IEEE TRANSACTIONS ON ANTENNAS AND PROPAGATION, VOL. 61, NO. 2, FEBRUARY 2013 627
Fast Pencil Beam Pattern Synthesis of Large
Unequally Spaced Antenna Arrays
Kai Yang, Student Member, IEEE, Zhiqin Zhao, Senior Member, IEEE, and Qing Huo Liu, Fellow, IEEE
Abstract—Reducing the computational cost of large array pat-
tern synthesis is attractive in many applications. In this paper, a
fast pencil beam pattern synthesis method for large nonu niform
antenna arrays is proposed. This method is based on an interpo-
lation in a least square sense and iterative fast Fourier transform
(FFT), i.e., interpolate the nonuniform distributed elements int
o
uniform virtual elements , and th en apply FFT to synthesis the uni-
form array. Thanks to the efficiency of FFT, the proposed method
is much faster and can handle much larger arrays than the existing
methods. To guarantee the interpolation accuracy, the choice of
the interpolation parameters is discussed. Both linear and planar
nonuniform array examples are shown to validate the adv
antages
of the proposed method.
Index Terms—Array pattern synthesis, nonuniform fast Fourier
transform, unequally spaced array.
I. INTRODUCTION
N
ONUNIFORM or unequally spaced arrays have a
wide range of applications in radar, sonar and wireless
communication, and othe
r areas. One example of using a
nonuniform array is to prevent grating lobes or obtain a better
pattern performance than a uni form array. Another example is
the use of a nonunifo
rm array may be more suitable when the
antennas are restricted into a limited space on the body of flight
vehicles. In recent literature, several novel algorithms were
developed to op
timize the element positions and excitations of
nonuniform array s to improve pattern performance [1]–[3] and
reduce the computational cost [4]–[6] or obtain a given pattern
[7]–[10].
This paper focuses on the problem of optimization of the an-
tenna elem ent excitations with all the element positions fixed to
obtain a d
esired array pattern performance, especially for large
linear
elements and planar elements arrays.
Many modern applications require arrays that are capable of
achie
ving very narrow main beam. Therefore, large array pat-
tern synthesis methods are requ ired [11], [12]. During the past
Manuscript received March 13, 2012; revised June 16, 2012; accepted Au-
gust 30, 2012. Date of publication September 21, 2012; date of current version
January 30, 2013. This work was supporte d in part by the National Natural Sci-
ence Foundation of China under Grants 60927002 and 61171044.
K. Yang is with the School of Electronic Engineering, University of Elec-
tronic Science and Technology of China (UESTC), Chengdu 611731, China,
and also with the Department of Electrical and Compute r Engineering, Duke
University, Durham, NC 27708 USA (e-mail: eekaiyang@gmail.com).
Z. Zhao is with the School of Electronic Engin eering, University of Electronic
Science and Technology of China (UESTC), Chengdu 611731, China (e-mail:
zqzhao@uestc.edu.cn).
Q. H. Liu is with the Department of Electronic Engineering, Duke University,
Durham, NC 27708 U SA (e-mail: qh liu@ee.duke.ed u).
Di
gital Object Identifier 10.1109/TAP.2012.2220319
50 years, sev eral array pattern synthesis techniques, such as
analytical methods, stochastic methods, adaptive array theory
method and convex optimization methods, have been presented
for obtaining low side lobe levels. The analytical methods [13]
use the matrix relationship between the elements and the far-
zone array pattern of the array. This kind of method is a non-iter-
ative process, thus time efficient. However, this method requires
a specified array pattern and is difficult to be applied to larg
e
arrays. The stochastic m e tho ds, such as g enetic algorithm (GA)
[14], [15], particle swarm optimization (PSO) method [16], [17]
and simulated annealing (SA) method [18]–[20], a
re global op-
timization methods which can escape from local optimal so-
lutions. T he stochastic methods can solve large optimization
problem, but the computation time is huge and
will grow rapidly
with the p roblem size. The adaptive array theory method [21]
is an it erat ive method to minimize the errors between synth e-
sized and desired patterns by recursi
vely adjust ing the inter-
ference-to-noise-r ati o. S imilar to the stochastic met hods, this
method is difficult to be applied to large arrays due to the com-
putational cost. Element excita
tions optimization can be posed
in a convex form [22]– [24], and then it can be solved using a
convex problem solver (for example, SeDuMi [25] and CV X
[26]). However, most conve
x optim izat ion solvers are unlikely
to work well (or at all) for very large problems [26].
A promising iterative f ast F ou rier transform (FFT) method
[27], [28] was presented
to synthesize large planar arrays. This
method is based on the Fourier transform-pair relationship be-
tween the array excitations and the array factor. Thanks to the ef-
ficiency of FFT, this
method can solve large array synthesis with
much less time compared with the stochastic methods. How-
ever, this method requires all the elem e nts i n a uniform grid.
In this pa per, we e
xtend the iterative FFT method to nonuni-
form array through exponential in terpolation. Each element in
the nonuniform array is interpolated by several elements in a vir-
tual uniform a
rray, a nd then we optim ize the excitations of this
virtual array to arrive at a desired array pattern. Finally, calcu-
late the real element excitations by the optimized virtual element
excitatio
ns. To ensure the agreement between the virtual array
pattern and the real array pattern, t he exponential interpolation
error should be small. N umerical simulation results show that
the int
erpolation strategy can make the interpolation error very
small if we choose the interpolation parameters appropriately.
In addition, we should make sure th e virtual element excitations
are
always in the subspace of the real element excitations during
the iterative process; otherwise, there will be no exact solutions
for obtaining the real element excitations from virtual element
e
xcitations.
0018-926X/$31.00 © 2012 IEEE