Nikolova 2012

1
LECTURE 15: LINEAR ARRAYS PART III
(N-element linear arrays with uniform spacing and non-uniform amplitude:
Binomial array; DolphTschebyscheff array. Directivity and design.)

1. Advantages of linear array with nonuniform amplitude
The most often met BSAs, classified according to the type of their excitation
amplitude, are:
a) the uniform BSA relatively high directivity, but the side-lobe levels are
high;
b) DolphTschebyscheff (Chebyshev, ) BSA for a given number
of elements maximum directivity is next after that of the uniform BSA;
side-lobe levels are the lowest in comparison with the other two types of
arrays for a given directivity;
c) binomial BSA does not have good directivity but has very low side-lobe
levels (when / 2 d = , there are no side lobes at all).

2. Array factor (AF) of a linear array with nonuniform amplitude
distribution
Let us consider a linear array with an even number (2M) of elements,
located symmetrically along the z-axis, with excitation, which is also
symmetrical with respect to 0 z = . For a broadside array ( 0) | = ,

1 3 2 1
cos cos cos
2 2 2
1 2
1 3 2 1
cos cos cos
2 2 2
1 2
...
... ,
M
j kd j kd j kd
e
M
M
j kd j kd j kd
M
AF a e a e a e
a e a e a e
u u u
u u u

= + + + +
+ + + +
(15.1)

1
2 1
2 cos cos
2
M
e
n
n
n
AF a kd u
=
(
| |
=
|
(
\ .

. (15.2)
If the linear array consists of an odd number (2M+1) of elements, located
symmetrically along the z-axis, the array factor is

cos 2 cos cos
1 2 3 1
cos 2 cos cos
2 3 1
2 ...
... ,
o jkd j kd jMkd
M
jkd j kd jMkd
M
AF a a e a e a e
a e a e a e
u u u
u u u
+

+
= + + + + +
+ + + +
(15.3)

( )
1
1
2 cos 1 cos
M
o
n
n
AF a n kd u
+
=
= (

. (15.4)
Nikolova 2012
2
EVEN- AND ODD-NUMBER ARRAYS



Fig. 6.17, p. 291, Balanis
Nikolova 2012
3
The normalized AF derived from (15.2) and (15.4) can be written in the form

| |
1
cos (2 1) ,
M
e
n
n
AF a n u
=
=

for 2 N M = , (15.5)

| |
1
1
cos 2( 1) ,
M
o
n
n
AF a n u
+
=
=

for 2 1 N M = + , (15.6)
where
1
cos cos
2
d
u kd
t
u u

= = .

Examples of AFs of arrays of nonuniform amplitude distribution

a) uniform amplitude distribution (N = 5, / 2 d = , max. at
0
90 u = )

p. 148-149, Stutzman
Nikolova 2012
4
b) triangular (1:2:3:2:1) amplitude distribution (N = 5, / 2 d = , max. at
0
90 u = )



p. 148-149, Stutzman










Nikolova 2012
5
c) binomial (1:4:6:4:1) amplitude distribution (N = 5, / 2 d = , max. at
0
90 u = )



p. 148-149, Stutzman

Nikolova 2012
6
d) Dolph-Tschebyschev (1:1.61:1.94:1.61:1) amplitude distribution (N = 5,
/ 2 d = , max. at
0
90 u = )




p. 148-149, Stutzman








Nikolova 2012
7
e) Dolph-Tschebyschev (1:2.41:3.14:2.41:1) amplitude distribution (N = 5,
/ 2 d = , max. at
0
90 u = )


p. 148-149, Stutzman


Notice that as the current amplitude is tapered more toward the edges of the
array, the side lobes tend to decrease and the beamwidth tends to increase.





Nikolova 2012
8
3. Binomial broadside array
The binomial BSA was investigated and proposed by J. S. Stone (US
Patents #1,643,323, #1,715,433) to synthesize patterns without side lobes. First,
consider a 2element array (along the z-axis).
z
y
x
d

The elements of the array are identical and their excitations are the same. The
array factor is of the form
1 AF Z = + , where
( ) cos j kd j
Z e e
u | +
= = . (15.7)
If the spacing is / 2 d s and 0 | = (broad-side maximum), the array pattern
|AF| has no side lobes at all. This is proven as follows.

2 2 2 2
| | (1 cos ) sin 2(1 cos ) 4cos ( / 2) AF = + + = + = (15.8)
where cos kd u = . The first null of the array factor is obtained from (15.8) as

1,2 1,2
1 2
cos arccos
2 2 2
n n
d
d
t t
u u

| |
= =
|
\ .
. (15.9)
As long as / 2 d < , the first null does not exist. If / 2 d = , then
1,2
0,
n
u =
180. Thus, in the visible range of , all secondary lobes are eliminated.
Second, consider a 2element array whose elements are identical and the
same as the array given above. The distance between the two arrays is again d.
d
z
y
x

Nikolova 2012
9
This new array has an AF of the form

2
(1 )(1 ) 1 2 AF Z Z Z Z = + + = + + . (15.10)
Since (1 ) Z + has no side lobes,
2
(1 ) Z + does not have side lobes either.
Continuing the process for an N-element array produces

1
(1 )
N
AF Z

= + . (15.11)
If / 2 d s , the above AF does not have side lobes regardless of the number of
elements N. The excitation amplitude distribution can be obtained easily by the
expansion of the binome in (15.11). Making use of Pascals triangle,

1
1 1
1 2 1
1 3 3 1
1 4 6 4 1
1 5 10 10 5 1
..............................


the relative excitation amplitudes at each element of an (N+1)-element array
can be determined. An array with a binomial distribution of the excitation
amplitudes is called a binomial array. The excitation distribution as given by
the binomial expansion gives the relative values of the amplitudes. It is
immediately seen that there is too wide variation of the amplitude, which is a
disadvantage of the BAs. The overall efficiency of such an antenna would be
low. Besides, the BA has relatively wide beam. Its HPBW is the largest as
compared to the uniform BSA or the DolphChebyshev array.
An approximate closed-form expression for the HPBW of a BA with
/ 2 d = is

1.06 1.06 1.75
1 2
HPBW
N L L
= = =

, (15.12)
where ( 1) L N d = is the arrays length. The AFs of 10-element broadside
binomial arrays (N = 10) are given below.
Nikolova 2012
10

Fig. 6.18, p.293, Balanis

Nikolova 2012
11
The directivity of a broadside BA with spacing / 2 d = can be calculated
as

0
2( 1)
0
2
cos cos
2
N
D
d
t
t
u u

=
(
| |
|
(
\ .

}
, (15.13)

0
(2 2) (2 4) ... 2
(2 3) (2 5) ... 1
N N
D
N N

=

, (15.14)

0
1.77 1.77 1 2 D N L ~ = + . (15.15)

4. DolphChebyshev array (DCA)
Dolph proposed (in 1946) a method to design arrays with any desired side-
lobe levels and any HPBWs. This method is based on the approximation of the
pattern of the array by a Chebyshev polynomial of order m, high enough to
meet the requirement for the side-lobe levels. A DCA with no side lobes (side-
lobe level of - dB) reduces to the binomial design.
4.1. Chebyshev polynomials
The Chebyshev polynomial of order m is defined by

( )
( )
( 1) cosh( arccosh | |), 1,
( ) cos arccos( ) , 1 1,
cosh arccosh( ) , 1.
m
m
m z z
T z m z z
m z z

= s s

>

(15.16)
A Chebyshev polynomial T
m
(z) of any order m can be derived via a recursion
formula, provided T
m1
(z) and T
m2
(z) are known:

1 2
( ) 2 ( ) ( )
m m m
T z zT z T z

= . (15.17)
Explicitly, (15.17) produces:
0
0, ( ) 1 m T z = =
1
1, ( ) m T z z = =
2
2
2, ( ) 2 1 m T z z = =
Nikolova 2012
12
3
3
3, ( ) 4 3 m T z z z = =
4 2
4
4, ( ) 8 8 1 m T z z z = = +
5 3
5
5, ( ) 16 20 5 , etc. m T z z z z = = +
If | | 1 z s , then the Chebyshev polynomials are related to the cosine
functions, see (15.16). We can always expand the function cos(mx) as a
polynomial of cos(x) of order m, e.g., for 2 m = ,

2
cos 2 2cos 1 x x = . (15.18)
This can be done by observing that ( )
jx m jmx
e e = and making use of Eulers
formula to expand the exponents as
(cos sin ) cos( ) sin( )
m
x j x mx j mx + = + . (15.19)
The left side of the equation is then expanded and its real and imaginary parts
are equated to those on the right. Similar relations hold for the hyperbolic
cosine function.
Comparing the trigonometric relation in (15.18) with the expression for
2
( ) T z above, we see that the Chebyshev argument z is related to the cosine
argument x by
cos or arccos z x x z = = . (15.20)
For example, (15.18) can be written as:

| |
2
cos(2arccos ) 2 cos(arccos ) 1 z z = ,

2
2
cos(2arccos ) 2 1 ( ) z z T z = = . (15.21)
Properties of the Chebyshev polynomials:
1) All polynomials of any order m pass through the point (1,1).
2) Within the range 1 1 z s s , the polynomials have values within [1,1].
3) All nulls occur within 1 1 z s s .
4) The maxima and minima in the [ 1,1] z e range have values +1 and 1,
respectively.
5) The higher the order of the polynomial, the steeper the slope for | | 1 z > .

Nikolova 2012
13

Fig. 6.19, pp. 296, Balanis

Nikolova 2012
14
4.2. Chebyshev array design
The main goal is to approximate the desired AF with a Chebyshev
polynomial such that
- the side-lobe level meets the requirements, and
- the main beam width is as small as possible.
An array of N elements has an AF approximated with a Chebyshev polynomial
of order m, which is always
1 m N = , (15.22)
where 2 N M = , if N is even; and 2 1 N M = + , if N is odd.
In general, for a given side-lobe level, the higher the order m of the
polynomial, the narrower the beamwidth. However, for m > 10, the difference
is not substantial see the slopes of ( )
m
T z in the previous figure. The AF of an
N-element array (15.5) or (15.6) is identical with a Chebyshev polynomial if

| |
| |
1
1
1
1
cos (2 1) , 2 , even
( )
cos 2( 1) , 2 1, odd
M
n
n
N
M
n
n
a n u N M
T z
a n u N M
=

+
=

= +

(15.23)
Here, ( / )cos u d t u = . Let the side-lobe level be

max
0
1
sl sl
E
R
E AF
= = (voltage ratio). (15.24)
Then, the maximum of
1 N
T

is fixed at an argument
0
z (
0
| | 1 z > ), where

max
0 0
1
( )
N
T z R

= . (15.25)
Equation (15.25) corresponds to
max
0
( ) ( ) AF u AF u = . Obviously,
0
z must
satisfy the condition:

0
| | 1 z > , (15.26)
where
1
1
N
T

> . The maxima of
1
| ( ) |
N
T z

for | | 1 z s are equal to unity and they

correspond to the side lobes of the AF. Thus, ( ) AF u has side-lobe levels equal
to
0
R . The AF is a polynomial of cosu , and the
1
( )
N
T z

is a polynomial of z
where the limits for z are

0
1 z z s s . (15.27)
Nikolova 2012
15
Since
1 cos 1 u s s , (15.28)
the relation between z and cosu must be normalized as

0
cos / u z z = . (15.29)

Design of a DCA of N elements general procedure:
1) Expand the AF as given by (15.5) or (15.6) by replacing each cos( ) mu
term ( 1, 2,..., m M = ) with the power series of cosu .
2) Determine
0
z such that
0 0
1
( )
N
T z R

= (voltage ratio).
3) Substitute
0
cos / u z z = in the AF as found in step 1.
4) Equate the AF found in Step 3 to
1
( )
N
T z

and determine the coefficients

for each power of z.

Example: Design a DCA (broadside) of N=10 elements with a major-to-minor
lobe ratio of
0
26 R = dB. Find the excitation coefficients and form the AF.

Solution:
The order of the Chebyshev polynomial is 1 9 m N = = . The AF for an even-
number array is:
| |
5
2
1
cos (2 1) , cos
M n
n
d
AF a n u u
t
u

=
= =

, 5 M = .
Step 1: Write
10
AF explicitly:
10 1 2 3 4 5
cos cos3 cos5 cos7 cos9 AF a u a u a u a u a u = + + + + .
Expand the cos( ) mu terms of powers of cosu :
3
cos3 4cos 3cos u u u = ,
5 3
cos5 16cos 20cos 5cos u u u u = + ,
7 5 3
cos7 64cos 112cos 56cos 7cos u u u u u = + ,
9 7 5 3
cos9 256cos 576cos 432cos 120cos 9cos u u u u u u = + + .

Nikolova 2012
16
Step 2: Determine
0
z :
0
26 dB R =
26
20
0
10 20 R =
9 0
( ) 20 T z = ,
| |
0
cosh 9arccosh( ) 20 z = ,
0
9arccosh( ) arccosh20 3.69 z = = ,
0
arccosh( ) 0.41 z = ,
0
cosh0.41 z =
0
1.08515 z = .

Step 3: Express the AF from Step 1 in terms of
0
cos / u z z = :
( )
( )
( )
( )
( )
9
10 1 2 3 4 5
0
3
2 3 4 5
3
0
5
3 4 5
5
0
7
4 5
7
0
9
5
9
0
3 5 7 9
( )
3 5 7 9
4 20 56 120
16 112 432
64 576
256
=9 120 432 576 256
T z
z
AF a a a a a
z
z
a a a a
z
z
a a a
z
z
a a
z
z
a
z
z z z z z
= + +
+ +
+ +
+
+ =
+ +

Step 4: Finding the coefficients by matching the power terms:
9
5 0 5
256 256 2.0860 a z a = =
7
4 5 4
0
64 576 576 2.8308 a a z a = =
5
3 4 5 3
0
16 112 432 432 4.1184 a a a z a + = =
7
2 3 4 5 2
0
4 20 56 120 120 5.2073 a a a a z a + = =
9
1 2 3 4 5 1
0
3 5 7 9 9 5.8377 a a a a a z a + + = =
Normalize coefficients with respect to edge element (N=5):
5 4 3 2 1
1; 1.357; 1.974; 2.496; 2.789 a a a a a = = = = =
Nikolova 2012
17
10
2.789cos( ) 2.496cos(3 ) 1.974cos(5 ) 1.357cos(7 ) cos(9 ) AF u u u u u = + + + +
where cos
d
u
t
u

= .



Fig. 6.20b, p. 298, Balanis

Nikolova 2012
18

Fig. 6.21, p. 300, Balanis




Nikolova 2012
19
4.3. Maximum affordable d for Dolph-Chebyshev arrays
This restriction arises from the requirement for a single major lobe see
also equation (15.27):
1 z > ,

0
cos cos 1
d
z
t
u

| |
>
|
\ .
. (15.30)
For a given array, when u varies from 0 to 180, the argument z assumes
values
from
0
( 0 )
cos
d
z z
u
t

=
= (15.31)
through
0
( 90 )
z z
u =
= (15.32)
to
0
( 180 ) ( 0 )
cos
d
z z z
u u
t

= =
| |
= =
|
\ .
. (15.33)
The extreme value of z to the left on the abscissa corresponds to the end-fire
directions of the AF. This value must not go beyond 1 z = . Otherwise, minor
lobes of levels higher than 1 (higher than
0
R ) will appear. Therefore, the
inequality (15.30) must hold for 0 u = or 180:

max max
0
0
1
cos 1 cos
d d
z
z
t t

| | | |
> >
| |
\ . \ .
. (15.34)
Let

1
0
arccos z

= . (15.35)
Then,

max
0
1
arccos
d
z
t
t t

| |
< =
|
\ .
max
0
1 1
1 arccos
d
z t
| |
<
|
\ .
. (15.36)
In the previous example,
max
1 1 0.39879
1 arccos 1 0.873
1.08515
d
t t
| |
< = =
|
\ .
,
max
0.873 d < .
Nikolova 2012
20
0
1
z

cos( / ) d t
1
1
1
max
d t

ILLUSTRATION OF EQUATION (15.34) AND THE REQUIREMENT IN (15.36)

5. Directivity of non-uniform arrays
It is difficult to derive closed form expressions for the directivity of non-
uniform arrays. Here, we derive expressions in the form of series in the most
general case of a linear array when the excitation coefficients are known.
The non-normalized array factor is

1
cos
0
n n
N
j jkz
n
n
AF a e e
| u

=
=

, (15.37)
where
n
a is the amplitude of the excitation of the n-th element;
n
| is the phase angle of the excitation of the n-th element;
n
z is the z-coordinate of the n-th element.
The maximum AF is

1
max
0
N
n
n
AF a

=
=

. (15.38)
The normalized AF is

1
cos
0
1
max
0
n n
N
j jkz
n
n
n
N
n
n
a e e
AF
AF
AF
a
| u

=
= =

. (15.39)
Nikolova 2012
21
The beam solid angle is derived as

( )
2
0
2 sin
A n
AF d
t
t u u u O =
}
,

( )
1 1
( ) cos
2
1
0 0
0
0
2
sin
m p m p
N N
j jk z z
A m p
N
m p
n
n
a a e e d
a
t
| | u
t
u u

= =
=
O =
| |
|
\ .

, (15.40)
where
( )
cos
0
2sin ( )
sin
( )
m p
m p
jk z z
m p
k z z
e d
k z z
t
u
u u

(

=

}
.

( )
1 1
2
1
0 0
0
sin ( )
4
( )
m p
N N
m p
j
A m p
N
m p
m p
n
n
k z z
a a e
k z z
a
| |
t

= =
=
(

O =

| |
|
\ .

. (15.41)
From

0
4
A
D
t
=
O
,
we obtain

( )
2
1
0
0
1 1
0 0
sin ( )
( )
m p
N
n
n
N N
m p
j
m p
m p
m p
a
D
k z z
a a e
k z z
| |

= =
| |
|
\ .
=
(

. (15.42)
For equispaced linear (
n
z nd = ) arrays, (15.42) reduces to

( )
| |
2
1
0
0
1 1
0 0
sin ( )
( )
m p
N
n
n
N N
j
m p
m p
a
D
m p kd
a a e
m p kd
| |

= =
| |
|
\ .
=

. (15.43)
Nikolova 2012
22
For equispaced broadside arrays, where
m p
| | = for any (m,p), (15.43)
reduces to

| |
2
1
0
0
1 1
0 0
sin ( )
( )
N
n
n
N N
m p
m p
a
D
m p kd
a a
m p kd

=

= =
| |
|
\ .
=

. (15.44)
For equispaced broadside uniform arrays,

| |
2
0
1 1
0 0
sin ( )
( )
N N
m p
N
D
m p kd
m p kd

= =
=

. (15.45)
When the spacing d is a multiple of / 2 , equation (15.44) reduces to

2
1
0
0
1
2
0
, , ,...
2
( )
N
n
n
N
n
n
a
D d
a

=
| |
|
\ .
= =

. (15.46)


Example: Calculate the directivity of the DolphChebyshev array designed in
the previous example if / 2 d = .

The 10-element DCA has the following amplitude distribution:

5 4 3 2 1
1; 1.357; 1.974; 2.496; 2.798 a a a a a = = = = = .
We make use of (15.46):

( )
2
5
2
1
0
5
2
1
4
9.625
2 8.9090 (9.5 dB)
20.797
2 ( )
n
n
n
n
a
D
a
=
=
| |
|
\ .
= = =

.
Output from ARRAYS.m:
0
8.9276 D = .
Nikolova 2012
23
6. Half-power beamwidth of a BS DCA
For large DCA with side lobes in the range from 20 dB to 60 dB, the
HPBW can be found by introducing a beam-broadening factor f given by

2
2 2
0
0
2
1 0.636 cosh (arccosh ) f R
R
t

(
= +
`

)
. (15.47)
The HPBW of the DCA is equal to the product of the broadening factor by the
HPBW of the respective uniform linear array:

DCA UA
HPBW f HPBW = . (15.48)
In (15.47),
0
R denotes the side-lobe level (voltage ratio).