Você está na página 1de 4

Proceedings of the 3rd European Radar Conference

Design considerations in sparse array antennas


Cristian I. Coman, Joan E. Lager, and Leonardus P. Ligthart
International Research Centre for Telecommunications and Radar, Delft University of Technology, 2628 CD, Delft, the Netherlands, Phone: +31 15 278 1034, E-mail: c.coman@irctr.tudelft.nl
Abstract- The design aspects concerning the sparse array antennas are discussed. Commonly, the beamwidth, the side-lobes level and the gain are the relevant features in antenna design procedures. These parameters are evaluated for different sparse topologies and generalised design relations are provided. The capabilities of sparse architectures are evaluated by comparison with uniform arrays for a similar level of performance. Index Terms- Antenna arrays, radiation properties.
I. INTRODUCTION

The idea of developing electronically controlled antenna radiation patterns (RP) goes back to 1940. It stemmed from the requirement to replace the mechanical steering of the antenna main beam by a more reliable, electronic method. Later, the fact that the elementary antennas used in array configurations act as sampling devices in the spatial domain was also employed for enhancing antenna capabilities. The most important benefits from using array antennas are: reduced (or even completely eliminated) mechanical movement; wellcontrolled beamwidth and side-lobes region; fast scanning in the field of view; ability to control the positions of the nulls in the radiation pattern; electronically re-configurable antenna radiation pattern. While phased array antennas have such attractive features, their realization represents a challenging task. In this respect, it is known that for applications where a high resolution (narrow beamwidth) is required the antenna should have a large spatial dimension. At the same time, when scanning in the entire half-space is aimed at, preventing the apparition of grating lobes (i.e. alias effects due to spatial under-sampling) requires the inter-element spacing in uniform configurations not to exceed Amin/2, with Amin denoting the wavelength at the maximum operational frequency. The combination of the fine resolution requirement with the Amin/2 condition will often result into excessively large numbers of elements. As each primary radiator is associated with a transmitting/ receiving module, the cost of a fully populated, large aperture array antenna becomes quickly prohibitive. The manufacturing of array antennas is subject to technical limitations, as well. Commonly, the characteristic size of radiators is approximately equal to half of the wavelength at the minimum operational frequency, that is obviously conflicting with the Amin/2 condition. Consequently, the individual elements to be employed in fully sampled arrays must be miniaturised, an often technologically difficult solution. Furthermore, the individual radiators are positioned very close to each other, the mutual coupling being, thus, very high. This phenomenon

affects detrimentally the antenna performance, most notably as concerns the side-lobes level and the maximum scanning angle. The problems mentioned above (cost, conflicting dimensional requirements, mutual coupling) can be more effectively solved by reducing the number of elements in the array configuration (with its accompanying increase of the interelement spacing). The resulting antenna will be a sparse array. When the sparse configuration follows from eliminating radiators from a fully populated array, the antenna is referred to as a thinned array [1]. (Note that sparse arrays are also referred to as space tapered, random, non-uniform, aperiodic or arbitrary arrays [2], [3], [4].) The benefits of employing sparse array antennas are not restricted to cost and mutual coupling. Such aspects as: bandwidth, weight, power consumption, heat dissipation and (multi) functionality are also expected to improve when the distance between radiators is larger than Amin/2. The main drawback of these antenna configurations is their high level of the side-lobes in the radiation pattern, which is considered unacceptable in many applications. However, by controlling the number of elements, their positions and their relevant weights (both in amplitude and in phase) it becomes possible to ensure adequate RP properties for most practical applications. Many techniques have been proposed in the literature for the realisation of sparse array antennas. To assure a unified treatment of these methods a rigorous formulation of the array design problem is proposed in the following.
II. PROBLEM FORMULATION

Consider the configuration in Fig. 1 concerning a planar array antenna that consists of N elementary radiators distributed at known locations in the xOy plane. Let P be the observation point and let {ro, 0, O} be the coordinates in a spherical reference frame, with the elevation angle 0 being measured with respect to the Oz axis and the azimuth angle X being measured in the xOy plane, with respect to the Ox axis. The radiation properties of the array are determined by superposing the contributions of the primary radiators at different observation points, typically located on a (hemi)sphere [5, p. 249]. For ensuring a maximum flexibility, each element is allowed to have a different, known, spatial response (radiation pattern). Furthermore, the array is taken to transmit a narrow band signal, this justifying the single frequency analysis to be henceforth carried out.

2-9600551-7-9 (D 2006 EuMA

72

September 2006, Manchester UK

elementary radiators

nth radiator|
Fig. 1. Geometry of a generic, sparse array antenna.

can then be expressed as

With these choices, the field radiation pattern of the array

and as {Xn, Yn } (n = 1, ..., N) their locations in the plane, the entries in the vector v(0, o) read
g1(0, o) exp j

O ) = WTV(0, (1) f(0, where w is a vector of complex weights w, (n =1,.. , N) applied to each element and v(O, o) is a vector that is representative for the array's architecture, for the elements' radiation patterns and for the direction of observation, the superscript T indicating the transposition operator. By denoting as gn (n = 1 ... . N) the radiation patterns of the elements

xOy

XY

(k,x1 + kyyl)]
(kxX2
+

< )) v(,

92(0, o) exp

Ej

kyY2)]

imposed to the array, and Ci and C2 are two N-dimensional sets of complex numbers that define the possible values of the elements' positions and weights. The solution to this problem consists of the number of elements, their positions and the relevant weights that, together, provide the best approximation f(O, o) fd(0,0), V{O,o} C D. In most practical cases, an exact solution f(, ) = fd (0, o) to the hereby formulated problem does not exist. Consequently, attaining the design goal will call upon the solution of a constrained optimisation problem. For an array with a given number of elements, some representative formulations of this problem are as follows: . Determine the locations of the elements such that the side-lobes level of a narrow beamwidth radiation pattern is minimised. Note that this is the classical formulation referring to the design of sparse array antennas. . Determine the elements' weights for different objective functions such as maximum signal to noise ratio (SNR) or minimum approximation error with respect to the desired pattern. * Determine the locations of the elements and their weights in order to satisfy a twofold objective: the minimisation of the number of elements and the minimisation of the approximation error with respect to the desired pattern. The problem of finding the optimum positions of the radiators for a given aperture is often reduced to a thinning procedure. This allows for the reformulation of the optimisation problem as: Find the minimum number of elements from a fully populated array (with given dimensions) that, together, constitute an array meeting the radiation pattern specifications. This simplified problem has quite well known mathematical solutions, which estimate (truly, but with limited accuracy) the behaviour of the array in practical applications. A more accurate approximation of the performances can be obtained by adding supplementary constraints to the optimisation problem.
III. BEAMWIDTH Usually, the antenna beamwidth is the relevant parameter used for estimating the physical dimensions of antenna systems. This parameter can be easily measured from the antenna radiation pattern by selecting a certain threshold in the amplitude. However these patterns are not available to the

gN

(0, o)

exp

[j XN, +,Y (kxXN + kyYN)]

in which k = k sin(O) cos(o) and ky = k sin(O) sin(o), with k = 27/A denoting the propagation constant (the wavenumber). It is worth observing that in most references (see, for example, [5], [6]) the origin of the reference frame is chosen such that the quantities (xn + Yn ) / 2 + y2 with n =1, 2,... , N are approximately one. Finally, it is stressed that (2) also covers the case of the linear array configurations by taking, for example, Yn = 0 for n = 1, ..., N. By now slecting a desired field of view D (that is taken in this case as D = {{O,o} 0 <_ 0 <_ 7/2, 0 <_ < 27r), the design goal is to adjust f(0, o) such that it mimics a beforehand specified radiation pattern fd (0, so): minimise: F(w, v, 0) <0, subject to: H(w,}V, weC1a) <d 0C w C C, and v CE C2, f{,0o W} C

(4) where A denotes the wavelength at the operational frequency, N the number of elementary radiators and d the inter-element spacing. For d = A/2 spaced arrays, (4) can be expressed as
=

designer during the initial design phase. To circumvent this limitation, a good practice is to make use of the approximation that holds for (large) uniform, linear arrays and that states that the 3 dB beamwidth 03dB reads [5, pp. 260-261]: 2.782 A

Nd'

where, F is the objective function that measures the deviation between f and fd, H is representative for the constraints

(3)

03dB

1.771
N

(5)

that clearly demonstrates that for such arrays the beamwidth is inversely related to the number of their elements.

73

By turning now to the sparse arrays, one can assume that, in general, the antenna beamwidth equals the one of the fully populated configurations. This assumption is violated in the case of some design techniques (for example, by the amplitude tapering method [3]) when the specific deployment of the primary radiators results into a moderate enlargement of the antenna's main beam. Note that the algorithms used to shape the antenna pattern by placing the radiators at non-uniform locations are not covered by this study. The figures 2, 3 and 4 depict the radiation patterns of three planar, sparse array antennas assembled by means of the following techniques: combinatorial approach, fractal theory and space tapering method, respectively. By resorting to (5) for evaluating the approximate value of the beamwidth, one can easily notice that in the cases of the combinatorial and of the fractal solutions the beamwidth of the fully populated array is preserved. Nonetheless, the radiation pattern of the spatially tapered array exhibits a slightly enlarged beamwidth when compared to that of the fully populated configuration. IV. SIDE-LOBES LEVEL The side-lobes level is, possibly, the most frequently quoted parameter in the design and the analysis of sparse array antennas. To start with, in the case of uniform array antennas there are two distinct aspects related to this feature. One of these is the level of the first side-lobe, that is highly relevant for radar applications. The other one is the mean level in the side-lobes region, that has a large impact in the power budged analysis. Following a similar reasoning, sparse configurations are also characterised by two quantities, namely the peak side-lobes level (not necessarily occurring in the vicinity of the main lobe) and the mean value of the side-lobes level. From this perspective, the thinning techniques can either aim at the reduction of the maximum side-lobes level or attempt to minimise the mean value in the side-lobes region. The problem of finding an a priori approximation of the side-lobes level has been addressed in the literature from two perspectives. The first approach is strictly associated to a particular thinning algorithm. In fact, many of the authors that proposed such methods augmented their design strategies with (more or less) accurate side-lobes level estimators. However, the applicability of those estimators is confined to the sparse configurations obtained by using the relevant thinning procedure. Alternatively, [7] discusses a different, more general, approach that focuses on the study of arrays having randomly located elements. The approximations proposed in that work are formulated using the statistical theory and, therefore, they are applicable for estimating the performance of a large variety of sparse configurations. This conclusion relies on the fact that the statistical properties of the thinned arrays are not design method dependent. Using common sense and some straightforward physical arguments, a simple expression for the mean side-lobes level SLLpower was derived in [8] and it reads

with NS denoting the number of elementary antennas in the sparse configuration. This expression can be employed for providing an initial approximation of the side-lobes level in sparse array antennas. Obtaining more accurate solutions warrant the use of method related estimators but, evidently, this will require selecting a specific thinning algorithm.
V. GAIN To a large extend, the directivity of an antenna can be easily evaluated once the beamwidth and the side-lobes level were determined. During the design phase a common practice is to assume that there are no losses in the antenna system. With this condition, the gain becomes equivalent to the directivity and hence, closely related to the beamwidth and the side-lobes level. The antenna gain has a relevant energetic signification. Making use of this observation it can be stated that the gain of array antennas (either uniform of sparse) approximately equals the sum of the gains of the consisting elementary radiators. Assuming that the elementary antennas exhibit isotropic radiation patterns, the gain of the array can be taken to be directly proportional to the number of elements. VI. CONCLUSIONS An engineering point of view on the performance of sparse array antennas was formulated. Simple approximations were provided for the relevant parameters of these antennas. The beamwidth of a sparse configuration was shown to be comparable to that of the fully populated configuration having a similar size (and, correspondingly, a larger number of elements). The side-lobes level in non-uniform arrays was proven to be also comparable to that of the fully populated configurations having the same number of elementary radiators. Finally, it was stated that the gain can be taken in a first approximation to be proportional to the number of radiators deployed in the sparse architecture. Summarising, it can be affirmed that sparse configurations are similar to uniform arrays having a reduced number of elements and large inter-elements spacing, the energy of the grating lobes to be expected in such cases being uniformly spread out over the side-lobe region.
REFERENCES
[1] R. L. Haupt, "Thinned Arrays Using Genetic Algorithms," IEEE Transactions on Antennas and Propagation, vol. AP-42, no. 7, pp. 993-999, July 1994. [2] C.I. Coman, I.E. Lager, L.P. Ligthart, "Sparse Array Antennas", in Workshop EuMC WS 5. Multifunction Radar Front End Design and Management, pp. 19-22, 2003. [3] R. E. Willey, "Space Tapering of Linear and Planar Arrays" IEEE Transactions on Antennas and Propagation, vol. 10, no. 4, pp. 369377, July 1962. [4] R. Redlich, "Iterative least-squares synthesis of nonuniformly spaced linear arrays," IEEE Transactions on Antennas and Propagation, vol. AP-21, no. 1, pp. 106-108, January 1973. [5] C.A. Balanis, Antenna Theory: Analysis and Design, John Wiley & Sons Inc.,1997. [6] R. C. Hansen, Phased array antennas, John Wiley & Sons Inc.,1998. [7] B. D. Steinberg,"The Peak Sidelobe of the Phased Array Having Randomly Located Elements," in IEEE Transactions on Antennas and Propagation, vol. AP-20, no. 2, pp. 129-136, March 1972. [8] E. Brookner, Practical phased-array antenna systems, Artech House, Inc. ,1991.

SLLpower

Ns

(6)

74

55

45

40
35
.

3015~~~~~~~~ i h
20 1zX1=>X

toLsWMY

10

Fig. 2. factor.

Array factor pertaining to a thinned array assembled by using a combinatorial technique: (a) deployment of the elements; (b) corresponding array

~10~r~#:L~l~ >~:~ ~B<~L+IbI b M 1 m A1m~Iun


_._. 1.
III

-107

-o~

-20

-30j# 40
! -40

sH11

5l

20

30

40

50

60

70

Elevation angle 0 [degrees]

70)

-10 7

60~~~~~~~~~~~~~~~~~~2 -20
40

30

Ct~ ~ ~ ~ ~ ~ 3

-3075

20~~~~~~~~~~~~~~~~~~4
-50
20 40

60

80

-90

-60

a
Fig. 3. Array factor pertaining to
a

0 30 -30 Elevation angle 0 [degrees]

60

90

thinned array assembled by using

fractal technique: (a) deployment of the elements; (b) corresponding array factor.

0111111

2 111

2 111111

70

<4

1111

11111111111111

WX

0F

60tMlltxfFm um T r llfFmlfg rrr z larzXTr r rm:lam r rrmlr m : iX 40

l^ 1vSSXIlA 30-1 20 XSSy00a-<4;4t toI _ ^iIH1Xm

-o

Ct
ct

20

40

60

80

uU90

-60

a
Fig. 4. Array factor pertaining to
a

0 30 -30 Elevation angle 0 [degrees]

60

90

thinned array assembled by using

Taylor tapering: (a) deployment of the elements; (b) corresponding array factor.

75

Você também pode gostar