Vous êtes sur la page 1sur 34

THE INSTITUTION OF ELECTRICAL ENGINEERS

PHASED ARRAY FUNDAMENTALS:



M S SMITH

STC TECHNOLOGY

.----- ----------1

PHASED ARRA Y FUNDAMENTALS

MS Smith

STC Technology Ltd, Harlow, Essex

Coruems
1. Introduction
2. Radiating apertures
2.1 Linear apertures
2.2 Planar apertures
3. Linear array analysis
3.1 Introduction
3.2 Radiation patterns
3.3 Beam steering
3.4 Nulls 4. Array realisation
4.1 Array elements
4.2 Beamforming
4.3 Systematic and random error effects
4.4 Calibration and BITE
4.5 Limited sector scanning
5. Planar arrays
5.1 Planar array analaysis
5.2 Beamforming
5.3 Mutual coupling
6. Active arrays - a brief introduction
6.1 Transmit/receive modules
6.2 I.F. beamforming
6.3 Baseband beamforming 1/1

_C"<J

\ r ~ x (x , 0). ~ ,)< ~ ( j k x; ~L'L<:x I d x .

'\ J_«>

(::2. i )

1. Introduction

Phased arrays are a major class of antennas which offer many important features such as electronic beam scanning. Their flexibility has to be assessed against their complexity and cost. In this lecture the fundamental principles of a phased array are described, to form a basis for the later lectures on applications.

The properties of radiating apertures are given first, as these are basic to both aperture antennas and arrays as sampled apertures. Linear array analysis then introduces beam steering and null steering. Array beamfonning shows how array excitation is controlled. This is followed by discussion of the effects of systematic or random excitation errors.

Planar array analysis includes different array grid structures and the important topic of mutual coupling. Finally active arrays are briefly introduced.

2.1 Linear apertures

The basic result of aperture theory for linear apertures is that the radiation pattern P(sin..q is the Fourier transfonn of the aperture distribution E (x,O). Mathematically

x

The calculation of radiation patterns for some standard aperture distributions will both illustrate the technique and derive several very important antenna properties. The fIrst case is a unifonn distribution, that is

- o../;;_ L.."X -< 0../ ';L \-;c.\ ). :J.../';L .

Then the radiation pattern is

,...-t-O/!;l.

j Eo . -0'.:f' Uk -oc S) d.x

. -o../;I._

.\

Figure l(a) shows peS). The radiation pattern peak is at S = 0, nonnal to the aperture, and the main beam extends from S =-;~ to s = +\8. The first sidelobe peak is at S = ±i~ and zeros of

1/2

the pattern occur at S = ±rjt Negative values for P(S) indicate a phase difference of 1800 relative to the main beam.

Figure l(b) shows the same pattern expressed on a dB scale, as 20 10glOi iP(s)II/Pim4 The flISt (and largest) sidelobe level is at -13.2 dB relative to the main beam peak. The two main features of a directive radiation pattern are the beamwidth (usually at the -3 dB level) and the sidelobe levels. Here the zero to zero beamwidth is 2'A/a on a sin« scale, while the 3 dB beamwidth is - 0.9>-/a.

The above case illustrates one very imponant antenna property, namely the inverse relationship between aperture size in wavelengths and the beamwidth. Thus narrow beams require large apertures.

A second example is a half-cosine distribution:

~ Eo - ( ~~<) _ .:].. ~ ~ tx: S CLI")..
CL'':;t __
~: • .;C (x,6) co .a.
1 zc. I Ct_/ ';l_
I ,-.. »
~.-'
, The radiation pattem is:

~_'X~ (j r: ~ let) -!- eJxp (_ j rrx/ a) Jx. :2.

- EcCl...

(after some algebra). This radiation pattern is plotted in figure 2 (linear field scale). The flISt zero of the pattern is not where kaS/2 =Tf/2, as both numerator and denominator become zero simultaneously. The zero-to-zero beamwidth is now 3'!1a on a sin scale, 50 per cent larger than for a uniform aperture distribution. The 3 dB beamwidth is also increased correspondingly. The sidelobe levels have decreased significantly - the rust is now -23 dB (compared with -13 dB).

1/3

This second case illustrates another important antenna property; using a tapered aperture distribution (an 'amplitude taper) tends to reduce sidelobe levels (thete are limits), at the expense of an increase in main beam width. Low sidelobes are often a very desirable property in various applications.

Table 1 lists four standard aperture distributions and the corresponding radiation patterns.

Table 1 First
sidelobe
Aperture distribution Radiation pattern (normalised) level
I. fix) = 1 sin(ka 512) -13 dB
]x; < al2 PIS) = (ka 5/2)
2. fix) = cos (1Tx.'a) cos(ka 5/2) -23 dB
'xI < al2 PIS) = 1 - (ka 5/-rr)'
3. fix) = COs:(1Tx/a) sin(ka 512) I -32 dB
[r] < al2 P(S) = (ka 5/2) I - (ka S/21T)'
~. fix) = 1 - 2ixl/a ein(ka 514)}, -26 dB
,xI < al2 PIS) =
(ka 514) As a further example, consider a general amplitude distribution A(x). The corresponding radiation pattern is

?(s) =

00

+ f MX)""'r (jkxs) do<.

-..0

(~.b)

Now 'add' a linear phase variation across the aperture - that is, the aperture distribution becomes A(x) exp ~). Then the new radiation pattern is

pies) ,::

,,-

-+ J A(X) ~'~f (.j (Il 5 + f):C)

-_

=

(~.7)

The new pattem therefore has an identical shape to the original one (on a sinCt(scale) but is shifted by ~(sin o() - -f"'- (This is an application of the Fourier shift theorem). If the

1/4

1 -

£.

fust pattern has its main beam peak at S = 0, the second pattern has its peak where S = fk. As an example, if P= f' the beam maximum is steered from c/ = 0 tQ;(= _300.

Because the pattern has the same shape on a sin:( scale, its beamwidth is constant on this scale. This is not constant on an.xscale. Let the beamwidth be 2(\" on the sin« scale, and let the beam steering angle be "0' Then the beamwidth on aru(scale is

'XI - oX2_.

where

Assume that£" is small (a narrow beam), and let 0< I ~ cX.~ + f. Then

- J( ....

Then o=Etdi~d the beamwidth on f1f!j!JCale becomes

The beamwidth therefore increases as the steering angle increases. Physically this corresponds to the reduction in projected aperture as the beam is steered (a smaller aperture giving a wider beam).

The principle of beam steering by applying a phase gradient to a radiating aperture is very important. as it allows electronic control of beam direction.

2.2 Planar apertures

We now consider the three-dimensional generalisations of the formulae for two dimensions. There the fields had no j dependence, and an aperture of width a. was infinitely long in the y direction. We now have an aperture such as shown in figure 3, and Cartesian coordinates x, y, z, or spherical polar coordinates I, S, ¢ , may be appropriate. As in the two-dimensional case, there are two independent polarisations, and we shall give results for the ~case (E = 0 in

y

the aperture).

1/5

The two-dimensional and three-dimensional radiation patterns of linear and planar apertures are

pes) z:

..... Otj

-+- j Ex (x)~,) ~xF (j ~xs) Jx

A ~

-sa

p (5" ,,) z: :' l~f""E"x (x, IJ ,0) £xp (j~ (x ~f + .:35";2.)).Jx cL.J. (~.<~)

, v'

-ue -_.:)

where 5, = 5\:.\. (; W!.?I S!=5I1OGSU.fusing conventional spherical polar coordinates as in figure 3.

In the far field, in two-dimensions and three-dimensions respectively:

(tUc)

E (A-,&)¢) = A ~,;x.f(-j(~Y--ITI;1»)(c:o<,:¢~&-~U(L~ca€~¢)

r-

The dependence on r in the three-dimensional case has now changed to that of a spherical, rather than cylindrical, wave. The nf2 phase shift is half the rr shift through a spherical focus. The factor (..-~e-·:!;!~efines the polarisation of the wave.

The evaluation of radiation patterns for some important cases can now be made. consider a unifonnly illuminated rectangular aperture, as in figure 3, such that

Then

(Some simple aperture distributions are separable - that is, Ex (x, y, 0) = El (x). E2(y), which allows the x and y integrations to be separated as in this case.)

H

I

?=

and the pattern in the x-z plane is

Also,if

.-;:: - r .

-I - .. ~;'

and the pattern in the y-z plane is

>tA ( -¥- S~~&)

kb SI~V'~ 6 2

(:::. I b)

These patterns are of the same fonn as for a unifonn aperture in two dimensions.

Note that the first zero in the x-z plane is where sin e- = ±>ja. and the first zero in the y-z plane is where sin ~ = ± >-jb. The beamwidths in the two principal planes are separately controlled by the aperture dimensions in each plane.

As a second example (a hom antenna), let

{ 6=:0 coos ('rr~~ / b ) () ) (j'th.c_rt-~-Ue.

I x I ..:. ,1./::z! I V I L oj-;_

(~. (1)

The radiation pattern is then:

1/7

This is again separable, with the two integrals similar to the two dimensional examples given earlier. Thence

[- ~ \'J

-0 .,.

9:"'_ (~fA ,; \ (2::)

bO~'I/2.

rr/:_ (<o:;(\-:d'~7/L) .(:..1~) -r:'-:/4- - R?-b '-S;j 4-

In the x-% plane (S2 = 0), the first zero of the pattern is where sine= ~/a, whereas in the y-z plane (S 1 = 0), the first zero is where sin e = 3/2 f.-/b. The sidelobes in the y-z plane are at a lower level than in the x-z plane.

A third example is a unifonn illumination over a circular aperture with diameter D. The radiation pattern in any plane is a Bessel function, shown in figure 4. The [lISt sidelobe level is -17.6 dB, and the 3 dB beamwidth is 1.02,VD. The (noanalised) pattern function is

(where D = 2a). IT an amplitude taper is superimposed on this field distribution, the beamwidth increases and the sidelobes are (generally) reduced, just as for a linear aperture.

The directional characteristics of an antenna are frequently expressed in tenns of a gain function 0(&, t/i), whose maximwn value is the gain. (Here.Q andfare spherical polar angular coordinates.) This is now discussed for the transmit case, for convenience. The gain is

. defined as the ratio of the maximum radiation intensity from the antenna to the maximum from a reference antenna with the same input power. The reference is usually a hypotheticallossless isotropic radiator, and the gain can then be expressed in dBi. At a distance r from an

isotropic source, the power transmitted is evenly spread over a spherical surface of area 4.:/.

For an antenna with gain, it follows that the power P incident on an area A at a distance r in the direction of most intense radiation is

p

where P t is the transmitter power, and G is now used for the gain, the maximwn value of G (-6; rp). Directivity, D, is defined as the ratio of the maximwn radiation intensity to the average radiation intensity (averaged over all angles). For an antenna which is 100 per cent efficient

1/8

and has no copper, dielectric or mismatch loss, directivity and gain are the same. For an antenna with losses, G will be less than D by a factor corresponding to the efficiency, :

G=7·D.

Equation (2.21) can be rearranged as

G =

L..,...,..'!.(P!A) PT

(~), :n)

Consider the unifonnly illuminated rectangular aperture defined by (2.12). Its radiation pattern P(Sl' S2)' is given by 2.(l3). The far field, E(r,9,¢), is found from (2.11). The

maximum value of E for a given r is found where SI = S2 = 0; this value is

The power flux density is

lEI :l

=

The power transmitted, P T' is found by assuming that E/H = Zo in the aperture - this is a fair . assumption if a, b »A. Then

Substituting (2.25) and (2.26) into (2.23), the gain of a unifonnly illuminated rectangular aperture is

c;.. :::

1/9

----T

r

where A is the area of the aperture. This result is in fact true for any shape of aperture, provided the illwnination is uniform.

A second example is the calculation of the gain of a hom antenna. The maximum value of IE I is now found from (2.19) and (2.11).

I ~I =

(~.:;.:s)

The power flux density is now

"'/' ,

r H

=

The power transmitted, again using the assumption E/H = Zo' is

.~ r-~h

;'T:CC ""\ ~:'" . A . J c.0!' ~ (TI ~ h,) :kzr

.,_ ~o -L/2

The gain now becomes

__ . ~1To-b '6 t z: .31;
b'" ~
\'- -rr2 ' . Compared with the uniformly illuminated aperture case, the maximum power flux density is multiplied by 4;#; however the power transmitted' is multiplied by 1/2 so the net gain is 8/; times that for the uniform aperture. The effective area (c.f. equation (2.27) is therefore - 80 per cent of the geometrical area for a hom antenna. The 'aperture efficiency' of aperture type antennas is defined as the ratio of the effective area to the geometrical area. A typical figure for a front-fed parabolic reflector antenna is about SS per cent. Aperture efficiency should not be confused with antenna efficiency, (equation (2.22).

1110

, 0

4.

'"

I.'

I. (a) Uniform aperture radiation pattern. (b) Radiation pattern in decibels

x

"

y

Ibl

3. (a) Uniform rectangular aperture. (b) Coordinate system

'.0

cos u

flul", _

1 -1121"UI2

~1J •• .s{),

;2.. Half-cosine aperture radiation pattern

SIM'-

4- Uniform circular aperture radiation pattern (lOA diameter)

1/11

(3.1 )

3. war array analYsis

3.1 Introduction

In the previous section the radiation pattern of an aperture antenna was given as the spatial Fourier transfonn of the aperture distribution. Antenna radiation patterns can therefore, in principle, be synthesised by control of the aperture excitation. IT the aperture excitation can also be varied electronically, an adaptable radiation pattem is available. Electronic beam steering or null steering (to minimise interference) could therefore be provided. There are limitations on beamwidth imposed by the aperture size, as the Fourier transfonn relationship implies. In practice, it is very difficult to implement control of the excitation of a continous apenure. An array antenna allows far more control by using a number of individual antenna elements grouped together to fonn a sampled aperture. The array elements themselves may not be very much more expensive than a continuous aperture antenna (such as a reflector), but in general the control devices are a significant cost penalty. This has to be traded off against the various advantages of an array antenna system.

Because an array is a sampled aperture, there are limitations of the spacing of the may elements. A typical spacing is of the order A/2 (this will be explained below), and so fairly small antennas are required as elements. Dipoles, slots, patches and open-ended waveguides can all be used.

The tlieory given in this section will be for a linear array of identical, equispaced elements. This is the simplest case, and is also commonly used in practice. The effects of mutual coupling between elements will be ignored in this section. It will be assumed that individual phase and amplitude control of each element is available. There are various ways of implementing this control, and some of these will be explained in section 4.

3.2 The radiation Pattern of a linear array

Figure 5 shows a linear array of equispaeed antenna elements. Firstly we shall consider the radiation pattern of an array of (hypothetical) isotropic radiators, F(9). This pattem is called the 'array factor'. Note that the angle ~ is now defined relative to the array axis, rather than the normal to the aperture, as in section 2. This is because the array line is an axis of symmetry for the array factor, which is then a function of ~ only.

The array in figure 5 contains N elements with spacing d. In the direction ~ the path from the nth element to a distant point is ndeos ~ shorter than that from the zeroth (reference) element to the same distant point. IT the nth element is excited with amplitude a and phase ¢ ,the

n n

contribution from that element to the field at a great distance at the angle ~ is proportional to

The total field is then

1/12

~ 0'0 + r.1.., z .._ .::t;tZ-

+ c.J..N_1 2.

(~, ' ) \. ."0'.4-

N -I

F = 2. i\,~ -tJ)(I' (j y, kd ': cS & )

n. -= 0

where a is a complex number (= la I exp(j ¢). A useful method of analysing linear arrays, due

n Inn

to Schelkunoff, consists of associating a polynomial with an array. Let

'Z =

Then (3.2) becomes

rJ-1

The radiation pattern (array factor) is F(z) = F (exp(jkd cos -&». The complex coefficients of the polynomial F(z) are the excitations of the individual elements in the array. In the general case, the form (3.2) is often used directly to evaluate F(-&).

As an example, consider an array where all the elements are excited with unit amplitude and zero phase. This is the sampled equivalent of a unifonnly excited linear aperture. The radiation pattern for N elements is

Fe z.) =

z.

+ 2. + Z ---

"'-1 +z

~r>.,)(.p (J ~ do. N ctls8) «(-i , s)

..t_)(P (jRdc.o~G)

~_tr\.. \F(Z.)!2. = 2 _ ~ c,-,s(rJk!u;5,;b) := s0","-(-:k Nkjc-o~f,)

a - ~ UD~ (J~d. c.c6&) sv,"a: (1:- IH;t: L.c~G)

\ Fez; \ == ~.~Vo_ ( 1: N k.c4 cose) / ~:"'. ( t kd r..cS~) (? . f:, )

Figure 6 shows the function sin Nx/N sin x (with N = 10). This is a normalised form (peak value of unity) for (3.6) if

x ==

rr d CC''C f:.\ ».

(3./)

The result (3.6) can be compared with the radiation pattern of a uniform aperture of length a:

$A~VL (~..i. ka. W. o()

J_ ~C\. S L"'\. ,.,(_ ?

The numerator is exactly the same if a = Nd, noting that sind= cos &. but are similar if

S,~f\. ( t ld (.4)5 b)

The denominators differ,

1/13

N-I

;- Ct.7) ¢). L Ct". J:'l<.f (jl" kJ co~e).

("3. I~ .. )

that is

/ ~ ,.,)

\. ., ..

(They also differ by a factor N, but that is simply a constant of proportionality.) If d = ).,/2, this condition required cos ~« 1, so that the patterns are similar near broadside to the aperture or array. The zero-to-zero beamwidth is the same for the two cases. In general, the pattern is similar to that of a diffraction grating in optics, and the denominator can cause 'grating lobes'. The condition for these is

1... 1.) ,.0<>6 = J,.. TI
-, -
/'"
Ccs & -= 4- Aid (;.IC) At such angles, the denominator in (3) becomes zero, simultaneous with the numerator, and a repeated main lobe occurs in the array factor radiation pattern. Referring to figure 6, the fust zeros of the pattern occur where x =1ftN, that is cos & (i sino9 = 4Nd), while the first grating lobes occur where x =11;' that is, cos ~= f}/d. The main beam widtli is thus governed by the complete array length. while the grating lobes are governed by the spacing of the array elements. The relation (3.10) implies that. if grating lobes are to be avoided, d < A is required. When beam steering is considered, further restrictions on d occur.

Now consider non-isotropic elements. The radiation pattem now becomes ",-I

~ ~". f", lb,)2")' "''.Y.p (j¥'kd. cc<;e)

iI. =- C

Here f .... (&,~) is the radiation pattern of the nth array element. If the array elements have identical patterns, and similar alignment. the element pattern can be factored out, so that

VI:. 0

This is simply the product of the element pattern and the array factor. The array factor is generally more directive than the element pattern, so that it dominates properties such as main beamwidth and sidelobe levels. The element pattem will generally reduce far out sidelobe levels, including any grating lobes wbich may be present.

1/14

F.(Z) =

N-1

2~

a" z", where z = exp (j kd cos 9)

(3.13)

3.3 Beam steering

Consider a general array excitation a" (n The corresponding radiation pattern is

0,1, ... N-l).

n=o

Now apply a linear phase gradient to the array - that is, let the array excitations become

ao, at5q,), a2 exp(j2q,) ,etc.

Let

bn = a" exp(jnq,)

(3.14)

Then the new radiation pattern is

exp (jnq,) zn

(3.15)

where

z '

exp(j(kdcos 9 + 41»

(3.16)

Then, if

kd cos I1b + 41 = kd cos 9.:

Fb (cos 6b) = Fa (cos 6a)

The radiation pattern keeps the same form but with an angular shift related to the incremental phase 41. From (3.17):

(3/17)

cos 9~ - cos 6. = -q,/kd

(3.18 )

The right hand side of (3.18) is proportional to the phase gradient (dq,/dx) along the array.

If the main beam peak is initially at 11 = 90·, broadside to the array, then cos 6. = 0, and cos 9b = sin a, where a is the beam deflection angle.

1/15

------ -------

cos 8 - cos 8m = ± ~/d

(3.24)

Then

sin a = -q,/kd

(3.19)

If d

~/2, and a is small so that sin a ~ a, then

a (radians) ~ -q,(radians)/~

a· ~ -4>. /~ (3.20)

Thus, for an incremental phase value of 30·, the deflection angle 1S ~ 10·. using (3.19), an incremental phase of 90· for ~/2 spaced elements gives a deflection angle of 30'.

The use of electronically controlled phase shifters on each element of th~ 'phased' array allows the beam-pointing direction of an array antenna to be varied rapidly and without any mechanical movement of the antenna. This is in contrast to a fixed beam antenna, such as a simple horn-fed parabolic reflector, where the antenna is rotated mechanically to 'scan' the beam. However, the latter is in general much cheaperl

The question of grating lobes arose earlier where d <~ was required to avoid them, for a beam with a broadside peak. The radiation pattern of a uniformly excited N element array is

IF.(Z) I = sinnN kd cos 8.) sin (~ kd cos /I.)

Now consider this array with a phase gradient superimposed. Using equation (3.17):

IFb(z) I = sin(~N(kd cos /lb + 41»

sinO (kd cos 9b+5Il»

The peak of this radiation pattern is at 8

(3.21)

/1m, where

cos /1m "" -q,/kd

Equation (3.21) can therefore be written as

(3.22 )

I F.(Z) I

sin(~N kd(cos /I - cos 8m»

sin (~kd(cos 8 - cos /I,.»

(3.23)

Grating lobes will appear if the denominator of (3.23) becomes zero for some /I ~ /1m, This will occur if

by comparison with equation (3.10).

1/16

z = exp (jkd cos 0)

(3.29)

Let the deflection angle from broadside (6= 90·) be a, so

that

9 = 90' + a

(3.25)

Equation (3.24) then becomes sin ~ - sin a = ± Aid

(3.26)

The extreme values of sin a are ±l. If we consider am >0, the magnitude of the left hand side of (3.26) has a maximum value of

sin am + 1

(3.27 )

Thus, if

Aid > 1 + sin am

dl A < 11 (1 + s in am)

(3.28)

grating lobes will not appear. For example, if the maximum beam steering angle is 30· from broadside, (3.28) requires d<2/3A. If the beam can be steered by 90·, to 'end fire', d < ~ A is needed to avoid grating lobes.

3 • 4 ID.!.ll.§

A null of the radiation pattern is here taken to be synonymous with a zero - that is, no transmission or reception at a particular angle. A null can also mean a minimum of the pattern, but in the present analysis exact zeros can be achieved. The analysis and synthesis of nulls can be achieved using Schelkunoff's representation.

Consider the complex variable:

As 9 varies from 0 to ~, z moves along a locus in the Argand diagram, as shown in figure 7. The modulus of z is 1, while its phase (argument) is

2~d cos 0 A

which therefore varies between -2~d/A and +2~d/A. The locus of z is therefore an arc of the unit circle in the Argand diagram. For a large value of d, z may move several times round the unit circle as 0 varies from 0 to~. If d = A/2, the locus of z as o varies from 0 to ~ just close to form~ complete circle. If d « A/2 the locus of z is a small arc of~unit circle.

1/17

Now consider the polynomial representation (3.4). Any polynomial can be expressed as a product of linear factors. The polynomial (3.4) will have N-l roots Zl' Z2 •••• ZK-l' and it can be written:

Since only the relative radiation pattern is important, we can set a.l = 1 and then

F(z) = (z - zd (z - Z2) ••• (z - ZM-l) (3.30)

The roots Zi' zz, etc. (of F(z) = 0) are complex numbers, but do not necessarily lie on the unit circle. If they do lie on the unit circle, and d is such that that part of the unit circle is traversed by z as 9 varies from 0 to ", a null will be present in the radiation pattern. As there are N-l roots for an' N element array, there can be up to N-l independent nulls. An example is now used to illustrate null synthesis.

A four-element array is to provide a radiation pattern with nulls at 9 = 45·, 90· and 135·. Then

Zl exp(jkd cos 91) = exp(jkd/j2» Zz = 1, Z3 = exp(-jkdjj2)

and

F(z) = (z - exp(jkdjj2» (z-I) (z - ..(.'I'-f' C -jkd/J£))

Z3 - z2(1 +2 cos(kd/j2» + z(l + 2 cos (kd/j2» - 1 (3.31)

The array coefficients are therefore:

a, = 1, a2 = - (1 + 2 cos (kd/ k> )

a1 = (1 + 2 cos(kd/j2», ao = -1

The coefficients a1 vary with the spacing d, but the nulls always exist: the synthesis starts from 9 values and so ensures that the Zi are traversed by the locus of z. The coefficients are all real, and this is due to the symmetry of the radiation pattern about broadside. An array distribution with real coefficients always gives a radiation pattern whose amplitude (modulus) is' symmetrical about broadside.

1/18

n

N- 1

\

\

\

\ \

o

nd

5 Linear array of equispaced antenna dements

10

:t I

1:

-06

-0.4

6. The function sin(Nx)INsin x (with N = 10)

Imlll

',. ... /2

7. Argand diagram for z = exp(jkd cos 6)

1/19

1/20

4. ARRAY REALISATION

4.1 Array Elements

Many different types of radiating elements are used in phased arrays due to a variety of different requirements placed upon them. Some of the more important of these requirements include the following:

a) Operating frequency

b) Polarisation

c) Bandwidth

d) Scan angle

e) Power handling

f) Constructional compatibility

In a large phased array it is possible to deduce the ideal radiation pattern for each element of the array. Thus, for an array of N elements which yields an overall gain of G at broadside, the broadside gain of each element should be GIN. Also the power radiation pattern should follow a cosine law over the required scan angles of the array and be zero outside this range. This ideal pattern cannot be realised in practice and can only be approximated to with varying degrees of success with different types of element.

It is also important to distinguish between two different element radiation patterns. These are the pattern of an isolated element and the pattern of the same element when it is incorporated into its array environment. It is the second of these two patterns that is relevant to the present discussion. The two patterns differ primarily due to the effects of mutual coupling, with differences being influenced by the type and spacing in wavelengths of the array lattice. Mutual coupling is an array feature that cannot be avoided and has to be allowed for in the construction of most arrays.

significant radiating elements which have been used in phased arrays include:

a) Open ended waveguides (rectangular and circular)

b) Dipoles

c) slots

d) Various printed elements

The properties of the first three element types have been extensively investigated and have been well characterised for planar array geometries. Their behaviour in non-planar array configurations (i.e. in a conformal application) is not so well known. The fourth category of radiating element has only recently been investigated and consequently is not as well characterised as the first three types. Further effort is needed to establish the optimum design of this group of elements where the effect of the substrate needs to be taken into account.

1/21

Two element types of this class have received particular attention recently, they are the printed dipole and the microstrip patch. So far it is not clear whether these two types will satisfy all possible requirements or whether there is scope for new types of radiating elements.

Circular or dual polarisation is required for

a) b) and c)

Satellite communications Dual polarised radar Conformal array design.

In these cases the choice of possible radiating element is limited considerably. One solution is to use pairs of orthogonal elements (e.g. dipoles), others require elements with two feed points, one for each polarisation. Dual polarisation also creates extra complexity by either requiring separate feeds for each polarisation or additional control of the radiating element. satisfactory solutions to this situation are not readily available and further work is needed in this area, particularly for printed elements.

A variety of other elements types have also been used to construct phased arrays, though their use has been restricted in the main to specialised applications. For instance the following element types have received limited application:

a) Helical radiators

b) Polyrod radiators

c) Yagi radiators

d) Log periodic radiators

e) Vivaldi radiators

4.2. RF Beamforming

A standard phased array consists of a number of antenna elements distributed in space, for example, equispaced along a straight line - a linear array. A single beam is formed by feeding the array elements with the amplitudes and phases needed to create the desired beam shape. The 'beam forming' techniques described in this section operate directly at the radio frequency (rf). Both intermediate frequency (if) and baseband beam forming can also be used (particularly on reception).

The desired amplitude and phase distribution can be formed using a fixed or variable feed network. Figure Sea) shows a standard configuration for a linear array with a 'corporate' or 'parallel' amplitude distribution network, and a phase shifter for each array element. The power divisions at the junctions may be fixed or variable - the former is usual. In that case, a fixed amplitude distribution results, with corresponding main beam width and sidelobe levels. The phase shifters can then be used to vary the array phase distribution - for example, to steer the beam electronically. The phase shifters are normally composed of a series of discrete valued switched elements (for

----~-------~---~-- --- ----

(4.2)

example, using PIN diodes), although continuously variable units also exist.

Figure 8(b) shows a 'series' distribution network, where power is progressively coupled out of a feed line into the array elements to give the amplitude distribution. The rf path length between successive elements is one wavelength, Ao, at a particular frequency, fa (or A~2 plus a phase-reversed connection). At that frequency, the array elements are co-phased and a broadside beam is produced. At another frequency f, wavelength A, there is a phase difference ~ between elements, given by

~ = 271' ( ~ - 1) . .\

(4.1)

For a given array spacing d, the beam is then deflected from broadside by an angle a; using equation (3.19):

Depending on the application, this effect can either be a benefit or a drawback. A series feed is generally cheaper and requires less space than a parallel feed. 'Frequency scanning', as this effect is called, can be used to steer a beam using a simple (series) feed and no phase shifters. On the other hand, when the communication direction is fixed, the effect can limit the use of a series feed. The deflection angle a is independent of the array length, from equation (4.2) ; however, the beamwidth reduces as the array is lengthened, so that the deflection becomes a larger fraction of a beamwidth. If the beam is correctly aligned for f., it will be misaligned at other frequencies. Once the misalignment is half a beamwidth, the signal level will be down by 3 dB. Clearly this is a significant limitation on either the array length (and hence the beamwidth and ga. in) or the system bandwidth. A numerical example can be used to illustrate this effect.

Consider an antenna array which has to provide a specific narrow beamwidth. Say that at fo(Ao), the beam points correctly, and this is broadside to the array. Then at f = ~± ~f/2, the beam is deflected from broadside by a small angle a, where

I a I ::= (L _ 1 ) • ~ =

x, d

Let d be Ao/2 and the beamwidth be 2·. Then, if a = ±l·, there will be a 3 dB reduction in signal at the band edges. Then

fJ.f/ fo ::= 71'/180 ::= 1. 8 per cent

Another type of amplitude distribution network is shown in figure 8 (c). This is an equal path length, series feed. This does not exhibit frequency scanning, but is less compact than a standard series feed as in figure 8 (b).

1/22

The ease with which both sum and difference beams for tracking applications can be produced in a phased array depends on the beamforming technique which is used. In the case of digital or IF beamforming ( ~6 below) good performance can be obtained for both the sum and difference beams since the signal to noise ratio is established before beamforming takes place. However, in the case of RF beamforming, the most common and simplest way of achieving the sum and difference patterns is to split the array in half (in the plane of the difference pattern) and feed the two halves through a hybrid junction. Unfortunately the resulting difference pattern has high sidelobes due to the large discontinuity in the aperture distribution at the centre of the array. A number of techniques exist for improving the difference pattern sidelobes and these involve some increase in design complexity together with a marginal increase in cost.

4.3 systematic and random error effects

In order to steer a beam to a desired angle, a linear phase gradient is required across the array. However, with a digital phase shifter with n bits, the phase of one element can only be set to discrete values, for example 0, 45', 90· ..... 315· with a 3 bit phase shifter. This means that phase errors are introduced for most scan angles and in a large array, the errors are periodic across the array. This systematic error effect produces additional sidelobes at particular angles (related to the spatial period), called quantisation lobes. The peak quantisation lobe level is approximately -6n dB. For low sidelobe arrays, this is a significant limitation. The quantisation lobe level can be reduced to approximately - 12n dB by deliberately introducing small random phases at each element. (This does raise the average far out sidelobe level).

In many applications one of the most difficult tasks will be to maintain the tolerances required to meet low, wide angle sidelobe specifications, particularly over a wide range of scan angles, and it is likely that this will require the incorporation of a good calibration system. The exact relationship between sidelobe levels and tolerance is somewhat complex but if the ampli tude and phase errors in the aperture are assumed to be random then the average sidelobe level can be calculated from a knowledge of the desired aperture distribution and the rms errors. Typical examples are shown in figure 9 for a uniformly excited array. These curves provide a useful guide to the permitted level of errors for a given sidelobe specification, however it should be borne in mind that the errors in a real array will have a systematic (as well as a random) component which will raise the peak sidelobes in some angular regions (depending on the correlation interval of the errors) well above the mean level. It can be observed from figure 9 that the sidelobe levels decrease with increasing numbers of elements. Thus it may be easier to achieve a given sidelobe level with a two dimensional array than with a one dimensional array, provided that the errors are random.

1/23

T

Increasingly, the reliability and cost of ownership of modern radar systems is becoming as important as their performance. In mechanically scanned systems employing a single transmitter, a single scanner and a single antenna, failure of anyone of these components is generally catastrophic and can lead to expensive repairs. On the other hand, electronically scanned antennas with their many hundreds, or even thousands, of elements will generally have a graceful degradation of performance with component failure.

In an electronically scanned antenna employing a reasonable number of elements, failure of a single element will have a barely noticeable effect on performance. As the number of element failures increases then there will be a gradual degradation of performance and it is not until the number of failures produces an unacceptable degradation in performance that the equipment needs to be withdrawn from service and repaired. In addition, if a good calibration and BITE facility is incorporated, then component failures can be detected and steps can be taken to minimise their effect. For example, if an element has failed in such a way that it is radiating with an uncontrolled phase it may be better to swi tch the element off completely. The use of adaptive techniques could also be used in order to optimise the performance from the remaining working elements.

4.4 Calibration and BITE

In order to obtain good performance from an array it is necessary to maintains tight control of all phase and amplitude errors as discussed in the previous section. Many potential sources of error exist within an array such as element to element variations, mechanical· assembly inaccuracies, mutual coupling effects, etc. The effects of many of these errors can be reduced or eliminated by the use of appropriate calibration techniques which, in essence, require the measurement of phase and amplitude through the various paths of the array.

In all military systems and many commercial systems it is necessary to incorporate some form of built in test equipment (BITE). In the transmit mode of the array then, ideally, the provision of BITE requires the ability to monitor the signal being transmitted by each element of the array and in the receive mode it is necessary to be able to inject a signal at some point close to the face of the array and to monitor the signal through the receive system and into the processor.

Calibration and BITE therefore require similar processes and can utilise a common network. If each element of the array has the ability to control its amplitude, as well as its phase, then calibration and BITE can be carried out using the mutual coupling which exists between elements of the array.

Calibration can also be carried out by illuminating the array with a plane wave and monitoring the signal received by each element. This will permit initial setting of the phase and

1/24

1/25

amplitude controls within the array but will not permit in service calibration which may be required due, for example, to changes in temperature.

It is apparent from the above that the technique to be used for calibration and BITE must be determined at an early stage and incorporated in the initial design of the array.

4.5 Limited Sector Scanning

So far we have considered array antennas which are capable of beam scanning over a wide angular sector. If a very narrow, high gain scanned beam is required, a very large phased array could be needed. High gain beams can be readily provided by reflector antennas, but then require mechanical scanning. Figure 10 shows a compromise arrangement for 'limited sector scanning'. A small phased array with a wide range of scan angles is combined with an imaging dual reflector to provide a high gain beam scanned over a limited angular sector. The sub-reflector and the main reflector have a common focus. When the array is used to transmit, the sub-reflector aperture distribution is provided by the small array. The imaging arrangement then produces the main reflector aperture distribution. This is a large aperture, and hence a narrow beam is radiated. The phase gradient across the larger aperture is reduced in proportion to the change in aperture size. The scan angle from the main aperture is therefore reduced compared with the scan provided by the small array. The number of beamwidths of scan is maintained; if the main reflector's effective aperture is H times larger than the array aperture, the beamwidth is reduced by a factor of H, but so is the angle of scan.

Antenna elements

A.ntenna elements

Yfllli

'- __ .r-'- _ ___l Beam con

Antenna elements

'CI

l? (a) Corporate (parallel) feed network. (b) Series feed network. (c) Equal path length. series feed

U'

, ..

" ...

~~~.I -i- __ ---

UI'

" 0'

'lD

"0

I''''

tu .. un

- s· , 0'

."

ID

"

10, Limited sector scanning

q. Effect of Tolerances on Array Sidelobes

1/26

CPx = -k d,.510, CPy = -k dy 520

in order to steer the beam to (80, CPo).

(5.2)

5. PLANAR ARRAY5

5.1 Planar Array Analysis

as

Just~a linear array can be considered as a sampled aperture in two dimensions, a planar array can be thought of as a sampled aperture in three dimensions. The analysis in section 2.2 of planar apertures is therefore appropriate to derive many of the properties of planar arrays,provided they are at least several wavelengths across and are composed of elements which radiate predominantly into one half-space, such as open-ended waveguides, or dipoles in front of a planar reflecting sheet. Typical aperture shapes are rectangular, elliptical or circular. Figure 11 shows two standard array 'lattices': (a) a rectangular grid and (b) a triangular grid.

Beamwidths and first sidelobe levels can generally be found using continuous aperture theory (section 202). The 'far out' sidelobe levels need array calculations. Beam steering to a direction (510, 520) can be accomplished using the appropriate phase distribution. The direction cosines 51 and 52 are

51 = sin 8 cos cP, 52 = sin 9 sin cP

(5.1)

where 8 and cP are the usual spherical polar coordinates with respect to z as the normal to the array aperture. Let the element lattice spacings in the x and y directions be d,. and dy (for either type of lattice in figure 11). Then the interelement phase shifts in the x and y directions need to be

As the beam is scanned, its beamwidth changes as the inverse of the projected aperture. For a rectangular aperture, with an aperture distribution separable in x and y, and scanned in one of the two principal planes (x-z or y-z), the beamwidth varies approximately as l/cos a, as described at the end of section 2.1. For scanning in intermediate planes, or for other aperture shapes, the calculation is more complicated.

In section 3, the conditions for the appearance of grating lobes were analysed for a linear array. For a planar array, there are different results for a rectangular lattice and a triangular lattice (see figure 11). Consider a rectangular lattice with inter-element spacings d,. and dy. The array factor for an unsteered beam is

1/27

P(Sl' 52) = If the beam is now

2~ 2~ aim exp{jk (1 d,

;_ (YI

steered to (510, 520):

(5.3) (5.4)

The main lobe is where s, appear if

S10' S2 = S20' A grating lobe will

211'n

or

(5.5)

where nand p are negative or positive integers. The grating lobe conditions are thus:

or (5.6)

82 - 820 = L·P

e,

plane

A rectangular lattice scanned in a principalLbehavessimilarly to a linear array. For arbitrary scan angles, the diagram shown as figure 12~)can be used. This shows the 'inverse lattice' for the rectangular grid array, with lattice points spaced by A/~ and A/dy. The inverse lattice coordinates are 81 - 510 and 82 - S20' The inverse lattice points correspond to the main beam and ·grating lobes.

For a broadside beam, S10 = 820 = 0, and the coordina~es become 81, 82, For arbitrary observation angles 8, ¢, S12 +~<l, and a circle of unit radius encompasses all possible cases. If this circle does not include any inverse lattice points (other than 0,0), grating lobes will not appear. Then the condition for no grating lobes becomes

d, < x and dy < >.

(5.7)

for an unsteered beam.

Now let the main beam be scanned to a direction (S10' S2~' The unit circle is centred where 81 = 0, 82 = 0, which is now at coordinates (-S10' - S20) in figure 12[..:1. If the circle now includes inverse lattice points other than (0, 0), grating lobes will occur. If the scan angle from broadside is 60, the centre of the

1/28

d < 2A/[j(3){1 + sin Ao}]

(5.10)

circle is displaced from (0,0) by a radial distance sin B 0 (=j (S102 + S2/»' Now if the beam can be steered anywhere within a cone about broadside with maximum scan angle Bo, a circle centred at (0,0) with radius 1 + sin 80 encompasses all the possible cases. The condition for no grating lobes is then

d, < Vel + sin 80) and ely < A/el + sin Bo)

(5.8)

Now consider a special case of figure 11 'b), where the triangles are made equilateral, with nearest neighbour spacing d. This has the inverse lattice shown in figure l2'~. The 'x' spacing in the inverse lattice is 2 A/d, because the lines of elements are d/2 apart in the x direction. Similarly, the 'y' spacing is 2 A/j(3)d, because the lines of elements are spaced by j(3)d/2 in the y direction. The other points in the inverse lattice follow from symmetry considerations. For the unscanned case, S10 = S20 = 0, a circle of unit radius is used, and the condition for no grating lobes is

d < 2 Vj3

(5.9)

For scanning within a cone of semi-angle 60, the condition becomes

A s~are lattice with spacing d has an area per element, A.1, of d ,whereas an equilateral triangle lattice with spacing d has an area per element of [J (3) /2 ] d2 • I f we consider the condi tions for no grating lobes, without scanning in the two cases:

(square ) lattice

(triangular) lattice

(5.11) [2/j(3) p2

For scannin~ out to Bo, both areas are reduced by the same factor (1 + sin 80), The equilateral triangular lattice thus allows a greater area per element than a square lattice, for grating lobe free performance. Fewer array elements, 86.6 percent, can then be used.

1/29

5.2 Beamforming

A planar array can be fed in various ways. Figure 13 shows a corporate (parallel) amplitude distribution network for a planar array. This could be used with phase shifters at each element to provide a steerable single beam. This approach has several drawbacks, notably the high cost of phase shifters, and an interconnection problem. As well as connections to the array elements, the phase shifters need control lines, and a thin planar structure for the complete antenna assembly is difficult to achieve.

A planar array is often fed with separable 'beam forming' in two orthogonal planes. Here 'beam forming' means either providing the amplitude and phase distribution required for a single beam or forming simultaneous multiple beams. As an example, phase shift scanning in one plane can be combined with frequency scanning in the other. The corporate feed and phase shifters are then similar to those for a linear array of n elements, for a planar array with N = n x m elements. There would then be n series-fed arrays of m elements.

A second example is an array where each row has a fixed corporate feed, and the rows are combined using a simultaneous multiple beam former. Such an array can be scanned by mechanical rotation in azimuth, while covering a wide elevation sector with simultaneous high gain beams.

5.3 Mutual coupling

In the analysis of linear and planar arrays, it has been indicated that the radiation pattern of the array is the product of the array factor and the element pattern. Two caveats are needed, Firstly, the element pattern is not in general that of the element in isolation, due to the effects of its neighbours. If one element is fed, its neighb~urs are excited due to mutual coupling, hence giving a different radiation pattern to that of the element in isolation. secondly, there are 'edge' effects, particularly in a small array, because elements near the edges of the array do not see the same distribution of neighbours as near the centre, and therefore the effects of mutual coupling differ.

If we ignore edge effects, the array radiation pattern can be written as the product of the array factor and the 'active element pattern'. The active element pattern can be measured wi th that element excited and all other elements present and terminated in matched loads. Let the array factor be N h (51' 52), where the peak value of h is unity (and N is the number of the elements). The array gain pattern is then:

(5.12)

where g. is the active element gain pattern.

1/30

---~-----~ -----

Active element patterns can exhibit several notable features, (i) flattening of the 'beam' peak, and (ii) deep dips in the pattern. Neither of these effects are present in the isolated element pattern.

Consider what happens when the whole array is excited, with the main beam scanned to the angle of a deep dip. The elements will be unable to radiate any significant fraction of the available power. There must therefore be a severe mismatch, resulting in most of the incident power being reflected. Such a scan direction is called a 'blind spot'. Clearly it is important to design a phased array such that blind spots only occur outside the desired range of scan angles.

Mutual coupling analysis is complicated: either an elementby-element approach, or a receive analysis using the periodic properties of an infinite array, may be used for certain standard configurations. Experimental measurement of active element patterns' may be required. How significant the effects are depends on the type of element, the array lattice type, the array spacings and which pattern cut/plane of scan is being considered.

Provided that blind spots, if they occur, are outside the required range of scan angles, array analysis ignoring mutual coupling can make reasonable predictions. The array factor tends to dominate the radiation pattern and gain, so that the precise pattern of the element becomes less important. Clearly there are potential problems for wide angle scanning, or for very low sidelobe systems where precise predictions are needed. Feed network design may have to cope with significant mismatch reflections.

1/31

0 0 0 0 0 0
0 0 0 0
0 0 0 0 0 0 0
0 0 0
d
x
a, ! 0 0 0 0 0 0 0 0 0 0
0,\/ 0 0 0
0 0 0 0 0 0
Elements d,
lal (b)
II. (a) Rectangular lattice. (b) Triangular lattice " "
Aid,! " )( " x 2Ax
,,(3k1
" "
x 2Ald
x
" ..
" "
- Ibl
Aid, (al

12.. (a) Inverse lattice for a rectangular grid. (b} Inverse lattice for an equilateral triangle grid

) [

1 ~

13, Corporate (parallel) feed for a planar array

1/32

In order to use an i.f. beam forming scheme, it is to equip each array element with a complete receiver (mixer, i.f. amplifier and local oscillator drive). these front ends will be identical, especially as tracking in phase and amplitude.

necessary front end Ideally,

regards

6. ACTIVE ARRAYS - A BRIEF INTRODUCTION

6.1 Transmit/receive modules

A single scanning beam for a planar array can be provided using r. f. phase shifters at each array element. These are usually digitally controlled and can scan the beam in two planes. The array amplitude distribution is usually fixed, with a network distributing the power from a single source. Al ternati vely, individual elements can have complete integrated transmit and receive modules. The main problem with these approaches is high cost, as the number of phase shifters or modules is equal to the number of array elements.

Figure 14 shows a block diagram of a transmit/receive module.

6.2 I.f. Beamforming

The fractional bandwidth at the i.f. is, of course, much higher than the fractional r.f. bandwidth. This can lead to significant difficulties if wide bandwidths are necessary.

6.3 Baseband Beamforming

A fully digital baseband beam former has separate I and Q digital processing for each element in the phased array aperture. Analogue baseband beamforming is also possible. In a system where the number of elements is high, this has obvious problems of cost and complexity. In practice, the full advantages offered by an optimum performance system r·,O¥ not be required, and a hybrid approach with baseband beamforming confined to one dimension, may be advantageous.

The biggest advantage of digital baseband beamforming is the flexibility it offers in beam pattern control.

FROM RADAR .:

EXCITER

TOANTfSNA RADIATING ELEMENT

TRANSMIT AMPLIFIER

1/33