Anda di halaman 1dari 4

PROCEEDINGS OF THE 2008 IEEE INTERNATIONAL CONFERENCE ON ULTRA-WIDEBAND (ICUWB2008), VOL.

Differential Time Delay Patterns


for UWB Antennas
James S. McLean, Heinrich Foltz, and Robert Sutton

2
 +∞ 
Abstract—The time delay pattern for a UWB antenna is  ( E ( R, θ , φ , t ) ⋅ a (θ , φ )T (t − τ )) Rdt 
considered. Many different definitions of time delay are possible.
  (1)
Possibilities include (a) time delay for maximum correlation with ρ (θ , φ ) = max  − ∞+ ∞ +∞ 
a template function, either in terms radiated field or port-to-port τ 2
 
 − ∞ − ∞
2
transfer function, (b) time delays extracted from the linear phase T (t ) dt E ( R , θ , φ , t ) R 2
dt

part of a transfer function, and (c) time delays based on time-  
G
domain waveform features. Each of these three general where E ( R,θ , φ , t ) is the radiated electric field and
definitions may be computed by multiple means. The delays can
G
be absolute or differential, that is, taken relative to the respective a (θ , φ ) T ( t ) is the template function.
delay in a reference direction. It is shown that some of
definitions are equal under certain very restrictive conditions.
Example data is given for a broadband (1-18 GHz) double ridged As has been noted previously, there is ambiguity in the
horn antenna. determination of this quantity. First, the template
G
a (θ , φ ) T ( t ) to which the radiated field is correlated is not
Index Terms—antenna measurements, antenna radiation
specified and clearly many different situations calling for
patterns, time delay pattern, ultra-wideband antennas.
different templates can be envisioned. For example, in
describing antennas designed for short pulse reproduction
I. INTRODUCTION applications, it might be useful to use the time derivative of
the input voltage or current as the template. For
S EVERAL radiation pattern descriptors appropriate for UWB
antennas have been published including energy gain,
correlation coefficient, and correlated energy gain [1-9].
communications antennas, on the other hand, a useful template
might be the radiated field in a specified direction, such as the
bore-sight direction. This template gives some idea of how the
Here, we present a somewhat different descriptor, the
antenna will perform with fixed equalization, since fixed
differential time-delay pattern, which is computed using two
equalization can compensate only for one specific pulse
diverse approaches. Under strict conditions the differential
distortion.
delay patterns computed using these two approaches would be
similar. Via Parseval’s Theorem, the correlation may be carried out in
either the time or frequency domain, although truncation of
For a given antenna, the extent to which these two patterns the data will cause the two approaches to produce slightly
differ from one another gives some indication of the departure different results. It is also possible to compute the correlation
of the antenna’s behavior from ideal. The variation in the time from the antenna transfer function as defined in [1,7,11].
delay with angle may also be useful in determining the When the template is taken as the radiated electric field of the
uncertainty in precision radiolocation applications, and in antenna in a particular reference direction [8], and the
assessing beam scanning performance in circular or other non- normalized correlation pattern is expressed in terms of the
linear arrays. The time delay pattern may also aid antenna frequency domain antenna transfer function:
design by helping to identify and localize radiation  G G ∗ 
2
Re  a (ω ) ω 2 H • H 0 e − jωτ dω 
∞ 2

mechanisms.   − ∞   (2)
ρ (θ , φ ) = max  
τ  ∞ 2 2 G 2 ∞ 2 2 G 2

 −∞
a (ω ) ω H dω a (ω ) ω H 0 dω 
−∞ 
II. COMPUTATION OF TIME DELAY FROM CORRELATION where the reference direction is taken as (θ0,φ0),
H = H (θ , φ , ω )
Correlation from Antenna Transfer Function. The is antenna transfer function,
normalized correlation coefficient is defined in [3] by:
H 0 = H (θ 0 , φ 0 , ω )
is the transfer function in the reference direction, and a(ω) is
the input pulse (incident wave at the antenna input port) in the
James McLean and Robert Sutton are with TDK R&D Corp. frequency domain.
Heinrich Foltz is with the University of Texas-Pan American.

978-1-4244-1827-5/08/$25.00 ©2008 IEEE 9


PROCEEDINGS OF THE 2008 IEEE INTERNATIONAL CONFERENCE ON ULTRA-WIDEBAND (ICUWB2008), VOL. 1

Correlation patterns have been presented in [2,9-11]. Least Squares Approach From a port-to-port measurement
Typically, the value of the correlation coefficient is presented one can easily define a phase function
as a function of angle thus describing how different the Φ(θ , φ , ω ) = arg(S 21 (θ , φ , ω )) (6)
radiated fields of the horn are off axis compared to those on
Once the phase function is properly unwrapped, the linear
axis. What is often not shown is the time delay required to
component can be defined as the slope of simple linear
maximize the correlation. One can plot a pattern
regression. Unlike the correlation case, one can directly use
τ(θ,φ), where τ is the value needed to maximize equation (2) S21 data without worrying about the effect of the probe
as a function of angle. antenna, as long as only the differential time delay pattern is
desired. Since the probe antenna is not rotated it only adds a
Correlation from Port-to-Port Transfer Function. A simpler constant time delay offset that is removed in a differential
but less rigorous approach to defining a time delay is to pattern referenced to the boresight direction.
driectly use S21 data from a port-to-port antenna measurement.
The boresight S21 measurement versus frequency is converted One shortcoming of the least squares approach is that a
to the time domain, either directly as an impulse response, or function that has true time delay, that is no first order
after multiplication by a theoretical frequency domain input. component in the Taylor series, can still have a non-zero
This is used as a time-domain reference waveform f0(t): slope.

f 0 (t ) = F −1{S 21 (θ 0 , φ 0 , ω )a (ω )} (3) Bode-Hilbert Approach A minimum phase transfer


function contains, by definition, no linear phase component.
At any given angle θ,φ, off-axis one then calculates For any measured transfer function we can find an associated
minimum phase function using the Bode-Hilbert relation:
2ω ln H (ω ′ ) − ln H (ω )

f (t ) = F −1{S 21 (θ , φ , ω )a (ω )}
Φ m (ω ) =
(4)
π 0
dω ′ . (7)
and then finds the time-delay τ to maximize ω ′2 − ω 2


For many passive, reciprocal antennas the difference in the
f 0 (t − τ ) f (t )dt (5)
phase of the transfer function and the Bode-Hilbert derived
The flaw in this approach is that this data contains information phase is a linear function of frequency:
from both the probe antenna and the antenna under test. For Φ lin (θ , φ , ω ) = ∠H (θ , φ , ω ) − Φ m (θ , φ , ω )
the measurements reported here we used nominally identical (8)
antennas for both transmit and receive. Although this method ≈ τ (θ , φ )ω
is very different in principle from the one previously where Φ m (θ , φ , ω ) is the associated minimum phase
described, the results shown in the last section are remarkably
function. In Figure 1, the non-linear component of the phase
similar.
of the on-axis transfer function of the double-ridged horn
Differential Time Delay Pattern. The time delay associated described in [11] as computed with a least-squares linear
with the correlation computed in the manner described here phase extraction is shown. In the same figure the phase of the
will naturally be zero in the reference direction. More associated minimum phase function as computed with the
generally, when the template function is different from the Bode-Hilbert integral relationships is also presented. It can be
response in the bore-sight direction, the definition of the time seen that the least-squares linear phase extraction
delay can be generalized to a differential delay computed overestimates the linear phase.
between that required for maximization of the correlation in
the reference direction and that required for maximization of The procedure is as follows: measure the magnitude and
the correlation in another direction. phase of the antenna transfer function. From the magnitude,
compute a minimum phase function. Subtract the minimum
III. COMPUTATION OF TIME DELAY USING LINEAR PHASE phase from the measured phase. Assuming the result is a
VARIATION nearly ideal linear function of frequency (as has been the case
in our measurements), the time delay is computed as
A second approach is to find the time delay between two ∠H (θ , φ , ω ) − Φ m (θ , φ , ω )
antennas by examining the linear component of the phase τ (θ , φ ) = (9)
variation with frequency, and equating this to an equivalent
ω
time. As in the case of correlation, there are multiple ways
As in the previous methods, the difference between time
one could define and measure the "linear component." The
delays computed in this manner for an antenna’s response in
first and most straightforward method is based on a least-
the reference direction and its response in some other direction
squares fit, the second approach is based on Hilbert
provides a differential time delay pattern. This differential
transforms.
time delay pattern, like the differential time delay pattern
computed for the maximization of the correlation, will

10
PROCEEDINGS OF THE 2008 IEEE INTERNATIONAL CONFERENCE ON ULTRA-WIDEBAND (ICUWB2008), VOL. 1

naturally be zero in the reference direction. Thus, the two


patterns can easily be compared. When the pulse shape V. EXPERIMENTAL RESULTS
distortion is minimal, this time delay pattern will be very
similar to that computed by maximizing the correlation, but in In Figures 2 and 3, differential time delays as computed via
other cases it can be quite different. several different approaches from measured transfer function
200 data for the double-ridged horn described in [10] are plotted.
Measurements between two nominally identical horns were
150
made in an anechoic chamber at 1601 frequency points from
100 least squares linear phase extraction 0.01 to 20 GHz. The antenna has a frequency null slightly
Bode-Hilbert linear phase extraction
Phase (degrees)

50 around 18.5 GHz, at which phase and relative amplitude data


have reduced accuracy. Below 1 GHz the ridged waveguide
0 cutoff also leads to rapidly decreasing accuracy. Therefore,
-50 truncated data was used in the calculations.

-100 As can be seen, each of the approaches gives a slightly


-150 different result. It is thought that the distance extracted using
the minimum phase transfer function is the most accurate, but
-200
2 4 6 8 10 12 14 16 18 20 the variation in the data arrived at using the different
Frequency (GHz) approaches is larger than one would expect. One reason for
Figure 1: Comparison of nonlinear phase computed using this is an unfortunate coincidence involving the truncation of
least squares extraction of linear phase and Bode-Hilbert the frequency domain data at 20 GHz and a very deep
extraction of linear phase. transmission null in the transfer function of the horn at 18.5
GHz making extrapolation of the frequency domain data
difficult.
The minimum phase transfer function involves the removal of
all of the linear phase from the transfer function. In the time Some of the differential time delay observed is simply because
domain, this is equivalent to translating the response such that the antenna is not rotated about its center. As noted in [10], the
it begins exactly at the origin; that is, would begin “center” of many practical antennas is not a uniquely defined
simultaneously with the stimulus function. Of course, in the quantity. The data here was obtained by rotating the antenna
presence of noise, it is occasionally difficult to discern about a convenient point; since the antenna does not exhibit a
precisely when the time domain response begins. Thus, the true static phase center—the phase of the radiated fields varies
linear phase extraction using the minimum phase function with angle and frequency in such a way that phase center
might be useful for providing a more accurate estimate of this defined from the curvature of the equi-phase surfaces would
time. be a function of frequency and both rotation angles. Such
behavior is typical for most broadband antennas.
IV. TIME DELAY FROM WAVEFORM FEATURES
The effective “center” of a UWB antenna has been the subject
Differential time delays can also be found from various of some discussion. In [6], a center is defined through
features on time domain waveforms, which can in turn be extrapolation of the far field amplitude decay. In [10,12], an
found from either the antenna transfer function or directly effective center is determined from the phase variation of the
from S21 data. One possible choice is the time to the first far fields with distance. Specifically, a least-squares approach
significant peak. However, this neglects the rise time and thus is employed with measured transfer function data In [13], this
the zero just prior to the first peak in the impulse response concept was carried further using Hilbert transform
could be considered to provide a better estimate of the techniques.
beginning of the time response. In the results shown here, the
S21 data on-axis and off-axis were converted to directly to
time-domain waveforms (a port-to-port impulse response). VI. DISCUSSION AND CONCLUSION
The time to the desired feature on-axis is taken as the
The differential time delay pattern complements the
reference. Unfortunately, as can be seen in [10], the impulse
maximized correlation pattern and thus is part of the complete
response is difficult to work with due to artifacts from
description of a UWB antenna. It can be computed using
truncation and noise. Thus, the first peak and the zero just
several approaches including the time required to maximize
prior to the first peak in the step response are also considered
correlation and also from the linear phase component
as indicators of the start of the time response.
associated with the antenna transfer function. The Bode-
Hilbert integral relationships appear to be the most useful tool
for separating the transfer function phase into linear and
nonlinear components. However, when working directly with
time domain data, locating the beginning of the time domain

11
PROCEEDINGS OF THE 2008 IEEE INTERNATIONAL CONFERENCE ON ULTRA-WIDEBAND (ICUWB2008), VOL. 1

response might be more straightforward. ACKNOWLEDGEMENTS


In the hypothetical case in which the antenna does not This work was supported in part by the TDK R&D Corp., the
distort a pulse, but rather the pattern consists of linear phase or Texas Higher Education Coordinating Board, NSF MRI Grant
delay, the two time delays (that derived from the time delay 0421352, and Army Research Office Grant W911NF-06-1-
required to maximize the correlation and that computed from 0420.
the linear phase component) will be exactly the same. This is
the case when an antenna that exhibits a frequency REFERENCES
independent phase center and exhibits low distortion is rotated
about a point other than the phase center. [1] E. G. Farr and C. E. Baum, “Extending the Definitions of Antenna Gain
Finally, the use of the Bode-Hilbert integral relationships to and Radiation Pattern into the Time Domain,” Sensor and Simulation
Notes, Note SSN 350, Nov. 1992.
separate the transfer function phase into linear and nonlinear [2] D. Lamensdorf and L. Susman, “Baseband-Pulse-Antenna Techniques,”
components shows that the correlation pattern and hence the IEEE Antennas and Prop. Magazine, vol. 36, No. 1, pp. 20-30, 1994.
correlated gain can be computed from magnitude data only. [3] E. G. Farr and C. E. Baum, “Time Domain Characterization of
Antennas with TEM Feeds,” Sensor and Simulation Notes, Note 426,
October 1998.
[4] A. Sibille, “A framework for analysis of antenna effects in UWB
communications,” 2005 IEEE Vehicular Technology Conference, vol. 1,
0
pp. 48-52.
[5] A. Shlivinski, E. Heyman, and R. Kastner, “Antenna characterization in
-0.02 the time domain,” IEEE Trans. Antennas Propagat., vol. 45, no. 7, pp.
1140-1149, July 1997, see also correction in IEEE Trans. Antennas
-0.04
differential time delay (nanoseconds)

Propagat., vol.45, no. 8, p.1323, August 1997.


-0.06 [6] B. Scheers, M. Acheroy, A. Vander Vorst, “Time Domain simulation
and charactisation of TEM horns using a normalized impulse response,”
-0.08
IEE Proc. Microwaves, Antennas and Prop., vol. 147, No. 6, pp. 463-
-0.1 468, Dec. 2000.
Least-squares linear phase extraction: s21 [7] W. Sörgel, F. Pivit, and W. Wiesbeck, “Comparison of Frequency
-0.12 First peak of impulse resp.: s21 Domain and Time Domain Measurement Procedures for Ultra Wideband
Zero prior to first peak of impulse resp.: s21
-0.14 First peak of step resp.: s21 Antennas,” in Proc. 25th Annual Meeting and Symposium of the
Zero prior to first peak of step resp.: s21 Antenna Measurement Techniques Association (AMTA ’03), pp. 72–76,
-0.16 Bode-Hilbert linear phase extraction: s21 Irvine, Calif, USA, October 2003.
Max. correlation of impulse resp.: s21
Max. correlation of step resp.: s21 [8] J. S. McLean, H. Foltz, and R. Sutton, “Pattern Descriptors for UWB
-0.18
Max. correlation of cosine pulse: H Antennas,” IEEE Trans. Antennas and Propagat., vol. 53, no. 1, pp.
Max. correlation of sine pulse: H 553-559, Jan. 2005.
-0.2
0 10 20 30 40 50 60 [9] P. Miskovsky, J.M. Gonzalez-Arbesu, J. Romeu, “Application of UWB
Angle off of bore sight in E-plane (degrees) antenna Descriptors to Lossy Dipole Performance Assessment,” In Proc.
IEEE Antennas and Propagation Society International Symposium 2006,
Figure 2: Differential time delay plotted as a function of pp. 175-178, 9-14 July 2006.
the angle off of bore sight in the E-plane. [10] J. S. McLean, R. Sutton, A. Medina, H. Foltz, J. Li, “The Experimental
Characterization of UWB Antennas,” IEEE Antennas and Prop.
Magazine, vol. 49, No. 6, pp. 20-30, Dec. 2007.
0.4 [11] J. S. McLean, A. Medina, R. Sutton, H. D. Foltz, Junfei Li,
“Measurements of UWB Antenna Pattern Descriptors, ” presented at the
0.2 2006 AMTA Europe Symposium, Munich, May 4, 2006.
[12] J. S. McLean, A. Medina, R. Sutton, H. D. Foltz, J. Li, “Complex
0 Antenna Transfer Function Measurements with Emphasis on High
Time delay (nanoseconds)

Positional Resolution,” presented at the 2006 AMTA Symposium,


-0.2 Austin, TX, Oct., 16, 2006.
[13] H. Foltz, J. McLean, A. Medina, J. Li, and R. Sutton, “UWB Antenna
-0.4 Least squares linear phase extraction: s21 Transfer Functions Using Minimum Phase Functions,” in Proc. 2007
First peak of impulse resp.: s21 IEEE Int. Antennas and Propagation Symposium, Honolulu, June 2007,
Zero prior to first peak of impulse resp.: s21
-0.6 First peak of step resp. s21 pp. 1413-1416.
Zero prior to first peak of step resp.: s21 [14] H. W. Bode, Network Analysis and Feedback Amplifier Design, D. Van
-0.8 Bode-Hilbert linear phase extraction: s21 Nostrand, New York, 1945.
Max. correlation of impulse resp.: s21
Max correlation of step resp. s21
-1 Max. correlation of cosine pulse: H
Max. correlation of sine pulse: H
-1.2
0 10 20 30 40 50 60
Angle off of bore sight: H-plane (degrees)

Figure 3: Differential time delay plotted as a function of


the angle off of bore sight in the H-plane.

12