A transformation between the phasing techniques

required for linear and circular aerial arrays

D. E. N. Davies, M.Sc, Ph.D., Graduate I.E.E.

Synopsis
The paper.describes a method for producing an angular deflection, or continuous rotation, of any directional
pattern formed from a circular (or cylindrical) array, by means of a simple linear progressive phase shift
across the- input ports of a multiple-port matrix network, whose output ports feed the array elements.
If the progresive phase shift has a value of <f> radians between adjacent ports, the directional pattern is
rotated through an angle of <f> radians. The principal advantage of the technique is that it enables any
aperture distribution of a circular array to be electronically rotated by the simple phasing techniques
normally associated with a linear array. It is also shown that the matrix network may take the form of a
Butler-type matrix, which may be easily realised using only hybrid networks and phase shifters.

1 Introduction This paper describes circuits for feeding a circular array

Circular or cylindrical aerial arrays appear to possess which enable the beam to be electrically rotated by the use of
the attractive feature of being able to produce a narrow the simple phasing techniques normally associated with linear
directional beam, which can be electronically rotated without arrays.
change of beam shape, by means of either switching or
phase-shifting of the signals associated with each aerial
element. Despite this apparent advantage, little use appears 2 Electronic beam deflection
to have been made of such arrays. Experimental arrays with Any required aperture distribution round a circular
switching or electronic phase-shifting have been constructed,1 array may be considered as a continuous function of amplitude
but operational use of such arrays appears to have been and phase, which, owing to the adoption of a finite number of
mainly restricted to direction-finding applications, in which elements, may be approximated by a series of discrete values
only a small arc of the circle is used at any one time. The of both amplitude and phase: Axyx, /42y2> • • • ^nVn- The
Wullenweber array10 is one such example using a small arc process of beam deflection or beam rotation in the plane of
of a circle, but in principle this might more accurately be the circle is merely one of rotation of this aperture distribution
considered as a bent linear array, with appropriate phasing through appropriate angle. If we exclude the use of complex
used to compensate for the physical deformation. One dis- switching systems, the above rotation can be achieved elec-
advantage of circular arrays is that, once the directional tronically by the use of a phase shifter plus an envelope
pattern in the plane of the circle has been fixed, the pattern modulator (or amplifier with voltage-controlled gain) in the
in planes normal to the circle can only be altered to a limited feed to each aerial element. Since it is quite likely that ampli-
extent by controlling the directional patterns of the aerial tude tapers would involve considerable attenuation, or even
elements.5 This, however, is not an overwhelming disadvan- complete blanking of certain portions of the array, the
tage, and it appears that several of the problems preventing envelope modulator would have to be able to produce 100%
further exploitation of these arrays are associated with control of signal amplitude. Furthermore, the peak phase
problems of array-phasing techniques and pattern-synthesis deviation required from the phase modulator would involve
procedures. quite large deviations. For instance, many proposed aperture
There are two principal reasons why circular aerial arrays distributions for circular arrays involve cophasal beam disr
have not received much attention for possible applications tributions; i.e. aperture distributions in which the contribution
to communications, direction-finding or radar. These are: from each element adds in phase in the main beam. For such
arrays, the maximum phase deviation at each element required
(a) Until quite recently, very little work had been done on to rotate such a pattern is given by the diameter of the array
methods for synthetising attractive directional patterns, measured in wavelengths.
involving narrow beams and low side lobes, from circular To summarise, the rotation of the directional pattern of a
arrays. 1>2i3>4 In fact, simple synthesis techniques are still not circular array is complicated by the following three factors:
available.
(b) In order to utilise the symmetry of such arrays, it would (a) the necessity of providing 100% control of amplitude in
clearly be useful if the directional pattern could be electroni- each element feed
cally rotated. Methods have been proposed for achieving such (6) the need for control of phase shift over a large deviation
effects,1*5 but the methods concerned are, in general, far of phase
more complicated than the simple phasing required to elec- (c) the difference in amplitude and phase for most elements
tronically deflect the pattern of a linear array. Not only are of the array and the variation of these differences around
the phasing techniques for circular arrays more complicated, the array.
but, in addition, the use of amplitude tapers on such arrays
involves considerable further complication, since these also
must be electronically rotated with the directional pattern. 3 Properties of matrix networks
An unpublished report by Chadwick and Glass6
Paper 4834 E,firstreceived 21st May and in revised form 6th July 1965 describes a method for synthetising directional patterns from
Dr. Davies is with the Department of Electronic and Electrical
Engineering, University of Birmingham, Birmingham, War. a circular array in terms of the summation of cosine waves.
PROC. IEE, Vol. 112, No. 11, NOVEMBER 1965 2041
 It is shown that an approximation to each cosine component If input port 2 were energised, the outputs will be
can be generated by feeding an appropriate input port of a
multiple-port beam-forming matrix network with the output Eol = - e\pj(cot + a), Eo2 = - expy(co/ + 0 + a),
ports of the network connected to the elements of the circular n n
array. This proposal also affords the advantage of enabling the
synthetised pattern to be rotated by simple phasing of the Eo3 = -expj(ajt + a + 20). . .
n
inputs to the ports of the matrix. Phase shifts of the form
<f>, 2(f>, 30 . . . etc. radians result in an angular displacement Eon = - expy{a>/ + « + ( « - • (2)
of the directional pattern by <j> radians. It appears that the
disadvantage of the proposal concerns the imperfections in where B = — radians
the approximation of the cosine components of the directional n • (3)
pattern. However, this paper will show that the multiple-beam
matrix network may be used to simplify considerably the E
01 E 02 E 03 E 04
problem of electronic rotation of the pattern of a circular
array, and that this may be so, irrespective of the form of the
amplitude and phase distribution required round the array.
t t tt
The networks required for these applications are similar to
the multiple-beam networks used in conjunction with linear
arrays.
The use of multiple-beam matrix networks for linear arrays
has been well described.7-8 Such networks have n input and
n output ports, as shown in Fig. 1, and the best known type
is the Butler matrix, which is shown in schematic form in
Fig. 2. The operation may be briefly described as follows:
if a signal expyco/ is applied to input port 1, the output from
all the n output ports will be identical, and given by E
111 i
i1 E i2 E i3 E i4
r
Fig. 1
Eo\ = Eo2 = Eo3 = . . . Eon = - expj(u)t + a) . (1)
n Matrix network with n input and n output ports

hybrid ring

hr hr hr hr

E.i 8

Fig. 2
Schematic diagram of the Butler-type matrix used for the production of multiple beams from linear arrays
Hybrid rings may also be replaced by hybrid T or 3 dB couplers for wideband operation
2042 PROC. IEE, Vol. 112, No. 11, NOVEMBER 1965
Consequently, if the rth input port were energised, the outputs
Ei3 = — exp;(o>r + y2 - a)
would be n
Eol = - expj(wt + a), Eo2 = - expj(wt + a + rjS) A2
. (6)
n n n
Eo3 = - expj(wt + a + 2rjS). . .
n
Eon = - Gxpj{a>t + a + (n - l)/-jS} • (4)
n
It is therefore evident that, if such a network were connected
to an n-element linear array, the directional patterns asso-
ciated with the series of input ports to the network would
constitute a set of sin nx/s'm x-type patterns with differing
angular deflections, representing an orthogonal* set of
patterns in space.9
Since these networks may be constructed from linear
passive elements, reciprocity applies to the system, and
consequently, if all the input ports are fed in phase, the
output appears at output port 1 only; also, if, across the 12 3 4 5
input ports, there is set up a set of signals providing progres-
sively increasing phase shifts, the output will appear at a
different output port. In fact, the input and output ports may lilll E.
be considered to represent spatial samples of two functions
which are themselves a Fourier transform pair. It follows that, Fig. 3
for progressive phase shifts between multiples of j3, the output Matrix network with output ports feeding the elements of a circular
will not emerge from one port alone, but the principal output aerial array
will be from two ports, with the other ports representing a Circular array with n elements
sampled values of the side lobes of a sin nx/s'm x function. m Matrix network
The significant property of these networks is that a progressive
phase tilt across the input ports results in a transfer of output
energy from a given port to the adjacent one. The physical In a similar way, the required signals for element r may be
process of electronically deflecting the beam of a linear array obtained by feeding the input ports with signals
is one of producing a phase tilt across the aperture, but the
process of deflecting the beam of a circular array is to transfer En = ot + y r -
the signals feeding each element to the adjacent element. It n • « )

would therefore appear that it should be possible to use the

sample phasing of a linear array to cause beam rotation of En + y r -f (r - 1)
the pattern of a circular array, by using the properties of a
A
multiple-beam matrix. = f expy{o>/
Ein f (r - 1]<« " 1),
n
• • (7)
4 Use of matrix networks with circular Therefore, if it were required to feed all the elements in
arrays order to produce the required far-field pattern, the required
Fig. 3 shows an «-element circular array fed from an inputs to the matrix would be
H-input-, /i-output-port multiple-beam matrix. Suppose that,
for each element of the array, a suitable variation of ampli- En = yx - a) + A2 expj(ojt + y2 — a)
tude and phase has been computed to produce a required
far-field directional pattern. Let these amplitudes and phases
be given by expj(tot + y3 — a) • (8)
A
\Y\> Aiyi* Ai7i • • • An7n Ei2 = -{Ax expj(a)t + yt — a) + A2 expy(a>/ + y2 -a)
The required amplitude and phase for element 1, i.e. Alyl,
may be obtained by feeding each input port of the matrix f /43 txpjiuit + y 3 + 2j8 — a) + . • (9)
with an identical signal of the form
E expj(u)t + yi — a)
n = E/2 = Ei3 = • • • m = —
n
exp (<x)t + yx - a) . (5)

The required amplitude and phase for element 2 may be 2 expy{cu/ + y 2 + (n — l)jS — a}
obtained by feeding the input ports with signals 3 expy{co/ + y 3 + (n — I)2j3 — a} (10)

En = —? expy(o>/ + y2 - a), The feed for the input port r may be more conveniently
expressed as a summation

El2 = —2 expy(a>/ + y 2 - a), 1 m=n n

Eir = - 2 Am expj{a>t + ym + (r - l)(m - 1)£ - a}
• The orthogonality condition is a direct consequence of the choice of progressive
phase shift at the input ports denoted by eqn. 3 = Brexpj((*)t + ifjr) . . . (11)
PROC. IEE, Vol. 112, No. 11, NOVEMBER 1965 2043
It is quite evident that the result of the summation indicates The particular angular deflection covered by the above
the necessary amplitude and phase required at the inputs to equation, corresponding to the excitation of each element
the matrix, to yield the required far-field directional pattern. changing to that of its neighbour, represents an angular
There is no reason to suppose that these sets of signals denoted deflection of
by Bn tfjn for port r, are any easier to generate than the
original set denoted by Aryn specified at the elements of the (16)
array. However, the significant feature of the use of the matrix
is the simplification of the phase- and amplitude-modulation and.the incremental phase shift between ports required to
processes necessary to cause the directional pattern to rotate. produce this result is JS, which is given in eqn. 3 as
(17)
5 Electronic deflection of directional
pattern formed by matrix Therefore the resultant angular deflection is given directly
by the incremental phase shift between adjacent ports at the
To simplify the problem of electronic rotation of .the input to the matrix network.
directional pattern formed by the combination of matrix and It is therefore evident that, by the use of an appropriate
array, let us first return to the excitation required at the matrix multiple-beam matrix, it is possible to deflect any directional
inputs, given by eqn. 5, which energises only array element 1. pattern of a circular array through any angle, merely by the
If the excitation is modified to produce a progressive phase application of linear progressive phase shifts to the inputs of
shift of j8 = 2TT/« radians between elements, then, by the the matrix, as shown in Fig. 4. This process overcomes many
properties of a multiple-beam matrix, the excitation Aiy1 will
transfer from element 1 to element 2. The problem of elec-
tronic rotation of the directional pattern is therefore the • i n
application of this approach to the general distribution of
eqn. 11.
The extension of the above process to a general distribution
is quite simple, since the multiple-beam network is a linear
passive network, and therefore the superposition theorem may
be applied to synthetise the distribution of eqn. 11. The process
of rotating the complex aperture distribution is merely to
apply the appropriate phase progression to each component
of the input signals to the network associated with the feed
to each element of the array. If we insert the values of these Fig. 4
phase progressions in eqns. 8-10, we obtain Schematic diagram of arrangement for electronic angular deflection
of any directional pattern formed by a circular aerial array
a Phase shifts to deflect pattern through angle <t>
E
n = ~ + y{ - a) + A2 expj(ajt + y2 ~ a) b Values of amplitude and phase shift required to synthetise directional pattern
m Matrix network
A3 expy(cu/ + y 3 - a)} . . . (12)
of the problems associated with the electronic rotation of
patterns mentioned in Section 2. It will be noticed that the
E
i2 =
~niAi + y, + j8 - a) phase variation required to produce continuous rotation is
merely a uniform rate of change of phase (which may also be
A2 expj(a)t + y2 — a) produced by a frequency translation). A more important
73 + 3 0 - 0 0 . . . } . . (13) result is that there is no longer any need to use amplitude
1
modulation to rotate an amplitude taper. The taper Au A2,
expy{o>f + yx + (n - a} . . . An has been replaced by the taper Bx, B2, . . . Bn at the
input to the matrix, and this remains fixed, irrespective of the
+ A2 Qxpj{ojt + y2 + (n - 1)2)8 - a} direction of the main beam.
At this stage it is probably worth emphasising that the
+ A3 expj{cot + y3+(n- 1)3)8 - a} . (14) technique described is quite general and can be applied to
The value of the progressive phase shift chosen for these any synthetised pattern. However, it is only a scanning tech-
equations corresponds to that required for the excitation of nique, and, unlike the proposals of Chadwick and Glass,6
each element to change to that of its neighbour. it does not involve a synthesis procedure. Recent papers on
Inspection of eqns. 12-14 reveals that the additional phase circular arrays2-3>4 have gone a long way towards indicating
shifts that have been introduced are dependent on the input attractive forms of directional patterns that can be formed
port only, and consequently these phase, shifts may be taken from such arrays. For instance, throughout this paper it has
outside the square brackets, thus giving the input to the rth been assumed that electronic rotation of a directional pattern
port as merely requires a rotation of the element excitations. This is
only true for an infinite number of elements, since there is a
slight difference between the directional patterns when the
main beam passes through an element and when it is directed
between two elements. Reference 3 calculates the element
[k Amexpj{ojt + ym + (r - 0|8-«}l. (15) spacings required for this effect to be negligible.

It follows from the above that the rotation of the complex

aperture distribution (and consequent rotation of the corre- 7 Application of linear-array phasing
sponding directional pattern) may be achieved by the applica- techniques to circular arrays
tion of a linear progressive phase shift across the input ports It will be noticed that the form of phasing required
of the matrix network. at the input to the matrix is identical to that required for
2044 PROC. IEE, Vol. 112, No. 11, NOVEMBER 1965
electronic beam deflection of a linear array. However, the The principal application of the technique described in this
latter involves a pattern deflection on a sin 6 scale, compared paper would be to direction-finding or radar systems. The
with a uniform angular deflection for a circular array. It engineering practicability of the scheme is assured, since
follows that the proposed matrix network, together with the several electronic and mechanical phasing techniques for
set of networks required to produce the amplitude and phase- linear arrays have been used in operational systems for several
shift changes denoted by Brifjn may be considered to perform years, and multiple-beam matrix networks have been success-
a transformation from the phasing techniques necessary for fully built in several different frequency bands. However,
beam deflection in linear arrays to those in circular arrays. although some practical experience has been obtained relating
Furthermore, since the networks are both linear and passive, to problems of constructing multiple-beam matrix networks
the technique may be applied equally to both receiving and to given tolerance limits, there is, at present, little information
transmitting arrays. This means that the results of much of about the effects of phase and amplitude errors, mutual
the extensive work on electronic scanning techniques and coupling between elements, and shadowing of elements, upon
multiple beam-forming techniques, for linear arrays, may be the directional pattern of circular arrays. These studies are
applied directly to circular arrays. An interesting example of consequently included in a current research programme on
the application of linear-array techniques to a circular array circular arrays in progress at the Univerysity of Birmingham.
would be to consider a 'frequency-scanned' circular array, as An important limitation on the proposed scheme relates
shown in Fig. 5. The inputs to the matrix are fed from tapping to the increase of complication with the number of elements
in the array. The Butler matrix in Fig. 2 requires 12 hybrid
units for feeding an 8-element array. If we take the more
i 3i 4•
2
f
n
general case of a network for an w-element array (where
n = 2s) using the Butler-type circuitry, then the number of
• m
hybrid units required is given by
1J L 1 ±ns, or %n log2 n, where 5 is an integer
B
*2% 3'

I j d
I The consequence of this result is that the complication of the
matrix network increases faster than the number of elements
in the array. Multiple-beam matrix networks may also.be
synthetised using tapped delay lines," and this approach
Fig. 5 involves n2 tappings or directional couplers.
Schematic diagram of an arrangement for frequency scanning of the It can be concluded that the proposed technique would
directional pattern of a circular array become very complicated for circular arrays involving several
d Tapped delay line hundred elements, but it would be quite feasible for arrays with
/ Variable-frequency transmitting oscillator
m Wide-bandwidth matrix network
less than about 100 elements. If we take a typical element
spacing as one half wavelength, this number of elements
corresponds to a diameter of about 16. wavelengths, and
points on a delay line which is fed from a variable-frequency arrays of this order of size would probably have application,
source. In this case, angular deflection of the direction of either to metre-wave radar (where very large arrays are not
transmission of the main beam will be directly proportional to practical), direction-finding systems, or electronic-beam
the change of transmitter frequency, so that one frequency stabilisation for satellite-mounted arrays.
sweep of the transmitter may be made to correspond to any
required scan angle, including, if required, several rotations.
However, it is also necessary to design the multiple-beam 9 Acknowledgments
matrix to maintain its required performance over the total The author would like to thank Mr. E. D. R. Shearman
transmitted bandwidth. In practice, this is not difficult, since and Mr. M. J. Withers for helpful criticism of this paper.
bandwidths of the order of 10% of the carrier have been
achieved, and much wider bands could be obtained by the use
of appropriate time delays in place of phase shifts in the matrix
network, together with coaxial-line or stripline techniques. 10 References
1 NEFF, H. p., and TILLMAN, J. D. : 'An electronically scanned circular
antenna array', IRE Internat. Convention Record, 1960, 8, Pt. 1
2 JAMES, p. w.: 'Polar patterns of phase corrected circular arrays'.
RAE Technical Note RAD 872, June 1964
8 Conclusions 3 FENBY, R. p . : 'Limitations on directional patterns of phase com-
pensated circular arrays', Radio Electronic Engr. (to be published)
The paper describes a technique for producing a 4 ZIEHM, G. : 'Optimum directional pattern synthesis of circular
deflection or continuous rotation of the directional pattern of arrays', ibid., 1964, 28, p. 341
5 DAVIES, D. E. N., and MCCARTNEY, B. s.: 'Cylindrical arrays with
a circular array, by means of the phasing system normally electronic beam scanning', Proc. IEE, 1965, 112, (3), p. 497
associated with linear arrays. This results in a simplification 6 CHADwiCK, c , and GLASS, J.: 'Investigation of a multiple beam
scanning circular array'. Radiation Systems Inc. Report, Alexan-
of the complex phase-shifting circuits required to produce dria, V., USA
electronic rotation of the directional pattern of a circular 7 SHELTON, J. p., and KELLCHER, K. s.: 'Multiple-beams from linear
array;' and removes the need for modulator circuits for arrays', IRE Trans., 1961, AP-9, p. 154
8 BUTLER, j . , and LOWE, R.: 'Beam forming matrix simplifies design
producing electronic rotation of amplitude tapers, which are of electronically scanned antennas', Electronics Design, 1961, 9,
often necessary for the synthesis of low-side-lobe directional p. 170
9 ALLEN, j . L. : 'Theoretical limitation on the formation of lossless
patterns. It follows that several techniques for either electronic multiple beams in linear arrays', IRE Trans., 1961, AP-9, p. 350
beam scanning or multiple beam formation, developed for 10 CRAMPTON, c.: 'Naval radio direction-finding', / . IEE, 1947, 94,
linear arrays, may be applied directly to circular arrays. Pt. IIIA, p. 132
11 OGG, F. c.: 'Steerable array radars', IRE Trans., 1961, M1L-5, p. 80

PROC. IEE, Vol. 112, No. 11, NOVEMBER 1965 2045