11b

20. W. Rotman and R.F.
Turner, Wide-angle microwave lens for

line source applications, IEEE Trans. Anten. Propag.
11(6):623632 (Nov. 1963).
21. L. Musa and M. S. Smith, Microstrip port design and sidewall
absorption for printed Rotman lenses, IEE Proc. H, Micro-
waves Anten. Propag. 136(1):5358 (Feb. 1989).
22. D. Archer, Lens-fed multiple beam arrays, Microwave J.
3742 (Oct. 1975).
23. D. T. Thomas, Multiple beam synthesis of low sidelobe
patterns in lens fed arrays, IEEE Trans. Anten. Propag.
26(6):883886 (Nov. 1978).
24. R. C. Hansen, Phased Array Antennas, Wiley, New York;
1998.
25. K. K. Chan and S. K. Rao, Design of a Rotman lens
feed network to generate a hexagonal lattice of multiple
beams, IEEE Trans. Anten. Propag. 50(8):10991108
(Aug. 2002).
26. J. B. L. Rao, Multifocal three-dimensional bootlace lenses,
IEEE Trans. Anten. Propag. 30(6):10501056 (Nov. 1982).
27. C. M. Rappaport and A. I. Zaghloul, Optimized three-dimen-
sional lenses for wide-angle scanning, IEEE Trans. Anten.
Propag. 33(11):12271236 (Nov. 1985).
28. R. C. Hansen, Design trades for Rotman lenses, IEEE Trans.
Anten. Propag. 39(4):464472 (April 1991).
29. A. F. Peterson and E. O. Rausch, Scattering matrix integral
equation analysis for the design of a waveguide Rotman
lens, IEEE Trans. Anten. Propag. 47(5):870878 (May 1999).
30. N. Yuan, J. S. Kot and A. J. Partt, Analysis of Rotman lenses
using a hybrid least squares FEM/transnite element
method, IEE Proc. Microwaves Anten. Propag. 148(3):
193198 (June 2001).
31. R. E. Clapp, Extending the R-2R Lens to 3601, IEEE Trans.
Anten. Propag. 32(7):661671 (July 1984).
32. G. D. M. Peeler and D. H. Archer, A two-dimensional micro-
wave Luneburg lens, Trans. IRE Prof. Group Anten. Propag.
1(1):1223 (July 1953).
33. J. S. Hollis and M. W. Long, A Lunenburg lens scanning
system, IEEE Trans. Anten. Propag. 5(1):2125 (Jan. 1957).
34. G. A. Thiele and R. C. Rudduck, Geodesic lens antennas
for low-angle radiation, IEEE Trans. Anten. Propag. 13(4)
514521 (July 1965).
35. E. H. Braun, Radiation characteristics of the spherical Lune-
burg lens, IEEE Trans. Anten. Propag. 4(2):132138 (April
1956).
36. G. D. M. Peeler and H. G. Coleman, Microwave stepped-index
Luneburg lenses, IEEE Trans. Anten. Propag. 6(2):202207
(April 1958).
37. H. Mosallaei and Y. Rahmat-Samii, Nonuniform Luneburg
and two-shell lens antennas: Radiation characteristics and
design optimization, IEEE Trans. Anten. Propag. 49(1):6069
(Jan. 2001).
38. S. Rondineau, M. Himdi, and J Sorieux, A sliced spherical
Luneburg lens, IEEE Anten. Wireless Propag. Lett. 2:163166
(2003).
39. K. Kikuchi, Y. Kido, and T. Ogawa, Millimeter wave sensor
using cylindrical Luneburg lens with at sides, IEEE Trans.
Instrum. Meas. 47(2):423425 (April 1998).
40. G. T. Ruck, D. E. Barick, W. D. Stuart and C. K. Krichbaum,
Radar Cross Section Handbook, Vol. 2, Plenum, New York;
1970, pp. 597600.
41. G. M. Rebeiz, Millimeter-wave and tetrahertz integrated
circuit antenna, Proc. IEEE, 80(11):17481770 (Nov. 1992).
42. D. F. Filipovic, S. S. Gearhart, and G. M. Rebeiz, Double-slot
antennas on extended hemispherical and elliptical silicon
dielectric lenses, IEEE Trans. Microwave Theory Tech.
41(10):17381749 (Oct. 1993).
43. A. Neto, L. Borselli, S. Maci, and P. J. I. de Maagt, Input
impedance of integrated elliptical lens antennas, IEE Proc.
Microwaves Anten. Propag. 146(3):181186 (June 1999).
44. M. J. M. van der Vorst and P. J. I de Maagt, Efcient body of
revolution nite-difference time-domain modelling of inte-
grated lens antennas, IEEE Microwave Wireless Compon.
Lett. 12(7):258260 (July 2002).
45. X. Wu, G. V. Eleftheriades, and T. E. van Deventer-Perkins,
Design and characterization of single- and multiple-beam
mm-wave circularly polarized substrate lens antennas for
wireless communications, IEEE Trans. Microwave Theory
Tech. 49(3):431441 (March 2001).
46. C. A. Fernandes, Shaped dielectric lenses for wireless milli-
meter-wave communications, IEEE Anten. Propag. Mag.
41(5):141150 (Oct. 1999).
47. D. N. Black and J.C. Wiltse, Millimeter-wave characteristics
of phase-correcting Fresnel zone plates, IEEE Trans. Micro-
wave Theory Tech. 35(12):11221129 (Dec. 1987).
48. A. Petosa and A. Itippiboon, Shadow blockage effects on the
aperture efciency of dielectric Fresnel lenses, IEE Proc.
Microwaves Anten. Propag. 147(6):451454 (Dec. 2000).
49. H. D. Hristov and M. H. A. J. Herben, Millimeter-wave
Fresnel-zone plate lens and antenna, IEEE Trans. Microwave
Theory Tech. 43(12):27792785 (Dec. 1995).
50. A. Petosa, A. Itippiboon, and S. Thirakoune, Shadow blockage
improvement using a perforated dielectric Fresnel lens, Proc.
2003 IEEE Antennas and Propagation Society Int. Conf.
Vol. 4, June 2227, 2003, pp. 514517.
LINEAR ANTENNAS
ASOKE K. BHATTACHARYYA
Lincoln University
Historically, using a piece of radiating straight wire as an
aerial, or antenna, was a natural choice for wireless com-
munications. This radiating piece of straight wire without
curvature is the linear antenna. A simple example of a
linear antenna is a two-wire transmission line carrying
equal currents in opposite directions and hence no resul-
tant radiation. A two-wire transmission line may be bent
to create an efcient radiator such as a dipole. The linear
antennas have been treated in numerous references. Some
of them are in Refs. 113. We will describe key features of
linear antennas in this article.
1. SOME RELEVANT TERMS
Before we proceed to discuss linear antennas, we need to
dene and discuss certain terms in accordance with the
Institute of Electrical and Electronics Engineers (IEEE)
standard definitions of antenna terminology.
2336 LINEAR ANTENNAS
Previous Page
1.1. Power Radiated, Radiation Intensity, and
Radiation Resistance
Electromagnetic waves, by virtue of their transverse na-
ture, propagate in a direction perpendicular to the plane
containing the electric led E and magnetic eld H. The
instantaneous Poynting vector P, which is a measure of
the power density associated with the electromagnetic
wave, is given by
PEH 1
where P, E, and H are instantaneous Poynting vector in
watts per square meter, electric eld in V/m, and magnetic
eld in amperes per meter.
The total power P crossing a sphere enclosing the
source (antenna/scatterer) at its center is obtained by in-
tegrating the power density over the sphere and is given
by
PW ^ nn
.
dSWda 2
where W is the instantaneous power crossing the sphere
per unit area held perpendicular to the direction of the
ow, ^ nn is the positive outwardly drawn at the point of in-
cidence, and dS is the unit area arbitrarily oriented at the
point of incidence. With exp(jot) variation assumed, the
average power density is given by the time-average Poyn-
ting vector P
av
:
W
av
u; v; w
1
2
ReEH
3
The average radiated power is given by
P
av

1
2
_ _
ReEH
da 4
The radiation intensity U is dened by the product of
power density P
rad
and the square of the far-eld range (r)
and is expressed as
Ur
2
P
rad
5
The radiation resistance (R
r
) is dened as the positive re-
sistance across which the real power radiated (P
rad
) can be
thought of as being dissipated. The relationship between
P
r
, R
r
, the input resistance, and the current I is
R
r

P
r
I
2
6
The input resistance of an antenna is a sum of radiation
resistance plus the positive resistance due to ohmic losses.
1.2. Radiation Intensity, Directivity, and Gain
The antenna radiates real power in the far zone in space
over a solid angle of 4p radians. The radiation intensity
U(y, f), the real power radiated per unit solid angle, is a
product of the radiation intensity P
rad
, the real power per
unit solid area on the surface, multiplied by the square (r
2
)
of the distance and is given by
Uy; f r
2
P
rad
y; f 7
The total power can be estimated by integrating the radi-
ation over a large sphere enclosing the antenna over 4p
radians:
P
rad

_
S
UdO
_
p
y 0
_
2p
f0
U sin y dy df 8
An isotropic source, such as an ideal point source, radiates
uniformly in all directions and is independent of y and f,
and the radiation intensity U
0
is related to the real power
radiated by the simple formula:
U
0

P
rad
4p
9
The directivity is a measure of how efciently the antenna
is directing the radiation in space, according to the 1983
IEEE standard [14]. The directivity D, a dimensionless
quantity, of an antenna is given by
D
U
U
0
4pU
P
rad
10
The directivity is dependent on the direction. If the direc-
tion is not specied, the default is the direction of maxi-
mum radiation intensity.
The dimensionless maximum directivity D
max
, denoted
by D
0
, is expressed as
D
0

U
max
U

4pU
max
P
rad
11
Many practical antennas work with dual polarizations in
mutually perpendicular directions, and then the directiv-
ity is dened in that particular direction; the total max-
imum directivity is a sum of directivities in mutually
perpendicular directions and is expressed as
D
0
D
k
D
?
12
The mathematical expressions for D
k
and D
?
are
D
k

4pU
k
P
radk
P
rad?
13a
D
?

4pU
?
P
radk
P
rad?
13b
1.3. Antenna Gain and Radiation Efciency
An antenna is a passive device, but it can be designed to
radiate more energy in a desired direction. The gain (G) of
LINEAR ANTENNAS 2337
an antenna is dened as
G
radiationintensity inmaximumdirectionof radiationU
0
radiationintensity of a lossless isotropic source withsame input

14
All practical antennas have losses, and therefore efcien-
cies of practical antennas are less than 100%. The antenna
efciency (Z) is dened as the ratio of the real power ra-
diated in space by the antenna to the real power input at
its feed terminals:
Radiationefficiency Z
real power radiated by test antennaP

rad
total real input at antennafeedterminals P

in
15
The antenna efciency Z is related to the directivity D and
the gain G through the relationship
GZD 16
1.3.1. The Vector and Scalar Potentials and Field Calcu-
lations Using Potentials. Most of the time a direct solution
of Maxwells equations subject to the boundary conditions
for a practical problem becomes difcult. Therefore, it is
customary to use intermediatory (or auxiliary) functions,
called potential functions, to obtain solutions of electro-
magnetic problems. There are four such functions; two of
them are scalar (one electric and one magnetic) and two of
them are vector (one electric and one magnetic) potentials.
The magnetic vector potential A is related to the mag-
netic ux density through the relation BrA and the
electric scalar potential V is related to E and A through
the relation E rVA.
The steps to determine the elds at any point due to the
linear antenna are as follows: (1) dene the current dis-
tribution on the dipole; (2) nd expressions for the four
potentials; (3) transfer the Cartesian components of the
magnetic vector potentials to those in spherical polar co-
ordinates; (4) once the magnetic vector potential is deter-
mined, the magnetic eld at any point is obtained; and
(5) what remains to be done is to use Maxwells equation to
determine the electric elds at any point from the mag-
netic eld obtained.
Before we proceed to determine radiated elds, let us
discuss the four potentials for this example. The magnetic
current I
m
is equal to zero since the wire carries a la-
mentary electric current and hence the electric vector po-
tential F is zero since it is a function of magnetic current
only. In this situation, the magnetic vector potential A is
given by
A
m
0
4p
_
dl=2
dl=2
Jx
0
; y
0
; z
0
expjkR
R
dz
0
17
where (x
0
, y
0
, z
0
) are source coordinates, (x, y, z) are the eld
coordinates, R is the distance between the observation
point and any point on the source (Fig. 1). J
z
is the z-
directed electric current element, and the linear path C is
along the length of the source.
2. THE INFINITESIMAL, OR HERTZIAN, DIPOLE
Before we do the analysis for a practical antenna, namely,
a linear antenna, let us establish the analysis procedure
for an innitesimal, elementary, or Hertzian dipole. These
are building blocks for more complex antenna systems.
Since the dipole is innitesimal, the current is assumed to
be constant.
For the innitesimal dipole (Fig. 1), the current on the
innitesimal dipole is given by
J
e
x
0
; y
0
; z
0
^ zzJ
0
18
where
x
0
y
0
0; since the lengthof the
dipole is infinitesimal andof lengthdl
R
x x
0
2
y y
0
2
z z
0
x
2
y
2
z
2
_
rlet
l /2
(a)
(b)
l /2
y
r
z
z
x
x
y
E
r
E
0
E
0
[
0
r
0
[
Figure 1. (a) The innitesimal dipole and (b) its coordinate sys-
tem. This gure geometrically shows how the eld at any obser-
vation point from an innitesimal dipole, which is a building
block, can be estimated.
With these, the magnetic vector potential A is given by
Ax; y; z ^ zz
m
0
4r
expjkr dz
0
^ zz
m
0
I
0
dl
4pr
expjkr for rO0
excluding the source
19
The components of A are given by
A
r
A
z
cos y
m
0
I
0
dl
4pr
expjkr cos y 20a
A
y
A
z
sin y
m
0
I
0
dl
4pr
expjkr sin y 20b
A
f
0 20c
Due to symmetry of the radiating dipole, we have @/@f0;
thus we obtain
Hf
1
4pr
@
@r
rAf
@A
r
@y
_ _
21
The expressions for magnetic elds are given by
H
r
H
y
0
H
f
j
k
0
I
0
dl
4pr
1
1
jkr
_ _
sin y expjkr
22
The electric eld can be found from a curl relationship:
E
1
joe
rH 23
This gives the three longitudinal and transverse electric
eld components as
E
r

m
0
I
0
dl
2pr
2
1
1
jkr
_ _
cos y expjkr 24a
E
y
jZ
0
k
0
I
0
sin y
4pr
1
1
jkr
1
kr
2
_ _
expjkr 24b
E
f
0 24c
2.1. Near and Far Fields
The near-eld region are at a close enough distance such
that kr51:
E
r
jZ
0
I
0
dl
expjkr
2pk
0
r
3
cos y 25a
E
y
jZ
0
I
0
dl
expjkr
4pk
0
r
3
sin y 25b
H
f
I
0
dl
expjkr
4pr
2
sin y 25c
E
f
H
r
H
y
0 25d
Several observations are in order. E
r
and E
y
have (1/r
2
)
variation as distance and therefore decays very fast. These
are induction components and die down rapidly with
distance. The electric eld components E
r
and E
y
are
in time phase, but the magnetic eld component H
f
is in time quadrature with them. Therefore, there is no
time-average power ow associated with them. Hence, the
average power radiated will be zero, and the Poynting
vector is imaginary. This can easily be veried by inte-
grating the average power density over a sphere in the
near region.
The space surrounding the antenna can be divided into
three regions, namely, induction, near-eld (Fresnel), and
far-eld regions. The induction region has 1/r
3
space vari-
ation, the near eld has 1/r
2
variation, and the far eld
has a 1/r variation with distance r.
2.2. Far Field
The far-eld expression can be obtained with krb1 and by
extracting the (1/r) term and is given by
E
y
jZ
0
k
0
I
0
dl
expjkr
4pr
sin y 26a
E
r
E
f
H
r
H
y
0 26b
H
f
jk
0
I
0
dl sin y
expjkr
4pr
26c
The intrinsic impedance Z
m
of the medium is dened as
the ratio of the tangential electric and magnetic elds and
is given by
Z
m

E
y
H
f
Z
m
27
2.3. Intermediate-Field Region
For expression for eld components in the intermediate
region (kr41), the reader is referred to any standard text
on antennas [1].
2.4. Directivity
The radiation intensity U is given by
Ur
2
W
av
28
where
W
av

1
2
ReEH
29
and
U
r
2
2Z
0
jE
y
r; y; fj
2
30
The maximum directivity D
0
turns out to be equal to 1.5.
2.5. Radiation Resistance
The radiation resistance is obtained by dividing total pow-
er radiated by the lossless antenna by |I
0
|
2
/2 and is given
by
R
r
80p
2
dl
l
_ _
2
31
3. THE THIN LINEAR ANTENNA
This section deals with the analysis and properties of a
nite-length dipole. The wire is considered to be thin such
that tangential currents can be neglected and the current
can be considered as only linear. The thin linear antenna
and its geometry are shown in Fig. 2. The boundary con-
ditions of the current are that the currents are zero at the
two ends and maximum at the center. There is experi-
mental evidence that the current distribution is sinusoi-
dal. The current distributions are for a dipole l and for
length varying from l/2 to l. Thick dipoles will be treated
in a subsequent section.
3.1. The Current Distribution
The current distribution on the thin dipole is given by
I
x
x
0
0; y
0
0; z
0
^ zzI
0
sinkl=2 z
0
; 0 z
0
l=2
^ zzI
0
sinkl=2 z
0
; l=2 z
0
0
_
_
_
32
The current distributions on the linear dipoles for differ-
ent lengths are shown in Fig. 3, and Fig. 4 shows the cur-
rent distributions on a half-wave dipole at different times.
3.1.1. Fields and Radiation Patterns. To determine the
eld due to the dipole, it can be subdivided into small seg-
ments. The eld at any point is a superposition of the con-
tributions from each of the segments. Since the wire is
very thin, we have x
0
0 and y
0
0. The electric and mag-
netic eld components due to the elementary innitesimal
dipole segment of length dz
0
at an arbitrary point are
given by
dE
0
jZ
0
k
0
I
z
x
0
; y
0
; z
0
expjkR
4pR
sin y dz
0
33a
dE
r
dEfdH
r
dH
y
0 33b
dH
f
jk
0
I
z
x
0
; y
0
; z
0
expjkR
4pR
sin y dz
0
33c
where
R
x
2
y
2
z z
0
2
_

r
2
2rz
0
cos y z
02
_
with
r
2
x
2
y
2
z
2
; z r cos y
The expression for R can be expanded binomially as
Rr z
0
cos y higher-order terms decaying
very fast withrbz
0
34
x
x
y
P(r,0,[)
P(r,0,[)
R
r
R
z cos 0
r
z
z
l /2
(a)
(b)
dz
l /2
l /2
l /2
y
=
=

z
dz
z
[ [
[ [
0
0
0
0
Figure 2. (a) Thin linear antenna and (b) its coordinate system.
This gure geometrically shows how the eld at any observation
point can be formulated using the basic building block, namely,
the innitesimal dipole.
The total eld due to the dipole is obtained by integrating
over the whole length. Omitting the straightforward steps,
it turns out that the eld components E
y
and H
f
are given
by
E
y

jZ
0
I
0
2p
expjkr
r
coskl=2cos y coskl=2
sin y
_ _
35a
H
f
E
y
=Z
0
35b
where Z
0
is the intrinsic impedance of free space.
To save space we will not describe the derivation of
power radiated, which involves C
i
(x), S
i
(x), and C
in
(x)
functions.
3.1.2. The Radiation Resistance. The resistance can be
shown to be given by
R
r

Z
2p
fC lnkl Ckl
1
2
sinkl
.
S2kl 2Skl
1
2
cosklC lnkl=2
C2kl 2Cklg
where C(X) and S(X) are well-known functions constitut-
ing Fresnel integrals (see App. IV in Ref. 1).
4. THE METHOD-OF-MOMENTS SOLUTION
For many practical antennas and scatterers, including
linear antennas, the desired current distribution is ob-
tained by numerically solving the integral equations. The
method of moments (MoM) is a technique to convert an
integral equation to a matrix equation and hence solve the
linear system by standard matrix inversion techniques.
The MoM is very well documented in the literature [15],
and only the basic steps will be briefly discussed below.
The magnetic eld integral equation (MFIE) for the
unknown current density can be rewritten as an inhomo-
geneous equation in operator form as follows
LJ
s
2nHr 36
where the right-hand side is a known quantity and L(J
s
) is
an integrodifferential linear operator dened as
LJ
s
J
s
^ nn
1
2p
_
C
J
s
r
0
r
0
Gds
0
37
where J
s
is the electric current on the wire and G is the
free-space Green function.
Let us discuss the solution of inhomogeneous scalar
equation given by
Lf g 38
where f f(x) is the unknown function to be determined,
g(x) is a known function and L is a linear operator. The
seven steps [16] in obtaining the solution of Eq. (38) are
the same as the steps for the solution of Eq. (37).

I
l <<
(a)
z
z z
z
/2<l <
(c)
l = /2
(b)
<l<3z/2
(d)
I
in
I
I
0
= I
in
I
I
in
I
0
I
I
in
I
0
Figure 3. Current distribution on dipoles of different lengths.
Different physical lengths of the dipole support different current
distributions with varying number of half-sinusoids. This is be-
cause at the two ends of the wire, the boundary condition that the
current must be zero has to be satised.
I I I
I
0
t = 0
(a)
t = 3T/8
(d)
t = T/2
(e)
t = T/8 t = T/8
(b) (c)
0.707I
0
0.707I
0
I
0
I I
/4
/4
/4
/4
/4
/4
/4
/4
/4
/4 z
z
z
z
z
z
z
z
z
z
Figure 4. Current distribution on a l/2 wire antenna for differ-
ent times. The current, which is alternating, changes with time.
This gure shows the current changes on a half-wavelength wire
antenna at different time instants.
The steps are as follows:
1. Expand f as
f
N
n1
a
n
f
n
39
where the a
n
values are the unknown coefcients,
and the f
n
values are known functions of x, known as
expansion, or basis, functions.
2. Using Eq. (38) in Eq. (37), we get
N
n1
a
n
Lf
n
g 40
3. Dene a suitable inner product hf ; gi dened in the
range L of x:
hf ; gi
_
D
f xgx dx 41
4. Dene a set of testing or weighting functions w
m
,
m1, 2,y, N, in the range. Taking the inner prod-
uct of Eq. (40) with each w
m
and obtain
N
n1
a
n
hW
m
; Lfni hW
m
; gi 42
where m1, 2,y, N; h
.
i is the inner product, the
product of the two functions integrated over the
domain.
5. Express the set of algebraic equations given by
Eq. (40) in the matrix form
I
mn
a
n
g
m
43a
where the matrix is given by
I
mn

W
1
; Lf
1
W
1
; Lf
2
W
1
; Lf
1
W
2
; Lf
2
W
2
; Lf
2
W
2
; Lf
2

W
N
; Lf
N
W
N
; Lf
N
W
N
; Lf
N
_
_
_
43b
where a
N
and g
N
are the column vectors given by
a
N

a
1
a
2
.
.
.
a
N
_
_
_
_
44a
g
N

hW
1
; gi
hW
2
; gi

hW
N
; gi
_
_
_
_
44b
4.1. Moment-Method Solution for Radiation from Thin Wire
Two types of volume integral equations are used for the
linear antenna and wire scatterer problem. These are the
integral equation of Hallen type and the integral equation
of Pocklington type. Hallens integral equation usually ne-
cessitates the postulation of a delta-gap voltage at the
feedpoint and also requires the inversion of an (N1)-
order matrix. The advantages of Pocklingtons integral
equation is that it is possible to incorporate different
source congurations and it requires inversion of a
matrix of order N.
For a current-carrying perfectly conducting wire, the
Hallens integral equation obtained by solving the scalar
wave equation is given by (1)
I
z
z
0
expjkR
4pR
dz
0
j
e=m
_
B
1
coskz B
2
sinkz
45
where
I
z
z
0
current owing through elementary length of
wire
Rdistance of the observation point from ele-
mentary length
I total length of center- fed wire
e, m, s0 primary constants of medium in which the
antenna is radiating k is the derived second-
ary constant, namely, the wavevector of the
medium
B
1
and B
2
are constants to be determined
The Pocklingtons integral equation is given by
I
z
z
0
@
2
@z
2
k
2
_ _
Gz; z
0
_ _
dz
0
joeE
i
z
at r a
46
where for thin wire the radius a5l, the free-space Green
function G(z, z
0
) is given by
Gz; z
0
GR
expjkR
4pR
with
R
a
2
z z
0
2
_
E
i
z
the incident fieldthis is thesource fieldfor
antennas andscattered fieldfor scatterers
47
The availability of high-speed computers, graphics, and
software packages, along with the enormous development
of microcomputers, has made electromagnetic numerical
computation extremely viable. An attractive feature of
numerical methods is their ability to handle arbitrarily
shaped and electrically large bodies and bodies with non-
uniformity and anisotropy where exact solutions can only
at best provide some physical insight. For large problems,
it is possible to get a linear system with a minimum set of
equations to achieve a certain accuracy.
An account of the area of numerical computation of thin
wire problem is well-documented in the literature [1,3,6
13]. As described in Ref. 8, there are many important
computational issues involved in thin wire problems.
These are (1) segmentation of the structure and the cor-
rect number of segments, (2) choice of right current ex-
pansion functions, (3) the thin-wire approximation (radius
a5l), (4) roundoff error due to matrix factorization, (5)
near-eld numerical anomaly, (6) treatment of the junc-
tions of the segments, (7) wire-grid modeling, and (8) com-
puter time required. Also, the errors [7] involved are of
concern. There are two types of errors encountered: (1) the
physical modeling error, because in the absence of an ex-
act solution for a variety of semicomplex and complex
structures, it is the natural departure of the assumed
structural details from the actual structure; and (2) the
numerical modeling error, since all numerical methods are
approximate but sufciently accurate for the application.
4.2. Formulation
Before we discuss the formulation of the thin-wire integral
equation, comments on the expansion functions used in
this study are in order. There are broadly two types of
expansion functions:
1. Entire-Domain Functions. As the name suggests,
these functions are dened over the entire integra-
tion path C, and the subdomain is dened over piec-
es on the integration path C. Some of these are
Fourier, MacLaurin, Chebyshev, Hermite, Legend-
re, and so on. Tables available in Richmonds work
[17] and also in Ref. 8 are reproduced in Table 1.
2. Subdomain Expansion Functions. Subdomain ex-
pansions are attractive, convenient, and less expen-
sive in terms of computer time. This stems from the
fact that the current is matched on part of the inte-
gration path, whereas for the entire domain the in-
tegration is performed over the whole path and for
all N terms of the expansions and coefcients deter-
mined.
Miller and Deadrick [8] provide a table containing the
many basis and weighting functions that have been tested
in computer runs. This table also compares the suitability
of the use of different functions in different problems.
The table is too big to reproduce here, and it is left to
the reader to look up. It tabulates the method, integral
equation type, basis function, current conditions for inte-
rior and end segments, the basis function in terms of un-
known for unknown and end segments, weighting
function for interior and end segments, number of un-
knowns, and specific comments on the applicability of the
expansion functions.
4.3. The Electric Field Integral Equation and Its
Matrix Representation
Figure 5 gives the geometry of the arbitrarily oriented
straight wire. The wire is broken into segments, or sub-
sections. The mininumerical electromagnetic code (MI-
NINEC) relates the current distribution on the wire due
to the incident led. The integral equation relating the
incident eld E
i
, magnetic vector potential A, and electric
scalar potential f are given by
E
i
. ^
tt joA
. ^
tt
^
tt
.
rf 48a
where
A
m
0
4p
_
IsSskr ds 48b
f
1
4oe
_
qskr ds 48c
Table 1. Entire-Domain Current Expansions Using Different Polynomials
A. The Polynomials
Fourier: I(z) I
1
cos(px/2) I
2
cos(3px/2) I
3
cos(5px/2)
MacLaurin: I(z) I
1
I
2
x
2
I
3
x
4

Chebyshev: I(z) I
1
T
0
(x) I
2
T
2
(x) I
3
T
4
(x)
Hermite: I(z) I
1
H
0
(x) I
2
H
2
(x) I
3
H
4
(x)
Legendre: I(z) I
1
P
0
(x) I
2
P
2
(x) I
3
P
3
(x)
where 1/2rx 2z/Lr1/2
B. Typical Results for Functions
L0.5l; a0.005l; y
i
901
In Fourier MacLaurin Chebyshev Hermite Legendre
1 3.476 3.374 1.7589 8.2929 2.2763
2 0.170 4.037 1.5581 14.3644 2.1005
3 0.085 3.128 0.0319 4.4135 0.0655
4 0.055 4.101 0.0112 0.3453 0.0421
5 0.040 1.871 0.0146 0.0073 0.0322
t is a unit vector tangential to the wire at any distance
along the path of integration which is the length of the
wire and k(R) is given by
kR
1
2p
_
2p
f0
expjkR
R
df 49
The continuity equation given below determines the rela-
tionship between the charge q(s) and the rate of change of
current with distance:
qs
1
jo
dl
ds
50
The MININEC solves the integral equation using the
following steps:
1. The wires are divided into small equal segments
such that the length of the segment is still large
compared to the radius of the wire (a5l, typically
1
100
th of a wavelength). The radius vectors m, n0,
1, 2,yare dened with reference to a global origin.
2. The unit vectors are dened as
s
n1=2

r
n1
r
n
r
n1
r
n
51
The testing and expansion functions are pulse func-
tions that are dened by
P
n
s
1 for s
n1=2
osos
n1=2
0 otherwise
_
52
where the points s
n1/2
and s
n1/2
are the segment mid-
points and are given by
s
n1=2

s
n1
s
n
2
; s
n1=2

s
n
s
n1
2
53
In terms of global coordinates, we obtain
r
n1=2

r
n1
r
n
2
; r
n1=2

r
n
r
n1
2
54
When the pulse functions of Eq. (52) are inserted in pa-
rentheses, we obtain
E
i
s
m
.
s
m
s
m1
2
_ _
S
m1=2

s
m1
s
m
2
_ _
S
m1=2
_ _
joAs
m
.
s
m
s
m1
2
_ _
S
m1=2
_
S
m1
s
m
2
_ _
S
m1=2
_
55
The exact kernel treatment developed above is for obser-
vation points on source segments. For observation points
near but not on the source, a segment has been developed
by Wilton, and MININEC has incorporated it (16).
4.4. Expansion of Currents
The currents are expanded in terms of pulse functions as
shown in Fig. 5, excluding the endpoints where the cur-
rents are chosen as zeros to satisfy boundary conditions at
the ends. The current expansion is given by
Is
N
n1
I
n
P
n
s 56
Using this current expansion in Eq. (48b) and after math-
ematical manipulations which are available in Ref. 1 and
are not detailed here, we get the linear system matrix
equation in N unknowns
V
m
Z
mn
I
n
57
where m, n1, 2,y, N, [Z
mn
] is the square impedance
matrix, and [V
m
] and [I
n
] are applied voltage and current
column vectors:
Z
m

1
4pjoe
k
2
r
m1=2
r
m1=2
.
s
n1=2
c
m;n;n1=2
_
s
n1=2
c
m;n1=2;n

c
m1=2;n;n1
s
n1
s
n
c
m1=2;n1;n
s
n
s
n1
c
m1=2;n;n1
s
n1
s
n
c
m1=2;n1;n
s
n
s
n1
_
58
This matrix has elliptical integrals involved in it. These
elliptical integrals can be evaluated numerically.
These equations are valid for any radius other than
small, for which the expression for c breaks down, and
Harrington [19] provided an approximate formula for c.
0
1 0 n 1
n 1
n n +1
n +1 N 1
N 1 N
.........
.........
p
1
p
n 1
p
n +1
p
N 1
p
n
......... .........
.........
.........
1 n
(a)
(b)
N
Figure 5. Wire segmentation with pulses for current and charges:
(a) unweighted current pulses; (b) unweighted charge representa-
tion. The whole length is broken into several segments. Each seg-
ment is assigned a pulse, and the pulses represent the assumed
current distribution.
This is given by
c
1
2pDs
ln D
s
a
_ _
j
k
0
4p
for mn 59a
expjkr
m
4pr
m
for mOn 59b
The integral is given by
c
m;u;v

_
k
0
s
m
s
0
ds
0
60
4.5. Inclusion of Nonradiating Structures
4.5.1. The Ground Plane. When the wire structure near
the ground plane is assumed to be perfectly conducting, an
image is created. The structure and the ground plane is
equivalent to the structure and the image.
The voltage and current relationship is given by
V
m
Z
0
mn
I
n
61
where
Z
0
mn
Z
mn
Z
m;2Nn1
4.5.2. Wire Attached to Ground. When a wire is at-
tached to the ground on one or both sides, there will be a
residual component of current at one or both ends. In this
case, a current pulse is automatically added to the end
point in the formulation.
4.5.3. Lumped-Parameter Loading. If an additional com-
plex load is added to the perfectly conducting wire (Fig. 6),
there will be an additional voltage drop created at that
point if the location of the load (Z
l
R
l
jX
l
) is at a point
of nonzero pulse function. The impedance matrix is
modied to
V
m
Z
0
mn
I
n
where Z
0
mn
is the modied impedance matrix and is given
by
Z
0
mn

Z
mn
Z
load
for mn
Z
mn
for mOn
_
4.6. Validation of the MININEC code
Extensive work has been reported on the validation of the
MININEC. Numerous validation runs have been carried
out to meet the following requirements: (1) the segmen-
tation density, (2) thin-wire criteria, (3) small radius cal-
culation, (4) step changes in wire radius, (5) spacing of
wires, (6) loop antenna, and (7) monopoles and antennas
above ground.
4.7. Operation of CurrentsLU Decomposition
The operation is oriented around the menu shown below.
Here we describe the DOS version [19,20], but the Win-
dows version is also available [2123]:
MENU
1 -- COMPUTE/DISPLAY CURRENTS
2 -- CHANGE EXCITATION
3 -- CHANGE FREQUENCY
4 -- CHANGE LOADING
5 -- LOAD GEOMETRY
6 -- SELECT OUTPUT DEVICE
7 -- RETURN TO SYSTEM SUPERVISOR
0 -- EXIT TO DOS
SELECTION (1--7 OR 0) ?
4.8. Some Examples Using MININEC
4.8.1. Tee Antenna. Figure 7 shows the geometry of
the tee antenna fed from the base by a coaxial line. The
Wire 1
Wire 2
Wire 3
Wire 2
Overlapping pulse
Overlapping pulse
(a)
(b)
(c)
Overlapping pulse
Wire 1
Figure 6. Overlap scheme used at a multiple junction of wires:
(a) wire 1 with no end connections; (b) wire 2 overlaps onto wire 1;
(c) wire 3 overlaps onto wire 1.
y
z
x
Wire 2
Wire 3
Wire 1
0.04 m
0
.
1
2

m
0
.
1
2

m
Figure 7. Geometry of the tee antenna. Typical dimensions are
shown.
impedance calculations of the tee antenna using different
computer programs, including CURLU in MININEC and
have been compared with measurements [25].
4.8.2. Near and Far Fields. The near- and far-eld pro-
grams (FIELDS) calculate near and far elds using the
current distribution on the structure obtained by integral
equation formulations. The current distribution can be
computed using three programs: CURLU, CURTE, and
CURRO. The current distribution can be computed using
perfect and imperfect ground, although the real ground
corrections are applied to the far eld only. The real
ground correction is included in the form of reection co-
efcients for parallel and perpendicular polarizations. For
details, the reader is referred to Chap. 8 of Ref. 7. The
menu is given below. User input(UI) means the user is
expected to respond at that point.
MENU
1--COMPUTE NEAR FIELDS
2--COMPUTE FAR FIELDS
3--SELECT/CHANGE ENVIRONMENT
4--SELECT/CHANGE CURRENTS FILE
6--RETURN TO SYSTEM SUPERVISOR 7--EXIT TO DOS
SELECTION (1--6 OR 0) ? User Input
NAME OF INPUT CURRENT FILE? User Input (UI)
ELECTRIC OR MAGNETIC NEAR FIELDS (E/M) ? User
Input
X-COORDINATE Y-COORDINATE
INITIAL VALUE? UI INITIAL VALUE?
INCREMENT? UI INCREMENT?
NO. OF PTS? UI NO. OF PTS?
Z-COORDINATE
UI INITIAL VALUE? UI
UI INCREMENT? UI
UI NO. OF PTS? UI
PRESENT POWER LEVEL IS : CURRENT VALUE
CHANGE POWER LEVEL (Y/N) ? UI
NEW POWER LEVEL (WATTS) ? UI
Once the parameters are specied, the near- and far-
eld results are printed out in words. Figure 815 show
the near-eld characteristics of the monopole.
5. THE THICK LINEAR ANTENNA
The thin linear antenna is frequency-sensitive. In prac-
tical communication scenarios, the transponders use
10
1
10
2
10
3
10
4
Vertical electric field
at 1 m above perfect ground
1 kW radiated at 2.0 MHz
MININEC
Measured
Antenna:
10.67 m monopole
0.102 m wire radius
20-Segment model
NEC
20 16 12 8
Horizontal distance (m)
4 0
E
l
e
c
t
r
i
c

f
i
e
l
d

(
V
/
m
)
Figure 8. Monopole near elds: E
z
versus horizontal distance.
10
5
10
4
10
3
10
2
10
1
10
0
Near fields:
Magnitude of E
z
and E
r
at z = 1.5 m
Radiated power = 8.775 mW at f = 1 MHz
Antenna:
MININEC
MININEC
NEC
NEC E
z
E
r
Height = h = 75 m
Radius = a = 0.3 m
15-Segment model
500 400 300 200
Radial distance (m)
100 0
E
l
e
c
t
r
i
c

f
i
e
l
d

(
V
/
m
)
Figure 9. Monopole near elds: electric elds E
z
and E
r
versus
radial distance.
10
6
10
5
10
4
10
3
10
2
Near fields:
H[ component at z = 1.5 m
Radiated power = 8.775 mW at f = 1 MHz
Antenna:
Height = h = 75 m
Radius = a = 0.3 m
15-Segment model
500 400 300
MININEC
NEC
200
Radial distance (m)
100 0
M
a
g
n
e
t
i
c

f
i
e
l
d

(
A
/
m
)
Figure 10. Monopole near elds: H
f
versus radial distance.
wideband signals to increase the channel capacity and
therefore needs antennas that can handle a large band of
frequencies. One way of increasing the bandwidth is to use
electrically thick wires.
A thick cylindrical dipole (Fig. 16) is the inexpensive
way to increase the bandwidth of linear antennas. The
increase in thickness leads to circumferential component
I
f
of the otherwise linear current. This can be handled
with the integral equation formulation. Figures 17a and
17b, respectively, show the variation of input resistance
and reactance of the dipole with l/2a ratios 25 (thick), 50,
and 10,000 (thin), where 2a is the diameter of the wire.
6. THE SLEEVE DIPOLE
The structure that closely resembles an asymmetric dipole
is the sleeve dipole (Fig. 18), which is a base-fed monopole
0
1
2
3
4
5
500 450 400 350 300 250
Near fields:
Vertical and radial components of E-fields
z = 1.5 m
Power radiated = 8.775 mW
f = 1 MHz
Antenna:
Height = h = 75 m
Radius = a = 0.3 m
15-Segment model
Radial distance (m)
MININEC compared to NEC
E
z
E
r
200 150 100 50 0
P
e
r
c
e
n
t

d
i
f
f
e
r
e
n
c
e

i
n

e
l
e
c
t
r
i
c

f
i
e
l
d
s
Figure 11. Percentage difference in electric elds E
z
and E
r
ver-
sus radial distance.
1.5
1
0.5
0
0.5
1
500 450 400 350 300 250
Near fields:
H
at z = 1.5 m
Radiated power =
8.775 mW at
f = 1 MHz
Antenna:
Height = h = 75 m
Radius = a = 0.3 m
15-Segment model
Radial distance (m)
MININEC compared to NEC
200 150 100 50 0
P
e
r
c
e
n
t

d
i
f
f
e
r
e
n
c
e

i
n

H
[

[
component
Figure 12. Monopole near elds: percentage difference in H
f
versus radial distance.
10
1
10
2
10
3
10
4
10
5
10
0
10
1
180 144 108
Antenna:
Height = h = 75 m
Radius = a = 0.3 m
15-Segment model
Antenna:
z - Component at wire surface
Radiated power =
8.775 mW at
f = 1 MHz
NEC
MININEC
Vertical distance (m)
72 36 0
V
e
r
t
i
c
a
l

e
l
e
c
t
r
i
c

f
i
e
l
d

(
V
/
m
)
Figure 13. Monopole near elds: vertical electric eld versus
vertical distance.
10
5
10
4
10
3
10
2
10
1
10
0
10
1
180 144
Near fields:
x-Component of E field
Radiated power =
8.775 mW at
f =1 MHz
108 72
NEC
MININEC
Antenna:
Height = h = 75 m
Radius = a = 0.3 m
15-Segment model
36 0
R
a
d
i
a
l

c
o
m
p
o
n
e
n
t

o
f

e
l
e
c
t
r
i
c
a
l

f
i
e
l
d

(
V
/
m
)
Figure 14. Radial component of electric eld at wire surface
versus vertical distance.
on a ground plane. The outer conductor of the coax line is
extended to give wideband characteristics apart from im-
pedance control and mechanical strength. Another way of
increasing the bandwidth will be to coat the metal wire
with a low-loss dielectric or just push the metal wire into a
sleeve (a thin dielectric pipe). It can be fed symmetrically,
but the problem of impedance matching and using a prop-
er matching network will increase the complexity and cost
of the device. For further discussion on sleeve dipoles, the
reader is referred to Ref. 1.
7. COMPUTER CODES
7.1. A Summary of Different Available Codes
The following subsections summarize the scope of the
codes currently used for linear antenna analysis.
7.1.1. The Electromagnetic Surface Patch Code. The
electromagnetic surface patch (ESP) code version IV
[26], which uses the method of moments (MoM), is writ-
ten for the analysis of radiation and scattering from three-
dimensional geometries. These geometries include the in-
terconnection of thin wires and perfectly conducting and
thin dielectric polygonal plates. This code works for open
as well as closed surfaces since the formulation is based on
electric eld integral equation. It uses polygonal plates
modeling, which can generate an object as complex as an
aircraft with only 12 polygonal plates. The ESP code is
capable of doing the following:
1. It can model an arbitrary thin wire by using a num-
ber of piecewise straight segments.
2. It can be model an arbitrary perfectly conducting
surface by a combination of polygonal plates. The
plates can be of thin dielectric sheets, which have
been modeled using impedance approximation.
3. The code incorporates the wirejunction with the
restriction that the junction must be at least one-
tenth wavelength away from the nearest edge; it can
take care of several plates generated from a common
edge.
In order to use this code, the user species the (x,y,z) co-
ordinates of the corners of each plate by entering the (x,y,z)
coordinates of the corners of each plate and the maximum
segment size in wavelengths for calculating the MoM
modes. The code automatically takes care of the frequen-
cy independence of the models.
In summary, the ESP code can treat a variety of geom-
etries: (1) thin wires with nite conductivity and lumped
loads, (2) perfectly conducting or thin dielectric polygonal
plates (30 wireplate junctions) at least 0.1l from the edge
of a plate, (3) plateplate junctions, including several
plates of different sizes that intersect along a common
edge, and (4) excitation by either a voltage generator or an
incident plane wave.
ESP can compute the key characteristics of engineering
interest of an antenna, namely, current distributions, an-
tenna input impedance, radiation efciency, mutual cou-
pling, near- or far-zone radiation patterns for all
polarizations, and near- or far-zone back, bistatic, and for-
ward scattering patterns (full scattering matrix). These
are still bugs in the ESP code, and new features are being
10
8
10
7
10
6
10
5
10
4
10
3
10
2
10
1
180 144 108 72
Antenna:
Height = h = 75 m
Radius = a = 0.3 m
15-Segment model
MININEC
NEC
Near fields:
At x = 0.3 m
Radiated power =
8.775 mW at
f = 1 MHz
36 0
M
a
g
n
e
t
i
c

f
i
e
l
d

H
[

(
A
/
m
)

Figure 15. Monopole magnetic eld H
f
at surface versus vertical
distance.
y
x
A
B
R
r
j
dz
z
l /2
l /2
d = 2a
0
z
Figure 16. The center-fed thick dipole. A thick dipole has a fre-
quency characteristic over a much wider band than a thin anten-
na. The diagram shows how such a dipole will receive power from
the feeder network.
added to the codes. Nevertheless, it should be possible to
customize the code for a specific application.
7.1.2. The Mininumerical Electromagnetic Code
(MININEC). The MININEC [20,21] is a computer program
prepared in BASIC language using the method of mo-
ments for the analysis of linear antennas. It uses a Galer-
kin procedure [19] to solve for wire currents using an
integral equation formulation that relates the vector and
scalar potentials to the electric eld. This formulation can
easily be programmed for use in microcomputers. The code
can solve for impedance and currents on arbitrarily wires
including congurations with multiple wire junctions. The
code incorporates lumped impedance loading and near-
zone electric and magnetic elds when the wire is in free
space as well as over a perfectly conducting ground. MI-
NINEC is written in a IBM PC-DOS-compatible BASIC
language.
8. FEED FOR LINEAR ANTENNAS
In all practical antennas, there exists a mechanism of
feeding the antenna to launch the desired current distri-
bution. This requires a nite gap with a nonzero current
and alters the normally assumed current distribution.
Hence, a correction factor should be introduced to accu-
rately model the feedpoint current. Several methods of
feeding half-wave antennas are shown in Figs. 19a19d.
Normally, since the complex input impedance of the dipole
is quite different from the real input resistance (under
high frequency approximation), there is a need to use a
matching network to match the input impedance to the
feeder and keeping the voltage standing-wave ratio
(VSWR) close to unity in the frequency range of opera-
tion. Figure 19a shows the balanced line-type center feed.
Figure 19b shows a delta match or shunt feed feeder
that provides a good impedance mismatch and low VSWR
600
400
200
0
200
400
0
200
400
600
800
1000
0.8 1.0 1.2 1.4 1.6 0.6 0.4 0.2 0.0
Dipole length (wavelengths)
(a)
0.8 1.0 1.2 1.4 1.6 0.6 0.4 0.2 0.0
Dipole length (wavelengths)
(b)
l /d = 25
(Moment method)
l /d = 50
(Moment method)
l /d = 10
4
(Sinusoidal current)
I
n
p
u
t

r
e
s
i
s
t
a
n
c
e

(

)
I
n
p
u
t

r
e
a
c
t
a
n
c
e

(

)
l /d = 25
(Moment
method)
l /d = 50
(Moment
method)
l /d = 10
4
(Sinusoidal
current)
Figure 17. The input impedance of wire di-
poles with dipole length in wavelengths: (a) in-
put resistance in ohms; (b) input reactance in
ohms. Every antenna presents an input com-
plex impedance at its feed terminals. It is very
important to know how the real and imaginary
parts of the impedance vary with parameters
so that antenna can be matched to the feeder.
on the feeder line. The simplest feeding arrangement is
the end feed (Fig. 19c), where the vertical transmission
line radiates energy. The conguration in Fig. 19d can be
made efcient by tapping the vertical line at an appropri-
ate location.
To account for the nonzero current at the nite gap feed
terminals and consequent distortion in the current distri-
bution.
The modied current is given by [28]
I
z
x
0
; y
0
; z
0
a
z
fI
0
sinkl=2 z
0
jaI
0
coskz
0
coskl=2g
for 0z
0
l=2
a
z
fI
0
sinkl=2 z
0
jaI
0
coskz
0
coskl=2g
for l=2z
0
0
_
_
62
z = 0
z = h
z
Figure 18. The sleeve dipole. Apart from making the wire thick,
another way of increasing the bandwidth of a thin wire is to put it
in a sleeve, which is basically a dielectric coating, and terminate
by a ground plane on one side.
(a) (b)
(c) (d)
Figure 19. Methods of exciting linear antennas: (a) center feed;
(b) delta match or shunt feed; (c) end feed; (d) impedance-matched
taping. Excitation mechanisms are very important in order to
launch maximumpower. Not shown are matching transformers to
ensure maximum power transfer.
0
2
4
6
8
L
2a
2b
2
= 8"
= 0.025"
= 0.146"
= 2.36
0
Dielectric coated
dipole
10
12
14
16
Calculated
Measured
at O.S.U.
18
0.8 0.7 0.6 0.5 0.4
Dipole length
L
/ z
z Dipole length
L
/
(a)
0.3 0.2 0.1 0
G

(
m
m
h
o
s
)
6
4
2
0
2
4
6
8
10
0.8 0.7 0.6 0.5 0.4
(b)
0.3 0.2 0.1 0
B

(
m
m
h
o
s
)
L
c
Figure 20. Comparison of measured and calculated admittance
values for a coated dipole length L8 in.; 2a0.025 in.; 2b
0.146 in.; e
2
2:3e
0
. The gures show the bandwidth of antennas:
(a) conductance; (b) susceptance of an insulated dipole versus
length in wavelengths.
where a the coefcient dependent on the length of the
gap and the antenna. If the radius of the wire and the gap
are small the value of this coefcient is small.
For a half-wave dipole, we obtain
I
0
x
0
; y
0
; z
0
a
z
I
0
1ja coskz
0
for 0 z
0
l=4
The shape of the current does not change for a half-
wave dipole, but it does for other lengths, such as a wave-
length.
9. BANDWIDTH OF LINEAR ANTENNAS
The bandwidth of an antenna is the frequency range over
which the performance of the antennas meets a specific
requirement, such as that the gain does not fall below 3dB
of the maximum, or the midfrequency value. There is no
unique characterization of bandwidth since the different
properties such as input impedance, radiation pattern,
polarization, and gain of an antenna vary in entirely dif-
ferent ways in a frequency range. The definition meets the
requirement of specific application. Nevertheless, the
bandwidth is dened in three different ways: (1) half-pow-
er bandwidth is the frequency range within which the
gain does not fall by more than 3 dB, (2) the percentage
bandwidth normally dened for a narrowband antenna is
dened as the bandwidth (the difference between upper
and lower frequencies of operation) divided by the center
frequency and then multiplied by 100, and (3) for wide-
band- or frequency-independent antennas it is the ratio of
higher and lower frequencies of operation.
The thin linear antenna that we have dealt with in this
article is based on the assumption that the radius is small
compared with the wavelength of operation. This neces-
sarily means that the current on the wire is linear and has
no tangential component. In a practical situation an an-
tenna has to operate in a frequency band. This assumption
is no longer valid and the thin linear antenna becomes
frequency-sensitive. Therefore, something needs to be
done to the thin antenna so that it develops the capabil-
ity of handling a wide frequency band.
The way of achieving wider bandwidth will be to use
electrically thick dipoles or to coat thin metallic wires with
lossy dielectrics. A thick dipole and a dielectric coated di-
pole with their centrally located feeding source are shown
in Figs. 20a and 20b, respectively. The effect of coating the
thin linear antenna with a layer of electrically and mag-
netically lossless and lossy material is discussed is two
papers in (28) and (29) and summarized in (1). The ana-
lytical techniques used is a moment method solution. Two
parameters, P and Q [28,29], involving the electrical and
magnetic parameters, inner and outer radii, are found to
be of interest in moment method solutions and help in de-
signing antenna characteristics using coating of electri-
cally and magnetically lossless and lossy material. They
are given by
P
e
r
1
e
r
_ _
ln
b
a
_ _
63a
Qm
r
1 ln
b
a
_ _
63b
where aradius of the thin wire, bathickness of the
coating, e
r
the relative permittivity of the coating or the
medium in which the thin wire is embedded, and m
r
rel-
ative permeability of the coating or the medium in which
the thin wire is embedded.
It turns out that the effects of increasing the real part of
both P and Q are to increase the peak value of input im-
pedance, to increase the electric length which lowers the
resonant frequency and to reduce the bandwidth of oper-
ation. The effects of increasing the imaginary part of P and
Q are to decrease the peak value of input impedance, to
decrease the electrical length which increases the reso-
nant frequency, and to increase the bandwidth. This
means a proper choice of a lossy dielectric with maximum
imaginary parts of P and Q and minimum real parts of P
and Q can achieve an optimum bandwidth. But this will
be at the cost of antenna efciency because of the lossy
coating.
BIBLIOGRAPHY
1. C. A. Balanis, Antenna Theory: Analysis and Design, Wiley,
New York, 1997.
2. C. A. Balanis, Advanced Engineering Electromagnetics, Wiley,
New York, 1979, Chap. 12, pp. 670742.
3. J. Moore and R. Pozer, eds., Moment Methods in Electromag-
netics: Techniques and Applications, Research Studies Press,
Letchworth, UK, 1984; Wiley, New York, 1984.
4. W. A. Imbraile, Applications of the Method of Moments to
Thin-Wire Elements, in R. Mittra, ed., Computer Techniques
for Electromagnetics, Pergamon, New York, 1973, Chap. 1.
5. C. A. Balanis, Antenna Theory: Analysis and Design, Wiley,
New York, 1997, Chaps. 4 and 8.
6. G. A. Thiele, Wire antenna, in R. Mittra, ed., Computer Tech-
niques for Electromagnetics, Pergamon, New York, 1973,
Chap. 2, pp. 770.
7. J. Rockway et al., The MININEC System Microcomputer
Analysis of Wire Antenna, Artech House, Norwood, MA, 1988.
8. E. K. Miller and F. J. Deadrick, Some computational aspects
of thin-wire modeling, in R. Mittra, ed., Numerical and
Asymptotic Techniques in Electromagnetics, Springer-Verlag,
New York, 1975, Chap. 4, pp. 89127.
9. J. J. H. Wang, Electromagnetics: Formulation and Computer
Solution of Integral Equation, Wiley, New York, 1991, Chap. 5.
10. E. K. Miller, L. Medguesi-Mitchang, and E. H. Newman, Thin
Wires, Computational Electromagnetics: Frequency-Domain
Method of Moments, IEEE Press, New York, 1992, Part 3.
11. R. Mittra, ed., Numerical and Asymptotic in Electromagnet-
ics, Springer-Verlag, Berlin, 1975.
12. C. A. Balanis, Advanced Engineering Electromagnetics,
Wiley, New York, 1979, pp. 150.
13. K. K. Mei, On the integral equations of thin wire antennas,
IEEE Trans. Anten. Propag. AP-13 (1965).
14. C. M. Butler and D. R. Wilton, Analysis of various techniques
applied to thin-wire scatterers, IEEE Trans. Anten. Propag.
AP-23:534540 (1975).
15. G. J. Burke and A. J. Poggio, Numerical Electromagnetic Code
NEC-Method of Moments, Tech. Document 116, Naval Ocean
Systems Center, San Diego, CA, Sept. 6, 1982.
16. IEEE Standard definition for terms for antennas (IEEE Std.
145-1983) IEEE Trans. Anten. Propag. AP-17 (1969); AP-22
(1974); AP-31(Part II) (1983).
17. A. K. Bhattacharyya and D. L. Sengupta, Radar Cross Sec-
tion Analysis and Control, Artech House, Norwood, MA, 1991.
18. J. H. Richmond, Proc. IRE 53:796 (1965).
19. R. F. Harrington, Field Computation by Moment Methods,
Macmillan, New York, 1968.
20. J. C. Logan and J. W. Rockway, The New MININEC Version 3:
A Mininumerical Electromagnetic Code, Tech. Document 93B,
Naval Ocean Systems Center, San Diego, CA, Sept. 1986.
21. A. J. Julian, J. C. Logan, and J. W. Rockway, MININEC: A
Mininumerical Electromagnetics Code, NOSC TD 516, Sept.
1982.
22. Windows Version of MININEC, EM Scientific, available by e-
mail: emsi@aol.com.
23. Windows versions of NEC2 and NEC4, Nitany-Scientific
Available (online), www.nittany-scientific.com.
24. A. R. Djordjevic et al., AWAS for Windows: Analysis of Wire
Antennas and ScatterersSoftware and Users Manual, Ar-
tech House, Norwood, MA, 1995.
25. S. Prasad and R. W. P. King, Experimental study of inverted
L-, T- and related transmission line antenna, J. Res. Natl. Bur.
Standard-D Ratio Propag. 65D(5) (1961).
26. E. H. Newman, A Users Manual for Electromagnetic Surface
Patch Code ESP, OSUESL, Dept. Electrical Engineering, Re-
port 713602-1, prepared under Contract DAAG-39-81-K-0020
for the Department of the Army, U.S. Army Res. Ofce, Res.
Triangle Park, NC, July 1981.
27. J. H. Richmond, Radiation and Scattering by Thin-Wire
Structures in the Complex Frequency Domain, OSU Research
Foundation Report RF-2902-10, 1973.
28. S. A. Schelkmoff and H. T. Fris, Antennas: Theory and Prac-
tice, Wiley, New York, 1952, pp. 229244, 351353.
29. J. H. Richmond and E. H. Newman, Dielectric coated wire
antennas, Radio Sci. 11(1):1320 (1976).
30. J. Y. P. Lee and K. G. Balmain, Wire antennas coated with
magnetically and electrically lossy material, Radio Sci.
14(3):437445 (1979).
LOCAL-AREA NETWORKS (LANs)
JOSEPH B. EVANS
University of Kansas
Lawrence, Kansas
Local-area networks (LANs) are data communications
networks that are restricted in extent to an ofce, home,
building, or in some cases, areas as large as a college or
university campus. Because of the spectacular growth in
networking, LANs can be found deployed in almost every
organization. Various established and evolving technolo-
gies are used in LANs, based on physical facilities ranging
from copper and optical ber to radio. The characteristics
of commonly used LAN technologies will be discussed
below.
1. LAN TOPOLOGIES
LANs can be logically organized in several topologies, the
most popular of which are the bus, star, and ring.
In the bus structure, illustrated in Fig. 1, nodes (com-
puters, printers, or similar devices) are interconnected to
a common, shared physical resource, typically a wire or
cable. This topology is inexpensive, since wiring expenses
are shared among the nodes. Unfortunately, this scheme
involves sharing the limited bandwidth resources of the
bus, and can also be somewhat unreliable, as bus failures
in the vicinity of one node can affect the others on the same
bus. IEEE 802.3 10base5 and 10base5 Ethernet are ex-
amples of networking standards based on a bus topology
at the physical layer. Despite its shortcomings, this re-
mains a common topology because of the simplicity of de-
ployment. Because of the nature of the wireless medium, it
is the topology used by wireless LANs such as those based
on IEEE 802.11.
An alternative is the star topology, shown in Fig. 2, in
which each node has dedicated resources to some central
switching site. This has the advantage of dedicated band-
width to the interconnection point, but the attendant ca-
bling costs are often higher than those for the bus
topologies. Switched Ethernet and asynchronous transfer
mode (ATM) are examples of networking standards based
on a star topology. There is increasing dependence on star
topologies because the limited bandwidth on a cable is not
shared and trafc is not subject to internode arbitration
delays for access to the medium.
Another option is the ring topology, shown in Fig. 3, in
which each node is interconnected to its neighbor. The
IEEE 802.5 Token Ring and the Fiber Distributed Data
Interface (FDDI) are examples of networking standards
based on a ring topology. This topology shares many of the
advantages and disadvantages of the bus topologyinex-
pensive wiring, but with reliability problems if the ring
should be broken. Ring-based LANs have been designed to
overcome the reliability issues by using counterrotating
rings (e.g., FDDI). Depending on the protocols in use,
bandwidth in a ring-based network can be reused (since
the ring is not physically contiguous), and hence such a
Figure 1. Bus topology.
2352 LOCAL-AREA NETWORKS (LANs)
topology can have a capacity greater than the equivalent
bus.
2. IEEE 802 LAN STANDARDS
Much of the growth in deployment of LAN technology can
be attributed to the standardization of selected technology
options, which has enabled multivendor interoperability
and has spawned a highly competitive market. The IEEE
802 LAN standards are among the most widely used data
protocols yet developed.
The IEEE 802.2 standard species the logical link con-
trol (LLC) protocols used by the other IEEE LAN stan-
dards. IEEE 802.2 allows the lower-level protocols to
interface with higher-level protocols in a consistent man-
ner. Using this approach, for example, the Internet Pro-
tocol (IP) need not know the type of underlying hardware
being used on a particular host, which implies that soft-
ware can be simplied and made more reliable. Note that
certain other protocol suites (e.g., IP over ATM) use the
IEEE LLC SAP (service access point) codes for protocol
multiplexing and demultiplexing, so that similar benets
can be obtained. IEEE 802.2 provides several services;
which services are used and the extent to which they are
used depends on the needs of the other protocols involved.
The IEEE 802.3 standard has been one of the most
successful in the IEEE LAN suite. This standard describes
the Carrier Sense Multiple Access with Collision Detec-
tion (CSMA/CD) protocol that forms the basis for the
Ethernet family (note that the 802.3 standard and Ether-
net differ slightly, but can be made to interoperate). The
IEEE 802.3 standard consists of several related protocols,
for different physical media. Included are the original
10base5 standard for CSMA/CD on 50-O-thick coaxial ca-
ble, the 10base2 standard for lighter 50-O coaxial cable,
and the 10baseT standard for unshielded twisted-pair ca-
bles. No longer commonly used are the 1base5 StarLAN
standard and the 10broad36 standard for more widely
dispersed networks. Fiber extension options are available
for distributed site interconnection (within protocol dis-
tance limits). Ethernets can be found in almost all corpo-
rate data networks. Although networks based on the 10-
Mbps (megabits per second) data rate are still deployed,
higher-rate Ethernet protocols are becoming increasingly
popular, particularly 100-Mbps Fast Ethernet, Gigabit
Ethernet, and most recently 10-Gbps Ethernet. The 100-
Mbps and 1-Gbps Ethernet varieties utilize both ber and
copper; the latter are restricted to server rooms and other
facilities with small distances between nodes.
The IEEE 802.4 standard species the token bus pro-
tocol. This protocol has been the basis for several net-
working technologies, including the MAP/TOP suite.
Multiple physical layers are dened for token bus on 75-
O coaxial cable, including systems at 1, 5, and 10 Mbps.
These are all broadband systems. The original 1-Mbps
system was quite popular because of its low cost and rel-
ative simplicity.
The IEEE 802.5 standard species the token ring pro-
tocol. This standard was extensively deployed in PC-based
networks and became second only to Ethernet in ubiquity,
although it is no longer widely used. It uses unshielded
Figure 2. Star topology.
Figure 3. Ring topology.
LOCAL-AREA NETWORKS (LANs) 2353
twisted-pair cabling, with data rates at 4 and 16 Mbps. It
has several very desirable features, including robust be-
havior in the presence of high trafc loads and bounded
delay (to transmit) times.
3. PROTOCOL LAYERING
For standardization purposes, networking protocols are
most often conceptually partitioned into several layers. In
the case of LAN technologies, the physical layer (PHY),
media access layer (MAC), and logical link layer (LLC) are
commonly specied. The latter two are often grouped to-
gether to form the data link layer in standard layering
schemes.
3.1. PHY Layer
The PHY, or physical layer, is the lowest layer of a protocol
stack. The standards for this layer typically describe the
medium to be used (e.g., cable, ber, wireless), modulation
schemes, and encoding schemes used to transmit infor-
mation across the medium.
The PHY layers of LAN protocols generally fall into two
categories: baseband and broadband. A baseband PHY
layer is one in which the information bearing signals are
digital signals, typically encoded using simple level-based
keying, Manchester encoding, or differential Manchester
encoding. This is the most common type of PHY layer in
current LANs, as it is relatively inexpensive and suf-
ciently robust for most local environments. The disadvan-
tages are distance limitations, typically 100m to at most
1000 m on copper, and bandwidth no more than about
1 Gbps over copper using current technologies. Baseband
techniques may be used over optical ber at much greater
distances and rates, but with the attendant installation
and network equipment costs. For typical LAN installa-
tions, however, baseband systems on copper are sufcient.
Encoding schemes are another key element of the PHY
layer. A variety of encodings, tailored to the physical me-
dium for a given protocol, have been developed. Some typ-
ical encodings are depicted in Fig. 4. These can be broadly
classed as non-return-to-zero (NRZ) techniques and
biphase techniques.
The conceptually simplest encodings are the NRZ
methods. In the NRZ-level approach, for example, zeros
(0s) are encoded as low voltage level, and ones (1s) are
encoded as a high voltage level. In optical ber systems,
the corresponding encodings may be that ones are the
presence of optical power; zeros, the lack of light. In the
NRZI (NRZ with invert on ones) approach, a transition
(either falling or rising edge) denotes a one, and the lack of
a transition signies a zero.
While simple, the NRZ schemes have several shortcom-
ings. Most significantly, recovery of bit timing at the re-
ceiver can be difcultthe moment in time at which to
sample a bit to determine whether it is a zero or one is
seldom apparent in the presence of noise and other such
impairments. A technique that provides an unambiguous
timing reference is highly desirable. Furthermore, the oc-
currence of a long string of zeros or ones can result in an
undesirable DC voltage bias on the transmission medium,
which may cause threshold-related errors and problems
with the use of transformers. There are several approach-
es to resolving these related problems, which center
around the need for signal transitions.
The 4B/5B (4-byte/5-byte) and related techniques (4B/
6B and 8B/10B are also common) involves guaranteeing
sufcient transitions by inserting extra bits into the signal
stream. Data symbols, 4 bits in this case, are mapped into
a 5-bit code, which is then transmitted using NRZI, for
example. This is illustrated in Table 1. An inspection of
this table will prove that strings with a maximum of three
consecutive zeros are possible, even when codewords are
concatenated. Multiple ones are not an issue if NRZI is
used for transmission, as ones force a transition to occur.
The cost of a 4B/5B mapping, of course, is that only 80%
efciency is possible.
The biphase encodings are benecial with respect to
signal balance maintenance, and bit timing recovery is
particularly easy to implement in this scheme. These are
based on signal transitions at a rate double that of the bit
rate. A transition (rising edge or falling edge) is guaran-
teed to occur at the center of the bit period. The absence
0 0 0 1 1 1 1 0 0
Data
NRZ-level
NRZI
Manchester
Differential
Manchester
Figure 4. Baseband PHYencoding schemes.
Table 1. 4B/5B Encoding
Data Symbol Codeword
0000 11110
0001 01001
0010 10100
0011 10101
0100 01010
0101 01011
0110 01110
0111 01111
1000 10010
1001 10011
1010 10110
1011 10111
1100 11010
1101 11011
1110 11100
1111 11101
of such a transition can be used as an error detection
method.
In Manchester encoding, a zero is encoded as a rising
edge at the center of the bit period, and a one as a falling
edge at such a time. The encoding mechanism can be im-
plemented as an exclusive-OR operation between the data
and the clock. This is the encoding used for most of
the original IEEE 802.3 protocols (10base5, 10base2,
10baseT). Differential Manchester encoding uses the mid-
period transition for a clocking reference only, and uses
the presence (denoting a zero) or absence (denoting a one)
of a transition at the beginning of the bit period to encode
the information. This is the method used for the IEEE
802.5 token ring standard.
The primary disadvantage of biphase signaling is that
transitions happen at twice the data rate, which means
that the bandwidth required is greater than that of the
equivalent NRZ system, and the hardware need operate
twice as quickly. The former is particularly critical in
wireless systems.
A broadband PHY layer is one in which the infor-
mation is coupled into the medium by analog signals,
which are modulated by some carrier, and encoded using
frequency shift keying (FSK), amplitude shift keying
(ASK), phase shift keying (PSK), or some similar scheme.
This type of PHY layer is most often used in situations
where longer distances need be served or additional
bandwidth is required. Much greater bandwidths
may be supported on one cable using broadband schemes,
as multiple frequencies can be used. The primary
disadvantage of the broadband approach is the cost of
modulators, demodulators, and the associated analog
hardware.
3.2. MAC Layer
The MAC, or media access layer, is used to arbitrate access
to the PHY layer. For example, in the case of Ethernet,
there is a shared medium (cable) that must be used by
several nodes, and only one of the nodes can be permitted
to access the cable at a particular time.
The MAC layer inuences the effective throughput over
a given physical layer and should be efcient in its use of
the available bandwidth. This includes minimizing the
overhead due to factors such as protocol headers and
deadtime between transmissions, while at the same time
maximizing the successful transmissions on a busy shared
medium network. In addition, the MAC layer is often de-
signed to ensure that errors are not propagated to the
higher-layer protocols.
Various MAC schemes have been developed for the
LAN protocols. The two most common are the Carrier
Sense Multiple Access with Collision Detection (CSMA/
CD) protocol used in Ethernet, the token ring protocols,
and the token bus protocol.
The CSMA/CD MAC protocol involves detecting the use
of the medium by another station by checking the state of
the carrier. If a station has data to transmit, it rst at-
tempts to verify whether the medium is unused. If it is
available, the station transmits. If the medium is not
available, the station waits until it becomes idle and
then immediately begins to transmit (note that this is
the IEEE 802.3 solution, but other options are possible in
the general case of CSMA). The success or failure of trans-
missions is monitored on the shared medium, and if a
transmission is unsuccessful, that is, a collision is detect-
ed, the station waits a prescribed random amount of time
(binary exponential backoff) and attempts to transmit
again. This procedure is repeated until the transmission
is successful, or the limit to the number of transmission
attempts (16 in IEEE 802.3) is reached. CSMA/CD is sim-
ple, inexpensive, and performs well under light loads. Un-
fortunately, it can perform poorly under heavy loads, and
be sensitive to physical layer errors. The CSMA/CD pro-
tocol, however, need not be invoked in a full-duplex
switched Ethernet network. In a full-duplex switched net-
work, each segment from switch to station or from switch
to switch is its own isolated domain, and no collisions that
require CSMA/CD can occur.
Token ring protocols use a token to arbitrate access to
the transmission medium. A token is a small frame that is
exchanged between stations in order to gain the right to
transmit. If a station has data to transmit, it waits until a
token is seen on the medium. This station then modies
the token and appends the necessary elds as well as its
data. When this frame returns around the ring to the
originating station, it is purged from the medium. When
data transmission is complete, the station inserts a new
token onto the ring. Token rings support fair, controlled
access to the medium, and perform well under heavy load
conditions. A disadvantage is the need for careful token
maintenance, particular in the presence of errors. Several
varieties of token ring exist; some of these will be dis-
cussed in subsequent sections.
The token bus protocol is closely related to the token
ring, but with an underlying physical bus topology. The
token exchange mechanism, however, does in fact use a
logical ring for token passing. This logical ring is simply
an ordering of stations on the bus. Once the logical ring is
in place, token passing can proceed as in a ring-based sys-
tem. This system provides controlled access to the bus,
and is robust under heavy loads. One of the disadvantages
of this approach is that ring initialization and mainte-
nance is more complex than in a physical ringthe or-
dering of stations must be determined through some
algorithm, and stations additions and deletions must be
managed.
3.3. Logical Link Control (LLC) Layer
The LLC, or logical link control layer, can be viewed as the
upper part of the data link layer. It is used to provide data
services to the higher layers. In particular, two types of
services, connectionless and connection-oriented, are de-
ned. In LANs supporting complex higher layer protocols,
such as TCP/IP, only the simplest LLC services are com-
monly used.
An example of an LLC protocol is the IEEE 802.2 layer.
This provides both connectionless and connection-oriented
services. The unacknowledged connectionless service pro-
vides simple datagram support for the multiplexing
and demultiplexing of higher layer protocols. In addition,
a connectionless service with acknowledgments (for
monitoring systems, for example) is supported, as well
as a connection-oriented service that furnishes ow con-
trol and error recovery capabilities based on the lower-
layer CRC (cyclic redundancy check) and a go-back-N
strategy. The IEEE 802.2 LLC frame format is depicted in
Fig. 5. The SAP and SSAP (direction and source service
access point) elds are used to indicate the service type
[e.g., IP or IPX (Internet protocol and internetwork packet
exchange)] to higher layers. The control eld is used for
the LLC service support, including indication of the type
of service.
4. ETHERNET (IEEE 802.3)
The Ethernet, and the closely related IEEE 802.3 stan-
dard, has been one of the most successful LAN protocols
developed to date. This technology is based on CSMA/CD,
and takes a variety of forms at the PHY layer. A typical
Ethernet installation is depicted in Fig. 6.
The Ethernet frame format is illustrated in Fig. 7. The
preamble is used for frame delineation. The destination
and source address elds (48 bits each) are globally unique
identiers for each Ethernet adapter, and are used for
station-to-station communication, as well as broadcast (all
ones) and multicast (the rst bit is one). It should be noted
that the 48-bit addresses used in Ethernet have become a
common feature in IEEE 802-based LANs. The type eld
(Ethernet) can be used for higher-layer demultiplexing, as
in an LLC protocol. The length eld (IEEE 802.3) is used
to aid in end-of-frame detection. A 32-bit CRC is used for
error detection, and is followed by a postamble for end-of-
frame detection.
5. TOKEN-PASSING BUS (IEEE 802.4)
The IEEE 802.4 token bus standard has been widely used
in manufacturing systems and early ofce automation
products. Because it is based on a broadband physical me-
dium, it is somewhat more resistant to the low-frequency
EM noise that might arise on a factory oor.
A token-passing bus is a LAN with a bus topology that
operates of the principle that a token will be received prior
to the transmission of data by a station. The token bus
frame format includes a preamble, frame control byte for
denoting whether a particular frame is a token or data,
the destination and source addresses (48 bits, as in 802.3),
the data (an LLC frame), a error detection eld (CRC-32,
as in 802.3), and the postamble.
The token bus operates by rst establishing a logical
ring that overlays the physical bus topology. Station ad-
ditions and deletions require reconguration of the logical
ring. When a token is received, a station is permitted to
transmit multiple packets, until its token holding time has
expired. The token bus offers optional support for multiple
classes of service through the use of complex timer spec-
ications that enable per-class bandwidth guarantees.
Support for simpler nontoken stations is included to al-
low low cost devices to respond to polling requests using
this medium.
6. MAP/TOP
MAP is the Manufacturing Automation Protocol devel-
oped by General Motors Corporation for communication
among automated manufacturing devices, including robo-
tic equipment and the associated controllers. It was pri-
marily designed to support communication between very
different sorts of devices, in real time with low, predictable
DSAP SSAP Control Data
8 8 8-16 n
Figure 5. IEEE 802.2 LLC frame format.
Preamble
Destination
address
Source
address
Type or
length
Data CRC Postamble
64 48 48 16 32 8
Figure 7. Ethernet frame format.
delays. It supports applications as varied as word process-
ing and equipment telemetry (temperature measurement,
for example).
TOP is the Technical and Ofce Protocol developed by
Boeing Corporation for communication between ofce au-
tomation devices such as word processing systems and
printers. Interoperability between devices from a variety
of manufacturers was a key design goal of this protocol.
The MAP/TOP protocol suite is based on the IEEE
802.4 token bus protocols. As such, MAP/TOP networks
are often interconnected with some variety of token pass-
ing network for ease of interface design.
7. TOKEN RING (IEEE 802.5)
The token ring protocol was widely deployed in networks
based on PCs before Ethernet became the dominant LAN
technology. Token ring operates on the principle of the ex-
change of a token to a station before it is permitted to
transmit. Only one token is allowed on the ring at one time.
The IEEE 802.5 token ring frame formats are illustrat-
ed in Fig. 8. The rst format is used for token frames, and
only includes start and end delimiters, and the access con-
trol eld with priorities and reservation information. The
second format includes start and end delimiters, a frame
control word for optional LC support, source and destina-
tion addresses (in 802.3 format), the LLC (data) frame, a
CRC-32, and a frame status word used by transmitting
stations to verify reception.
8. OTHER TOKEN RINGS
Another example of token ring technology in wide use
today is the Fiber Distributed Data Interface (FDDI) stan-
dard. This technology supports multiple packets on the
ring at one time, with rates of 100 Mbps. Provisions are
made for multiple service classes (synchronous and asyn-
chronous) with differing throughput and delay require-
ments. Further, reliability support is provided through the
capability for optional redundant counter-rotating rings,
which can mask a station or ber failure.
8.1. Slotted Ring
Slotted-ring technology uses multiple slots that rotate
around the ring to arbitrate access. Each slot is a small
frame that can be marked empty or full. When an empty
slot arrives at a station with data to transmit, the slot is
marked full and data is injected. The slot is marked empty
when it returns around the ring to its source. A given sta-
tion cannot transmit again when it has an outstanding
slot. The provision for multiple packets from different
sources on the ring at one time assists in fair utilization
and quality of service support.
The Cambridge ring is an early example of such tech-
nology (some claim that it is also the ancestor of the ATM
protocols, also based on small xed frame sizes). A slot
contains one octet each for the source and destination ad-
dresses, ve control bits, and two, four, six, or eight data
octets, and thus slot sizes are extremely small. This im-
plies that higher-layer packet data are almost always seg-
mented into small units prior to transmission. Stations
could choose not to receive packets from particular sourc-
es; some of the control bits support this through response
codes. The Cambridge ring was simple to implement, but
somewhat wasteful of bandwidth due to the header over-
head in such small datagrams.
8.2. Register Insertion Ring
Register insertion rings are a common LAN technology,
and can be used to provide high performance through
their support for multiple packets on the ring at one time.
The register insertion ring uses a small shift register at
each station to control forwarding and insertion onto the
ring. The shift register is at least as large as the maximum
frame size. This allows a station to store a frame as it
passes. If the station has no data to send, a passing frame
is buffered long enough to determine whether it is des-
tined for the local station. If it is destined locally, a typical
implementation will both copy the frame into adapter
memory and forward the frame back around the ring in
order to support acknowledgments. Transmission when
the medium is available is handled by simply copying the
data onto the ring. If a frame arrives during this time, it is
buffered in the insertion register. The register insertion
method provides excellent ring utilization due to the mul-
tiple simultaneous packets on the ring without the over-
head penalty of the slotted ring. The disadvantage of this
technology is that the purge mechanism, that is, the tech-
nique used to remove problematic packets from the ring, is
generally more complex than in other systems.
9. HYPER CHANNEL, HIPPI, AND FIBRE CHANNEL
A number of LAN protocols are designed for very-high-
speed interconnection of computers and their peripherals.
Start
delimiter
Start
delimiter
Access
control
Access
control
Frame
control
Destination
address
Source
address
Data CRC
End
delimiter
End
delimiter
Frame
status
8
8 8 8 48 48 32
8 8
8 8
Figure 8. Token ring frame format.
HYPERchannel, developed by Network Systems Cor-
poration, is one of these. This protocol was developed in
the mid-1980s for the interconnection of supercomputers
and high-performance peripherals, and has been used
with Cray and Amdahl systems, among others. It sup-
ports data rates of up to 275Mbps over a variety of phys-
ical layers.
HIPPI (high-performance parallel interface) is another
of the protocols developed primarily for interconnection of
supercomputers. This protocol supports 800-Mbps or 1.6-
Gbps links over a large parallel cable, which is either 32 or
64 lines wide and runs at 25 MHz. The distances over
which HIPPI can be used is quite limited, but large
enough for a typical supercomputer center equipment
oor. Interconnection of sites can be accomplished using
ber extension options. Simple ow control features are
provided to lessen problems with computers and periph-
erals of widely different I/O (input/output) bandwidth. Al-
though simple and effective, this ow control scheme does
contribute to the problems of extending HIPPI networks
over larger distances while maintaining high throughput.
In order to build HIPPI networks of nontrivial size, simple
switches are used to interconnect devices. These switches
are typically not designed to switch between sources and
destinations at high rates, as with routers and packet
switches, but rather act as interconnection panels that
may be recongured at reasonable rates for the sharing
peripherals.
Fibre Channel is a LAN protocol suite designed for
high-speed communication between nodes using optical
ber. Rates of up to 4 Gbps are supported. Fibre Channel
is designed to combine the features of channel protocols
(such as HIPPI) and switched LAN protocols in order to
provide a scalable and reliable interconnection with the
high speeds and low latency required for servers and their
peripherals in storage-area networks (SANs).
10. OTHER LAN PROTOCOLS
A number of new, exible, and high-performance LAN
protocols have been developed. These developments in-
clude FireWire (IEEE 1394) and USB (universal serial
bus), high-speed protocol suites based on serial intercon-
nection technology. There are several varieties of Fire-
Wire, for example, with support for bandwidths of up to
800 Mbps with up to 63 devices (with no more than 16
cable hops) per bus in the more commonly deployed ver-
sions. More recent FireWire standards offer bandwidths
up to 1.2 Gbps with continued increases expected. Fire-
Wire is most commonly used for interconnection of digital
media (video and audio) devices. USB 1.1, a 12-Mbps se-
rial protocol with chaining support, was originally de-
signed primarily as an improvement over traditional
serial port technologies. Peripherals such as keyboards
and mice now most often use USB, and more exotic USB
peripherals such as storage media devices are increasing-
ly commonplace. With the introduction of USB 2.0, which
supports speeds of up to 480 Mbps, it is expected that USB
video and audio devices will also be widespread.
Asynchronous transfer mode (ATM) technology was
also deployed in LANs. ATM is a switch-based technology
that uses small packets (53 bytes) called cells. Intercon-
nection of nodes is through virtual circuits, which are
analogous to circuits in voice telephony. Multiple physical
layers are supported, including both copper and ber in-
frastructure options. Although ATM is often viewed as a
wide-area networking (WAN) technology, it does provide
support for features that are not available in other tech-
nologies. For example, ATM allows the definition of virtual
LANs that provide network administrators with options
that are not available in less sophisticated technologies.
Virtual or emulated LANs are interconnections of LANs,
perhaps widely separated, which are congured to emu-
late a single local-area network. Furthermore, multiple
logical local area networks can be supported over a single
physical infrastructure using this capability. Because of
its complex administration, ATM LAN deployments have
largely been restricted to networks with applications
needing its unique capabilities.
11. WIRELESS LANs
Wireless LANs use radio or infrared as the transmission
medium, as opposed to the traditional wire or ber. This
has significant advantages, particularly for deployment
in older buildings where wiring costs are high, as well
as environments in which workers may be moving
frequently.
Many of the initial wireless LANs in the United States
have used radiofrequencies in one of the ISM (instrumen-
tation, scientific, medical) bands, which generally may be
used without individual site licensing subject to restric-
tions on power output. The data rates on these systems
range from tens of kilobits per second to tens of megabits
per second, with typical ranges of a few hundreds of me-
ters. Wireless LANs are still evolving, but instances of
several standards are now the basis for extremely popular
products.
The most significant developments in this area are
based on the IEEE 802.11 standard. It is based on a
CSMA/CA (CSMA with Collision Avoidance) MAC layer
with multiple physical layers, including direct sequence
spread spectrum, orthogonal frequency-division multi-
plexing, frequency-hopping spread spectrum, and infra-
red. The widely deployed IEEE 802.11b standard provides
data rates of 11 Mbps over a range of approximately 100m
in typical radio congurations using the 2.4-GHz ISM
band. This has been supplemented by the IEEE 802.11g
standard that provides data rates up to 22 Mbps at these
same frequencies in a backward compatible manner. The
IEEE 802.11a standard provides 54 Mbps rates using
radios in the 5-GHz unlicensed bands.
Another wireless local area networking technology that
is increasingly popular is Bluetooth. This is intended to be
a very-local-area-network, often referred to as a personal-
area network, with ranges typically up to only 10 m and
data rates below 1 Mbps. This standard also uses the
2.4-GHz ISM frequency band, with frequency hopping
spread spectrum. Although Bluetooth is often called a
cable replacement technology, it also offers computer-to-
computer connectivity within its local region.
Wireless LANs will offer the opportunity to connect all
manners of devices such as sensors and instruments to the
increasingly pervasive network.
BIBLIOGRAPHY
1. W. Stallings, Local and Metropolitan Area Networks, 6th ed.,
Prentice-Hall, New York, 2000.
2. L. L. Peterson and B. S. Davie, Computer Networksa Systems
Approach, 3rd ed., Morgan Kaufmann, San Francisco, 2003.
3. P. T. Davis and C. R. McGufn, Wireless Local Area Networks,
McGraw-Hill, New York, 1995.
4. J. Walrand and P. Varaiya, High Performance Communication
Networks, 2nd ed., Morgan Kaufmann, San Francisco, 1999.
5. J. Wobus, LAN Technology Scorecard; available online at
http://www.mcgeoch.com/other/lan-technology.html, 1999.
6. J. Geier, Wireless LANs, Sams Publishing, Indianapolis,
2001.
LOOP ANTENNAS
KAZIMIERZ SIWIAK
TimeDerivative, Inc.
Coral Springs, Florida
The IEEE Standard Definitions of Terms for Antennas [1]
denes the loop as an antenna whose conguration is
that of a loop, further noting that if the current in the
loop, or in the multiple parallel turns of the loop, is es-
sentially uniform and the loop circumference is small com-
pared with the wavelength, the radiation pattern
approximates that of a magnetic dipole. That definition
and the further note imply the two basic realms of loop
antennas: electrically small, and electrically large struc-
tures.
There are hundreds of millions of loop antennas cur-
rently in use [2] by subscribers of personal communication
devices, primarily pagers. Furthermore, loops have ap-
peared as transmitting arrays, like the massive multiele-
ment loop array at shortwave station HCJB in Quito,
Ecuador, and as fractional wavelength-size tunable HF
transmitting antennas. The loop is indeed an important
and pervasive communication antenna.
The following analysis of loop antennas reveals that the
loop, when small compared with a wavelength, exhibits a
radiation resistance proportional to the square of the en-
closed area. Extremely low values of radiation resistance
are encountered for such loops, and extreme care must be
taken to effect efcient antenna designs. Furthermore,
when the small loop is implemented as a transmitting
resonant circuit, surprisingly high voltages can exist
across the resonating capacitor even for modest applied
transmitter power levels. The wave impedance in the im-
mediate vicinity of the loop is low, but at close distances
(0.12 wavelengths) exceeds the intrinsic free-space im-
pedance before approaching that value.
A loop analysis that applies to loops of arbitrary circu-
lar diameter and of arbitrary wire thickness is summa-
rized here. The analysis leads to some detail regarding the
current density in the cross section of the wire. Loops of
shapes other than circular are less easily analyzed, and
are best handled by numerical methods such as moment
method described in Ref. 3.
Loops are the antennas of choice in pager receivers, and
appear as both ferrite loaded loops and as single-turn
rectangular shaped structure within the radio housing.
Body-worn loops benet from a eld enhancement due to
the resonant behavior of human body with respect to ver-
tically polarized waves. In the high-frequency bands, the
loop is used as a series resonant circuit fed by a secondary
loop. The structure can be tuned over a very large
frequency band while maintaining a relatively constant
feedpoint impedance. Large loop arrays consisting of one-
wavelength-perimeter square loops have been successfully
implemented as high-gain transmitting structures at
high-power shortwave stations.
1. ANALYSIS OF LOOP ANTENNAS
Loop antennas, particularly circular loops, were among
the rst radiating structures analyzed beginning as early
as 1897 with Pocklingtons analysis [4] of the thin wire
loop excited by a plane wave. Later, Hallen [5] and Storer
[6] studied driven loops. All these authors used a Fourier
expansion of the loop current, and the latter two authors
discovered numerical difculties with the approach. The
difculties could be avoided, as pointed out by Wu [7], by
integrating the Green function over the toroidal surface of
the surface of the wire. The present author coauthored an
improved theory [8,9] that specifically takes into account
the nite dimension of the loop wire and extends the va-
lidity of the solution to wires thicker than previously con-
sidered. Additionally, the work revealed some detail of the
loop current around the loop cross section. Arbitrarily
shaped loops, such as triangular loops and square loops,
as well as loop arrays, can be conveniently analyzed using
numerical methods.
1.1. The Innitesimal Loop Antenna
The innitesimal current loop consists of a circulating
current I enclosing an innitesimal surface area S, and is
solved by analogy to the innitesimal dipole. The elds of
an elementary loop element of radius b can be written in
terms of the loop enclosed area Spb
2
and a constant ex-
citation current I (when I is RMS, then the elds are also
RMS quantities). The elds are near in the sense that
the distance parameter r is far smaller than the wave-
length, but far larger than the loop dimension 2b. Hence,
this is not the close near-eld region. The term kIS is often
called the loop moment and is analogous to the similar
term Ih associated with the dipole moment. The innites-
imally small loop is pictured in Fig. 1a next to its elemen-
tary dipole analog (Fig. 1b). The dipole uniform current
I owing over an elemental length h is the dual of a
LOOP ANTENNAS 2359
magnetic current M
z
SIh and the surface area is S
h/k. The elds due to the innitesimal loop are then found
from the vector and scalar potentials.
1.1.1. Vector and Scalar Potentials. The wave equation,
in the form of the inhomogeneous Helmholtz equation, is
used here with most of the underlying vector arithmetic
omitted (see Refs. 1012 for more details). For a magnetic
current element source, the electric displacement D is al-
ways solenoidal (the eld lines do not originate or termi-
nate on sources); that is, in the absence of source charges
the divergence is zero
r
.
D0 1
and the electric displacement eld can be represented by
the curl of an arbitrary vector F
De
0
ErF 2
where F is the vector potential and obeys the vector iden-
tity r
.
rF0: Using Amperes law in the absence of
electric sources
rHjoe
0
E 3
and with the vector identity rrF 0; where F rep-
resents an arbitrary scalar function of position, it follows
that
H rF joF 4
and for a homogeneous medium, after some manipulation,
we get
r
2
Fk
2
F e
0
Mrr
.
Fjom
0
e
0
F 5
where k is the wavenumber and k
2
o
2
m
0
e
0
: Although
Eq. (2) denes the curl of F, the divergence of F can be
independently dened and the Lorentz condition is chosen
jom
0
e
0
F r
.
F 6
where =
2
is the Laplacian operator given by
r
2
@
2
@x
2

@
2
@y
2

@
2
@z
2
7
Substituting the simplication of Eq. (6) into (5) leads to
the inhomogeneous Helmholtz equation
r
2
Fk
2
F e
0
M 8
Similarly, using Eqs. (6) and (4), it is seen that
r
2
Fk
2
F0 9
Using Eq. (4) and the Lorentz condition of Eq. (6), we can
nd the electric eld solely in terms of the vector potential
F. The utility of that definition becomes apparent when we
consider a magnetic current source aligned along a single
vector direction, for example, MzM
z
, for which the vec-
tor potential is FzF
z
, where z is the unit vector aligned
with the z-axis, and Eq. (8) becomes a scalar equation.
1.1.2. Radiation from a Magnetic Current Element. The
solution to the wave equation [Eq. (8)] presented here,
with the details suppressed, is a spherical wave. The re-
sults are used to derive the radiation properties of the in-
nitesimal current loop as the dual of the innitesimal
current element. The innitesimal magnetic current ele-
ment MzM
z
located at the origin satises a one-dimen-
sional, hence scalar, form of Eq. (8). At points excluding
the origin where the innitesimal current element is
located, Eq. (8) is source-free and is written as a function
of radial distance r
r
2
F
z
r k
2
F
z
r
1
r
2
@
@r
r
2
@F
z
r
@r
_ _
k
2
F
z
r 0 10
which can be reduced to
d
2
F
z
r
dr
2

2
r
dF
z
r
dr
k
2
F
z
r 0 11
Since F
z
is a function of only the radial coordinate, the
partial derivative in Eq. (10) was replaced with the ordi-
nary derivative. Equation (11) has a solution
F
z
C
1
e
jkr
r
12
There is a second solution where the exponent of the
phasor quantity is positive; however, we are interested
here in outward traveling waves so we discard that solu-
tion. In the static case the phasor quantity is unity. The
constant C
1
is related to the strength of the source cur-
rent, and is found by integrating Eq. (8) over the volume
including the source, giving
C
1

e
0
4p
kIS 13
and the solution for the vector potential is in the z unit
vector direction
F
e
0
4p
kIS
e
jkr
r
z 14
2b
S
h I
I
(a) (b)
Figure 1. Small-antenna geometry showing (a) the parameters
of the innitesimal loop moment, (b) its elementary dipole dual.
Source: After Siwiak [2].
2360 LOOP ANTENNAS
which is an outward propagating spherical wave with in-
creasing phase delay (increasingly negative phase) and
with amplitude decreasing as the inverse of distance. We
may now solve for the magnetic elds of an innitesimal
current element by inserting Eq. (14) into Eq. (4) with
Eq. (6) and then for the electric eld by using Eq. (2). The
elds, after sufcient manipulation, and for rbkS, are
H
r

kIS
2p
e
jkr
k
2
j
kr
2

1
kr
3
_ _
cosy 15
H
y

kIS
4p
e
jkr
k
2
1
kr

j
kr
2

1
kr
3
_ _
siny 16
E
f
Z
0
kIS
4p
e
jkr
k
2
1
kr
j
kr
2
_ _
siny 17
where Z
0
cm
0
376.730313 is the intrinsic free-space
impedance, c is the velocity of propagation (see Ref. 13
for definitions of constants), and I is the loop current.
The equations for the magnetic elds H
r
and H
y
of the
innitesimal [Eqs. (15) and (16)] loop have exactly the
same form as the electric elds E
r
and E
y
for the inni-
tesimal dipole, while Eq. (17) for the electric eld of the
loop E
f
has exactly the same form as the magnetic eld H
f
of the dipole when the term kIS of the loop expressions is
replaced with Ih for the innitesimal ideal (uniform cur-
rent element) dipole. In the case where the loop moment
kIS is superimposed on, and equals the dipole moment Ih,
the elds in all space will be circularly polarized.
Equations (15)(17) describe a particularly complex
eld behavior for what is a very idealized selection of
sources: a simple linear magnetic current M representing
a current loop I encompassing an innitesimal surface
Spb
2
. Expressions (15)(17) are valid only in the region
sufciently far (rbkS) from the region of the magnetic
current source M.
1.1.3. The Wave Impedance of Loop Radiation. The wave
impedance can be dened as ratio of the total energy in the
electric eld divided by the total energy in the magnetic
eld. We can study the wave impedance of the loop elds
by using Eqs. (15)(17) for the innitesimal loop elds,
along with their dual quantities for the ideal electric di-
pole. Figure 2 shows the loop eld wave impedance as a
function of distance kr from the loop along the direction of
maximum far-eld radiation. The wave impedance for the
elementary dipole is shown for comparison. At distances
near kr 1 the wave impedance of loop radiation exceeds
Z
0
376.73O, the intrinsic free space impedance, while
that of the innitesimal loop is below 376.73O. In this
region, the electric elds of the loop dominate.
1.1.4. The Radiation Regions of Loops. Inspection of Eqs.
(15)(17) for the loop reveal a very complex eld structure.
There are components of the elds that vary as the inverse
third power of distance r, the inverse square of r, and the
inverse of r. In the near eld or induction region of the
idealized innitesimal loop, that is, for kr51 (however,
rbkS for the loop and rbh for the dipole), the magnetic
elds vary as the inverse third power of distance.
The region where kr is nearly unity is part of the ra-
diating near eld of the Fresnel zone. The inner boundary
of that zone is taken by Jordan [12] to be r
2
40.38D
3
/l and
the outer boundary is ro2D
2
/l where D is the largest di-
mension of the antenna, here equal to 2b. The outer
boundary criterion is based on a maximum phase error
of p/8. There is a significant radial component of the eld
in the Fresnel zone.
The far eld or Fraunhofer zone is region of the eld for
which the angular radiation pattern is essentially inde-
pendent of distance. That region is usually dened as ex-
tending from ro2D
2
/l to innity, and the eld amplitudes
there are essentially proportional to the inverse of dis-
tance from the source. The far-zone behavior is identied
with the basic free-space propagation law.
1.1.5. The Induction Zone of Loops. We can study the
induction zone in comparison to the far eld by con-
sidering induction zone coupling which was investigated
by Hazeltine [14], and which was applied to low-frequency
radio receiver designs of his time. Today the problem
might be applied to the design of a miniature radio mod-
ule where inductors must be oriented for minimum cou-
pling. The problem Hazeltine solved was one of nding the
geometric orientation for which two loops in parallel
planes have minimum coupling in the induction zone of
their near elds and serves to illustrate that the near-eld
behavior differs fundamentally and significantly from far-
eld behavior. To study the problem, we invoke the prin-
ciple of reciprocity which states
_
V
E
b
.
J
a
H
b
.
M
a
dV
_
V
E
a
.
J
b
H
a
.
M
b
dV 18
0.1 1 10
10
100
1,000
10,000
p
o
Small dipole antenna
W
a
v
e

i
m
p
e
d
a
n
c
e
,

o
h
m
s
Small loop antenna
kr
Figure 2. Small loop antenna and dipole antenna wave imped-
ances compared. Source: After Siwiak [2].
LOOP ANTENNAS 2361
That is, the reaction on antenna a of sources b equals the
reaction on antenna b of sources a. For two loops with loop
moments parallel to the z axis, we want to nd the angle y
for which the coupling between the loops vanishes, that is,
both sides of Eq. (18) are zero. The reference geometry is
shown in Fig. 3. In the case of the loop, there are no elec-
tric sources in Eq. (18), so J
a
J
b
0, and both M
a
and M
b
are aligned with z, the unit vector parallel to the z axis.
Retaining only the inductive eld, components and clear-
ing common constants in Eqs. (15) and (17) are placed
into (18). We require that (H
r
r H
y
h)z 0. Since r
.
z
sin(y) and r
.
z cos(y), we are left with 2 cos
2
(y)
sin
2
(y) 0, for which y 54.7361. When oriented as shown
in Fig. 3, two loops parallel to the xy plane whose centers
are displaced by an angle of 54.7361 with respect to the z
axis will not couple in their near elds. To be sure, the
angle determined above is exactly correct for innitesi-
mally small loops; however, that angle will be nominally
the same for larger loops. Hazeltine [14] used this princi-
ple, placing the axes of the inductors in a common plane
each at an angle of 54.71 with respect to the normal form
the radio chassis, to minimize the coupling between the
inductors.
The same principle can be exploited in the design of a
metal detector, as depicted in Fig. 4. The loop a is driven
with an audiofrequency oscillator. Loop b, in a parallel
plane and displaced so that nominally y 54.71, is con-
nected to a detector that might comprise an audio ampli-
er that feeds a set of headphones. Any conductive object
near loop a will disrupt the balance of the system and re-
sult in an increased coupling between the two loops, thus
indicating the presence of a conducting object near a.
1.1.6. The Intermediate and Far-Field Zones of Loops. The
loop coupling problem provides us with a way to investi-
gate the intermediate and far-eld coupling by applying
Eq. (18) with Eqs. (15) and (16) for various loop separations
kr. In the far-eld region only the H
y
term of the magnetic
eld survives, and by inspection of Eq. (16), the minimum
coupling occurs for y 0 or 1801. Figure 5 compares the
coupling (normalized to their peak values) for loops in
parallel planes whose elds are given by Eqs. (15)(17).
Figure 5 shows the coupling as a function of angle y for an
intermediate region (kr 2) and for the far-eld case (kr
1000) in comparison with the induction-zone case (kr
0.001). The patterns are fundamentally and significantly
different. The coupling null at y 54.71 is clearly evident
for the induction-zone case kr 0.001 and for which the
(1/kr)
3
terms dominate. Equally evident is the far-eld
coupling null for parallel loops on a common axis when
the 1/kr terms dominate. The intermediate zone coupling
shows a transitional behavior where all the terms in kr are
comparable.
1.1.7. The Directivity and Impedance of Small Loops. The
directive gain of the small loop can be found from the far-
eld radially directed Poynting vector in a ratio to the
average Poynting vector over the radian sphere:
Dy; f
jEH
.
rj
1
4p
_
2p
0
_
p
0
jEH
.
rjsiny dy df
19
Only the y component of H and the f component of E sur-
vive into the far-eld. Using Eq. (16) for H
y
and (17) for E
f
and retaining only the 1/kr terms, Eq. (19) yields D
1.5sin
2
(y) by noting that the functional form of the product
of E and H is simply sin
2
(y) and by carrying out the simple
integration in the denominator of Eq. (19).
0
a
b
y
x
z
Figure 3. Two small loops in parallel planes and with y
54.736 degrees will not couple in their near elds. Source: After
Siwiak [2].
0
b
oscillator
a
detector
Figure 4. A metal detector employs two loops initially oriented to
minimize coupling in their near elds.
0
90 270
0, deg
180
kr = 1,000
kr = 2
kr = 0.001
Figure 5. Normalized induction zone, intermediate zone, and far
zone coupling between loops in parallel planes. Source: After
Siwiak [2].
2362 LOOP ANTENNAS
Taking into account the directive gain, the far-eld
power density P
d
in the peak of the pattern is
P
d

1:5 I
2
R
radiation
4pr
2
H
2
y
Z
0

kS
4p
k
r
I
_ _
2
Z
0
20
for radiated power I
2
R
rad
; hence, we can solve for the
radiation resistance
R
rad

kS
2
6p
Z
0
Z
0
p
6
kb
4
21
for the innitesimal loop of loop radius b.
When fed by a gap, there is a dipole moment that adds
terms not only to the impedance of the loop but also to the
close near elds. For the geometry shown in Fig. 6, and
using the analysis of King [15], the electrically small loop,
having a diameter 2b and wire diameter 2a, exhibits a
feedpoint impedance given by
Z
loop
Z
0
p
6
kb
4
18kb
2
1
a
2
b
2
_ _

jZkb
_
ln
8b
a
_ _
2
2
3
kb
2
_
1 2kb
2
22
including dipole-mode terms valid for kb50.1. The leading
term of Eq. (22) is the same as derived in Eq. (21) for the
innitesimal loop. Expression (22) adds the detail of terms
considering the dipole moment of the gap-fed loop as well
as renements for loop wire radius a. The small loop an-
tenna is characterized by a radiation resistance that is
proportional to the fourth power of the loop radius b. The
reactance is inductive and hence is proportional to the
antenna radius. It follows that the Q is inversely propor-
tional to the third power of the loop radius, a result that is
consistent with the fundamental limit behavior for small
antennas.
Using Eq. (22), and ignoring the dipole mode terms and
second-order terms in a/b, the unloaded Q of the loop
antenna, is
Q
loop

6
p
_
ln
8b
a
_ _
2
_
kb
3
23
which for b/a6 becomes
Q
loop

3:6
kb
3
24
which has the proper limiting behavior for the small loop
radius. The Q of the small loop given by Eq. (23) is indeed
larger than the minimum possible Q
min
(kb)
3
predicted
in Siwiak [2] for a structure of its size. It must be empha-
sized that the actual Q of such an antenna will be smaller
than given by Eq. (24) because of unavoidable dissipative
losses not represented in Eq. (22)(24). We can approach
the minimum Q but never be smaller, except by introduc-
ing dissipative losses.
1.2. The Gap-Fed Loop
The analysis of arbitrarily fat wire loops follows the meth-
od in Ref. 8, shown in simplied form in Ref. 9 and sum-
marized here. The toroid geometry of the loop is expressed
in cylindrical coordinates r, f, and z with the toroid locat-
ed symmetrically in the z 0 plane. The relevant geome-
try is shown in Fig. 6.
1.2.1. Loop Surface Current Density. The current densi-
ty on the surface of the toroidal surface of the loop is given
by
J
f

1
n1
1
p1
A
n;p
e
jnf
F
p
25
where the functions F
p
are symmetric about the z axis and
are simple functions of cos(nc), where c is in the cross
section of the wire as shown in Fig. 6 and is related to the
cylindrical coordinate by z asin(c). These functions are
orthonormalized over the conductor surface using the
GramSchmidt method described in Ref. 16, yielding
F
0

1
2p
ab
p 26
and
F
1
F
0
2
1 a=2b
2
cos c
a
2b
_ _
27
The higher-order functions are lengthy but simple func-
tions of sin(pc) and cos(pc).
1.2.2. Scalar and Vector Potentials. The electric eld is
obtained from the following vector and scalar potentials:
E rF joA 28
The boundary conditions require that E
f
, E
c
, and E
r
be
zero on the surface of the loop everywhere except at the
feed gap |f|re. Because this analysis will be limited to
wire diameters significantly smaller than a wavelength,
the boundary conditions on E
c
and E
r
will not be enforced.
In the gap E
f
V
0
/2er, where V
0
is the gap excitation
voltage.
[
2b
2a
Figure 6. Parameters of the fat wire loop. Source: After Siwiak
[2].
LOOP ANTENNAS 2363
The components of the vector potential are simply
A
f

1
4p
_
S
_
J
f
cos f f
0
dS 29
and
A
r

1
4p
_
S
_
J
f
sinf f
0
dS 30
and the vector potential is
F
jZ
0
4pk
_
S
_
1
r
@J
f
@f
GdS 31
where the value of dS[basin(c)]adc. Greens func-
tion G is expressed in terms of cylindrical waves to match
the rotational symmetry of the loop
G
1
2j
1
m1
e
jmff
0
_
1
1
J
m
r
1
nH
2
m
r
2
n
e
jzzz
0
dz
32
where
n
k
2
z
2
_
r
1
r acos c
r
2
r acos c
and where J
m
(nr) and H
m
(2)
(nr) are the Bessel and Hankel
functions, respectively.
1.2.3. Matching the Boundary Conditions. Expression
(8) is now inserted into Eqs. (5)(8), and the electric eld
is then found from Eq. (2) and the boundary condition is
enforced. For constant r on the wire
_
p
p
E
f
e
jmf
df
V
0
r
sinme
me
33
This condition is enforced on the wire as many times as
there are harmonics in c. Truncating the index p as de-
scribed in Ref. 9 to a small nite number P, we force E
f
0
except in the feeding gap along the lines of constant r on
the surface of the toroid. If we truncate to P, the number of
harmonics F
p
in c, and to Mthe number of harmonics in f,
we nd the radiation current by solving M systems of P
P algebraic equations in A
m,p
. In Ref. 9, P2 and M in the
several hundreds was found to be a reasonable computa-
tional task that led to useful solutions.
1.2.4. Loop Fields and Impedance. With the harmonic
amplitudes A
m,p
known, the current density is found from
Eq. (1). The electric eld is found next from Eq. (2) and the
magnetic eld is given by
H
r

@A
f
@z
34
H
f

@A
r
@z
35
H
z

@A
f
@r

A
f
r

1
r
@A
r
@r
36
The loop current across a section of the wire is found by
integrating the function J
f
in Eq. (25) around the wire
cross section. The loop radiation impedance is then the
applied voltage V
0
in the gap divided by the current in the
gap. Figure 7 shows the loop feed radiation resistance, and
Fig. 8 shows the corresponding loop reactance, as a func-
tion of loop radius kr for a thin-wire (O15) loop and a fat
wire (O10) loop where O2 ln(2pb/a). The thin-wire
loop has very sharp resonant behavior compared with
the fat-wire loop, especially for a half-wavelength-diame-
ter (kb 0.5) structure. The higher resonances are less
pronounced for both loops. Fat-wire loops exhibit an in-
teresting behavior in that at a diameter of about a half-
wavelength, the reactance is essentially always capacitive
and the total impedance remains well behaved.
1.2.5. Small Gap-Fed Loops. The detailed analysis of
the fat, gap-fed wire loop, as shown in Refs. 8 and 9, re-
veals that the current density around the circumference of
the wire, angle c in Fig. 6, is not constant. An approxi-
mation to the current density along the wire circum-
ference for a small diameter loop is
J
f

I
f
2pa
1 2 cosfkb
2
1 Y cosc 37
where I
f
is the loop current, which has cosine variation
along the loop circumference, and where the variation
0.5 1 1.5 2 2.5
0
200
400
600
800
1,000
kb
R
a
d
i
a
t
i
o
n

r
e
s
i
s
t
a
n
c
e
,

o
h
m
s
= 10
= 15
Figure 7. Loop radiation resistance.
2364 LOOP ANTENNAS
around the wire circumference is shown as a function of
the angle c. Y is the ratio of the rst-order to the zero-
order modes in f, and is not a simple function of loop di-
mensions a and b, but can be found numerically [2] and
from the analysis of the previous section. For the small
loop Yis negative and of order a/b, so Eq. (37) predicts that
there is current bunching along the inner contour (c
1801) of the wire loop. Table 1 gives representative values
for Y as a function of a/b.
This increased current density results in a correspond-
ing increase in dissipative losses in the small loop. We can
infer that the cross-sectional shape of the conductor
formed into a loop antenna will impact the loss perfor-
mance in a small loop.
The small loop fed with a voltage gap has a charge ac-
cumulation at the gap and will exhibit a close near electric
eld. For a small loop of radius b and centered in the xy
plane, the elds at (x,y) (0,0) are derived in Ref. 9 and
given here as
E
f
j
Z
0
kI
2
38
where I is the loop current and
H
z

I
2b
39
Expression (39) is recognized as the classic expression for
the static magnetic eld within a single-turn solenoid.
Note that the electric eld given by Eq. (38) does not de-
pend on any loop dimensions, but was derived for an elec-
trically small loop. The wave impedance Z
w
at the origin is
the ratio of E
f
to H
z
and from Eqs. (38) and (39) is
Z
w
jZ
0
kb 40
In addition to providing insight into the behavior of loop
probes, Eqs. (38)(40) are useful in testing the results of
numerical codes like the Numerical Electromagnetic Code
(NEC) described in Ref. 3, and are often used in the
numerical analysis of wire antenna structures.
When the small loop is used as an untuned and un-
shielded eld probe, the current induced in the loop will
have a component due to the magnetic eld normal to the
loop plane as well as a component due to the electric eld
in the plane of the loop. A measure of sensitivity of the E
eld to the H eld is apparent from expression (40). The
electric eld to magnetic eld sensitivity of a simple small
loop probe is proportional to the loop diameter. The small
gap-fed loop, then, has a dipole moment, which compli-
cates its use as a purely magnetic eld probe.
2. LOOP APPLICATIONS
Loop antennas appear in pager receivers as both ferrite-
loaded loops and as single-turn rectangular structures
within the radio housing. When worn on the belt, the
loop benets from coupling to the vertically resonant hu-
man body. In the high-frequency bands, the loop has been
implemented as a series resonant circuit fed by a second-
ary loop. The structure can be tuned over a very large
frequency band while maintaining a relatively constant
feedpoint impedance. One-wavelength-perimeter square
loops have been successfully implemented as high-gain
transmitting structures.
2.1. The Ferrite-Loaded Loop Antenna: A Magnetic Dipole
Let us examine a small ferrite-loaded loop antenna with
dimensions, 2h2.4cm and 2a0.4cm, and at a wave-
length of about l 8.6m as depicted in Fig. 9. When the
permeability of the ferrite is sufciently high, this anten-
na behaves like a magnetic dipole. The magnetic elds are
strongly conned to the magnetic medium, especially near
the midsection of the ferrite rod, and behave as the dual of
the electric dipole excited by a triangular current distri-
bution. We can therefore analyze its behavior using a
small dipole analysis as shown by Siwiak [2].
The impedance at the midpoint of a short dipole having
a current uniformly decreasing from the feed point across
its length 2h is
Z
dipole

Z
0
6p
kh
2
j
Z
0
2p
_
ln
2h
a
_ _
1
_
kh
41
1 1.5 2 2.5
1,000
500
0
500
1,000
kb
1,500
1,500
0.5
R
e
a
c
t
a
n
c
e
,

o
h
m
s
= 10
= 15
Figure 8. Loop reactance.
Table 1. Parameter Y for Various Loop Thickness and
b0.01 Wavelengths
O a/l Y
19.899 0.000003 0.0039
17.491 0.00001 0.0090
15.294 0.00003 0.020
12.886 0.0001 0.048
10.689 0.0003 0.098
8.2809 0.001 0.179
LOOP ANTENNAS 2365
The corresponding unloaded Q of the dipole antenna is
Q
dipole

3
_
ln
2h
a
_ _
1
_
kh
3
42
Equation (42) has the expected inverse third power with
size behavior for small antennas, and for h/a6
Q
dipole

4:5
kh
3
43
Comparing the Q for a small dipole given by Eq. (43) with
the Q of a small loop of Eq. (24), we see that the loop Q is
small even though the same ratio of antenna dimension to
wire radius was used. We conclude that the small loop
utilizes the smallest sphere that encloses it more efcient-
ly than does the small dipole. Indeed, the thin dipole is
essentially a one-dimensional structure, while the small
loop is essentially a two-dimensional structure.
We can use Eqs. (41) and (42) for the elementary dipole
to examine the ferrite-loaded loop antenna since it resem-
bles a magnetic dipole. The minimum ideal Q of this an-
tenna is given by Eq. (42), 1.0 10
6
. The corresponding
bandwidth of such an antenna having no dissipative losses
would be 2 35f/Q70 MHz/1.3 10
6
69 Hz. A practical
ferrite antenna at this frequency has an actual unloaded
Q
A
of nearer to 100, as can be inferred from the perfor-
mance of belt-mounted radios shown in Table 2. Hence, an
estimate of the actual antenna efciency is
10 log
Q
A
Q
40 dB 44
and the actual resultant 3dB bandwidth is about 700kHz.
Such an antenna is typical of the type that would be used
in a body-mounted paging receiver application. As detailed
in Siwiak [2], the body exhibits an average magnetic eld
enhancement of about 6 dB at this frequency, so the aver-
age belt-mounted antenna gain is 34 dBi. This is typical
of a front-position body-mounted paging or personal
communication receiver performance in this frequency
range.
2.2. Body Enhancement in Body-Worn Loops
Loops are often implemented as internal antennas in
pager receiver applications spanning the frequency bands
from 30 to 960MHz. Pagers are often worn at belt level,
and benet from the body enhancement effect. The
standing adult human body resembles a lossy wire anten-
na that resonates in the range of 4080 MHz. The
frequency response, as seen in Fig. 10, is broad, and for
belt-mounted loop antennas polarized in the body axis di-
rection, enhances the loop antenna azimuth-averaged
gain at frequencies below B500MHz.
The far-eld radiation pattern of a body-worn receiver
is nearly omnidirectional at very low frequency. As fre-
quency is increased, the pattern behind the body develops
a shadow that is manifest as a deepening null with in-
creasing frequency. In the high-frequency limit, there is
only a forward lobe with the back half-space essentially
completely blocked by the body. For horizontal incident
polarization there is no longitudinal body resonance and
there is only slight enhancement above 100MHz.
2.3. The Small Resonated High-Frequency Loop
The simple loop may be resonated with a series capa-
citor having a magnitude of reactance equal to the loop
2
h
2
a
>> 1
Figure 9. A Ferrite loaded loop antenna. Source: After Siwiak
[2].
Table 2. Paging Receiver Performance Using Loops
Frequency
Band (MHz)
Paging
Receiver, at Belt
Average Gain
(dBi)
Field Strength
Sensitivity
(dB
.
mV/m)
3050 32 to 37 1217
85 26 13
160 19 to 23 1014
280300 16 10
460 12 12
800960 9 1828
3
0
3
6
9
12
15
10 100 1000 10000
Frequency, MHz
d
B
Vertical
Horizontal
Figure 10. Gain-averaged body-enhanced loop response. Source:
After Siwiak [2].
2366 LOOP ANTENNAS
reactance, and indeed, is effectively implemented that way
for use in the HF bands as discovered by Dunlavy [17].
When fed by a second untuned loop, this antenna will ex-
hibit a nearly constant feedpoint impedance over a 3 : 1 or
4 : 1 bandwidth by simply adjusting the capacitor to the
desired resonant frequency. The reactive part of the loop
impedance is inductive, where the inductance is given by
Im{Z
L
} oL, so ignoring the higher-order terms
L
Z
0
kb
_
ln
8b
a
_ _
2
_
o
45
which, with the substitution Z
0
k/om
0
, becomes
Lm
0
b
_
ln
8b
a
_ _
2
_
46
The capacitance required to resonate this small loop at
frequency f is
C
1
2pf
2
L
47
The loop may be coupled to a radio circuit in many differ-
ent ways, including methods given in Refs. 17 and 18.
When used in transmitter applications, the small loop an-
tenna is capable of impressing a substantial voltage across
the resonating capacitor. For a power P delivered to a
small loop with unloaded Qof Eq. (23) and with resonating
the reactance X
C
given by the reactive part of Eq. (22), it is
easy to show that the peak voltage across the resonating
capacitor is
V
p

X
C
QP
_
48
by recognizing that
V
p

2
p
I
RMS
X
C
49
where I
RMS
is the total RMS loop current
I
RMS

P
RefZ
loop
g
50
along with Q at the resonant frequency in Eq. (23).
Transmitter power levels as low as one watt delivered
to a moderately efcient small-diameter (l/100) loop can
result in peak values of several hundred volts across the
resonating capacitor. This is not intuitively expected; the
small loop is often viewed as a high-current circuit that is
often described as a short-circuited ring. However, be-
cause it is usually implemented as a resonant circuit with
a resonating capacitor, it can also be an extremely high-
voltage circuit, as will be shown below. Care must be ex-
ercised in selecting the voltage rating of the resonating
capacitor even for modest transmitting power levels, just
as care must be taken to keep resistive losses low in the
loop structure.
As an example, consider the Q and bandwidth of a
small loop antenna, 2b 10 cm in diameter, resonated by a
series capacitor and operating at 30 MHz. The example
loop is constructed of 2a1-cm-diameter copper tubing
with conductivity s 5.710
7
S/m. The resistance per
unit length of round wire of diameter 2a with conductiv-
ity s is
R
s

1
2pad
s
s

1
2pa
om
0
2s
_
51
where d
s
is the skin depth for good conductors, o is the
radian frequency, and m
0
4p10
7
H/m is the permeabil-
ity of free space, so R
s
0.046 O. From Eq. (22) the loop
impedance is Z0.00792j71.41. Hence the loop efcien-
cy can be found by comparing the loop radiation resistance
with loss resistance. The loop efciency is R
s
/(R
s
Re{Z})
0.147 or 14.7%. From Eqs. (46) and (47) we nd the
resonating capacitance C74.3mF. From Eqs. (48)(50),
we see that if 1W is supplied to the loop, the peak voltage
across the resonating capacitor will be 308V, and the loop
current will be 4.3A. The resonated loop is by no means
the low-impedance structure that we normally imagine
it to be.
2.4. The Rectangular Loop
Pager and other miniature receiver antennas used in the
30940 MHz frequency range are usually implemented as
electrically small rectangular loops. For a rectangle
dimensioned b
1
b
2
of comparable length, and constru-
cted with 2a-diameter round wire, the loop impedance is
given in Ref. 19 as
Z
rect

Z
0
6p
k
2
A
2
j
Z
0
p
_
b
1
ln
2A
ab
1
b
c
_ _
:
b
2
ln
2A
ab
2
b
c
_ _
2ab
c
b
1
b
2
_ _
_
52
where Ab
1
b
2
and b
c
(b
1
2
b
2
2
)
1/2
. The loss resistance is
found by multiplying R
s
in Eq. (51) by perimeter length of
the loop, 2(b
1
b
2
). For a given antenna size the lowest
loss occurs for the circular loop.
2.5. The Quad Loop Antenna
The quad loop antenna, sometimes called the cubical
quad, was developed by Clarence C. Moore in the 1940s
as a replacement for a four element parasitic dipole array
(YagiUda array). The dipole array exhibited corona arc-
ing at the element tips severe enough to damage the an-
tenna when operated at high power levels (10 kW) in a
high-altitude (10,000-f) shortwave broadcasting applica-
tion in the 25-m band. Moore sought an antenna design
with no tips that would support extremely high electric
eld strengths, in order to avoid destructive arcing. His
solution was a 1l-perimeter square loop, later with a loop
LOOP ANTENNAS 2367
director element as shown in Fig. 11. The conguration
exhibited no arcing tendencies, and a new shortwave
antenna conguration was born.
As pictured in Fig. 11, the driven element is approxi-
mately one-quarter wavelength on an edge. Actually, res-
onance occurs when the antenna perimeter is about 3%
greater than a wavelength. The reector element perim-
eter is approximately 6% larger than a wavelength, and
may be implemented with a stub tuning arrangement.
Typical element spacing is 0.140.25 wavelength. The
directivity of a quad loop is approximately 2 dB greater
than that of a Yagi antenna with the same element spacing.
BIBLIOGRAPHY
1. IEEE Standard Definitions of Terms for Antennas, IEEE Std
145-1993, SH16279, March 18, 1993.
2. K. Siwiak, Radiowave Propagation and Antennas for Person-
al Communications, 2nd ed., Artech House, Norwood, MA,
1998.
3. G. J. Burke and A. J. Poggio, Numerical Electromagnetics
Code (NEC)Method of Moments, Lawrence Livermore Lab-
oratory, NOSC Technical Document 116 (TD 116), Vols. 1 and
2, Jan. 1981.
4. H. C. Pocklington, Electrical oscillations in wires, Proc. Cam-
bridge Physical Society London, England, 1897, Vol. 9, pp.
324333.
5. E. Hallen, Theoretical investigation into transmitting and re-
ceiving qualities of antennae, Nova Acta Regiae Societatis
Scientiarum Upsaliensis. Ser. IV, II(4):144 (Nov. 4, 1938).
6. J. E. Storer, Impedance of thin-wire loop antennas, Trans. of
AIEE, 75:609619 (Nov. 4, 1965).
7. T. T. Wu, Theory of the thin circular antenna, J. Math. Phys.
3:13011304 (Nov.Dec. 1962).
8. Q. Balzano and K. Siwiak, The near eld of annular antennas,
IEEE Trans. Vehic. Technol. VT36(4):173183 (Nov. 1987).
9. Q. Balzano and K. Siwiak, Radiation of annular antennas,
Correlations, Motorola Eng. Bull., (Motorola Inc., Schaum-
burg, IL), VI(2) (winter 1987).
10. C. A. Balanis, Advanced Engineering Electromagnetics,
Wiley, New York, 1989.
11. R. E. Collin, Antennas and Radiowave Propagation, McGraw-
Hill, New York, 1985.
12. E. C. Jordan and K. G. Balmain, Electromagnetic Waves and
Radiating Systems, 2nd ed., Prentice-Hall, Englewood Cliffs,
NJ, 1968.
13. 2002 CODATA Recommended Values of the Fundamental
Physics Constants, December 31, 2002, (Online): http://phys-
ics.nist.gov/cuu/Constants/ December 2003.
14. L. A. Hazeltine, Means for Eliminating Magnetic Coupling
between Coils, US Patent 1,577,421 (March 16, 1926).
15. R. W. P. King and C. W. Harrison, Jr., Antennas and Waves:
A Modern Approach, MIT Press, Cambridge, MA, 1969.
16. R. Courant and D. Hibert, Methods of Mathematical Physics,
Interscience Publishers, New York, 1953.
17. J. H. Dunlavy, Jr., Wide Range Tunable Transmitting Loop,
US Patent 3,588,905 (June 28, 1971).
18. T. Hart, Small, high-efciency loop antennas, QST, 3336
(June 1986).
19. K. Fujimoto, A. Henderson, K. Hirasawa, and J. R. James,
Small Antennas, Wiley, New York, 1987.
LOW NOISE AMPLIFIERS
VIJAY NAIR
Intel Corporation
Hillsboro, Oregon
1. INTRODUCTION
Low-noise ampliers (LNAs) are one of the most critical
components of the todays communication receivers. In
wireless systems, the input signal from the antenna rst
passes through an LNA that provides sufcient gain and
noise reduction for the subsequent stages for the RF re-
ceiver at the frequency band of interest. This LNA should
operate at low voltage, consume very low power, have wide
dynamic range, and exhibit excellent low-noise properties.
It should also have sufcient linearity for application in
systems that employ digital modulation schemes. Very-
low-power-consumption low-noise ampliers are neces-
sary for increasing talk time of wireless communication
products. Semiconductor devices having high linearity
and better noise guregain performance are needed for
digital wireless systems. Front-end circuits with good per-
formance have been fabricated using silicon CMOS, BJT,
BiCMOS, GaAs MESFET, HFET, HBT, and SiGe technol-
ogies [1]. In this article we will discuss the low-noise am-
plier design fundamentals, with special emphasis on low-
power communication circuits.
2. DEVICE MODELING
A low-noise amplier consists of transistors to provide
the gain and matching circuits to tune the performance
at the frequency range of interest. The gain and noise
z/4
z/4
Feed
Reflector
element
Driven
element
0.14 to 0.25 z
Stub
tuner
Figure 11. Two element loop array.
2368 LOW NOISE AMPLIFIERS
characteristics of a transistor can be completely specied
by the transistors S-parameters and noise parameters.
When an active device, such as a transistor, is employed in
a two-port system, the amount of noise that it generates
has to be considered carefully. This is often characterized
by the ratio of the output signal power to output noise
power. A term called noise factor is used to express this
noise deterioration of the systems. The noise factor F is
dened as
F
S=Npower ratio at input
S=Npower ratio at output
1
where N is noise and S is signal.
If G is the gain of the device, then the signal at the
output is given by
S
output
GS
input
and the noise at the output is given by
N
output
GN
input
noise generated by the device.
Hence, (1) can be rewritten as
F
N
output
GN
input
2
The noise gure in decibels is given by
FdB 10 logF 3
The major sources of noise in active device are thermal
noise generated by the resistance and the shot noise gen-
erated by the current owing through the active-device
junctions. Noise properties of bipolar transistors can be
described by an equivalent circuit as shown in Fig. 1a.
Three noise generators are introduced into this common-
emitter equivalent circuit to analyze the bipolar transistor
noise properties. These noise generators are due to the
uctuations in the DC base current, collector current and
thermal noise of the base resistance.
A simple noise equivalent circuit of a common-source
FET, as shown in Fig 1b, considers the noise contributions
of the gate resistance R
g
, source resistance R
s
, gate-to-
source capacitance C
gs
, and associated charging resistance
R
i
. Noise contributions due to other reactive parasitic
elements such as the gate and source lead inductance
are ignored, since they are negligible.
The noise gure in the presence of the source admit-
tance Y
s
can be written as
F F
min

R
n
G
s
jY
s
Y
opt
j
2
4
where Y
s
G
s
jB
s
, connected at the input port. F
min
, R
n
,
and Y
opt
characterize the device and are independent of its
output terminations [2]. The optimum source admittance,
Y
opt
G
opt
jB
opt
, is that particular value of source ad-
mittance Y
s
for which the optimum noise factor F
min
is
realized. Thus the noise performance of the device can be
characterized by a set of four noise parameters: the min-
imum noise gure F
min
, the noise resistance R
n
, and the
optimum source admittance Y
opt
G
opt
jB
opt
, where G
opt
and B
opt
individually affect the noise gure.
Thus (4) can be rewritten as follows:
F F
min

R
n
G
s
_ _
G
s
G
opt
2
B
s
B
opt
2
5
Equation (5) can be converted to arbitrary impedance form
as
F F
min

R
n
R
ss
R
opt
2
X
ss
X
opt
R
ss
R
2
opt
X
2
opt
6
where
Y
s
G
s
jB
s

1
Z
s
1
R
ss
jX
ss
7
Y
opt
G
opt
jB
opt

1
Z
opt
1
R
opt
jX
opt
8
Thus, knowing the noise parameters F
min
, R
n
, R
opt
, and
X
opt
, the noise gure of the device can be calculated using
(4) when the device is matched to Z
s
instead of Z
opt
.
3. LOW-NOISE BEHAVIOR OF BJTs AND FETs
The noise behavior of a bipolar device can be analyzed by
introducing noise generators due to the uctuations in the
DC base current, collector current, and thermal noise of
the base resistance into the common-emitter equivalent
circuit shown in Fig. 1a. The mean-square values of these
generators are given by [3]
i
2
b
2qI
b
Df 9
i
2
c
2qI
c
Df 10
e
2
b
4kTr
b
Df 11
where i
b
is the base current noise generator, i
c
is the col-
lector current noise generator, e
b
is the thermal noise volt-
age of the base resistor, I
b
is the DC base current, I
c
is the
dc collector current, Df is the narrow frequency interval, r
b
is the base resistance, q is the electronic charge, k is the
Boltzmann constant, and T is the absolute temperature.
Theoretical and experimental study of noise character-
istics of a GaAs MESFET were performed by several re-
searchers [4,5]. Fukui [6,7] developed a set of simple
analytical expressions for the noise parameters of a
MESFET. The Fukui equations shown below are used
LOW NOISE AMPLIFIERS 2369
extensively to estimate the noise parameters of a device
F
min
1 k
1
fC
gs
R
s
R
g
g
m
_ _
R
n

k
2
g
m
R
opt
k
3
1
4g
m
R
s
R
g
_ _
X
opt

k
4
fC
gs
12
where k
1
, k
2
, k
3
, and k
4
are tting factors and f is the
operating frequency.
These tting factors, estimated on the basis of the mea-
sured data on microwave transistors, are k
1
0.016, k
2
0.8, k
3
2.2, and k
4
160, provided the resistance is mea-
sured in ohms; transconductance g
m,
in siemens; gate-to-
source capacitance C
gs
, in picofarads; and frequency f, in
gigahertz. These empirical constants depend on the device
manufacturing process and give a good estimate of the
capabilities of different types of semiconductor device
technologies. These constants have to be determined by
measuring the S parameters and noise parameters of the
devices.
The noise parameters and the S parameters should be
taken into account while choosing a device for a low-noise
amplier design. A method for experimentally determin-
ing the noise parameters of a FET is described in Ref. 8.
The noise characteristics of heterostructure devices such
as HEMT and P-HEMT have been studied by Cappy
et al. [9].
Advances in CMOS technology enabled the fabrication
of MOSFETs having gate lengths below 100 nm. These
devices have realized very high transit frequencies and
very low noise gures. CMOS ampliers with very good
low-noise performance have been demonstrated for wire-
less system applications [10,11]. Figure 2a shows a sche-
matic cross section of a MOSFET that depicts the origin
properties of several parasitic elements. Although MOS-
FETs and MESFETS are very different in their operation
and large-signal properties, the small-signal operations
are very similar. Therefore theoretical and experimental
studies of MESFET behavior can be used to analyze the
MOSFET small-signal and high-frequency noise proper-
ties. A simplied equivalent circuit of a MOSFET suitable
for noise analysis is shown in Fig. 3. Just as in the ME-
SFET, the most important parasitic parameters are the
gate resistance R
g
, source resistance R
s
, channel conduc-
tance g
0
, and gate capacitance C
gs
. The dominant noise
sources are the thermal noise associated with the gate
noise, source resistance, and channel conductance. A noisy
two-port network analysis can be used to study the MOS-
FET noise properties. Following an analysis similar to
MESFET noise analysis, one can show that [12]
F
min
1 2gg
d0
f
f
T
R
g
R
s
gg
d0

f
f
T
R
g
R
s
_ _
13
where F
min
is the minimum noise gure, g
d0
is the equiv-
alent noise conductance of the channel, g is a bias-depen-
dent parameter, and R
s
and R
g
are the source and gate
resistances, respectively. Since the second term inside the
brackets is typically much smaller than the rst, we can
Figure 1. (a) Small-signal equivalent circuit of
common-emitter BJT; (b) noise-equivalent cir-
cuit of a common-source FET.
approximate (13) as
F
min
% 1 2
f
f
T
gg
d0
R
g
R
s
_
14
The cutoff frequency f
T
is given by
f
T

g
m
2pC
gs
C
gd
15
where g
m
is the transconductance of the device and C
gs
and C
gd
are the gate-to-source and gate-to-drain capaci-
tances, respectively.
The noise conductance G
n
can be written as
G
n

f
f
T
_ _
2
gg
d0
16
The optimum impedance is given by
Z
opt
R
opt
X
opt
17
where the optimum resistance is
R
opt
%
f
T
f
R
g
R
s
gg
d0
18
and the optimum reactance is
X
opt

1
oC
gs
C
gd
19
For low-noise amplier design, the device should be bi-
ased at the minimum noise gure point and the device
width should be chosen so that optimum resistance is close
to the driving resistance, which is typically 50 O. Also note
that the minimum noise gure is independent of the gate
width. In order to achieve a minimum noise gure, the
R
d
- Drain resistance
B - Base
D - Drain
S - Source
G - Gate
R
s
- Source resistance
R
g
- Gate resistance
R
g
R
dsb
G S
R
S
D
sb
Trench
B
p+
p+
Trench Trench Trench
n
n+
n+
n
D B
R
d
C
gs0
C
ds0
D
db
R
sb
R
db
p_sub
R
ds
- Drain-to-source resistance
R
db
- Drain-to-substrate resistance
R
sb
- Source-to-substrate resistance
R
dsb
- Drain-source to substrate resistance
C
gs0
- Gate-source overlap capacitance
C
ds0
- Gate-drain overlap capacitance
D
sb
- Source-to-bulk junction diode
D
db
- Drain-to-bulk junction diode
Figure 2. Small-signal equivalent circuit of a CMOS device showing the origins of the parasitic
components. (This gure is available in full color at http://www.mrw.interscience.wiley.com/erfme.)
R
g
i
d
R
1
g
m
V
gs
C
G
S
+
+
g
ds
R
S
C
gs
V
gs
C
gd
V
g
V
S
Figure 3. Noise-equivalent circuit of a common-source CMOS.
cutoff frequency of the device should be much higher than
the operating frequency.
4. AMPLIFIER POWER GAIN EQUATIONS
Many power gain equations are derived from the two-port
S parameters of the amplier. Let us consider the power
ow from the source to the load. Figure 4 shows a RF am-
plier circuit block diagram depicting the different power
gains and reection coefcients.
The transducer power gain G
T
of an amplier is dened
as the ratio of the output power P
L
delivered to the load Z
L
and the input power P
av s
available from the source to the
network.
G
T
G
T
G
G
; G
L
; S
P
L
P
av s
20
This can be expressed in terms of two-port S parameters
by
G
T

1 jG
s
j
2
jS
21
j
2
1 jG
L
j
2
j1 S
11
G
s
j
2
j1 G
out
G
L
j
2
21
or
G
T

1 jG
s
j
2
jS
21
j
2
1 jG
L
j
2
j1 G
in
G
s
j
2
j1 S
22
G
L
j
2
22
where
G
in
S
11

S
12
S
21
G
L
1 S
22
G
L
23
G
out
S
22

S
12
S
21
G
s
1 S
11
G
s
24
When both the input and output networks are perfectly
matched to the source impedance and the load impedance,
respectively, the transducer power gain is given by
G
Tm
jS
21
j
2
25
The unilateral transducer power gain (i.e., |S
12
|0) G
Tu
is given by
G
tu
1jGsj
2
j1S
11
Gsj
2
jS
21
j
2
1jG
L
j
2
j1S
22
G
L
j
2
26
We can rewrite this in the form
G
tu
G
s
G
0
G
L
27
Term G
s
, which depends only on the S
11
parameter of the
transistor and the source reection coefcient, represents
the gain or loss of the input matching circuits. Similarly
G
L
depends on the S
22
parameter of the transistor and the
load reection coefcient. This represents the gain or loss
of the output matching circuits. A block diagram of max-
imum unilateral gain amplier is shown in Fig. 5. The
middle term depends only on the S
21
parameter of the
transistor.
Maximum unilateral gain G
Tumax
is achieved when
G
s
S
11
and G
L
S
22
28
Figure 4. RF amplier circuit block diagram
depicting different power gains and reection
coefcients.
Figure 5. A block diagram of the maximum uni-
lateral gain conguration of an RF amplier.
and is given by
G
Tu max

jS
21
j
2
1 jS
11
j
2
1 jS
22
j
2
29
5. AMPLIFIER STABILITY
Stability of RF ampliers is one of the most important cri-
teria in the circuit design. The amplier stability can be
determined from the S parameters, input and output
matching networks, and circuit terminations. The ampli-
er can be conditionally or unconditionally stable. A two-
port network as shown in Fig. 4 is unconditionally stable
at a given frequency if the input and output resistances
(the real part of Z
in
and Z
out
) of the device are positive for
all passive source and load terminations. This also means
that jG
s
j1 and jG
L
j1 in this case. An amplier is condi-
tionally stable if the real part of input impedance and
output impedance of the amplier are greater than zero
for some positive part of real source and load impedances
at a specific frequency.
5.1. Stability Circles
The conditions for unconditional stability can be written
in terms of reection coefcients as follows:
jG
s
j1; jG
L
j1 30
jG
in
j S
11

S
12
S
21
G
L
1 S
22
G
L
o1 31
jG
out
j S
22

S
12
S
21
G
s
1 S
22
G
s
o1 32
Solutions to these equations give the required conditions
for an amplier to be unconditionally stable. When a two-
port network is potentially unstable (or conditionally sta-
ble), there may be values of G
s
and G
L
for which the real
parts of Z
in
and Z
out
are positive.
If we set jG
in
j and jG
out
j to unity, a boundary is estab-
lished beyond which the amplier becomes unstable. The
boundary condition is given by
S
11

S
12
S
21
G
L
1 S
22
G
L
1 33
S
22

S
12
S
21
G
s
1 S
11
G
s
1 34
Solutions to these two equations provide the values of G
s
and G
L
. It can be shown that the solutions for G
L
and G
s
lie
on circles [13]. These circles are known as the stability
circles. The equations for these are given by
G
L

S
22
DS
11
jS
22
j
2
jDj
2
S
12
S
21
jS
22
j
2
jDj
2
35
G
s

S
11
DS
22
jS
11
j
2
jDj
2
S
12
S
21
jS
11
j
2
jDj
2
36
where
DS
11
S
22
S
12
S
21
37
The circles in the G
L
plane where jG
in
j 1 are called out-
put stability circles and the circles in the G
s
plane where
jG
out
j 1, the input stability circles. The radii and centers
of input stability circles are given by
r
s
radius of G
s
circles
S
12
S
21
jS
11
j
2
jDj
2
38
C
s
center of G
s
circles
S
11
DS
22
jS
11
j
2
jDj
2
39
The radii and centers of output stability circles are given
by
R
L
radius of G
L
circles
S
12
S
21
jS
22
j
2
jDj
2
40
C
L
center of G
L
circles
S
22
DS
22
jS
22
j
2
jDj
2
41
These equations can be plotted on a Smith chart provided
the S parameters of the two-port network are known. Fig-
ure 6 illustrates the graphical construction of the stability
circles where |G
in
|1 and |G
out
|1 in the G
L
and G
s
planes, respectively. On one side of the stability circle
boundary in the G
L
plane |G
in
|o1, and on the other side
|G
in
|41.
Similarly, on one side of the stability circle boundary in
the G
s
plane |G
out
|o1, and on the other side |G
out
|41.
Figure 6. Graphical illustration of construction of stability cir-
cles where |T
in
|1 and |G
out
|1: (a) G
L
plane; (b) G
s
plane.
Stability circles can be plotted on the Smith chart di-
rectly. These circles separate the output or input planes
into stable or unstable regions. A stability circle plotted on
the input plane indicates the values of all the loads that
provide negative real part of output impedances. Similarly
stability circle plotted on the output plane indicates the
values of all the loads that provide negative real parts of
input impedances. A negative real part of impedance is
dened as a reection coefcient that has a magnitude
that is greater than unity. The presence of the negative
part of impedances at amplier input or output causes the
amplier to oscillate. The next step is to determine which
area in the Smith chart region is stable:
If Z
L
Z
0
, then G
L
0, and G
in
|S
11
| [from (30)].
If |S
11
|o1, then |G
in
|o1 when G
L
0.
This means that the center of the Smith chart repre-
sents a stable operating point in the G
L
plane:
If |S
11
|41 when Z
L
Z
0
, then |G
in
|41 when G
L
0.
This means that the center of the Smith chart repre-
sents an unstable operating point. These two cases
are shown in Figs. 7a and 7b. The shaded area repre-
sents values of G
L
that produce a stable region. Simi-
larly, Figs. 8a and 8b show stable and unstable regions in
the G
s
plane.
For unconditional stability, any passive source or load
impedances in a two-port network must produce stability
circles completely outside the Smith chart as shown in
Figs. 9a and 9b. The conditions for unconditional stability
for all passive loads and sources can be expressed in the
form:
jjC
L
j r
L
j > 1 for jS
11
j 42
jjC
s
j r
s
j > 1 for jS
22
j 43
The necessary and sufcient stability conditions can be
summarized as follows. A stability factor K for an ampli-
er can be expressed as
K
1 jS
11
j
2
jS
22
j
2
jDj
2
2jS
12
S
21
j
> 1 44
1 jS
11
j
2
> jS
12
S
21
j 45
1 jS
22
j
2
> jS
12
S
21
j 46
For unconditional stability
K > 1 47
jDjo1 48
5.2. Constant-Gain Circles
These are contours mapped on a Smith chart showing the
impedance values that produce a constant gain for a given
operating condition of an amplier.
5.2.1. Unilateral Case (|S
12
|0). In the unilateral
case, for G
s
S
11
or G
L
S
22
, the power gain G
s
or G
L
are at maximum value, and for |G
s
|1 or |G
L
|1, the
power gain G
s
or G
L
is zero. A general expression for the
G
s
and G
L
can be written in the following form:
G
s

1 jG
s
j
2
j1 S
11
G
s
j
2
49
G
L

1 jG
L
j
2
j1 S
22
G
L
j
2
50
Figure 7. Smith chart showing stable and unstable regions in
the G
L
plane: (a) |S
11
|o1; (b) |S
22
|41. The shaded area repre-
sents the stable region.
Figure 8. Smith chart showing stable and unstable regions in
the G
s
plane: (a) |S
22
|o1; (b) |S
22
|41. The shaded area repre-
sents the stable region.
Figure 9. Smith chart showing stability circles for an uncondi-
tionally stable amplier.
A normalized gain factor g
i
can be dened as
g
i

G
i
G
i max
1 jG
i
j
2
1
j1 S
ii
G
i
j
2
1 jS
ii
j
2
51
where i s, (ii 11), and i L (ii 22). These equations
can be rearranged to prove that they represent a family
of circles [14]. It is quite often convenient to draw these
circles on a Smith chart. The centers of these circles
are located on the vector drawn from the center of the
Smith chart to the point S
11
or S
22
. The distance from
the center of the Smith chart to the center of the constant-
gain circles along the vector S
11
for the source side is
given by
d
s

g
s
jS
11
j
1 jS
11
j
2
1 g
s
52
The radius of the constant-gain circles is expressed as
r
s

1 jS
11
j
2
1 g
s
_
1 jS
11
j
2
1 g
s
53
Similar expressions for load-side constant-gain circles can
be derived by replacing the S parameters of the signal side
with that of the load side.
5.2.2. Simultaneous Conjugate Match: Bilateral Case
S
12
O0. This case occurs when S
12
cannot be neglected.
The transducer power gain is given by Eq. (22). Conditions
required for maximum transducer gain are given by
G
in
G
s
; G
out
G
L
54
The simultaneous conjugate match condition thus ob-
tained is shown in Fig. 10.
The gain values G
s
and G
L
can be written as follows:
G
s
S
11

S
12
S
21
G
L
1 S
11
G
L
55
G
L
S
22

S
12
S
21
G
s
1 S
22
G
s
56
By solving these last two equations simultaneously, we
can get
G
Ms

B
1

B
2
1
4jC
1
j
2
_
2C
1
57
G
ML

B
2

B
2
2
4jC
2
j
2
_
2C
2
58
B
1
1 jS
11
j
2
jS
22
j
2
jDj
2
59
B
2
1 jS
22
j
2
jS
11
j
2
jDj
2
60
C
1
S
11
DS
22
61
C
2
S
22
DS
11
62
where G
Ms
and G
ML
are the values of input and output
reection coefcients when the amplier is conjugate-
matched simultaneously.
The maximum transducer power gain, under simulta-
neous conjugate-matched conditions, is achieved when
G
s
G
Ms
and G
L
G
ML
. Applying these conditions to equa-
tion (22), we get
G
t;max

1 jG
Ms
j
2
jS
21
j
2
1 jG
ML
j
2
j1 S
11
G
Ms
1 S
22
G
ML
S
21
S
12
G
Ms
G
ML
j
2
63
This can be simplied to
G
t;max

jS
21
j
jS
12
j
_ _
K
K
2
1
p
64
where the stability factor K is as dened in (44).
The maximum stable gain G
msg
, dened as the value of
G
T,max
when K1, is given by
G
msg

jS
21
j
jS
12
j
65
This G
msg
is a gure of merit that represents the maxi-
mum value of G
T,max
.
Figure 10. A block diagram of the conjugately
matched RF amplier. (Simultaneous conjugate
match is achieved when G
in
G
s
and G
out
G
L
.)
5.3. Constant-Noise-Figure Circles
In some applications the major design objective is to ob-
tain the lowest possible noise gure for the circuit. Since
minimum noise gure and maximum power gain cannot
be obtained simultaneously, constant-noise-gure circles
should be plotted on the same Smith chart along with the
constant-power-gain circles. The reection coefcient can
then be selected that give the optimum performance in
terms of the noise gure and power gain.
The noise gure equation (4) discussed in Section 2 can
be rewritten as
F F
min

r
n
g
s
jY
s
Y
opt
j
2
66
where r
n
is the normalized noise resistance of the two-port
network, Y
s
g
s
jb
s
is the normalized source admit-
tance, and Y
opt
g
opt
jb
opt
is the normalized source ad-
mittance, which results in the minimum noise gure.
Usually the normalizing impedance is 50 O.
We can express the normalized source admittances Y
s
and Y
opt
in terms of the source reection coefcients as
Y
s

1 G
s
1G
s
67
and
Y
opt

1 G
opt
1G
opt
68
Substitution of these quantities into (66) gives the noise
gure equations as
F F
min

4r
n
jG
s
G
opt
j
2
1 jG
s
j
2
j1 G
opt
j
2
69
The resistance r
n
can be found by measuring noise gure
for G
s
0, when a 50 O resistance is used. Then
r
n
F
50
F
min
j1G
opt
j
2
4jG
opt
j
2
70
To determine the noise gure circle for a given noise gure
F
i
, we dene a noise gure parameter N
i
as
N
i

jG
s
G
opt
j
2
1 jG
s
j
2

F
i
F
min
4r
n
_ _
j1G
opt
j
2
71
which can be rewritten in the following form [18]:
G
s

G
opt
1 N
i
N
2
i
N
i
1 jG
opt
j
2
1N
i
2
72
These equations represent a family of circles in terms of
N
i
. The center of the circle is given by
C
F

G
opt
1 N
i
73
and the radius is given by
r
1
1 N
i
N
2
i
N
i
1 jG
opt
j
_
74
A set of constant-gain circles and constant-noise-gure
circles, as shown in Fig. 11, are drawn on the same Smith
chart for the amplier design. This plot shows the tradeoff
between the noise gure and the gain of a device.
6. NARROWBAND AMPLIFIERS
Typical narrowband ampliers have bandwidth less than
10% of the center frequency. Most of the ampliers de-
signed for portable communication ampliers fall into this
category. These types of ampliers can be further divided
into maximum power gain ampliers and low-noise
ampliers. A different design procedure has to be followed
for each case.
6.1. Maximum Power Gain Design
From the discussions in Section 5.2.2, the reection coef-
cient of the source and load impedance required to con-
jugately match the amplier are given by Eqs. (57)(62).
Knowing the values of G
Ms
and G
ML
provided by the device
manufacturers, the inputoutput matching network for
the amplier can be designed.
The source equivalent circuit for maximum power
transfer can be written as
Z
in

1 G
Ms
1 G
Ms
75
Figure 11. Smith chart showing the constant-gain circles and
constant-noise-gure circles for optimum impedance determina-
tion. (With permission from Wiley [1].)
where
G
Ms
jG
Ms
j cosangG
Ms
jjG
Ms
j sinangG
Ms
76
The input impedance for minimum noise gure can be
written as follows:
Z
inopt
Z
0
1 jG
Ms
j
2
j2jG
Ms
j sinangG
Ms
1 jG
Ms
j
2
2jG
Ms
j cosangG
Ms
77
The load impedance for maximum power transfer can be
written as
Z
out
Z
0
1 jG
ML
j
2
j2jG
ML
j sinangG
ML
1 jG
ML
j
2
2jG
ML
j cosangG
ML
78
Once these impedances are determined, CAD tools may be
used to compute matching elements of the amplier.
6.2. Low-Noise Amplier Design
In this case the minimum noise gure is more important
than determining the maximum gain. The noise parame-
ters of the devices are included in the data sheets provided
by the vendor or foundry. Figure 12 shows a block diagram
of a low-noise narrowband amplier. The impedances at
various points along the circuits are illustrated in this
gure. The design procedure is similar to the maximum
power gain design, except that we start with the input
reection coefcients for the minimum noise gure. This
reection coefcient is then converted to the equivalent
input impedance or admittance. The matching elements
are then designed to transform these impedances to 50 O.
Let G
0
be the optimum reection coefcient for the
minimum noise gure. From Eq. (77), we obtain
Z
opt
Z
0
1 jG
0
j
2
j2jG
0
j sinangG
0
1jG
0
j
2
2jG
0
j cosangG
0
79
We may compute the output impedance computed simi-
larly if we know the load reection coefcient for the min-
imum noise gure using Eq. (78).
7. WIDEBAND AMPLIFIER DESIGN
Wideband LNAs have many applications in instru-
mentation, high-speed microwave, and wireless commu-
nication systems. The design methodology for wideband
LNA is more complex than the narrowband ampliers
because of the power gain rolloff characteristics of the
device. The matching circuit of the amplier has to be
carefully designed to achieve a constant gain over a
broad frequency of interest. The variation in the device
S parameters has to be carefully considered. As the fre-
quency increases, the forward gain S
21
of the device de-
creases at the rate of 6 dB/octave and the reverse gain
S
12
increases at about the same rate. The input and
output impedances S
11
and S
22
also vary significantly
with frequency.
Several circuit design approaches are employed to
achieve the desired broadband characteristics of the
amplier. Broadband amplier design techniques using
negative feedback, compensated impedance matching,
and balanced amplications are discussed in the next
section.
7.1. Negative-Feedback Amplier Design
In this conguration a feedback network is used to com-
pensate for the gain rolloff of the device. Figures 13a and
13b show the shunt feedback amplier conguration for a
bipolar device and a FET, respectively. Similarly, Figs. 14a
and 14b show the series feedback conguration for a bi-
polar device and a FET, respectively. The series feedback
enhances the input match of the amplier, while the shunt
feedback attens the forward gain S
21
of the amplier
over the frequency range of interest. Typically a resistor is
used as the primary feedback element. A capacitor is often
added to the shunt feedback circuit to decouple the DC
bias of the transistor.
An equivalent circuit of a bipolar transistor can be rep-
resented as shown in Fig. 15a and that of a FET, as in Fig.
15b. Other parasitic elements are neglected for simplicity
and because their contributions to the amplier perfor-
mance are relatively insignificant. The admittance matrix
of the network can be written as
I
1
I
2
_ _
1
R
1
1
R
1
g
m
1 g
m
R
2
1
R
1
1
R
1
_
_
_
_
V
1
V
2
_ _
80
Figure 12. A block diagram of the minimum
noise gure narrowband RF amplier.
By converting this Yparameter equation to S parameters,
we can get
S
11
S
22

1
D
1
g
m
Z
2
0
R
1
1 g
m
R
2
_ _
81
S
21

1
D
2g
m
Z
0
1 g
m
R
2

2Z
0
R
1
_ _
82
S
12

2Z
0
DR
1
83
where
D1
2Z
0
R
1
g
m
Z
2
0
R
1
1 g
m
R
2
84
For the ideal case of VSWR1, S
11
S
22
0
1 g
m
R
2

g
m
Z
2
0
R
1
85
or
R
2

Z
2
0
R
1
1
g
m
86
S
21

Z
0
R
1
Z
0
87
S
12

Z
0
Z
0
R
1
88
Also note that when the transconductance g
m
b1, we
obtain
Z
0

R
2
R
1
_
89
Equation (87) shows that S
21
depends on R
1
only. This
shows that gain attening can be achieved for an amplier
by using parallel feedback.
It can also be seen from Eq. (87) that the minimum
transconductance is realized when R
2
is zero:
g
m
min
R
1
Z
2
0
90
From Eq. (82), it can also be shown that
g
m
min
1 S
21
Z
2
0
91
R
1
Z
0
1 jS
21
j 92
The phase of S
21
is an important consideration in nega-
tive-feedback amplier design. At lower frequencies, the
phase of S
21
is close to 1801, but as the operating frequen-
cy increases, the phase increases rapidly. The phase dif-
ference between the input and output voltages could
become zero at some frequency, creating positive feedback.
This problem can be avoided by adding a reactive element
to the feedback circuit.
7.2. Compensated Matching Technique
In this technique the input and output matching imped-
ances are mismatched to compensate for the variation in
the magnitude of S
21
with frequency. Since the amplier
has to operate over a wide bandwidth, VSWR optimization
must cover a wide frequency range. Even though the an-
alytical technique using Smith chart can be used to de-
termine a matching circuit, a more accurate design can be
achieved by using computer-aided design procedures.
7.3. Balanced Amplier Design
It is difcult to achieve good VSWR for broadband ampli-
ers that employ a compensated matching technique. The
mismatching of the input and output impedances causes
the VSWR to degrade significantly. Balanced amplier de-
sign can be used to achieve good gain atness and excel-
lent VSWR. Figure 16 shows the schematic of the balanced
amplier. The input coupler circuit divides input power
equally between the amplier A and B input ports. The
output coupler circuit combines the amplied signal from
each amplier. The unused ports of the couplers are ter-
minated at 50 O impedance to eliminate reections. In this
approach the input and output matching of each amplier
(a) (b)
B
R
1
C
E
G
S
D
R
1
Figure 13. Shunt feedback amplier: (a) bipolar transistor;
(b) FET.
(a) (b)
D G
S
R
2
B C
R
2
E
Figure 14. Series feedback amplier: (a) bipolar transistor;
(b) FET.
(a) (b)
R
2
g
m
V
gs
S
D R
1
G
R
be
R
1
R
2
C
g
m
V
be
B
E
Figure 15. Equivalent circuit of feedback amplier: (a) bipolar
transistor; (b) FET.
can be mismatched to determine the gain atness. The
combined amplier will be matched when parallel ampli-
ers A and B are combined. This will provide excellent
VSWR at the input and output of the circuit. The coupler
can be realized in many different ways and is discussed
elsewhere in this book and in Ref. 1.
For the balanced amplier that uses a 3-dB coupler, we
obtain
S
11
0:5S
11a
S
11b
93
S
22
0:5S
22a
S
22b
94
The forward power gain is
jS
21
j
2
0:25jS
21a
S
21b
j
2
95
The reverse power loss is
jS
12
j
2
0:25jS
12a
S
12b
j
2
96
The input standing-wave ratio is
VSWR
in

1 jS
11
j
1 jS
11
j
97
The input standing-wave ratio is
VSWR
out

1jS
22
j
1 jS
22
j
98
These types of ampliers achieve high-gain atness over a
wide range and a high degree of stability and are easy to
cascade. However, to achieve this performance, two am-
plier chains are required. Hence this amplier has high
power consumption and a larger chip size.
BIBLIOGRAPHY
1. K. Chang, I. Bahl, and V. Nair, RFand Microwave Circuit and
Component Design for Wireless Applications, Wiley, New
York, 2002.
2. IRE standards on methods of measuring noise in linear two-
ports, 1959, Proc. IRE 48:6068 (Jan. 1960).
3. H. Fukui, The noise performance of microwave transistors,
IEEE Trans. Electron. Devices ED-13:329341 (March 1966).
4. H. Statz, H. A. Haus, and R. A. Pucel, Noise characteristics of
gallium arsenide eld-effect transistors, IEEE Trans. Elec-
tron. Devices ED-21:549562 (Sept. 1974).
5. R. A. Pucel, D. J. Masse, and C. F. Krumm, Noise performance
of galliumarsenide eld-effect transistors, IEEEJ Solid-State
Electron. SC-11:243295 (April 1976).
6. H. Fukui, Design of GaAs MESFETs for broad-band lownoise
ampliers, IEEE Trans. Microwave Theory Tech. MTT-
27:643650 (July 1979).
7. H. Fukui, IEEE Trans. Microwave Theory Tech. MTT-29:1119
(Oct. 1981).
8. R. Q. Lane, Determination of device noise parameters, Proc.
IEEE 57:14611462 (Aug. 1969).
9. A. Cappy, M. Schortgen, and G. Salmar, A new method for
calculating the noise parameters of MESFETs and TEG-
FETs, IEEE Electron. Device Lett. EDL-6(6):270272 (June
1985).
10. B. Razavi, A 2.4 GHz CMOS receiver for IEEE802.11 wireless
LANs, IEEE J. Solid-State Circ. 34(10):13821385 (Oct. 1999).
11. T. H. Lee, 5-GHz CMOS wireless LANs, IEEE Trans. Micro-
wave Theory Tech. 50(1):268280 (Jan. 2002).
12. E. Abou-Allam and T. Manku, A low voltage design technique
for low noise RF integrated circuits, Proc. IEEE Int. Symp.
Circuits and Systems, 1998, pp. 373377.
13. S. Y. Liao, Microwave Circuit Analysis and Amplier Design,
Prentice-Hall, Englewood Cliffs, NJ, 1987.
14. G. Gonzalez, Microwave Transistor Ampliers, Analysis and
Design, Prentice-Hall, Englewood Cliffs, NJ, 1984.
LOW NOISE AMPLIFIERS: DEVICE NOISE
CHARACTERIZATION AND DESIGN
SUNGJAE LEE
KEVIN J. WEBB
Purdue University
West Lafayette, Indiana
1. INTRODUCTION
Receiver sensitivity is limited by noise associated with the
elements of the receiver. The most critical element is the
50
Input
Output
50
1 2
3 4
Input
matching
Input
matching
Output
matching
Output
matching
Amplifier A
Amplifier B
S
11a
S
11b
Input
coupler
Output
coupler
Figure 16. Balanced FET amplier conguration.
LOW NOISE AMPLIFIERS: DEVICE NOISE CHARACTERIZATION AND DESIGN 2379
one that rst processes the detected signal, usually an
amplier. Therefore, achieving low noise amplier (LNA)
designs is critical in many receiver systems.
Given the input signal-to-noise ratio, the receiver
noise, which is the noise added by the receiver compo-
nent, degrades the signal-to-noise ratio. It is common to
use noise gure (F) as a system parameter that charac-
terizes the ability of the component (e.g., the amplier)
to process low-level signals [1]. The noise gure is the
ratio of the input to output signal-to-noise ratios, resulting
in FZ1. Assuming that dissipative circuit losses are
small, the active device (e.g., the transistor) dominates
the receiver noise. An effective receiver amplier thus
requires a suitable low-noise device. It is also important to
present an appropriate impedance through matching
networks.
The device can be characterized using two-port noise
and scattering (S)-parameter measurements. The mea-
surement of the noise gure for various source admit-
tances yields three noise parameters, the minimum noise
gure (F
min
), the noise resistance (R
n
, or conductance G
n
),
and the optimum source admittance for minimum noise
(Y
opt
G
opt
jB
opt
) [2]. The S-parameter measurements
can be used to extract the values for an equivalent-circuit
representation for the transistor. The measured noise pa-
rameters can then be used to construct a noise equivalent
network that, coupled with the device equivalent circuit,
can be used to predict the noise performance of this device
in, for example, an amplier conguration, and hence for
the design of LNAs.
An empirical relation between the noise gure and the
circuit parameters that employs tting parameters was
proposed for GaAs FETs. Fukui describes the dependence
of F
min
on the operating frequency and the model elements
in the equivalent circuit [3]. More quantitative equivalent
circuit modeling has been pursued in several works [47],
where Nyquist noise sources were represented equiva-
lently as temperature or resistance/conductance [8].
Within the active device, there are parasitic noise con-
tributions due to contact metal and contact resistance,
along with intrinsic noise mechanisms. Low-noise silicon
device technology (e.g., CMOS) has made great strides,
showing fairly good performance [9]. However, the lowest-
noise-gure device is the high-electron-mobility transistor
(HEMT), based on compound semiconductor materials. To
rst order, a higher-speed device will have lower noise.
2. NOISE FIGURE OF A MULTISTAGE SYSTEM
The analysis of the noise gure in a multistage system
demonstrates the importance of good LNA design in re-
ceivers. Knowing the noise gure and gain of the individ-
ual stages, one can determine the noise gure of the
multistage system as [1]
F
tot
F
1

F
2
1
G
1
F
3
1
G
1
G
2
1
where the rst three stages have gains G
1
,G
2
,G
3
, and
noise gures F
1
,F
2
,F
3
. Equation (1) clearly shows that the
noise performance of a multistage system is dominated by
the noise gure and the gain of the rst stage, since the
noise of the later stages is reduced by the gain of the pre-
vious stages. Thus, high gain in the rst stage (the LNA)
reduces the inuence of the noise of later stages, making
the noise of the rst stage critical.
3. NOISE MECHANISMS IN HEMTS
High-gain and low-noise characteristics have made
HEMTs the device of choice for building low-noise tran-
sistor ampliers. GaAs-based HEMTs have long been used
in microwave low-noise ampliers [10]. When grown on an
InP substrate, GaAs-based HEMTs have demonstrated
low noise gure at millimeter-wave frequencies [11].
More recently demonstrated AlGaN/GaN HEMTs on SiC
offer promising microwave noise performance [12]. Table 1
summarizes the noise performance achieved with these
devices.
One can consider three uncorrelated noise mechanisms
in the intrinsic HEMT, associated with velocity uctua-
tion, gate leakage, and traps, as shown in the schematic of
Fig. 1. The velocity uctuation or diffusion noise is due to
electron scatter from the heterojunction, the lattice (pho-
non), and impurities. This diffusion noise contracts to
thermal or Johnson noise if the Einstein relation holds
[13]. The bandwidth of this process is proportional to the
HEMT (Substrate) Frequency F
min
Gatelength
AlInAs/GaInAs (GaAs) [10] 18 GHz 0.48 dB 0.1 jm
AlInAs/GaInAs (InP) [11] 63 GHz 0.8 dB 0.2 jm
AlGaN/GaN (SiC) [12] 18 GHz 0.75 dB 0.25 jm
V
g
+
GaN
AlGaN Leakage
1/f noise
S D
G
Surface or
AlGaN/GaN interface states
Electron
v fluctuation
Channel
Figure 1. Fundamental noise mechanisms in AlGaN/GaN
HEMTs.
Table 1. Microwave and Millimeter-Wave Noise
Performance (Measured) of GaAs, InP, and GaN-
based HEMTs
HEMT (Substrate)
Frequency
(GHz) F
min
(dB)
Gate Length
(mm)
AlInAs/GaInAs (GaAs) [10] 18 0.48 0.1
AlInAs/GaInAs (InP) [11] 63 0.8 0.2
AlGaN/GaN (SiC) [12] 18 0.75 0.25
2380 LOW NOISE AMPLIFIERS: DEVICE NOISE CHARACTERIZATION AND DESIGN
inverse of the scattering time, making it frequency-inde-
pendent (i.e., white noise) in the microwave regime
where the device is typically operated. This fundamental
noise process results in a drain current noise variance
(/|i
d
|
2
S) that will be ideally white and dominated by the
channel conduction process. The associated gate voltage
variance (/|v
g
|
2
S) is coupled to the channel random pro-
cess by the gate capacitance (C
gs
). Physically, one could
view this as a random variation in the electric eld lines
that terminate on the local random variation in channel
charge, due to velocity or charge density uctuation, and
the induced gate charge. The line integral of this eld
gives the gate voltage. A frequency independent /|i
d
|
2
S
will therefore result in /|v
g
|
2
S being frequency-indepen-
dent. This model suggests that the correlation between
the gate (v
g
) and drain (i
d
) noise sources will be a function
of the degree of electron scatter in the channel, and that it
would tend to reduce with increasing channel length.
The gate leakage noise process is associated with
electron injection into the channel over the gate Schottky
barrier. The randomness of this emission process leads
to classical shot noise (/|i|
2
S2eI
g
A
2
/Hz, with e
1.610
19
C and I
g
the gate DC current) [13]. This gate
leakage noise is also ideally independent of frequency.
Finally, a 1/f
a
noise process (where a 0.52) occurs as
a result of the trapping of electrons in surface states (near
the gate) or AlGaN/GaN interface states (near the chan-
nel). This lower frequency noise is particularly important
when the device operation is nonlinear, and leads to the
noise sidebands on an oscillator signal.
All three of these noise processes will contribute to both
the gate and drain noise variances. In addition to the in-
trinsic noise sources, there is thermal noise associated
with various resistances, including source and drain con-
tact resistance and gate nger resistance. All of these par-
asitic noise sources are uncorrelated with each other and
with the intrinsic noise.
4. TWO-PORT NOISE FIGURE
As suggested by Rothe and Dahlke [2], all noise within a
linear two-port device can be represented as two equiva-
lent and partially correlated noise sources, and two forms
are shown in Fig. 2, with the same two-port device now
considered noiseless. In the circuit of Fig. 2(a), the random
current i
A
and random voltage v
A
can be related by
i
A
Y
c
v
A
i
u
2
where v
A
and i
u
are uncorrelated and Y
c
( G
c
jB
c
) is a
correlation admittance. In this way, two independent
noise sources (v
A
, i
u
) can be used, as in Fig. 3. This allows
the convenient expression of the two-port noise gure and
other noise parameters using superposition.
A zero-mean Gaussian noise model is assumed. Also,
white noise is assumed over the frequency range of inter-
est. The variance of the noise sources in Fig. 3 can be
written using Nyquists theorem [8] as
jv
A
j
2
_
4kT
0
BR
n
3
ji
u
j
2
_
4kT
0
BG
n
4
where / S denotes the statistical average, k is Boltzm-
anns constant, T
0
is room temperature (e.g., 290K), B is
the noise bandwidth, and the values of R
n
and G
n
deter-
mine the variances. The two-port noise gure, F [ 1
(/|i|
2
S
G
0 //|i|
2
S
G
) in Fig. 3], with source admittance
Y
s
( G
s
jB
s
), becomes [2]
F 1
1
G
s
G
n
R
n
jY
s
Y
c
j
2
5
Partial derivatives of (5) with respect to G
s
and B
s
[(@F/
@G
s
) 0, (@F/@B
s
) 0] give the optimum value of Y
s
for the
minimum F. This optimum input admittance, Y
opt
( G
opt
jB
opt
), is given by
Y
opt

G
n
R
n
G
2
c
jB
c
6
Substituting Y
s
Y
opt
into (5) gives
F
min
1 2R
n
G
c
2R
n
G
n
R
n
G
2
c
7
Extrinsic
Noiseless
Device
Extrinsic
Noiseless
Device
G
S
Correlated
D
S
i
+
_
A
v
A
(a) (b)
G
S
D
S
i
1 i
2
Figure 2. (a) Equivalent noise circuit with noise
sources (v
A
and i
A
) at the input (ABCD represen-
tation); (b) equivalent noise circuit with current
noise sources (i
1
,i
2
) at input and output (Y repre-
sentation).
Extrinsic
Noiseless
Device
G
S
D
S
i
v
A
Y
c
+
_
Y
c
u
Uncorrelated
G
Figure 3. Equivalent noise circuit with two independent noise
sources and a correlation admittance (i
A
Y
c
v
A
i
u
).
LOW NOISE AMPLIFIERS: DEVICE NOISE CHARACTERIZATION AND DESIGN 2381
Rather than expressing F as a function of G
n
and Y
c
, as in
(5), it can be expressed in terms of F
min
[using (7)] and Y
opt
as
F F
min

R
n
G
s
jY
s
Y
opt
j
2
8
A noise-gure meter is commonly used for noise mea-
surements [1416]. Measurement of the noise gure for
various Y
s
allows the determination of F
min
, R
n
, and Y
opt
in
(8) by means of a least-mean-square error t to F(Y
s
) [17].
Then, the parameters G
c
, B
c
, and G
n
can be evaluated
from (6) and (7). Measurements of this type are common,
and work has been done, for example, to relate the data to
noise sources within GaAs, GaN, and CMOS devices
[7,18,19].
5. GATEDRAIN NOISE CORRELATION EFFECT
In the simple FET circuit model of Fig. 4, one can gener-
ally anticipate correlation between the gate and drain
noise current by capacitive coupling [7]. For example,
i
d
induces the gate voltage v
g
through the transconduc-
tance g
m
. Then, v
g
i
d
/g
m
if perfect correlation can be
assumed, and i
g
joC
gs
v
g
joC
gs
i
d
/g
m
. Finally, ji
g
j
2
_
o
2
C
2
gs
ji
d
j
2
_
=g
2
m
. It has been demonstrated experimentally
[7] that, in the case of velocity uctuation, /|i
d
|
2
S is in-
dependent of frequency in the microwave operating range
of the transistor (white noise), and /|i
g
|
2
Spo
2
, thus
establishing the concept of capacitive coupling between
the gate and drain noise. In order to assess the effect of
the correlation between i
g
and i
d
, the correlation coef-
cient C is dened in normalized form as [4]
C
i
g
i
d
_
ji
g
j
2
_
ji
d
j
2
_
_ 9
With perfect correlation assumed in the ideal noise model
of Fig. 4, Cj1.
To investigate the inuence of C on the total noise, the
ideal FET noise model in Fig. 4 is considered with velocity
uctuation noise only, to calculate the minimum noise g-
ure, F
min
, in terms of /|i
g
|
2
S, /|i
d
|
2
S, and the correla-
tion coefcient C, by setting i
g
i
d
_
jjCj
ji
g
j
2
_
ji
d
j
2
_
_
(imaginary correlation only). Then, F
min
can be expressed
as [4,7]
F
min
1
ji
g
j
2
_
ji
d
j
2
_
_
2kT
0
Bg
m
1 jCj
2
_
10
Equation (10) indicates that, with no parasitics, perfect
correlation (|C|1) cancels the noise contribution from
/|i
g
|
2
S and /|i
d
|
2
S completely, giving F
min
1. There-
fore, high correlation (j1 is the theoretical maximum)
would be desirable for a low-noise device, since correla-
tion is a cancellation term in calculating the total noise of
the FET.
6. LOW-NOISE AMPLIFIER DESIGN
There are design tradeoffs between noise gure, gain, and
bandwidth. Tradeoffs exist between noise gure and gain
because the desired source reection coefcients for low-
noise gure, the optimum being F
min
, and maximum avail-
able gain values are different. The stability of the LNA
circuit is also important. If the transistor used is condi-
tionally unstable, source and load impedances presented
by the input and the output matching networks should
stabilize the circuit rst over all frequencies [20]. Addi-
tionally, the linearity performance (third-order intercept
point, i.e., IP3) must be adequate to ensure sufcient dy-
namic range of the receiver system.
Low-noise amplier design techniques have been in-
vestigated by many authors, resulting from various tech-
nologies (especially GaAs-based transistors) and circuit
architectures. Examples of low-noise ampliers, including
design techniques, can be found in a series of reprint ar-
ticles edited by Fukui [21]. Computer-aided design syn-
thesis tools provide a convenient platform.
In general, in order to achieve the lowest possible noise
gure, the input matching network is designed to provide
the source admittance for low-noise gure, and the output
matching network is optimized for high gain. The noise
parameters F
min
, R
n
, and Y
opt
should be determined rst
for the particular device being used (either from the mea-
surement or given by the manufacturer). Then, these pa-
rameters can be used in low-noise transistor amplier
design and simulation with the S parameters.
BIBLIOGRAPHY
1. H. T. Friis, Noise gure of radio receivers, Proc. IRE
32(7):419422 (July 1944).
2. H. Rothe and W. Dahlke, Theory of noisy fourpoles, Proc. IRE
44:811818 (June 1956).
3. H. Fukui, The noise performance of microwave transistors,
IEEE Trans. Electron. Devices 13:329341 (March 1966).
4. H. Statz, H. A. Haus, and R. A. Pucel, Noise characteristics of
gallium arsenide eld-effect transistors, IEEE Trans. Elec-
tron. Devices 21:549562 (Sept. 1974).
5. R. A. Pucel, H. A. Haus, and H. Statz, Advances in Electronics
and Electron Physics, Vol. 38, Academic Press, New York,
1975.
Gate
Drain
Source
v
i
d
= g
m
v
g
i
g
g C
gs
+
_
Figure 4. Ideal FET model with white-noise random processes i
d
and v
g
(i
g
).
2382 LOW NOISE AMPLIFIERS: DEVICE NOISE CHARACTERIZATION AND DESIGN
6. M. W. Pospieszalski, Modeling of noise parameters of ME-
SFETs and MODFETs and their frequency and temperature
dependence, IEEE Trans. Microwave Theory Tech. 36(1):110
(Jan. 1988).
7. S. Lee, K. J. Webb, V. Tilak, and L. F. Eastman, Intrinsic noise
equivalent circuit parameters for AlGaN/GaN HEMTs, IEEE
Trans. Microwave Theory Tech. 51(5):15671577 (May 2003).
8. H. Nyquist, Thermal agitation of electric charge in conduc-
tors, Phys. Rev. 32:110113 (July 1928).
9. T. H. Lee, The Design of CMOS Radio-Frequency Integrated
Circuits, 2nd ed., Cambridge Univ. Press, 2004.
10. M. Kawano, T. Kuzuhara, H. Kawasaki, F. Sasaki, and H.
Tokuda, InAlAs/InGaAs metamorphic low-noise HEMT, IEEE
Microwave Guided Wave Lett. 7(1):68 (Jan. 1997).
11. U. K. Mishra, A. S. Brown, S. E. Rosenbaum, C. E. Hooper, M.
W. Pierce, M. J. Delaney, S. Vaughn, and K. White, Microwave
performance of AlInAs-GaInAs HEMTs with 0.2- and 0.1-mm
gate length, IEEE Electron. Device Lett. 9(12):647649 (Dec.
1988).
12. J. W. Lee, A. Kuliev, V. Kumar, R. Schwindt, and I. Adesida,
Microwave noise characteristics of AlGaN/GaN HEMTs on
SiC substrates for broad-band low-noise ampliers, IEEE Mi-
crowave Wireless Compon. Lett. 14(6):259261 (June 2004).
13. A. van der Ziel, Noise in Solid State Devices and Circuits,
Wiley, New York, 1986.
14. A. Przadka, K. J. Webb, and D. B. Janes, Two-port noise and
impedance measurements for two-terminal devices with a
resonant tunneling diode example, IEEE Trans. Microwave
Theory Tech. 46(9):12151220 (Sept. 1998).
15. Agilent Technologies, Fundamentals of RF and Microwave
Noise Figure Measurement, Application Note 57-1.
16. Agilent Technologies, Noise Figure Measurement Accuracy,
Application Note 57-2.
17. M. Mitama and H. Katoh, An improved computational meth-
od for noise parameter measurement, IEEETrans. Microwave
Theory Tech. 27(6):612615 (June 1979).
18. A. Cappy, Noise modeling and measurement techniques,
IEEE Trans. Microwave Theory Tech. 36(1):110 (Jan. 1988).
19. P. Heymann, M. Rudolph, H. Prinzler, R. Doerner, L. Klapp-
roth, and G. Bock, Experimental evaluation of microwave
eld-effect-transistor noise models, IEEE Trans. Microwave
Theory Tech. 47(2):156163 (Feb. 1999).
20. G. Gonzalez, Microwave Transistor Ampliers: Analysis and
Design, 2nd ed., Prentice-Hall, Englewood Cliffs, NJ, 1997.
21. H. Fukui, ed., Low-Noise Microwave Transistors and Ampli-
ers, IEEE Press, New York, 1981.
LOW-PASS FILTERS
CHIU H. CHOI
University of North Florida
Jacksonville, Florida
A lowpass lter suppresses the high-frequency compo-
nents of a signal, leaving intact the low-frequency ones.
A lowpass lter specication can be expressed as shown in
Fig. 1. In the stopband (above f
2
Hz), the attenuation is at
least A
s
dB. In the passband (below f
1
Hz), the attenuation
is at most A
p
dB. The band from f
1
to f
2
is called the tran-
sition band.
A rst-order lowpass lter is shown in Fig. 2. The
transfer function is
V
out
s
V
in
s

1
RC
s
1
RC
The polezero diagram is shown in Fig. 3.
An active second-order lowpass lter is shown in Fig. 4.
The circuit is known as the SallenKey lowpass circuit.
The transfer function is
V
out
s
V
in
s

a
R
1
R
2
C
1
C
2
s
2
1
R
1
C
1
1
R
2
C
1
1 a
R
2
C
2
_ _
s
1
R
1
R
2
C
1
C
2
1
where a 1 r
2
/r
1
. The polezero diagram is shown in
Fig. 5.
An approach to design a circuit (a lowpass lter) whose
frequency response satises the lowpass requirements
shown in Fig. 1 consists of two steps: approximation of
the requirements by a transfer function and synthesis of
the transfer function.
There are several approximation methods, for instance,
Butterworth, Chebyshev, inverse Chebyshev, and Cauer
approximations. A transfer function obtained by one
method is different from those obtained by the others,
and has different properties. However, the frequency re-
sponse of each of the transfer functions satises the low-
pass requirements. The Butterworth approximation
method is described below.
Compute the scaling factor k given by the equation
below:
k
10
0:1 Ap
1
_
Attenuation (dB)
Frequency (Hz)
f
2
f
1
A
p
A
s
Figure 1. Lowpass lter requirements.
+
R
V
in
V
out
C
Figure 2. First-order lowpass lter.
LOW-PASS FILTERS 2383
Next, compute the quantity
log
10
10
0:1A
s
1
k
2
_ _
2 log
10
f
2
f
1
_ _
Choose the order n of the lter to be the smallest integer
not smaller than the quantity given above. Solve for all the
left-half-plane roots Z
i
s of the equation
1
n
S
2n
10
The Butterworth lowpass transfer function is formed as
follows:
T
LP
s
1
P
n
i 1
S Z
i
s k
1=ns=2pf
1
Example 1. Find the Butterworth transfer function for a
lowpass lter with A
s
15 dB, A
p
0.5 dB, f
1
1 kHz, and
f
2
5kHz:
k
10
0:10:5
1
p
0:35
log
10
10
0:115
1
0:35
2
_ _
2log
10
5k
1k
_ _ 1:72
Choose n2. The left-half-plane roots of the equation
1
2
s
22
1 0
are 1=
2
p
1 j. Therefore, the Butterworth transfer
function is
T
LP
s
1
s
2
2s
p
1
0:35s
p
=2p1000
which simplies to
T
LP
s
1:12810
8
s
2
1:502 10
4
s 1:12810
8
These calculations have all been performed before, and the
results and available in tabular and computer program
forms.
The SallenKey lowpass circuit can be used to realize a
second-order lowpass transfer function of the form
Ts
K
s
2
as b
Compare T(s) with Eq. (1):
K
a
R
1
R
2
C
1
C
2
a
1
R
1
C
1
1
R
2
C
1
1 a
R
2
C
2
b
1
R
1
R
2
C
1
C
2
Since there are more unknowns than equations, one can
assign values to certain unknowns and then solve for the
remaining unknowns. As an example, choose C
1
1, C
2
1, a 2. Then
K 2b R
1

1
a
R
2

a
b
The impedance-scaling method can be used to scale the
values of Rs and Cs into the practical ranges.
In general, a higher-order lowpass transfer function
obtained by Butterworth and Chebyshev approximations
can be factorized into a product of biquadratic functions
and possibly one rst-order expression. Each of the biqua-
dratic functions can be synthesized by using the Sallen
and Key or other lowpass circuits. The rst-order transfer
function can be realized by an active or passive circuit. By
cascading all the biquad and rst-order circuits together,
the lowpass lter is realized.
1
RC
j
Figure 3. Polezero diagram of the rst-order lowpass lter.

+
V
in
R
1
R
2
r
2
r
1
C
2
C
1
V
out
Figure 4. SallenKey lowpass circuit.
j
Figure 5. Polezero diagram of the SallenKey lowpass lter.

2384 LOW-PASS FILTERS
FURTHER READING
G. Daryanani, Principles of Active Network Synthesis and Design,
Wiley, New York, Wiley, 1976.
M. E. Van Valkenburg, Analog Filter Design, Holt, Rinehart, and
Winston, New York, 1982.
R. W. Daniels, Approximation Methods for Electronic Filter
Design, McGraw-Hill, New York, 1974.
G. Morchytz and P. Horn, Active Filter Design Handbook, Wiley,
New York, 1981.
LOW-POWER BROADCASTING
CHIP MORGAN
CMBE, Inc.
Bridport, Vermont
Low-power broadcasting systems are very interesting
from a technical perspective for several reasons. Although
they appear to be simple to build and easy to design, they
can actually be more critical than higher-powered sta-
tions. In fact, since the signal is local and likely to receive
or cause interference, design and optimization methods
for low-power systems can make the difference between a
successful operation and a failure.
It is a mistake to think that a low-powered facility can-
not perform as well (within its intended coverage area) as
a full-powered station. Around the world, there are many
facilities operating at low power that provide excellent
full-time service to communities every day.
Modern techniques allow some very interesting project
work, such as synchronized systems, directional systems
to ll in specific areas, and local systems (such as in tun-
nels, buildings, or temporary high-density populated ar-
eas such as stadiums). These techniques also have
application for ll-in service for satellite-delivered servic-
es as well as terrestrial digital broadcasting.
1. DEFINITIONS
A low-power station is typically a facility with a transmit-
ter power of 1kW output or less. Although antenna gain
plays a large role in the effective radiated power (ERP) of a
given facility, it is easier to dene the power of a station in
general terms by transmitter power. However, when regu-
lating the location and coverage of a given facility, ERP and
height above average terrain (HAAT) are often specied.
1.1. Translator
A translator is a radiofrequency (RF) device that retrans-
mits a television or FM signal within a specific broadcast
band. The translator alters none of the signal character-
istics except frequency and amplitude with a main pur-
pose to extend or ll in the coverage of a transmitted
signal. It typically receives a signal, changes its carrier
frequency, and then rebroadcasts the changed carrier at
substantially higher levels of radiated energy. A trans-
lator is essentially a repeater.
As an example, a broadcast station might have a zone
within its trade area that receives a poor signal from its
main transmitter. The technical consultant of the station
prepares a study to identify the existence of frequencies
for a translator. The consultant recommends a site and a
specific design.
Listeners and/or viewers within the coverage area of
the translator receive primary station programming on a
different frequency from the primary station. Thus, the
station signal is translated in that area.
1.2. Booster
A booster is an RF device that retransmits a TV or FM
signal within a specific broadcast band. It alters none of its
characteristics, including frequency or amplitude, with a
main purpose to extend or ll in the coverage of a trans-
mitted signal. A booster operates on the same frequency as
the primary signal and rebroadcasts it at substantially
higher levels of radiated energy. The signal of a booster is
the same as the primary station. Boosters are complicated,
and a poorly designed or built booster can cause much
more harm than good. A booster is an isofrequency trans-
mitter.
As an example, a station might have a location within
its trade area that receives a poor signal from its main
transmitter. The consultant of the station prepares a
study to identify a site with appropriate terrain shielding
or to allow the operation of a booster without substantial
interference to or from the main signal. Listeners and/or
viewers receive the programming from the primary sta-
tion on the same frequency as the primary station. Thus,
the station signal is boosted in that area.
1.3. LPTV
An LPTV (low-power television) station can be dened as
one that may operate as a translator or originate pro-
gramming and operate as a subscription service low-pow-
ered TV system intended to serve a local area. It functions
as a full-powered TV station, but with reduced output
power.
1.4. ITFS, MDS, and MMDS
ITFS (instructional television xed service), MDS (multi-
point distribution service), and MMDS (multichannel
MDS) are television transmission systems authorized to
provide specific programming to subscribers. Technically,
ITFS and MDS are quite similar to each other as well as to
broadcast television transmission.
2. GENERAL GUIDELINES
Following are some tips and points to consider when de-
signing a low-power system. Some of these items are sim-
ple common sense; others require high-powered computer
system analysis to compare options and identify predicted
results. These guidelines also apply to high-powered
systems. In most cases, low-powered systems have very
LOW-POWER BROADCASTING 2385
stringent requirements due to specific signal challenges.
Before designing a system, a complete understanding of
all the technical issues will minimize problems. The eld
is quite complex. For example, propagation and signal
analysis is a career unto itself, but understanding the ba-
sics will help you identify potential problems in the initial
design of a system. Use the services of a professional who
designs systems regularly for the best results.
2.1. Location of Antenna in Relation to Population
Locate the transmitting antenna site as close to the target
population as possible. Building penetration and received
signal strength decreases rapidly as distance between
transmitter and receiver increases. RF energy diminishes
with the square of the distance. If the transmitter site is
distant from the target population, system design is much
more critical and difcult.
2.2. Height of Antenna above Population
More antenna height usually means better coverage.
However, one should specify an antenna height as high
as is needed for the specific requirements, but not any
higher. Excessive antenna height leads to unnecessary in-
terference, increased cost of construction and operation,
and decreased reliability. The antenna should be just high
enough for adequate Fresnel ellipsoid clearance to the
target area but not much above that height.
2.3. Fresnel Ellipsoid Clearance of Signal into Target Market
The antenna signal is not a narrow ray of energy like a
laser beam. The signal is a noncoherent three-dimensional
beam that gets wider as it leaves the transmitting anten-
na, just like the beam from a ashlight. For maximum re-
ceived signal strength, all the energy in the Fresnel
ellipsoid must pass from the transmitter to the receiver
with no obstacles affecting the path of transmission. The
receiving antenna works exactly opposite from the trans-
mitting antenna. It gathers energy from a three-dimen-
sional area and narrows it down to the location of the
receiver, just as a telescope gathers light from a wide area
and focuses it at on small area. At the midpoint of the
transmission/reception path, the signal beam is its widest.
Therefore, even if you can see the population you want to
serve from the transmitter site, the broadcast signal may
not be able to reach it due to obstructions belowthe line-of-
sight path.
2.4. Tuning and Installing an Antenna
A transmitting antenna is just as sensitive to its immedi-
ate environment as a receiving antenna. If you have ever
adjusted a rabbit ear antenna on a TV set or the rod
antenna on a boombox, you understand the concept. A
very small change in the antenna orientation or environ-
ment has a drastic effect on its performance. Install the
antenna and tune it properly from the beginning and the
facility will perform as predicted. The supporting struc-
ture, the antenna design, and the way the antenna is
mounted all affect the quality of the signal.
2.5. Elevation Pattern of an Antenna
There is often population below the height of a transmit-
ting antennaespecially if it is on a tall tower or hill. The
elevation pattern of the antenna identies the amount of
signal radiated toward the horizon as well as the amount
radiated below the horizon, into low-lying population. The
elevation pattern is critical in systems design because
control of energy in specific areas can be an important
factor in system performance. Conversely, a receiving an-
tenna often receives from a location above its horizontal
plane, and the elevation pattern above the horizon is im-
portant. The number, spacing, and type of elements in the
antenna array typically control the elevation pattern of a
wire-type antenna.
2.6. Interference from Adjacent-Channel and
Cochannel Stations
Unless the station is located in an area with very few sig-
nals, certain portions of the service area will receive in-
terference from other nearby or more powerful
transmitters. This can come from stations as far as three
channels above or below the carrier frequency. Consider
the potential of interference in the design of a facility.
Most complaints of poor performance are a result of in-
terference, due to poor signal from the intended transmit-
ter in an area or excessive signal from an interferor or
jammer.
2.7. Receiver-Induced Third-Order Harmonic Interference
When a receiver is near other transmitters, even if the
desired station has a strong local signal, the receiver will
overload or be blanketed by the relatively higher signal
levels of the jamming transmitter. The selectivity (ability
to reject strong local interference) of the receiver deter-
mines its susceptibility to this kind of interference. In
other cases, harmonic mixes occur within the receiver
itself, causing the desired signal to be jammed. This
situation is receiver-induced third-order harmonic inter-
ference.
2.8. Condition of Antenna System
Most transmitters have a test meter that indicates if the
antenna has failed to the point that it can harm the trans-
mitter. However, there is typically no accurate test equip-
ment available at a transmission facility to show
qualitatively how well the transmitter or antenna is per-
forming. Most antenna complaints are very subjective;
you just cannot get the station as you used toit does
not sound right or look right any more, and so on. The only
way to tell if an antenna is working correctly is to test it
with appropriate test equipment used by antenna manu-
facturers and consultants.
2.9. Quality of Modulation
A commonly overlooked cause of signal problems is poor
source material. In this case, no amount of improvement
to the antenna or transmission system will correct the
problem.
2386 LOW-POWER BROADCASTING
3. LOW-POWER SYSTEM DESIGN
Site selection is the rst step in designing a system. The
details of the site determine many other design criteria, so
an overall review of the impact of any potential site must
be exhaustive. Choice of the best site for a transmission
system is often a series of compromises. Strike a balance
between economical issues, environmental issues, and
performance issues of the particular site under review.
In addition to these restrictions, the site must be avail-
able; must satisfy all regulatory, practical, and engineer-
ing requirements; and must be accessible during the time
construction is planned to take place.
The primary technical consideration of site selection is
the performance of a transmission system that is located
there. Computer analysis should be performed to predict
the performance at all intended receiving locations, to
predict the interference generated to other operators from
the proposed site, and to demonstrate the relative quality
of any particular site with others being considered. In ad-
dition to the performance of the transmitter(s) located at
the proposed site, receiving capability of the primary sig-
nal must be possible. This may mean satellite reception,
microwave link reception, or off-air reception of the modu-
lating signal.
When using off-air rebroadcast, the site selection is
critical. For example, if a booster receives its primary sig-
nal from a receiver tuned to the primary FM transmitter,
the boosters transmitted signal generally ranges from 90
to 110 dB stronger than the intended input signal. If feed-
back contamination is to be limited to less than 30 dB, the
transmitting antennas and receiving antennas have to
have between 120 and 140dB isolation.
A high degree of shielding between the antennas is re-
quired. Sometimes this can be accomplished by terrain or
even by a building. Depending on system power levels, a
physical separation between highly directional antennas
may be enough.
It is good engineering practice to choose a site that is
accessible in varying weather conditions. That is, it cannot
be subject to factors such as excessive winds, erosion,
snow, heat, or water.
3.1. Antennas
Some system designers mistakenly believe that running
all the power the license permits and using the largest
antenna they get is the best course of action. The problem
with higher-gain antennas is that they have small vertical
beamwidths that can cause deadspots in the minor lobes
due to the nulls in the radiation pattern. Since the major
lobe provides the most gain to the detriment of areas near
the antenna, signals in the remote areas will likely be ex-
cellent, at the cost of local signals. Sometimes, however,
this is an intended effect when areas near the antenna are
lightly populated.
Beamtilt will not solve a problem caused by insufcient
Fresnel ellipsoid clearance (shadowing). Antenna tuning
will also not solve this kind of problem. If a site has poor
performance due to radio shadowing, no amount of ma-
nipulation of the antenna system will make a substantial
improvement over a properly operating antenna system.
Some operators are willing to pay for minor improve-
ments, but a better use of nances is to analyze carefully
a potential site before building it. Examine every proposed
system with care to analyze accurately the magnitude of
the potential problem areas. Antenna pattern control
equals better system performance, less co-channel and
adjacent channel interference, and better spectrum utili-
zation.
3.1.1. Transmitting Antennas. There are a number of
factors to consider when it comes to designing the anten-
na system for a low-power operation. Careful attention to
materials selected is key. A poor-quality antenna will ad-
versely affect the best system. For omnidirectional usage,
use low-power versions of the standard higher-powered
transmitting antennas. (They are available in horizontal,
vertical, and elliptically polarized models.) Virtually all
transmitting and receiving antennas are available in 50-
or 75-O models. The relationship between the tower and
the radiating elements is essential. The best designs are
custom tuned for the specific situation.
When designing an antenna system, make sure that
antenna and tower manufacturers or owners are kept
abreast of any possible problem areas, such as wind resis-
tance, the antennas postinstallation directivity patterns,
and mounting procedures.
Calculate ERP by multiplying the antennas power gain
by the transmitter output in watts less losses (hybrid,
coax, duplexer, circulators, etc.). Be aware of the antennas
vertical plane pattern as well as the probable signal levels
at various points in the service area, because the measur-
ing gain on VHF antennas is usually set at zero degrees.
(The gain at other elevation angles may be considerably
less.)
Antenna beamwidth equals the number of degrees be-
tween the major lobe half-power points. If it is necessary
to have coverage in local and distant areas, choose the
highest possible antenna location and the antenna cong-
uration that will be the best compromise. A high antenna
with no obstructions between the transmission site and
the receiving sites is required for consistent service to the
intended audience.
If local coverage is of paramount importance, utilize a
high-gain antenna with a moderate degree of beamtilt, or
lower power and an omnidirectional antenna. This will
make the best use of radiated power and reduce inter-
ference.
3.1.2. Receiving Antennas for Retransmission. The input
signals delivery to the transmitter is an essential factor in
overall system design when retransmitting a signal. A
poor-quality input signal makes all the difference in the
entire operation of the system.
A simple antenna on a nearby tower is sometimes suf-
cient for off-air reception. More often, measures that are
more intricate are necessary. Sometimes, increasing an-
tenna gain or narrowing the pattern of the arrangement is
an option. A 301 horizontal pattern width from a single
receiving antenna has been known to allow for reception of
an interfering signal. A number of methods can resolve
retransmission reception difculties, including signal l-
tering to minimize out-of-band products, antenna-mount-
ed preamplication to increase signal levels, and
horizontal and vertical antenna stacking techniques to
increase antenna gain and directivity.
The majority of difculties related to the input signal
have to do with weak input signal strength, measured in
microvolts on the receiver front panel. Depending on the
manufacturer, equipment specications usually specify
that 210mV will result in a good output signal, but those
numbers are reective of equipment capabilities only.
Practical limits at a site tend to be much higher for
several reasons. Depending on local conditions, weak sig-
nals can result in fades. Upward of 100300mV is usual,
but even this level of signal can have problems during a
fade. Carefully monitoring proposed sites helps avoid the
problem from the start. The level of RF noise near a re-
ceiving antenna can inuence input. Noise levels can be
higher than the signal strength of the desired signal. Sub-
sequent ltering or antenna location can help resolve the
issue.
In many cases, one must take special measures to en-
sure that reception and retransmission signals stay clean.
A weak signal with poor signal-to-noise ratio is generally
the problem. The systems receiving section must contain
a superior signal conditioning system, preamp, and
receiving antenna to maximize clarity.
4. ANTENNA STACKING
Antenna stacking increases the gain of an antenna array.
Stacking can be either horizontal or vertical. Stacking ap-
plies in transmission and reception and can be helpful in
solving certain reception difculties.
Vertical antenna stacking inuences vertical beam-
width. It is effective in amplifying the gain of the array.
Horizontal stacking narrows the arrays beamwidth from
side to side. If the antennas are in phase, this method adds
3 dB to antenna gain for every doubling of the stack. (A
single Yagi antenna has a horizontal acceptance angle of
approximately 301 between the 3 dB down points off its
front.) Note that horizontal beamwidth decreases consid-
erably and vertical beamwidth remains unaffected.
A variation in stacking resolves certain interference
and reception difculties. There are myriad ways to make
use of this technique, including offset antenna arrange-
ments, which allow phase reinforcement off the front of
the array and phase cancellation off its rear. The process
involves vertical stacking of the antennas (i.e., one anten-
na forward of another by precisely
1
4
l at the center fre-
quency of the rejected channel). Use this formula to
calculate stacking:
Din: 2951 dividedby frequency of
undesired signal MHz
RF signals travel faster in free space than in cable, so
connect the rear antenna to the common junction with a
short feedline. Then connect the front antenna to the com-
mon junction with a feedline whose length results from
the formula multiplied by the velocity of propagation of
the cable used, plus the amount of cable that is used for
the rearmost antenna.
As another example, set the space from one antenna
boom to another so that it will cancel an unwanted signal
from a specific forward direction. Install the antennas
with the center of each spaced so the unwanted signal is
out of phase at the connection between the two antennas,
and the interfering signals will negate each other at the
combined output.
4.1. Transmitters
Transmitters provide the RF power for the antenna. Prop-
er design of a low-power facility includes selection of the
appropriate transmitter for the job. A primary consider-
ation is adequate power in order to achieve desired effec-
tive radiated power. The gain of the antenna(s) and the
length of the transmission line(s) affect this power level
requirement. The available electrical power and the cool-
ing system available at the transmission site also affect
the RF power level design. Generally, low-power facilities
do not need more complex electrical power systems, such
as three-phase power or high-voltage power systems. As in
any design, the transmitter should be well designed and
easy to maintain and provide diagnostic information about
its status and condition.
4.1.1. ITFS and MDS Transmitters. Usually, transmit-
ters for MDS and ITFS rate at 10100 W visual power
and 10% aural capability; 10-W transmitters are typically
solid state. Higher-powered transmitters typically operate
with vacuum tubes.
Internally and externally diplexed transmitters are
widely used. Low-level internal systems work well in un-
congested areas due to their ease of use and cost efciency.
Externally diplexed systems reduce cross modulation of
video synchronization components onto the aural carrier.
They also offer better rejection of products caused by the
intermodulation of aural and visual carriers. Aural carrier
signal integrity is a considerable element in system per-
formance with multichannel sound and pay television en-
coding and decoding. Amplitude, phase precorrection
methods, and contemporary linear amplier design
ensure good signal performance.
4.1.2. LPTV Transmitters. When choosing an LPTV
transmitter, there are a few terms to remember. Linear
waveform distortion, called the 2T K factor, measures
the distortion of a pictures ne detail. The 2T sine-
squared pulse and the 2 T bar dene the K factor, or ac-
tual distortion. The K factor must be less than 3% to meet
LPTV standards.
Envelope delay is the delay within the system of the
modulation envelope. It is usually a frequency function,
with higher frequencies equaling shorter delays. For
transmitters, the standard is built on a baseline of the
delay within the equipment between 0.05 and 0.2MHz.
For up to 3MHz, that delay is to be maintained. Past that,
the delay should linearly decrease up to 4.18, so the delay
is 170ns with respect to the baseline at 3.58 MHz.
Differential gain is the difference in gain of the LPTV
systems translator for a small, high-frequency signal
(chrominance) at two specific levels of a low-frequency
signal (luminance) on which the high-frequency signal is
superimposed. Differential gain cannot exceed 10% at an
average picture level ranging from 10% to 90%.
Differential phase is the difference in the output phase
of a signal such as that used in differential gain measure-
ments. It should never exceed 71 past the range of blank-
ing to white.
4.1.3. FM Transmitters. Low-power FM transmitters
are usually either a standard FM exciter with lowpass l-
ter or an exciter with a low-power amplier. Many of the
low-power systems are solid state; others operate with
vacuum tubes. In a translator or rebroadcast application,
several systems are available as complete solutions, with
built-in receivers and audio processors. In most cases, the
transmitter is simple to install and operate.
4.2. Transmission Lines
Transmission lines play a major role in the design of a low-
power system. They provide the connection between the
antennas and the equipmentwhether transmitters or
receivers. It is common for designers to pay little attention
to this important part of the design process. In fact, often,
low-power systems designers completely forget about the
existence of feedline losses. Since these installations oper-
ate at low power and are usually on a tight budget, they
generally use inexpensive small-diameter coaxial cables.
The savings in transmission line costs tend to result in a
compromised signal level.
Detailed engineering data are available to allow the
user to calculate accurately and compensate for feedline
losses. Use the best cable in low-power applications to
maximize the power delivered from the transmitter to the
antenna or from the antenna to the receiver. The antenna
can radiate only the power it receives.
Avoid using a foam-type dielectric cable unless it has a
rigid outer connector. Foam dielectric is exible and has no
need for pressurization, but extended exposure to high
temperatures can result in migration of center conductors
and impedance variations. In some cases, rigid outer con-
ductor foam dielectric cable may be more cost efcient
than air dielectric lines if it is installed carefully without
sharp bends or kinks. Keep in mind that air dielectric ca-
ble, or some alternate means of delivering nitrogen gas,
must be provided if the antenna requires pressurization.
An air dielectric transmission-line has a spiral-wound
spacer that runs along its length to hold the center con-
ductor in place. It is harder to handle and install than
foam because it is so stiff, but it is much sturdier. Air di-
electric lines have diameters ranging from
1
2
in. to 5 in., but
the most common sizes for low-power operations are
7
8
in.
and 1
5
8
in.
Here are some important points to remember about co-
axial cabling:
The coaxial cable and connector quality is just as im-
portant as that of the quality of the system compo-
nents. All it takes is one bad connector to make a
system worthless.
Use a semirigid line for optimum performance. Never
use braided cables. If there is any cross-coupling be-
tween receiver cables and transmitter cables, all the
isolation available will be unusable. Since shield
movement can cause noise in systems, secure all
exible cables to keep them safe from strain and
unnecessary motion.
Become familiar with the dielectric material, as con-
nectors are the weakest link in every coaxial cable
system. Avoid nylon connectorsthey can soak up
moisture as well as high-frequency RF energy. Poly-
styrene and Teflon are excellent choices. Only buy
superior-quality connectors from a company with a
respected name.
Avoid permanently installed adapters. Adapter con-
struction is a compromise providing greater loss
and inferior stability than using the correct cable
terminations. It can introduce high VSWR into the
system, thereby deteriorating overall performance
and decreasing isolation.
Above all, pay attention to manufacturer instructions.
Just about any good-quality connector works well if
protected from cable motion and properly installed.
4.3. Towers
The supporting structure for a low-power antenna may be
very short, or it can be 500m tall. Tower design details are
beyond the scope of this article, but it is important for the
designer of a low-power system to be aware of various as-
pects of tower use.
Towers are expensive and dangerous and have a tre-
mendous impact on the surrounding environment. Delays
in construction can last for years if proper planning and
coordination does not occur. In the initial design phase,
use of only the best-quality tower can help avoid many
devastating problems in the end.
The loss of a tower can cause more time off the air than
the loss of any other major component and can cause se-
vere damage as well. It is essential to give proper atten-
tion to the tower supporting any broadcast antenna. While
purchasing the best tower money can buy sounds like an
expensive proposition, it is nothing compared to the cost of
two towers plus any damage resulting from failure.
4.4. Frequency Separation
The strategy of frequency or channel selection for low-
power systems, particularly translators, is an art form of
itself. According to a common theory, the output frequency
of a translator should be as close to the input as possible
and ideally an adjacent channel. The reality is that prac-
tical and technical issues may preclude this possibility.
The majority of quality equipment should be able to oper-
ate well on adjacent channels, but there are other more
complex issues. For example, operation of a translator on
an adjacent frequency always causes interference to the
main station. If the translator is located in the center of a
large population, it could actually cause more harm than
good.
On the other side of the coin, there is no limit to the
frequency separation. Somewhere between the two ex-
tremes lies the solution. If maximum quality is the design
goal, use at least 2 MHz separation for an FM translator
and 60 MHz for a TV translator so that lters and tradi-
tional engineering solutions can be used in case of instal-
lation difculties.
4.5. Signal Treatments
If every installation were perfect, there would be no need
for signal treatment. The only time when it is unlikely
that special treatment for reception or transmission of a
signal will be needed is when the facility is located away
from all other transmission and reception systems and is
not near a populated area. In most of the real world, plan-
ning for signal treatment is necessary and important in
system design.
4.5.1. Receiving Preampliers. Many manufacturers
produce tower-mounted preamps with gains ranging
from 10 dB to more than 60 dB. Since the received signal
may experience increased noise and loss as it travels from
the cable to the translator, mount the preamp as close to
the antenna as possible. A high-quality amplier should
have no effect on signal reception except to increase am-
plitude to conquer cable and/or system losses as well as
render the signal useful. A preamp with greater gain does
not necessarily mean better performance than one with a
smaller gain. The reason is that high-gain preamps be-
come more subject to overload as signal input increases.
Select a preamp so that installation and input signal re-
quirements are compatible. Overloading during input will
result in distortion that can never be cleaned up after-
ward. Moreover, if a preamp experiences a poor signal-to-
noise ratio at the input, an equal or slightly worse signal-
to-noise ratio will be on the output.
4.5.2. Receiver Filtering. There are many ways to im-
plement ltering. When a low-power receiver is tuned to a
frequency near its output, increased out-of-band products
result. In early gain stages of receiving equipment, tuning
is broader, and a strong rst or adjacent second signal may
overload the input or mix in the receiver. Increased sharp-
ness of input tuning can result in an increase in system
noise.
Traps or bandpass lters at the input or at the tower
may help in certain situations. Usually, a bandpass lter
attenuates everything but the desired signal and a trap
may attenuate a specific signal, causing interference.
4.5.3. Transmitter Filtering. Bandpass cavities are com-
monly used to reduce a transmitters sideband noise. They
pass the desired signal with a minimum of loss while
sharply attenuating those frequencies that lie above and
below the passband.
Notch cavities have a response curve that is the reverse
of a bandpass lterit possesses the same general
conguration. Some people call a notch cavity a trap l-
ter or a reject because of its ability to pass the desired
frequency while it suppresses the rest.
Pass reject (sometimes called pass notch) cavities are
a combination of the two other cavities. When the inter-
fering frequency is extremely close to the undesired fre-
quency, a pass reject is used.
4.5.4. Combiners. Wireless communications usage in-
creases every day, and tower or system space is becoming
more difcult to arrange due to environmental pressures,
lack of real estate, and costs. Transmission sites are be-
coming crowded, and these trends are resulting in a great-
er need to mix two or more transmitters into one antenna.
Combiners do a number of good things at the same time.
They lter potential interference and intermodulation,
they provide isolation between transmitters, and they al-
low two or more transmitters to be mixed into one coax
run and antenna. Although combiners themselves are rel-
atively expensive, the cost savings to operators of tower
sites are high.
Cavity combiners are generally manufactured from in-
tercabled bandpass cavities. The isolation is the result of
the cavitys resonant response curve. Sometimes pass re-
ject cavities provide for closer frequency spacing or better
isolation.
Advantages of cavity combiners include exible cong-
uration, cost efciency, and low insertion loss. Negative
features include large size, reduced channel capacity, and
possible redesign of a combiner when frequencies are
changed or added.
Use a combiner when the frequency spacing between
channels is extremely wide. The minimum spacing for the
FM band is usually about 1 MHz.
5. INSTALLATION AND TROUBLESHOOTING
Every transmitter site has a potential for problems. While
it is impossible to anticipate every conceivable difculty,
proper foresight and planning can help avoid the likeli-
hood of many disasters.
5.1. Boosters
Boosters may experience two kinds of major problems. The
rst problem is when the boosters output corrupts the
input with the boosters signal. The second example occurs
at a location that receives signals from the booster and the
original station.
When isolation between receive and transmit antennas
is less than the boosters overall gain, the result is a power
oscillator rather than a power amplier. Logperiodic an-
tennas rather than omnidirectional antennas are highly
recommended. Self-oscillation takes place when the boost-
er is transmitting an unmodulated signal and the boost-
ers power meter indicates normal operation. Try turning
the RF gain control down. If the transmitted signal sounds
clean, advance the gain control until the system breaks
into oscillation and nd a compromise gain setting.
If that does not solve the problem, then the installation
is to blame. Disconnect the receive antenna from the down
converter. If the oscillation persists after the antenna is
disconnected, there are a few things to do:
The transmission line is carrying energy to the down
converter. Burying the interconnecting transmis-
sion line underground often resolves the situation.
If there is no energy present at the downconverter end
of the transmission line, relocate the transmitting or
receiving antennas or both. Or shield the receiving
antenna by locating it over the edge of a hill or be-
hind an obstruction such as a water tower. One more
solution is to increase the separation between the
transmit and receive antennas until the system is
stable.
5.2. Translators
If an installed translator fails to operate correctly from the
beginning, the cause is nearly always site related. Problems
can generally be traced to at least one of the following:
Receive and Transmit Antennas Are Too Close. Trans-
lator output power is usually in the millions of mi-
crovolts while the input signal is just a few
microvolts. Poor translator performance is the result
of spacing the antennas a few meters apart because
the relatively high-powered output puts too much
strain on the receive section. A distance of at least
15 vertical meters is recommended. To ascertain
whether antenna separation is the problem, replace
the transmit antenna with a dummy load.
Incoming Signal Is Poor-Quality. Translators cannot
improve the quality of signal they receive. Some
translators are operational with signals from prima-
ry stations as much as 160 km away, usually when
the elevation is high at the transmit and receive
points and low at the transmission path. It is impor-
tant to monitor the received signal for a good length
of time using recording equipment in such cases,
since signal quality can come and go.
An Adjacent Channel Is Present. This can be exasper-
ating, because the adjacent channel can sound weak
and yet be powerful enough to create distortion in the
desired signal. When modulation components overlap
on adjacent channels, mixing occurs, resulting in de-
ceptive product generation. The problem can be xed
through special intermediate lters or external lters
on the receive antenna.
A High-Powered Station Is Nearby. Serious perfor-
mance issues can arise when the FM stations output
energy mixes with that of the translator. This is a
common problem. Tubes and output transistors are
not totally linear and thus tend to operate as mixers,
producing false signal generation that interferes with
sensitive receiving equipment. Because the problem
goes away when the translator is shut off, inexperi-
enced technicians tend to blame the translator. The
truth is that all translators react this way under the
same conditions. The goal is to keep external signals
from the nal RF stage of the translator. Installing an
isolator in the output transmission line is a cost-ef-
cient way to solve the problem. A second option is to
employ a high Q cavity in the output, which bypasses
the translator output and rejects the offending signal.
Input and Output Channels Are Too Close in Frequency.
The majority of transmitting equipment (including FM
translators) produces spurious emissions extending
several megahertz from the principal carrier. These
signals possess an innite amount of signals with the
same frequencies as the desired signal so they inl-
trate the translator and are amplied. Solve the prob-
lem with good antenna and frequency separation,
usually with at least 1MHz between input and output.
Harmonic Problems Cause Interference. The transla-
tors main carrier is almost always the cause. It over-
burdens the front of the translator or the preamp
associated with it. Although the translator suppress-
es harmonics by at least 60 dB, the seemingly obvious
addition of an extra external harmonic lter will not
do a bit of good. By placing a bandpass lter into the
translator in front of any preamplication, the inter-
fering signal will be too feeble to cause difculties.
Alternatively, try moving the translator installation
far enough away that its carrier will not be powerful
enough to do harm.
5.3. ITFS and MDS
There are a number of potential RF system difculties
with IFTS and MDS. Both services use 100W or less per
channel, so eld intensities are fairly low in contrast to
full-service broadcast. The ITFS and MDS services rely
heavily on radio line-of-sight coverage, so the use of fre-
quencies greater than 2000 MHz creates a need for better
consideration of signal path clearances. The receive sys-
tem needs a low-noise-gure downconverter (which gen-
erally changes a block of the ITFS/MDS band frequency to
VHF) and a relatively high-gain antenna. Other problem
areas include adjacent-channel and cochannel interfer-
ence. As the spectrum becomes more crowded, the proba-
bility of interference grows in relation. In response, many
cities are planning adjacent channel systems. A number
of MDS systems, in particular, are already experiencing
cochannel interference problems. As the ITFS band grows,
so does the likelihood that it, too, will develop problems.
Luckily, a good number of the common engineering
principles that solve UHF and VHF broadcast difculties
are proving helpful. Suppressing out-of-band products is
essential, as older transmit systems frequently display
high levels of lower sideband reinsertion and out-of-band
product formation. Externally diplexed linear transmit-
ters and waveguide lters can bring these products to more
workable levels. Offset frequency operation can be utilized
with relatively high levels of cochannel interference.
The area of best signal coverage is within the areas re-
ceiving the entire Fresnel ellipsoid. These areas can ex-
tend beyond visible line of sight due to the refraction of
electromagnetic waves in the atmosphere. Use a path
clearance of at least 0.6 Fresnel zone to avoid excessive
diffraction loss due to path obstructions. The formula is as
follows:
Rwavelengthd
1
d
2
=d
1
d
2
1=2
d
1
the distance fromthe transmitter to the obstruction
d
2
the distance fromthe obstructionto the receiver
Path clearances above 1.3 Fresnel zones can yield multi-
path propagation, nulls in received levels, and/or picture
ghosts in the received pictures. Moving the antenna
slightly can help alleviate the problem.
5.4. Spurious Emissions
Looking at the data that come with the translator is a good
way to ensure that the system is clean and clear of un-
wanted emissions, except for harmonic products, which
are undetectable by the test data measuring output per-
formance. Low levels of distortion and high signal-to-noise
ratios generally indicate clean transmission. The problem
with unwanted emissions is that the undesirable products
they generate fall inside and outside the assigned output
channel.
When there are several stations located at one trans-
mitter site where you are receiving interference, there
may be harmonic mixes between the other stations that
fall on or near your frequency. Sometimes these harmonics
originate within the transmitters involved in the mix
and if the spurious harmonic signals are above a certain
threshold, they may be in violation of governmental rules.
The existence of this type of interference is veried by us-
ing a spectrum analyzer at the towers of the offending
stations. The source of this kind of interference can be
veried only by connecting test equipment to the various
transmitters involved.
Most complaints are about interference to nearby tele-
vision translators. Usually the harmonic stems from the
television translator, not the FM translator. Any overload-
ed amplier is subject to harmonic generation. Installing a
highpass lter at the output of the receive antenna might
help. The advantage to going this route is that low loss of
the lter does not compromise the integrity of the TV
translator. This method could fail if the FM and TV trans-
lator antennas are just a few meters apart. If a lter does
not work and the antennas are properly spaced, cabinet
radiation or power line feedthrough is the culprit.
5.5. Spectrum Analyzers
Spurious emissions are not necessarily transmitter relat-
ed. Spectrum analyzers can be unreliable when it comes to
harmonic measurements if not used properly. Towers and
other metallic objects can reradiate signals and cause
nonlinear distortion, resulting in spurious emission.
Use a spectrum analyzer to examine the output channel.
If the translator datasheet reveals that solid data yet un-
wanted emissions are present, the cause is usually an adja-
cent channel. There are several courses of actions to resolve
the issue. Use narrowband lters in the IF section of the
translator, reduce the signal level of the interfering adjacent
channel, or increase the level of the desired channel.
5.6. Desensing
Desensing is the result of brute-power RF transmissions
in the surrounding area. It is entirely possible to have a
translator operating on 10 mV, for example, and a 100-kW
station 1km away. Intermodulation occurs when the
translator tries in vain to accommodate the lower-level
input signal while being hit by the much stronger signal,
which runs into the rst translator it sees. Filtering on the
input line is the best way to solve the problem before the
signals mix in the rst place. If the site lacks a strong
enough input signal to overcome the insertion loss of a
selective lter, consider using special preamps designed
just for that purpose. Relocating or redesigning the input
antenna structure is another viable option. Other methods
include raising the receive antenna, adding a preamp to
overcome line loss, adding or stacking receive antennas,
and relocating receive antennas (or perhaps the entire
site).
After translator installation is complete and the power
is turned on, there is nothing but hash and noise. But once
the translator is off and the receiver is hooked up to the
receive antenna, there is no problem. This is because any
translator conveys undesirable products as well as the de-
sired modulation. These products possess a wide band-
width, which is usually several megahertz from the main
carrier frequency. Generated products are composed of the
energy caused by the random motion of electrons (Johnson
noise) and sideband components from the desired modu-
lation. Generally, the undesirable products are quite
weak, so they cannot be observed on a spectrum analyz-
er without suppressing the main carrier. Regardless of
whether the undesirable products are modulation compo-
nents or noise, some of these undesirable products will be
in phase between the input and the output of the trans-
lator.
5.7. Problem Solving
Providing a lter at the translators output can solve
many problems. The purpose of the lter is to subdue
those undesirable products that fall on the translators
input channel, but execution is more difcult than it
sounds. Note that the quality factor, known simply as Q,
ascertains how lossy the lter will be in operation.
If a translator operates at 100.1 MHz, it would need a
lter to prevent undesirable products from falling on the
input channel at 102.1 MHz. Two megahertz is the sepa-
ration factor, so to make sure that no products from output
fall on the input, the bandwidth cannot be more than
1 MHz. So Q100.1/1 100.1 is the minimum value. To
have a lter with a low insertion loss and a minimum of
energy loss, the Q for every element inside the lter has to
be 100 times the minimum, or 10,000 times. But the re-
alization of such a Q factor necessitates the use of costly
cavity-type lters.
It is much more cost-efcient to use Q factors of around
800 for the lter elements. The lter uses a combination of
two techniques: bandstop and bandpass. The bandstop
section is a slot about 300kHz wide, adjusted to the input
channel frequency.
The bandpass section passes carrier frequency and use-
ful modulation data. This low-loss device adds more selec-
tivity that optimizes the systems performance. It can also
decrease the radiation of out-of-band signals. A good
illustration would be between an antenna and transmit-
ter combiner, where the lters job is to overpower any
harmonics originating in the ferrite isolators and down-
grade any broadband transmitter noise so that receive
sensitivity remains intact. Energy from the translator
output that falls on the input frequency is attenuated by
approximately 50 dB. Using an example of 5 mV worth of
undesirable products falling on the input after passing
through the lter, the magnitude is reduced by 50 dB
(0.0158 mV the SNR10/0.015863256 dB. Now per-
formance is excellent.
BIBLIOGRAPHY
1. G. Bostick and R. E. Mohar, Introduction to Low Power Tele-
vision Broadcasting, Communications & Energy Corp., East
Syracuse, NY, 1998.
2. G. W. Durling and C. Cuevas, Lower power Ku-band direct to
home video broadcasting in rainy regions, Space Commun.
12(34):175184 (1994).
3. K. Nakamura and T. Nakamura, The helical loaded monopole
antenna, Electron. Commun. Jpn., Part 1 (Commun.)
74(6):3741 (1991).
4. G. G. Luettgenau, Low-power television gets solid-state boost,
Microwaves RF 23(3):165171 (1984).
5. International Telecommunication Union. Troisie`me edition de
lAnnexe l a` lAccord regional (Gene`ve, 1975): contenant, le
plan dassignations de frequence pour les stations de radio-
diffusion a` ondes kilometriques et hectometriques (regions 1
et 3), lappendice au plan, canaux pour emetteurs de faible
puissance, lappendice 2 au plan, gain de lantenne (en dB),
janvier 1987/Union internationale des telecommunications
third edition of Annex 1 to the Regional agreement (Geneva,
1975): containing, frequency assignment plan for LF/MF
broadcasting stations (regions 1 and 3), appendix to the
plan, low-power channels, appendix 2 to the plan, antenna
gain (dB), January 1987/International Telecommunication
Union, Gene`ve: LUnion, 1987.
LOW-TEMPERATURE COFIRED CERAMIC
(LTCC) TECHNOLOGY IN RF AND
MICROWAVE ENGINEERING
KE-LI WU
The Chinese University of
Hong Kong
Shatin, NT, Hong Kong
1. INTRODUCTION
LTCC is a rapidly developing passive integration and
package technology that offers an attractive solution to
RF and microwave products in three aspects: (1) em-
bedded passives, (2) packaging and system integration,
and (3) low manufacturing cost for high-volume produc-
tion, featuring excellent electrical performance at RF and
microwave frequencies.
LTCC has distinct advantages in embedded passives
and packaging in that it provides a harmonic carrier for
ICs (digital and analog) with embedded passives, control
and power lines, and functional accessories. As far as
interconnect is concerned, LTCC offers the capability of
ne pitch on each layer. Since each layer is processed
individually, LTCC is more cost-effective than any other
sequentially processed technologies.
As shown in Table 1, another advantage of LTCC is its
stability and relative insensitivity to environmental
stress. The dense structure of a red ceramic body has
very low permeability to moisture, which allows LTCC
parts to perform in high humidity without special packa-
ging or protection [1].
The three key areas for LTCC technology are the
development of LTCC tape and conductive paste materi-
als; the manufacturing process that leads to high-preci-
sion, high-reliability, and low-cost products; and the
design technology that needs to take all the necessary
electromagnetic couplings and parasitics into account.
In developing dielectric material for an LTCC system,
the electrical and thermomechanical aspects must be
investigated for enabling the production and usage of
reliable components. In the case of RF applications, the
dielectric properties such as low-loss tangent [tan(d)],
permittivity (e
r
), and coefcient of thermal expansion
(CTE) are important. The typical properties of three major
commercial LTCC tape systems are listed in Table 2. After
a tape system is developed, matched conductive paste
must be developed for different usages such as solderable
and wirebondable paste for surface-mounted components
and IC dies, solderable paste for BGA (ball grid array)
attachment, and conducting paste for inner layer and via
lling.
LTCC tapes need to withstand handling and via punch-
ing and produce, after lamination and low temperature
coring, a solid material with the desired properties. A
good surface smoothness and compatibility with em-
bedded conductive lines are also essential properties [2].
The current LTCC products can be divided into two
major categories: functional components and modules.
Chip LC lters, baluns, couplers, and chip antennas are
good examples of functional components. On the other
hand, modules usually integrate active devices such as IC
dies with other required passive components to form a
subsystem. In designing LTCC modules or components,
electromagnetic phenomena must be taken into account as
Table 1. Performance Characteristics of Dielectric
Materials
Materials
Process
Temperature
(1C)
Expansion
Coefcient
(ppm/1C)
Moisture
Permeability
Polymers o200 25100 Medium to high
Metals 4600 2025 Low
Ceramics 8001700 410 Low to medium
LOW-TEMPERATURE COFIRED CERAMIC (LTCC) TECHNOLOGY IN RF AND MICROWAVE ENGINEERING 2393
accurately as possible. This demands the design tools to
incorporate electromagnetic (EM) simulation, in conjunc-
tion with the standard circuit-level simulation function for
system-level design. Since the LTCC circuit is a multilayer
conguration with vias connecting the circuitry on differ-
ent layers, the EM simulation tools have to be able to
handle the current owing in both horizontal and vertical
directions. Additionally, if the thickness of printed con-
ductors is noticeable as compared to the thickness of the
dielectric layer, the conductor thickness must be ac-
counted for in an accurate design. The research for
effective EM design of LTCC module and design meth-
odologies has attracted a great deal of attention in the
eld.
2. HISTORY OF LTCC [3]
Like other good invention, LTCC was born of necessity. In
1982, Hughes Aircraft was concerned with developing a
process for a group of high complexity multilayer sub-
strates. The solution to this was a dielectric layer that was
tape-cast. The tape could be via-punched, lled, and
printed with conductor and then red in place. This
process became known as Ttran and formed the basis for
the three generations of LTCC that we have gone through
to date.
The rsthand cast formulations of LTCC tape were
based on existing multilayer dielectric pastes from Du-
Pont, ESL, Ferro, and other manufacturers. While made
with the most primitive methods to demonstrate the
technology and with the aid of some sewing needles for
via formation the rst real LTCC parts were made in the
lab. This grew into the DuPont 851 Green Tape. The
dielectric loss of this material at microwave frequencies
prompted the development and commercialization of the
Ferro A6M system. There were a great deal of renements
and development of additional inks (both metallic and
resistive), and this completed the development of LTCC
generation 1.
The environmental movement was key in spurring
DuPont to develop the 951 formulation. This eliminated
the made with chlorinated solvents stigma and allowed
continued marketing in Europe and Asia. Improvements
in the organic portion of the formulation were made as
part of this transition greatly improving produceability
and reducing costs. At about the same time Ferro sought
to broaden their offering by creating lower cost-versions of
their agship A6M product, including A6C, A6B, and the
still available (as of Sept. 2003) A6S.
Nearly concurrent with this effort, Heraeus developed
the CT600 and CT800 materials that are nearly equiva-
lent to the Ttran and 951 products. Additionally Heraeus
worked toward constrained sintering for zero shrink
production process. The market for LTCC was growing
rapidly during this time with many commercial adoptions
in areas such as automotive applications and modules for
wireless communication system. The bursting of the In-
ternet and wireless bubbles slowed down the worldwide
economy enough to end the development phase of LTCC
generation 2 by mid-2001.
Generation 3 of LTCC technology marks the maturing
of the LTCC industry. New compositions targeted not at
increasing technical merit but lowering system cost. In
one instance, Motorola introduced T2000 material, which
was designed with its physical properties balanced to have
a zero T
f
. This material has been commercialized by
Heraeus as CT2000. Heraeus has also developed this
material in a self-constrained version as HL2000 under
their HeraLock patent as a way to increase array preci-
sion, thus allowing larger arrays for lower assembly cost.
Another aspect of LTCC generation 3 is the emergence of
low-cost LTCC foundries in the Far East, Taiwan, and
mainland China. This further enables low-cost manufac-
turing but at the same time requires the designer to have
the expertise to do the full design and engineering of the
product.
Table 2. Typical LTCC Tape Properties
Property Dupont 951 Dupont 943 Ferro A6M Ferro A6B
Heraeus
CT2000
Heralock
HL2000
Color Blue Blue White Black Light blue Light blue
Available red thickness
(mils)
1.7, 3.5, 5.1, 7.8 4.4 3.7, 7.4 3.3, 6.7 0.8, 1.6, 3.1, 4.1,
8.0
3.5 to 3.75
Dielectric constant 7.8 7.5 5.9 6.5 9.1 7.3
Loss tangent (%) 0.15 o0.1 o0.2 o0.5 o0.2 o0.26
Insulation resistance 410
12
O per
layer
410
12
O per
layer
410
12
O per
layer
410
12
O per
layer
410
13
O cm 410
13
O cm
Breakdown voltage 41000V per
mil
41000V per
mil
4900V per mil 41000V per
mil
41000V per
mil
43000V per
mil
Fired density (g/cm
2
) 3.1g/cm
2
3.2g/cm
2
2.50g/cm
2
2.50g/cm
2
2.45g/cm
2
2.9 g/cm
2
TCE (ppm/1C) 5.8 6.0 7.5 910 5.6 6.1
Flexural strength 320MPa 230MPa 4210MPa 4210MPa 310MPa 4200MPa
Thermal conductivity 3.0W/mK 4.4W/mK 2.0W/mK 3.0W/mK 3.0 W/mK
Surface roughness o0.7 mm o0.7mm o0.7mm
Flexural strength 28.3kpsi 33.4kpsi 17.1kpsi 17.1kpsi 310MPa 4200MPa
XY shrinkage (%) 12.770.3 9.570.3 15.070.2 14.570.2 10.670.3 0.160.24
Z shrinkage (%) 1570.5 10.370.5 25.070.5% 35.070.5% 16.071.5 32.0
Metallizations Au/Ag-Ag-Au Au/Ag-Ag-Au Au/Ag-Ag-Au Au/Ag-Ag-Au Au/Ag-Ag-Au Au/Ag-Ag-Au
2394 LOW-TEMPERATURE COFIRED CERAMIC (LTCC) TECHNOLOGY IN RF AND MICROWAVE ENGINEERING
3. BASICS OF LTCC DIELECTRIC MATERIALS AND
PROCESS
Dielectric materials used for LTCC components must have
proper ring, dielectric, and thermomechanical properties
so that mechanically reliable multilayer structures with
the desired electrical performance can be produced.
Proper ring properties means that the dielectric ma-
terial has to achieve a dense, nonporous microstructure
below 9501C so that it can be cored with low resistive
conductor materials such as silver or copper. As revealed
by Table 3, unlike traditional cored ceramics with ring
temperature in excess of 14001C and only those low
conductive with high melting point metals such as tung-
sten can be embedded, by keeping the ring temperature
below 9501C, all the metals with high conductivity can be
used.
For commercial LTCC materials, the ring prole is
specied accurately in the information given by the man-
ufacturers, and for new dielectric materials, this is an
essential part of their development. Figure 1 shows a
typical ring prole for LTCCs. Steps 1, 3, and 5 are
heating and cooling periods, step 2 is for burning out the
organic additives, and step 4 is the actual sintering period.
For Heraeus CT2000, for instance, the recommended
ring prole is 3.01C/min to 1001C, 3.01C/min to 4501C,
5.51C/min to 8658751C, held for 2030 min. The cooling
rate is approximately 6101C/min (furnace cooling rate).
Firing must be done on at setter material since the tape
will conform to the setter material.
The permittivity of LTCC dielectric is particularly
important because the length of a resonator is roughly
inversely proportional to the square root of the permittiv-
ity. The LTCC materials available commercially have
permittivities of 510, but higher values (r85) are also
available for noncommercial materials.
The number of possible different materials and compo-
sitions to decrease the ring temperature of crystalline
ceramic is countless. The most common method is to use
SiO
2
-based glass compositions.
The tape-casting process consists of a slurry prepara-
tion and its casting as shown in Fig. 2. The basic formula
for the tape casting slurry includes the ceramic powder,
solvents, and organic additives. In the red product the
ceramic material is all that remains, and thus the solvents
and additives have to be removed in earlier process steps.
These ingredients are added only to facilitate the fabri-
cation of the tape with an appropriate thickness and
density and to make the tape strong enough for subse-
quent processing [4].
After drying the tape in air, it is further processed to
form multilayer components as shown in Fig. 3. The tape
is rst blanked to size and the required vias are formed
using mechanical or laser punching. After via lling and
screen printing of the conductive patterns on each sepa-
rate layer, they are stacked together, laminated, and
cored.
4. LTCC EMBEDDED PASSIVE COMPONENTS
The ability to embed passive elements is one of the most
attractive features of LTCC technology. Simple passive
elements include inductors, resistors, capacitors, LC
shunt elements, LC series elements, and various func-
tional circuits such as lters. Some common congurations
of embedded passive components are listed in Table 4. The
main parasitic capacitors in each circuit layout are marked
as C
p
in the circuit models.
The circular series inductor on a single layer shown in
(a) may provide about 1015% higher Q value than its
rectangular shaped counterpart shown in (b). The induc-
tors in (c) and (d) are the implementation of the inductors
with multilayer windings to increase the inductance va-
lue. Usually, the inductor winding layers should be kept as
far away from the ground as possible to minimize the loss
[5]. To trade off the occupied area and the Q value, the
ovum-shaped inductor shown in (e) would be a good
choice. Capacitors are usually in rectangular shape as
shown in (f) and (g). Multilayered interdigital capacitors
in (h) are commonly used for high-value capacitance. By
combining the basic elements of inductors and capacitors,
we can construct various simple functional LC networks
such as capacitor p network (i), LC series network (j), and
LC resonator tank (k). In constructing the LC resonator
tank it is wise to lay the inductors above the shunt
capacitor to minimize the loss caused by the ground and
the parasitic capacitance on the inductor. This guideline is
also applicable to the design of the lowpass network (l) and
the highpass network (m). In the bandpass lter with one
transmission zero (n) [6], the mutual coupling M is
realized by two side-by-side inductor strips. The mutual
coupling M in the bandpass lter (o), which has one
transmission zero at the lower stopband and a transmis-
sion zero at the upper stopband, can be created by a
overlapped two inductor strips [7]. It is worth mentioning
Table 3. Conductivity and Melting Points for Some
Commonly Used Conductors
Conductor Conductivity s (S/m) Melting Point (1C)
Silver (Ag) 6.1710
7
961
Copper (Cu) 5.810
7
1083
Gold (Au) 4.1010
7
1063
Tungsten (W) 1.81510
7
3370
1
2
3
4
5
T
e
m
p
e
r
a
t
u
r
e
Time
Figure 1. Typical ring prole for LTCC components.
that in designing high-performance lters, the parasitic
coupling between the input and the output must be care-
fully taken into account [8].
5. LTCC RF MODULES
In most radiofrequency (RF) applications, in order to meet
size requirements, a large part of the circuit is usually
integrated in a semiconductor die. However, the required
functional passive circuits in the system such as lters
and baluns form an exception due to the power handling
and the requirement for low signal losses. The losses in
transmit and receive channels directly inuence the
power consumption and receiver sensitivity. With this
regard, LTCC provides an option to integrate those re-
quired high-Q passive circuits while providing a packa-
ging substrate to the Si or GaAs IC dies. LTCC enables
low-loss stripline designs and geometries realizable with
standard screen-printing techniques. In addition, the use
of thick-lm silver metallization in LTCC substrate en-
sures lower conductor loss than building circuits on Si or
GaAs.
A good example of LTCC modules is the antenna switch
diplexer front-end module for GSM/DCS/PCS applica-
tions, in which the diplexer is realized using lowpass
and highpass lters. In addition to a lowpass lter at the
switch end of DCS band, some elaborately designed re-
sonant circuits are built in the diplexer circuit to trap the
second, third, and fourth harmonics. Figure 4a shows the
lumped-element model for an LTCC antenna switch di-
plexer. The 3D layout of a realization of the lumped-
element model is illustrated in Fig. 4b with its response
shown in Fig. 4c. Due to the strong EM couplings in the
embedded passives, an accurate EM modeling and ef-
cient optimization scheme must be used in the design [9].
6. LTCC LAMINATED WAVEGUIDE AT
MICROWAVE FREQUENCY
At microwave frequencies, the dielectric loss becomes
more signicant as compared to the cases in RF frequen-
cies. A part of the loss is contributed by the surface wave.
On the other hand, due to the lling effect of a high-
dielectric-constant material, some low-loss transmission
lines whose dimension is considered as bulky in conven-
tional applications become suitable in LTCC applications.
A good example of this type of transmission line is the
laminated waveguide [10], a derivation of rectangular
Filtering & de-airing
Tap casting
Mixing
Milling
Characterization
Solvents
Plasticizer
Ceramic powder preparation
Viscosity control
Binder
Dispersants
Figure 2. General production process of cera-
mic tape casting. (This gure is available
in full color at http://www.mrw.interscience.
wiley.com/erfme.)
1
2
3
4
5
6
!
n
Layer
index
Via
punching
Via
filling
Conductor
printing
Stacking Lamination Burnout &
firing
Figure 3. General production process of
LTCC multilayer components. (This gure
is available in full color at http://www.mrw.
interscience.wiley.com/erfme.)
Table 4. Basic Passive Components that Can be Embedded in LTCC Substrate and Their Equivalent-Circuit Models
Description Typical Circuit Layout Equivalent-Circuit Model
(a) A single-layer circular inductor
Cp Cp
Ls
(b) A single-layer rectangular inductor
Cp
Cp
Ls
(c) A multilayer rectangular inductor
Cp Cp
Cp
Ls
(d) A multilayer circular inductor
Cp Cp
Cp
Ls
(e) An ovum-shaped inductor
Cp Cp
Cp
Ls
(f) A two-layer shunt capacitor
Cs
(g) A single-layer series capacitor
Cp Cp
Cs
(h) A four-layer series capacitor
Cp Cp
Cs
(i) A capacitor PI network
Cp Cp
Cs
(j) An LC series network
Cs
Ls
(k) An LC resonator tank
Ls
Cs
(l) A low pass network
Ls
Cs
Cp
Cs
(m) A highpass network with one transmission zero
Cs
Cp
Cs
Ls
(n) A bandpass lter with one transmission zero
M
(o) A bandpass lter with two transmission zeros
M
waveguide. As shown in Fig. 5, a laminated waveguide is
composed of an LTCC substrate, a pair of main conductive
layers on the upper and lower surfaces of the substrate,
and a pair of conducting walls formed by a plurality of via
holes extending in a thickness direction. The inner con-
ductive layers connecting the via holes play a very im-
portant rule in reducing the leakage of electromagnetic
waves. The laminated waveguide has also been used in
ridge waveguide bandpass lters [11] and surface-
mounted bandpass chip lters [12].
In developing an LTCC integrated module at micro-
wave frequencies, people always need to use various
transition and packaging structures to connect one type
of transmission line to another. For example, some com-
mon transitions between laminated waveguide to stri-
plines were proposed in a U.S. patent [13]. A waveguide
microstrip line transition was proposed in another U.S.
patent [14]. A broadband transition between a waveguide
and an LTCC laminated waveguide was developed by
Huang and Wu [15].
7. CONCLUDING REMARKS
It has been suggested that LTCC as a high-density packa-
ging and passive integration technology offers many ben-
ets over other competing technologies. However, three
major limitations need to be mentioned: shrinkage control,
strength, and thermal conductivity. These concerns have
been addressed by the industry by introducing zero
shrinkage materials, improving materials, fabrication
process, and thermal via/heatsink technology. For high-
power applications, LTCC-on-metal (LTCC-M) technology
is a good candidate. Another cost-effective way to dissipate
heat is the use of cavities in the LTCC structure, allowing
access to the underlying heat spreader.
While continuing reduction of material cost, develop-
ment of a large format process will be another direction
toward cutting manufacturing cost.
Today, LTCC products can be found in radar, space,
medical, and wireless communication systems; computer
hardware; biotechnology; satellite communication; and
automobile manufacturing. It is expected that with steady
improvement of LTCC materials and process technologies,
an increasing number of new applications will appear in
our lives.
50
40
30
20
10
0
0.5 2 3.5 5 6.5 8
Freq (GHz)
|
S
i
j
|
(
d
B
)
(c)
SWITCH
SWITCH
Antenna (PORT1)
PORT2
PORT3
LPF (GSM850/900) HPF (GSM1800/1900) LPF (GSM1800/1900)
SM Inductor
(a)
LPF
(850/900)
HPF
(1800/1900)
LPF
(1800/1900)
Switch
(b)
Figure 4. (a) The lumped-element circuit model for an LTCC GSM antenna switch diplexer;
(b) layout design of the schematic shown in (a) with size of dimensions 4.53.20.5 mm
3
;
(c) response of the antenna switch diplexer. (This gure is available in full color at http://
www.mrw.interscience.wiley.com/erfme.)
Via-hole
Pitch size
Via-hole
Dielectric
Layers
Conductive
layers
Port
Port
Figure 5. Schematic diagram of laminated waveguide.
BIBLIOGRAPHY
1. R. L. Brown and W. R. Smith, Embedded passive functions for
RFand mixed-signal circuits, Proc. 6th Int. Conf. Multichip
Modules, 1997, pp. 351356.
2. P. Barnwell, W. Zhang, J. Lebowitz, K. Jones, N. MacDonald,
C. Free, and Z. Tian, An investigation of the properties of
LTCCmaterials and compatible conductors for their use in
wireless applications, Proc. Int. Symp. Microelectronics,
IMAPS, Boston, 2000, pp. 659664.
3. M. Ehlert, private communication, 2003.
4. C. Berry, D. Parlow, T. Vasilow, S. Gurkovich, and A. Bailey, A
design of experiment for a tape casting process, Proc. Int.
Symp. Microelectronics, IMAPS, Boston, 2000, pp. 150155.
5. A. Sutono, D. Heo, Y.-J. E. Chen, and J. Laskar, High-QLTCC-
based passive library for wireless system-on-package (SOP)
module development, IEEE Trans. Microwave Theory Tech.
49(10): 17151724 (2001).
6. A. Sutono, J. Laskar, and W. R. Smith, Development of
integrated three dimensional Bluetooth image reject lter.
Proc. Int. Microwave Symp., IEEE, Vol. 1 2000 1116.
7. L. K. Yeung, and K.-L. Wu, A compact second-order
LTCCband-pass lter with two nite transmission zeros,
IEEE Trans. Microwave Theory Tech. 51(2): 337341 (2003).
8. C.-W. Tang, Y.-C. Lin, and C.-Y. Chang, Realization of trans-
mission zeros in combine lters using an auxiliary inductively
coupled ground plane, IEEE Trans. Microwave Theory Tech.
51(10) (2003).
9. K.-L. Wu, Y.-J. Zhao, J. Wang, and M. K. K. Cheng, An
effective dynamic coarse model for optimization design of
LTCC RF circuits with aggressive space mapping, IEEE
Trans. Microwave Theory Tech. 52(1) (2004).
10. H. Uchimura, T. Takenoshita, and M. Fujii, Development of a
laminated waveguide, IEEE Trans. Microwave Theory Tech.
46(12): 24382443 (1998).
11. Y. Rong, K. A. Zaki, M. Hageman, D. Stevens, and J. Gipprich,
Low-temperature cored ceramic (LTCC) ridge waveguide
bandpass chip lters, IEEE Trans. Microwave Theory Tech.
47(12): 23172324 (1999).
12. J. Hayashi and Y. Nikawa, Improvement of characteristics
using cross patch in the LTCC lter, Proc. Int. Microwave
Symp., IEEE, 2003, pp. 531534.
13. H. Uchimura and T. Takenoshita, Wiring board Equipped
with a Line for Transmitting a High Frequency Signal, U.S.
Patent 5,982,256 (1999).
14. H. A. Huang, R. J. Duprey, and R. T. Villeages, Millimeter-
Wave LTCC Package, U.S. Patent (1999) 5,982,250.
15. Y. Huang and K.-L. Wu, A broadband LTCC integrated
transition of laminated waveguide to air-lled waveguide
for millimeter wave applications, IEEE Trans. Microwave
Theory Tech. 51(5): 16131617 (2003).

11b

Diunggah oleh

Informasi Dokumen

Judul Asli

Hak Cipta

Format Tersedia

Bagikan dokumen Ini

Bagikan atau Tanam Dokumen

Opsi Berbagi

Apakah menurut Anda dokumen ini bermanfaat?

Apakah konten ini tidak pantas?

Hak Cipta:

Format Tersedia

11b

Diunggah oleh

Hak Cipta:

Format Tersedia

20. W. Rotman and R.F.

Turner, Wide-angle microwave lens for

radiationintensity of a lossless isotropic source withsame input

real power radiated by test antennaP

total real input at antennafeedterminals P

Figure 3. Polezero diagram of the rst-order lowpass lter.

Figure 5. Polezero diagram of the SallenKey lowpass lter.

Anda mungkin juga menyukai