Mutual coupling effects and optimum

architecture of a sparse antenna array

IRINA TECSOR

KTH Information and

Communication Technology

Master of Science Thesis

Stockholm, Sweden 2013

TRITA-ICT-EX-2013:171
Mutual coupling effects and optimum
architecture of a sparse antenna array

IRINA TECSOR

Master of Science Thesis performed at

the Radio Communication Systems Group, KTH.
June 2013

Examiner: Associate Professor Anders Västberg

Supervisor: Professor Claes Beckman
KTH School of Information and Communications Technology (ICT)
Radio Communication Systems (RCS)
TRITA-ICT-EX-2013:X
c Irina Tecsor, June 2013
Abstract

The objective of this thesis is to investigate the performance of the EISCAT 3D

Array antenna for different architectures and different types of elements. The
thesis focuses on the type of element designed by Luleå Technical University
and build by GELAB Company. Another objective is to find the optimum
architecture that will make the array meet the requirements of a 3D imaging
radar and result in a minimum number of elements for achieving the performance
goals set by the specifications.
The EISCAT 3D Array requires a high number of elements and this implies a
high level of interaction between the elements, also known as mutual coupling.
In order to simulate the performance of the array including the coupling, a
program was written in MATLAB using an element radiation pattern simulated
in NEC2. This pattern includes the mutual coupling effects of the single element
designed by Luleå Technical University.
A method called thinning is applied to reduce the number of elements and, con-
sequently, the coupling. The results show that the thinning parameters change
the performance of the array with a fixed pattern. Increasing the number of
elements, the key performance indicators (Directivity and Side lobe level) show
an improvement. It was found that one of the thinning parameters, the design
side lobe ratio, has more weight on the indicators. This parameter indicates
the maximum level of the side lobe and for increasing values, the Directivity de-
creases but the Peak Side lobe improves to a higher level. The Average Side lobe
level, however, decreases when the design side lobe ratio parameter increases.
As a conclusion, it can be said that a thinned array with a circular aperture
meets the requirements better than one with a square aperture. The circular
aperture leads to improved results (higher Directivity) with less elements. An-
other important conclusion is that the directivity of the main beam and the side
lobe level is strongly dependent on both the azimuthal and elevation angle. In
order to fully understand the consequences of mutual coupling and thinning, a
more in-depth investigation into all directions in the field of view needs to be
conducted before a final decision on the element design can be made.

iii
Acknowledgements

This work was done at the Communication Systems Department, Royal Institute
of Technology, to complete my Master of Science studies. First of all, I would
like to express my sincere gratitude to my supervisor, Professor Claes Beckman,
who guided me into the vast and wonderful area of Antenna Array, gave me
valuable advices and shared his knowledge. Also, I would like to thank Göran
Andersson for his help in mathematics and Mathematica and Anders Västberg
for easing the administrative process.
This thesis also investigates the EISCAT 3D Array and I would want to extend
my gratitude to Gunnar Isaksson, who provided me his work and time. I also
want to thank the EISCAT Community for making me feel welcomed and letting
me be part of their inspiring project.
Last of all, I want to thank my friends from home and abroad who flourished my
life experience. Ultimately, I want to say thank you to my parents and family
for always believing in me and supporting my decisions.

v
Contents

1 Introduction 1
1.1 Background. EISCAT Radar . . . . . . . . . . . . . . . . . . . . 2
1.2 EISCAT 3D Radar . . . . . . . . . . . . . . . . . . . . . . . . . . 2
1.3 Problem description . . . . . . . . . . . . . . . . . . . . . . . . . 5

2 Antenna Theory 7
2.1 Parameters for designing an Antenna . . . . . . . . . . . . . . . . 7
2.2 Antenna type . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10
2.3 Key performance indicators . . . . . . . . . . . . . . . . . . . . . 13
2.3.1 Radiation pattern and Side lobe level . . . . . . . . . . . 13
2.3.2 Antenna Directivity . . . . . . . . . . . . . . . . . . . . . 14

3 Mutual coupling in Antenna Arrays 17

4 Thinning method 21

5 Implementation 25
5.1 Implementation of the Array Factor . . . . . . . . . . . . . . . . 25
5.2 Implementation of SDT . . . . . . . . . . . . . . . . . . . . . . . 26
5.2.1 Taylor’s amplitude distribution for a linear array . . . . . 27
5.2.2 Taylor’s amplitude distribution for circular aperture . . . 29
5.3 Implementation of Radiation pattern . . . . . . . . . . . . . . . . 30
5.4 Method for calculating the Side lobe level . . . . . . . . . . . . . 32
5.5 Implementation of Directivity . . . . . . . . . . . . . . . . . . . . 32

6 Results and Discussion 35

6.1 Directivity results . . . . . . . . . . . . . . . . . . . . . . . . . . . 36
6.2 Key performance indicators for a 625 elements array . . . . . . . 37
6.3 Influence of thinning parameters . . . . . . . . . . . . . . . . . . 39
6.4 Array with 10000 elements . . . . . . . . . . . . . . . . . . . . . . 40
6.5 Influence of direction of main beam . . . . . . . . . . . . . . . . . 43

7 Conclusions 47
7.1 Future work . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 48

References 51

A Appendix 53

vii
Chapter 1

Introduction

Earth is surrounded by a composition of gases where Nitrogen and Oxygen are

predominant. Near the surface of Earth, almost all gases have a molecular form,
but with increasing altitude, the gases’ density decreases [1]. The atmosphere’s
layers surrounding the Earth are presented in figure 1.1. The ionosphere is the
upper layer of the atmosphere and extends from 90 km to approximately 400
km above the Earths surface. Also called thermosphere, this layer gets its name
from the type of atoms that populate it: ionized atoms. By gaining or loosing
electrons, atoms become ionized. The ionosphere protects the Earth against the
Suns Ultraviolet Radiation by absorbing the radiation (photons) and releasing
ions by liberating electrons.

The ionosphere is of great importance for communication and navigation. Wire-

less communications use it for signal reflection while satellites must transmit
signals that can pass through it. Researchers are interested in studying this
layer because the ionosphere possesses gas and plasma proprieties. At northern
and southern altitudes, a beautiful natural phenomenon can be seen because of
the ionosphere: the collision of solar wind particles with atoms from ionosphere
creates the Aurora.

Figure 1.1: Atmosphere’s layers. Source [2]

1
2 Chapter 1. Introduction

1.1 Background. EISCAT Radar

For studying all the three layers of the ionosphere and the interactions between
the Earth and Sun, the European Incoherent Scatter Scientific Association (EIS-
CAT) was founded in 1975 and built three incoherent scatter radars. They are
located in the northern hemisphere, two radar systems on the mainland and one
on the Spitzbergen island, in the Svalbard archipelago. Figure 1.2 presents the
location of the EISCAT radar systems. From the radar systems, information
about ion and electron temperature, electron density and ion velocity can be
detected [3].
The incoherent radar is based on incoherent scattering. The free electrons from
the ionosphere, liberated after the interaction between Sun’s radiations and
atoms, scatter the electromagnetic waves originated from the radar. Usually,
the radar has the transmitter and receiver on the same physical structure. This
is called a monostatic radar. When the transmitting and receiving sites are
separated by a large distance, the radar is called a bistatic radar. Two of the
EISCAT radar systems, VHF1 and ESR2 , are monostatic and the third one,
UHF3 , is a tristatic one. The UHF radar system has the transmitter and one
receiver in Norway and two receiving sites, in Sweden (Kiruna) and Finland
(Sodankyl).

1.2 EISCAT 3D Radar

The technological progress made in the field of incoherent scatter radars, the
new issues arising from the EISCAT radar research and also the need to im-
prove the efficiency of the existing radar, drove scientists to implement a new
incoherent scatter radar, called EISCAT 3D. The Fenno-Scandinavian Arctic
atmosphere and the coupling between space and Earth’s atmosphere will be
studied with a three-dimensional imaging radar. A three-dimensional radar is
able to create a 3-D image of a three dimensional quantity and to distinguish
between temporal and spatial variations. The project is in Preparatory Phase
starting from October 2010 until September 2014. Following this phase, the
Implementation Phase and Construction Phase are expected to be finished in
2018, when the first observations can be made.
The new generation radar will offer [5] information about induced changes in
the ionosphere, small scale plasma physics, micrometeors, planetology, and so-
lar wind acceleration. These measurements can be done by combining several
techniques [6] that were not used until now in a single radar:
• volumetric imaging and tracking (the beam can be moved in just a few
milliseconds),
• interferometric imaging (the elements of the array can be grouped in clus-
ters for aperture synthesis imaging),
1 Very High Frequency radar, operating in the 224 MHz band.
2 EISCAT Svalbard Radar, operating in the 500 MHz band
3 Ultra High Frequency radar, operating in the 931 MHz band
1.2. EISCAT 3D Radar 3

Figure 1.2: Location of the three radar systems. Source [4]

• multistatic configuration,
• better sensitivity (higher antenna gains and larger active arrays)
• and flexibility for transmitter (there will be a signal generator for every
transmitter unit).
The center frequency is at 233 MHz, in the 200 to 240 MHz band [6].
The existing EISCAT radars operate with big diameter antennas. The trans-
mitter for UHF and ESR radars is a 32 m steerable parabolic dish antenna and
for VHF radar, the transmitter is a 120 x 40 m parabolic cylinder antenna. In
figure 1.3 the EISCAT UHF radar is shown. Having one element to transmit,
the resolution is low and the radiation pattern has only one beam. To perform
the advanced measurements, EISCAT 3D radar needs to have a high resolution
and multiple beams. To achieve this, one solution is to increase the diameter of
the parabolic antennas, but is hard to realize it physically. The solution used
for the 3D radar is an Array Antenna. With this configuration, multiple beams
and high resolution can be achieved.
Using just one antenna the radiation pattern is a known constant but placing
a group of antennas in a certain way leads to a different radiation pattern.
Multiple antennas positioned on a straight line are known as a Linear Array of
antennas. If the elements are placed on a plane, then that is a Planar Array of
antennas. Figure 1.4 exemplifies the difference between a Linear Array and a
4 Chapter 1. Introduction

Figure 1.3: EISCAT UHF Radar. Source [7]

Planar Array of antennas.

Figure 1.4: Left: Linear Array along z axis. Right: Planar Array

For radar applications, a high number of beams used for imaging increases the
performance. By properly choosing the excitation and phase of the array’s
antenna elements, multiple beams can be formed. To search and track the
particles in the ionosphere, the beams have to be steered (swung) [7] very fast.
The new EISCAT 3D radar is expected to achieve instantaneous beam-swinging
and multiple simultaneous transmit and receive beams, without moving the
physical structures [3]. The transmitting and receiving beams are used for
different signals steered in different directions. To create beams strong enough
to be detected by a radar, after the signal was scattered in the ionosphere, a
high power aperture is needed. Also, a big aperture is desired to collect the
signals and for the 3D radar, the aperture will be built with a high number of
antennas in the Planar Array of Antennas. Figure 1.5 shows a possible solution
1.3. Problem description 5

proposed by GELAB, where the antennas are elevated for a protection against
the snow. For this radar project, 5 sites with approximately 10.000 elements
each are needed, for transmission and reception. The core site will be located
at the border of Sweden, Norway and Finland and the four receivers placed 50
to 250 km away from the core.

Figure 1.5: Possible configuration of antenna proposed by

GELAB Company. Source [7]

1.3 Problem description

For incoherent scatter radar applications, it is required a high number of ele-

ments to achieve a good performance. But placing a high number of elements
close to each other will lead to electromagnetic interactions between them also
known as mutual coupling. The coupling becomes difficult to handle as the
number of elements increases. It is no wonder that a technique to remove ele-
ments (thinning technique) is applied to these arrays to reduce the number of
elements and implicitly the cost. The half power beamwidth, see section 2.3.1,
for the antenna array is related to the length of the array [8] and by removing
elements, but keeping the same length of the antenna, the beamwidth stays
approximately unchanged. The down size of the thinning technique is that the
beam will loose its narrow pattern, as the directivity, see 2.3.2, is proportional
with the number of elements. As the mutual coupling affects the antenna pa-
rameters, by reducing the elements, the mutual coupling between array elements
is also reduced.
As the newly thinned array has fewer elements, the spacing between radiators
is changed from a uniform spacing to a non-linear one. The main beam is not
affected by the position of the elements, but the mutual coupling is highly de-
pendent on the position and on the relative position between each other. The
challenge when thinning the array is not to cause a greater coupling. In the
literature, there are a few methods to thin the array, using statistical density
taper or quantized elements weights, for example. It is known that some meth-
ods lead to fewer removed elements but the coupling is not investigated. The
6 Chapter 1. Introduction

studies of the thinning methods are done assuming no mutual coupling in the
array, but in real scenarios coupling cant be ignored.
Different thinning techniques lead to different architectures of the antenna array
and therefore different values for the directivity and side lobe level. The chal-
lenge will be to apply a thinning technique that will remove enough elements to
lower the cost, but will keep an unchanged beam and a low coupling.
Chapter 2

Antenna Theory

Antenna is used as a means to communicate, its basic function is to radiate

and receive electromagnetic waves. The radio waves coming from a guiding
device are radiated by the antenna into free space and received by another
antenna. This chapter presents the parameters that can influence the radiation
characteristics of an antenna, the type of elements investigated in this thesis
and the indicators of the performance of an antenna.

2.1 Parameters for designing an Antenna

To be able to transmit information at long distances, the antenna has to be

highly directive. While a single antenna usually has a broad radiation pattern
and a low directivity, a linear array of antennas can achieve a narrower pattern
and higher directivity through the Array Factor. This factor quantifies the
combination of radiating elements without taking into account the radiation
pattern of the single element. The electric field of the whole array in the far
zone [9] will then be the product of the array factor and the field of the single
radiating element, presented in (2.1). If the overall radiation pattern leads to
an improved directivity, the gain will also be improved.

E(total) = E(single element at reference point) ∗ Array Factor (2.1)

where E is the far field radiated by an antenna.

Figure 2.1 presents the product of the Element Factor and Array Factor to
calculate the Antenna Pattern, for an array with 25 elements, phase shift of 0◦
and a distance of λ/2 between the elements.
Now let’s have a closer look at this product. The field radiated by the single
antenna is constant, but the array factor can be changed as required. An array
with identical elements, same excitation and progressive phase is called a uni-
form array. To compute the Array Factor, the radiating elements are considered

7
8 Chapter 2. Antenna Theory

Element Array Factor

0° 0°
15° 345° 15° 345°
0 0
30° 330° 30° 330°

45° -5 315° 45° -5 315°

-10 -10
60° 300° 60° 300°

-15 -15
75° 285° 75° 285°
-20 -20

90° -25 270° 90° -25 270°

105° 255° 105° 255°

120° 240° 120° 240°

135° 225° 135° 225°

150° 210° 150° 210°

165° 195° 165° 195°
180° 180°

Total

0°
15° 345°
0
30° 330°

45° -5 315°

-10
60° 300°

-15
75° 285°
-20

90° -25 270°

105° 255°

120° 240°

135° 225°

150° 210°
165° 195°
180°

Figure 2.1: Calculation of the Radiation Pattern [dB]. Source [9]

to be point sources. For a linear array, where the elements are placed along z
axis, the Array Factor (2.2) is :

N
X
AF(θ) = ej(n−1)(kdcosθ+β) (2.2)
n=1

where k is the wave number, d the distance between elements and β the pro-
gressive shift applied to the elements.
When placing the elements on a rectangular grid, a Planar Array is formed. It
has more parameters compared to the Linear array that can adapt and control
the pattern. For a Planar Array with M elements on x axis and N elements on
y axis, the Array Factor is presented in (2.3):

M
X N
X
AF(θ, φ) = Im1 ej(m−1)(kdx sinθcosφ+βx ) I1n ej(n−1)(kdy sinθsinφ+βy ) (2.3)
m=1 n=1
2.1. Parameters for designing an Antenna 9

In some applications, it is of interest to have the main beam toward a certain

direction. For example, if the beam should be in the (θ0 , φ0 ) direction, the
progressive phase between the elements should be:

βx = −kdx sinθ0 cosφ0 (2.4)

βy = −kdy sinθ0 sinφ0 (2.5)

A number of parameters are used to control the array [10] such as:
1. Number of elements,
2. Element type and pattern of individual element,
3. Position,
4. Signal excitation.
1. Number of elements. The number of radiating elements in the array is
inversely proportional to the beam [11]. When a high number of elements is
used, the beam becomes narrower, more directive. Figure 2.2 exemplifies the
relation between the number of elements in the array and the field pattern of
the Array factor. The configuration for the two Planar Arrays is also presented
in figure 2.2. The distance between the elements is λ/2, the phase shift is zero
and the cut is made in the xz plane (φ = 0◦ ).
2. Element type and pattern of individual element. Depending on the
application, different types of antenna are used: Wire antennas for frequencies
up to hundred of MHz, Aperture antennas for frequencies higher than 1 GHz
and if the radiated power should have a certain direction, Reflector antennas
for radio astronomy to increase the aperture, Microstrip antennas for integrated
systems because they are easy to manufacture and conformable with planar
surfaces. The radiation pattern with a certain element factor is presented in
figure 2.3.
3. Position. The relative distance between the elements plays a well know role.
Increasing the spacing of elements, the beam becomes narrower. If, however,
a broad radiation pattern is desired, the spacing must decrease until the limit
where the mutual coupling becomes considerable. Usually elements are designed
with a distance of half a wavelength between. Figure 2.4 presents the radiation
pattern for a planar array with 25 elements and a distance between them of λ/4
and λ/2.
4. Signal excitation. When a relative delay of excitation currents or time
delay is introduced between the elements of the array, a relative phase shift
occurs between the radiation vectors. The phase shift can cause the vectors to
add constructively and destructively in different directions, steering the radia-
tion beam in the desired direction. Figure 2.5 shows√the beam of an array with
no phase shift and the beam of an array with -π/2 2 phase shift between the
elements. The elements are λ/2 spaced and the cut is made in the xz plane.
Because of the aggregation of elements, a controlled radiation pattern can be
achieved and arrays are sometimes preferred, instead of fixed aperture antennas.
10 Chapter 2. Antenna Theory

0° 0°
15° 345° 15° 345°
0 0
30° 330° 30° 330°

45° -5 315° 45° -5 315°

-10 -10
60° 300° 60° 300°

-15 -15
75° 285° 75° 285°
-20 -20

90° -25 270° 90° -25 270°

105° 255° 105° 255°

120° 240° 120° 240°

135° 225° 135° 225°

150° 210° 150° 210°

165° 195° 165° 195°
180° 180°

0 0

0.5 1

1 2

1.5 3

2 4

2.5 5

3 6
0 0.5 1 1.5 2 2.5 3 0 1 2 3 4 5 6
nz = 4 nz = 25

Figure 2.2: Radiation Pattern [dB] of Planar Array Factor and the
corresponding array configuration. Left: 4 elements. Right: 25 elements

By changing the amplitude of the currents, the shape of the radiation pattern
can be controlled. When the same amplitude is used for all elements, then the
array is said to be uniformly illuminated.
For controlling the excitation currents, for a single beam array, [12] a circuit
of power splitters and delay lines, also referred as a passive fed array, is used.
Electronic control can be used to excite the array elements and dynamically
adjust the radiation pattern. The amplitude and phase of excitation current
are controlled [13] by a variable attenuator and a variable phase shifter. The
array using this method of excitation is called an active array or electronically
steered beamforming array.
The combination of element excitation, amplitude and phase can be viewed as a
complex voltage coefficient. This complex coefficient is often referred to as the
array weight of the element array. Through the use of the array weight, antenna
designers can implement a wide range of radiation pattern for array antennas.

2.2 Antenna type

If the element has a wide radiation pattern or on the contrary, has a narrow
pattern, it will make the array change its pattern. In this thesis, 3 different
2.2. Antenna type 11

Element Array Factor

0° 0°
15° 345° 15° 345°
0 0
30° 330° 30° 330°

45° -5 315° 45° -5 315°

-10 -10
60° 300° 60° 300°

-15 -15
75° 285° 75° 285°
-20 -20

90° -25 270° 90° -25 270°

105° 255° 105° 255°

120° 240° 120° 240°

135° 225° 135° 225°

150° 210° 150° 210°

165° 195° 165° 195°
180° 180°

Total

0°
15° 345°
0
30° 330°

45° -5 315°

-10
60° 300°

-15
75° 285°
-20

90° -25 270°

105° 255°

120° 240°

135° 225°

150° 210°
165° 195°
180°

Figure 2.3: Calculation of radiation pattern [dB] with a certain element factor

types of antenna will be investigated:

1. Isotropic antenna,
2. Half wavelength dipole antenna,
3. Antenna designed by the EISCAT 3D Consortium (simulated in NEC21 )
Isotropic antenna. It is an ideal model of an antenna, which doesn’t exist in
reality. It is a tool used to reference the directivity because an isotropic radiator
radiates in all direction with the same intensity, thus having the directivity 1.
The directivity of real antennas is compared with the directivity of an isotropic
radiator and will always be higher than 1. An actual antenna will radiate and
receive energy in some directions better than other directions.
Half wavelength dipole antenna. A dipole is a wire antenna and usually is
positioned along the z axis, simmetrically at the origin, as presented in figure
2.6. If the wire has the length of half the wavelength (l = λ/2), then it is called
a half wavelength dipole.
1 Numerical Electromagnetics Code, software for antenna modeling
12 Chapter 2. Antenna Theory

0° 0°
15° 345° 15° 345°
0 0
30° 330° 30° 330°

45° -5 315° 45° -5 315°

-10 -10
60° 300° 60° 300°

-15 -15
75° 285° 75° 285°
-20 -20

90° -25 270° 90° -25 270°

105° 255° 105° 255°

120° 240° 120° 240°

135° 225° 135° 225°

150° 210° 150° 210°

165° 195° 165° 195°
180° 180°

Figure 2.4: Radiation pattern [dB] of Array Factor, for 25 elements planar
array. Left: λ/4 distance. Right: λ/2 distance between elements

0° 0°
15° 345° 15° 345°
0 0
30° 330° 30° 330°

45° -5 315° 45° -5 315°

-10 -10
60° 300° 60° 300°

-15 -15
75° 285° 75° 285°
-20 -20

90° -25 270° 90° -25 270°

105° 255° 105° 255°

120° 240° 120° 240°

135° 225° 135° 225°

150° 210° 150° 210°

165° 195° 165° 195°
180° 180°

Figure 2.5: Radiation pattern

√ [dB]; Left: no phase shift.
Right: -π/2 2 phase shift

The electric field of the half wavelength dipole is found with 2.6 and shown in
2.6:

"
#
π
Io e−jkr cos 2 cos θ
E = jη (2.6)
2πr sin θ

where η is the intrinsic impedance of the medium and I0 a constant.

Antenna designed by the EISCAT 3D Consortium. The antenna is de-

signed at Luleå Technical University to meet the requirements for an incoherent
scatter radar. To account for the mutual coupling effects, a 3x3 Planar array is
simulated in NEC2 where only the middle element is fed with 1 Volt. The other
elements are terminated with a 50 Ohms resistor.
2.3. Key performance indicators 13

Figure 2.6: λ/2 dipole. Left: Geometrical position.

Right: Three dimensional pattern of electric field

2.3 Key performance indicators

To evaluate the performance of the simulated antenna array, certain parameters
are chosen to be calculated. They can be calculated in a theoretical way or
can be measured from the radiation pattern. The latter is a better way, having
in mind that the actual properties of an antenna are related to the radiation
pattern. The following parameters are investigated in this thesis:
• Side lobe level
• Directivity

2.3.1 Radiation pattern and Side lobe level

An antenna receives and transmits energy in some directions better than for
other directions. According to [9], the radiation pattern expresses the radiation
properties of a radiating element, in the shape of a mathematical function or
graphical representation, as a function of space coordonates.
If the received electric field is measured at a constant radius, then this is known
as amplitude field pattern. When the function of power density is constructed at
a constant radius, with variating space coordonates, this is known as amplitude
power pattern. Usually the following patterns are used:
• field pattern; the function of the magnitude of the electric (magnetic) field,
in linear scale, with variating space coordonates,
• power pattern; the function of the squared field (magnetic or electric), in
linear scale, with variating space coordonates,
• power pattern; the same as the linear power pattern, but in a logarithmic
scale, in dB.
The radiation has a maximum and the portion of the pattern that incorporates
the maximum is called a main lobe. A lobe is when the pattern is bordered by
parts of the radiation pattern where the magnitude is small. In figure 2.7 the
14 Chapter 2. Antenna Theory

main lobe can be seen to be the part where the radiation intensity is maximum.
For some applications, it is of interest to have more than one main lobe.

Figure 2.7: The lobes of the radiation pattern

Beside the main lobe, all the regions with a small intensity as a border are
called a minor lobe. A common name for the lobe that is not in the maximum
direction is side lobe. When the lobe is 180◦ away from the main lobe, then it
is called a back lobe. The minor lobes are unwanted energy in certain direction
and the lower they are, the better. Among the minor lobes, the side lobes have
the highest levels. In figure 2.7 can be seen that the side lobe is different than
the main lobe and that is adjacent to it. It is common to express the level of the
minor lobe as the ratio of the power density in that lobe to the level of the main
lobe. This value is called side lobe level (SLL) and it is wanted to be higher
than 20 dB.

2.3.2 Antenna Directivity

The directivity is a function of the radiation intensity. The radiation intensity,

U (θ, φ) is the radiated power per unit solid angle. According to [9], if the electric
field in the far zone of the antenna and intrinsic impedance (η) of the medium
where the antenna radiates are known, then the radiation intensity (2.7) is

1 
Eθ (θ, φ)|2 + Eφ (θ, φ)|2

U (θ, φ) = (2.7)
2η

The Directivity of an antenna is defined in [9] as the ratio of the radiation inten-
sity in a given direction to the radiation intensity averaged over all directions.
It shows how well an antenna radiates the power into a certain direction. If
there is no specification about the direction, then it is implied the direction of
the maximum radiation intensity. The general expression for computing the
directivity is presented in (2.8)
2.3. Key performance indicators 15

4πU (θ, φ)max

D(θ, φ) = R 2π R π (2.8)
0 0
U (θ, φ)sinθdθdφ

From equation 2.8, one can easily see that directivity expresses the directional
proprieties of a radiating element and is only a function of the radiation pattern.
In consequence, the radiation pattern controls the directivity.
Another parameter related to the antenna pattern, is the beamwidth. The angle
distance between two equivalent points, diametrically opposed, of the antenna
pattern is known as beamwidth [9]. There are different types of beamwidth, but
it is most common to refer to the Half power beamwidth, or in short HPBW.
It is measured in the plane where a maximum of the pattern occurs, as the
angular distance between two directions where the radiated power is half of the
specified beam. A small beamwidth means high directivity and high directivity
translates into high level of side lobes. It is an inverse proportional relationship
between the beamwidth and the side lobe: when the beamwidth increases, the
side lobe decreases and the other way around.
Chapter 3

Mutual coupling in
Antenna Arrays

When an object is placed in the vicinity of a radiating element, it will affect

the current distribution of that element and in consequence the radiated fields.
Therefore, the current of close elements together with own currents of the radia-
tor alter the characteristics of the antenna. Energy exchange or electromagnetic
interaction between the array elements of an antenna array is known as Mutual
Coupling and is an unwanted effect. It complicates the antenna design and
analysis, influences the radiation pattern of the whole array and is difficult to
generalize. Mutual coupling is influenced by the type of the antenna element,
the position of elements and their relative orientation [9].
In an array, the object mentioned before is another radiating element. If at
least two antennas are in the receiving or transmitting mode, a part of the
energy for one will be absorbed by the second one. When the two antennas
are transmitting, as the consequence of a non-ideal directional proprieties, the
energy radiated by one is received by the second one and vice versa. One or
both antennas act as secondary transmitters by rescattering some of the received
energy in different directions. If two antennas are assumed to be placed near
each other, the coupling can be investigated. To describe the mutual coupling, a
scenario with two antennas, called X and Y, placed in each other’s vicinity, will
be analyzed. Whether the antennas are transmitting or receiving, the mutual
coupling is:
1. Coupling in Receiving Mode,
2. Coupling in Transmitting Mode.
Coupling in Receiving Mode. The antenna has the function of converting an
electromagnetic field into an induced current or voltage. Each antenna element
will provide a value for the measured voltage that depends on the incident
field and also on the voltages on the other array elements. Some part of the
received plane wave at antenna X will be rescattered into the medium, some
part will travel into its generator and another part will travel toward antenna
Y. At antenna Y, the scattered energy coming from antenna X will be summed

17
18 Chapter 3. Mutual coupling in Antenna Arrays

vectorially with the plane wave coming incidentally. Therefore, the received
energy at one antenna is the sum of the incident wave and the wave scattered
from other antennas in the vicinity.
Coupling in Transmitting Mode. When antenna X is fed, it will radiate
energy into the medium and also toward antenna Y. The unwanted incident
received energy at antenna Y will enable currents that will rescatter some of
the energy back into space and will let the remaining energy to go toward the
source of antenna Y. This behavior is the same if antenna Y is transmitting
and can repeat endlessly. When the two radiators are radiating at the same
time, the radiated and rescattered energy of each antenna has to be summed
vectorially to find the total field at a certain point of observation.
When designing a large antenna array, mutual coupling comes as an inherent
effect that can only be reduced to a small value so that the real radiation
pattern is close to the designed pattern. Over the years, different methods
to compensate this effect on antenna arrays were suggested. One of the first
methods was the Open Circuit Voltage Method, where the mutual coupling is
modeled as a mutual impedance. The idea comes from analyzing an electrical
circuit or to be specific, the network analysis with its Z parameters.
The authors in [14] use an analysis with Y parameters - admittance, where the
array is treated as a N x N linear network and the exciting voltages are related
to the port currents by an admittance matrix. To describe the input admittance
when mutual coupling is present, a circuit model of self-admittance and mutual
admittance terms is proposed. For reinforcing their model two examples of half
wavelength dipoles array are presented, where only the spacing between the
elements is changed. With the specified method, the edge elements admittance
is improved more than the one of the elements positioned in the middle of
the array, implying that the coupling of the edge elements is a consequence of
the coupling of the neighboring elements. However, if the element spacing is
increased, the admittance of center elements is improved more than the one of
the edge elements, because the center elements have more adjacent elements
than the edge elements. Dealing with admittance instead of impedance, this
method is suited for large array, having an advantage that there is no need of a
complex matrix inversion.
When designing large arrays, often the coupling effects between single elements
are neglected. The coupling introduces a change in the whole radiation pattern
and for each radiating element, in the input impedance. To design the feeding
network, a precise understanding of the input impedance is extremely crucial.
A high frequency analysis of a large array by solving Maxwell’s equations is very
difficult to perform and requires a substantial amount of computing resources.
Next, a simplified example of how the mutual coupling effects are included in
this thesis is presented.
To investigate the mutual coupling effects, first the characteristics of a single
antenna have to be investigated. At Luleå Technical University, the element
presented in figure 3.1 was designed and its gain was simulated in NEC2 when
the element is fed 2.8 m above earth, with a metallic ground plane in between
and a mesh of size λ/20. The normalized gain of the antenna, when φ = 0◦ is
shown in figure 3.1.
 19

0°
15° 345°
20.
30° 330°

45° 15. 315°

60° 300°
10.

75° 285°
5.

90° 0. 270°

105° 255°

120° 240°

135° 225°

150° 210°
165° 195°
180°

Figure 3.1: Left: Architecture of single antenna. Right: The normalized

gain of the single antenna, φ = 0◦

When constructing a 3x3 Planar Array with this element, the coupling effects
can be seen in the different shape of the gain at different angles. In the array
shown in figure 3.2, only the antenna in the center is fed with 1 Volt, the others
are terminated with a 50 ohm resistor. This configuration enables the investi-
gation of the coupling effects because the elements surrounding the antenna in
the middle will absorb the energy and will radiate it back towards the middle
element. From simulations made in NEC2 provided by [15], the electric field
of the antenna in the middle is taken to compute the Antenna pattern and the
gain. The normalized gain of the antenna in the middle, shown in figure 3.2, is
a little changed because of the coupling. The gain doesn’t have the maximum
at the same angles as the single antenna, but at higher angles. When θ is big-
ger than 60◦ the gain is decreasing, compared to the single antenna case. The
distance between elements is 0.54λ.

0°
15° 345°
20.
30° 330°

45° 15. 315°

60° 300°
10.

75° 285°
5.

90° 0. 270°

105° 255°

120° 240°

135° 225°

150° 210°
165° 195°
180°

Figure 3.2: Left: Architecture of the 3x3 array. Right: the normalized
gain of the antenna in the middle of the 3x3 array, φ = 0◦

Regularly the radiators in an array are placed on a square grid, but as the
20 Chapter 3. Mutual coupling in Antenna Arrays

relative distance between them influences the coupling, a hexagonal grid is in-
vestigated to see which antenna parameters are improving and which not. In
figure 3.3 the position of elements is presented.
√
The horizontal spacing is dx
and on the vertical axis, the spacing is dy = 23 dx .

Figure 3.3: Hexagonal grid

An advantage of using hexagonal grid is the circular symmetry of the beam. A

geostationary military satellite uses a hexagonal grid to produce multiple beams
[16]. The hexagonal grid makes the radiators to be placed in concentric circles.
Chapter 4

Thinning method

Thinning methods started being investigated in the sixties, with the work of
Skolnik and others. To decrease the high number of elements required for some
applications, but also the cost and power consumption, a process called Thinning
is applied. The thinning method reduces the element number by removing a part
of the elements, depending on a thinning factor, following a suitable procedure.
After the removal of antennas, the position of radiating elements will depend
on the applied thinning method. The degree of thinning is controlled with the
thinning factor k, which will be explained later. This factor implies how many
elements will be active in the array, referenced to the number of elements in the
filled amplitude tapered array. For example, if the thinning factor is less than
1, more than half of the elements will be removed. The same procedure applies
if k is bigger than 1, less than half of elements will be removed.
With thinning, the width of the main lobe remains almost unchanged but the
antenna gain will suffer a reduction [17]. Outside the main beam, the control
over the radiation pattern will decrease. Having this in mind, thinning is suc-
cessfully applied when control outside the main beam is not very important and
the main beam is narrow.
For large arrays, designing a narrow beam and low side lobes with amplitude
tapering takes a high degree of computational workload. Skolnik presents in
[18] a thinning method using density tapering, where the spacing between the
elements is varied, not the amplitude. The density of elements of the thinned
array is calculated using the amplitude taper of the filled array. The probability
density function used for placing the elements is calculated with the amplitude
taper and all the elements radiate with the same power. Therefore, amplitude
tapering is not necessary.
The Statistical density taper (SDT) uses the amplitude taper to decide whether
an element is active or not in the array. This method relies on Taylor’s amplitude
taper where the values of the excitation are normalized. As the distance between
the elements and the origin is increasing, the amplitude decreases, hence the
elements near the border will have a low excitation.
Taylor introduces a parameter that controls the number of side lobes with equal

21
22 Chapter 4. Thinning method

Figure 4.1: Taylor’s amplitude distribution, n=5, designed SLL=25 dB

level, n, and thinning factor k. If k is one then the thinning is natural and
the number of elements multiplied with their excitation is the same for the
filled array and thinned array. Figure 4.1 shows an example of the amplitude
excitation for an array with amplitude taper. After applying the density taper,
the number of elements will decrease, but their amplitude will be one, as opposed
to the filled case, where the amplitude is between 0 and 1. The probability of
an element being active is shown in (4.1):

VTaylor
P =k∗ (4.1)
Vmax
where VTaylor is the amplitude excitation of the element from the filled array
and Vmax is the maximum value of amplitude in the filled array.
In figure 4.2, for a 625 square array, the statistical density taper was used with
k=1. As a consequence, almost half of the elements remained active and in
this case, 304 elements are radiating. Because the method relies on a statistical
process, every time this taper is applied, the placing of the elements will differ,
as well as the number of active elements.

10

15

20

25
0 5 10 15 20 25
nz = 304

Figure 4.2: Distribution of active elements for a square 625 planar array
 23

The far field intensity is a function of the thinning factor k, accounting for the
percentage of removed elements and a statistically factor, random and indepen-
dent, Fn . Fn is 1 if the elements is active or 0 when is not active. Acording to
[18], if An is the amplitude of n-th element of the amplitude tapered antenna
and Ψn the signal’s phase of the n-th elements, the far field intensity (4.2) is:

N
X
E(θ, φ) = Fn ejΨn (4.2)
n=1

The radiation pattern is a function of the radiation pattern using the amplitude
taper and of an angle independent term. If the field intensity of an amplitude
tapered antenna (4.3) is

N
X
E0 (θ, φ) = An ejΨn (4.3)
n=1

then the radiation pattern of a statistically density tapered antenna (4.4) is

N
X
|E(θ, φ)|2 = k 2 |E0 (θ, φ)|2 + kAn (1 − kAn ) (4.4)
n=1

Using the method, the side lobe behavior is influenced by the removal of com-
ponents but the main beam is approximately the same, with or without missing
elements. Particularly the first side lobes neighboring the main beam are dom-
inated by the removal of elements and not by the amplitude taper technique.
The density tapered method is however unsuited for small arrays.
The averarge side lobe level normalized to the maximum directivity (4.5) is

PN
n=1kAn (1 − kAn )
SLL = PN (4.5)
2
n=1 |Fn |

Another thinning method, which will only be presented, is the Statistically

thinning with quantized element weights. The element weight is the excitation
that element has, comprising of the amplitude and the phase. Mailloux and
Cohen present in [19] a thinning technique based on Skolniks statistical density
tapering. Skolnik’s method is based on equally radiating elements, as opposed to
this paper, where the elements are excited with weights which can have multiple
quantized levels. A thinning method is needed when the excitation has more
than two levels, 1 for on and 0 for off, to reduce the side lobe levels. The ideal
excitation is considered in this paper to be according to the Taylor distribution.
The probability of positioning an element at a certain place depends on the
value and number of quantization levels.
The authors give 3 methods for thinning the array, after the excitation is chosen.
Using the first method, the weight is chosen, with a certain probability, above
the ideal amplitude and the weight can have a value different than 0 or 1. This
method, compared with Skolnik’s, reduces fewer elements because the elements
24 Chapter 4. Thinning method

have a low density at the end of the quantization level. With the second method,
if the excitation is closer to the next lower quantization level, the probability of
the element being excited by the weight of the lower level increases. It is then
understood that less elements are removed because the probability of being
excited with the lowest level, corresponding to weight of 0, decreases. The third
method resembles the second method, but has an increased number of quantized
weight levels. With a small number of levels, this method does not lead to good
results. The presented experiment attests that using one of the three methods
a side lobe reduction can be achieved.
Chapter 5

Implementation

This chapter offers a detailed explanation of how the antenna array was sim-
ulated and how its parameters of concern were evaluated. To construct the
radiation pattern for a linear array and a planar array, Mathematica software
was used. It offers a wide range of plotting functions, compared to Matlab,
essential to the design process and the actual radiation pattern can be easily
shown and understood. The field pattern of the single element in the middle
of the 3x3 planar array, designed by Luleå Technical University, is simulated in
NEC2 and the values are used in this thesis to construct the antenna pattern of
the single element of the EISCAT antenna. Further, when the array is thinned,
because multiple functions have to be implemented, it was found that Matlab
is a more appropriate software.

5.1 Implementation of the Array Factor

If the element factor is straightforward to implement, the array factor requires a
detailed attention because of its double parameter character (it has a variation
with θ and φ). The function calculating the electric field for the array factor has
two input values, the spherical angles θ and φ, which can be seen in figure 1.4,
but has only one output - the value of the array factor. In order to easily access
the values of the far field pattern, they will be stored in a matrix. Therefore,
the values of the array factor have to be stored in a matrix as well. A function
with two inputs - θ and φ in radians, is implemented. As stated in chapter
2.1, there are a number of parameters that influence the characteristics of the
antenna and the function which calculates the electric field can accept different
parameters such as:
• the number of elements on x and y axis (the total number of elements is
the product between them),
• the distance between elements expressed in multiples of λ, on x and y axis,
• the phase shift between elements on x and y axis,
• the amplitude of each element.

25
26 Chapter 5. Implementation

To create the matrix, two for loops are inserted where the function calculating
the electric field is called. As the far field values are stored in a matrix, the
rows represent the values of the pattern with increasing θ and the columns the
values with increasing φ. Usually, the pattern is investigated for θ from 0 to
π/2 and for φ from 0 to 2π. Using a 1◦ step, the radiation matrix will have 91
rows with θ from 0 to 90◦ and 360 columns with φ from 0 to 259◦ . For example,
the value stored at position (31,45) is the value of the antenna pattern when θ
is 30◦ and φ is 44◦ . The array factor is calculated for a square planar array and
for a circular aperture array.
To validate the function, the antenna pattern for a 25 square planar array √ was
computed. The distance between elements is λ/2, the phase shift is -π/2 2.
The normalized pattern is shown in figure 5.1, plotted in Mathematica, where
the graphic functions are more diverse. In the left of figure 5.1, the three
dimensional antenna pattern is shown using a function that plots the surface of
revolution with the height of the antenna pattern, radius θ and φ, varied from
0 to π/2. In the right is the shape of the actual pattern generated by the array,
plotted in spherical coordinates. It was found that the function gives correct
results.

Figure 5.1: Antenna pattern for a 25 square planar array.

Left: Three dimensional pattern. Right: Real pattern

5.2 Implementation of SDT

Relying on the Taylor taper, the first step in applying the SDT is to find the
Taylor amplitudes. Taylor defined an amplitude distribution for a linear array
and another distribution for a circular aperture. Both of them will be described
and investigated.
5.2. Implementation of SDT 27

5.2.1 Taylor’s amplitude distribution for a linear array

As the elements are placed on a rectangular grid, the Taylor distribution of

amplitude is based on a linear distribution. Taylor’s linear distribution designs
the minor lobes closest to the main lobe to have an equal and specified level
while the others are decaying monotonically. Taylor introduces 3 parameters:
• n, represents how many side lobes, adjacent to the main beam, have equal
amplitude,
• design side lobe ratio, the ratio of the main beam and adjacent side
lobes. Related to this ratio, A is a real parameter with the property
that cosh(πA) is the design side lobe ratio,
• σ, also called a scaling factor, is used to space the nulls of the pattern
function.
After choosing the n parameter and the designed side lobe ratio, according to
[8], the σ parameter can be found with (5.1):

n
σ=p (5.1)
A2 + (n − 0.5)2

Taylor defined a function for the antenna pattern which matches the desired
pattern - a number of side lobes have equal level and the rest of them are
decaying. Taylor’s pattern is shown in figure 5.2 for a 20 elements linear array
with n = 5 and designed SLL of 25 dB. Taylor’s function is presented in formula
(5.2)

2
n−1 z
sin πz Y 1 − zn 2
F (z, A) = (5.2)
πz n=1 1 − nz22

Relative Power@dBD
degree
20 40 60 80

-10

-20

-30

-40

-50

-60

Figure 5.2: Antenna pattern for a 20 elements linear array

with n = 5 and designed side lobe level of 25 dB

The nulls of the pattern function, formulated by [8], are found with (5.3):

 p
σ A2 + (n − 0.5)2 , n <= n
zn = (5.3)
n ,n > n
28 Chapter 5. Implementation

The coefficients of the function, as stated by [8] are calculated with (5.4):

n−1
(n − 1)!2 Y m2

F (m) = 1− 2 (5.4)
(n − 1 + m)!(n − 1 − m)! n=1 zn

Finally, the distribution function for a linear array, that produces Taylor pattern,
defined by [8] is found with (5.5):

n−1
X
g(x) = F (0) + 2 F (m) cos(mπx), − L/2 ≤ x ≤ L/2 (5.5)
m=1

The g function stretches from x = −L/2 to x = L/2 and has a maximum at

x = 0. If a normalization to the length of the antenna is done, g stretches from
x = −1 to x = 1 with a maximum again at x = 0. Depending on the number of
elements of the filled array, the amplitude for a linear array is ”sampled” and
the excitation of each element is calculated. In figure 5.3 the Taylor amplitude
distribution is presented, where 0 represents the middle of the antenna and the
borders are at -1 and 1. The pink squares represents the excitation of each
element, if 9 elements are chosen to build the linear array. For finding the
excitation of an element in the planar array, the amplitude corresponding to
the position on x axis is multiplied with the amplitude corresponding to the y
axis [20], which will lead to a matrix of amplitude values.

Figure 5.3: Taylor’s linear amplitude distribution for each element

in a 9 linear array, n = 5, SLL = −25 dB
5.2. Implementation of SDT 29

5.2.2 Taylor’s amplitude distribution for circular aperture

The aperture is the area of the radiating device. An antenna with circular
aperture is a radiating device with a circular shape, as the paraboloidal reflector,
for instance.
Taylor defined a pattern for circular aperture derived from the pattern of a cir-
cular aperture with uniform excitation, J1πz
(πz)
. Taylor’s pattern [21] is presented
in equation (5.6)

2
n−1 z
2J1 (πz) Y 1 − σ2 [A2 +(n−0.5)2 ]
F (z, A) = 2 (5.6)
πz n=1 1 − υzn 2

where J1 (x) is the order one Bessel function.

A new parameter is introduced for this distribution and is named µn . It is a
point on the real axis and is defined as:

J1 (πµn ) = 0 (5.7)

The σ parameter is a little different than the one for the linear distribution and
defined as:

µn
σ=p (5.8)
A2 + (n − 0.5)2

The nulls of the radiation function, formulated by [8], are found with (5.9):
 p
σ A2 + (n − 0.5)2 , n <= n
zn = (5.9)
µn ,n > n

Finally the distribution function for the aperture is given in (5.10)

n−1
2 X J0 (µm p) F (µm , A)
g(p) = (5.10)
π 2 m=0 [J0 (πµm )] 2

The coefficients of the function, as stated by [21] are calculated with (5.11):

n−1 µm 2
Y 1− σ 2 [A2 +(n−0.5)2 ]
F (µm , A) = −J0 (πµm ) 2 (5.11)
n=1 1 − µυm n
2

As the aperture distribution is a continuous function, to find the excitation for

each element in the array, the continuous function has to be sampled. First, the
elements are placed on a grid and the array will have a circular boundary to fit
the circular aperture.
For the (m,n) element, which is on the m-th row and n-th column, the distance
to the middle of the array is computed with equation (5.12)
30 Chapter 5. Implementation

 2  2
(Nele − 1) (Nele − 1)
dmn = −m−1 + −n−1 (5.12)
2 2

Nele−1
where the middle of the array is at 2 as the elements are placed starting
with position (0,0).
The distance to the middle element is normalized with the radius of the circular
aperture, assumed to be Nele−1
2 .

dmn
pmn = (5.13)
Radius

From [16], to find the excitation of each element, the circular distribution func-
tion is computed for the value of pm n

excitationmn = g (pmn ) (5.14)

After the excitation for each element is found, the elements can be removed
according to the method. To place an active element in the array, Sklonik
defines a probability function kAmn where k is the thinning factor and Amn is
the amplitude corresponding to the element located at (m,n). A matrix with
random numbers between 0 and 1 is generated. Each number of this matrix
is compared with the matrix of Taylor amplitudes. If the random number is
smaller than the product of k factor and Taylor amplitude, then that element
will be active and this result will be stored as an element of the matrix containing
the signal excitation for the Planar array. It is understood that at the edge of
the array, where the Taylor amplitude is smaller, the probability of an element
being removed is higher.
The thinnig factor influences the probability of being removed as well. A high
k multiplied with the Taylor amplitude will increase the chances of the random
number being smaller than the product. Thus, the probability of being removed
is smaller.

5.3 Implementation of Radiation pattern

The array factor, as it was presented earlier, represents the electric field at a
constant radius away from the antenna. The variations of the power density,
again at a constant radius away from the antenna, represents the amplitude
power pattern, according to [9]. Usually these patterns are normalized to their
maximum value and is not unusual for the power pattern to be plotted in a
logarithmic scale, or to be precise in a dB scale.
A distinction has to be made between Antenna pattern and Radiation pattern.
The Antenna pattern represents the magnitude variations of the electric field,
when the angular space is varied and the Radiation pattern represents the vari-
ations with angular space of the square of the magnitude of the electric field. In
5.3. Implementation of Radiation pattern 31

figure 5.4 is shown how the electric field for an array is computed, as the prood-
uct between the electric field of the single element and the electric field of the
array factor. All patterns presented and calculated in this thesis are normalised
to their maximum value.

Figure 5.4: Computation of the electric field

The interest is on the radiation pattern and so after the magnitude of the electric
field for the antenna element and the array factor is calculated, their magnitude
product is calculated. To have an efficient code, instead of normalizing both
magnitudes for the antenna element and array factor and multiply them, first
the product is calculated and later a normalization is done. Then the normalized
magnitude of the electric field is squared to find the radiation intensity.
32 Chapter 5. Implementation

5.4 Method for calculating the Side lobe level

The main beam is where the radiation is concentrated and usually is pencil
shaped. Everything outside the main beam is considered side lobe, after the
first minimum of the radiation pattern. In the literature, the side lobe level is
referred to as the Average Side lobe level or Peak Side lobe level.
The first local minimums surrounding the main beam have to be found to distin-
guish between the main beam and the side lobes. The radiation pattern values
are stored in a matrix, where the rows represent the θ variation and the columns
the pattern φ variations. The radiation is normalized and so the maximum is
represented by a one in the matrix. From the maximum’s position, the first
dips in the matrix are searched, with increasing and decreasing θ and φ. The
four new found coordinates for θ and φ represent the borders of the main beam.
The shape of the beam is considered to be square.
Average Sidelobe level
After the main beam is found, with a for loop the matrix is scanned and the
radiation pattern outside the main beam is summed and averaged with the
number of positions outside the main beam.
Peak Sidelobe level
To find the Peak, the same method used for calculating the Average Side lobe
level is applied. The first local minimums in the radiation are found. As the Peak
Side lobe level represents the highest level of any side lobe, the first maximum
outside the main beam is the Peak.

5.5 Implementation of Directivity

The expression for calculating the directivity with the radiation intensity from
chapter 2.3.2 will be repeated for an easier understanding.

4πU (θ, φ)max

D(θ, φ) = R 2π R π
0 0
U (θ, φ)sinθdθdφ

The above expression of directivity assumes a continuous radiation pattern but

the implemented pattern is discrete, with steps of 1◦ for both θ and φ. Therefore,
integrating the radiation intensity to find the radiated power is not a valid
option. The solution in this case lies in handling the expression with numerical
techniques.
To illustrate the method used for compiling the directivity, an example will be
given. If the sought radiation intensity is assumed to be separable, then it will
look like in (5.15)

U = C0 p(θ)r(φ) (5.15)
5.5. Implementation of Directivity 33

where C0 is a constant. Then the directivity of the antenna [9] is expressed in

(5.16)

4πUmax
D0 = (5.16)
Prad

and the radiated power is presented in 5.17

Z Z 
2π π
Prad = C0 p(θ)r(φ)sinθdθ dφ (5.17)
0 0

which is the same as writting

Z Z 
2π π
Prad = C0 r(φ) p(θ)sinθdθ dφ (5.18)
0 0

From [9], the last integral can be written as

Z N
X
π
p(θ)sinθdθ = [p (θi ) sinθi ] ∆θi (5.19)
0 i=1

When the interval is divided into N parts

π
∆θi = (5.20a)
N

and the step angle is

π
θi = i , i = 1, 2, 3, ..., N (5.20b)
N

In the same way the second integral can be written

Z M
X
2π
r(φ)dφ = r(φ)∆φj (5.21)
0 j=1

When the interval is divided into M parts

2π
∆φj = (5.22a)
M

and the step angle is

 
2π
φj = j , j = 1, 2, 3, ..., M (5.22b)
M
34 Chapter 5. Implementation

Using 5.19, 5.20a, 5.21 and 5.22a, 5.18 can be written as

M
( "N #)
 π   2π  X X
Prad = C0 r (φj ) p (θi ) sinθi (5.23)
N M j=1 i=1

If the radiation intensity is not a separable function of θ and φ, the radiated

power is

M
"N #
 π   2π  X X
Prad = U (θi , φj ) sinθi (5.24)
N M j=1 i=1

The double sum should be interpreted as follows: for each value of j from 1
to M, all values of i from 1 to N are added. This is implemented as two for
loops, one to do the summation for i and the other to do the summation for j.
Equation (5.24) shows how for each value of r(φ), when φ = φj , all p(θ)sinθ
values are summed, when θ = θi .
Equation (5.17) is a general equation which assumes the radiation intensity
exists for θ in [0, π] interval and for φ in [0, 2π] interval. The antenna designed
by EISCAT is investigated for θ in [0, π/2] interval and for φ in [0, 2π] interval
and thus equation (5.24) has to be changed accordingly. For simulations in this
thesis, (5.25) was used

M
"N #
 π   2π  X X
Prad = U (θi , φj ) sinθi (5.25)
2N M j=1 i=1

The radiation intensity is calculated for θ and φ starting from 0◦ with a step of
1◦ . Therefore, the π/2 interval, for θ, is divided into 90 uniform intervals and
2π interval, for φ, is divided into 360 uniform intervals.
Chapter 6

Results and Discussion

In this chapter, different architectures and different element types for the array
antenna are being investigated. An evaluation of the Statistical Density Taper
method is done, using two amplitude distributions. For making a comparison
of the key performance indicators, a square array and a circular array were
simulated. A square array has NxN elements but a circular array has less. In
order to compare both arrays, the circular array will be refered to as a circular
array with NxN elements. The spacing of the elements will be investigated but
a 0.54λ distance will be used for simulations.
Three important types of element for array design will be studied. The isotropic
element is chosen because it represents the ideal case and the behavior of the
array can be seen easily. The half wavelength dipole is a common and popular
element, chosen because it is practical, it can be constructed and measured
with measuring devices. The third element is designed at Luleå University to
match the requirements of EISCAT 3D radar. In the following, an array with
isoptropic elements will be refered to as isotropic array and an array with Luleå
designed elements, an EISCAT array.
This chapter presents the relationship between the directivity and the spacing
to asses the difference in using a spacing of 0.54λ and a spacing that will give the
maximum directivity. The layout of the array is changed when a rectangular grid
is compared to an equilateral grid. Isotropic elements and EISCAT elements are
used for the array to find the ideal performance and compare it when practical
elements are used. After, a 625 square array with different elements is analyzed.
The statistical density taper is applied that will remove a different number of
elements to find how the array behaves with fewer components.
When the side lobe level is measured, it is not indicated to find it when the
scan is only in a few directions. Because the elements are statistically active
and the array doesn’t have a uniform distribution of components, when using a
thinning technique, grating lobes can appear in a certain directions when certain
scanning angles are used. For instance, architecture X with M◦ scanning angle
can perform better than architecture Y and N◦ scan angle and at the same
time worst than architecture Y but with a different scan angle R◦ . Hence, a set
of scanning angles will be used when analyzing the architecture, with θ and φ

35
36 Chapter 6. Results and Discussion

angles from 0 to 90◦ . A combination of 10 scanning angles will be used and the
mean values for the key performance indicators will be calculated and presented
in this thesis.
In this section, the side lobe level is calculated and presented as Average side
lobe and Peak side lobe. Both levels are calculated in dB and presented as
magnitude. The radiation pattern is normalised and a Peak level of e.g. -20 dB
should be regarded as 20 dB below the maximum intensity.

6.1 Directivity results

Directivity is an important factor of the antenna’s performance, showing how
well the signal is radiated into a certain direction. For this reason, an extra
care has to be taken when finding the directivity, to ensure that this param-
eter is calculated using an appropriate method and that the method leads to
results close to the theoretical ones. In section Implementation of directivity
the whole process of finding this parameter is explained and now the results will
be presented.
For the isotropic antenna, radiating the same energy in all direction, the direc-
tivity is one. The next type of investigated antenna is the half dipole element.
From [9], the directivity of this element is 1.643 in linear (no unit measure) or,
compared to the directivity of an isotropic antenna, 2.15 dB. The calculations
made with the implemented code showed a directivity of 1.6648 or 2.21 dB.
The values for the theoretical and simulated directivity are almost the same
and the conclusion that the code calculates in a correct way this parameter,
can be drown. Since only the electric field of the antenna designed by Luleå is
available, its directivity is calculated by simulation. It was found that it has a
directivity of 2.75 in linear and 4.4 in dB scale.
The spacing between elements is another important parameter that changes the
radiation pattern. This thesis proposes to analyze an array with 10000 elements
and three different types of elements, and the emphasize is on the array with
Luleå designed elements. With Luleå elements, a 10000 array is formed and
the spacing is varied to find the distance that will bring the highest directivity.
All elements have the same amplitude and no phase shift. In table 6.1 can be
seen that with increasing spacing, the directivity increases and the peak side
lobe level decreases. The average side lobe level has an almost constant value
showing that it is not affected by the relative distance between the elements.
However, the directivity has a maximum value for 0.6λ and even if the spacing is
increased, the directivity will not increase. After the maximum value is reached,
the directivity decreases with increasing relative distance. Opting for a distance
of 0.6λ is not the within EISCAT goals. On all array configurations that will
follow, a distance of 0.54λ will be used.
It was discussed earlier that the spacing between elements and the type of
grid has much to say on the radiation pattern. For a number of elements,
directivity is evaluated when using a rectangular grid and an equilateral grid.
The difference between the two types of grid is the
√
spacing on y axis. For the
hexagonal grid, the elements on y axis are placed 23 dx apart and on the x axis
6.2. Key performance indicators for a 625 elements array 37

Distance (λ) Directivity (dB) Average Sidelobe level (dB) Peak Sidelobe level (dB)
0.75 50.91 -19.54 -30.17
0.70 51.14 -19.55 -28.41
0.65 51.77 -19.58 -31.55
0.60 53.25 -19.6 -29.60
0.54 49.55 -19.53 -29.58
0.50 46.36 -19.41 -29.01

Table 6.1: Table of key performance indicators for

a 10000 square array, uniform amplitude, no phase shift

rectangular hexagonal
No. of elements Isotropic EISCAT Isotropic EISCAT
100 170.6 181.3 44.3 97.6
625 961 1173.3 275.5 619.5
2500 5155 5229 1121.6 2499.6
10000 87561 90347 5296.4 15159

Table 6.2: Directivity for rectangular and hexagonal grid.

they are placed like for the rectangular grid. However, the first elements on
two consecutive rows are spaced d2x apart. The directivity for the two grids is
presented in table 6.2.
As a general trend, with increasing elements in the array, the directivity also
increases. The hexagonal grid gives lower values than the rectangular grid,
because the elements on y axis are closer. When the number of components is
low, the directivity with hexagonal grid is around half times lower than with the
rectangular grid. For 10000 elements, using a rectangular grid is the best option
as the directivity is 9 times greater than with hexagonal grid. A hexagonal grid
is suited for a smaller array, when the directivity is still big enough compared
with rectangular grid.

6.2 Key performance indicators for a 625 ele-

ments array

For a comparison between the three types of element, a smaller array is chosen
since the characteristics of the radiation pattern are scalable when using more
elements. Analyzing the same array with the same excitation, distance and
phase shift but different antenna type, will give a good evaluation of how the
pattern is influenced by the type of radiators. The array has 625 elements placed
on a square grid, the same amplitude and no phase shift.
The directivity for all three type of elements is presented in figure 6.1, with
increasing thinning factor (increasing number of elements). The array with
EISCAT designed elements performs the best between the three of them. The
array with isotropic elements is close to the performance of the EISCAT array.
The performance of the array with dipole elements is the lowest of all three
38 Chapter 6. Results and Discussion

because it has a lower directivity of the single element, compared to the EISCAT
element. The half wavelength dipole has a directivity of 1.66 and the EISCAT
element has 2.75.

Directivity
30
25
20
[dB]

15 EISCAT elements

10 Isotropic elements

5 dipole elements

0
0,8 1 1,2 1,4 1,6
Thinning factor

Figure 6.1: Directivity for a square 625 planar array with different elements

In terms of average side lobe, the isotropic array and EISCAT array have almost
the same behavior as can be seen in figure 6.2. When less than half on the
elements are active in the array, the isotropic array shows a smaller level than
for EISCAT array. If the number of elements exceeds half of the filled array the
EISCAT array performs better - has a lower level.

Average Side lobe Level

0
0,8 1 1,2 1,4 1,6
-5

-10
EISCAT elements
[dB]

-15
Isotropic elements
-20 dipole elements

-25

-30
Thinning factor

Figure 6.2: Average side lobe level for a square 625 planar
array with different elements

Though the directivity for the isotropic array is smaller than the other two
arrays, it is advantageous in terms of peak side lobe, as can be seen in figure
6.3. However, an isotropic element is an ideal element which does not exist
in practical applications. Between the dipole and EISCAT array, the latter
performs better, with a difference of 6-7 dB less.
6.3. Influence of thinning parameters 39

Peak Side lobe Level

0
0,8 1 1,2 1,4 1,6
-5

-10 EISCAT elements

[dB]

Isotropic elements
-15
dipole elements
-20

-25
Thinning factor

Figure 6.3: Peak side lobe level for a square 625 planar
array with different elements

6.3 Influence of thinning parameters

The position of the elements in the array is found in relation to the Taylor
amplitude distribution. It is then natural to analyze how the amplitude dis-
tribution affects the placing of the elements and further the characteristics of
the array. Taylor’s amplitude can be adjusted with the two parameters defined
earlier, the n and design side lobe level. For high design side lobe levels, the
pattern for the linear array will have side lobes decreasing with a high steep.
The n parameter defines the number of equally high side lobes, in the vicinity
of the main lobe. The amplitude distribution is based on Taylor’s linear ampli-
tude. In this section, Taylor’s parameters will be varied and the changes will be
examined.

In figure 6.4 and 6.5 are presented the variations of the key performance indica-
tors when the n and design side lobe level are changed. For the 10000 elements
array, placed on a rectangular grid and with a thinning factor of 1.8, the Direc-
tivity and Side lobe level are presented. The first parameter n is investigated
for values starting from n = 3 and finishing with n = 6. The second parameter
is investigated for values starting from 25 dB until 40 dB with a 5 dB step.

Modifying n from 3 to 6 doesn’t bring much differences as can be seen. The

performance indicators are very similar and the same general trend can be seen
for all 4 values of n. The n parameter defines the region of equally high side
lobes and when it increases, the region with uniform lobes increases as well.
Taylor advises to choose a not too smal n and so, for a design SLL of 25 dB, n
should be chosen bigger than 3 and for a design SLL of 40 dB, n should be at
least 6. When the design side lobe level is too high and the n parameter is too
low, the relationship between them is not good. The design SLL influences the
directivity and side lobe the same way. With a higher design SLL, the excitation
gets steeper and lower toward the border of the array and a higher number of
elements will be removed.
40 Chapter 6. Results and Discussion

- =3
n
50
40
30
20
10 Average SLL
[dB] 0
Peak SLL
-10 -25 -30 -35 -40
-20 Directivity
-30
-40
-50
Thinning factor

50 n- =4
40
30
20
10 Average SLL
[dB]

0
Peak SLL
-10 -25 -30 -35 -40
-20 Directivity
-30
-40
-50
Thinning factor

Figure 6.4: Key performance characteristics with n = 3, n = 4 and design SLL

The peak SLL decreases with increasing design SLL, but decreases more when
the design SLL is in a good relationship with the n. When n is 3 or 4, the peak
SLL decreases for design SLL of 40 dB.

For all values of n, with increasing design side lobe level, the number of removed
elements increases and as a consequence the directivity decreases. But the
removed elements are more from the outer border, where the excitation is low
because of the design SLL. The parameters change simultaneous with the design
SLL and with increasing n, the difference between the performance indicators
is visible when the design SLL changes from 25 dB to 30 dB. The peak SLL
improves with increasing design SLL, but the average SLL decreases.

6.4 Array with 10000 elements

A 10000 elements square array, placed on a rectangular grid will be studied.

The distance between the elements on x and y axis is 0.54λ. The performance
is calculated over a range of 10 scanning angles that the value is not affected by
the direction of the main beam. The array is simulated with EISCAT designed
elements with a square aperture and a circular aperture, as presented in figure
6.6.
6.4. Array with 10000 elements 41

-
50 n =5
40
30
20
10 Average SLL
[dB]

0
Peak SLL
-10 -25 -30 -35 -40
-20 Directivity
-30
-40
-50
Thinning factor

50 n- =6
40
30
20
10 Average SLL
[dB]

0
Peak SLL
-10 -25 -30 -35 -40
-20 Directivity
-30
-40
-50
Thinning factor

Figure 6.5: Key performance characteristics with n = 5, n = 6 and design SLL

Figure 6.6: Thinned 10000 planar. Left: square array. Right: circular aperture

When thinning is applied to the 10000 elements array, the statistical function
that decides if an element will be part of the array or not, is based on Taylor’s
distribution of amplitude. Taylor defined a distribution for the amplitude for
a linear array and a distribution of amplitude for a circular aperture. Both
42 Chapter 6. Results and Discussion

distribution are used for thinning the array and the results are shown in figure
6.7 and 6.8.

In figure 6.7 are shown the values when the thinning factor is increased and
the thinning is based on Taylor’s linear amplitude. The elements are placed
on a rectangular grid. Like the case of 625 elements, the indicators change
sinchronous with the thinning factor and all three have a constant change. The
improvement of all three parameters appears when more elements are active in
the array, or the thinning factor is increased. The indicator that has the best
improvement is the average SLL. Compared to the case when the thinning factor
is 0.8, when it increases to 1.6, the growth in average SLL is 5.3 dB. This can be
explained by the incorporation of more active elements, to be exact, 1.7 times
more elements. The better value of directivity can be explained with the high
number of elements as well. It is true that for a thinned array, the directivity
is proportional to the number of elements.

Key performance indicators

50
40
30
20
10 Directivity
[dB]

0 Average SLL
-10 0,8 1 1,2 1,4 1,6
Peak SLL
-20
-30
-40
-50

Figure 6.7: Key performance indicators for a square 10000 planar array with
different thinning

For increasing thinning factor, the performance indicators are being evaluated
when the circular amplitude distribution is used. The array is placed on a
rectangular grid and has a circular aperture. This means that all elements
outside the circle with the origin in the middle of the rectangular array, are
inactive. The statistic taper which decides whether an element is active or
inactive is based on Taylor’s amplitude for circular apertures.

For this array, the general trend of the indicators is the same as when Taylor’s
distribution for linear array is used. With increasing thinning factor and ele-
ments, the values are improving. The average SLL improves with almost the
same amount from the lowest to the biggest thinnig factor, 5.4 dB. Interesting
to remind is that the array with circular aperture has from the beggining fewer
elements than the square array. If a filled aquare array has M 2 elements, then
a circular array on a rectangular grid has π2 M 2 elements. In table 6.4 is shown
the number of active elements in the square and circular array.
6.5. Influence of direction of main beam 43

Key performance indicators

50
40
30
20
10 Directivity
[dB]

0 Average SLL
-10 0,8 1 1,2 1,4 1,6
Peak SLL
-20
-30
-40
-50

Figure 6.8: Key performance indicators for a circular 10000 planar array with
different thinning

Thinning factor Square Circular

0.8 3925 3715
1 4889 4627
1.2 5744 5464
1.4 6440 6121
1.6 7025 6685

Table 6.3: Number of active elements in a thinned array.

6.5 Influence of direction of main beam

It was established that better results are obtained when applying the thinning
method to an array with a circular shape and use Taylor’s amplitude distribution
for circular apertures. A circular array with 625 elements is investigated when
the beam is steered into a direction of θ = 60◦ . The circular array is placed on
a square rectangular grid, which, when filled has 625 elements. With a circular
aperture, the array has 441 elements.

Unlike the case when the array is square, it can be observed that a circular
shape makes the EISCAT array perform better in terms of Directivity than
the isotropic array. Not applying a phase shift to the array, the main beam is
directed toward θ = 0◦ and φ = 0◦ also known as zenith, and the directivity
is the maximum for an isotropic array. Steering the beam into a direction of
θ = 60◦ will decrease the directivity.

A comparison between an array with isotropic elements and EISCAT elements

is done in terms of directivity and is presented in figure 6.9. The results for
the Directivity of the isotropic and EISCAT array are shown in figure 6.11. For
the isotropic array choosing φ for steering the main beam from 0 to 90◦ does
not influence the directivity in a strong manner. For the EISCAT array tough,
when φ = 70◦ a decrease in directivity can be seen. This can be explained
by the electric field of the EISCAT antenna, which has different magnitude for
different angles.
44 Chapter 6. Results and Discussion

Directivity
30
25
20

[dB]
15 EISCAT zenith

10 EISCAT (60°,20°)

5 EISCAT (60°,70°)

0
0,8 1 1,2 1,4 1,6
Thinning factor

Directivity
30
25
20
[dB]

15 isotropic zenith

10 isotropic (60°,20°)

5 isotropic (60°,70°)

0
0,8 1 1,2 1,4 1,6
Thinning factor

Figure 6.9: Directivity for 625 circular array. Left: EISCAT array.
Right: Isotropic array

The Peak side lobe level for the EISCAT and isotropic array is investigated and
the results are presented in figure 6.10. When the main beam is toward the
zenith, the Peak side lobe level has close values for both types of array and as
a general trend, this level improves with increasing thinnig factor. For the case
when half of the elements are active and the main beam is toward θ = 60◦ , the
Peak level declines for both arrays. For the isotropic array, this level improves
when the thinning factor is higher than 1.4. For the EISCAT array, the level
is improving with increasing number of elements or, equivalently, the thinning
factor.
To investigate if placing more elements in the array will change the influence of
the thinning factor, an array with 1600 elements is simulated. In the circular
array, 1184 elements are active. The results for the Directivity of the isotropic
and EISCAT array are shown in figure 6.11. Even with an increased number of
elements, the array behaves the same way as the array with 625 elements. The
Directivity for the main beam in direction of θ = 60◦ and φ = 70◦ is lower than
if the beam is into θ = 60◦ and φ = 20◦ direction.
The Peak side lobe level for both isotropic and EISCAT array are presented in
figure 6.12. It can be seen that for the isotropic array with almost half elements
active, the peak level is around the same value for different scanning angles.
As the number of elements increases and the direction of main beam is toward
θ = 60◦ and φ = 90◦ , the peak improves to a more negative value. The same
happens to the EISCAT antenna but when half the elements are active, the
array with main beam toward zenith has a better Peak.
Increasing the grid of the array from a 25x25 to 40x40, the general behavior of
6.5. Influence of direction of main beam 45

Peak Side lobe Level

0
0,8 1 1,2 1,4 1,6
-5

-10 EISCAT zenith

[dB]

EISCAT (60°,20°)
-15
EISCAT (60°,70°)
-20

-25
Thinning factor

Peak Side lobe Level

0
0,8 1 1,2 1,4 1,6
-5

-10 Isotropic zenith

[dB]

Isotropic (60°,20°)
-15
Isotropic (60°,70°)
-20

-25
Thinning factor

Figure 6.10: Peak side lobe level for 625 circular array. Left: EISCAT array.
Right: Isotropic array

the EISCAT array does not change. Directing the beam toward θ = 60◦ and
φ = 20◦ or φ = 70◦ has the same influence on the circular array. The Directivity
is better for φ = 20◦ than for φ = 70◦ and the Peak side lobe level improves for
φ = 20◦ .
46 Chapter 6. Results and Discussion

Directivity
35
30
25
20

[dB]
EISCAT zenith
15
EISCAT (60°,20°)
10
EISCAT (60°,70°)
5
0
0,8 1 1,2 1,4 1,6
Thinning factor
0
0,8 1
Directivity
35
30
25
20
[dB]

Isotropic zenith
15
Isotropic (60°,20°)
10
Isotropic (60°,70°)
5
0
0,8 1 1,2 1,4 1,6
Thinning factor

Figure 6.11: Directivity for 1600 circular array. Left: EISCAT array.
Right: Isotropic array

Peak Side lobe Level

0
0,8 1 1,2 1,4 1,6
-5

-10 EISCAT zenith

[dB]

EISCAT (60°,20°)
-15
EISCAT (60°,70°)
-20

-25
Thinning factor

Peak Side lobe Level

0
0,8 1 1,2 1,4 1,6
-5

-10 Isotropic zenith

[dB]

Isotropic (60°,20°)
-15
Isotropic (60°,70°)
-20

-25
Thinning factor

Figure 6.12: Peak side lobe level for 1600 circular array. Left: EISCAT array.
Right: Isotropic array
Chapter 7

Conclusions

Two thinning methods were presented in this thesis, that rely on the same
thinning method, the Statistical density taper, with the only difference in a
distribution of the amplitude used for removing the elements. One distribution
is the Taylor’s linear distribution, defined for linear arrays, but interpolated to
match a planar array. The other distribution is Taylor’s amplitude distribution
for circular apertures. When the linear distribution was used, the array was
squared, with 10000 elements, and when the other distribution was applied,
the array was circular, where the aperture was made circular by removing the
elements that were outside the specified radius, with 7888 elements.
The thinning method in a way dictates the number of elements the array will
have. In order to opt for the best thinning method, the number of elements
should be already chosen. The type of application dictates the range of best
values for a certain parameter and depending on which parameter is chosen, the
type of thinning can be selected. The performance of the antenna is connected
to how many active components the array has.
In particular, the Directivity parameter is highly sensitive to the number of
elements. This can be seen that with increasing factor k (that increases the
number of active elements) the directivity becomes closer to the value of a filled
10000 array with uniform amplitude.
When the relationship between the design SLL and n parameter is good, the
directivity has a higher value than if the relantionship is not so good. An
example of good relationship is when the design SLL is 25 dB and n is 6. The
two parameters however don’t influence that much the directivity as can be seen
that with increasing design SLL, the directivity doesn’t have a steep fall. The
Peak SLL has a steeper fall when the design SLL increases from 25 dB to 30
dB, but this improvement is however masked by the fact that the average SLL
decreases with increasing design SLL.
As the incoherent radar will need to investigate the ionosphere and look into
different directions, it was investigated the performance of the EISCAT and
isotropic array when the beam is directed toward θ = 60◦ . The results show
that the performance is linked to the direction of the main beam. When the

47
48 Chapter 7. Conclusions

phase shift is introduced to steer the main beam into θ = 60◦ direction, the
directivity falls and is smaller for a value of φ closer to 90◦ than closer to 0◦ .
When more than half of the elements are active, the peak side lobe level improves
for a value of φ closer to 0◦ .
A good outcome is that it was found that using an amplitude taper for a circular
aperture gives a better performance in terms of Directivity with fewer elements
than the square array. If the directivity is chosen to be the main parameter
of concern, selecting a circular aperture is the best option because it offers a
higher directivity and an economy of elements is done.
When the Average SLL is the parameter that is wanted to be improved, choosing
the square array is a good option because a high number of elements will give
a more negative SLL. The Peak SLL performs the same, for both types of
amplitude distribution.
Although the circular aperture performs well when it comes to the directivity
properties, it reaches a point where it can’t be improved. The elements will
not be removed because the amplitude taper is too high. For a square array,
the elements at the border have a lower value and they can be removed by
the statistical taper, but the circular aperture has the border elements removed
from the beginning, as a requirement for the circular shape of the array.
It can be concluded that thinning a circular array on a square grid with Taylor’s
amplitude for circular apertures is a good option.

7.1 Future work

In this thesis two types of grids were used to perform the simulations, rectan-
gular grid and hexagonal grid and the array was constructed with a square and
circular shape. These parameters are not the only ones that can be investigated
to evaluate the performance of an array, other non uniform grids can be used.
Possible architectures that can be investigated further are the sunflower archi-
tecture where the elements are concentrated in the middle of the area and at
the borders, they are spread out like the petals of the flower and the concentric
rings where the elements are placed along the length of circles with the origin
the middle of the area and different radiuses. These should be investigated in
particular with the EISCAT antenna in order to compare the results with a
uniform grid and check if the indicators are improving or not.
For an array as large as 10000 elements, for thinning it, Skolnik uses Taylors
circular distribution but Taylor defined other distributions for amplitude taper-
ing and an interesting parameter to investigate is the amplitude distribution
used for removing the elements. In this thesis Taylors distribution for linear
and circular arrays was used, but choosing Taylors one parameter distribution
is a valid option.
To study the mutual coupling a 3x3 antenna array was simulated in NEC2 where
only the middle element was fed. For this simulation the array was placed 2.8 m
above a ground plane and simulated with a mesh size of λ/20. As the radiation
pattern of the element in the middle was used in simulating the array, the results
7.1. Future work 49

include the coupling for one scenario. In order to fully understand how the
coupling affects the performance of the array and what can be done to diminish
this effect, other scenarios should be simulated in NEC and the resulting field
pattern should be used to compute the performance indicators.
References

[1] The ionosphere. [Link] Last ac-

cessed June 20, 2013.
[2] Earth’s atmosphere. [Link]
earth/[Link]. Last accessed June 20, 2013.
[3] Eiscat status october 2012. [Link]
files/ESFRI_EISCAT_web.pdf. Last accessed June 20, 2013.
[4] Eiscat 2005 anual report. [Link]
Documentation/AnnualReport/[Link]/view. Last
accessed June 20, 2013.
[5] Eiscat3d design specification document. [Link]
sites/default/files/P_S_D_7.pdf,LastaccesedJune2013. Last ac-
cessed June 20, 2013.
[6] Eiscat3d science case. [Link]
EISCAT3D_ScienceCase_v2.pdf. Last accessed June 20, 2013.
[7] Eiscat3d blog. [Link] Last accessed June 20, 2013.
[8] Alan W Rudge. The handbook of antenna design. 2 (1983), volume 2. Peter
Peregrinus Ltd, 1983.
[9] Constantine A Balanis. Antenna theory: analysis and design. Wiley-
Interscience, 1997.
[10] Robert J Mailloux. Phased array antenna handbook. Artech House Boston,
MA, 2005.
[11] Thomas T Taylor. Design of line-source antennas for narrow beamwidth
and low side lobes. Antennas and Propagation, Transactions of the IRE
Professional Group on, 3(1):16–28, 1955.
[12] Array antenna. [Link]
pdf. Last accessed June 20, 2013.
[13] EMS TECHNOLOGIES. Passive phased arrays for radar antennas. White
paper, EMS TECHNOLOGIES, December 2005.
[14] Kun-Chou Lee and Tah-Hsiung Chu. A circuit model for mutual coupling
analysis of a finite antenna array. Electromagnetic Compatibility, IEEE
Transactions on, 38(3):483–489, 1996.

51
52 References

[15] Gunnar Isaksson. private conversation.

[16] Harry L Van Trees. Detection, Estimation, and Modulation Theory, Opti-
mum Array Processing. Wiley-Interscience, 2004.
[17] Rajashree Jain and GS Mani. Solving antenna array thinning problem using
genetic algorithm. Applied Computational Intelligence and Soft Computing,
2012:24, 2012.
[18] ML Skolnik, J Sherman III, and F Ogg Jr. Statistically designed
density-tapered arrays. Antennas and Propagation, IEEE Transactions on,
12(4):408–417, 1964.
[19] Robert J Mailloux and Edward Cohen. Statistically thinned arrays with
quantized element weights. Antennas and Propagation, IEEE Transactions
on, 39(4):436–447, 1991.
[20] Eyung W Kang. Radar system analysis, design, and simulation. Artech
House Publishers, 2008.
[21] T Taylor. Design of circular apertures for narrow beamwidth and low
sidelobes. Antennas and Propagation, IRE Transactions on, 8(1):17–22,
1960.
Appendix A

Appendix

In order to calculate the key performance indicators for different types of array,
the radiation pattern is calculated from the values of the electric field of the
array. To find the electric field of the array, the product between the electric
field of the single antenna and the electric field of the array factor is calculated.
To find the electric field of the array factor, presented in equation 2.3, the
function arrayf is implemented.
To find the amplitude distribution, according to Taylor, the gg function was
implemented for a linear array and taylorc irc function for the circular aperture
array.
The main code calculates the field pattern for the EISCAT antenna and dipole
antenna, finds the radiation pattern and the key performance indicators. The
indicators are found averaging the values for 10 different scanning angles and so
in the code, a loop is implemented in order to change the scanning angle. The
code can be changed to have a loop where the thinning factor is changing.
When generating a matrix with random numbers, in order to calculate the
amplitude distribution, different seeds are used. The seed is used to choose the
sequence of random numbers Matlab has implemented in its core. To make a
correct comparison, when thinning the array, the same sequence of seeds is used
or the same random matrix when thinning all 3 types of array.
Below are presented the functions used in this thesis.
1. Function to calculate the linear Amplitude taper, according to Taylor 2.3
function[output]=gg(x) %Finding taylor linear taper

%Calculating F(m,A,nbar)
%Design SLL is 25 dB, nbar is 5

nbar=5;
sll = (10^(-25/10)); % Design SLL of Taylor in linear

53
54 Appendix A. Appendix

ATaylor=acosh(sqrt(1/sll)) / pi;
sigma=nbar/(sqrt(ATaylor^2+(nbar-0.5)^2));

for m=1:nbar
z(m)=sigma*sqrt(ATaylor^2+(m-0.5)^2); %the roots of the function

end

%Calculate the coefficients of the functions

for m=1:nbar-1
po=1;
for n=1:nbar-1

po=po*(1-m^2/z(n)^2);
po1=[po,n,m];

end
f(m)=po*(factorial(nbar-1))^2/(factorial(nbar-1+m)*factorial(nbar-1-m));
s=(factorial(nbar-1))^2/(factorial(nbar-1+m)*factorial(nbar-1-m));

end

s=0;
for m=1:nbar-1

s=s+f(m)*cos(m*pi*x);

end

output=1+2*s;
2. Function to calculate the amplitude distribution for circular apertures, ac-
cording to Taylor
function[output]=taylor_circ(Nele,x)%Finding taylor circular taper

%Calculating F(m,A,nbar)

nbar=5;
sll = (10^(-25/10)); % Design SLL of Taylor
ATaylor=acosh(sqrt(1/sll)) / pi;

%parameter defined by Taylor

miu = [ 1.2196699 2.2331306 3.2383155 4.2410629 5.2439216 ...
6.2439216 7.2447598 8.2453948 9.2458927 10.2462933 ...
11.2466228 12.2468985 13.2471325 14.2473337 15.2475086 ...
16.2476619 17.2477974 18.2479181 19.2480262 20.2481237];
 55

sigma = miu(nbar) / sqrt(ATaylor^2 + (nbar - 0.5)^2);

%Calculate the coefficients of the function

for m=1:nbar-1
p1=1;
for i=1:nbar-1
p1=p1*(1-miu(m)^2/(sigma^2*(ATaylor^2+(i-0.5)^2) ) );
end
p2=1;
for j=1:nbar-1
if j~=m
p2=p2*(1-miu(m)^2/miu(j)^2) ;
end
end
F(m)=-besselj(0, pi*miu(m))*p1/p2;
end

%Calculate the function for each distance x from the center of the circle
gg=1;
for m = 1 : nbar-1
gg = gg+besselj(0, pi * miu(m) * x)*F(m)/ besselj(0, pi * miu(m))^2;
end

output=gg;

3. Function to calculate the Array Factor

function[output]=arrayf(A,Nele,x,y,theta0,phi0)
%A is the amplitude of elements, a matrix
%Nele is the number of elements on x and y axis
%x is the theta angle where the pattern is calculated
%y is the phi angle where the pattern is calculated
%theta0 is the theta angle of the main beam
%phi0 is the phi angle of the main beam

%k=1;
%distance between elements on x axis expressed in multiples of lambda
eledistx=0.54;

%distance between elements on y axis expressed in multiples of lambda

eledisty=0.54;
%eledisty=0.467; % for hexagonal grid
Betax=-2*pi*eledistx*sin(theta0)*cos(phi0); %phase shift
Betay=-2*pi*eledisty*sin(theta0)*sin(phi0);

% rectangular grid
output=0;
56 Appendix A. Appendix

for i=1:Nele
for k=1:Nele
if A(i,k)~=0
output=output+A(i,k)*exp(j*(i-1)*(2*pi*eledistx*cos(y)*sin(x)+Betax))...
*exp(j*(k-1)*(2*pi*eledisty*sin(x)*sin(y)+Betay));

% for hexagonal grid

%output=output+A(i,k)*exp(j*(0.27*mod(i,2)+(i-1)*eledistx)*(2*pi*cos(y)...
*sin(x)+Betax))*exp(j*(k-1)*(2*pi*eledisty*sin(x)*sin(y)+Betay));
end
end
end
output=abs(output);

.4. Main function to calculate the Radiation Pattern

clear all
Nele=100; % No. of elements on x and y axis
A=ones(Nele);
middle=(Nele-1)/2;

% for building the Matrix with amplitude taper for circular aperture
%for each position calculate the distance to the middle of the array

for m=1:Nele
for n=1:Nele
dist=sqrt((middle-m+1)^2+(middle-n+1)^2);

A(m,n)=taylor_circ(Nele,dist/middle);
if dist>middle %remove the elements outside the circle
A(m,n)=0;
end
end
end

% for building the Matrix with amplitude taper for planar array
%multiplying the amplitude of the linear array from x axis to
%the amplitude of the y axis.

% the length of the array is normalized and the

%borders are 1 and 1, and the middle is 0
%constructing Taylor planar amplitude

% xx=-1:2/(Nele-1):1;
%
% v=gg(xx)/gg(0); %normalize the amplitude to the maximum
%
% for i1=1:Nele
%
 57

% for j1=1:Nele
% vs(i1,j1)=v(i1)*v(j1);
% end
% end
% A=vs;

A=A/max(max(A));

%read from the NEC2 file

fid = fopen(’radiation [Link]’);
C = textscan(fid, ’%d %d %f %f %f %f %f %s %f %f %f %f’);

fclose(fid);
radi=0.017453292519943; % one radian
thetan = C{1}; %the theta values from the NEC2 file
phin = C{2};
absotheta = C{9}; % the magnitude of E_theta
phasetheta = C{10}; % the phase of E_theta
absophi = C{11};
phasephi = C{12};
Etheta=absotheta.*exp(j*phasetheta);
Ephi = absophi.*exp(j*phasephi);
efield = sqrt((abs(Etheta).^2 + abs(Ephi).^2)/(240*pi));

%transform the NEC2 output into a matrix

for i = 1:91
for j=1:360
rad(i,j) = efield(i+j-1+90*(j-1));

end

end

%The NEC2 file has the value of the field calculated

%for the following values of theta and phi

theta = deg2rad([Link]);
phi = deg2rad([Link]);
% [theta2 phi2 ] = meshgrid(phi,theta);
% [x,y,z] = sph2cart(theta2,phi2,rad); %plot the pattern
% figure(1)
% surf(x,y,z)

for i = 1:91
for j=1:360
Edipol(i,j)=60*cos(pi/2*cos(theta(i)) )/sin(theta(i));
if Edipol(i,j)==Inf Edipol(i,j)=0;
end
end
58 Appendix A. Appendix

end

%Array factor

%the values of theta and phi for the direction of main beam
%make the average of scanning angles
thet0=[0 0 20 20 30 30 40 40 50 60];
ph0=[0 20 30 30 40 40 50 60 70 80];
%k1=[0.8 1 1.2 1.4 1.5]; % the thinning factor when one seed
k1=1.2;
sedall=1:10; % the seed to generate the random matrix
iteration_nr=10;
vect=1:10; %vector with number of elements for each iteration

%do a loop to average the indicators for different scanning angles

for iterat=1:iteration_nr
elenr=0;
%mySeed = 1; %this use when only one seed is wanted

mySeed = sedall(iterat);
rng(mySeed); % Set the seed
rand([Nele,Nele]);

rng(mySeed); % Set the seed again to duplicate the results

m=rand([Nele,Nele]);

%construct the Amplitude matrix where 0 means the element is not active
for i1=1:Nele

for j1=1:Nele

if m(i1,j1)/k1<=A(i1,j1) B(i1,j1)=1; elenr=elenr+1;

else B(i1,j1)=0;
end
end

end

%Calculate the Array Factor

for i = 1:91
for j=1:360
ARF(i,j) = arrayf(B,Nele,(i-1)*radi,(j-1)*radi,thet0(iterat),ph0(iterat));

end

end
 59

vect(iterat)=elenr; %vector with active elements for each iteration

% figure(2) %the distribution of active elements

% spy(B)
%
% [xrf,yrf,zrf] = sph2cart(theta2,phi2,ARF);
%
% figure(3)
% surf(xrf,yrf,zrf)

%The normalised RADIATION PATTERN

maxpat=max(max(ARF));
patt=(ARF/maxpat).^2; %for isotropic array

% maxpat=max(max(rad.*ARF));
% patt=(rad.*ARF/maxpat).^2; %for EISCAT array

% maxpat=max(max(Edipol.*ARF));
% patt=(Edipol.*ARF/maxpat).^2; %for array with dipoles

%Calculate directivity
summ=0;
for j = 1:360
for i=1:90

summ=summ+patt(i,j)*sin(theta(i));

end

end

[maxxi,idx]=max(patt(:));
[mtheta,mphi]=ind2sub(size(patt),idx) % the position of the maximum

summ;
prad=summ*pi/360*2*pi/90; %radiated power
D(iterat)=4*pi/prad %Directivity

% [xpat,ypat,zpat] = sph2cart(theta2,phi2,patt);
%
% figure(4)
% surf(xpat,ypat,zpat)

sumefield=0; %sum of all radiation pattern

sumoutbeam=0; %sum of radiation pattern lower than half the main beam
nrefield=0;
nroutbeam=0;
for i=1:91
60 Appendix A. Appendix

for j=1:360
sumefield=sumefield+patt(i,j);
nrefield=nrefield+1;
if patt(i,j)<0.5
sumoutbeam=sumoutbeam+patt(i,j);
nroutbeam=nroutbeam+1;
end
end
end

%calculate position of dips in the up, down, left, right

%direction from the maximum
anteriorleft=1.1;
for jleft=mphi:-1:1
if patt(mtheta,jleft)<anteriorleft
anteriorleft=patt(mtheta,jleft);
posanteriorleft=jleft;
else
break
end
end

anteriorright=1.1;
for jright=mphi:1:360
if patt(mtheta,jright)<anteriorright
anteriorright=patt(mtheta,jright);
posanteriorright=jright;
else
break
end
end
anteriorup=1.1;
for iup=mtheta:-1:1
if patt(iup,mphi)<anteriorup
anteriorup=patt(iup,mphi);
posanteriorup=iup;
else
break
end
end

anteriordown=1.1;
for idown=mthet[Link]
if patt(idown,mphi)<anteriordown
anteriordown=patt(idown,mphi);
posanteriordown=idown;
else
break
end
end
 61

sumbeam=0; %sum of radiation pattern in the main lobe

nrbeam=0;
for im=posanteriorup:posanteriordown
for jm=posanteriorleft:posanteriorright
sumbeam=sumbeam+patt(im,jm);
nrbeam=nrbeam+1;
end
end

avgoutbeam(iterat)=10*log10((sumefield-sumbeam)/(nrefield-nrbeam))

%calculate peak as the first maximum outside the main beam

peak=0;
for i1=1:91
for j1=1:360
if i1<posanteriorup | i1>posanteriordown
if j1<posanteriorleft | j1>posanteriorright
if patt(i1,j1)>peak
peak=patt(i1,j1);
pospeak=[i1,j1];
end
end
end
end
end
peakdB(iterat)=10*log10(peak)
pospeak

end
D1=sum(D)/10
avgoutbeam1=sum(avgoutbeam)/10
peakdB1=sum(peakdB)/10
save(’[Link]’,’vect’,’D’,’avgoutbeam’,’peakdB’)

..
TRITA-ICT-EX-2013:171

[Link]