Kraus-38096 kra21032_ch20 May 4, 2006 12:4

20
Antenna Array
Analysis and Synthesis
The topics in this chapter include:
I Retarded potentials
I Array structures
I Weighting functions
I Actual source antennas
I Equivalent source antenna
I Linear array analysis
I Different forms of linear arrays
I Linear array synthesis
I Schelkunoff unit circle
I Sum and difference patterns
I Dolph-Chebyshev synthesis of sum
patterns
I Taylor synthesis of sum patterns
I Bayliss synthesis of difference patterns
I Planar array synthesis
201 Introduction
This chapter deals with some basic aspects of array antenna analysis and synthesis. The
approach adopted involves systematic building up of concepts ultimately leading to antenna
design. Since the matter involves many aspects only the most essential ones are briey
described. In order to understand the array analysis, rst the concepts relating to retarded
potentials are introduced, followed by introduction to weighting functions for Actual Source
Antennas and Equivalent Source Antennas. These weighting functions are then related to
array factors, which are taken to be the basis for antenna analysis and subsequently for
synthesis. Initially the analysis problem is directly addressed in terms of array factor. Later
the application of Schelkunoff Unit Circle, which is related to the array factor, has been
explored. Similarly there are many approaches to synthesize an antenna array. Other aspects
of antennas that need to be addressed, viz. sum and difference patterns, beam width and
location of the main lobe, location of nulls, number, levels and locations of side lobes,
directivity of antenna, etc. Since all the above aspects cannot be addressed in a single
introductory chapter, the matter is covered only to the extent of introducing the subject. For
detailed study reader is advised to go through the references given at the end of the chapter.
863
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
864 Chapter 20 Antenna Array Analysis and Synthesis
202 Retarded Potentials
The analysis of all antenna problems normally revolves around certain eld quantities. The
most important among them is magnetic vector potential denoted by A, which is dened
in terms of Magnetic ux density by:
B A (1)
The quantity A in itself is dened in terms of a source (current density J Amp/m
2
) by:
A =

Jdv
4R
Since, Idl = kds = Jdv, where I is the current in Amps, k is the linear current density in
Amp/m, dl, ds and dv are the elemental values of length, surface and volume respectively,
A can also be written as:
A =

Jdv
4R
=

kds
4R
=

Idl
4R
(2)
Parameter A satises steady magnetic led Laws (i.e. Biotsavart Law and Amperes
Circuital Law)
If the eld source is time varying, (2) can be written as
A =

v
[J]dv
4R
(3)
Where [J] indicates that every t appearing in the expression for J has been replaced by
a retarded time t

where,
t

= t
R
v
(4)
In view of (1) A can be related to other eld quantities (viz. H, E, D) and Poynting vector
P by using the Maxwells equations and other relations:
H = B/ (5)
H = E/t, (for free space) = 0, or J = 0) (6)
D = E (7)
|E / H| = =

(/) (8)
and
P = E H (9)
In all above equations and are the conventional media parameters, v is the velocity
of wave propagation and R is the distance from source to the point at which eld is to be
estimated
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
202 Retarded Potentials 865
A also satises the equation:

2
A = J +

2
A
t
2
(10)
If eld is time invariant (10) reduces to common form of Poissons equation:

2
A = J (11)
If J is also zero, it reduces to Laplaces equation:

2
A = 0 (12)
In view of Helmholtz Theorem, any vector eld due to a nite source is specied
uniquely if both the curl and the divergence of the eld are specied. Since the curl of A
has already been specied by (1), vector A can be specied uniquely only if its divergence
is also specied. The problem is resolved by using Lorentz Gauge condition, which relates
electric scalar potential with A by:
A =

t
(13)
A is further related to E and throught the equation
E =
A
t
(14)

2
=

t
2
(15)
Since A and are related by (13) (provided both are time variant), both or any of these
quantities may be used for the study of antenna behaviour. The magnetic vector potential
A however, nds more favour with antenna theory since it more conveniently leads to the
other eld quantities.
Normally eld problems are solved in terms of electric currents and charges and all the
above equations play their appropriate role. However, in many applications use of ctitious
magnetic currents and charges is also useful and all the above equations need to be re-written
in terms of such ctitious sources. The two sets of some of the above equations are produced
below:
1. If only Electric currents (J) and Electric charges () exists
H
e
= J +E
e
/t (16)
E
e
= H
e
/t (17)
B
e
= A (18)
E
e
= +A/t (19)
2. If only Magnetic currents (J
m
) and Magnetic charges (
m
) exists
H
m
= E
m
(20)
E
m
= H
m
J
m
(21)
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
866 Chapter 20 Antenna Array Analysis and Synthesis
LINEAR ARRAY (1D)
PLANAR ARRAY (2D)
FS ARRAY (3D)
Figure 201
D
m
= F (22)
H
m
= +F/t (23)
In above equations prex
e
and
m
refer to elds due to electric and magnetic sources.
3. It is often required to solve problems having both electric and magnetic distributions.
This is the case when both electric and magnetic sources simultaneously exist. Two of
the equations to be satised by E and H (without prexes) can be written as:
E = +A/t (1/) F (24)
H = +F/t (1/) A (25)
Similarly other equations can be accordingly modied.
In view of the assumption of sinusoidal time variation, all eld quantities can be char-
acterized by e
jt
. Further, all spatial variations are characterized by e
R
, where is
the propagation constant. The parameter normally consists two factors namely attenu-
ation constant and phase shift constant (sometimes also referred as wave number k,
k = = /v = 2/) and is written as = +j. In case of perfect dielectric = 0.
For free space is usually neglected and spatial variation is characterized only be e
jR
.
203 Array Structures
An array may be dened as a collection of elements. It may be composed of large number
of elements, which may run from few to thousands and are limited by practical constraints.
The elements in an array may be Dipoles, Polyrods, Helix, Spirals, Log Periodic Struc-
tures, Slots or Horns. Dipoles and Slots generate broad band pattern, and are widely used
in simple structures. Polyrods, Helices, Spirals and Log Periodic Structures have more
directivity. Slots and Horns are used at UHF and Microwave frequencies and are easy to
construct. The purpose of using array antennas is to achieve more directivity and to generate
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
203 Array Structures 867
( ) Cylindrical a ( ) Spherical b
Figure 202
Elements forming Curved Planes
steereable beams. Arrays may be made to generate many search beams, tracking beams or
simultaneously both.
Arrays may be linear (one-dimensional), planar (two-dimensional) or frequency sen-
sitive (FS) (three-dimensional) congurations. Figures 201 and 202 illustrate some such
congurations. In linear arrays the elements are aligned along a line. Planar arrays may
be Flat (square or rectangular), Curved (circular, semi-cylindrical/cylindrical or semi-
spherical/spherical) in shape. The elements may be arranged into different types of grids as
shown in Fig. 203. Rectangular grids, square grids and circular grids are, however, more
commonly used. Rectangular grids may be used with rectangular or circular boundaries.
Similarly circular grids may also be used with either type of boundaries. However, rect-
angular grids are more commonly used with rectangular boundaries and circular grid with
circular boundaries. These are illustrated in Fig. 204.
Rectangular planar arrays generate Fan-shaped beams. Square and Circular planar
arrays generate Pencil shaped beams. Planar arrays have lower side lobes (SL) than curved,
due to superior illumination. Planar arrays require less maintenance. Planar arrays can scan
angle (off broadside) limitedto60

70

, thus several faces are neededfor full hemispherical

coverage
In arrays the elements are normally equispaced. Non-equispaced structures are some
times also used. Element spacing governs lobe grating. There are certain advantages of
non-equispaced structures, e.g. these require less number of elements for comparable beam
width.
The major problems with arrays include mutual coupling between elements resulting
in changed radiation pattern and radiation resistance. Large coupling results in poor pattern,
raised side-lobes and mismatching with transmitting or receiving circuits.
Each element in an array acts as a source or radiator. As long as the element size is
so small that it can be regarded as a point source the problem of evaluation of distances
between different elements or an element and the eld point remains simple. The moment
size increases, the radiators will be regarded as collection of point sources, each point
source having its different geometrical location. Besides, if the shape of the radiator is also
arbitrary the problemfurther complicates. In such a situation the current distribution on these
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
868 Chapter 20 Antenna Array Analysis and Synthesis
Square grid square cell Square grid square cell Square grid rectanglar cell
Triangular grid hexagonal cell
Triangular grid square cell
Triangular grid rectanglar cell
Figure 203
Rectangular Boundary
Rectangular Grid
(commonly used)
( ) a
Rectangular Boundary
Circular Grid
Elements arranged in different forms
Circular Boundary
Rectangular Grid
Circular Boundary
Circular Grid
(commonly used)
( ) d
( ) c ( ) b
Figure 204
elements may or may not be accurately known. If the actual current distribution on radiators
is accurately known or assumed to be so, the antennas composed of such radiators can be
classied as Actual Source Antennas. This class of radiators include dipoles and helices.
Alternatively, if the assumption for current distribution used in actual source antennas does
not hold well but for which the close-in elds can be accurately described, for estimation
of eld distribution actual sources can be replaced by equivalent sources. Such antennas
are termed as Equivalent Source Antennas. Radiators for this case may include slots and
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
204 Weighting Functions 869
horns. These (equivalent) sources must be so chosen that they produce the same elds at all
points exterior to the surfaces as the actual antennas does.
Thus for the purpose of analysis, antenna arrays can be broadly classied into
(l) Actual Source Antennas and
(2) Equivalent Source Antennas. The method of analysis for these is briey described
in the following section.
204 Weighting Functions
204a Case - I: Actual Source Antennas
Consider Fig. 205, which shows an antenna array containing a number of elements num-
bered 1, 2, . . . , i, j, . . .. Each of these radiators is composed of a large number of point
sources A, B, C, . . . , P, Q. . .. These point sources conned to some nite volume V are
assumed to be oscillating harmonically with time at an angular frequency .
Also assume the origin to lie within the volume V. The volume V that contains all the
sources is dimensionally much smaller in comparison to the distance between any source-
point to the eld point. There are no sources left outside this volume or lying on any related
surface. These time varying sources with current density J(, , , t ) or charge density
(, , , t ) are assumed to exists in (or surrounded by) free space, which is characterized
by
0
and
0
. These sources are assumed to be arbitrarily located at (, , ) in rectangular
coordinate system. Figure 206 illustrates the coordinate system, location of volume V
therein, location of point source (at , , ) and the eld point (at x, y, z). The magnetic
vector potential A or electric scalar potential at arbitrary location of eld point (x, y, z)
can be written as:
A(x, y, z, t ) =

0
J(, , )
4R
e
j (t R)
dv (1)
(x, y, z, t ) =

v
(, , )
4
0
R
e
j (t R)
dv (2)
In terms of location , , and x, y, z the distance R can be given by
R = [(x )
2
+(y )
2
+(z )
2
]
1/2
(3)
B B P P
A A
Q Q
Point radiators Point radiators
Volume
with
Surface
V
S
Volume
with
Surface
V
S
Local
origin
Local
origin
Element 1 Element 2
B P
A
Q
Point radiators
Volume
with
Surface
V
S
Local
origin
Element i
B P
A
Q
Point radiators
Volume
with
Surface
V
S
Local
origin
Element j
C C C
Figure 205
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
870 Chapter 20 Antenna Array Analysis and Synthesis
Figure 206
( , , )
V
S

R
r
( , , ) x y z
Location of
point source
or radiator in
a volume
having
surface
V
S

y
z
Field point
(point at which
Field is to be
evaluated)
x
Further in view of the correspondence between cartisian and spherical coordinates:
x = r sin cos , y = r sin sin and z = r cos , and the binomial expansion:
R = r ( sin cos + sin sin + cos ) +0(r
1
)
= r , on neglecting the last term
Where, = sin cos + sin sin + cos
Thus A of (1) can now be written as
A(x, y, z) =
0
e
j (t r)
4r

v
J(, , )e
j
dv (4)
Equation (4) is written as a product of two terms. The rst one of these terms represents
an outgoing spherical wave and is given by:

e
j (t r)
4r
(5)
The second term, called directional weighting function (, ) is given by:
(, ) =

v
J(, , )e
j
ddd (6)
The average power with complex eld quantities E and H (both represented in spherical
coordinates containing r, and components, i.e. E = E
r
a
r
+ E

+ E

and H =
H
r
a
r
+H

+H

) can be obtained by using complex Poynting vector P(, ):

P(, ) =
1
2
Re[E H

] =

(4r)
2

(1/2)[

] (7)
The radiation pattern associated with E

is called the -polarized or vertically polarized

pattern. The radiation pattern associated with E

is called the -polarized or horizontally

Kraus-38096 kra21032_ch20 May 4, 2006 12:4
204 Weighting Functions 871
polarized pattern. Thus
P
r,
(, ) = (1/2)

(4r)
2

(, )]
2
(8)
P
r,
(, ) = (1/2)

(4r)
2

(, )]
2
(9)
where,

(, ) =

v
[cos cos J
x
(, , ) +cos sin J
y
(, , )
sin J
z
(, , )]e
j
ddd (10)

(, ) =

v
[sin J
x
(, , ) +cos J
y
(, , )]e
j
ddd (11)
Equations (10 and 11) form the basis of pattern analysis and synthesis of antennas. For
antenna analysis, if the actual current distribution is known, the weighting functions

and

can be determined from (10) and (11) and the power radiation patterns can be obtained
from (8) and (9). For synthesis if the desired patterns are specied, (10) and 11) can be used
to nd the current distributions.
204b Case - II: Equivalent Source Antennas
In this case electric sources are to be supplemented by ctitious magnetic sources for faithful
reproduction of external elds. If real electric sources ( and J) create electromagnetic elds
E
1
, B
1
and the ctitious magnetic sources (
m
and J
m
) create electromagnetic elds E
2
,
B
2
, all of these elds must satisfy Maxwells equations and conform to the same boundary
conditions, since the second set of sources is replacing the rst set. Away fromthese sources,
no distinction in characteristic features of elds can be made in order to determine the type
of source, which has resulted in this eld.
In case of actual source antennas the magnetic vector potential A and electric scalar
potential can be dened and related expressions can be derived. In this case we dene
electric vector potential F and magnetic scalar potential
m
. The solutions are obviously
similar. If the electric and magnetic sources are conned to reside in surfaces, then linear
current densities K Amp/m and K
m
magnetic Amp/m replaces J and J
m
. Similarly the real
charge densities
s
C/m
2
and
sm
magnetic C/m
2
replaces and
m
. The potential functions
can now be written as:
A(x, y, z, t ) =
0

s
K(, , )
4R
e
j (t R)
ds (12)
F(x, y, z, t ) =
0

s
K
m
(, , )
4R
e
j (t R)
ds (13)
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
872 Chapter 20 Antenna Array Analysis and Synthesis
(x, y, z, t ) =

s
(, , )
4
0
R
e
j (t R)
ds (14)

m
(x, y, z, t ) =

sm
(, , )
4
0
R
e
j (t R)
ds (15)
As before, the far eld form of vector potential functions A and F (12) and (13)/scalar
potential functions and
m
(14) and (15) can be written as the product of outgoing
spherical wave factor (same as (5) with the directional weighting functions (similar to (6)).
In this case weighting functions [(, ) for A and (, ) for F] for vector potential
functions are:
(, ) =

v
K(, , )e
j
ddd (16)
(, ) =

v
K
m
(, , )e
j
ddd (17)
Once vectors A and F are known the eld quantities E and H can be readily obtained
by using the dening equations of Section (202). The complex Poynting vector giving
the average power density can also be evaluated in terms of A and F, as before. Besides,
the transverse components of and can be obtained by expanding (16) and (l7). The
resulting equations are:

(, ) =

s
[cos cos K
x
(, , ) +cos sin K
y
(, , )
sin K
z
(, , )]e
j
ds (18)

(, ) =

s
[sin K
x
(, , ) +cos K
y
(, , )]e
j
ds (19)

(, ) =

s
[cos cos K
xm
(, , ) +cos sin K
ym
(, , )
sin k
zm
(, , )]e
j
ds (20)

(, ) =

s
[sin K
xm
(, , ) +cos K
ym
(, , )]e
j
ds (21)
205 Linear Array Analysis
The matter embodied in this section encompasses the most general cases of antennas em-
ployed in practical applications. In such antennas the elements (normally less than half
wavelength in their maximum dimensions) in an array are of the same type, equispacecd
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
205 Linear Array Analysis 873
and similarly oriented. The relative physical positioning of the elements and their relative
electrical excitations (in terms of current magnitudes and phases) are the parameters that
normally govern the shape of radiation pattern. The analysis of the effects on radiation
pattern caused by varying excitation reveals its ability:
1. to form a radiation pattern with main beam and side lobes.
2. to control angular placement of the main beam.
3. to select beam sharpness by choosing the length of the array.
4. to create a difference pattern.
5. to produce a shaped beam pattern without nulls.
As discussed earlier a current density distribution J(, , )e
jt
contained in a nite
volume V causes a far eld pattern (in and planes) and can be represented by weighting
functions given by (l0) and (11) and (18)(21) in the previous section. After evaluation of
various eld expressions for components of E in spherical coordinates it can be shown that
E

= E

. The weighting function

(, ) can be viewed as vertically polarized

component of the far eld and

(, ) as horizontally polarized component. It can also be

stated that if magnetic currents are introduced as secondary sources

(, ) and

(, )
will have the similar expressions if J
m
replaces J. When only that part of eld due to
magnetic sources is being considered then E

= E

. Thus

(, ) can be viewed
as vertically polarized component of the far eld and

(, ) as horizontally polarized
component. In view of the above, the weighting function concept applies equally well for
actual source and equivalent source antenna arrays.
Assume that an array contains 2N + 1 arbitrarily shaped, identical, similarly oriented,
discrete radiating elements. Since, these elements are not point sources, these may be
assumed to be composed of large number of point sources, as shown in Fig. 205. A point
P
i
(x
i
, y
i
, z
i
) in the i
th
element and a point P
j
(x
i
, y
i
, z
i
) in the j
th
element occupies the same
position vis-` a-vis the current distribution. This collection of reference points describes the
relative positions of different (N + 1) elements. Similarly another set of points Q
i
(, , )
and Q
j
(, , ) in the i
th
and j
th
elemnents can also be dened.
It is convenient to establish local coordinate system at each of these reference points.
In view of Fig. 207

i
= x
i
,
i
= y
i
,
i
= z
i
(1)
Let J
x
(, , ) = J
x
(x
i
+
i
, y
i
+
i
, +z
i
+
i
) andJ
x
(, , ) = J
x
(x
j
+
j
, y
j
+
j
+z
j
+
j
)
be the current densities in i
th
and j
th
elements, which happened to be the same as per our
initial assumption. Since all elements are identical and similarly oriented, if I
i
is the complex
total current at the terminals of the i
th
radiator and I
j
is the complex total current at the
terminals of the j
th
radiator, it leads to the following:
J
x
(x
i
+
i
, y
i
+
i
, z
i
+
i
)/J
x
(x
j
+
j
, y
j
+
j
, z
j
+
j
) = I
i
/I
j
(2)
provided
i
=
j
,
i
=
j
,
i
=
j
and J
x
belongs to corresponding physical points in
the two elements. The terminals referred to above may be the cross-section of a waveguide
feeding a slot, or a junction of coaxial cable and a helix along with its ground plane.
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
874 Chapter 20 Antenna Array Analysis and Synthesis
0
X,
( + + )
2 2 2
( + + ) x y z
2 2 2
y,
P
|( ) + ( ) + ( )| x y z
2 2 2

i
P
i

i
z,

i
( + + )
i
2 2 2
i i
( + + )
i
2 2 2
i i
( + + )
i
2 2 2
i i
i
th
element

i

i
Local origin
of elements i
th
Local origin
of elements i
th
P
i
P
j

i

i

i

i
Q
i
Q
i
i
th
elements j
th
elements
(a)
( + + ) x y z
i i i
2 2 2
(b)
Figure 207
From (2) and (22-4-11) the weighting function may be written as

(, ) =
N

V
i
[sin J
x
(x
i
+
i
, y
i
+
i
, z
i
+
i
) +cos J
y
(x
i
+
i
, y
i
+
i
, z
i
+
i
)]e
j
d
i
d
i
d
i
(3)
=
,a
(, )
,e
(, ) (4)
If the origin of principal coordinates is xed at a point P
0
(x
0
, y
0
, z
0
) in the above equation,
it will merely result in phase change in array factor and can lead to:

,a
(, ) =
N

i=0
I
i
I
0
e
j(x
i
sin cos +y
i
sin sin +z
i
cos )
(5)

,e
(, ) =

V
0
[sin J
x
(
0
,
0
,
0
) +cos J
y
(
0
,
0
,
0
)]
e
j(
0
sin cos +
0
sin sin +
0
cos )
d
0
d
0
d
0
(6)
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
205 Linear Array Analysis 875
An expression for

(, ) can also be obtained in a similar manner as

(, ) =
,a
(, )
,e
(, ) (7)
where,

,a
(, ) =
N

i=0
I
i
I
0
e
j(x
i
sin cos +y
i
sin sin +z
i
cos )
(8)

,e
(, ) =

V
0
[cos cos J
x
(
0
,
0
,
0
) +cos sin J
y
(
0
,
0
,
0
)
sin J
z
(
0
,
0
,
0
)] e
j(
0
sin cos +
0
sin sin +
0
cos )
d
0
d
0
d
0
(9)
The expressions of
,e
(, ) and
,e
(, ) involve the current distributions in only
one of the elements. The patterns thus obtainedare calledelement patterns or element factors.
These expressions are not only identical but involve relative current levels in different
elements and their relative placements. Since, subscripts and in (5) and (6) spell only
the orientation (polarization) of eld, and henceforth can be dropped. The resulting notations
retain no vector characteristics and can be named as array patterns or array factors. For
consideration of polarization one can revert to the original expressions of (5) and (6) and
evaluate

(, ) and

(, ) from (4) and (7).

205a Different Forms of Linear Arrays
Now let r
i
be the distance from P
0
(x
0
, y
0
, z
0
) to P
i
(x
i
, y
i
, z
i
) and the line connecting P
0
and P
i
have direction cosines cos
i
, cos
i
, cos
i
. If all the elements lie along a common
line, then = = for every P
i
. The antenna thus formed is termed a linear array and
its array factor can be written as:

a
(, ) =
N

n=0
I
n
I
0
e
jr
n
(cos sin cos +cos sin sin +cos cos )
(10)
Equation (10) can be taken as the starting point to study different forms of linear arrays,
and the modied versions of (10) can be accordingly obtained for the following cases.
1. If array elements are equispaced and odd (2N + 1) in number the zeroth element can
be taken in the mid of array. If now r
n
is replaced by n
d
, where d is the common
spacing between elements, (10) results in a general form of array factor expression
for a uniformly spaced linear array.

a
(, ) =
N

n=0
I
n
I
0
e
nd(cos sin cos +cos sin sin +cos cos )
(11)
If all elements lie along one of the axis, z-axis (say), i.e. at z
i
= 0, d, 2d,
Nd, cos = cos = 0, cos = 1 and the resulting pattern will be
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
876 Chapter 20 Antenna Array Analysis and Synthesis
-independent or rotationally symmetric. Equation (11) reduces to:

a
() =
N

n=0
I
n
I
0
e
jnd cos
(12)
(a) If all currents are equal and in phase the distribution results in uniformly excited
broadside array, with an array factor:

a
() =
N

n=0
e
jnd cos
(13)
Equation (13) represents a sum of phasors with common unit magnitude,
possessing phase angles which depend on Q, but for a given value of Q are
progressive multiples of the basic angle , where is given by:
= d cos = (2d/) cos (14)
Equation (14) is same as Equation (4-13-2) with replaced by and taken to be
zero. The term is referred as
z
in the subsequent text. If = 0, it requires
= (2k +1)/2, for k = 0, 1, 2, . . . and the radiation from such a conguration
of elements is perpendicular to the line in which elements are arranged, and is
maximum at = /2, 3/2, . . .. Such an array was identied as broadside array
in Case 1 of Section (413).
(b) Alternatively, if the phase change between adjacent elements is required to make
the eld a maximum in the direction of array ( = 0) the array is termed as end
re array. For this case and are zero and = (2d/). This is described in
Case 2 and 3 of Section (413).
(c) If the currents are unequal and tapered (i.e. maximum in the center and minimum
at the end elements of the array the distribution results in broadside array with
tapered excitation.
(d) If the phase angles ( or ) are progressively increasing, I
n
= I
0
e
jn
the
distribution results in a scanned beam array and the array factor for this case can
be written as:

a
() =
N

n=N
e
jn(d cos
z
)
(15)
(e) A combination of any of the above cases is also possible resulting in (say)
tapered scanned beam array.
2. If the number of elements in an equispaced linear array is even (say 2N) the array
elements can lie along z-axis at d/2, 3d/2, . . . the array factor can be written as

a
() =
N

n=N
I
N
I
1
e
j[(2N1)/2]d cos
+ +
I
1
I
1
e
j (1/2)d cos
+e
j (1/2)d cos
+
I
N
I
1
e
j[(2N1)/2]d cos
(16)
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
205 Linear Array Analysis 877
A number of different types of arrays can be derived from (16) in the similar manner
as obtained for the case of odd elements. A detailed discussion relating to different
types of arrays is included in Chapter 4.
205b Schelkunoff Unit Circle
To understand different arrays afresh consider a linear array containing N + 1 elements
(where N may be even or odd integer) lying along z-axis. The array factor of (15) can be
rewritten in the modied form as under.

a
() =
N

n=0
I
n
I
0
e
jn(d cos
z
)
(17)
Since, the ratio of currents will be a complex quantity, (17) may be further modied
by substituting:
w = e
j
(18)
where as before,
= d cos
z
(19)
Equation (19) and (4-13-2) are same except containing different symbols to represent
quantities. The array factor now transforms into:

a
(w) =
N

n=0
I
n
I
0
e
jn
=
N

n=0
I
n
I
0
w
n
=
I
N
I
0
N

n=0
I
n
I
N
w
n
(20)
On expanding the summation term and writing the expanded terms in reverse order,
the modulus of modied form of array factor becomes:
|
a
(w)| =

I
N
I
0

w
N
+
I
N1
I
N
w
N1
+ +
I
0
I
N
w
0

I
N
I
0

|f (w)| (21)
If the summation of (21) is replaced by a product term
N

n=1
(w w
n
), |f (w)| on ex-
pansion can be written as:
|f (w)| = |w w
1
| |w w
2
| |w w
3
| |w w
N
| (22)
Thus in view of the above |f (w)| cannot only be expressed as a polynomial in w of
degree one less than the number of elements in the equispaced array, it can also be used as
substitute to the array factor (as the two differ only by a multiplying factor). The roots of the
polynomial are linked to the array excitation, since current ratios comprise the coefcients
of the polynomials.
The following fundamental theorems (due to Schelkunoff) in fact lay the foundation
to the understanding of antenna arrays.
Theorem I Every linear array with the commensurable separations between the elements
can be represented by a polynomial, and every polynomial can be interpreted as a linear
array.
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
878 Chapter 20 Antenna Array Analysis and Synthesis
Since the product of two polynomials is again a polynomial, a corollary to theorem I
is:
Theorem II There exists a linear array with a space factor equal to the product of the
space factors of two linear arrays.
Theorem III The space factor of a linear array of n apparent elements is the product of
(n 1) virtual couplets with their null points at the zeroes of E.
The space factor of an array is dened as the radiation pattern of a similar array of
non-directive or isotropic elements. Thus the term space factor, array pattern or array factor
can be used to spell the same meaning.
The parameter E referred to in Theorem III is the electric eld intensity, which can
be obtained after knowing the magnetic potential A from the basic equations of Section
202. A in itself can be obtained in terms of array factor, which in turn is expressed as a
polynomial. Thus the product of (22) will result in a polynomial equation of the following
form. This equation for f (w) can represent A or E.
|f (w)| = |a
0
w
0
+a
1
w
1
+a
2
w
2
+ +a
m2
w
m2
+a
m1
w
m1
|
= |a
0
+a
1
w +a
2
w
2
+ +a
m2
w
m2
+a
m1
w
m1
|
Alternatively,
|f (w)| = |w
m
+a
1
w
m1
+a
2
w
m2
+ +1|
From (18) and (19), it is evident that w is a complex quantity and shall represent a
complex plane. Angle on the contrary represents real space. The roots of |f (w)| thus lie
in a complex w plane and are related, to the eld pattern in real space. Since varies from
0 to , w (in view of (18) and (19)) will vary from some initial value
i
= kd
z
to some
nal value
f
= kd
z
. Equation (18) if plotted, results in a circle as shown in Fig.
208(a). This gure illustrates real and imaginary axes of w, arbitrary location of and
corresponding e
j
, increasing direction of , arbitrary locations of = 0 and = , and
the corresponding positions of
i
and
f
(which in turn correspond to rst and last roots of
(22) on the periphery of a unit circle. From (22) it can be concluded that if the roots w
1
, w
2
,
w
3
, . . . w
m
(within the range of w) are placed on the unit circle then for |f (w
m
)| = 0, an
array pattern containing N nulls will result. If all the roots are placed off the unit circle or
are outside the range of w the resulting radiation pattern will be free from nulls. This circle
is known as Schelkunoff unit circle. Figure 208(b) illustrates placement of four roots
w
1
, w
2
, w
3
, and w
4
on the unit circle. These roots belong to a ve-element array problem.
If the excitation currents in amplitude and phases are same (i.e.
z
= 0) (19) reduces to
= d cos and (20) becomes:

a
(w) = f (w) =
N

n=0
w
n
=
1 w
N+1
1 w
(23)
In view of DeMovivers theorem the roots of f (w) can be evaluated from
w
m
= e
j2m/(N+1)
For m = 1, 2, . . . , N (24)
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
205 Linear Array Analysis 879
increasing
= 0
i

w e =
j
Im Im
=

f

i z
= k d

f z
= k d
Re
w
1
d
1
w
d
2
w
2
d
3 d
4
w
3
w
4
( ) a ( ) b
Re
Figure 208
Figure 209
d

( , ) x y
w
y
Im
Re
x
( , ) x y
w
Complex
- plane w
Equation (24) is illustrated by Fig. 208(b). Since number of roots have to be one less
than number of elements in an array, the four roots obtained from (24) are 2/5 = 72

and4/5 = 144

. The value of |f (w)| is obtainedbymultiplyingdistances d

1
, d
2
, d
3
and
d
4
shown in Fig. 208(b). As w moves along the unit circle these distances change and thus
their product also changes. Whenever w coincides with one of the roots the corresponding
distance d = 0 and hence |f (w
m
)| = 0, resulting in a null in the array factor.
Fig. 209 shows a circle with location of a root w

at an angle

, w is located at an
arbitrary angle . To estimate the distance d between w and w

the following procedure

can be followed.
If the real and imaginary axes of the complex w-plane lie along x and y axes, the coor-
dinates of w and w

can be given as (x, y) and (x

, y

). In fact x, y, x

and y

can themselves
be written as x = r cos = cos , y = r sin = sin , x

= r cos

= cos

and
y

= r sin

= sin

(since r = 1 for assumed unit circle). Thus d.

[cos

cos )
2
+
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
880 Chapter 20 Antenna Array Analysis and Synthesis
(sin

sin )
2
]. Thus any change in will result a change in the value of d. If there
are more roots located on the unit circle distances d
1
, d
2
. . . can be evaluated by using the
equation d
m
=

[(cos

m
cos )
2
+ (sin

m
sin )
2
] for m = 1, 2, . . .. As w moves
along the unit circle changes all the distances change and hence their product will also
change.
205c Different Forms of Linear Arrays
To obtain different forms of arrays the following steps may be followed:
1. Select the number of elements (N +1) and spacing between elements in terms of
wavelength (d = /2, 2/2, 3/2 . . .)
2. From Eqn. (w
m
= e
j2m/(N+1)
) calculate the root positions for m = 1, 2, . . . , N.
3. Place these roots w
1
, w
2
. . . on the unit circle
4. Obtain w
i
and w
f
from Eqn.(18).
5. Start moving clockwise from w
i
until w
f
is reached. On such movement w will have
many discrete locations. Measure distance from w to w
1
, w
2
. . . and list these as d
1
,
d
2
, . . .. Evaluate the product of these distances for every discrete location of w. Each
such product shall give value of |f (w)|.
6. Whenever w will coincide w
1
, w
2
, . . . one of the distances d
1
, d
2
, . . .. Will be zero,
indicating a null at that location.
7. Maximum values of product will give, maximas either of main lobe or that of side
lobes. These maximum values need not be equal.
8. Since w represents a complex plane and is related to which represents real space,
the values of corresponding to each discrete value of w can be obtained.
9. Plot the product of distances in the real space to get the radiation pattern.
Knowledge of the lobe heights (maximas) and the null positions is all that is needed to
produce a radiation pattern.
Adopting the above procedure for the problem of ve element array, if d = /2, then
from equation = d cos ,
i
= and
f
= (since varies from 0 to and
= 2/ or d = ) all the four roots are within the range of w (i.e. 0 to 2). Fig. 20
10 reveals the formation of radiation pattern and its relation to Fig. 208(b). As w moves
clockwise from e
j
to e
j
through e
j0
, the product of distances (d
1
, d
2
, etc.) rst gives a
half side lobe, followed by a null at w
2
, a side lobe, a null at w
1
, a main lobe, a null at w
4
,
a side lobe, a null at w
3
, and thereafter another half side lobe. The nulls in real space can
be determined from the relation:

m
= 2m/(N +1) = kd cos
m
If d = then
I
= 2 and
f
= 2, w ranges two full revolutions around the
unit circle. In such a case there will be extra main beams at = 0 and = . A suitable
restriction is to be imposed on
i
to avoid the extra lobes. The reader is advised to see the
references at the end of chapter for this and more involved cases.
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
206 Linear Array Synthesis 881
Figure 2010
e
j
e
j

W
2
W
1
W
4
W
3
Z
e
0 j
X Y - Plane
206 Linear Array Synthesis
In synthesis of antennas the requisite current distribution in an equispaced array is to be
obtained to satisfy a specied desired array pattern. In case of linear arrays this desired
pattern has to be a function of (or ) alone, so that the array factor can be expressed in the
form of
a
() (or
a
()). For planar arrays the desired pattern will be a function of both,
and . These functions [
a
() or
a
()] may encompass a variety of array congurations
including
(a) sum patterns with: (i) uniform side lobes, (ii) symmetrically tapered side lobes, (iii)
asymmetric side lobes
(b) difference patterns with same topographies of side lobes,
(c) patterns with neither nulls nor side lobes.
Since all these aspects cannot be covered in an introductory chapter the discussion
in this section is conned only to very essential features and techniques of synthesis. In
synthesis the termsumand difference patterns forms the basis for selection of the techniques
employed. The meaning of sum and difference patterns is explained as follows.
206a Sum and Difference Patterns:
If an antenna is illuminated by two offset feeds, two offset radiation patterns are generated.
Such an arrangement is frequently used in mono-pulse tracking radars, in which a parabolic
reector is fed by two offset horns. If tracking is to be accomplished in (say) azimuth only
two horns are to be placed side-by-side as shown in Fig. 2011a. In case the tracking is to
be accomplished in both, azimuth and elevation, four offset horns arranged in Fig. 2011b
are to be used. If in a two-horn case, the two offset beams generated Fig. 2011c are of the
same polarity they get summed, the beam shown by, Fig. 2011d will result Alternatively
if beams generated are of opposite polarity Fig. 2011e the difference of two beams will
result as in Fig. 2011f . The generation of same or opposite polarity beams will depend
on the relative polarity of feeds, i.e. the two feeds may be in phase or out of phase. The two
radiation patterns, so obtained, are termed as sum pattern and difference pattern.
These patterns nd extensive use in many practical applications. Tracking of Radars
Satellites and Space vehicles are some such applications. In Fig. 2011 signals to be summed
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
882 Chapter 20 Antenna Array Analysis and Synthesis
A B
B A
D C
+ + + + +
+
(a)
(b)
(c) (d) (e) (f)
Sum Pattern
Sum = +
Case (a)
or
+ + +
Case (b)
A B
A B C D
Difference Pattern
for (a)
or
( + ) ( + ) for elevation Case (b)
or
Difference
( + C) ( + ) for azimuth Case (b)
A B
A B C D
A B D
Feed Horn
Arrangement
Figure 2011
or differenced in case of azimuth or elevation are also mentioned. The creation of sum or
difference pattern is not conned to reectors or lenses; antenna arrays are equally suitable
for such ventures. Thus synthesis requires understanding of the meanings of these patterns,
since generation of these patterns is related to the number of elements in an array. If this
number is odd, an array is not suitable to generate a difference pattern. On the other hand, an
array with even number of elements is equally suitable for creation of sumand of difference
patterns. Further, the description given above is not conned to transmitting antennas; the
application of sumand difference pattern is equally valid for receiving antennas. In this case
the incoming signal are added or subtracted to get the sum or difference patterns. Normally
in monopulse tracking, sum pattern is used for transmission. In case of reception, it is used
to obtain range information whereas one or two difference patterns are used to track the
target in azimuth or elevation or in both.
To produce sum and difference patterns with main beam of sum pattern pointing at
an angle
0
, and the twin main beams of the difference pattern straddling
0
, both patterns
exhibiting a symmetrical side lobe structure with 2N equispaced elements in an array, the
array factor for separation d can be written as:

a
() =
1

n=N
I
n
I
1
e
j[(2n1)/2]d(cos cos
0
)
+
N

n=1
I
n
I
1
e
j[(2n1)/2]d(cos cos
0
)
(1)
If all current amplitudes are real, and I
n
= I
n
, Eqn (1) for sum pattern becomes:
S() = 2
N

n=1
I
n
I
1
cos[(2n 1)(d/)(cos cos
0
)] (2)
Similarly for difference pattern I
n
= I
n
, and Eqn (1) becomes:
D() = 2j
N

n=1
I
n
I
1
sin[(2n 1)(d/)(cos cos
0
)] (3)
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
206 Linear Array Synthesis 883
The sum pattern from an array with 2N + 1 (odd number) elements can be given by
modifying Eqn. (20511) as:
s() = 1 +2
N

n=1
I
n
I
1
cos[2n(d/)(cos cos
0
)] (4)
Amajor class of synthesis problems can be circumvented by (1)(4). To obtain specied
side lobe topography, the required current distribution, I
n
/I
1
or I
n
/I
0
for sum or difference
patterns can be obtained from these equations.
If the Schelkunoff unit circle is correlated to the design of equispaced linear arrays,
excited to give an array factor with main beam accompanied by side lobes the following
can be concluded.
1. For 2N +1 elements, if 2N roots (in complex conjugate pairs) are placed on the unit
circle, a symmetrical sum pattern results. The adjustment of the position of these root
pairs can alter the side lobe heights. Further by clustering the root pairs closer to
= can also reduce the levels of side lobes. This option will however, broaden the
main beam.
2. For 2N elements, if 2N 1 roots are placed on the unit circle with one root at
= and the remaining inN 1 complex conjugate pairs, a symmetrical sum
pattern will result. The side lobe heights again can be altered by adjusting root pair
positions. As before the level of side lobes can also be reduced by clustering the root
pairs closer to = but at the expense of main lobe widening.
3. Occurrence of roots in complex conjugate pairs, results in pure real coefcients of
polynomial f (w). These coefcients appear in symmetrical pairs and hence the
current distribution in an array is symmetrical in amplitude.
206b Dolph-Chebyshev Synthesis of Sum Pattern
Dolph addressed the problem of proper root positioning or proper excitation to give a sum
pattern with uniformside lobes of specied levels by exploring the properties of Chebyshev
polynomials. These polynomials denoted by T
m
(u), with index m and a variable argument
u, are the solutions of the differential equation:
i u
2
d
2
T
m
du
2
u
dT
m
du
+m
2
T
m
= 0 (5)
If m(= 2N) is an even integer:
T
2N
(u) =
N

n=0
(1)
Nn
N
N +n
(
N+n
2N
)(2u)
2n
(6)
If m(= 2N 1) is an odd integer:
T
2N1
(u) =
N

n=1
(1)
Nn
2N 1
2(N +n 1)
(
N+n1
2n1
)(2u)
2n1
(7)
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
884 Chapter 20 Antenna Array Analysis and Synthesis
Figure 2012
m = 6
m = 7
+1 u
( , ) u b
0
( , ) u b
0
T u
m
( )
T u
m
( )
+1
1
0
1
+1 u
0
1
+1
The termof the form

r
s

stands for binomial coefcient

r!
s!(r s)!
. Equation (6) and
(7) can also be written as
T
m
(u) = cos(mcos
1
u) for 1 u 1
= cosh(mcosh
1
u) for u 1 (8)
The plots of T
m
(u) are shown in Fig. 2012 for even and odd indices. These gures illustrate:
1. T
m
(u) oscillates cosinusoidally (for even m) and sinusoidally (for odd m). Also for
|u| < 1, T
m
(u) exhibits symmetry and anti-symmetry about u = 0 for respective
values of even and odd m.
2. For |u| > 1, T
m
(u) rises hyperbolically, in continuity, in the respective directions.
Larger the value of m steeper is the slope.
The task involved nowis to make a correspondence between variable u to the real angle
variable . Such correspondence will correlate T
m
(u) to
a
(). This was developed by
considering basic equation for array factor, with the assumption of a uniform progressive
phase and symmetrical amplitude distribution. As a result (4) and (2) for odd and even
number of elements can be rewritten in the form as given below:
S

= 1 +2
N

n=1
I
n
I
0
cos 2n

2
(9)
S

= 2
N

n=1
I
n
I
1
cos(2n 1)

2
(10)
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
206 Linear Array Synthesis 885
where,
= d(cos cos
0
) (11)
In view of the transformation
u = u
0
cos(/2) (12)
As ranges from 0 to
0
to , changes from
1
= d(1 cos
0
) to 0 to
f
=
d(1 +cos
0
), u will assume values fromu
i
= u
0
cos(
i
/2) to u
0
to u
f
= u
0
cos(
f
/2).
If u
0
is chosen to be T
m
(u
0
) = b with 20 log
10
b as the assumed side lobe level, then
a pattern consisting of a main beam at the relative lobe level b and a family of side lobes,
all at the same (unit) height, will result. The implementation of the Dolph design procedure
for a simple case of uniform array can thus be summarized as below.
1. Select the number of elements by which an array is to be implemented. This gives the
degree of Chebyshev polynomial m, which is one less than the number of elements
2. Select the desired side lobe level (SLL) in dBs (i.e. SLL = 20 log
10
b).
3. Find u
0
from equation T
m
(u
0
) = b.
cosh(mcosh
1
u
0
) = b . . . for|u| 1
or u
0
= cos((cos h
1
b)/m)
4. Determine the roots of T
m
(u) by using relation
T
m
(u) = cos(mcos
1
u) . . . For 1 u 1
Get the roots from cos(mcos
1
u) = 0 as u
p
= cos[{cos
1
[(2p 1)/2]}/m]
5. Evaluate u
p
for p = 1, 2, 3, . . .
6. From (12), obtain =
p
for different values of u
p
7. If
p
is known, values of w
p
(from w = e
j
) can be evaluated. This will lead to the
known value of f (w). From equation = d cos
z
,
p
can be evaluated
8. Since |f (w)| = |w w
1
||w w
2
||w w
3
| . . . the product will result in a
polynomial equation:
f (w) = w
m
+a
1
w
m1
+a
2
w
m2
+. . . +1
The coefcients a
1
, a
2
, . . . will represent the relative magnitude of currents in different
elements of array. These coefcients will be symmetrical about a central element, i.e. the
current in rst and last, second and last but one and so on will be equal.
It is possible to determine the Dolph current distribution for the general case without
going through the intermediate steps of nding u
p
,
p
and w
p
and the multiplication to get
f (w). The relative current distribution for an array of known side lobe level, after evaluating
u
0
can be calculated as below:
1. For any array with 2N +1 elements:
I
n
=
N

p=n
(1)
Np
N
N +p
N+p
C
2p
2p
C
pn
u
2p
0
(13)
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
886 Chapter 20 Antenna Array Analysis and Synthesis
2. For any array with 2N elements:
I
n
=
N

p=n
(1)
Np
2N 1
2(N +p 1)
N+p1
C
2p1
2p1
C
pn
u
2p1
0
(14)
These relations are obtained by inserting (12) in (6) and (7) and using the relations
cos
2n

2
=
1
2
2n
n

q=0

2n
nq

cos
2q
2
(15)
cos
2n1

2
=
1
2
2n2
n

q=1

2n1
nq

cos(2q 1)

2
(16)
where
q
= 1 if q = 0, otherwise
q
= 2
206c Taylor Synthesis of Sum Patterns
When a current distribution is to be obtained for producing a radiation pattern with a single
main beam of a prescribed width and specic scan position, along with a family of side
lobes having a common specied level, the problem is similar to that of a discrete linear
array. The synthesis procedure must permit determination of the line source distribution
corresponding a specic pattern. Dolph method leads to the solution of such a problemwith
discrete sources. The method is also applicable for the cases where a source conguration
can be viewed as a continuous line source. A horn of aperture size a b, with a >> and
b << belongs to this category.
The Taylor distributions can be sampled and applied to the design of discrete array.
This alternative can also be extended to planar apertures. To this extent it is a competitor
to the Dolph method. However, Taylor method offers an added advantage, since this can be
extended to deal with circular planar apertures. Such an extension is not possible in case of
Dolph.
For a continuous line source of small cross-section s stretching along (say) z-axis from
a to +a (20410) and (20411) can take the form

() =
a

a
s[cos cos J
x
( ) +cos sin J
y
( ) sin J
z
( )]e
j
d (17)

() =
a

a
s[sin J
x
( ) +cos J
y
( )]e
j
d (18)
where = cos . If the direction of current density is assumed to be the same in every
aperture element d , that is:
J( ) = (C
1
a
x
+C
2
a
y
+C
3
a
z
)g( ) (19)
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
206 Linear Array Synthesis 887
with C
1
, C
2
, and C
3
as constants and a
x
, a
y
, and a
z
are respective unit vectors.

() = (C
1
cos cos +C
2
cos sin C
3
sin )s
a

a
g( )e
jcos
d (20)

() = (C
1
sin +C
2
cos )s
a

a
g( )e
jcos
d (21)
The multiplying factors before integrals are called the element factors for

and

.
The integrals (fortunately same) give the array factors for the line source. The general array
factor can be written as
S() =
a

a
g( )e
jcos
d (22)
Thus for a given desired sum pattern the aperture distribution g( ) is to be obtained.
There can be many approaches to obtain the array factors g( ). It can be an assumed
value, e.g. g( ) = Ge
jb
with Gand b having constant values. Its substitution in (22) gives
the relation for sum pattern for a uniformly excited discrete array for L = 2a, provided the
limit d/ 0
S() = 2Ga[{sin a(cos b/)}/{a(cos b/)}] (23)
The substitution u = (2a/) (cos b/) in (23) leads to a universal power pattern
shown by Fig. 209, which is obtained by implementing the equation:
f (u) = 20 log
10

S(
0
)
S()

= 20 log
10

sin u
u

(24)
in which
0
= cos
1
b/ is the poynting angle of main beam.
A plot of f (u) results in a sum pattern with symmetrical side lobes where the eld
heights are inversely proportional to u. The closest pair of side lobes is down by 13.5 dB.
Since changes in real space from0 to
0
to , the value of u changes from(2a/)(1b/)
to 0 to (2a/) (1 + b/). The number of side lobes will depend on the aperture length
(2a/).
An approximate universal pattern can be constructed by selecting an integer n

such
that |u| n

. The nulls of new pattern have to occur at an integral value of u. In order to

depress the intervening side lobes a little the next pair of nulls in the vicinity of main beam
have to occur at u = u
n1
, where |u
n1
| > n

1. Also the penultimate (last but one) pair

of nulls needs to be shifted to u = u
n2
, where |u
n2
| > n

2 and so on. Such a pattern

can be expressed by the function:
s(u) =
sin u
u
n1

n=1
(1 u
2
/u
2
n
)
n1

n=1
(1 u
2
/n
2
)
(25)
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
888 Chapter 20 Antenna Array Analysis and Synthesis
0
10
20
30
40
50
dB
15 10 5 0 5 10 15
2a

u =
(cos cos )
0
sin u
u
Figure 2013
Thus the innermost n

1 pairs of nulls are removed from the original sin u/u

pattern and replaced by new pairs at modied positions u
n
. Taylor obtained the following
relation for determining the new null positions:
u
n
= n

[
A
2
+(n 1/2)
2
A
2
+(n

1/2)
2
]
1/2
(26)
From side lobe level = 20 log
10
b and b = cos hA gives value of A.
The aperture distribution g( ) given by (22) to satisfy pattern given by (25) and (26)
let g( ) = h(, )e
p
, where h( ) can be represented by:
h( ) =

m=0
B
m
cos(m/a) (27)
Equation (22) becomes
S(u) =

m=0
B
m
a

a
cos(m/a) e
ju/a
d (28)
Discarding of odd integrand results
S(u) =

m=0
B
m
a

a
cos(m/a) cos(u/a)d (29)
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
206 Linear Array Synthesis 889
Thus integral is zero unless u is an integer and equals m for such limiting case,
S(0) = 2aB
0
and S(m) = aB
m
, for m = 1, 2, . . . (30)
From (25) S(m) = 0 for m n

, thus this Fourier series truncates. The continuous

aperture distribution is obtained to be:
g( ) =
e
jp
2a
[S(0) +2
n1

m=1
S(m) cos(m/a)d] (31)
Modied Taylor patterns If equal importance is attached to all the directions Taylor
patterns may not be optimum in some applications. It is often desirable to shape the pat-
terns so as to allow high side lobe level in unimportant directions and low side lobe level
in regions adjacent to the main beam or in the desired direction. For such cases Taylor
distribution obviously needs modication, which can be achieved by using perturbation
procedure resulting in new form of equations.
206d Bayliss Synthesis of Difference Patterns
The Taylor line source distribution g( ) given by (31) can be seen as a product of a uniform
progressive phase term e
jp
and a pure real amplitude distribution function:
h( ) =
1
2a
[S(0) +2
n1

m=1
S(m) cos(m/a)] (32)
It is the rst term, which determines direction of the main beam. The expression for
h( ) contains all even Fourier series terms that are non zero at the end points. In the case
of uniform distribution h( ) is constant and corresponds to only rst term of this Fourier
series. It results in the pattern of Fig. 2013 and can be modied by root location. If a line
source distribution is obtained, which can be represented by a Fourier series containing all
odd terms that are non zero at the end points, (32) can be written as:
h( ) =

m=0
B
m
sin[(m+1/2)(/a)] (33)
If only the zeroth term of this Fourier series are taken a generic difference pattern
should result from the aperture distribution:
g( ) = sin

2a

e
jp
(34)
When (34) is used in the array factor common to (20) and (21), it leads to
D() =
a

a
g( )e
j
cos d
= j
a

a
[cos{(u 1/2)(/a)} cos{(u +1/2)(/a)}]d (35)
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
890 Chapter 20 Antenna Array Analysis and Synthesis
10
0
20
30
40
50
7.5 5 2.5 0 2.5 5 7.5
2a

u =
(cos cos )
0
D u ( ) =
u
u u
cos
( 1/2) ( + 1/2)
u
dB
Figure 2014
where u is given by (26). The integration and simplication of (26) ultimately leads to:
D(u) =
u cos u
(u 1/2)(u +1/2)
(36)
The plot of difference pattern obtained from (36) is shown in Fig. 2014. The nulls in
view of (36) will occur at u = 0 and cos u or when:
u = (n +1/2), for n = 1, 2, . . . (37)
If the lobe pairs in the vicinity of main lobes are to be depressed (36) is to be modied:
D(u) = u cos u
n1

n=1
(1 u
2
/u
2
n
)
n1

n=0
[1 {u/(n +1/2)}
2
]
(38)
The synthesis problem thus boils down to nding new null locations u
n
. In view of
a parametric study with the aid of computer Bayliss obtained relations given by Equation
(39) for root replacement:
u
n
= 0 for n = 0 (a)
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
207 Planar Arrays 891
u
n
= (n

+1/2)

2
n
A
2
+n
2

1/2
for n = 1, 2, 3, 4 (b)
u
n
= (n

+1/2)

A
2
+n
2
A
2
+n
2

1/2...
for n = 5, 6 (c) (39)
where parameter Aand
2
n
are related to side lobe level (SLL) and their appropriated values
are given in Table 201.
Table 201
(SLL) 15 20 25 30 35 40
A 1.0079 1.2247 1.4355 1.6413 1.8431 2.0415

1
1.5124 1.6962 1.8826 2.0708 2.2602 2.4504

2
2.2561 2.3698 2.4943 2.6275 2.7675 2.9123

3
3.1693 3.2473 3.3351 3.4314 3.5352 3.6452

4
4.1264 4.1854 4.2527 4.3276 4.4093 4.4973
Bayliss difference pattern can be obtained on multiplying (33) by e
jp
, inserting the
result in (35) and using (39). The resulting equation is:
D(u) = 2j

m
B
m
a

0
sin[(m+1/2)(/a)] sin(u/a)d (40)
If u is midway between two integers (say n + 1/2) the integral is zero unless n = m,
for which: D(m+1/2) = jaBm. Further D(m+1/2) = 0 for m n

, the Fourier series

truncates. The aperture distribution to produce Bayliss pattern is nally obtained to be:
g( ) = e
jp
n1

m=0
D(m+1/2) sin[(m+1/2)/2] (41)
207 Planar Arrays
In order to understand a planar array it is necessary to understand its structure vis-` a-vis the
nature of plane, distribution and spacing of elements. The planar array may be of square,
rectangular or circular shape. The elements used to shape a plane may be in the form of
a square grid, a rectangular grid or a circular grid. Rectangular grids are normally used
with rectangular boundaries and circular grids with circular boundaries. In these grids the
elements may be equispaced or non-equispaced. The spacings along the two axes of a plane
may be equal or unequal. The plane of an array may be divided into symmetrical quadrants
for an appropriate mode of excitation to generate sum and/or difference patterns.
If the aperture distribution in a planar array is such that its resulting pattern is the product
of patterns of two orthogonal linear arrays, such a distribution is termed as a separable
aperture distribution. If such a separation is not realizable the distribution may be termed
as inseparable aperture distribution. The synthesis in separable case is simply the extension
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
892 Chapter 20 Antenna Array Analysis and Synthesis
Figure 2015

d
y
d
x
x
y

r
P r ( , , )
z
of the techniques developed for linear arrays. In non-separable cases newtechniques need to
be developed to tackle the problems. In the following text only the separable distributions are
being discussed in brief. The issue of non-separable distributions and detailed description
of separable cases are thoroughly discussed by Elliot [1]. The two array congurations,
which have been discussed include:
1. Arrays with Rectangular Boundary
2. Arrays with Circular Boundary
207a Arrays with Rectangular Boundary
Figure 2015 illustrates a rectangular planar array in which the elements are arranged in
the form of a rectangular grid. The number of elements along x and y-axes are assumed
to be 2N
x
+ 1 and 2N
y
+ 1 and their separations as dx and dy respectively. An mn
th
,
element in an array is located at
m
= mdx and
n
= n dy where N
x
m N
x
and
N
y
n N
y
. The current of this mn
th
element can be designated by I
mn
. The array
factor given by (2058) can be modied to:

a
(, ) =
N
x

m=N
x
N
y

n=N
y
I
mn
I
00
e
j sin (mdx cos +ndy sin )
(1)
Equation (1) is equally valid if the representative current is magnetic, for which

a
(, ) will be replaced by
a
(, ).
If each row has the same current, even though the current levels in different rows are
different, i.e. if I
mn
/I
m0
= I
0n
/I
00
, the current distribution is said to be separable and (1)
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
207 Planar Arrays 893
can be written as:

a
(, ) =
x
(, )
y
(, ) (2)
where,

x
(, ) =
N
x

m=N
x
I
m
e
jm sin cos dx
(3)

y
(, ) =
N
y

n=N
y
I
n
e
jm sin sin dy
(4)
and
I
m
= I
m0
/I
00
, I
n
= I
0n
/I
00
(5)
are the normalized current distributions in a row of elements parallel to x axis and y axis
respectively.
Equation (2) indicates the principle of multiplication of patterns of two linear arrays
lying along x-axis (3) and y-axis (4). If I
mn
and I
00
differ in phase by the factor e
j (m
x
+n
y
)

a
(, ) =
N
x

N
x
I
m
e
jm(d
x
sin cos
x
)

N
y

N
y
I
n
e
jn(d
y
sin cos d
y
)
(6)
The amplitude distribution I
m
and I
n
are pure real and have uniform phase progression

x
and
y
in x and y directions. lf the distributions are also symmetric, the factors
x
and

y
represent patterns consisting conical main beams with side lobes, which are rotationally
symmetric about x and y axes respectively. The main beam of
x
makes an angle
x
with
positive x axis and satises the relation:
dx cos
x

x
= dx sin cos
x
= 0
cos
x
=

x
dx
sin cos (7)
Similarly for
y
:
cos
y
=

y
dy
sin cos (8)
Equation (2) spells that planar array factor is the product of two linear array factors. It
further reveals that the two conical main beams and associated side lobes of each conical
pattern intersect to forma planar radiation pattern of a planar antenna. Acriterion needs to be
developed to ensure an intersection. For given spacings dx and dy and given inter-element
phase shift
x
and
y
the equations, which provide unique pointing direction (
0
,
0
) in
z>0 plane are:
tan
0
= (
y
dx)/(
x
dy) (9)
sin
2

0
= (
x
/dx)
2
+(
y
/dy)
2
(10)
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
894 Chapter 20 Antenna Array Analysis and Synthesis
Figure 2016
a

y
d

r
P r ( , , ) z
x
Equation (10) indicates the situation of no intersection. If sin
2

0
= 1, (10) reduces to
an elliptic relation:
(
x
/dx)
2
+(
y
/dy)
2
= 1 (11)
which limits the ranges of a
x
(or
y
) for specied values of dx and dy.
Since, the desired pattern has to be a function of both, and , the problemboils down to
nding the proper current distribution across a nite width aperture so as to produce a desired
radiation pattern. There are several techniques of synthesis like Fourier Integral method,
Woodward technique that synthesizes pattern-requiring null lling, Dolph technique using
Chebyshev polynomials to deduce discrete current distributions that yield sumpatterns with
uniform side lobes and the Taylor procedure, which accomplishes basically the same goal
but for continuous line sources and modies the Dolph procedure.
It is alsopossible todetermine the Dolphcurrent distributionfor general (separable) case
without going through the intermediate steps as was done in Section (206). In equations
(20613) and (14), current I
n
were estimated since it was a one-dimensional problem.
In planar array (involving two dimensions) I
m
can also be similarly evaluated by making
suitable changes in (20613) and (20614).
Arrays with Circular Boundary: If a planar aperture with circular boundary of radius
a, contains a linear (unidirectional) current density distribution (20418) and (l9) or 20
42) and (22) can be written as:

(, ) = cos sin

s
k
y
(, )e
j
dd (12)

(, ) = cos

s
k
y
(, )e
j
dd (13)
where,
= sin cos + sin sin (14)
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
207 Planar Arrays 895
The integral in (12) and (13) can be taken to be the array factors for linearly polarized planar
aperture distribution. In view of Fig. (2016),
= cos , = sin (15)
If k
y
(, ) is designated as k(, ), the common integral becomes:
F(, ) =
a

0
2

0
k(, )e
j(sin cos()
dd (16)
The aperture distribution can be represented by Fourier Series
k(, ) =

n=
k
n
()e
jn
(17)
Also in view of Bessel expansion
e
j
sin cos( ) =

m=
j
m
J
m
(k sin )e
jm()
(18)
Thus (16) becomes
F(, ) =

m=

n=
2

0
a

0
k
n
()(j)
m
J
m
(k sin )e
jm
e
j (nm)
dd (19)
The integrals of (19) has a non zero value only when m = n, which leads to:
F(, ) = 2

n=
j
n
e
jn
a

0
k
n
()J
n
(k sin )d (20)
For getting a independent pattern n should be restricted to zero value. This restriction
leads to:
F() = 2
a

0
k
0
()J
0
(k sin )d (21)
On substituting u = 2a sin /, p = /a, g
0
(p) = 2a
2
k
0
()/, F() transforms to
F(u) =

0
pg
0
()J
0
(up)d (22)
For a uniformly excited circular aperture, g
0
(p) = 1 is the representative condition
and (22) results in a sum pattern given by:
S(u) = J
1
(u)/u (23)
The pattern obtained from Eqn. (23) is illustrated by (Fig. 2017) which contains main
beam plus a family of side lobes that decay in height as the side lobe position becomes
more remote from the main beam. Since this pattern is rotationally symmetric the main
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
896 Chapter 20 Antenna Array Analysis and Synthesis
l
Figure 2017
lobe is a pencil beam, surrounded by ring-side-lobes, the number of which in the visible
region depends on the aperture size. Since u = 2a sin /, the range of u corresponds to
0 < u < 2a/ of visible space.
The main problem now is to get the function that modies the side lobe structure so
that near in lobes are at controlled height. Taylor tackled the problem by moving innermost
side lobes to achieve the desired side lobe level for interleaving side lobes.
If the roots of J
1
(u) are dened by
J
1
(
ln
) = 0, n = 1, 2, . . . (24)
Then a modication of (24) can be written as
S(u) = [J
1
(u)/(u)]

(1 u
2
/u
2
n
)/(1 u
2
/
2
ln
) (25)
Equation (25) accomplishes the purpose of removing the rst n
1
1 root pairs of (24)
and replacing them by n 1 root pairs of new positions u
n
. Taylor obtained the new root
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
207 Planar Arrays 897
Figure 2018
position at
U
2
n
=
2
ln
{A
2
+(n 0.5)
2
}/{A
2
+(n
1
0.5)
2
} (26)
where, 20 log
10
(cos hA) is the desired side lobe level. A typical circular Taylor pattern
is shown in Fig. 2018, with n = 6 and desired side lobe level of 15 dB. The near in side
lobes are seen to be somewhat drooping.
To get the aperature distribution g
0
(p) to produce this type of pattern g
0
(p) can be
expressed in a series form
g
0
(p) =

B
m
J
0
(
1m
) (27)
Thus (23) becomes
S(u) =

B
m

J
0
(
1m
)J
0
(u)d (28)
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
898 Chapter 20 Antenna Array Analysis and Synthesis
0 n/4 n/2 3 /4 n n
0
1.0
Figure 2019
In (28) each m
th
term will contribute toward S(u). The contribution of k
th
term S(
1k
)
towards sum can be written as
S(
1k
) = B
k

J
2
0
(
1k
)d (29)
From which
B
k
= (2/
2
)S(
1k
)/J
0
(
1k
) (30)
Since, S(
1k
) = 0 for k > n
1
, the series truncates and the aperature distribution is given
by:
g
0
(p) = (2/
2
)

S(
1k
)J
0
(
1m
)/J
2
0
(
1k
) (31)
where, S(
1k
) can be computed from (27)
The aperture distribution corresponding to the pattern of Fig. 2018 is shown in Fig. 20
19.
Discretization of a circular planar aperture by employing a grid system comprising
concentric circles is a more natural arrangement, though it may not necessarily lead to a
practical system from feeding aspect. The radii of the family of concentric circles may be
given by

m
= (2m1)d/2 (32)
Kraus-38096 kra21032_ch20 May 4, 2006 12:4
References 899
where d is the inter radial spacing. If d is also the spacing between edges and elements
along any circle, then
2
m
= (2m1)d = N
m
d (33)
in which N
m
is the number of elements on the m
th
circle. The values of N
m
satisfying (33)
need not be integers. In such cases these can be rounded to the nearest integer. The sum
pattern can be obtained to be
S(, ) = 4

I
mn
cos(k
mn
sin cos ) cos(k
nm
sin sin ) (34)
References
Elliot, R. S: Antenna Theory and Design, Prentice Hall, New Delhi, 1985.
Jordon, E. C. and Balmain, K. G.: Electromagnetic Waves and Radiating Systems, Prentice Hall Ltd, New Delhi,
1987.
Jesic, H.:. Antenna Engineering Hand book, McGraw-Hill Book Co., 1981.
Weeks, W. L. Antenna Engineering", Tata McGraw-Hill, 1974.