________________________________________________________________________________

Design of a wideband conformal array antenna system with

beamforming and null steering, for application in a DVB-T
based passive radar




Master of Science Thesis

Vedaprabhu Basavarajappa
Student number: 4121910
03-July-2012


Supervisors:
Dr. Massimiliano Simeoni and Dr. Peter Knott







_________________________________________________________________________________
2





























3


Delft University of Technology
Department of Telecommunications

The undersigned hereby certify that they have read and recommend to the Faculty of
Electrical Engineering, Mathematics and Computer Science for acceptance a thesis entitled
Design of a wideband conformal array antenna system with beamforming and null
steering, for application in a DVB-T based passive radar, by Vedaprabhu Basavarajappa in
partial fulfilment of the requirements for the degree of Master of Science in Electrical
Engineering.


Dated: 03-July-2012
Committee members

____________________________
Prof. Dr. A. Yaravoy (Chair)
____________________________
Dr. M. Simeoni
____________________________
Dr. P. Knott
_____________________________
Dr. Ir. B. J. Kooij
____________________________
Dr.ing. I.E. Lager





4

Acknowledgments

I would like to express my sincerest thanks to Dr. Peter Knott who was my supervisor at
Fraunhofer FHR for his constant support and encouragement. I am very grateful to my
supervisor at TU Delft, Dr. Massimiliano Simeoni for having provided this opportunity and for
his constant support and motivation and for providing valuable feedback to my work at
Fraunhofer FHR. I would also like to extend my heartiest thanks to my graduation professor
Prof. Dr. Alexander Yaravoy for his support and for his valuable feedbacks. I would also like
to thank Dr. Daniel O Hagan of PSK group at FHR for his valuable inputs.
I would like to extend my thanks to my friends at Fraunhofer FHR for having provided me
with a wonderful atmosphere where I could carry out my thesis. I also thank my friends at
TU Delft for their support.
Last but not least, l would like to offer my heartiest thanks to my family whose constant
encouragement and support played a vital role.
Finally, I would like to thank everyone involved both directly and indirectly in the completion
of my thesis.

Vedaprabhu Basavarajappa
Wachtberg, Germany
04 June, 2012











5




























6

Abstract

The use of radars in detecting low flying, small targets is being explored for several decades
now. However radar with counter-stealth abilities namely the passive, multistatic, low
frequency radars are in the focus recently. Passive radar that uses Digital Video Broadcast
Terrestrial (DVB-T) signals as illuminator of opportunity is a major contender in this area. A
DVB-T based passive radar requires the development of an antenna array that performs
satisfactorily over the entire DVB-T band. At Fraunhofer FHR, there is currently a need for an
array antenna to be designed for operation over the 450-900 MHz range with wideband
beamforming and null steering capabilities. This would add to the ability of the passive radar
in detecting covert targets and would improve the performance of the system. The array
should require no mechanical adjustments to inter-element spacing to correspond to the
DVB-T carrier frequency used for any particular measurement. Such an array would have an
increased flexibility of operation in different environment or locations.
The design of such an array antenna and the applied techniques for wideband beamforming
and null steering are presented in the thesis. The interaction between the inter-element
spacing, the grating lobes and the mutual couplings had to be carefully studied and an
optimal solution was to be reached at that meets all the specifications of the antenna array
for wideband applications. Directional beams, nulls along interference directions, low
sidelobe levels, polarization aspects and operation along a wide bandwidth of 450-900 MHz
were some of the key considerations.












7





























8

Table of Contents

ACKNOWLEDGMENTS 4
ABSTRACT 6
TABLE OF CONTENTS 8
LIST OF FIGURES 10
LIST OF ACRONYMS 13
LIST OF SYMBOLS 14
CHAPTER 1- INTRODUCTION 15
1.1 About passive radar - A Brief history of passive radar 16
1.2 Motivation for the type of antenna design for passive radar 18
1.3 Identification of the problem 18
1.4 Challenge of the work and methodology adopted 19
1.5 Outline of the thesis 19
CHAPTER 2 THEORETICAL BACKGROUND 20
2.1. Design of the discone antenna element 20
2.2 Array antenna theory 22
2.3 Wideband beamforming and null steering 38
CHAPTER 3- BEAMFORMING ALGORITHMS AND THEIR COMPARISON 45
3.1 Conventional beamforming algorithms 45
3.2 The Bucci beamforming algorithm 47
3.3 The advantage of Bucci beamforming algorithm 49
CHAPTER 4- THE CORA RADAR SYSTEM AND ANTENNA ARRAY 50
4.1 Outline of the CORA radar system 50
9

4.2 The architecture of the array antenna 51
CHAPTER 5 SIMULATION STUDIES ON BEAMFORMING AND NULL STEERING 53
5.1 Preliminary studies with dipole circular antenna array 53
5.2 Uniform Circular Array of Discone antennas 62
5.3 Beamforming and null steering in 4NEC2 64
5.4 Need for CST Microwave studio validation 69
5.5 CST Microwave studio analysis 69
5.6 Discone antenna array designed using CST Microwave Studio 71
5.7 Supporting structure and its influence on the antenna performance 77
5.8 Polarimetric beamforming 80
CHAPTER 6- EXPERIMENTAL VALIDATION 83
6.1 Measurement results of Discone antenna 83
6.2 Experimental set up 84
CHAPTER 7-CONCLUSIONS AND FUTURE WORK 86
7.1 Conclusions 86
7.2 Future work 87
BIBLIOGRAPHY 88
PUBLICATION 91














10


List of Figures


FIGURE 1. CLASSICAL MONOSTATIC RADAR PRINCIPLE ........................................................................................ 15
FIGURE 2 . A PASSIVE RADAR OPERATING WITH SINGLE TARGET AND MULTIPLE TRANSMITTERS. .................... 16
FIGURE 3. DAVENTRY EXPERIMENT [5] ................................................................................................................. 16
FIGURE 4. PCR SENSOR CORA OF FRAUNHOFER FHR [5] ...................................................................................... 17
FIGURE 5. DISCONE ANTENNA AND PARAMETERS [9] .......................................................................................... 21
FIGURE 6. RETURN LOSS OF THE DISCONE ANTENNA [9] ..................................................................................... 21
FIGURE 7. LINEAR AND PLANAR ARRAY DEPICTION .............................................................................................. 22
FIGURE 8. TWO DIPOLES PLACED ALONG THE Z AXIS ........................................................................................... 23
FIGURE 9. TWO DIPOLES PLACED ALONG THE Z AXIS UNDER FARFIELD CONDITION ........................................... 24
FIGURE 10 ILLUSTRATION OF PRINCIPLE OF PATTERN MULTIPLICATION WHEN D= /4 AND = -90 [10] ......... 26
FIGURE 11. N ELEMENT ARRAY ............................................................................................................................. 26
FIGURE 12. UNIFORM LINEAR ARRAY ................................................................................................................... 27
FIGURE 13. ARRAY FACTOR PLOT FOR N=10 ......................................................................................................... 28
FIGURE 14. PHASED ARRAY RADIATION PATTERN WITH

30
0
= u , 0 . 3 / = d AND N=20 [39] ....................... 29
FIGURE 15. SCHEMATIC REPRESENTATION OF PHASED ARRAY [33] .................................................................... 30
FIGURE 16. PLANAR ARRAY DEPICTION ................................................................................................................ 31
FIGURE 17. PLANAR BROADSIDE ARRAY PATTERN [39] ....................................................................................... 32
FIGURE 18. PLANAR ARRAY PATTERN WITH BEAM ALONG ) , (
0 0
| u = ) 45 , 30 (

[39] ........................................ 32
FIGURE 19. UNIFORM CIRCULAR ARRAY [10]........................................................................................................ 33
FIGURE 20. LOCAL AND FAR FIELD COORDINATES OF THE CIRCULAR ARRAY [40] ............................................... 35
FIGURE 21. PART OF CIRCULAR ARRAY [40] .......................................................................................................... 36
FIGURE 22. GRATING LOBES DEPENDENCE ON INTER ELEMENT SPACING ......................................................... 37
FIGURE 23. ARRAY WITH FOUR SENSORS AND TWO IMPINGING SIGNALS .......................................................... 38
FIGURE 24. A PLANE WAVE PROPAGATING IN THE DIRECTION OF Z AXIS ........................................................... 39
FIGURE 25. GENERAL STRUCTURE OF NARROWBAND ARRAY ARCHITECTURE [24] ............................................. 41
FIGURE 26. A SIMPLE ON-BEAM RF BEAM FORMER[20]....................................................................................... 42
FIGURE 27. DIGITAL BEAMFORMER ARCHITECTURE [24] .................................................................................... 43
FIGURE 28. ILLUSTRATION OF NULL INTRODUCTION AND BEAMFORMING [40] ................................................. 44
FIGURE 29. THE MASKS AND THE BEAMFORMED PATTERN. ................................................................................ 48
FIGURE 30 CORA SYSTEM ARCHITECTURE [30] .................................................................................................... 50
FIGURE 31. DIPOLE ANTENNA DESIGNED IN 4NEC2. ........................................................................................... 54
FIGURE 32. RETURN LOSS (S11) AND VSWR OF THE DESIGNED DIPOLE ELEMENT. ............................................. 55
FIGURE 33. HORIZONTAL AND VERTICAL PATTERNS OF THE DESIGNED DIPOLE ELEMENT. ................................ 55
FIGURE 34. 3D RADIATION PATTERN DEPICTING THE TOTAL GAIN IN DBI OF THE DESIGNED DIPOLE ELEMENT. 56
FIGURE 35. GEOMETRY OF 11 ELEMENT CIRCULAR ARRAY BUILT IN 4NEC2 ...................................................... 57
FIGURE 36 RADIATION PATTERN IN THE HORIZONTAL AND VERTICAL PLANE AT THE CENTER FREQUENCY ..... 57
FIGURE 37. 3D EMBEDDED RADIATION PATTERN ................................................................................................. 58
FIGURE 38. EMBEDDED ELEMENT PATTERN ....................................................................................................... 59
FIGURE 39. SUPERPOSITION OF FIELDS DUE TO ROTATION OF THE ELEMENT PATTERN. ................................... 59
FIGURE 40. ARRAY MANIFOLD FOR HALF LAMBDA SPACING ( AMPLITUDE AND PHASE VS ELEMENT). .............. 60
FIGURE 41. MAIN BEAMS STEERED ALONG 56 AND 206 DEGREES USING THE MAXIMUM GAIN METHOD. ...... 60
FIGURE 42. BUCCI BEAMFORMING PERFORMED ON THE CIRCULAR DIPOLE ARRAY AT 675 MHZ ...................... 61
FIGURE 43: NULL INTRODUCED AT 50 AND 271 DEGREES AT THE DESIRED SIDELOBE LEVEL OF 20 DB .............. 61
FIGURE 44. THE WIRE MODEL AND THE 3D RADIATION PATTERN OF DISCONE ANTENNA USED IN 4NEC2 ........ 62
11

FIGURE 45. THE HORIZONTAL AND VERTICAL RADIATION PATTERN OF THE DISCONE ANTENNA ..................... 63
FIGURE 46. THE
u
E AND
|
E POLARIZATION COMPONENTS OF THE DISCONE ANTENNA . .............................. 63
FIGURE 47. THE RETURN LOSS CHARACTERISTICS OF THE EMBEDDED DISCONE ANTENNA ELEMENT ............... 64
FIGURE 48. FIGURE DEPICTING THE TOP VIEW OF THE DISCONE CIRCULAR ARRAY MODEL ............................. 64
FIGURE 49. BUCCI BEAMFORMING PERFORMED AT SELECTED FREQUENCIES. ................................................... 65
FIGURE 50. BUCCI BEAMFORMING PERFORMED OVER THE WIDEBAND FROM 450-900 MHZ ........................... 66
FIGURE 51. NULL STEERING PERFORMED ON THE DISCONE ANTENNA ARRAY AT 50 AND 271 DEGREES. .......... 66
FIGURE 52. NULL DEPENDENCE OVER FREQUENCY WITH COEFFICIENTS CALCULATED FOR 675MHZ APPLIED
OVER THE 450-900 MHZ BAND. NULL POSITION AT 230/675 MHZ. ........................................................... 67
FIGURE 53. ARRAY COEFFICIENTS AMPLITUDE AND PHASE. ................................................................................ 68
FIGURE 54. 11 ELEMENT CIRCULAR ARRAY ........................................................................................................... 68
FIGURE 55. EMBEDDED RADIATION PATTERN OF ELEMENT VS FREQUENCY ....................................................... 69
FIGURE 56. DISCONE ANTENNA ARRAY MODEL DESIGNED IN CST MWS. ARRAY DIAMETER IS 77 CM. THE TYPE
OF FEEDING IS COAXIAL FEED FROM BOTTOM OF DISCONE. THE CENTER FREQUENCY OF OPERATION IS
675 MHZ. ...................................................................................................................................................... 72
FIGURE 57(A). TOP VIEW OF THE ARRAY ANTENNA SHOWING A MAGNETIC SYMMETRY PLANE PASSING
THROUGH THE CENTER OF THE ARRAY ALONG ITS DIAMETER.(B) THE MAGNETIC FIELD IN THE
FUNDAMENTAL MODE OF COAXIAL CABLE .................................................................................................. 72
FIGURE 58. THE EMBEDDED RADIATION PATTERN OF THE SINGLE ELEMENT ANTENNA ALONG THE AZIMUTHAL
PLANE
) 360 0 , 90 (

< < = | u
AT 450, 675 AND 900 MHZ WITH THE CO AND CROSS POLARIZATION LEVELS IS
SHOWN IN (A), (B) AND (C) RESPECTIVELY. .................................................................................................. 73
FIGURE 59. THE ARRAY MANIFOLD AMPLITUDE AND PHASE PLOTTED VS THE ELEMENT NUMBER. THE
ROTATION OF THE FILED AMPLITUDES AND PHASES ALONG WITH THE ELEMENT NUMBER IS CLEARLY
SEEN. THE ROTATION IS BIN STEPS OF 360/11 DEGREES. ............................................................................ 74
FIGURE 60. A,B,C AND D SHOW THE BUCCI BEAMFORMING PERFORMED USING EMBEDDED ELEMENT
PATTERN OBTAINED USING CST MWS AT 450, 675, 875AND 900 MHZ RESPECTIVELY ............................... 75
FIGURE 61. BUCCI BEAMFORMING PERFORMED OVER THE WIDEBAND FROM 450-900 MHZ IN STEPS OF 25
MHZ. ............................................................................................................................................................. 76
FIGURE 62. NULL STEERING PERFORMED ON THE DISCONE ANTENNA ARRAY AT 145 AND 251 DEGREES USING
CST MWS RADIATION PATTERNS ................................................................................................................. 76
FIGURE 63. (A) BLINDSPOT SEEN AROUND 600 MHZ USING 4NEC2 EMBEDDED PATTERN. (B) THE BLINDSPOT
DISAPPEARS AROUND 600 MHZ ON USING CST MWS EMBEDDED PATTERN. ............................................. 77
FIGURE 64. SINGLE ELEMENT OF THE SUPPORT STRUCTURE WITH PARALLELOPIPED ALUMINIUM SUPPORTS
AND PVC CONNECTORS FROM DISCONE TO ALUMINIUM SUPPORT. .......................................................... 78
FIGURE 65. SUPPORT STRUCTURE FOR THE ARRAY ANTENNA BUILT USING ELEMENT IN FIGURE 64. ................ 78
FIGURE 66. 3D RADIATION PATTERN OF THE ARRAY ANTENNA WITH SUPPORT STRUCTURE ............................. 78
FIGURE 67. THE S11, S12 AND S13 FOR A SECTION OF THE ARRAY WITHOUT THE SUPPORTING STRUCTURE. ... 79
FIGURE 68. THE S11, S12 AND S13 FOR A SECTION OF THE ARRAY WITH THE SUPPORTING STRUCTURE. .......... 79
FIGURE 69. COMPARISON OF THE RADIATION PATTERNS WITHOUT AND WITH THE SUPPORT STRUCTURE AT A)
450 MHZ B) 675 MHZ AND C) 900 MHZ. IT CAN BE CONCLUDED THAT THE RADIATION PATTERN IS
COMPARABLE AND DOES NOT DEVIATE MUCH FROM ITS NORMAL PATTERN WHEN A SUPPORT
STRUCTURE IS USED. .................................................................................................................................... 80
FIGURE 70. POLARIMETRIC BEAMFORMING WITH SAME ARRAY EXCITATION COEFFICIENTS FOR CO-POLAR AND
CROSS POLARIZATIONS AT A) 450 MHZ B) 675 MHZ C) 875 MHZ ................................................................ 81
FIGURE 71. POLARIMETRIC BEAMFORMING INVOLVING CONCATENATION OF THE COPOLAR ARRAY MANIFOLD
MATRIX WITH THE CROSS POLAR ARRAY MANIFOLD MATRIX FOR FREQUENCIES A) 450 MHZ B) 675 MHZ
C) 900 MHZ ................................................................................................................................................... 82
FIGURE 72. THE FABRICATED DISCONE ANTENNA ................................................................................................ 83
FIGURE 73. THE MEASURED RETURN LOSS CHARACTERISTICS OF THE ANTENNA [8] .......................................... 83
12

FIGURE 74. THE MEASURED ANTENNA RADIATION PATTERN IN THE ELEVATION PLANE FOR DISCONE WITH
VERTICAL POLARIZATION AT THREE DIFFERENT FREQUENCIES [8]. ............................................................. 84
FIGURE 75. 11 DISCONE ANTENNA ELEMENTS ..................................................................................................... 85
FIGURE 76 (A). MOUNT FOR THE ARRAY ANTENNA IN THE ANECHOIC CHAMBER (B) THE PROBE ANTENNA ..... 85
FIGURE 77. CONTROL ROOM FOR THE ANECHOIC CHAMBER. ............................................................................. 85














































13

List of acronyms



ADC Analog to Digital Converter
AF Array Factor
CORA COvert RAdar
CST Computer Simulation Technology
DAB Digital Audio Broadcast
DOA Direction Of Arrival
DVB-T Digital Video Broadcast Terrestrial
FFT Fast Fourier Transform
FIT Finite Integration Technique
FM Frequency Modulation
FPGA Field Programmable Gate Array
GPU General Processing Unit
HF High Frequency
LMS Least Mean Squares
LNA Low-Noise Amplifier
MoM Method of Moments
MWS Microwave Studio
NEC Numerical Electromagnetic Code
NLMS Normalized Least Mean Squares
PCL Passive Coherent Location
PCR Passive Covert Radar
RCS Radar Cross Section
RDE Radar Data Extractor
RLS Recursive Least Squares
RF Radio Frequency
SIR Signal to Interference Ratio
SLL Side Lobe Level
SMI Sample Matrix Inversion
TR Transmit-Receive
UCA Uniform Circular Array
UHF Ultra High Frequency
VHF Very High Frequency














14

List of symbols

| Phase difference
d Spacing between antenna elements
t
E

Total electric field

k Wavenumber
u Elevation angle
| Azimuthal or horizontal angle
Wavelength
AF Array factor
M Number of elements on x-axis of planar array
N Number of elements on y-axis of planar array
0
D Maximum directivity
a Circular ring radius
) (t u
N
Received signal in the
th
N antenna
R(t) Covariance matrix
R

Correlation matrix
) (n e Error function
) (x E Expectation of x
Step size
) (u M
L
Lower mask
) (u M
U
Upper mask
E Field matrix























15

Chapter 1- Introduction

Radars have long been used for detection and tracking of flying targets. There are numerous
examples in the literature to substantiate the fact that a successful detection and tracking of the
target is possible to the highest accuracies. However many of these active radars are prone to be
detected as they inherently have to transmit signals to detect a target. Thus active radars are
vulnerable to localisation due to the fact that they cannot operate without transmitting energy,
though their need in the military as means to detect enemy air or naval targets cannot be
undermined. A radar location could be determined by receiving the radar signals and using
triangulation. The location being known could lead to jamming and destruction of radar systems.
Thus the advantage of silent operation without revealing ones position was obvious and therefore
the idea to have passive radars was implanted in construction during the early days of radars itself
[1].
A classical monostatic radar is shown in Figure 1. An electromagnetic pulse is sent by the radar
transmitter towards the target. The reflected pulse is measured by the radar receiver. The
transmitter and the receiver are collocated. The accurate position of the target can be detected with
the help of directional antennas. The velocity and Doppler of the target can also be estimated.




Figure 1. Classical monostatic radar principle

Passive radars operate on the principle of using either transmitters of opportunity or co-operative
transmitters. The signal received directly from a transmitter of opportunity is cross correlated with its
reflections from a target. Using a passive radar receiver and an independent transmitting antenna, a
target can be located along an ellipse. The antennas are located at the foci of the ellipse. To resolve a
target's location without much uncertainty, multiple such transmitter receiver pairs are needed,
allowing the target to be tracked at the intersection of the resulting ellipses [4]. The accuracy of the
16

target position strongly depends on the bandwidth of the used signal and the receiver antenna beam
width.

A classical passive radar scenario with a single target is shown in Figure 2. It is assumed that the
moving target is illuminated by multiple transmitters of opportunity and the reflected signal is
received by the passive radar. The direct signal from the transmitter is also received at the reference
antenna in the passive radar. These two signals are cross correlated to detect and track the target
efficiently.


Figure 2 . A passive radar operating with single target and multiple transmitters.
1.1 About passive radar - A Brief history of passive radar
The passive radar was first built before the World War II. It dates back to 1935 when Sir Robert
Watson-watt conducted an experiment to detect a Heyford bomber aircraft by the illumination from
a shortwave BBC Empire transmitter at Daventry which is shown in Figure 3



Figure 3. Daventry experiment [5]

17

Robert Watson-Watt and his assistant Arnold Wilkins conducted the experiment and as the Heyford
bomber flew overhead the signal of the transmitter which was being received and displayed on the
oscilloscope began to fluctuate indicating that a measurable and variable amount of radio signal was
being reflected from the flying bomber. Thus the successful implementation of passive radar was first
demonstrated [2].

After World War II, and recently many studies on passive radar have been done. The first commercial
Passive radar using FM radio broadcast emissions was developed by Lockheed and Martin and called
the Silent Sentry *3+. Thales developed a HA100 antenna for this purpose. The group at UCL,
London has been investigating use of analogue TV transmission for detection of aircrafts. Fraunhofer
FHR have recently developed PCR sensor CORA for exploitation of DAB and DVB-T signals as shown in
Figure 4.



Figure 4. PCR sensor CORA of Fraunhofer FHR [5]
Benefits of passive radar

The passive radar has many benefits. Some of them are listed here:

- Cost: The receivers use passive components which are usually cheaper. Operations and
maintenance costs are also low due to lack of transmitters and moving parts. Also passive
radars use a third partys transmitter and hence is cheaper and more reliable to run
- Stealth operation: The opponent being detected is unaware that they are being scanned.
- Wide area: Passive radar can detect targets over a wide area often measured in several
kilometres.
- Continuous detection: Passive radars detect targets continuously, typically once in a second.
- No frequency allocation is required for the radar frequency though it is required in the DVB-
T signal.
- Environment friendly: Allowing the radar to be setup in areas where normal radars would
not be allowed to as there is no exposure to extra EM radiation
- Relatively inexpensive: as they require nothing more than a digital receiver system, stable
oscillators and an adequate signal processing capability


18

1.2 Motivation for the type of antenna design for passive radar
There are several illuminators of opportunity available. DVB-T is one such digital TV signal that is
widely in use across Europe. Such multi-static transmitter networks are found in the digital broadcast
service nets for radio (DAB) and television (DVB-T) with bandwidths ranging from 1.5 to 7 MHz and
transmit powers between 1 and 10 kW. The receiver becomes difficult to locate when the DVB-T
signals are employed for radar purposes. The VHF/UHF frequency range has anti-stealth capability i.e
the ability to counter stealth. Radars that use these frequencies can counter the threat of stealth
aircraft as stealth shaping does not help against low frequency waves due to half-wave resonance
effects. These are returns generated when the wavelength of the wave is roughly twice the size of
the target, thus returning one full wavelength of measurement to the receiver antenna allowing a
picture to be obtained [6] Simultaneous multistatic illuminations under different aspect angles can
be done. Increased multistatic coverage adds to the detection of low flying targets. Passive, multi-
static, low frequency radars are thus of tremendous use in anti-stealth and detection of low flying
objects.

A passive radar that operates in the DVB-T frequency range requires an antenna that should have
beamforming capabilities like main beam steering and null steering in order to be able to direct a
main beam pointing in the desired signals direction and nulls along the direction of interference.
This can be best realized by using an array antenna. The array antenna element should be
operational over a wideband of DVB-T namely the 450MHz to 900 MHz. Such an antenna element
was built at the Fraunhofer FHR. It uses a Discone element. This Discone element is to be arranged in
a suitable array geometry.

One option in design of an array would be to have an array operating at 450 MHz, another array
optimized for 900 MHz and another possibly for 675 MHz. It would necessitate a mechanically
adjustable array one in which the inter element spacing can be adjusted with the frequency of
operation or to switch between the individually designed arrays. To have an array capable of
providing satisfactory performance without requiring mechanical adjustment would be of great value
for use in DVB-T based passive radar. Digital waveforms like DVB-T signals offer wide bandwidth
channels which allow achieving good spatial resolution. Moreover, they have spectral properties
which are nearly independent of the signal content [7]. During the course of the thesis the design
and implementation of such an antenna array was aimed at.
1.3 Identification of the problem
Considering the scenario described above, the problem can be identified as to design a circular
cylindrical antenna array for a passive radar system in the DVB-T frequency range using Discone
antenna element and to find the optimum array geometry that would allow for beamforming and
null steering to be performed on it. An investigation of the performance of the design is eventually
done.
The element to be used in the array was designed at Fraunhofer FHR earlier. The goal is to adopt the
existing element in an array to achieve wideband capabilities for the array. It is required to find array
dimensions allowing for beam steering with low side-lobes / deep nulls of required bandwidth and to
specify parameters for antenna fabrication such as the polarization and the type of array grid.
Approach adopted:
Step 1: Literature review of the general properties of antenna arrays, beam forming techniques and
adaptive null steering and implementation of algorithms for beamforming and null steering in
MATLAB.
19

Step 2: Used simplified NEC-2 model of Discone antenna for principal array analysis (proof-of-
concept). The 4NEC2 Discone model to be used was to operate in the 450 -900 MHz. The Discone
model available in the Library of NEC2 was operable in the 7 MHz range. This model had to be
adopted to operate in the 450 -900 MHz range.
Step 3: Full wave analysis using CST of arrangement of the Discone element in the circular array
fashion with the appropriate radius of the array and inter element spacing. This was done by a series
of parametric studies at different positions of the elements.
Step 4: The Discone array was simulated to get the radiation pattern in the azimuthal plane at steps
of 5 MHz between 450-900 MHz using full wave analysis using CST. The data of the simulation was
read into Matlab and then processed to perform the beam forming and null steering.
Step 5: Investigation of the influence of support structure and improvement of antenna mount.
Step 6: Fabrication and testing of the designed array antenna (to be performed in near future)
1.4 Challenge of the work and methodology adopted
Within the framework outlined in detail above the challenge was to find a suitable antenna array
topology for application in the DVB-T range that would also additionally allow for beamforming and
null steering. The interaction between the interelement spacing, the grating lobes and the mutual
coupling had to be carefully studied and an optimal solution was to be reached at that meets all the
specifications of the antenna array as described above. Directional beams, nulls along interference
directions, low sidelobe levels, polarization, and operation along a wide bandwidth of 450-900 MHz
were some of the key considerations. The major challenge was in the difficulty of satisfying all of
these constraints simultaneously.
1.5 Outline of the thesis
Chapter 1 introduces the topic and discusses about passive radar. It also defines the problem, the
aim of work and approach followed. Chapter 2 deals with the relevant theoretical background. It
discusses the different kinds of array and related array antenna theory. The choice of circular array
and the array architecture are explained. The phase mode theory of circular arrays and wideband
beamforming and null steering theory are discussed. Chapter 3 describes the beamforming
algorithms- the conventional ones and compares them. It also introduces the Bucci beamforming
algorithm. Chapter 4 details the CORA radar system and the architecture of the antenna array.
Chapter 5 presents simulation studies performed and design of the antenna array alongwith
beamforming and null steering. The results of 4NEC2 and CST MWS are presented. Chapter 6
describes the experimental setup. Chapter 7 concludes by listing the prominent outcomes and
possible suggestions for future work.




20


Chapter 2 Theoretical background
The work presented here has scope across multiple dimensions. It involves the concepts in individual
antenna element design to design of an array, while also covering the beamforming and null steering
concepts. The theory to be dealt with spans across these domains.

2.1. Design of the discone antenna element

The Discone antenna element [8] for the array was chosen due to several reasons. Prominent among
these were:

- The large bandwidth
- Omnidirectional coverage
- Low cost
- Ease of fabrication

A discone antenna is in principle a biconical antenna where one of the cones is replaced by a disk-
shaped ground plane. The disk is mounted on top of the cone. This structure was first developed by
Kandoian in 1945. Discones have linearly polarized radiation patterns like a monopole. Due to the
strict vertical polarization of the Discone antenna, it exhibits higher cross-polarization. The radiation
pattern is near omnidirectional in the azimuth plane for about one octave and the impedance
bandwidth can be upto 10:1. In the elevation plane radiation pattern, it is toroidal shaped.

The Discone antenna has three components: the disk, the cone and the insulator as in Figure 5. The
diameter of the top of the cone depends on the diameter of the coaxial cable used to feed the
antenna. A coaxial line outer conductor is connected to the cone and the inner one is connected to
the ground disk. The insulator keeps the disk and cone apart. The antennas input impedance
depends on the cone angle and disk to cone spacing.

Design rules

- The Length l of the cone elements should be a quarter wavelength of the minimum operating
frequency.
- The diameter of the top of the cone depends on the diameter of the coaxial cable and
determines the upper frequency limit of the antenna: the smaller the diameter the higher
the frequency.
- The disk elements should have an overall length r of 0.7 times a quarter wavelength at the
minimum frequency.
- The slant height of the cone is equal to a full free space quarter wavelength at the lowest
frequency of interest.


The Discone antenna used in the scope of this project was designed earlier by the group [8] at
Fraunhofer FHR to operate over a frequency range of 450 to 900 MHz for DVB-T system. It is a very
small antenna, but very effective. This broadband antenna which has omnidirectional H-plane
pattern and vertical polarization is meant to be element of the array.

The Discone antenna dimensions are shown in Figure 5 as derived in [9]

21








r=150 mm, l=210mm, d=194mm, s=8mm, w=20mm

Figure 5. Discone antenna and parameters [9]

The return loss (|S
11
|) characteristics of the antenna are shown in Figure 6.



Figure 6. Return loss of the Discone antenna [9]




22

2.2 Array antenna theory

In this section a review of all the important and relevant topics related to antenna array theory is
done. This includes the properties of linear array, phased array, planar array and the circular array.

Individual antennas generally have rather wide radiation patterns that lead to low directivity. Some
applications like radar, radio links, remote sensing require high directivity beams and therefore need
very directive antennas. There are two possibilities to achieve this:

1) Increasing the physical dimensions of the antennas such as a dipole antenna results in more
directional beams. An increase in electrical size of the antenna generally results in directive beams
2) Arranging the single antennas in assemblies thereby generating a complex radiating structure
called array antenna

Array antennas are arrangements of usually identical radiators assembled in a specific geometrical
configuration to produce highly directive radiation.



Figure 7. Linear and planar array depiction
The total field radiated by an array-antenna is the superposition of the fields radiated by the
individual radiators. To produce high directivity the fields radiated by the individual radiators have to
add up coherently in the desired direction and cancel out in the other directions [10]. Linear and
planar arrays are depicted in Figure 7.

In order to achieve this high directivity there are several degrees of freedom that can be used. These
are namely[10]:

The relative displacement within the geometrical arrangement
The excitation amplitude and phase
The relative pattern of different elements



23



2.2.1 Two element array

Consider two infinitesimal dipoles aligned along the z axis excited with a phase difference of
radians as shown in Figure 8.


Figure 8. Two dipoles placed along the z axis

The total field in the yz plane is given by

(

+
(

=
+ =
+
2
2
/2)] ( j[kr
0
1
1
/2)] ( j[kr
0
t
2 1 t
cos
r
e
4
l I
j a cos
r
e
4
l I
j a E
E E E
2 1
| |

(1)
Under far field conditions u u u ~ ~
2 1
, as in Figure 9

24


Figure 9. Two dipoles placed along the z axis under farfield condition
For the phase variations we have

u
u
cos
2
cos
2
2
1
d
r r
d
r r
+ ~
~
(2)

For the amplitude variations, under the assumption of Rayleigh distance criterion for far fields[10]

r r r ~ ~
2 1
(3)

The total field can now be approximated as [10]

| |
)
`

+ =
+ =

+ + +

) cos (
2
1
cos 2 cos
4

cos
4

0
2 / ) cos ( 2 / ) cos ( 0
| u u
t
q
u
t
q
u
| u | u
u
kd
r
le kI
j a E
e e
r
le kI
j a E
jkr
t
kd j kd j
jkr
t

(4)

In a two element array of identical elements, the principle of pattern multiplication applies:

(AF) factor Array x E E
element single t

= (5)

The pattern multiplication when performed for different d and values results in the following cases:








25

Case 1: d= /4, =0


Figure 10(a). Illustration of principle of pattern multiplication when d= /4 and a) = 0 [10]


Case 2: d= /4, =90


(b)
Figure 10(b). Illustration of principle of pattern multiplication when d= /4 and a) = 90 [10]


26

Case 3: d= /4, = -90



Figure 10(c). Illustration of principle of pattern multiplication when d= /4 and = -90 [10]
2.2.2 N element array

Consider N identicalelements aligned along the z axis at arbitrary locations z
1
,z
2
,z
3
,..z
N
. The elements
are fed with different amplitude and phase. Each element has a potentially different radiation
pattern. The arrangement is shown in Figure 11


Figure 11. N element array

The total radiated field in the far zone is
27

=
+

=
N
n
kz j
n n
jkr
t
n n
e f a
r
e
E
1
) cos (
) , (
4
) , (
| u
| u
t
| u

(6)
where ) , ( | u
n
f

is the electric field radiation pattern of the individual radiator

If the elementary radiators are identical[10]

=
+

=
N
n
kz j
n
jkr
t
n n
e a f
r
e
E
1
) cos (
) , (
4
) , (
| u
| u
t
| u

(7)

The term inside the summation is array factor (AF). The array factor would correspond to the
normalized radiation pattern of the array if the identical individual antennas were isotropic radiators.
The array factor is a function of the location of the antennas in the array and their amplitude and
phase excitation coefficients [10].

=
+
=
N
n
kz j
n
n n
e a AF
1
) cos (
) , (
| u
| u (8)


Consider a uniform array, one in which the N identical elements are equally spaced, aligned and fed
with identical amplitudes and progressive phase difference | as in Figure 12


Figure 12. Uniform linear array

Considering the elements to be point sources

=
+
+ + +
=
+ + + + =
N
n
kd n j
kd N j kd j kd j
e AF
e e e AF
1
) cos )( 1 (
) cos )( 1 ( ) cos ( 2 ) cos (
..... 1
| u
| u | u | u
(9)
The AF can be rewritten in the form
28



|
.
|

\
|
|
.
|

\
|
=

2
1
sin
2
sin
N
AF (10)

After normalization this becomes

(
(
(
(

|
.
|

\
|
|
.
|

\
|
=

2
1
sin
2
sin
1
N
N
AF
n
(11)


Array factor plot for N=10 is shown in Figure 13.




Figure 13. Array factor plot for N=10

The positions of the nulls are found for
(13)
2
2
cos
(12) 0
2
sin
1
(

|
.
|

\
|
=
=
|
.
|

\
|

t |
t

N
n
d
N
n

29

The positions of the maxima are found for

( ) (15) 2
2
cos
(14) ) cos (
2
1
2
1
(

=
= + =

t |
t

u
t | u

m
d
m kd
m


By controlling the progressive phase difference between the elements it is possible to steer the
maximum radiation in any desired direction
0
u

( )
(17) cos
(16) 2
2
cos
0
1
0
u |
t |
t

u
kd
m
d
=
(

=



An example of phased array radiation pattern with

30
0
= u , 0 . 3 / = d and N=20 is shown in Figure
14


Figure 14. Phased array radiation pattern with

30
0
= u , 0 . 3 / = d and N=20 [39]
It is possible to reconfigure the radiation pattern by dynamically changing the feeding phase of the
elements and point the maximum radiation in different directions.

A schematic representation of a phased array is shown in the figure 15:

30


Figure 15. Schematic representation of phased array [33]
In a phase array antenna, each transmit-receive (TR) element is essentially a miniature radar
antenna. The antenna computer then modulate individual TR power output and the entire system
exploits the wave superposition principle to effect beamforming and to move or electronically
'sweep' the main beam by varying the phases at the individual antenna elements. The amplitudes
and taper also influence the beam scanning. If the magnitudes of the excitation currents In are
chosen to be tapered across the array instead of equal in magnitude, so that
0
I is largest and
n
I

decrease in magnitude, then u must move farther from 2 / t to obtain a pattern null than for
the uniformly weighted array, and the main lobe is wider. In addition, the phasors add to a smaller
value for the sidelobes, and the sidelobe level decreases [11].
Mutual coupling in antenna arrays:

The electromagnetic interaction between the antenna elements in an antenna array is called mutual
coupling. The effect of mutual coupling is serious if the element spacing is small.
It will affect the antenna array mainly in the following ways:
1. Change the array radiation pattern
2. Change the array manifold (the received element voltages)
3. Change the matching characteristic of the antenna elements (change the input impedances)
The amount of mutual coupling depends on[10]
1. Radiation characteristics of each antenna
2. Relative separation between them
3. Relative orientation of each antenna
For most practical configurations mutual coupling is difficult to predict analytically but must be taken
into account because of its significant contribution.

Isolated pattern and embedded element pattern:

An isolated pattern is the radiation pattern of an element alone without the influence of any
neighboring antenna. It is the individual antenna radiation pattern, whereas the embedded element
pattern is the radiation pattern in the presence of other antenna elements of the array when only
31

the element under consideration is excited with other elements present passively around it in the
array. It is therefore called the embedded element pattern.

Grating lobes in antenna arrays
If t 2 > kd then for u large enough, t u 2 cos = kd and we get another angle in addition to
2 / t u = where all the phasors in the array factor sum are aligned, causing another pattern lobe with
maximum amplitude. These are called grating lobes, and are usually undesirable. For an array with
equal phase excitations, to avoid grating lobes the element spacing must be small enough
that s d . An array with widely spaced elements has many grating lobes, and the radiation pattern
has strong peaks at many angles. Grating lobes can be eliminated by reducing the element spacing or
suppressed by choosing an electrically large element with narrow radiation pattern [11].

If > d then at some angle u away from 2 / t , t u 2 cos = kd , and
n
V = 1 for all n. The samples of
the field at the array elements are too widely spaced to discriminate between the two different
plane waves. In order to avoid grating lobes, the elements must be spaced more closely than one
element per wavelength. For arrays with steered beams using progressive phase shifts of the
element excitations, we must have 2 / s d , or at least two elements per wavelength, to avoid
grating lobes for all beam steering angles [11]. Since the array essentially samples the incident field
at the element locations, grating lobes are simply a manifestation of aliasing.
2.2.3 Planar arrays

Consider a uniform planar array, with M elements along the x axis, N elements along the y axis with
element spacing d
x
, d
y
.The feeding signals have uniform amplitude and progressive phase
y x
| | , as
shown in the Figure 16.

Figure 16. Planar array depiction

The uniform planar array, array factor (AF) is given by [10]

)
`

=
) 2 / sin(
) 2 / sin(
1
) 2 / sin(
) 2 / sin( 1
) , (
y
y
x
x
n
N
N
M
M
AF

| u (18)
32


With,
y y y
x x x
kd
kd
| | u
| | u
+ =
+ =
sin sin
cos sin


An example of a planar broadside array pattern with M=10, N=20, d
x
=0.1, d
y
=0.2, 0 = =
y x
| | ,
f=1GHz is shown in Figure 17.

Figure 17. Planar broadside array pattern with M=10, N=20, d
x
=0.1, d
y
=0.2, 0 = =
y x
| | , f=1GHz [39]

By controlling the progressive phase difference between the elements it is possible to steer the
maximum radiation in any desired direction ) , (
0 0
| u

0 0
0 0
sin sin
cos sin
| u |
| u |
y y
x x
kd
kd
=
=

An example of a planar array pattern with beam along ) , (
0 0
| u = ) 45 , 30 (

is shown in Figure 18


Figure 18. Planar array pattern with beam along ) , (
0 0
| u = ) 45 , 30 (

[39]

The maximum directivity than can be achieved in planar arrays is an important factor. It is given by

33

| || |
| || |
} }
=
tt
| u u | u | u
| u | u t
2
0 0
0 0 0 0
0 0 0 0
0
sin * ) , ( ) , (
* ) , ( ) , ( 4
d d AF AF
AF AF
D (19)

For large planar arrays with maximum pointed at small scanning angles the directivity is given by:


y x
D D D
0 0
cosu t = (20)
where D
x
and D
y
represent the directivities of broadside linear arrays of M and N elements
respectively.

2.2.4 Circular array

Consider N equally spaced elements on a circular ring of radius a as shown in Figure 19


Figure 19. Uniform circular array [10]
The normalized total field is given by

=
N
n n
jkR
n n
R
e
a r E
n
1
) , , ( | u (21)

n
a are the complex excitation coefficients.

The distance R
n
between the n
th
element and the observation point is:


n n
ar a r R cos 2
2 2
+ = (22)


For : a r >>

34

) cos( sin
) . (
cos
n
r
n n
a r
a a a r
a r R
| | u

=
=
~
(23)

Where
n
| is the angular position of the n
th
element.

For amplitude variations:

|
.
|

\
|
=
=
~
~

=
+

N
n
e I a
with
e a
r
e
r E
r R
n
j
n n
N
n
n jka
n
jkr
n
n
n
t |
| u
o
| | u
2
) , , (
1
) cos( sin
(24)
The farfield now becomes
| |
| | (25) ) , ( ) , , (
) , , (
1
) cos( sin
| u | u
| u
o | | u
AF
r
e
r E
e I
r
e
r E
jkr
n
N
n
n ka j
n
jkr
n
n

=
+ +

~
~



Where the array factor is

| |

=
+ +
=
N
n
n ka j
n
n
e I AF
1
) cos( sin
) , (
o | | u
| u (26)

The array factor depends on [10]

- The number of elements (N)
- Their excitations ) , (
n n
I o
- The radius of the ring (a)

The main radiation beam can be directed towards ) , (
0 0
| u by setting the excitation phases to
) cos( sin
0 0 n n
jka | | u o = (27)
The Array factor for a directional circular array becomes


| |

=

=
N
n
jka
n
n n
e I AF
1
) cos( sin ) cos( sin
0 0
) , (
| | u | | u
| u (28)

One of the principal advantages of circular arrays is their ability to deflect beams electronically
through 360 degree with little change of either beam width or sidelobe level. There has been a
significant increase of interest in circular arrays for applications of direction finding and electronic
support measures systems [12].

35

The most significant development was made in the 1960s when the concept of phase mode
excitation was developed. The lack of applications in the past is due to the fact that the basic
problems of exciting circular arrays with the correct values of amplitude and phase are in general
more complex than for linear arrays. Also the electronic scanning of directional patterns for circular
arrays may be difficult to implement and may require both the amplitude and phase of each element
of the array to be changed.

Circular array applications have included Wullenweber arrays for direction finding [13], wide
bandwidth HF communication arrays [14], wrap around antennas for ship borne communications
[15], navigational aids, spacecraft antennas and null steering antennas for mobile communications
applications [16]

The theory of the uniform circular array (UCA) is reviewed in this section[12]. This generic array has
elements equally spaced along the periphery of a circle as in Figure 20(a), and the elements are
excited with equal phase and amplitude. We will start by an analysis of the radiation in the plane that
contains the array, which we refer to as the azimuth (or horizontal) plane. Local coordinates on the
circle are (R, ) , the radius and angle, respectively. Far-field coordinates are indicated by (r, , ) as
in Figure 20(b) . We consider element excitations (voltages or currents) as given or specified including
the effects of mutual coupling. Limiting the analysis to the azimuth plane only makes the circular
array problem a two-dimensional problem, that is, it has bearing on other two-dimensional problems
like cylindrical array structures.

(a) (b)

Figure 20. Local and far field coordinates of the circular array [40]
There are some elementary differences between linear and circular arrays. For the linear array the
radiating elements are identical, with equal spacing d, and they all point in the same direction. The
far-field radiation function in the azimuth plane is, given by

=
n
jknd
n
e V EL E
|
| |
sin
) ( ) ( (29)
where, V
n
is the excitation amplitude of element n and k is the propagation constant, k = 2/. The
radiation pattern EL() is common to all elements and was therefore brought outside the summation
sign in this equation. The radiation function is the product of the EL() and the array factor.
The corresponding far-field expression for the circular array shown in Figure 21, in the azimuth plane
is

A
A =
n
n jkR
n
e n EL V E
) cos(
) ( ) (
| |
| | | (30)

36

where the phase has been referenced to the center of the circle. The identical elements are spaced
R along the circle, each element pointing in the radial direction. The element function can,
therefore, in general not be brought outside the summation, since it is a function of the element
position; there is no common element factor. Consequently, we can, in general, not define an array
factor as in linear and planar arrays, unless the elements are isotropic, that is, have isotropic
(omnidirectional) radiation at least in the horizontal plane. A typical example of the latter is an array
of vertical dipoles with their axes perpendicular to the array plane. A further difference compared to
linear and planar arrays is found in the slightly more complicated phase expression because of the
array curvature.

Equation (1) for the linear case is essentially a Fourier relationship between excitation and pattern
functions. Much of the experience from Fourier analysis can be directly applied to linear (and planar)
arrays, and there are several analogies with well-known time/frequency spectral relationships.
Unfortunately, this knowledge is not so easily applied to the analysis of circular and conformal arrays.


Figure 21. Part of circular array [40]
With the linear array, we can point a focused beam in a direction
0
by applying a linear phase shift
(n) to the elements along the array:


0
sin ) ( knd n = (31)

With the circular array, we can similarly impose phase values to each element so that they add up
coherently in the direction
0
. We get the proper phase excitation for each element n by choosing

) cos( ) (
0
| A = n kR n (32)

Thus, the radiation function for the focused, circular case becomes

A A
A =
n
n n jkR
n
e n EL V E
)] cos( ) [cos(
0
) ( ) (
| | | |
| | | (33)

Sometimes, it is required to generate a beam with equal radiation in all directions in the azimuthal
plane (an omnidirectional beam). Circular arrays are particularly suitable for this by virtue of their
circular symmetry. We will now look more closely at circular array behaviour and investigate how
performance can be optimized. Indeed, much better results can be obtained by proper choice of
parameters.
37

Accordingly a parametric study was done on the inter-element spacing and grating lobe dependence
using the uniform circular array factor definition using Matlab and with uniform excitation using
eleven isotropic elements. Selected results are presented here for brevity. Consider for different
inter-element spacing d as shown in Figure 22.

d=0.25 d= 0.5

d= 0.75 d= 1.0

d= 2.0 d= 5.0

Figure 22. Grating lobes dependence on inter element spacing in a circular array of 11 isotropic elements
From the Matlab calculations above it can be concluded that the grating lobes, though difficult to
locate in a circular array pattern, become dominant at spacing equal to or more than 1 . At spacing
lesser than 0.5 the grating lobes are absent.
38

2.3 Wideband beamforming and null steering

The passive radar requires in its operation beam steering and null steering capabilities. These abilities
are needed because it has to have antenna that can direct a main beam in the direction of signal of
interest and a null in the direction of interference. The passive radar in which the designed antenna is
deployed exploits DVB-T signal which is spread over the 450-900 MHz range. This beamforming has
to therefore be accomplished over a wideband range of 450-900 MHz. This necessitates wideband
beamforming and null steering.

Beamforming
Beamforming refers to the ability of the antenna array to focus energy toward a specific direction
without the knowledge of the incoming signals direction of arrival (DOA). Adaptive beamforming
involves knowing the direction of desired signal and undesired signals and then adaptively calculating
the array excitation coefficients to point directional beam along the desired direction in space and
nulls in the undesired direction. It is often referred to as spatial filtering. It is one of the major areas
of signal processing and has been studied extensively in the past due to its wide applications in
various areas ranging from radar, sonar, microphone arrays, radio astronomy, seismology, medical
diagnosis and treatment, to communications. It involves multiple sensors (microphones, antennas,
etc.) placed at different positions in space to process the received signals arriving from different
directions. An example for a simple array system consisting of four sensors with two impinging
signals is shown in Figure 23 for illustrative purposes, [24] where the direction of arrival (DOA) of the
signals is characterized by two parameters: an elevation angle u and an azimuth angle| .



Figure 23. Array with four sensors and two impinging signals

For the impinging signals, we always assume that they are plane waves, i.e. the array is located in the
far field of the sources generating the waves and the received signals have a planar wavefront.

Consider a plane wave with a frequency f propagating in the direction of the z-axis of the Cartesian
coordinate system as shown in Figure 24.
39





Figure 24. A plane wave propagating in the direction of z axis of Cartesian coordinate system.


At the plane defined by z = constant, the phase of the signal can be expressed as:

kz ft z t = t | 2 ) , ( (34)

where t is time and the parameter k is referred to as the wavenumber and defined as

t e 2
= =
c
k (35)

where e is the (temporal) angular frequency, c denotes the speed of propagation in the specific
medium and is the wavelength. Similar toe , which means that in a temporal interval t the phase
of the signal accumulates to the value t e , the interpretation of k is that over a distance z, measured
in the propagation direction, the phase of the signal accumulates to kz radians. As a result, k can be
referred to as the spatial frequency of a signal [25].


Different from the temporal frequencye , which is one-dimensional, the spatial frequency k is three-
dimensional and its direction is opposite to the propagating direction of the signal. In a Cartesian
coordinate system, it can be denoted by a three-element vector:

| |
T
y x
k k k k , ,
z
= (36)
with a length of:

2 2 2
z y x
k k k k + + = (37)
This vector is referred to as the wavenumber vector. In the case shown in Figure 23, we have
0 = =
y x
k k and k k
z
= . Let
T
z ] 1 , 0 , 0 [ = denote the unit vector along the z-axis direction, then we
have z -k

= k .

These two quantities are not independent of each other and are related by the following equation:
40


c
f
k
t 2
= (38)
Any point in a 3-D space can be represented by a vector | |
T
y x
r r r r , ,
z
= , where r
x
, r
y
and r
z
are the
coordinates of this point in the Cartesian coordinate system. With the definition of the wavenumber
vector k, the phase function ) , ( r t | of a plane wave can be expressed in a general form [25]:

r
T
k ft r t + = t | 2 ) , ( (39)

For the case in Figure 24, we have:

z
T T
kr k = = r) z r k

( (40)
Therefore, as long as the points have the same coordinate r
z
in the z-axis direction, they have the
same phase value at a fixed time instant t.

For the general case, where the signal impinges upon the array from an elevation angle u and an
azimuth angle| , as shown in Figure 23, the wavenumber vector k is given by:


(
(
(

=
(
(
(

=
u
| u
| u
cos
sin sin
cos sin
k
k
k
k
z
y
x
k (41)

Then the time independent phase term r k
T
changes to:

) cos sin sin cos sin ( u | u | u
z y x
T
r r r k + + = r k (42)
The wavefront of the signal is still represented by the plane perpendicular to its propagation
direction.

Array signal processing involves [25]:
1. Detecting the presence of an impinging signal and determine the signal numbers.
2. Finding the DOA angles of the impinging signals.
3. Enhancing the signal of interest coming from some known/unknown directions and suppress
the interfering signals (if present) at the same time.

The third area is the task of beamforming, which can be divided into narrowband beamforming and
wideband beamforming depending on the bandwidth of the impinging signals.

Narrowband beamforming

In adaptive beamforming, we estimate the signal of interest arriving from some specific directions in
the presence of noise and interfering signals with the aid of an array of sensors. These sensors are
located at different spatial positions and sample the propagating waves in space. The collected
spatial samples are then processed to attenuate/null out the interfering signals and spatially extract
the desired signal. As a result, a specific spatial response of the array system is achieved with beams
pointing to the desired signals and nulls towards the interfering ones.
Figure 25 shows a simple beamforming structure based on a linear array,

41


Figure 25. General structure of narrowband array architecture [24]
where M sensors sample the wave field spatially and the output y(t) at time t is given by an
instantaneous linear combination of these spatial samples 1 ,... 1 , 0 ), ( = M m t x
m
as

=
=
1
0
*
) ( ) (
M
m
m m
w t x t y
(43)
The beamformer associated with this structure is only useful for sinusoidal or narrowband signals,
where the term narrowband means that the bandwidth of the impinging signal should be narrow
enough to make sure that the signals received by the opposite ends of the array are still correlated
with each other[26], and hence it is termed a narrowband beamformer.

Wideband beamforming

The beamforming structure discussed above works effectively only for narrowband signals. When the
signal bandwidth increases, its performance will degrade significantly. This can be explained as
follows.
Suppose there are in total M impinging signals ) (t s
m
, m = 0, 1, . . . , M 1, from directions of
m
, m =
0, 1, . . . , M 1, respectively.
i
t is the propagation delay for the signal from sensor 0 to sensor m and
is a function of u . The first one ) (
0
t s is the signal of interest and the others are interferences. Then
the arrays steering vector d
m
for these signals is given by [25]:

| |
T
j j
m
m M m
e e d
) ( ) (
1 1
... 1 ) , (
u et u et
u e


= (44)

Ideally, for beamforming, we aim to form a fixed response to the signal of interest and zero response
to the interfering signals. For simplicity, we do not consider the effect of noise here. This
requirement can be expressed as the following matrix equation:
|
|
|
|
|
.
|

\
|
=
|
|
|
|
|
.
|

\
|
|
|
|
|
|
.
|

\
|

0
0
constant
1
1
1
*
1
*
1
*
0
) ( ) (
) ( ) (
) ( ) (
1 1 1 1
1 1 1 1
0 1 0 1

M
j j
j j
j j
w
w
w
e e
e e
e e
M M M
M
M
u et u et
u et u et
u et u et
(45)

Obviously, as long as the matrix on the left has full rank, we can always find a set of array weights to
cancel the M 1 interfering signals and the exact value of the weights for complete cancellation of
the interfering signals is dependent on the signal frequency.

42

For wideband signals, since each of them consists of infinite number of different frequency
components, the value of the weights should be different for different frequencies and we can write
the weight vector in the following form:
| |
T
M
w w w w ) ( ..., ) ( ) ( ) (
1 1 0
e e e e

= (46)

This is why the narrowband beamforming structure with a single constant coefficient for each
received sensor signal will not work effectively in a wideband environment.

While designing a wideband array, the array designer must balance avoiding undesired mutual
coupling effects with eliminating grating lobes due to element spacing at the high end of the
frequency band. This limits the array bandwidth to an extent.

The Bartlett beamformer was the first to emerge during World War II [17]. Later adaptive beam
formers and classical time-delay estimation techniques were applied to enhance the ability to resolve
closely spaced signal sources [18, 19]. The conventional beam former has some fundamental
limitations connected to physical size of the array, available data collection time, and the signal to
noise ratio (SNR).

Analog beamforming

The figure 26 depicts the radio frequency (RF) beamformer example for creating only one beam at
the output [20]. RF beam formers can employ microstrip structures, transmission lines or microwave
guides.




Figure 26. A simple on-beam RF beam former[20]
Multiple beamformers are more complex configurations, based mathematically on the beamforming
matrix. The Butler matrix is the most well-known and widely used matrix [21, 22]. Typically the
number of beams is equal to the number of antenna elements in the arrays. Although the Butler
matrix was developed before the Fast fourier transform (FFT), they are completely equivalent. The
Buttler matrix is used in analog beamforming , whereas FFT is used in digital beamforming.

Digital beamforming

The incident RF signal at each antenna element is converted in to two streams of binary complex
baseband signals that represent the in-phase (I) and quadrature (Q) phase components. These
weighted signals from each element are sampled, stored and beams are then formed summing the
appropriate samples [23]. A digital beamformer of this form is shown in figure 27.

43




Figure 27. Digital beamformer architecture [24]

The array in Figure 27 has M elements. The output ) ( ), ( nT t y
n
= u is given by the sum of the
sampled baseband data at M sensors given by

=
=
1
0
*
) ( ) (
M
m
m m n
n x w y u (47)

Where
m
x is the signal from the m
th
element of the array,

*
m
w is the weight applied to the antenna element m
* denotes the complex conjugate operation

The above equation can also be written as

) ( ) ( n x w y
H
n
= u (48)

Where H represents the Hermitian transpose.

The n
th
sample of the output ) (u
n
y is given by the array snapshot multiplied by a set of weights
chosen to form a beam in desired directionu . By choosing the appropriate weight vectors, beam
steering, adaptive nulling and beam shaping can be achieved.

Null introduction and steering
The application for which the passive radar is built requires that a null is introduced in the direction
of the direct signal in the presence of reflected signal from target. This calls for simultaneous
existence of both aiding beam in one direction and canceling beam in the other direction. An
illustration is shown in Figure 28. Dotted curve in the figure is pattern before beamforming. Solid
curve in the figure is pattern after beamforming and nulling. The null as seen can be steered into any
direction along the horizontal plane of the circular array. It can be rotated all around the circular
array in a circle in any direction from 0 to 360 degrees.
44


Figure 28. Illustration of null introduction and beamforming [40]
The introduction of a null is part of adaptive beamforming which involves estimation of direction of
arrival/interference and calculation of optimum weights by using the concept of covariance matrix.
Consider covariance matrix R(t)
* * *
1 1 1 2 1
* * *
* 1 1 2 2 2
* * *
1 2
( ) ( ) ( ) ( ) ( ) ( )
( ) ( ) ( ) ( ) ( ) ( )
( )
( ) ( ) ( ) ( ) ( ) ( )
N
T N
N N N N
u t u t u t u t u t u t
u t u t u t u t u t u t
R t u u
u t u t u t u t u t u t
(
(

(
= =
(
(

(

(49)
Where u
N
(t) are the received signals in N
th
antenna. The covariance matrix describes the correlation
between the signals in different antennas. Then for the maximum signal to interference ratio (SIR)
we have the array excitation coefficients given by

* 1
0
(E )
T
a u u a

= + (50)
The null depth can be controlled by introducing a scaling factor o

* 1
0
(E * )
T
a u u a o

= + (51)











45

Chapter 3- Beamforming algorithms and their comparison

The beam formers coefficients often need to be adjusted according to the received array data in
order to achieve an optimum solution to the specified scenario. When the environment keeps
changing, the coefficients need to change too. This class of beamformers often employ all kinds of
adaptive algorithms to update their coefficients. For block-based adaptation, we normally assume
that the signals are relatively stationary for the duration of the received block of array data, so that
the statistics of the signals can be estimated accurately for the block period to calculate the optimum
solution. For a more time-varying environment or where the number of coefficients is too large, the
coefficients are updated continuously, i.e. they are adjusted for each new set of signal samples. In
this chapter the conventional beamforming algorithms are presented and compared for their
complexities and then the Bucci beamforming algorithm and its advantage is presented.

3.1 Conventional beamforming algorithms

There are a lot many conventional algorithms that have been used frequently in beamforming
applications. Prominent among them are the least mean squares, sample matrix inversion,
normalized least mean squares and recursive least squares. These are discussed in detail in the
section below.
3.1.1 Sample matrix inversion (SMI) algorithm

The SMI algorithm provides estimates for the array weights by replacing the correlation matrix R

an
estimate from multiple snapshots (N) of the array signal:

=
=
N
n
H
n x n x
N
R
1
) ( ) (
1

(52)

The estimate of R

can be updated when new samples arrive:

1
) 1 ( ) 1 ( ) (

) 1 (

+
+ + +
= +
n
n x n x n R n
n R
H
(53)

Reed developed the sample matrix inversion technique for a fast adaptive beamforming scheme [27]

3.1.2 Least mean squares (LMS) algorithm

The LMS algorithm the most commonly used adaptive algorithm was developed by [18]. The LMS
algorithm updates the weights at each iteration by estimating the gradient of the quadratic surface
and then moving the weights in the negative direction of the gradient by a constant, referred to as
step *21+. Let e(n) be the error function and E(x) be the expectation of x then we have


)) ) ( ( ( ) ( ) 1 (
) ( ) ( )) ) ( ( (
2
* 2
n e E step n w n w
n x n e n e E
V = +
= V
(54)

46

From these two equations it can be seen that

)] ( ) ( [ ) ( ) 1 (
*
n x n e step n w n w + = + (55)


When the step size is smaller than the maximum eigenvalue of the correlation matrix R, the
algorithm is stable and the mean values of the estimated weights converge to the optimal weights.
Since the sum of the eigenvalues of R is its trace, the step size is usually chosen as


) (
1
0
R Trace
step < < (56)

The LMS algorithm does not require prior knowledge of the signal statistics [18]. The algorithm
amounts to approximating the expected value by the product of the observation vector and the
estimation error, a technique known as stochastic approximation. As the filter coefficients depend on
new observations and are continuously updated via the estimate of the gradient, the LMS filter
coefficients do not converge to a stable set of values even when the observation is stationary. Rather
after an initial transient response to the set of observations, the coefficients settle towards a set of
values and vary randomly to some degree. The variation of the filters coefficients is related to the
adaptation parameter step but in the opposite way to the settling time. A small value for the
step results in slow convergence of the filter coefficients and their long term values do not vary
greatly about the optimal values.

The LMS algorithm is the least computationally complex weight adaptation algorithm. The rate of
convergence to the optimum weights depends on eigenvalues of correlation matrix R, that, is the
power of the desired and interfering signals.

3.1.3 Normalized least mean squares (NLMS) algorithm

The LMS algorithm is a simple adaptive algorithm, but with data dependent behaviour. The influence
of power of the input signal can be removed by normalization of the step size. This leads to the
NLMS. In the LMS algorithm, the step size should never exceed its upper bound. In a non-stationary
environment, the worst case has to be assumed and we have to choose a very small value for step
size, which leads to a rather slow convergence rate for the algorithm. It is possible to normalize the
step size to ensure an approximately constant rate of adaptation. Then a normalization of the step
size is given by [25]:


) ( ) (
0
n x n x
H

=

(57)

where
0
is the new step size. Substituting this equation into the LMS update equation yields a
constant convergence rate independent of the power of the input signal x.

3.1.4 Recursive least squares (RLS) algorithm

In [28] it is shown that the weights can be calculated in a recursive manner

47

(58) ) 0 (
) 1 ( ) ( ) 1 (
) ( ) 1 ( ) 1 ( ) (
) (
1
) 1 (
) 1 ( ) ( ) 1 (
) 1 ( ) 1 ( ) (
) ( ) 1 (
1 1
1
1 1
1 1
1
* 1
I R
n x n R n x
n R n x n x n R
n R n R
n x n R n x
n x n e n R
n w n w
H
H
H

=
(

+ + +
+ +
= +
+ + +
+ +
+ = +
o

Where o is a small positive constant. The forgetting factor depends on the fading rate of the
channel and the best value is close to 1. The RLS algorithm is extremely fast compared to the LMS
algorithm and does not suffer from poor convergence in variance as with LMS, but requires several
matrix multiplications at each iteration.


3.2 The Bucci beamforming algorithm

Of several beamforming algorithms available, the goal is to use an algorithm that has fast
convergence and lower computational complexity. O. M. Bucci et al. [29] proposed a general
projection method to array synthesis. This algorithm is discussed here.

An array factor synthesis procedure which allows to specify the side-lobe and shaped-zone ripple
levels as well as the fall rate of the main lobe is proposed. The problem is formulated as the search of
the intersection between two suitable sets and is solved by an iterative LMS projection method. The
presence of excitation constraints can be very easily dealt with.

The array synthesis problem aims to compute the array excitations in order to satisfy the prescribed
specifications. The array excitations are posed on the radiated field but also on the excitations
themselves, at least in an implicit way.

Consider an array factor ) (u F of a linear equally spaced array of N+ 1 elements;

=
=
N
n
jnu
n
e c u F
0
) ( (59)

u sin where kd u =

The pattern mask can be prescribed as:

) ( ) ( ) ( u M u F u M
U L
s s (60)
where ) ( ), ( u M u M
U L
are the lower mask and upper mask respectively.
Every feasible array pattern must lie within this mask. Two function sets can now be defined [29]

1) The set of all trigonometric polynomial which satisfy the excitation coefficients
2) The set of all functions which satisfy the pattern mask.

An array factor that belongs to the intersection of both the sets is a solution to the synthesis
problem. The synthesis problem is therefore formulated as the search of the intersection of the two
sets.
48


More specifically we are looking for the intersection of:

1) The set M of all functions which fit the assigned mask;
2) The set B of all array factors which can be radiated by a set of excitations satisfying the prescribed
constraints.

The projector over M maps every function f onto the function f according to:

(61)
) ( ) (
) (
) (
) (
) ( ) ( ) (
) ( ) (
) (
) (
) (
) (

<
s s
>
=
u M u f
u f
u f
u M
u M u f u M f(u)
u M u f
u f
u f
u M
u f
L L
U L
U U

Figure 29 below shows the masks and the beamformed pattern.


Figure 29. The masks and the beamformed pattern.
The projector over the set B is defined as a relation between Fourier coefficients [29].

The solution to this problem is found by the method of iterative LMS algorithm. For a dynamic range
constraint we can start from a pattern whose amplitude is constant. The iterative LMS algorithm uses
the principle of pattern inversion as discussed below.


f c E

2
1
2
1
2 1
2 22 21
1 12 11

=
(
(
(
(

=
(
(
(
(

(
(
(
(

n m nm n n
m
m
f
f
f
c
c
c
e e e
e e e
e e e
(62)

where E is the field matrix,
49

c

is array coefficient, and

f

is the electric field vector

The pattern inversion is done as:


inverse) (pseudo solution LMS f ] [ c
f ] [ c ] [
f c
1
1 1
==> =
=
=

T T
T T T T
T T
E E E
E E E E E E E
E E E
(63)

The pseudoinverse calculated using the LMS solution is then applied in multiple iterations until a
pattern fitting into the prescribed mask is obtained following the projection rules described above
and the corresponding array excitation coefficients are stored.


3.3 The advantage of Bucci beamforming algorithm


The Bucci beamforming algorithm has many advantages over the conventional approaches.
Prominent among them being:

1) The whole method lends itself to a very efficient implementation.
2) The algorithm does not require phase information. It performs Power synthesis;
3) It uses iterative LMS method which is both simple and computationally efficient;
4) The projection of farfield onto prescribed masks gives control over varied beamshapes and
sidelobe levels;
5) It has a fast convergence under physical limits;
6) The Algorithm uses the simple array factor in prescribing limits and this allows for a degree of
freedom in having a tradeoff between the achievable beamwidth and sidelobe level.

















50

Chapter 4- The CORA radar system and antenna array
4.1 Outline of the CORA radar system

Figure 3 shows the antenna, front-end and processing van of the DAB-DVB-T (Digital audio broadcast
and Digital video broadcast-terrestrial) radar using the CORA system during the set-up for a
measurement campaign in Germany in 2006. The digital waveforms employing COFDM modulation
are noise-like. They have almost constant spectral amplitude within the band limits of 1.5 MHz (DAB)
and 7.5 MHz (DVB-T). This provides a constant measurement quality at range resolutions of about
100m and 20 m, respectively [5]. This exceeds by far the range resolution of some km, which can be
achieved using FM-radio emission. On the other hand, DAB stations transmit at much lower power
than FM-stations; while the DVB-T transmit power levels are comparable to those of FM-radio [5].


Figure 30 .CORA system architecture [30]
The HF-front-end consists of 16 equal receiver channels. Each of the receiver channels comprises of a
LNA (low-noise amplifier), a tuneable or fixed filter and an adaptive gain control for optimum control
of the ADC (Analog to Digital Converter). The LNAs have a noise figure of 1.1 dB and a gain of 40 dB.
In the current configuration fixed DAB band-pass filters are being used with a pass band of 220 to
234 MHz. A chirp signal with a bandwidth of 1.536 MHz, centred around 227.36 MHz (226.592-
228.128 MHz, channel 12C), generated by a separate signal generator and transmitted to the front-
end by coaxial cable, is used for calibration. A bank of switches provides for calibration of each
receiver channel chain from the LNA to the ADC, excluding only the antenna element.
A/D conversion is realised with 4 FPGA boards, each processing 4 receiver channels. The FPGAs
currently provide only for serialising the 4 channels. Each FPGA board is equipped with 4 ADC-
modules with 14 bit 100 Msamp/sec maximum sample rate ADC-chips. For the processing of DAB
signals, a sample rate of 18.432Msamp/sec is used. It is matched to sample all sub-bands of TV-
channel-12 and is also a multiple of 512 kHz, the basic clock rate of the DAB signal [30].
Each ADC output is fed to an electro-optic converter and linked to the signal and data processing unit
via a fibre-optic cable. In the signal and data processing unit, the optical signals are converted back to
digital data streams of 4 serial channels, each. Four high performance Quad-Opteron computers
51

handle the four data streams. The opto-electric conversion and the feeding of the data to PCI-X-bus
are performed on 64-bit-boards, hosting 4 FPGAs. The FPGAs can additionally be used for pre-
processing the raw data. A data control process controls the storing of the raw data on two 1 TByte
hard disc drives per Quad-Opteron and drives a FiFo RAM, which serves the pre-view real time
visualisation process. Hence, the signal and data processing unit provides 8 Tbyte of hard disc
recording space. For further data storage a raid-array is used with 15.3 Tbyte hard disc space, where
the measured raw data can be saved after each measurement sequence. In the visualisation master
processor, which is served by the four data streams from the Quad-Opterons via 10 Gbit Infiniband
links, performs the processing control, beam forming and detection processing. The pre-view
display is processed on a separate high performance computer in the signal and data processing unit
[30].
The CORA-signal processing is based on determining the intersections of the RDEs and measuring the
direction of the target echo. Here, too, the exact knowledge of the location of the transmitters is
required to determine the RDEs and, together with the direction-of-arrival of the target echo, the
true target range in order to separate false targets from true ones. Figure 30 shows the total system
architecture of the CORA system.
4.2 The architecture of the array antenna
The Discone element being chosen to be the antenna array element, the next step in the antenna
design was the requirement for the design of an antenna array. The antenna array had to be
designed so that it supports several additional functionalities.

Choice of array geometry

Consider the scenario of unknown location of emitters/targets. In such a case equal performance of
the array in all directions must be ensured. There is a need to introduce nulls in the direction of the
direct signal and to be able to steer the main beam to any desired direction. When the direct signal
and reflected signal from the target are received there is a need to attenuate the direct signal by
several dB so as to make its magnitude comparable to the reflected signal form the target. For this
purpose nulls are introduced in the direction of the direct signal. It should also be possible to steer
this null in the entire azimuth plane for a 360 degree rotation. This can be achieved through the
circular array configuration. The linear array fails to operate satisfactorily for the entire 360 degree
rotation. Several other array configurations of the non-uniform nature were tried earlier which
turned out to have non-satisfactory performance for the concerned application.

The basic symmetry of circular arrays offers a number of advantages. It has an ability to compensate
for the effects of mutual coupling by breaking down the array excitation into a series of symmetrical
spatial components. This gives rise to directional patterns that remain constant in shape over broad
bandwidth. So the use of uniform circular array is self-evident.

The circular array thus chosen allows for beam steering and null steering capabilities in the entire 360
degree rotational plane.

Choice of number of elements in the array

The number of channels in the electronic components is 12. The signal processing involved for array
signal processing has an advantage if there be a prime number of elements. The question to be
investigated was, are 11 elements satisfactory.

52

The choice of 11 elements in the circular array can be explained based on two arguments:
1) The signal processing block that is used for the beamforming works best when the number of
signal vectors are prime and this is fixed at 11 because of the number of available channels in
the signal processing block.

2) Theoretically the number of elements in the Uniform circular array can be determined by
taking the ambiguity issues into account. A radio direction finder is said to have a rank-n
ambiguity if one steering vector is a linear combination of n other steering vectors. The
minimal number of sensors in each array can be set initially to not less than 5 (justifying 11)
to avoid the rank-1 ambiguities *34].

The dipole element thus developed was used in a circular array of 11 elements.


































53

Chapter 5 Simulation studies on beamforming and null
steering

The task of being able to steer the beam in the direction of Signal Of Interest (SOI) and to introduce
a null in the direction of interference can be performed using a series of steps involving the
computation of the array radiation patterns and post processing using Matlab. The step of
computing the array radiation pattern proceeds by first computing the embedded radiation pattern
of the element. The beam forming and null steering simulation results performed on 4NEC2 are
presented in the following sections. The study was first done on 4NEC2 to avoid complexity of full
wave analysis at first.
5.1 Preliminary studies with dipole circular antenna array

As mentioned above here we discuss the simulation results obtained using dipole as the antenna
element in 4NEC2. Before embarking into explaining and presenting the results a brief section on the
4NEC2 and its operation is presented below:

5.1.1. Numerical Electromagnetics Code (4NEC2)

4NEC2 is based on a numerical solution of electromagnetic field integrals for thin, perfectly
conducting wire segments using the Method of Moments (MoM). Such segments can be freely
arranged in three-dimensional space and excited in different ways. To calculate the electromagnetic
properties of antennas there are different analysis functions available [31].

The task of determining the current distribution on a wire antenna resulting from an arbitrary
excitation may be put in terms of an integral equation problem. The formulation begins with the
development of an integral expression which defines the electric field resulting from an arbitrary
current distribution on the wire. The integral expression will employ a Greens function which relates
the electric field at an arbitrary observation point to the current at an arbitrary source point. The
integral equation problem then employs the integral expression to relate known electric field
boundary conditions to an unknown current distribution on the wire. A well known formulation for
simple wire antennas is Pocklingtons integral equation. The MoM applies orthogonal expansions to
translate the integral equation statement into a system of circuit like simultaneous linear
equations[37].

The software Numerical Electromagnetics Code (NEC-2) has been developed in the 1970s in the
Lawrence Livermore Laboratory in Livermore, California. The program has later been released as
Public Domain and is available today in many different versions for almost all operating systems and
CPU platforms. 4NEC2 is a completely free windows based antenna modeling tool developed by the
Dutch radio amateur Arie Voors [31].

A NEC-2 input file consists of a number of commands specifying the geometry of a problem, the
excitation and the output data to be calculated. The usage of the different commands and the
different parameters is explained in a detailed Handbook, the complete NEC-2 Manual is available
over the internet [32]. For wire geometries some design criteria have to be considered. To obtain
accurate and reliable results, certain rules for the dimensions and alignment of the wires must be
observed [31]:


54

Segment length: l < /10
Segment radius: r < /100, r < l/8

4NEC2 is an efficient tool to compute radiation pattern of electrically large structures such as arrays.
This feature of 4NEC2 was exploited to make parametric studies on deciding on the array radius of
the final array geometry.


5.1.2 Design of the dipole element

As an example the dipole was designed for operation in the center frequency of at 675 MHz. The
dipole single arm length has to be /4 that turns out around 11cm. The radius of wire is less than
/30 =0.01, =0.44m. The radius was kept to 1/10 of a mm. The number of segments was adjusted
to 7 to get impedance near to the ideal dipole impedance of 73 ohms. The current distribution was
sinusoidal. The dipole antenna is shown in Figure 31.




Figure 31. Dipole antenna designed in 4NEC2. Pink line is the dipole and the green curve shows the sinusoidal
current distribution along the wire











55

The return loss and VSWR characteristics of the dipole are shown in the Figure 32:




Figure 32. Return loss (S11) and VSWR of the designed dipole element.
It can be seen from Figure 31 and Figure 32 that the dipole element designed has a sinusoidal current
distribution and resonates around the center frequency of 675 MHz with a good return loss greater
than 10 dB. The VSWR is also less than 2. These results will further improve if normalization is done
to 73 ohm than 50 ohm. The radiation pattern of the dipole along the horizontal and vertical plane is
shown next in Figure 33.




Figure 33. Horizontal and vertical patterns of the designed dipole element.

The 3D radiation pattern of the dipole element is shown in Figure 34.

56



Figure 34. 3D radiation pattern depicting the total gain in dBi of the designed dipole element.
From Figure 32 and 33, it can be seen that the radiation pattern obtained is omnidirectional, which is
the requirement in our current scenario of beamforming. The pattern is almost circular in the
azimuthal plane which is an added advantage.

The circular dipole array simulation results are next presented.

5.1.3 Design of the dipole circular array

The design of the dipole circular array needs the following important steps to be performed in order
for successful operation.

1) Decision on the number of elements in the array (Explained above)
2) Decision on the right inter-element spacing to reduce mutual coupling
3) Decision on the array radius to be able to avoid grating lobes.

Keeping the above criteria in focus the circular array antenna was designed and the array simulation
results are presented next.
In the array a single element is excited with a voltage source and the other 10 elements are retained
in a passive way as shown in Figure 35. The radiation pattern thus obtained is called the embedded
radiation pattern as in Figure 36 and 37

57



Figure 35. Geometry of 11 element circular array built in 4NEC2 with 0.5 inter-element spacing




Figure 36 Radiation pattern in the horizontal ) 360 0 , 90 (

< < = | u and vertical plane
) 360 0 , 0 (

< < = | u at the center frequency


58


Figure 37. 3D embedded radiation pattern
The embedded radiation pattern, which is the radiation pattern of the excited element when the
other elements of the array are passively located around it along the circle, is omnidirectional as
desired. The embedded element pattern takes into account the effect of the coupling between the
excited element and the passively located elements. . The embedded radiation pattern of a single
element is then used in generating the array manifold matrix of the entire array by rotating and
superposing the embedded element pattern 11 times. This array manifold matrix is then used in the
signal processing steps. This leads to beamforming and null steering which is discussed in the next
section.
5.1.4 Beamforming using the dipole circular array
The circular array geometry has rotational symmetry. This property can be exploited by calculating
the embedded radiation pattern of a single element and using the same to generate the embedded
radiation of other elements in the array by simple rotation and then superposing all the patterns to
get the array manifold. The array manifold matrix thus generated is then processed in Matlab for
beamforming and null steering.
One method of beamforming is the maximum gain method where the goal is to maximize the gain
along the desired direction while keeping the gain along other directions relatively low.
Steps involved in maximum gain method beamforming
Algorithm
1) Import the embedded element pattern from 4NEC2 to Matlab (Figure 38)
2) Calculate the array manifold using the embedded element pattern.(Figure 39)
3) The array manifold matrix is obtained by rotation and superposition of the embedded
element pattern 11 times. (Figure 40)
4) Calculate the array excitation coefficients using the following method
The array excitation coefficients are the conjugate transpose of the row corresponding to
the desired phi direction in the Array manifold matrix
59

5) Calculate the far field as a matrix multiplication of the Array manifold matrix with the array
excitation coefficients a
i

6) The total field is given by

=
=
N
i
i i total
E a E
1
) (|
7) Maximum gain for direction
0
is obtained for array excitation coefficients
i
a

8) Get the corresponding radiation pattern.

Figure 38. Embedded element pattern (Pattern on excitation by a single embedded element). The numerical
value of the electric field represents the E-theta component of the Electric field in dB.

Figure 39. Superposition of fields due to rotation of the element pattern.The E-theta component of the
electric field is plotted against the azimuthal angle for all the 11 elements.
60


Figure 40. Array manifold for half lambda spacing ( Amplitude and phase vs element).
In figure 40 on the left is the plot of the amplitude of E-theta component vs element number. It can
be seen that there is a progressive rotational shift in the radiation pattern on moving sequentially
along the number of element. The figure on the right is a plot of the phase shift in radians vs.
element number. It can be seen that there is a progressive rotational shift in the phases on moving
sequentially along the number of element. The scaling has been chosen so as to accommodate a
grading between the maximum and minimum values in the picture.
The maximum gain method described above is employed in steering the main beam (that is obtained
from the array manifold matrix) in the direction of interest of maximum gain. Presented in Figure 41
are main beams steered to 56 and 206 degrees.

Figure 41. Main beams steered along 56 and 206 degrees using the maximum gain method. The E-theta
component of the electric field is plotted against the azimuthal angle
The algorithm described above can be used to steer the main beam along any direction. However
there is no control over the sidelobe levels. An algorithm that can direct main beams along particular
directions with fixed sidelobe levels is developed using the generalized Bucci projection method. The
61

Bucci algorithm was discussed in detail in chapter 3. Here the results of Bucci beamforming are
presented.
The sidelobe level that can be obtained from a given radiation pattern has a physical limit. The
sidelobe level attainable is a tradeoff with the required beamwidth. The wider the beamwidth the
lower is the sidelobe level. In the Bucci beamforming performed on the Dipole array at 675 MHz a
sidelobe level of 20 dB was aimed at. The results are shown in Figure 42.

Figure 42. Bucci beamforming performed on the circular dipole array at 675 MHz
The upper mask defines the beamwidth and the location of the main lobe and the lower mask is
introduced to get a pattern nearing a flat table top pattern. The extent to which this can be achieved
depends on the physical limits controlled by the array radiation pattern.
5.1.5 Null introduction and steering
Using the null steering principles discussed in section 2.3 the null was introduced and steered along
the desired sidelobe level as shown in the Figure 43 below:

Figure 43: Null introduced at 50 and 271 degrees at the desired sidelobe level of 20 dB
62


- The null can only be present in directions other than the direction of the received signal i.e in
the direction of the direct signals in other directions.
- The introduction of nulls seems to depend on frequency of operation. A study of this is
presented in the section on the actual Discone antenna array.
- The null depth can be easily controlled by the above method by just introducing a proper
weight and scaling it in the formula.

5.2 Uniform Circular Array of Discone antennas

The design of the Discone antenna element was discussed in chapter 2. The Discone antenna model
in 4NEC2 library was used for the initial simulation studies. The Discone antenna model available in
the library was designed for operation around 7 MHz. This antenna had to be scaled down for
operation in the 450-900 MHz range. The Discone antenna element thus designed was adapted into
the circular array antenna of 11 elements. The 4NEC2 simulation results of the same are presented.
The Discone antenna model in 4NEC2 is shown in Figure 44.



Figure 44. The wire model and the 3D radiation pattern of Discone antenna used in 4NEC2
The radiation pattern of the Discone antenna in the horizontal and vertical plane is shown in Figure
45 and the
u
E and
|
E polarization components are shown in Figure 46.

63



Figure 45. The horizontal and vertical radiation pattern of the Discone antenna at the center frequency of
675 MHz.


Figure 46. The
u
E and
|
E polarization components of the Discone antenna at the center frequency of 675
MHz.


It can be seen from figures 44 and 45 that the Discone antenna element exhibits omnidirectional
pattern characteristics as desired. The return loss characteristics of the embedded Discone antenna
element are shown in Figure 47.



64



Figure 47. The return loss characteristics of the embedded Discone antenna element

The characteristics of the 4NEC2 model of the Discone meet the required criteria and therefore it
was used in a uniform circular array system. The top view of the Discone array system is shown in
Figure 48.



Figure 48. Figure depicting the top view of the Discone circular array model with the inter element spacing =
22 cm, and array radius of 77 cm derived through parametric studies, corresponding to center frequency of
675 MHz. Desired operational band is 450-900 MHz.

5.3 Beamforming and null steering in 4NEC2
The array thus modelled in 4NEC2 was simulated and the embedded radiation pattern obtained for
all frequencies in steps of 25 MHz between 450 MHz and 900 MHz. The embedded radiation pattern
was then used in generating the array manifold matrix in Matlab. Post processing in Matlab was
65

performed to achieve beamforming and null steering. The Bucci algorithm described earlier was
applied to obtain the required beamwidth and sidelobe levels. The beamforming at various
frequencies is shown in Figure 49.




Figure 49. Bucci beamforming performed at selected frequencies.

It can be seen from Figure 49 that the beamforming is more accurate around the center frequency
than at the extremes, this can be attributed to the fact that the inter-element spacing was set
corresponding to the enter frequency.

1) It is seen that the inter-element spacing corresponding to half lambda is good enough for the
Bucci algorithm to converge on a wide bandwidth i.e 450-900 MHz.
2) There are no grating lobes present at this spacing.
3) The convergence is very sharp at the 675 MHz operation as expected. This is because the
spacing corresponds to half wavelength at 675 MHz.
4) The spacing could not be set corresponding to 900 MHz (the highest frequency) because of
the limitation of the Discone element dimensions. The interelement spacing turns out much
smaller than the realizable center to center spacing between the discone elements.
The Bucci beamforming algorithm involves calculating array excitation coefficients that generate the
required shaped radiation pattern. The algorithm runs in iterations before it converges to a good
solution (fitting the mask). The array excitation coefficients were calculated at each frequency
between 450-900 MHz in intervals of 5 MHz and the sidelobe level was set to 20 dB. The results of
such wideband beamforming performed are shown in Figure 33.

It can be seen that the beamforming maintains an almost constant sidelobe level of 20 dB
throughout the band except at the extreme ends. This can be attributed to the physical limits set on
the radiation pattern due to the interelement spacing. It can also be seen form Figure 47 that a
66

prominent main beam with a fixed beamwidth is set for the entire frequency range from 450-900
MHz. The main beamwidth can be changed to any desired value by setting masks and so is the
corresponding sidelobe level. There is a trade-off between the two. In general the wider the
beamwidth the lower the sidelobe level (SLL).



Figure 50. Bucci beamforming performed over the wideband from 450-900 MHz in steps of 5 MHz.
As stated earlier it is the requirement of the DVB-T based passive radar to have ability to introduce
null in a particular direction and be able to steer it. The nulls as explained earlier are introduced by
the covariance matric concept and the same was applied to the Discone antenna element array .The
results are shown in Figure 51.



Figure 51. Null steering performed on the Discone antenna array at 50 and 271 degrees.

The null introduced at a particular frequency generated array excitation coefficients at that particular
frequency. It is of interest to see if the same array excitation coefficients generated, at the center
frequency (675MHz) can be used in applying nulls at other frequencies. This gives rise to a study on
the null dependence on frequency. The result of this study is shown in Figure 52.

67








Figure 52. Null dependence over frequency with coefficients calculated for 675MHz applied over the 450-
900 MHz band. Null position at 230/675 MHz.

It can be seen from Figure 35 that the null bandwidth does not operate over the entire range when
the same coefficient is used for other frequencies. The null is displaced from its location partially
when approaching the frequencies lower than 675 MHz and the null gradually disappears for
frequencies higher than 675 MHz.

The array excitation coefficients that are generated by the Bucci algorithm have their relative
amplitudes and phases dependent on the beam direction. For example consider the beam direction
along 140 at 675 MHz. The array excitation coefficients amplitude and phase is plotted vs the
elements in Figure 53.
68


Figure 53. Array coefficients amplitude and phase.

As seen in Figure 53 the amplitude of the array coefficients is maximum along the element number 5
which corresponds to scan direction =140 as shown in Figure 54.It can be seen that the Bucci
masking essentially results in dominant array elements along the desired direction and receded ones
along the other directions in case of a circular array.



Figure 54. 11 element circular array
The embedded element radiation pattern when plotted vs the 450-900 MHz shows a blind spot
around 600 MHz as shown in Figure 55. Since this is due to the limit of the 4NEC2 wire grid model, it
will be compared with full wave analysis of the same in the next section.
69

f [MHz]
|

[

]


500 600 700 800 900
50
100
150
200
250
300
350
-40
-35
-30
-25
-20
-15
-10
-5
0
5

Figure 55. Embedded radiation pattern of element vs frequency
5.4 Need for CST Microwave studio validation

As discussed in the sections above the wideband array antenna has been designed using 4NEC2 and
beamforming and null steering performed on it. The array design procedure is operational with
wideband beamforming and null steering capabilities. Simulation has considered the EM
characteristics of the Dipole and Discone via the 4NEC2 software including mutual coupling. Inter-
element spacing corresponding to half lambda (at 675 MHz) is good enough for the Bucci algorithm
to converge on a wide bandwidth i.e 450-900 MHz and there are no grating lobes present at this
spacing. However there is a blind spot at around 600 MHz as in Figure 38. The existence of which
has to be checked with full wave analysis using CST Microwave studio. CST Microwave studio
provides better options for modeling the Discone antenna when compared to 4NEC2. 4NEC2 Discone
antenna model is a wire grid model and is only an approximation to the actual solid Discone antenna.
The solid Discone antenna model can be more realistically designed and simulated in CST Microwave
studio.

The 4NEC2 Discone antenna model is an approximate model to the actual Discone antenna discussed
in chapter 2. Therefore an actual model of the Discone antenna built in CST Microwave studio will be
used in further studies.

In view of the above two requirements the Discone antenna model was used in CST Microwave
studio and the beamforming and null steering studies were performed on it. The CST Microwave
studio simulation results and analysis is presented in the next section.

5.5 CST Microwave studio analysis

In any design the ability to have a superior and user friendly Graphical User Interface (GUI) always
aids the design process and allows for better handling of the structure in the CAD tool. CST
70

Microwave studio is a CAD tool that combines the power of efficient EM solvers with extremely user
friendly GUI.

CST MICROWAVE STUDIO 2012 is a full-featured software package for electromagnetic analysis and
design in the high frequency range. It simplifies the process of inputting the structure by providing a
powerful solid 3D modelling front end. Strong graphic feedback simplifies the definition of your
device even further. After the component has been modeled, a fully automatic meshing procedure is
applied before a simulation engine is started [36].

CST MICROWAVE STUDIO is part of the CST DESIGN STUDIO suite [35] and offers a number of
different solvers for different types of application. Since no method works equally well in all
application domains, the software contains four different simulation techniques (transient solver,
frequency domain solver, integral equation solver, eigenmode solver) to best fit their particular
applications.
A brief description of the four solvers as defined by CST Studio Suite is given below [35]
- Transient solver: This is a very flexible time domain simulation module that is capable of
solving any kind of S-parameter or antenna problem. It will stimulate the structure at a
previously defined port using a broadband signal. Broadband stimulation enables you to
receive the S-parameters for your entire desired frequency range and, optionally, the
electromagnetic field patterns at various desired frequencies from only one calculation run.
- Frequency Domain Solver: Like the transient solver, the main task for the frequency domain
solver module is to calculate S-parameters. Due to the fact that each frequency sample
requires a new simulation run, the relationship between calculation time and frequency
steps is linear unless special methods are applied to accelerate subsequent frequency
domain solver runs. Therefore, the frequency domain solver usually is fastest when only a
small number of frequency samples need to be calculated. Hence, a broadband S-parameter
simulation with adaptively chosen frequency samples is performed to minimize the number
of solver runs. Especially for lower frequency problems with a small number of mesh cells
(e.g. 50,000) the frequency domain solver may be an interesting alternative to the transient
solver.
- Eigenmode Solver: In cases of strongly resonant loss-free structures, where the fields (the
modes) are to be calculated, the eigenmode solver is very efficient. This kind of analysis is
often useful for determining the poles of a highly resonant filter structure. But of course
there are different applications for the eigenmode solver: with periodic boundaries and a
non-zero phase shift for instance, the eigenmodes are traveling waves. The eigenmode solver
directly calculates the first N frequencies for which fields may exist in the structure, and the
corresponding field patterns.
- Integral Equation Solver: The areas of application for the integral equation solver are S-
Parameter and Farfield/ RCS calculations. The integral equation solver is of special interest
for electrically large models. The discretization of the calculation area is reduced to the
object boundaries and thus leads to a linear equation system with less unknowns than
volume methods. The system matrix is dense. For calculation efficiency the equation system
is solved by the Multi Level Fast Multipole Method (MLFMM). The integral equation solver is
available for plane wave excitation and discrete face ports. Electric and open boundaries are
supported. Far field monitors and surface current monitors can be set in the Frequency
Domain Solver with surface mesh.
The most flexible tool is the transient solver, which can obtain the entire broadband frequency
behavior of the simulated device from only one calculation run (in contrast to the frequency step
approach of many other simulators). It is based on the Finite Integration Technique (FIT) introduced
71

in electrodynamics [36]. This solver is efficient for most kinds of high frequency applications such as
connectors, transmission lines, filters, antennas and more. For the current array antenna purpose the
Time Domain solver was used.

The General CST simulation workflow is listed below:

- Define the Units

- Define the Background Material: The background material has to be set. For an antenna
problem, these settings have to be modified because the structure typically radiates in an
unbounded/open space or half-space.

- Modelling the Structure

- Define the Frequency Range

- Define Ports: The ports define the excitation to the structure. The correct definition of the
port is essential to get accurate S parameters.

- Define Boundary and Symmetry Conditions: The ability to define symmetry planes is very
useful especially when there is a large structure with large number of mesh cells.

- Set Field Monitors: To calculate the farfield at desired frequencies field monitors have to be
set at those frequencies.

- Start the Simulation

When the simulation run completes successfully the output data such as the far field patterns at the
field monitors and the calculated S parameters are available for analysis and interpretation.

5.6 Discone antenna array designed using CST Microwave Studio

The Discone antenna array was designed in CST MWS 2012. The Discone antenna model described in
chapter 2 was used in the array design. The array antenna was built with a diameter of 77 cm as
found by 4NEC2 simulations. One of the Discone antennas was excited with a waveguide port and
the rest of the antenna ports were matched terminations. As the number of mesh cells obtained for
such an antenna was very large - 90 million cells, a symmetry plane was defined to ease the
computation load. The antenna array structure is shown in Figure 56.

72


Figure 56. Discone antenna array model designed in CST MWS. Array diameter is 77 cm. The type of feeding
is coaxial feed from bottom of Discone. The center frequency of operation is 675 MHz.


Figure 57(a). Top view of the array antenna showing a magnetic symmetry plane passing through the center
of the array along its diameter.(b) The magnetic field in the fundamental mode of coaxial cable

As in Figure 57(a) the symmetry plane chosen was a magnetic one and it passes right through the
antenna that is excited thereby dividing the array into two symmetrical halves. The decision on the
magnetic symmetry plane was made on the fact that the single antenna is excited by a coaxial cable
and the excitation of the field will be performed by the fundamental mode in the coaxial cable for
which the magnetic field is shown in Figure 57(b).The magnetic field has no component tangential to
the plane of the structures symmetry. The electric fields are parallel to the boundary and magnetic
fluxes are normal to the boundary. Thus the symmetry plane chosen was magnetic. The definition of
such a symmetry plane reduced the computational effort on the CPU by almost half.


73


(a)

(b)


(c)

Figure 58. The embedded radiation pattern of the single element antenna along the azimuthal plane
) 360 0 , 90 (

< < = | u
at 450, 675 and 900 MHz with the Co and Cross polarization levels is shown in (a), (b) and
(c) respectively.
74


The embedded radiation pattern of the array antenna has omnidirectional characteristics in the
azimuthal plane and it is seen for all the three frequencies of 450,675, and 900 MHz in Figure 58.

The embedded radiation pattern of the array antenna is read from CST into Matlab and then the post
processing is performed in Matlab. The embedded array pattern is rotated and superposed to obtain
the array manifold matrix. The array manifold amplitude and phase plotted against the element
number at the center frequency of 675 MHz is shown in Figure 59.


Figure 59. The array manifold amplitude and phase plotted vs the element number. The rotation of the filed
amplitudes and phases along with the element number is clearly seen. The rotation is bin steps of 360/11
degrees.

The array manifold matrix derived using the procedure described above is then subject to Bucci
beamforming by setting masks and projecting the farfields onto this mask in an iterative fashion as
shown in Figure 60




















75















(a) (b)














(c) (d)

Figure 60. a,b,c and d show the Bucci beamforming performed using embedded element pattern obtained
using CST MWS at 450, 675, 875and 900 MHz respectively

The Bucci beamforming algorithm involves calculating array excitation coefficients that generate the
required shaped radiation pattern. The algorithm runs in iterations before it converges to a good
solution (fitting the mask). The array excitation coefficients were calculated at each frequency
between 450-900 MHz in intervals of 25 MHz and the sidelobe level was set to 20 dB. The results of
such wideband beamforming performed are shown in Figure 58.

76


Figure 61. Bucci beamforming performed over the wideband from 450-900 MHz in steps of 25 MHz.

It can be seen from Figure 61 that the beamforming performs good at all frequencies except around
the higher frequencies where though the required main beam is obtained in the desired direction,
the sidelobe levels (SLL) are a bit high. This can be attributed to the physical limits set on the array
due to its inter element spacing. However the sidelobe levels can be made lower by increasing the
beamwidth at higher frequencies.
The nulls as explained earlier are introduced by the covariance matrix concept and the same was
applied to the Discone antenna element array. Two typical null steering examples are shown in
Figure 62.















Figure 62. Null steering performed on the Discone antenna array at 145 and 251 degrees using CST MWS
radiation patterns
As seen in the post processing results performed using 4NEC2 the embedded element radiation
pattern vs frequency exhibited a blindspot at around 600 MHz. When the post processing was
performed using the CST MWS embedded element pattern the blindspot at 600 MHz was no longer
77

seen. The results are shown in Figure 63. This can be explained by attributing the appearance of
blindspot in 4NEC2, to the approximation done in 4NEC2 modeling of the Discone antenna as a wire
grid model. The blindspot is often caused by array mutual coupling, which tends to direct the
radiation in the plane of the array, rather than as a wave propagating away from the array. Careful
design of the array element shape, size, and spacing can prevent the occurrence of blind spots [38].
Since the Discone antenna array is in free space, there is little scope for the existence of surface
waves, which hints that there should be no blindspot expected. Thus the existence of blindspot in
4NEC2 could be a numerical artefact and this had to be investigated further using CST Microwave
studio. The Discone antenna could be modeled more accurately in CST MWS. CST MWS models and
solves the array in a different manner using FIT (Finite integration technique). The much more
accurate Discone model designed in CST explains the disappearance of the blindspot in CST MWS. It
can also be seen from the CST pattern, Figure 63(b), that there is a more directive pattern at high
frequency.

f [MHz]
|

[

]


500 600 700 800 900
50
100
150
200
250
300
350
-40
-35
-30
-25
-20
-15
-10
-5
0
5


(a) (b)
Figure 63. (a) Blindspot seen around 600 MHz using 4NEC2 embedded pattern. (b) The Blindspot
disappears around 600 MHz on using CST MWS embedded pattern.

5.7 Supporting structure and its influence on the antenna performance

The wideband array antenna designed above needs a supporting structure to rest on. This supporting
structure is expected to have influence over the radiation pattern. Therefore the supporting
structure was designed using CST MWS and its effect on the radiation pattern and mutual coupling
was studied. In addition to the aluminium support rods arranged radially along the array, the support
structure consists of a central mast that extends from the array to the ground. Since the mast is long,
the cylindrical space around it with diameter equal to the array diameter takes up quite a lot of
volume and thus leads to huge number of mesh cells which makes it impossible to solve on the CPU.
However it is well known fact that the mast should not have much influence on the operation as it is
anyway located along the nulls of the radiation pattern. For this reason the studies were done
without the mast and consisting of the support structure only. A single element of the support
structure is shown in Figure 64.

78


Figure 64. Single element of the support structure with parallelopiped aluminium supports and PVC
connectors from Discone to aluminium support.
The support structure designed using the element in Figure 64 is shown in Figure 65.



Figure 65. Support structure for the array antenna built using element in Figure 64.

The 3D radiation pattern with the support structure is shown in Figure 66. The computation was
highly intensive and it was solved using GPU acceleration. Time taken to solve using GPU was 15 hrs.



Figure 66. 3D radiation pattern of the array antenna with support structure
79


In order to assess the coupling effects of the aluminium support rods on the antenna pattern, a full
wave solution of the antenna array with the supporting structure is required. However this is difficult
to compute on the CPU due to memory limits. Therefore a section of the array was considered to
study mutual coupling effects. The section consists of 5 antennas with the center antenna fed and
the mutual coupling along the other antennas on either side studied. This was performed with and
without the supporting structure to see its influence. The results are shown in Figure 67 and 68. It
can be seen from the figures 67 and 68 that the supporting structure does not much influence the
coupling.




Figure 67. The S11, S12 and S13 for a section of the array without the supporting structure.



Figure 68. The S11, S12 and S13 for a section of the array with the supporting structure.

80


It now follows to compare the radiation pattern for this section at different frequencies with and
without the support structure. A comparison of the radiation pattern is presented at 450,675 and
900 MHz in Figure 69.




(a) (b)

(c)
Figure 69. Comparison of the radiation patterns without and with the support structure at a) 450 MHz b) 675
MHz and c) 900 MHz. It can be concluded that the radiation pattern is comparable and does not deviate
much from its normal pattern when a support structure is used.
5.8 Polarimetric beamforming

The DVB-T signals are mostly vertically polarized. The reception of these vertically polarized signals
by the antenna becomes a crucial factor in deciding the polarization purity of the received signals.
The antenna has to have very low levels of cross polarization and good co-polarization patterns. The
beamforming performed keeping in consideration the polarization aspects of the received signals is
called polarimetric beamforming.

The polarimetric beamforming was realized as detailed below:

Beamforming involving, applying the same array excitation coefficients, generated for the
copolarization patterns, on the cross polarization patterns too. The results of such beamforming
done on the Discone antenna array is shown in Figure 70.

81



(a) (b)

(c)
Figure 70. Polarimetric beamforming with same array excitation coefficients for Co-polar and Cross
polarizations at a) 450 MHz b) 675 MHz c) 875 MHz

Polarimetric beamforming involving concatenation of the copolar array manifold matrix with the
cross polar array manifold matrix and then applying masks for both together and doing iterative
Bucci beamforming. The results of such beamforming is shown in Figure 71


(a)
82


(b)


(c)

Figure 71. Polarimetric beamforming involving concatenation of the copolar array manifold matrix with the
cross polar array manifold matrix for frequencies a) 450 MHz b) 675 MHz c) 900 MHz

On comparison between Figure 70 and 71 it can be seen that the polarimetric beamforming involving
concatenation of the copolar array manifold matrix with the cross polar array manifold matrix works
better at the desired frequencies though the performance is satisfactory at higher frequencies.












83

Chapter 6- Experimental validation

6.1 Measurement results of Discone antenna

The Discone antenna elements were fabricated and measured individually for their performance
characteristics. The measurement results of these are presented in this section.

The fabricated Discone antenna is shown in Figure 72.



Figure 72. The fabricated Discone antenna
The construction and design rules for these Discone antennas have already been presented in section
2.1.The measured antenna input reflection coefficient for two different feeding mechanisms is
shown in Figure 73. Type A is for top feeding connector and Type is for bottom feeding connector. It
can be seen that the return loss is well below 10 dB throughout the desired range.

Figure 73. The measured return loss characteristics of the antenna [8]
84

The measured antenna radiation pattern in the elevation plane for Discone with Vertical Polarization
is shown in Figure 74 for three different frequencies.



Figure 74. The measured antenna radiation pattern in the elevation plane for Discone with Vertical
Polarization at three different frequencies [8].
6.2 Experimental set up
The fabricated Discone antenna array along with the supporting structure is placed in the anechoic
chamber for measurement of far field patterns. A log periodic antenna will be used as probe to
evaluate the near fields of the antenna array and the far fields are calculated from them by a
transformation. The probe antenna is able to scan the array along azimuthal and elevation directions.
The transmitting and receiving antennas namely the probe and the antenna array can be used to
calculate the gain of the antenna array too by applying well known Friis formula or modified version
of it:

2
4
|
.
|

\
|
=
R
G G
P
P
r t
t
r
t

(64)
where
r
P is the power available at the input of the receiving antenna,
t
P is the power at the
transmitting antenna and
r t
G G , are gain of antennas at transmitter and receiver respectively and R
is the distance between the antennas.
The antenna can also be checked for beamforming capabilities by exciting the individual Discone
antenna elements with different voltages one at a time and then superposing the patterns and
performing Bucci masking. At the time of writing of the thesis the supporting structure for the array
was being built. So a picture showing the 11 manufactured Discone antenna elements arranged in a
circle is presented in Figure 75. The antenna anechoic chamber for the setup is shown in Figure 76.
Figure 77 shows the control room for the anechoic chamber.
85


Figure 75. 11 Discone antenna elements

(a) (b)
Figure 76 (a). Mount for the array antenna in the anechoic chamber (b) The probe antenna



Figure 77. Control room for the anechoic chamber.
86


Chapter 7-Conclusions and future work

7.1 Conclusions
The goal of this work was to design and develop a wideband antenna array for application in the
DVB-T passive radar with necessary algorithm implementation. The design of the wideband antenna
array was successfully carried out with wideband beamforming and null steering capabilities. The
design studies were started with a dipole antenna array. Several parametric studies were made to
decide on the appropriate inter-element spacing between the elements of the array taking into
account the interaction between the grating lobes and mutual coupling. A full wave EM analysis was
conducted and using these results , later the Discone antenna element was used for the actual design
of the antenna array. The embedded Discone antenna element pattern was used in generating the
array manifold matrix and beamforming and null steering were performed. The phenomenon of
occurrence of blindspot usually observed in arrays was thoroughly investigated using both 4NEC2
and CST microwave studio and it was concluded through simulations that the blindspot does not
actually exist in the designed array.
Extensive simulation studies were performed on the designed antenna array to test its performance
when a main beam and a null are introduced in different directions. Studies were done taking into
account the polarization aspect and correspondingly polarimetric beamforming was performed. The
beamforming done with the Bucci algorithm has shown good results over a wide range of
frequencies around the center frequency. However at the upper and lower end of the frequency
range some problems with high side lobes are seen. - There may be several possible explanations for
this behavior: On one hand side, the antenna elements have a half-wavelength spacing only around
the center frequency of 675 MHz. For lower frequencies where the wavelength increases, the
antenna array becomes electrically small and can hardly satisfy the given main lobe width and side
lobe level requirements. At the upper frequencies where the wavelength is smaller, the spacing
between neighboring antenna elements becomes larger and effects similar to the grating lobes of
linear / planar arrays may occur.
On the other side is the strong increase of cross polarization due to the effects of mutual coupling.
The close inter-element spacing causes strong coupling which affects the current distribution on the
discone antennas and alters the embedded element radiation pattern. It is shown in section 5.6 that
the amount of horizontal polarization in the far-field increases and that, for some directions, it is in
the same order of magnitude as the main polarization. This deteriorates the quality of the far-field
radiation patterns and makes beamforming more difficult, since both orthogonal polarizations must
be controlled with the same number of degrees of freedom, i.e. antenna excitation coefficients.
An additional issue is the beamforming method itself. The projection method proposed by Bucci has
the advantage of a relatively fast convergence but may not necessarily find the best solution
physically possible. Since its original publication several modifications to the algorithm have been
proposed. Also beamforming methods based on global optimization techniques may provide better
results, e.g. Genetic Algorithms. In the scope of the present thesis it was, however, not possible to
test the performance of other beamforming methods. The simulation results indicate that the
87

performance of the antenna array will be sufficiently high for the major part of the DVB-T frequency
range and that passive radar operation will be possible.
Despite the issues mentioned above, the following were the major outcomes of the thesis work:
1) Design of an algorithm to steer the null to any desire direction along the azimuthal plane in a
circular array. The algorithm helps to steer a null of fixed beamwidth throughout the 360
degree azimuthal range while maintaining the same sidelobe level. (the sidelobe levels raise
when the null is in the direction of the main beam which is not relevant for practical
experiments or applications). It involves first performing the beam shaping by introducing
masks and then introducing the null for steering which also maintains the required sidelobe
levels.
2) Array topology optimization (radius and inter-element spacing) for UWB digital beamforming
realized with Bucci algorithm.
3) Study of null dependence on frequency over the operating range.
4) Study of mutual coupling impact on performance of the wideband antenna element
(Discone) to be fully operable on the wideband and the corresponding inter element spacing
when used in an array.
5) Investigation of the influence of the support structure on the array performance.
7.2 Future work
The wideband antenna array was designed and its operation was verified through simulations. The
actual implementation of the designed array antenna in the CORA radar system and its performance
with respect to beamforming and null steering will be performed in the near future. The
performance of the array antenna will be first checked in the anechoic chamber and field
measurements will be done.
Another significant step is the calibration of the array where the antenna is connected to the RF
components of the CORA system. By measuring each channel against a pre-defined reference signal,
fabrication tolerances of the antennas and variations between individual channels of the electronic
back-end may be characterized and compensated for. These data are important for the calculation of
the correct array excitation coefficients and will be stored in the system memory.
As for the beamforming procedure itself, it is recommended to compare the Bucci projection method
used in the present thesis to other methods which may provide better radiation pattern results, as
indicated above.





88

Bibliography

[1] Robert Watson-Watt, Three Steps To Victory, Odhams Press, London, 1958.

*2+ S. S. Swords, Technical history of the beginning of radar, The institute of engineering and
technology 1986

*3+ D.Freeman, the Silent Sentry System, in AoC 5th Multinational Conference of Passive Covert
Radar, November 2007

[4] M.Tobias and A.D.Lanterman, A probability hypothesis density based multitarget tracker using
multiple bistatic range and velocity measurements. Proc. 36th Southeastern Symp. on System
Theory, Atlanta, GA, March 2004, pp. 205209

[5] H. Kuschel and D. O'Hagan, "Passive radar from history to future,"11th international Radar
symposium, pp.1-4, June 2010

[6] LTJ Thornton, Passive and Low-Frequency Radar: Theory in Detection of Stealth Aircraft

[7] M.Conti, F.Berizzi, D. Petri, A.Capria and M. Martorella, "High range resolution DVB-T Passive
Radar," Radar Conference (EuRAD), 2010 European , vol., no., pp.109-112, Sept. 30 2010-Oct. 1 2010

[8] P.Knott, T.Nowicki, H.Kuschel , "Design of a disc-cone antenna for Passive radar in the DVB-T
frequency range," Microwave Conference (GeMIC), 2011 German , pp.1-4, March 2011

[9] T.Nowicki , Internship report, Discone antenna element for circular array construction, August
2011

*10+ C. A. Balanis, Antenna Theory - Analysis and Design 3rd Edition, John Wiley & Sons, 2005

*11+ K.F.Warnick,Antennas and Propagation for Wireless Communications,2005

[12] A.W.Rudge, K. Milne , A.D. Olver and P.Knight, The handbook of antenna design Vol 2, IET
1983

[13] P.J.D Gething, High frequency direction finding,Proc IEE, 113,Jan 1966

[14] J.T. Starbuck , A multiple beam HF receiving aerial system, Radio Electron Engr, April 1969

[15] J.E.G. Wyatt, The design of an omnidirectional UHF wrap around antenna, Journal of Royal
Naval Scientific Service, pp 251-258

[16] D.E.N.Davies and M.S.A.S Rizk, A broadband experimental null steering antenna system for
mobile communications, Radio Electron Engr,48 , Oct 1978, pp 511-517

[17] M.S.Bartlet, An introduction to Stochastic process , Cambridge university press, Cambridge, UK
1956

[18] B. Widrow, P.E Mantey, L.J Griffiths and B.B.Goode, Adaptive antenna systems, IEEE proc, vol
55,No 12, December 1967

89

[19] S.P. Applebaum, Adaptive arrays, IEEE Transactions on Antenna propagation, Vol 24, No 5,
September 1976

[20] L.C. Godara,Application of antenna arrays to mobile communication,Proc IEEE,Vol 85,pp1195-
1245, August 1997

[21] J. Butler and R.Lowe,Beamforming matrix simplifies design of electronically scanned antennas
Electron design ,April 1961.

[22] J.P. Shelton and K.S.Kelleher, Multiple beams from linear arrays, IRE transactions Antennas
propagation, March 1961.

[23] R.G Pridham and R.A.Mucci , Digital interpolation beamforming for low pass and bandpass
signals ,Proc IEEE,Vol 67,No 5 ,1979

[24] S.Anderson, M. Millnert, M.Viberg ,An adaptive array for mobile communication systems, IEEE
transactions on Vehicular technology, Vol 40,No 2,1991

[25] Wei Liu, Stephan Weiss, Wideband beamforming concepts and techniques, Wiley series 2010

[26] R.T .Compton ,Adaptive Antennas: Concepts and Performance, Prentice-Hall, Englewood Cliffs,
NJ, USA, 1988

[27] I.S.Reed, J.D.Mallett and L.E.Brennan, Rapid convergence rate in adaptive arrays, IEEE
Transactions, AES, Vol 10, pp 853-863, 1974

[28] S.Haykin,Adaptive filter theory, Prentice Hall, Englewood Cliffs, NJ, 2002

[29] O.M.Bucci ,G.Franceschetti,G.Mazzarella, G.Panariello,A general projection approach to array
synthesis,1989 Antennas Propagation symposium, San Jose, USA June 1989

*30+ J.Heckenbach and H.Kuschel, Exploitation of DVB-T signals for passive radar, Internal report,
Fraunhofer FHR

[31] Peter Knott, Tutorial wire antennas, Antenna Engineering

[32] NEC-2 User Manual, ftp://ftp.qsl.net/pub/wb6tpu/NEC/nec2prt1.pdf

[33] http://www.microwaves101.com/encyclopedia/phasedarrays.cfm

[34] I.Liberal, D.Caratelli and A. Yaravoy , Conformal antenna array for ultra-wideband
direction-of-arrival estimation, International Journal of Microwave and Wireless Technologies, 2011

[35] CST Studio Suite 2012 , Online guide

[36] Peter Knott, Tutorial patch antenna design, Antenna Engineering

*37+ W.D.Rawle, The MoM: a numerical technique for wire antenna design, High frequency design

[38] A. J. Fenn , D. H. Temme , W. P. Delaney and W. E. Courtney "The development of phased-array
radar technology", Lincoln Lab. J., vol. 12, pp.321 2000

90

[39] M.Simeoni, Antenna systems , ET4015 , TU Delft

[40]L.Josefsson and P.Persson,Conformal Array Antenna Theory and Design, IEEE Press Series on
Electromagnetic Wave Theory, Wiley-IEEE Press, 2005











































91

Publication
D.W.OHagan, V.Basavarajappa, P.Knott, H.Kuschel, M.Ummehofer and M.Simeoni , "A wideband
antenna array for DVB-T based passive bistatic radar applications" , Accepted in the IET Radar
conference, Glasgow, UK, October 2012