LECTURE 15: LINEAR ARRAYS – PART III

(N-element linear arrays with uniform spacing and non-uniform amplitude:

Binomial array; Dolph–Tschebyscheff array. Directivity and design.)

1. Advantages of Linear Arrays with Nonuniform Amplitude Distribution

The most often met BSAs, classified according to the type of their excitation
amplitudes, are:
a) the uniform BSA – relatively high directivity, but the side-lobe levels are
high;
b) Dolph–Tschebyscheff (or Chebyshev)1 BSA – for a given directivity with a
fixed number of array elements, achieves the lowest side-lobe level;
c) binomial BSA – does not have good directivity (for a given number of
elements) but has low side-lobes (if d =  / 2 , no side lobes at all).

2. Array Factor of Linear Arrays 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 z = 0 . For a broadside array (  = 0) ,
1 3 2 M −1
j kd cos j kd cos j kd cos
AF e = a1e 2 + a2 e 2 + + aM e 2 +
(15.1)
1 3 2 M −1
− j kd cos − j kd cos −j kd cos
+ a1e 2 + a2 e 2
+ + aM e 2 ,
M
 2n − 1  
 AF = 2 an cos 
e
 kd cos  . (15.2)
n =1   2  
If the linear array consists of an odd number (2M+1) of elements, located
symmetrically along the z-axis, the array factor is
AF o = 2a1 + a2e jkd cos + a3e j 2kd cos + ... + aM +1e jMkd cos +
(15.3)
+ a2e− jkd cos + a3e− j 2 kd cos + ... + aM +1e− jMkd cos ,
M +1
 AF o = 2  an cos ( n − 1) kd cos  . (15.4)
n =1

1
Russian spelling is Чебышёв.

Nikolova 2020 1
 EVEN- AND ODD-NUMBER ARRAYS

Fig. 6.17, p. 291, Balanis

Nikolova 2020 2
The normalized AF derived from (15.2) and (15.4) can be written in the form
M
AF e =  an cos  (2n − 1)u , for N = 2M , (15.5)
n =1
M +1
AF o =  an cos  2(n − 1)u , for N = 2M + 1, (15.6)
n =1

1 d
where u = kd cos = cos .
2 

Examples of AFs of arrays of nonuniform amplitude distribution

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

pp. 148-149, Stutzman

Nikolova 2020 3
 b) triangular (1:2:3:2:1) amplitude distribution (N = 5, d =  / 2 , max. at
0 = 90 °)

pp. 148-149, Stutzman

Nikolova 2020 4
 c) binomial (1:4:6:4:1) amplitude distribution (N = 5, d =  / 2 , max. at
0 = 90 °)

pp. 148-149, Stutzman

Nikolova 2020 5
 d) Dolph-Tschebyschev (1:1.61:1.94:1.61:1) amplitude distribution (N = 5,
d =  / 2 , max. at 0 = 90 °)

pp. 148-149, Stutzman

Nikolova 2020 6
 e) Dolph-Tschebyschev (1:2.41:3.14:2.41:1) amplitude distribution (N = 5,
d =  / 2 , max. at 0 = 90 °)

pp. 148-149, Stutzman

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

Nikolova 2020 7
3. Binomial Broadside Array
The binomial BSA was investigated and proposed by J. S. Stone2 to
synthesize patterns without side lobes. First, consider a 2–element array (along
the z-axis).
z

x
The elements of the array are identical and their excitations are the same. The
array factor is of the form
AF = 1 + Z , where Z = e j = e j ( kd cos +  ) . (15.7)
If the spacing is d   / 2 and  = 0 (broad-side maximum), the array pattern
|AF| has no side lobes at all. This is proven as follows.
| AF |2 = (1 + cos ) 2 + sin 2  = 2(1 + cos ) = 4cos 2 ( / 2) (15.8)
where  = kd cos . The first null of the array factor is obtained from (15.8) as
1 2    
  d cos n1,2 =    n1,2 =  arccos  . (15.9)
2  2  2d 
As long as d   / 2 , the first null does not exist. If d =  / 2 , then  n1,2 = 0,
180 °. Thus, in the “visible” range of θ, all secondary lobes are eliminated.
Second, consider a 2–element array whose elements are identical and the
same as the array given above. The distance between the two arrays is again d.
z

d
y

2
US Patents #1,643,323, #1,715,433.

Nikolova 2020 8
This new array has an AF of the form
AF = (1 + Z )(1 + Z ) = 1 + 2Z + Z 2 . (15.10)
Since (1 + Z ) has no side lobes, (1 + Z )2 does not have side lobes either.
Continuing the process for an N-element array produces
AF = (1 + Z ) N −1 . (15.11)
If d   / 2 , 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 Pascal’s 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 a fairly wide variation of the amplitude, which is
a disadvantage of the BAs. The overall efficiency of such an array would be
low. Besides, the BA has relatively wide beam. Its HPBW is the largest as
compared to the uniform BSA or the DCA for a give number of elements.
An approximate closed-form expression for the HPBW of a BA with
d =  / 2 is
1.06 1.06 1.75
HPBW  = = , (15.12)
N −1 2L  L
where L = ( N − 1)d is the array’s length. The AFs of 10-element broadside
binomial arrays (N = 10) are given below.

Nikolova 2020 9
 d =
d = 3 / 4
d =/2 d = 3 / 4

d =/4

d =

Fig. 6.18, p.293, Balanis

Nikolova 2020 10
 The directivity of a broadside BA with spacing d =  / 2 can be calculated
as
4 2
D0 = = 2( N −1)
, (15.13)
Prad     
   2
cos  cos  

d
0

(2 N − 2)  (2 N − 4)  ...  2
D0 = , (15.14)
(2 N − 3)  (2 N − 5)  ...  1
D0  1.77 N = 1.77 1 + 2 L  . (15.15)

4. Dolph–Chebyshev Array (DCA)

Dolph proposed (in 1946) a method for designing arrays with any desired
side-lobe level for a given HPBW. 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) m cosh(m  arccosh | z |), z  −1,

Tm ( z ) = cos ( m  arccos( z ) ) , −1  z  1, (15.16)

cosh ( m  arccosh(z ) ) , z  1.
A Chebyshev polynomial Tm(z) of any order m can be derived via a recursion
formula, provided Tm−1(z) and Tm−2(z) are known:
Tm ( z ) = 2 zTm −1 ( z ) − Tm − 2 ( z ) . (15.17)
Explicitly, from (15.16) we see that
m = 0, T0 ( z ) = 1
m = 1, T1 ( z ) = z .
Then, (15.17) produces:

Nikolova 2020 11
 m = 2, T2 ( z ) = 2 z 2 − 1
m = 3, T3 ( z ) = 4 z 3 − 3z
m = 4, T4 ( z ) = 8 z 4 − 8 z 2 + 1
m = 5, T5 ( z ) = 16 z 5 − 20 z 3 + 5 z , etc.
If | z | 1, then the Chebyshev polynomials are related to the cosine functions
through z = cos x so that Tm ( x) = cos(mx) ; see (15.16). We can expand the
function cos(mx) as a polynomial of cos(x) of order m. For example, for m = 2 ,
cos 2 x = 2cos 2 x − 1 . (15.18)
The expansion of cos(mx) can be done by observing that ( e jx )m = e jmx and by
making use of Euler’s formula, which leads to
(cos x + j sin x )m = cos( mx ) + j sin(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 cosh.
Comparing the trigonometric relation in (15.18) with the expression for
T2 ( z ) above (see the expanded Chebybshev polynomials after (15.17)), we see
that the Chebyshev argument z is related to the cosine argument x by
z = cos x or x = arccos z ,| z | 1
. (15.20)
z = cosh x or x = arccosh z ,| z | 1
For example, (15.18) can be written as:
cos(2 arccos z ) = 2  cos( arccos z )  −1cos(2 arccos z ) = 2 z 2 −1=T2 ( z ) . (15.21)
2

Properties of the Chebyshev polynomials:

1) All polynomials of any order m pass through the point (1,1).
2) Within the range −1  z  1, the polynomials have values within [–1,1].
3) All nulls occur within −1  z  1.
4) The maxima and minima in the z  [−1,1] range have values +1 and –1,
respectively.
5) The higher the order of the polynomial, the steeper the slope for | z | 1.

Nikolova 2020 12
 Fig. 6.19, pp. 296, Balanis

Nikolova 2020 13
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
m = N − 1. (15.22)
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 Tm ( z ) in the previous figure. The AF of an N-
element array (15.5) or (15.6) is shaped by a Chebyshev polynomial by
requiring that
 e M
 AF =  an cos  (2n − 1)u , M = N / 2, even
 n =1
TN −1 ( z ) =  (15.23)
M +1
 AF o =
  an cos  2(n − 1)u , M = ( N − 1) / 2, odd
n =1

where u = ( d /  )cos . Let the side-lobe level be

Emax 1
R0 = = (aka voltage ratio). (15.24)
Esl AFsl
Then, we require that the maximum of TN −1 is fixed at an argument z0 where
−1 ( z0 ) = R0 , | z0 | 1 .
TNmax (15.25)
Equation (15.25) corresponds to AF (u ) = AF max (u0 ) and z0 must satisfy the
condition | z0 | 1 so that TN −1  1. The maxima of | TN −1 ( z ) | for | z | 1 are equal
to unity and they correspond to the side lobes of the AF. Thus, AF (u ) has side-
lobe levels equal to 1 / R0 .
The AF is a polynomial of cosu , and the TN −1 ( z ) is a polynomial of z where
z is limited to the range
−1  z  z0  1. (15.26)
Since −1  cos u  1 , the relation between z and cosu must be normalized as
Nikolova 2020 14
 cos u = z / z0 , u = ( d /  )cos . (15.27)

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
( m = 1, 2,..., M ) with the power series of cosu .
2) Determine z0 such that TN −1 ( z0 ) = R0 (voltage ratio).
3) Substitute cos u = z / z0 in the AF found in step 1.
4) Equate the AF found in Step 3 to TN −1 ( 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 R0 = 26 dB. Find the excitation coefficients and the AF.

Solution:
The order of the Chebyshev polynomial is m = N − 1 = 9 . The AF for an even-
number array is:
5
d
AF2 M =  an cos  (2n − 1)u , u = cos , M = 5 .
n =1 
Step 1: Write AF10 (see sum above) explicitly:
AF10 = a1 cos u + a2 cos3u + a3 cos5u + a4 cos7u + a5 cos9u .
Expand the cos(mu ) terms as powers of cosu :
cos3u = 4 cos3 u − 3cos u , ( a2 terms)
cos5u = 16 cos5 u − 20 cos3 u + 5cos u , ( a3 terms)
cos7u = 64 cos7 u − 112 cos5 u + 56 cos3 u − 7cos u , ( a4 terms)
cos 9u = 256 cos9 u − 576 cos7 u + 432 cos5 u − 120 cos3 u + 9cos u . ( a5 terms)
Note that the above expansions can be readily obtained from the recursive
Chebyshev relation (15.17) and the substitution z = cos u . For example,
m = 3, T3 ( z ) = 4 z 3 − 3z
translates into: cos(3u ) = 4cos3 u − 3cos u .

Nikolova 2020 15
Step 2: Determine z0 :
26
R0 = 26 dB  R0 = 10 20  20  T9 ( z0 ) = 20 ,
cosh 9arccosh( z0 )  = 20 ,
9arccosh( z0 ) = arccosh20 = 3.69 ,
arccosh( z0 ) = 0.41,
z0 = cosh 0.41  z0 = 1.08515 .

Step 3: Express the AF from Step 1 in terms of cos u = z / z0 and make equal to
the Chebyshev polynomial:
z
AF10 = ( a1 − 3a2 + 5a3 − 7a4 + 9a5 )
z0
z3
+ 3 ( 4a2 − 20a3 + 56a4 − 120a5 )
z0
z5
+ 5 (16a3 − 112a4 + 432a5 )
z0
z7
+ 7 ( 64a4 − 576a5 )
z0
z9
+ 9 ( 256a5 ) =
z0
= 9 z − 120 z 3 + 432 z 5 − 576 z 7 + 256 z 9
T9 ( z )

Step 4: Find the coefficients by matching the power terms:

256a5 = 256 z09  a5 = 2.0860
64a4 − 576a5 = −576 z07  a4 = 2.8308
16a3 − 112a4 + 432a5 = 432 z05  a3 = 4.1184
4a2 − 20a3 + 56a4 − 120a5 = −120 z07  a2 = 5.2073
a1 − 3a2 + 5a3 − 7a4 + 9a5 = 9 z09  a1 = 5.8377
Normalize coefficients with respect to edge element (N=5):

Nikolova 2020 16
 a5 = 1; a4 = 1.357; a3 = 1.974; a2 = 2.496; a1 = 2.789
 AF10 = 2.789cos(u ) + 2.496cos(3u ) +1.974cos(5u ) +1.357cos(7u ) + cos(9u )
d
where u = cos .


Fig. 6.20b, p. 298, Balanis

Nikolova 2020 17
 Fig. 6.21, p. 300, Balanis

Nikolova 2020 18
4.3. Maximum affordable d for Dolph-Chebyshev arrays
This restriction arises from the requirement for a single major lobe – see
also equation (15.26), −1  z  z0 :
d
z  −1, cos u = z / z0 , u = cos ,

d 
 z = z0 cos  cos    −1 . (15.28)
  
For a broadside array, when  varies from 0 to 180 , the argument z assumes
values
d 
from z( =0 ) = z0 cos   (15.29)
  
to z( =90 ) = z0 (15.30)
 d 
back to z( =180 ) = z0 cos  −  = z( =0 ) . (15.31)
  
The extreme value of z to the left on the abscissa corresponds to the end-fire
directions of the AF ( = 0,180 ). This value must not go beyond z = −1 .
Otherwise, end-fire lobes of levels higher than 1 (higher than R0 ) will appear.
Therefore, the inequality (15.28) must hold for  = 0 or 180 :
d  d  1
z0 cos    −1  cos    − . (15.32)
      z0
Let the angle  be such that cos  = 1 / z0 (see figure below). Then,
 = arccos ( z0−1 ) and
cos ( d /  )  − cos  . (15.33)
1

 d max

−1 cos   1
1 allowed cos( d /  ) cos( d /  )
−
z0
ILLUSTRATION OF EQUATION (15.32) AND THE REQUIREMENT IN (15.33) AND (15.34)

Nikolova 2020 19
Remember that z0  1; thus  is a real-positive angle. Then, from (15.33),
 d /    −  =  − arccos ( z0−1 ) (15.34)
or
 d max 1
=  − arccos ( z0−1 )  max = 1 − arccos  
d 1
(15.35)
    z0 


For the case of the previous example,

d 1  1  0.39879
 1− arccos   = 1 − = 0.873 ,
   1.08515  
d max = 0.873 .

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
N −1
AF =  an e j n e jkzn cos , (15.36)
n =0

where
an is the amplitude of the excitation of the n-th element;
 n is the phase angle of the excitation of the n-th element;
zn is the z-coordinate of the n-th element.
The maximum AF is
N −1
AFmax =  an . (15.37)
n =0

The normalized AF is

Nikolova 2020 20
 N −1

AF
 ane j e jkz cos
n n

n =0
AFn = = N −1
. (15.38)
AFmax
 an
n =0

The beam solid angle of a linear array along z is


 A = 2  AFn ( ) sin  d ,
2

0
N −1 N −1 
2 j ( m −  p )
A = 2   am a p e  e jk ( z m − z p ) cos  sin  d , (15.39)
 N −1  m=0 p =0
 
0
an
 n =0 
where

2sin  k ( zm − z p ) 
 e ( m p ) sin  d =
jk z − z cos 
.
0
k ( zm − z p )
4 N −1 N −1 sin  k ( zm − z p ) 
 A = 2   am a p e j( m −  p )  k ( zm − z p )
. (15.40)
 N −1  m=0 p =0
 n a
 n =0 
From
D0 = 4 /  A ,
we obtain
2
 N −1 
  an 
 D0 = N −1 N −1  n =0  . (15.41)
 m − p 
  am a p e j( m −  p )  k( zm − z p ) 
sin k ( z z )
m =0 p =0

For equispaced linear ( zn = nd ) arrays, (15.41) reduces to

Nikolova 2020 21
 2
 N −1 
  an 
D0 = N −1 N −1  n =0  . (15.42)
sin  ( m − p ) kd 
  am a p e j( m −  p )  (m − p)kd
m=0 p =0

For equispaced broadside arrays, where  m =  p for any (m,p), (15.42)

reduces to
2
 N −1 
  an 
D0 = N −1 N −1  n =0  . (15.43)
sin  (m − p )kd 
  am a p  (m − p)kd
m =0 p =0

For equispaced broadside uniform arrays,

N2
D0 = N −1 N −1 . (15.44)
sin  (m − p)kd 
  (m − p)kd
m =0 p =0

When the spacing d is a multiple of  / 2 , equation (15.43) reduces to

2
 N −1 
  an 

D0 = Nn−=10  , d = ,  ,... . (15.45)
2
 ( an ) 2

n =0

Example: Calculate the directivity of the Dolph–Chebyshev array designed in

the previous example if d =  / 2 .

The 10-element DCA has the following amplitude distribution:

a5 = 1; a4 = 1.357; a3 = 1.974; a2 = 2.496; a1 = 2.798 .
We make use of (15.45):

Nikolova 2020 22
 2
 5 
4   an 
D0 = 5 n =1  = 2
( 9.625 )
2
= 8.9090 (9.5 dB) .
20.797
2 (an ) 2
n =1

Output from ARRAYS.m: D0 = 8.9276 .

6. Half-power Beamwidth of a Broadside DCA

For large DCAs with side lobes in the range from –20 dB to –60 dB, the
HPBW HPBWDCA can be found from the HPBW of a uniform array HPBWUA
by introducing a beam-broadening factor f given by
2
2 
f = 1 + 0.636  cosh  (arccoshR0 ) 2 −  2   , (15.46)
 R0  
so that
HPBWDCA = f  HPBWUA . (15.47)
In (15.46), R0 denotes the side-lobe level (the voltage ratio).

Nikolova 2020 23