Antenna Arrays Page 1

Antenna Arrays
Recall from our previous discussion of the 2-element array that the total field pattern from an array
could be found by multiplying the element factor (the pattern produced by a single element) by
the array factor. We found that that array factor represents the response of an array of isotropic
elements, allowing us to treat the element and the array separately.
We now consider more general examples of arrays. First, we will consider arrays with any number
of elements. Then, we will consider the case where the elements are not necessarily excited by
the same signal, but excited by signals with different amplitudes and/or phases.

1 Array Factor for N Elements

Consider a single isotropic radiator that has a radiated field that is proportional to
ejkr
. (1)
4r
Hence, the radiation intensity associated with this is constant (isotropic). We can write the
unnormalized array factor of such a situation as
AF = 1 (2)

Now consider the array shown in Figure 1, which is receiving a signal from a plane wave incident
at angle to the plane of the array.
Each element is excited with a signal at an amplitude of 1, but because the transmission paths
between elements are not equal, the phase shift of each element will be different. Hence, we can
write the array factor as
AF = ej0 + ej1 + ej2 + + ejN 1 (3)
where m are the phases of an incoming plane wave at the element locations m = 0, 1, . . .,
referenced to some point such as the origin. Hence, the phase of the wave arriving at element m
leads the phase of the wave arriving at the origin by m .
Now we consider the case where all the array elements are separated by the same distance d,
leading to a linear array of total length D = (N 1)d. Such an array is called an equally spaced
linear array (ESLA), and since the excitation is uniform, we called it a uniformly excited ESLA.
The geometry of the array then resembles the figure shown in Figure 1.
From the figure, we can see that the phase of element m + 1 leads that phase of element m by
kd cos , since the path length to element m + 1 is d cos metres longer than that to m. If we
arbitrarily set the reference point to element 0, so that 0 = 0, we can write the array factor1 as
AF = 1 + ejkd cos + ejk2d cos + ejk(N 1)d cos (4)
N 1 N 1
X X d
= ejkmd cos = ejkm N 1 cos (5)
m=0 m=0
1
Note the Balmain notes use indices starting at 1, so m = n 1

Prof. Sean Victor Hum ECE422: Radio and Microwave Wireless Systems
Antenna Arrays Page 2

Figure 1: Ray diagram for linear receiving array

Figure 2: Linear array geometry

Prof. Sean Victor Hum ECE422: Radio and Microwave Wireless Systems
Antenna Arrays Page 3

Defining
= kd cos , (6)
the expression for array factor becomes
N
X 1
AF = ejm = 1 + ej + ej2 + + ej(N 1) (7)
m=0

This function is a function of and resembles a Fourier Series where the array factor is composed
of a set of sinusoids at multiples of a fundamental frequency . More on this later.
Note that because of reciprocity, the array works similarly in transmit mode except the direction
of the phase gradient is reversed to produce a plane wave leaving the array in the direction shown.

2 Plotting the Array Factor

It is not obvious what the radiation pattern produced by the array factor looks like by examining
Equation (7). Here we will present a simple graphical procedure for plotting the array factor.
Lets multiply Equation (7) by ej to obtain

AF ej = ej + ej2 + ej3 + + ejN (8)

Subtracting Equation (8) from Equation (7) results in

AF (1 ej ) = 1 ejN . (9)

Rearranging,

1 ejN
AF = (10)
1 ej
ejN /2 ejN /2 ejN /2
= (11)
ej/2 ej/2 ej/2
sin(N /2)
= ej(N 1)/2 (12)
sin(/2)

The maximum value of AF occurs when = 0, resulting in AF = N . Therefore, disregarding

the phase factor and normalizing we obtain

sin(N /2)
f () = (13)
N sin(/2)

To plot the array factor, we note that Equation (6) defines the polar equation of a circle, and is
used to relate to . Lets take an example for the 2-element case we have already discussed
earlier, and set the array spacing to d = /2.
To plot the radiation pattern pattern as a function of , we plot |f ()| and a circle of radius
= kd = 2
2
= below the plot, as shown in the illustration in Figure 2. Then, by sweeping
Prof. Sean Victor Hum ECE422: Radio and Microwave Wireless Systems
Antenna Arrays Page 4

from to , we trace out the projected array pattern inside the circle as follows. For a given
angle , determine the point of intersection of a radial line from the origin with the perimeter of
the circle. Then draw a vertical line up from this point and determine the value of f () at this
point. Since the circle has radius 1, which is the maximum value of |f ()|, the projection of the
point back inside the circle is at a distance |f ()| from the origin. This is repeated for as many
points as necessary to construct the radiation pattern.

Figure 3: Geometric array factor construction

The region 1 < cos < 1 or kd < kd cos < kd, or the horizontal extent of the circle, is
known as the visible region, since |f ()| is only evaluated for values in this region.
Evaluating |f ()| for various values of N yields the curves shown in Figure 2. The exact shape of
the radiation pattern depends on the size of the visible region, but we can may several important
observations about this plot.

1. There is always a maximum at = 0, corresponding to = 90 , which is called the

broadside direction as it is normal to the plane of the array.

2. As N increases, the width of the main lobe decreases. In fact, the first null beamwidth
can be found by considering where the numerator of Equation (13) go to zero, or where

Prof. Sean Victor Hum ECE422: Radio and Microwave Wireless Systems
Antenna Arrays Page 5

1
N=3
N=5
0.9 N = 10

0.8

0.7

0.6
f()

0.5

0.4

0.3

0.2

0.1

0
4 3 2 1 0 1 2 3 4

Figure 4: Plots of |f ()| for various N

N F N /2 = F N = 2/N . Since = kd cos ,

2 2
= d cos F N (14)
N  
1
F N = cos (15)
Nd
The first null beamwidth is then found as
F N BW = |F N,lef t F N,right | (16)
    
 1 1 
= cos
 cos (17)
Nd Nd 
  
  2
 + = (18)
2 Nd 2 Nd  Nd
where the approximation holds for long arrays, L = N d >> .
3. The number of sidelobes increases as N is increased. In one period of |f ()| there are
N 2 sidelobes.
4. The width of these minor lobes (in terms of ) is 2/N . The width of the major lobe is
twice that.
5. The sidelobe level (SLL) decreases with N .
6. |f ()| is symmetric about = 0 for = . . . .
Prof. Sean Victor Hum ECE422: Radio and Microwave Wireless Systems
Antenna Arrays Page 6

Example: N = 2, d = .
Lets try plotting the 2-element pattern using the technique just described. Note kd = 2.

0.8

0.6
f()

0.4

0.2

0
2 1 0 1 2
/

90 1
120 60

150 0.5 30

180 0

210 330

240 300
270

Figure 5: Pattern construction for N = 2, d =

Note that with this technique we plot what looks like half the pattern from = 0 . . . .
Technically, these are the bounds of theta, and the plot has been produced in a plane = 0 in
front of the array. To plot the response behind the array, we would need to consider a plane
= 180 , but since we are considering isotropic radiators here, obviously the pattern will be
symmetric. Hence, the 3-D pattern shown can be found by revolving the 2-D pattern about the
z-axis.
From the resulting pattern we can see we have a combination of a broadside and a so-called
end-fire pattern. End-fire patterns produce radiation along the axis of the array, which contrasts
to normal to the axis for broadside patterns. Usually, we want a broadside pattern, or and end-fire
pattern, or something in between, but not usually both. In fact, the end-fire lobes here are not
desired and are called grating lobes. If you were to compute the directivity of such an array, it
would not be very high since the pattern is quite unfocused.

Example: N = 5, d = 2 .
Here, we have a strongly directive, broadside pattern: at = 90 , |AF | = 5 which improves our
signal strength by a factor of D = 5. The sidelobes are very low. This pattern is very useful for
focusing energy to/from a certain direction, and is called a pencil beam because of the shape of
the pattern.
Prof. Sean Victor Hum ECE422: Radio and Microwave Wireless Systems
Antenna Arrays Page 7

0.8

0.6

f()
0.4

0.2

0
1 0.5 0 0.5 1
/

90 1
120 60

150 0.5 30

180 0

210 330

240 300
270

Figure 6: Pattern construction for N = 5, d = /2

2.1 Directivity of a Broadside Linear Array

The maximum directivity (which occurs in the broadside direction) of a uniformly excited ESLA
can be found as follows. First, the normalizedarray factor is

sin( N2 kd cos )
(AF )n = . (19)
N sin( 12 kd cos )

If d  , we can employ the small angle approximation for the denominator, yielding

sin( N2 kd cos )
(AF )n N
. (20)
2
kd cos

The normalized radiation intensity produced by this array factor is

" #2
N
sin( kd cos )
U () = [(AF )n ]2 = N
2
. (21)
2
kd cos

The average radiation intensity is found using

" #2
Prad 1 2
sin( N2 kd cos )
U0 = = N
sin dd, (22)
4 4 0 0 2
kd cos

Prof. Sean Victor Hum ECE422: Radio and Microwave Wireless Systems
Antenna Arrays Page 8

which can be evaluated by making the subsitution

N N 2
Z= kd cos dZ = kd sin d sin d = dZ. (23)
2 2 N kd
This yields
N kd/2  2
1 2 sin Z
U0 = dZ. (24)
2 N kd N kd/2 Z
If the array is very long (N kd very large), then we can approximate U0 as
 2
1 2 sin Z
U0 dZ , (25)
2 N kd Z
| {z }

since the integrand tends to zero for large N kd values. The integral evaluates to as shown, so

U0 , (26)
N kd
and
Umax 1 N kd Nd
D= = = =2 . (27)
U0 U0
This result assumes the array length L = N d is very long (L = N d >> ). Note that for a
special case of half-wavelength spacing,
D=N (28)
and the directivity is simply equal to the number of elements in the array. Remember that this is
the result for an array of isotropic radiatiors (i.e., the array factor), and that the incorporation of
real elements (via pattern multiplication) will increase the overall directivity of the array.

3 Different Array Excitations

One of the biggest advantages of antenna arrays as that they allow many different array patterns
to be synthesized. We have considered only one case so far, where each element is excited with
an identical signal. Obviously, this does not necessarily need to be the case. Though we wont
this situations much in this course, it is good to be aware of it and how it affects the radiation
pattern.
We can introduce arbitrary element excitation by re-writing the array factor expression as

AF = I0 + I1 ejkd cos + I2 ejk2d cos + IN 1 ejk(N 1)d cos (29)

N
X 1
= Im ejkmd cos (30)
m=0

Prof. Sean Victor Hum ECE422: Radio and Microwave Wireless Systems
Antenna Arrays Page 9

Linear Phase Progression Between Elements

One very simple and useful case is where the magnitudes of all the element signals are the same
(|I0 | = |I1 | = |I2 | = = |IN 1 |), which we will arbitrarily set to 1, but there is a progressive
phase shift between elements. Hence, we write the element excitation as

Im = ejm (31)

Hence,
N
X 1
AF = ejkmd cos +jm (32)
m=0

and redefining = kd cos + , Pwe can plot the array pattern using the graphical technique
N 1 jm
we studied already since AF = m=0 e . We plot f () for = 0, and to account for the
inter-element phase shift, the circle = kd cos is shifted from the origin by an amount . This
is shown in Figure 3 for the 2-element d = /2 case discussed earlier.

Figure 7: Geometric AF construction technique for 6= 0

The effect of this linear phase gradient is to steer the beam away from broadside (which is when
= 0 ). As an example, lets consider the 5 element case we discussed earlier, for various values
of . Figure 3 show the visible region of |f ()| with the pattern drawn below.
Prof. Sean Victor Hum ECE422: Radio and Microwave Wireless Systems
Antenna Arrays Page 10

1 1 1

0.8 0.8 0.8

0.6 0.6 0.6

f()

f()

f()
0.4 0.4 0.4

0.2 0.2 0.2

0 0 0
0.5 0 0.5 1 0.5 0 0.5 1 0.5 0 0.5 1 1.5
/ / /

90 1 90 1 90 1
120 60 120 60 120 60

150 0.5 30 150 0.5 30 150 0.5 30

180 0 180 0 180 0

210 330 210 330 210 330

240 300 240 300 240 300

270 270 270
= 30 = 60 = 90
1

0.8

0.6
f()

0.4

0.2

0
0 0.5 1 1.5 2
/

90 1
120 60

150 0.5 30

180 0

210 330

240 300
270
= 180

Figure 8: AF constructions for various values of

Prof. Sean Victor Hum ECE422: Radio and Microwave Wireless Systems
Antenna Arrays Page 11

Clearly the scanning action of the array is observable. In fact, if = 180 , the beam resembles
an end-fire pattern with beams at = 0 , 180 . This is highly useful behaviour because the beam
of the array can be scanned without mechanically turning the antenna array; instead, electronic
phase shifters can be introduced into the element paths and used to produce the phase gradient
across the array to point the beam in the desired direction.

Amplitude Weighting
In addition introducing phase shift between elements to scan the beam, different amplitudes can be
applied to the elements as well. We will not discuss this case in detail in this course. The graphical
technique cannot be used in this case and the AF must be plotted directly using Equation (30),
noting that
Im = Am ejm (33)
represents the element excitation if the linear phase gradient across the array is preserved.
The most common technique is to introduce an amplitude taper across the array to help reduce
sidelobe levels, at the expense of a small reduction in the overall array directivity. Some examples
of amplitude tapers are shown in Figure 9, and the corresponding impact on the array factor
shown in Figure 10. Notice how all forms of tapering reduce the sidelobe level as at the expense
of increasing the main lobe beamwidth (or decreasing the directivity of the array). That is, the
best possible directivity of a broadside array occurs when there is no tapering across the elements
(a). A simple triangular taper (b) reduces the sidelobes significantly, with the first sidelobe still
being the strongest. Sidelobes can be eliminated altogether using a Binomial distribution (c) at
the expense of a very wide main lobe. Finally, all the sidelobes can be set to be the same level
in the case of a Dolph-Chebyshev array, for example, -20 dB relative to the main lobe (d) or -30
dB (e). While the synthesis of such arrays is beyond the scope of this course, we see that any
reduction in the effective usage of the array results in lower directivities but better sidelobe levels.

4 Generalized Array Factor

So far, we have restricted our discussion to linear arrays, and in that analysis by itself, we have
been assuming uniform spacing between elements. The more general form of the array factor
expression is found using the following expression:
N 1
0
X
AF (, ) = Im ej(krrm ) , (34)
M =1

0
where ~rm is a position vector to the mth element, and r is a unit vector pointing in the direction
of interest, i.e.,
r = sin cos x
+ sin sin y + cos z. (35)
Im is the complex amplitude of the excitation of the mth element.
We note that for a linear, uniformly spaced array along the z axis, r 0m = md z, yielding

r r 0m = (sin cos x
+ sin sin y + cos , z) md
z = md cos , (36)

Prof. Sean Victor Hum ECE422: Radio and Microwave Wireless Systems
Antenna Arrays Page 12

Figure 9: Amplitude tapers

Figure 10: |AF | patterns for various amplitude tapers

Prof. Sean Victor Hum ECE422: Radio and Microwave Wireless Systems
Antenna Arrays Page 13

which when used in the generalized AF expression, gives

N
X 1
AF = ejkmd cos , (37)
m=0

a familiar result.
We now use this array factor to derive the response of a few new array topologies.

4.1 Uniformly Spaced Linear Array Along Other Cardinal Directions

We can easily write array factors for linear arrays along other cardinal directions besides the z-
0
direction by using an appropriate expression for the position vector ~rm . For example, for an evenly
spaced array along the x-axis,
r 0m = md x
(38)
and
M
X 1
AF (, ) = Im ejkmd sin cos . (39)
m=0

Notice that the AF is now a function of both and . Graphical techniques introduced earlier
can also be applied to uniformly excited arrays, since the array factor often takes the general form

f () = f (kd cos + ) (40)

where cos = r0 r is the dot product between the array axis direction and the direction of
interest, and is the inter-element phase shift. The normalized array factor is then
 
 sin(N /2) 
f () =   (41)
sin(/2) 

which can be plotted using the following procedure:

1. Plot f () on a rectilinear graph with = 0.

2. Draw a circle of radius kd below the graph, whose centre is offset from the origin ( = 0)
by an amount .

3. Vertical lines can be drawn between points of interest in f () and the circle to determine
the corresponding polar pattern angles of these features.

4. f () can be evaluated at several points around the polar graph and the points connected
to form the pattern.

Note that making such plots, as usual, requires one to define a cut in which one of or is kept
constant while the other variable is varied over an appropriate range. This leads to the following
cases:

Prof. Sean Victor Hum ECE422: Radio and Microwave Wireless Systems
Antenna Arrays Page 14

Array along x-axis:

r r 0m = md sin cos ,
k = kd sin cos (42)

xy-cut: = 90 , = kd cos . This is similar to the result for an array along the z axis
but with a cos instead of a cos . Therefore, one can graphically construct the pattern in
the same way as for an array along the z-axis, but intepret the physical angle as instead
of .

xz-cut: = 0, = kd sin . This is similar to the result for an array along the z axis
but with a sin instead of a cos . Therefore, the pattern must also be rotated 90 degrees
after the construction process, since sin() = cos( /2).

Array along y-axis:

r r 0m = md sin sin ,
k = kd sin sin (43)

xy-cut: = 90 , = kd sin . This is similar to the result for an array along the x axis
but with a sin instead of a cos . Therefore, the pattern is versus and must be rotated
90 after its construction.

yz-cut: = 90, = kd sin . This is similar to the result for an array along the z axis
but with a sin instead of a cos . Therefore, the pattern must also be rotated 90 degrees
after the construction process.

4.2 Uniformly Spaced Planar Array

An example of a 2D planar array is shown in Figure 4.2. The elements are arranged uniformly
along a rectangular grid in the xy-plane, with an element spacing dx in the x-direction and an
element spacing dy in the y-direction.

Figure 11: Two-dimensional planar array

Prof. Sean Victor Hum ECE422: Radio and Microwave Wireless Systems
Antenna Arrays Page 15

Since the arrangement is Cartesian, or matrix-like, it is useful to use two indices to refer to the
elements: a row index and a column index. Grid indices in the x and y direction are denoted as
m and n, respectively. The position vector of the mnth element is then given by
0
~rmn = x0mn x 0
+ ymn 0
y + zmn z (44)

Assuming we have the spacings indicated, and the array starts at the origin, we can rewrite the
position vector as
r 0mn = mdx x
+ ndy y (45)
where m, n = 0, 1, 2, . . .. The array factor expression is then written as follows, where we have
split the summation into two summations along each dimension:
N
X 1 M
X 1
AF (, ) = Imn ejk(mdx sin cos +ndy sin sin ) (46)
n=0 m=0

Here, Imn denotes the excitation amplitude of the mnth element of the array, and is assumed to
be a real number yielding broadside radiation.
The array factor is said to be separable if the excitations are such that

Imn = Imx Iyn . (47)

That is, the excitation is the product of two functions, one describing variation in the x-direction
and the other, the y-direction. Most commonly, we use uniform amplitude but progressive phase
shifts in each direction such that

Imx = I0 ejmx (48)

Iyn = I0 ejny , (49)

where x and y are the phase gradients in the respective directions. Then,

M
X 1 N
X 1
AF (, ) = I0 ejk(mdx sin cos +x ) ejk(ndy sin sin +y ) , (50)
m=0 n=0

which we see is simply the product of two linear array factors. This means that the beamwidths
in each of the principal directions of the array will be determined by a linear array along the
corresponding direction.

4.3 Circular Arrays

A circular array is another arrangement that is commonly found in phased arrays and recently,
microwave beacon arrays. A general diagram of a circular array is shown in Figure 4.3 for a
8-element array, although any number of elements is possible. The radius of the array is a. The
angle between elements is assumed to be uniform, such that
2
= n+1 n = . (51)
N
Prof. Sean Victor Hum ECE422: Radio and Microwave Wireless Systems
Antenna Arrays Page 16

Figure 12: Circular array

Prof. Sean Victor Hum ECE422: Radio and Microwave Wireless Systems
Antenna Arrays Page 17

If this is the case, the position vector of the nth element can be written as

r 0n = a cos n x
+ a sin n y (52)

where
2
n = n = n (53)
N
The corresponding array factor expression is
N
X 1
AF = In ejka[cos(2n/N ) sin cos +a sin(2n/N ) sin sin ] (54)
n=0

5 Feeding Array Antennas

There are a variety of ways to excite array antennas, depending on what the characteristics in the
array are desired. Generally, transmission lines must be run to each and every antenna element.
The amplitude and phase of the element excitations can be varied using discrete amplitude/phase
shifting devices, or the feed network, or a combination of both. A sampling of some common feed
networks are shown in Figure 5.

Parallel or corporate feed Series feed Hybrid feed

Figure 13: Array feeds

In a parallel or corporate feed network, all the elements are feed in parallel from a single source.
Practically, the power splitters are realized using special RF power dividers, such as Wilkinson
power dividers, or lossless combiners. Tricks can often be employed to combine matching and
power division in the feed network; for example, instead of tuning the antenna to 50 , it can
be tuned to 100 and two antenna combined in parallel using a lossless combiner to produce a
50 input impedance to a 2-element sub-array. In any case, at the antenna level, phase and/or
amplitude shifters can be added to control the element excitations, allowing a wide variety of
patterns and/or beam scanning to be introduced to the array.
Prof. Sean Victor Hum ECE422: Radio and Microwave Wireless Systems
Antenna Arrays Page 18

In a series-fed array, antennas are feed in series from a common source. As the signal travels away
from the source, antennas tap off the power, which usually results uneven power distribution to
the antennas that must be accounted for in array factor calculations. The inter-element phasing
is controlled by changing the length of feed line between the elements; for example, choosing a
length equal to one wavelength will result in all elements being fed in phase. An advantage of this
type of array structure is that it implements something called frequency scanning, since the beam
will scan with frequency. For example, using the one wavelength example, all the element will be
in phase at the frequency where the feed-line lengths are one wavelength. However, when the
frequency changes, the electrical length of the transmission line changes, resulting in shortening
(if the frequency is decreased) or lengthening (if the frequency is increased) of the feed-lines. This
in turn changes the inter-element phase shift which we know causes the beam to tilt away from
broadside. In some applications, this can be very useful. It is also quite easy to realize series-fed
arrays, depending on the transmission line technology. Periodic radiating slots cut in a waveguide
are an example.
Both concepts can be combined into a hybrid feed or parallel-series feed. Sub-arrays are formed as
series-fed groups of elements, fed by a common signal from a parallel feed structure. The parallel
structure in turn allows the individual amplitudes/phases of the sub-arrays to be controlled if
necessary.

Prof. Sean Victor Hum ECE422: Radio and Microwave Wireless Systems