1015
Hydrocarbon, 951 1127 current, 116,726,762 Microstrip field diagnostics, 1193
Hydrolysis, 875 Lattice arrays, 848 materials, 1292 Microstrip line, 537
Hydrophobic, 941 Launchers, 1287 source, 1260 Microstrip line feeds, 29
Hyperthermia applicator, 1293 Layer, resistive, 947 walls, 113 Millimetre wave hybrid antennas, 1285
Leaky cavity, 119 Main beam direction, 781 Minimum coverage gain, 793
Lecher line feeds, 353 Mandrel, 931 Mismatch thermal, 953
Image, 116 Light, ultra-violet, 929 Manpack radars, 1079 Mixed potential integral equation, 400
Impedance Limiting oxygen index, 919 Manufacture, 5,14 Mobile communications base stations, 1081
active, 732 Line analysis Marine radars, 1079 Mobile communications antenna, 1083
cylindrical array, 1243 dielectric Green's function, 1020 Maritime satellite communications, 1127 Mobile satellite communications, 800, 1142
input, 9, 118,439, 590, 708 Fourier transform, 1020 Matched terminations, 964 Mobile systems, 1079
matching, 655, 1190 integral equations, 1020 Matching, 5 Modal expansional method, 235
matrix, 708,716 TEM models, 1020 Matching circuits Mode ambiguous, 901
port matrix, 737,739 variational techniques, 1020 external, 28 Modelling mathematical, 14
surface, 400 Line antenna, circularly polarised, 762 gaps, 28 Modelling accuracy, 16
transformer, 1064 Linear arrays, 345,449 Mated connector test method, 963 Modes
 1306 lndex
lndex 1307
azimuthal, 53 Off centred pin transitions, 967
degenerate, 756 dual-fed, 220 triangular, 25, 235, 1209
Offset fed oatches. 221
higher order, 356, 787, 796, 1058 dual-frequency, 30, 188, 197, 200, truncated comer, 27
Ohmic lo-, 101'6
orthogonal, 221 312,313,796,802 truncated corner square, 499
Omnidirectional arrays, 382, 765
parallel-plate, 8 16 dual-polarisation, 3 12, 3 18 two port, 51 1
Open circuit end, 534 efficiency, 554
resonant, 900 Open circuited lines, 1212 wide, bandwidth, 320
suppressing pins, 1065 electromagnetically coupled, 31,207, wideband, 28, 796
Open microstrip antennas, 580
transverse electric field, 887 797 Pattern, active element, 739
Operational factors, 5 elliptical, 25, 182, 235, 756
unwanted, 258 Optical modulator probe, 983 Pattern synthesis of cylindrical arrays, 1251
Modularity, 744 equitriangular, 111 Peel strength, 937
Optically tuned patches, 192 gain, 554
Modulus, 953 Peel-test, 878
Orthogonal fields, 787 Green's functions, 462
Moisture, 875,941 Orthogonal polarisation, 30 Pentagon patches, 25,505,756, 1218
Moments method, 593 H shaped, 26 Performance trade-offs, 7
Overlaid feeds, 1263 half-wavelength, 313
Monopole probes, 1197 Overlaid patches, 35, 1277 Peripheral choke, 104
Monopulse arrays, 849, 1068, 1153 input conductance, 554 Permittivity
Monopulse comparator, 859 input resistance, 324 complex, 872
Moulding vacuum-bag, 926 multiple feed points, 262 effective, 48 1
Pagers, 1092
Multi-terminal antenna, 236 multiple frequency, 320 relative, 871, 878, 881, 886, 895
Paired elements, 263,270, 758, 788, 804
Multibeam antenna, 1083, 1146 multiresonator, 507 very-high, 1293
Parallel feeds, 335, 757, 816, 825, 828
Multilayer substrates, 679,944 mutual impedance, 378 Perpendicular feeds, 747
Parallel date resonator test method, 959 notched, 27,796
Multipath fading, 1096 ~arallel-bu~led lines, 856 Persulfate, 949
Multiple beam-forming networks, 817, 859 offset fed, 221 Perturbation segment, 224
Parallel-plate modes, 8 16
Multiple dipoles, 299 optically tuned, 192 Perturbation cavity, 914
Parallel-plate polariser, 766
Multiple feed point patches, 262 overlaid, 1277 Phase centre, 96, 106, 1161
Parallel-plate waveguide, 858
Multiple frequency patches, 320 paired, 263, 270, 804 Phase constant, 779
~arasiticrarra~s, 80i
Multiple layer patches, 30 parasitic, 214, 797 Phase shifters, 864, 1081,1241
Parasitic patches, 29,214,264, 797, 825,
Multiple tuning, 341 pentagon, 756 Phase shifters, Schiffmann, 848
833,1083,1086,1129 pentagonal, 25,505, 1218
Multiport network approach, 455 Parasitically coupled array, 37 Phased arrays, 378, 741, 802, 810, 862,
Multiresonator patch, 507 piggy-back, 313 1241
Passivate, 950
Mutual coupling, 249, 306, 337.445. 561. polarisation, 378 Phased arrays
Patch arrays, 822, 1263
post-tuned, 3 15 infinite, 698
Patches
probe-fed, 713,720 integrated, 742
annular ring, 25, 111
Mutual coupling in a cylindrical array, 1244 quarter wavelength, 313, 1154 Photolithographic techniques, 1002
apertureaupled, 330, 723 rectangular, 111, 224,235,436, 553,
Mutual coupling network, 464,482 bandwidth, 111, 554 Photomask, 888
Mutual impedance, 237,291,359,378 580, 1215,1234 adhesion, 9 18
cavity model, 1228 rectangular, comer-fed, 756
circular, 45,63, 11I, 232, 580, 720, Photoplots, 1175
rectangular ring, 26 Photoresist, 1029
1058, 1202 resonant frequency, 324
Narrow pin transitions, 967 circulardisc, 756 Piggy-back patches, 313
Near field mapping 989 short circuit, 374 Pinholes, 880
circularly polarised, 27,499, 755, short circuited, 353
Near-field probes, 98 1 821, 1218 Pits, 917
Network analysis techniques, 488 short circuited ring, 346 Planar arrays, 345, 776, 790
composite type, 222
Networks shorted, 1281 Planar near field scanning, 981
conically depressed, 29
multiple-beam-forming, 817 singly fed, 221 Planar segments, characterisation by
coolanar strioline, 3 1 slotted, 27
special-purpose, 859 corner fed, 847 Z matrix, 467
Nitrogen, 930 square, 25 Plane wave spectrum method, 285
cross, 501 square ring, 501
Nodules, 879 cylindrical, 1227 Plastic substrates, 14, 1025
Non radiating edge admittance network, stacked, 29,320 Platen-press, 929
design, 557 stacked circular-disc, 197
482 diagonal slot square, 499 Plating holes, 916
Non radiating edge characterisation, 481 star, 26 Point dipole approximation, 287
diaaonallv fed nearly square, 499 stepped, 29
Non uniform arrays, 630 directivity, 554 Poisson sum formula, 699
Notched patch, 27,796 strip line, 111 Polarisation, 5, 743, 783
disc, 1258
Numerical analysis, 16 surface-current model, 1232 Polarisation
dual aperture-fed, 823
Numerical techniques, 417 thick, 253 45 deg, 821
dual band circularly polarised, 1061 tilted slot, 756 axial, 1239
 1308 lndex lndex 1309

circular, 5, 130,219, 722, 744, 755, wafer, 968 1 553, 1215, 1234 Scan angle, 785
846,848,1261, 1273 Processing, 916 Rectangular ring patch, 26 Scan blindness, 700,709,719,723,731
circumferential, 1234 Proximity coupling, 8 18 Rectangular-slot array, 1118 Scan range, 743
cross, 68, 74,76, 100, 104,590, 1234 F'TFE, 873,881,919 Reflect array, 33 Scanning a cylindrical array, 1249
ellipse, 1187 ceramic, 884 Reflection coefficient active, 708, 739 Scanning arrays, 792
elliptical, 186 glass fibre, 883 Reflection losses, 816 Scanning losses, 792
ellipticity, 231 woven glass, 884 Reflector feeds, 96, 1258 Scanning network probes, 1195
linear, 130 Pulse basis functions, 282 Relaxation synthesis, 662 Scattering matrix, 702, 783
longitudinal, 822, 846 Pyrophosphate, copper, 936 Residue, 278 Scatterometer, 1073
orthogonal, 30 Resin Schartz-Christoffel transform, 1011
tracking, 755 poly(tetraflouroethy1ene)(PTFE), 879 Schiffmann phase shifters, 848
transverse, 82 1, 846 Q factor, 111, 119, 128, 174, 324, 590, polycyanate, 878 Schotky barrier diode, 1194
Poles, surface wave, 697 593,796,887,895 polyetherimide, 879 Secondary su~eillanceradar, 1068
Polyethylene, 873 Quarter wave resonance, 353 polyethersulfone, 879 Sector beam array, 1081
Polymer fume fever, 919 Quarter-wave patch, 3 13, 8 17 polyimide, 878 Sector patterns, 664
Polymer systems, thermoset, 878 Quasi-log-periodic, 1279 polystyrene, 878 Segmentation method, 456,488
Polymerisation, 88 1 polysulfone, 879 Semi infinite substrate, 731
Polymides, 951 triazine, 878 Semiconductor substrates, 1027
Polypropylene substrates, 678 Radar, 1105 Resin laminates, 929 Sequential arrays, 12,36,263, 788, 805,
Polypropylene-ethylenesubstrates, 679 Radar reflector, 1107 Resistance, radiation, 413 828,859,1142,1265
Polytetraflouroethylene, 873 Radial waveguide feeds, 859 Resistive box terminations, 964 Series arrays, 816
Post-tuned patches, 3 15 radiated, electric field, 41 1 Resistive layer, 950 Series compensated feeds, 36
Posts, 235 Radiation Resistivity, surface, 948 Series fed patch arrays, 5 15
Potential, 41 1 circular polarised, 824 Resistors, 1040 Series feeds, 335,758,767,816,832,789,
diffracted field, 397 conductance, 8 17 Resonant arrays, 769,816 818
scalar, 399, 405 cosmic, 940 Resonant cavity, 111 Series-fed circularly polarised arrays, 770
vector, 50, 399,403,405 damage, 941 Resonant feed networks, 835, 1074 Series-feed design procedure, 836
Power combiners, 8 16 dose, 941 Resonant frequency, 115,324,593 Serpent array, 38,816, 833, 843
Power dividers efficiency, 9, 126, 598, 734 Resonant peak, 887 Shear, 937
isolated, 852 exposure, 939 Resonant ring test method, 961 Shipbourne antenna, 1136
rat race, 1069 feeds, 12 Resonant-mode, 117 Shock, thermal, 934
split-tee, 852 fields, 112 Resonator Short circuit patches, 25,346,353, 374,
three-port, 852 high-energy, 940 microstrip, 914 1087,1281
Wilkinson, 307, 852, 947, 1069 losses, 788,791, 1016, 1075 stripline, 914 Short monopole probe, 982
in-line, 850 nuclear, 940 Resonator strip test method, 960 Shorting pin, 85
Precision, 891 patterns, 122, 335, 449 Richmonds reaction equation, 597 Sidelobe level, 5, 774
Press, platen, 926 resistance, 413 Rotation, sequential, 12,263, 828, 859 Simplex synthesis, 662
Pressure vessel, 929 slot, 222 Rotman lenses, 860 Single feeds, 219
Printed dipoles, 706,732,765 spurious, 353,743,816,824 Routers, 922 Singly diffracted field, 1069
Probe coupling, 822 ultra-violet, 941 Rutile substrates, 1025 Skin effect, 943
Probe-fed patches, 71 3 unwanted, 787 Slot antenna, 1136
Probes Radiometric measurement, 993 Slot combiner, 1084
coaxial, 433, 1194 Radius, effective, 65 S-matrix, 1240 Slotline transitions, 962
coplanar line, 968 Radome, 14,926,945, 1273 Safety, 916, 919 Slots
errors, 1200 Railway antennas, 1087 Sapphire substrates, 1025 annular, 6 11
monopole, 1197 Rampart array, 38, 763, 816, 844, 1222 Satellite crossed, 28
near-field, 981 Rat race hybrid, 855,1069 ERS-I, 1073 folded, 353
optical modulator, 983 Reactance compensation, 761,823 ETS-V, 1112 linear, 606
scanning network, 1195 Real space integration method, 286 NAVSTAR (GPS),1124 Slotted line measurements, 975
short monopole, 982 Rear feeds, 758 antennas, 1146 Slotted patch, 27
small loop, 982 Reciprocity method, 278 communications, 1136 Small loop probe, 982
split coaxial balun, 982 Rectangular array lattice, 706 systems, 1079 Smear, 923,934
square law, 983 Rectangular patch, 7,25, 111, 235,436, Scalar potential, 399, 405 Sodium-bisulphite, 950
 1310 lndex lndex 1311

Solder reflow, 938 St~ctWes resistivity, 117 mated connector, 963

Solvents, 875 array, 36 tangent, 52 microstripresonator, 893
Sommerfeld equation, 276,285,407,417 feed, 35 wave, 406,407,4 10 parallel plate resonator, 959
Sources, hybrid, 353 Subarrays Surface analytical techniques, 1194 resonant ring, 961
Space diversity, 939, 1084 2*2,759 Surface current model, 1228 resonant-cavity perturbation, 906
Sparse arrays, 1273 four element, 805 Surface currents, equivalent, 49 resonator strip, 960
Spatial phase delay, 805 two element, 804 Surface field metrology, 1193 stripline-resonator, 882
Special-purpose networks, 859 Subdomain basis functions, 282,437 Surface impedance, 400 stripline resonator test, 886
Specimen, test, 909,907 Subsectional basis functions, 423,440 Surface resistivity, 815 Testing functions, 53,427
Spearal domain method, 265,327,404, Substrates, 15 Surface treatment, 879,929 Thermal paint, 1065
590 alumina, 14, 1025 Surface wave conductance, 476 Thermalexpansion coefficient, 878, 936
Spectral domain Green's function, 327 bending, 1177 Surface wave losses, 816 Thermoplastic, 927
Spherical arrays, 34, 792, 793, 1134, 1239 beryllia, 1025 Surface wave poles, 593 Thermoset, 927,941
Spherical dielectric overlays, 1267 copolymer, 679 Surface waves, 9, 12, 592, 600, 703, 728, Thick film, 1031
Spherical near field scanning, 981 effects, 279 731,734,740,751 Thick patches, 253
Spill-over efficiency, 104 environmental effects, 679 Surface waves Thick substrate, 28,275
Spiral, 30 ferrimagnetic, 1027 circle diagram, 706 Thin film pad terminations, 964
Spiral slot, 33 foam, 1288 excitation, 116 Three-faced array, 1111
Split coaxial balun probe, 982 GaAs, 332 loss, 1075 Three-port power dividers, 852
Split-tee power dividers, 852 honeycomb, 796 poles, 697 Throwing power, 936
Splitters, Wilkinson, 356 low cost, low loss, 674 excitation, 127 Tilt angle, 1188
Spurious radiation, 332, 353, 743,816, 824 materials, 871 Surveillance radars, 1079 Time-domain-reflectometry,979
Square law probe, 983 measurements, 958 Susceptance, edge, 479 TM210-mode microstrip antenna, 1129
Square patch, 25 metal deposition, 1028 Suspended stripline, 816, 823,857 Total mass loss (TML), 940
Square patch array, 1073, 1115 metallisation, 1027 Suspension, 881 Toxity, 919
Square ring patch, 501 multilayer, 679 Switchedelement spherical array, 1132 Tracking slope, 1188
Square-loop-type arrays, 764 non-woven glass-PTFE, 892 Synthesis, 361 Train antenna, 1087
Squintless arrays, 758, 826, 832, 841 perpendicular, 337 Synthesis, lines, 1015 Transistors, 1041
Stacked antenna, 312 plastic, 14, 1025 Synthesis methods, 662 Transition, 942
Stacked circulardisc patches, 197 polypropylene, 678 Synthetic aperture radar antenna, 790, Transition, crystalline, 881
Stacked dipoles, 299 polypropyleneethylene, 679 1146 Transition phase, 875
Stacked patches, 29,320 mtile, 1025 Transitions, 1288
Stainless-steel, 946 sapphire, 1025 compensating hole, 967
Standing wave distribution, 887, 1199 semi infinite, 731 T junction, 977, 850, 1166 GaAs, 967
Star patch, 26 semiconductor, 1027 Tab patches, 30 inserted connector, 967
Steepest descent method, 278,408 technology, 14 Tapered absorbing film terminations, 964 K connector, 967
Stepped patch, 29 thick, 28, 113,275, 356 Tapered arrays, 650 metallic ring, 967
Strain, internal, 920 thickness, 74, 791 Taperedslot elements, 751 narrow pin, 967
Strain relief, 920 thick metal backed, 678 Taylor distribution, 830 off centred pin, 967
Stress riser, 934 thin, 112 Teflon superstrate, 281 slotline, 962
Strip combiner, 1084 Sum radiation pattern, 1068 Telemetry, 924 stripline, 962
Strip conductors, 275 Summary, of chapters, 21 Temperature, 875,941 thermal, 881
Strip dipole, 761, 1132 Summary, of topic areas, 18 transition, 921 waveguide, 962
Strip line patches, 111 Superstrate, 275, 597 variation of K' with, 942 Transmission line analysis, 295, 527
Strip slot, 1132 GaAs, 281 Terminations Transmission line model, 112,317, 456,
Stripline, balanced, 1003 teflon, 28 1 matched, 964 769
Stripline Superstrate effects, 281 resistive box, 964 Transmission lines, 1001
suspended, 8 16,823,857 Surface tapered absorbing film, 964 Transmission-line matrix method, 1023
thickness, 1011 charge, 1196 thin film pad, 964 Transportable earth station, 1125
triplate, 816, 823 current density, 1196 Test method Transposed arrays, 821
feeds, 353 currents, 441 evaluation, 914 Transverse polarisation, 821, 846
transitions, 962 fields, 67 fluiddisplacement, 883 Travelling wave arrays,' 13, 762, 765, 769,
resonator, 884 gradient, 52 full-sheet-resonance.897 816, 1222
1312 Index

Travelling wave feeds, 832 electric, 113

Travelling-wave array design, 777 magnetic, 113
Triangle line array, 38 Waveguide, 906
Triangular array lattice, 706, 1243 air-filled, 858
Triangular functions, 54 dielectric, 944
Triangular patch, 25,235,1209 model, 817, 1023
Triazene, 95 1 parallel-plate, 858,897
Triplate, 857 simulator, 711,719
Triplate, balanced, 1003 transitions, 962
Triplate CAD, 1001 Waves
Truncated corner patch, 27 cylindrical, 1072
Truncated corner square patch, 499 surface, 406,407,410,592,600,703,
Tuned circuit, 9 73 1,734,740,75 1
Two element sub-arrays, 804 unwanted, 762
Two port patch, 511 Weathering, 941
Two sided arrays, 746 Welding, 937
Two-dimensional feeds, 839 electron-beam, 938
Two-line feeds, 276 laser, 938
Two-sided configuration, 332 parallel gap, 938
percussive arc, 938
resistance, 938
UHF pagers, 1079 ultrasonic, 938
Uniform lines, 1006 Wettability, 935
Untransposed arrays, 821 Wheeled vehicles, 1085
Unwanted modes, 258 Wicking, 875,941
Unwanted radiation, 787 Wide bandwidth arrays, 372
Unwanted waves, 762 Wide bandwidth patches, 28, 320,796
Urban mobile communications, 1086 Widebandwidth, 324, 1273
Wideband arrays, 804
Wideband baluns, 968
Vacuum outgassing, 939 Wideband techniques, 253
Vacuum-bag, 929 Wiener-Hopf method, 478
Variational method, 235 Wilkinson power dividers, 307, 356, 852,
Vector potential, 50, 399 1069
Vehicle antennas, 1086 Window antennas, 1283
Very-high permittivity, 1293 Wire grid arrays, 848
Vim, 945 Wrap around antenna, 34,85
plated-through hole, 953
reliability of, 953
Visw-elastic, 873,920,924 X-ray, 933
Voids, 935
Voltage maxima, 88
Voltage vector, 716 Z matrix
VSWR bandwidth, arbitrary segments, 472
circular segments, 471
from Greens functions, 468
Wafer probing, 968 planer segments, 467
Walls rectangular segments, 469