NAVAL

POSTGRADUATE
SCHOOL
MONTEREY, CALIFORNIA

THESIS

DEVELOPMENT OF A RECEIVER PROCESSOR FOR

UAV VIDEO SIGNAL ACQUISTION AND TRACKING
USING DIGITAL PHASED ARRAY ANTENNA

by

Fevzi Aytaç Kaya

September 2010

Thesis Advisor: David C. Jenn

Second Reader: Terry Smith

Approved for public release; distribution is unlimited

THIS PAGE INTENTIONALLY LEFT BLANK
 REPORT DOCUMENTATION PAGE Form Approved OMB No. 0704-0188
Public reporting burden for this collection of information is estimated to average 1 hour per response, including the time for reviewing instruction,
searching existing data sources, gathering and maintaining the data needed, and completing and reviewing the collection of information. Send
comments regarding this burden estimate or any other aspect of this collection of information, including suggestions for reducing this burden, to
Washington headquarters Services, Directorate for Information Operations and Reports, 1215 Jefferson Davis Highway, Suite 1204, Arlington, VA
22202-4302, and to the Office of Management and Budget, Paperwork Reduction Project (0704-0188) Washington DC 20503.
1. AGENCY USE ONLY (Leave blank) 2. REPORT DATE 3. REPORT TYPE AND DATES COVERED
September 2010 Master’s Thesis
4. TITLE AND SUBTITLE 5. FUNDING NUMBERS
Development of a Receiver Processor For UAV Video Signal Acquisition and
Tracking Using Digital Phased Array Antenna
6. AUTHOR(S) Fevzi Aytaç Kaya
7. PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 8. PERFORMING ORGANIZATION
Naval Postgraduate School REPORT NUMBER
Monterey, CA 93943-5000
9. SPONSORING /MONITORING AGENCY NAME(S) AND ADDRESS(ES) 10. SPONSORING/MONITORING
N/A AGENCY REPORT NUMBER

11. SUPPLEMENTARY NOTES The views expressed in this thesis are those of the author and do not reflect the official policy
or position of the Department of Defense or the U.S. Government. IRB Protocol number ______N/A__________.
12a. DISTRIBUTION / AVAILABILITY STATEMENT 12b. DISTRIBUTION CODE
Approved for public release; distribution is unlimited
13. ABSTRACT
Air dominance is a key factor concerning today’s warfare. Obtaining air dominance requires having a high degree of situational
awareness. Unmanned aerial vehicles (UAVs) have gained popularity for surveillance and reconnaissance missions and provide
situational awareness to ground-based military units. During operations it is necessary to maintain an uninterrupted data and
control link between the UAV and the ground control station (GCS). This requires GCS antennas with signal-tracking capability.
The work on this research was based on an ongoing project that originally started in 2002. The ultimate purpose is to
design and build a digital phased-array antenna system that can automatically acquire, track, demodulate and decode video signals
from a UAV using commercial-off-the-shelf (COTS) equipment. Previous work done includes integration of hardware components
and development of software modules that allow the array system to auto-track signals from a UAV as well as decode the video
signals in a standard format. The research in this thesis focused on allowing the system to demodulate the video signals acquired by
the digital tracking array. The baseband demodulation technique implemented was previously tested with video signals. A new
technique utilizing tangent-type demodulation of signals was also implemented and tested using a bench-top test setup.

14. SUBJECT TERMS Air Dominance, Situational Awareness, UAV, Digital Phased Array 15. NUMBER OF
Antenna, Tracking, Video Signal, Commercial-Off-The-Shelf, Hardware, Software Modules, NTSC, PAGES
Baseband Demodulation, Tangent Type Demodulation 177
16. PRICE CODE
17. SECURITY 18. SECURITY 19. SECURITY 20. LIMITATION OF
CLASSIFICATION OF CLASSIFICATION OF THIS CLASSIFICATION OF ABSTRACT
REPORT PAGE ABSTRACT
Unclassified Unclassified Unclassified UU
NSN 7540-01-280-5500 Standard Form 298 (Rev. 2-89)
Prescribed by ANSI Std. 239-18

i
THIS PAGE INTENTIONALLY LEFT BLANK

ii
 Approved for public release; distribution is unlimited

DEVELOPMENT OF A RECEIVER PROCESSOR

FOR UAV VIDEO SIGNAL ACQUISITION AND TRACKING
USING DIGITAL PHASED ARRAY ANTENNA

Fevzi Aytaç Kaya

1st Lieutenant, Turkish Air Force
B.S., Gazi University, 2005

Submitted in partial fulfillment of the

requirements for the degrees of

MASTER OF SCIENCE IN ELECTRONIC WARFARE

SYSTEMS ENGINEERING
and
MASTER OF SCIENCE IN ELECTRICAL ENGINEERING

from the

NAVAL POSTGRADUATE SCHOOL

September 2010

Author: Fevzi Aytaç Kaya

Approved by: David C. Jenn

Thesis Advisor

Terry Smith
Second Reader

Dan C. Boger
Chairman, Department of Information Sciences

Clark Robertson
Chairman, Department of Electrical and Computer Engineering

iii
THIS PAGE INTENTIONALLY LEFT BLANK

iv
 ABSTRACT

Air dominance is a key factor concerning today’s warfare. Obtaining air dominance
requires having a high degree of situational awareness. Unmanned aerial vehicles
(UAVs) have gained popularity for surveillance and reconnaissance missions and provide
situational awareness to ground-based military units. During operations it is necessary to
maintain an uninterrupted data and control link between the UAV and the ground control
station (GCS). This requires GCS antennas with signal-tracking capability.

The work on this research was based on an ongoing project that originally started
in 2002. The ultimate purpose is to design and build a digital phased-array antenna
system that can automatically acquire, track, demodulate and decode video signals from a
UAV using commercial-off-the-shelf (COTS) equipment. Previous work done includes
integration of hardware components and development of software modules that allow the
array system to auto-track signals from a UAV as well as decode the video signals in a
standard format. The research in this thesis focused on allowing the system to demodulate
the video signals acquired by the digital tracking array. The baseband demodulation
technique implemented was previously tested with video signals. A new technique
utilizing tangent-type demodulation of signals was also implemented and tested using a
bench-top test setup.

v
THIS PAGE INTENTIONALLY LEFT BLANK

vi
 TABLE OF CONTENTS

I. INTRODUCTION........................................................................................................1
A. BACKGROUND ..............................................................................................1
B. PREVIOUS WORK.........................................................................................6
C. SCOPE OF RESEARCH ................................................................................7
D. ORGANIZATION OF THESIS .....................................................................7
II. ACQUISITION AND TRACKING TECHNIQUES................................................9
A. TARGET TRACKING....................................................................................9
1. Tracking Types.....................................................................................9
a. Single-Target Tracking (STT) ..................................................9
b. Track While Scan (TWS)........................................................10
c. Automatic Detection and Track (ADT) ..................................10
d. Phased Array Radar Tracking................................................11
2. Tracking Techniques .........................................................................11
a. Sequential Lobing ...................................................................11
b. Conical Scan ...........................................................................12
c. Monopulse Tracking ...............................................................13
B. ROBUST SYMMETRICAL NUMBER SYSTEM (RSNS) .......................16
1. RSNS Theory......................................................................................16
2. RSNS Virtual Spacing .......................................................................19
3. Direction Finding Using RSNS with Moduli [5, 9] .........................20
C. DIGITAL BEAMFORMING AND PHASE INTERFEROMETRY........21
D. SUMMARY ....................................................................................................23
III. DIRECT CONVERSION RECEIVER ARCHITECTURE ..................................25
A. QUADRATURE DETECTION THEORY .................................................25
B. DIRECT CONVERSION ADVANTAGES.................................................30
C. DIRECT CONVERSION PROBLEMS ......................................................31
D. SUMMARY ....................................................................................................33
IV. UAV VIDEO SIGNAL PROCESSING ...................................................................35
A. VIDEO SIGNALS..........................................................................................35
1. NTSC Video Signals...........................................................................36
2. Video Frame Format .........................................................................39
3. Digital Video Signals..........................................................................41
B. VIDEO SIGNAL MODULATION AND DEMODULATION ..................42
1. Frequency Modulation (FM) ............................................................42
2. FM Demodulation Techniques..........................................................45
a. Baseband Demodulation.........................................................46
b. Tangent-Type Demodulation..................................................48
c. Demodulation Using a Kalman Filter....................................51
C. SUMMARY ....................................................................................................52
V. ANTENNA SYSTEM STRUCTURE AND COMPONENTS ...............................53

vii
 A. SYSTEM STRUCTURE ...............................................................................53
B. HARDWARE COMPONENTS....................................................................54
1. Antenna...............................................................................................54
2. Down-conversion Block .....................................................................57
a. Low-Noise Amplifier (LNA) ...................................................58
b. Quadrature Demodulator .......................................................59
c. Differential Amplifier .............................................................62
3. Local Oscillator (LO).........................................................................62
4. A/D Converter ....................................................................................63
C. SOFTWARE COMPONENTS .....................................................................64
1. NI LabView Software Overview.......................................................64
2. Tracking Module................................................................................65
3. NTSC Decoding Module....................................................................72
a. Open and Fetch from 5112 SubVI .........................................75
b. Fetch From 5112 SubVI .........................................................76
c. NTSC Processor SubVI ..........................................................78
4. FM Demodulation Module ................................................................81
D. SYSTEM POWER SPECIFICATIONS ......................................................85
E. SUMMARY ....................................................................................................86
VI. DESIGN VALIDATION AND VERIFICATION...................................................87
A. VALIDATION OF FM DEMODULATION TECHNIQUES ...................87
1. Baseband FM Demodulation Simulations .......................................88
a. MATLAB Simulations ............................................................88
b. LabView Simulations ..............................................................93
2. Tangent-type FM Demodulation Simulations .................................96
a. MATLAB Simulations ............................................................97
b. LabView Simulations ............................................................102
B. SOFTWARE MODULES PERFORMANCE VERIFICATION
TESTS AND RESULTS ..............................................................................106
1. NTSC Decoding Module Tests........................................................106
2. FM Demodulation Module Tests ....................................................110
a. Baseband Demodulation Module Tests................................111
b. Tangent-type Demodulator Tests..........................................118
C. SUMMARY OF RESULTS ........................................................................123
VII. CONCLUSIONS AND RECOMMENDATIONS.................................................125
A. SUMMARY AND CONCLUSIONS ..........................................................125
B. RECOMMENDATIONS FOR FUTURE WORK....................................127
1. FM Demodulation Modules ............................................................127
2. NTSC Decoding Module..................................................................128
3. Quality of the FM Signal .................................................................128
4. Other System Enhancements ..........................................................128
APPENDIX: MATLAB CODES ........................................................................................131
A. MATLAB CODE FOR THE SIMULATION OF BASEBAND
DEMODULATION USING A 50 KHZ SINE WAVE .............................131
viii
 B. MATLAB CODE FOR DISPLAYING VIDEO FRAMES......................133
C. MATLAB CODE FOR THE SIMULATION OF BASEBAND
DEMODULATION OF A VIDEO SIGNAL.............................................137
D. MATLAB CODE FOR THE SIMULATION OF TANGENT-TYPE
DEMODULATION OF A SINUSOIDAL WAVEFORM .......................139
E. MATLAB CODE FOR THE SIMULATION OF TANGENT-TYPE
DEMODULATION OF A VIDEO SIGNAL.............................................141
LIST OF REFERENCES ....................................................................................................145
INITIAL DISTRIBUTION LIST .......................................................................................149

ix
THIS PAGE INTENTIONALLY LEFT BLANK

x
 LIST OF FIGURES

Figure 1. Pioneer UAV used by the U.S. in Operation Desert Storm (From [5]). ............3
Figure 2. Conventional (a) and digital (b) phased-array architectures (From [6])............4
Figure 3. UAV tracking beam formation according to range............................................5
Figure 4. AN/FPQ-6 Long-range precision tracking radar (From [15]). ........................10
Figure 5. Rectangular representation of the antenna beam in two switched positions
(From [14]).......................................................................................................12
Figure 6. Conical scan tracking (From [14])...................................................................13
Figure 7. Sum and difference patterns of two squinted beams (From [16]). ..................14
Figure 8. Typical plot of Δ Σ in the vicinity of the null (From [16]). ...........................15
Figure 9. Folding waveforms mapped into RSNS threshold values (After [11])............18
Figure 10. Digital array beamforming on reception (From [19])......................................22
Figure 11. Quadrature detection process (After [20]). ......................................................26
Figure 12. Frequency spectrum for the mixing and LPF process for the I channel
(From [13]).......................................................................................................28
Figure 13. Frequency spectrum for the mixing and LPF process for the Q channel
(From [13]).......................................................................................................29
Figure 14. Frequency domain representation for the filtered output signal (From
[13])..................................................................................................................30
Figure 15. Down-conversion of signal harmonics (From [21]). .......................................33
Figure 16. Left-to-right, top-to-bottom video scanning scheme (After [22]). ..................36
Figure 17. Interlaced scanning process (From [23]). ........................................................37
Figure 18. NTSC analog video signal for a single horizontal line (From [22]). ...............38
Figure 19. Utilization of horizontal and vertical sync signals in a video signal (From
[22])..................................................................................................................39
Figure 20. Timing of an NTSC video line (From [13]).....................................................40
Figure 21. Timing signals for digital video (From [22])...................................................41
Figure 22. Angle modulated waveform in the frequency domain (After [24]). ................44
Figure 23. Baseband Demodulation with I and Q signals...............................................47
Figure 24. Tangent-type FM demodulator using DSP (After [26]). .................................49
Figure 25. System generic block diagram. ........................................................................53
Figure 26. Image of the phased array antenna...................................................................54
Figure 27. PCB layout of the two element subarray (After [13])......................................55
Figure 28. Measured azimuth gain pattern of the six-element array (From [13]).............57
Figure 29. Down-conversion block detailed block diagram of an array channel.............58
Figure 30. Quadrature demodulator operation block diagram. .........................................59
Figure 31. LabView automated dc offset calibration program user interface...................61
Figure 32. Uncalibrated (left) and calibrated (right) I / Q circle plots..............................61
Figure 33. LO (Lab Brick signal generator) connected to the quadrature
demodulator. ....................................................................................................63
Figure 34. VI front panel (left) and block diagram (right)................................................65
Figure 35. Tracking module process flow chart (After [13]). ...........................................67

xi
Figure 36. Data acquisition from demodulator boards by the PXI-5112 board. (Only
the master and a single slave channel shown.) ................................................68
Figure 37. Application of dc offsets for channel 0 and channel 1 of the six array
channels............................................................................................................69
Figure 38. Computation of correction angle from array factor and sum/difference
beams. ..............................................................................................................71
Figure 39. Front panel of the tracking algorithm (From [13]). .........................................72
Figure 40. Process flow diagram for the NTSC decoding module (After [13])................73
Figure 41. NTSC decoding module VI block diagram. ....................................................74
Figure 42. Open and Fetch from 5112 subVI block diagram............................................75
Figure 43. Fetch from 5112 subVI block diagram data acquisition section......................76
Figure 44. Fetch from 5112 subVI block diagram, calculation of vertical sync...............77
Figure 45. NTSC video signal timing diagram (After [13])..............................................78
Figure 46. NTSC Processor subVI block diagram section for sync pulse detection.........79
Figure 47. Field type determination and parameter configuration in NTSC Processor
subVI................................................................................................................80
Figure 48. Video data stream reading process in NTSC Processor subVI........................81
Figure 49. FM Demod Single Fetch VI front panel. .........................................................82
Figure 50. FM Demod Single Fetch VI block diagram.....................................................83
Figure 51. FM Demod Single Fetch Tangent VI block diagram.......................................84
Figure 52. Tangent Demod subVI block diagram. ............................................................84
Figure 53. Block diagram for baseband demodulation simulation. ..................................88
Figure 54. Comparison of original and demodulated signals using baseband
demodulation in MATLAB..............................................................................89
Figure 55. Original video frame used as the modulating signal........................................90
Figure 56. Video signal used as the modulating signal.....................................................91
Figure 57. Demodulated video signal................................................................................91
Figure 58. Demodulated video frame................................................................................92
Figure 59. I vs. Q plot for the video signal........................................................................93
Figure 60. Video signal construction from data file in LabView......................................94
Figure 61. Modulation, mixing, demodulation and amplification process. ......................94
Figure 62. Video signal used as the modulating signal.....................................................95
Figure 63. Demodulated video signal................................................................................95
Figure 64. Attenuation of demodulated signal in the simulation due to high-frequency
deviation...........................................................................................................96
Figure 65. Block diagram for tangent-type FM demodulation simulation........................97
Figure 66. Demodulated vs. original message signal for tangent-type demodulation. ....98
Figure 67. Original video signal......................................................................................100
Figure 68. Demodulated video signal (tangent-type demodulator).................................100
Figure 69. Demodulated video frame..............................................................................101
Figure 70. I / Q circle obtained from tangent-type demodulation of the video signal. ...102
Figure 71. Partial VI block diagram (Part 1) showing the file read and signal
mixing process. ..............................................................................................103
Figure 72. Partial VI block diagram (Part 2) showing the filtering, demodulation,
improvement, dc offset and amplification process. .......................................104

xii
Figure 73. Original video signal before modulation and demodulation..........................105
Figure 74. Demodulated video signal obtained from tangent-type demodulator............105
Figure 75. Snapshot of the NTSC decoding module front panel while running. ............107
Figure 76. Tilted video image due to inappropriate setting of the horizontal sync
adjust. .............................................................................................................108
Figure 77. Video signal region showing the beginning of a new field. ..........................108
Figure 78. Digital camera image of the location captured by the video camera.............109
Figure 79. Demodulation test set up................................................................................110
Figure 80. Output obtained from the FM Demod Single Fetch VI. ................................111
Figure 81. Demodulated sine wave with 50-kHz frequency and 0.6-V amplitude. .......112
Figure 82. Demodulated square wave obtained from the baseband demodulation
module............................................................................................................113
Figure 83. Demodulated ramp wave obtained from the baseband demodulation
module............................................................................................................113
Figure 84. Snapshot of the baseband demodulation module running with video signal
input. ..............................................................................................................114
Figure 85. Demodulated video signal obtained from the baseband demodulation
module with fcut-off = 9 MHz. .........................................................................115
Figure 86. Zoomed video signal obtained from the baseband demodulation module
with fcut-off = 9 MHz. ......................................................................................115
Figure 87. Demodulated video signal obtained from the baseband demodulation
module with fcut-off = 6 MHz...........................................................................116
Figure 88. Demodulated video signal obtained from the baseband demodulation
module with fcut-off = 100 kHz........................................................................117
Figure 89. Zoomed video signal obtained from the baseband demodulation module
with fcut-off = 100 kHz zoomed. ......................................................................117
Figure 90. Output obtained from the FM Demod Single Fetch Tangent VI. ..................118
Figure 91. Demodulated sine wave with 50-kHz frequency. ..........................................119
Figure 92. Demodulated square wave obtained from the tangent-type demodulation
module............................................................................................................120
Figure 93. Demodulated ramp wave obtained from the tangent-type demodulation
module............................................................................................................120
Figure 94. Snapshot of the tangent-type demodulation module with video input (fcut-
off = 9 MHz). ..................................................................................................121
Figure 95. Demodulated video signal obtained from the tangent-type demodulation
module (fcut-off = 9 MHz)................................................................................122
Figure 96. Demodulated video signal obtained from the tangent-type demodulation
module (fcut-off = 100 kHz). .............................................................................122
Figure 97. Zoomed view of the video signal obtained from the tangent-type
demodulation module (fcut-off = 100 kHz).......................................................123

xiii
THIS PAGE INTENTIONALLY LEFT BLANK

xiv
 LIST OF TABLES

Table 1. Voltage threshold values for RSNS with moduli [5, 9] (From [13]). ..............20
Table 2. Properties of the PCB layout of the two-element sub arrays (From [13]). ......56

xv
THIS PAGE INTENTIONALLY LEFT BLANK

xvi
 LIST OF ACRONYMS AND ABBREVIATIONS

A/D Analog to Digital

ADT Automatic Detection and Track

AF Array Factor

AM Amplitude Modulation

AOA Angle of Arrival

BER Bit Error Rate

COTS Commercial off The Shelf

DBF Digital Beamforming

DF Direction Finding

DSP Digital Signal Processing

EMI Electromagnetic Interference

ESA Electronically Scanned Array Antennas

EW Electronic Warfare

FM Frequency Modulated/Modulation

FOV Field of View

GCS Ground Control Station

HPA High Power Amplifier

HPBW Half-Power Beam Width

JDAM Joint Direct Attack Munitions

I In-phase

IC Integrated Circuit

IP Intercept Point

xvii
IRE Institute of Radio Engineers

LNA Low Noise Amplifier

LO Local Oscillator

LPF Low Pass Filter

MGTOW Maximum Gross Takeoff Weight

MMIC Monolithic Microwave Integrated Circuits

MTBF Mean-time Between Failures

NI National Instruments

NTSC National Television Standards Committee

OODA Observation-Orientation-Decision-Action

PAL Phase Alternating Line

PCB Printed Circuit Board

PLL Phase Locked Loop

PM Phase Modulation

Q Quadrature

RF Radio Frequency

RSNS Robust Symmetric Numeric System

SAR Synthetic Aperture Radar

SECAM Systeme Electronic Pour Avec Memoire

SNR Signal to Noise Ratio

STT Single Target Tracking

TBM Tactical Ballistic Missile

T/R Transmit/Receive

TWS Track While Scan

xviii
UAV Unmanned Aerial Vehicle

UCAV Unmanned Combat Aerial Vehicle

USA Unites States of America

USAF Unites States Air Force

USB Universal Serial Bus

VCO Voltage Controlled Oscillator

VI Virtual Instrument

xix
THIS PAGE INTENTIONALLY LEFT BLANK

xx
 EXECUTIVE SUMMARY

In today’s warfare the concept of air dominance has emerged as a key factor. The ability
to gain air dominance is crucial in the sense that it saves lives, affects the duration of the
combat and ensures the accomplishment of national and military objectives. The
capability of obtaining air dominance requires a high degree of situational awareness
during combat. Research and developments in several areas have led to sensor systems
that can increase the level of situational awareness in a combat environment. Unmanned
aerial vehicles (UAVs) are among the most important of these systems. From their
operational beginnings as converted target drones to the more sophisticated battlefield
observation platforms, UAVs have finally gained acceptance by the military community.
During military operations, UAVs provide a high degree of situational awareness by
carrying out surveillance and reconnaissance missions that lead to gaining air dominance
against adversaries.

One of the biggest challenges encountered in UAV operations is the ability to

establish secure and reliable data links between the UAV and ground stations. Disruption
in data links can easily occur and the quality and speed of the data received may be
adversely affected. This necessitates an antenna, possibly with an electronic beam
steering and tracking capability, to sustain a reliable data link between the UAV and the
ground control station (GCS). When the signal is received at the GCS, signal processing
techniques are used to retrieve the necessary data (e.g., video).

The advancements in the field of signal processing have provided several

advantages. Many tasks that used to be done via hardware can now be done using
software. With this opportunity, today’s GCSs can be built at relatively lower costs, in
smaller spaces, and are easily upgraded. Another important technological advancement is
the phased-array antenna. The advent of relatively low-cost monolithic microwave
integrated circuits (MMICs), automated assembly of microwave components, and new
low-cost high-speed digital processors led to the development and optimization of these

xxi
antennas. Phased arrays have the capability to steer their beams electronically in
nanoseconds by using processor-controlled phase adjustments for each individual antenna
element.

The objective of the research presented in this thesis, as part of an ongoing

project, is to develop a digital tracking array for use in communications with a UAV
using commercial-off-the-shelf (COTS) components that reduce overall cost. The
research previously done in key areas comprised a robust symmetric number system
(RSNS) direction-finding processor, and digital beamforming (DBF) as well as
demodulation and decoding techniques for video data transmission. The previous
research focused on the optimization of the demodulation techniques in order to bring the
digital tracking array the capability to successfully retrieve and display the video data
embedded in the frequency modulated (FM) signals received by the antenna. An in-depth
review of the properties and functionalities of the current system and its components was
also made in order to provide a useful reference for future studies.

The feasibility of utilizing new demodulation techniques for the proper

demodulation of the video signals was previously investigated. The tangent-type
demodulation technique was studied as a novel approach for the demodulation process.
The baseband FM demodulation technique implemented previously was limited to simple
sinusoidal input waveforms. This technique was enhanced to handle video signals. The
applicability of these two techniques in the receiver system was validated by computer
simulations done in both MATLAB and LabView. It was seen that FM video signals
received from a UAV can be successfully demodulated by using either the baseband or
tangent-type demodulation techniques.

The development of the software modules for the receiver processor was done
using a modular approach to complement the previous work done. Based on the results of
the simulations, an FM demodulation module utilizing the tangent-type demodulation
technique for the demodulation of FM video signals was developed. Additionally, the
previous FM demodulation module, using the baseband demodulation technique was
modified for video signal demodulation. Both modules were developed for operation on a
single array channel. Subsequently, both modules were tested using a bench-top test
xxii
setup. Different waveforms were used as input-modulating signals. The waveforms tested
were chosen to range from simple to complex signal content. Hence they included
sinusoidal, square, and ramp waves and, for the more complex representations, real-time
streaming video signals.

The newly developed tangent-type FM demodulation module showed similar

performance compared to the previous baseband demodulation module. Both modules
were able to demodulate the sine wave. The demodulated square and ramp waves by both
modules had distortions due to the attenuation of high-frequency components by the
filters. However, the fundamental frequencies of both waves were retrieved. The
baseband video signals were frequency modulated and then successfully demodulated
using both demodulation modules. However, the video signals obtained had excessive
distortion levels which prevented decoding of the video data. The filter cut-off
frequencies were decreased to eliminate the high-frequency noise. A slight improvement
was obtained; however, the video data still could not be decoded and displayed by the
video decoder. This was due to the corruption of the synchronization (sync) pulses within
the video signals. The video decoder used measures the duration of the sync pulses and
compares the measured value with the minimum duration required for a valid sync pulse.
The distortions on the demodulated signals prevented the decoder to validate the sync
pulses. As a result they were considered as noise and the video could not be displayed.

The development process also included the modification of the user interfaces of
the previously developed software modules. The previous user interfaces of the National
Television Standards Committee (NTSC) decoding module and the baseband FM
demodulation module were examined, analyzed and modified to provide more efficient
functionality and analysis. The functionality of the modified NTSC decoding module was
tested using a bench-top test setup and verified using unmodulated streaming video
obtained from a video camera as the input. However, when integrated within the
demodulation modules, the module failed to decode and display the demodulated video
obtained at the output of the demodulators due to the excessive noise level of the
demodulated signals.

xxiii
THIS PAGE INTENTIONALLY LEFT BLANK

xxiv
 ACKNOWLEDGMENTS

This thesis has been the consequence of hard work, cooperation and high
motivation. As the author of this thesis, I would like to express my gratitude to my thesis
advisor, Prof. David C. Jenn from the ECE Department of the Naval Postgraduate School
(NPS), who made this work a great learning experience for me. His outstanding guidance
through his extensive knowledge had great influence in the success of this thesis. It has
been a great pleasure and honor to work with him. I would also like to thank my program
officer, Lt. Col. Terry Smith for his contributions to this work as the second reader as
well as his sincere leadership and guidance through my postgraduate education as my
program officer.

Special thanks also go to Robert D. Broadston, Director of ECE Microwave

Laboratory, for his extremely generous support in my studies with both the hardware and
software used in this thesis. I appreciate all the time he spent in helping me set up the
equipment and troubleshoot the problems that I encountered during my studies.

I would like to thank my father, Macit Kaya; my mother, Aynur Kaya; and my
sister, Özge Kaya, for their endless support, belief and trust in me during my entire
military career. I would also like to express my special gratitude to my fiancée, Gökselin,
for her admirable love, support and patience during my time at NPS.

I would like to express my appreciation to the Turkish Air Force Command for
giving me the opportunity to complete my postgraduate education at NPS.

Finally, I would like to thank Mr. Breen Dix, of Digital Consulting Services, for
editing the thesis, Mrs. Donna Cuadrez from the NPS thesis processing office for helping
me format the thesis, and Mrs. Pam Silva for her guidance through the preparation
process.

xxv
THIS PAGE INTENTIONALLY LEFT BLANK

xxvi
 I. INTRODUCTION

A. BACKGROUND

Air dominance has emerged as one of the most important factors concerning
today’s warfare. The ability to gain air dominance is crucial in the sense that it saves
lives, affects the length of the conflict and the quality of the peace, and ensures the
accomplishment of national and military objectives [1]. Obtaining air dominance
requires a high degree of situational awareness during combat. In the last few decades,
various systems have been developed to increase the level of situational awareness in a
combat environment. Unmanned aerial vehicles (UAVs) are among the most important
of these systems. From their operational beginnings as converted target drones to today’s
more sophisticated battlefield observation platforms, UAVs have finally gained
acceptance by the military community. During Operation Desert Storm in 1991, it was
even reported that an Iraqi ground unit attempted to surrender to a U.S. Pioneer UAV as
it buzzed overhead [2].

Today’s UAVs can carry out a variety of military and civilian applications. One
of the most common missions in which UAVs are extensively used is remote sensing.
This includes surveillance and reconnaissance, target recognition, search and rescue and
even traffic monitoring. These kinds of UAVs employ various kinds of electromagnetic
sensors in order to achieve their missions. Electromagnetic sensors include visual
sensors, infrared or near-infrared cameras, and radar systems [3].

UAVs can also be used for transporting goods. For example, a UAV can be used
to transport emergency medical supplies to frontline troops in a battlefield environment.
UAVs are uniquely capable of penetrating into areas where it is highly dangerous for
human beings to approach. This feature makes UAVs capable of monitoring
contaminated areas (such as the Chernobyl disaster) without risk to human life.

1
 Today, new military applications are opening to UAVs—beyond their traditional
surveillance and reconnaissance missions. Currently, UAVs are effectively used in
electronic warfare (EW) as airborne jammers, decoys, and emitter locators. Future roles
are likely to include anti-radar hard kill, communications relay, pre-launch and boost-
phase tactical ballistic missile (TBM), and both lethal and non-lethal submunition
dispensing.

Being a complex aerial system, UAVs have several design challenges. One of the
biggest problems encountered in UAV operations is establishing secure and reliable data
links between the UAV and ground stations. Disruption in the data link can easily occur,
and the quality and speed of the data received may be adversely affected. Of the 36
Pioneer UAVs (Figure 1) lost by the U.S. during Operation Desert Storm, only one was
lost to enemy fire. Eleven of them, however, were lost due to loss of data link. From the
Pioneer example, it can be understood that it is important for the ground control station
(GCS) to accurately track the signals coming from a UAV. This necessitates a directional
antenna, possibly with electronic beam steering and with tracking capability to sustain a
reliable data link between the UAV and the GCS. When the UAV signal is received at
the GCS, signal processing techniques are used to retrieve the necessary data (e.g.,
video).

Considering the operational environment, we see that the observation-orientation-

decision-action (OODA) loop, which is described in [4], has great importance in all
kinds of military operations. The OODA loop is a function of the characteristic tempo of
operations and, hence, situational awareness. A war-fighting enterprise that can process
the entire OODA cycle quickly, observing and reacting to unfolding events more rapidly
than an opponent, can “get inside” the opponent’s decision cycle and gain a military
advantage. From this perspective, establishing a reliable and uninterrupted data link
between a GCS and a UAV will decrease the time needed in the decision-making process
(operational tempo). Once disrupted, an enemy’s operational tempo may be stretched in
time, thus providing additional opportunities to exploit the enemy’s weaknesses.
2
Figure 1. Pioneer UAV used by the U.S. in Operation Desert Storm (From [5]).

There has been much technological advancement in the field of signal processing
in the last few decades. Many tasks that used to be done via hardware can now be done
using software that employs digital signal-processing techniques. With these
developments, today’s GCSs can be built at relatively lower costs and occupying less
space. Another important technological advancement is the phased array antenna. The
advent of relatively low-cost gallium arsenide (GaAs) monolithic microwave integrated
circuits (MMICs), automated assembly of microwave components and new low-cost,
high-speed digital processors led to the development and optimization of these antennas.
They have the capability to steer their beams electronically in nanoseconds by using
phase adjustments for each individual antenna element.

Phased arrays consist of a collection of individual antennas that are geometrically

arranged and excited so as to provide the desired radiation characteristics. Traditionally,
the elements are connected by a system of microwave transmission lines and power

3
dividers (the beamforming network). The array radiates or receives energy over a spatial
angle (beam) that is measured by its half-power beam width (HPBW), which is also
referred to as the field-of-view (FOV) [6].

In Figure 2, block diagrams of a conventional and a digital phased array antenna

are shown. For the data transmission between a UAV and a GCS using a conventional
phased array, the data is encoded onto a baseband waveform that is up-converted to the
operating band—2.4 gigahertz (GHz) for the proposed system—then amplified and sent
to the antenna. On the receive side (i.e., GCS), the signal out of the antenna is down-
converted to baseband in the receiver and the information contained in the received
message is reconstructed. For digital phased arrays, each element has its own transmitter
and receiver combined into a transmit-receive (T/R) module. This eliminates the need for
a beamforming network. The received signal processing is performed in a digital
beamformer. Advancements in the development of low-cost miniaturized electronics and
fast, powerful, computer processors have made this architecture practical for many
applications [6].

Figure 2. Conventional (a) and digital (b) phased-array architectures (From [6]).

4
 Another important advantage that comes with phased-array antennas is reliability.
With their fixed structure, phased-array antennas eliminate the need for gimbal systems,
servos and rotary joints to steer the beam of the antenna. These components are possible
sources of mission failures because of breakdowns even though there is nothing wrong
with the processors or the array itself. Using electronic beam steering eliminates these
problems and provides an improved mean-time between failures (MTBF), which may
even extend to a period as long as the lifetime of the antenna.
In a phased array, digital processing techniques can also be applied to track the
transmitting UAV. For a UAV located at a close range to the GCS, the beam should be
formed wider in order to keep up with the faster change in the angle of arrival (AOA) of
the signal. Similarly, if the range between the GCS and the UAV is long, a narrow beam
is preferred since more power density in a narrow angular range is required. This concept
is illustrated in Figure 3.

Figure 3. UAV tracking beam formation according to range.

With the help of high-speed digital processing techniques, real-time video

streams can be relayed from a UAV to personnel on the ground, providing better
situational awareness. This video data can also be transmitted to the decision-makers
located in the headquarters away from the battlefield and the decision process can be
efficiently supported.
5
B. PREVIOUS WORK

This research is part of an ongoing project that originally started with Bartee [7]
in 2002. Ong [8] studied the receiver part of the project and investigated the periodic
phase error attributed to the modulator boards. He was able to eliminate the phase errors
by modifying the bias voltage conditions that were used. Burgstaller [9] characterized the
critical commercial-off-the-shelf (COTS) hardware to reduce cost in 2006.

Using the device information obtained in the studies mentioned above, Gezer
[10] designed and built a digital array employing the National Instruments (NI) cRIO
system to process down-converted data from the COTS hardware. Gezer also determined
system design parameters, such as element spacing, scan angle, HPBW, gain, number of
elements, signal-to-noise ratio (SNR) and operating range.

Later, direction finding (DF) methods were studied by Lee [11] and Benveniste
[12] using the Robust Symmetrical Number System (RSNS). DF is of interest in this
study for the purpose of acquisition. Lee developed a three-channel phase interferometer
using RSNS and tested the antenna array using relatively prime moduli for direction
finding. Benveniste developed a single-channel RSNS system with virtual spacing for
direction finding.

Finally, Tan and Pandya [13] researched advanced techniques of the RSNS and
Digital Beamforming (DBF) to accurately acquire and track a UAV signal using the
monopulse technique. They determined optimum power levels for the direct-conversion
receiver to ensure optimum performance and integrated the antenna array and the
receivers into a single chassis. They also tested the antenna array in an anechoic
chamber.

For the software part, using LabView, Tan and Pandya implemented the
monopulse technique and software modules to retrieve a frequency modulated (FM)
video signal encoded using the National Television System Committee (NTSC) standard
format. They also tested the antenna using different transmit powers and different AOAs.

6
C. SCOPE OF RESEARCH

This research effort furthers the ultimate goal of designing, building and testing
an antenna array and digital receiver that utilizes DBF for tracking and the RSNS to
accurately acquire an FM video signal coming from a UAV. The hardware used was
selected from COTS equipment in order to reduce cost.

Software modules were developed and tested to retrieve an FM video signal

which was encoded using the NTSC format. Simulations were designed using MATLAB
and LabView software to verify the methods used for demodulation. Software module
tests were conducted using initially basic signals, such as sine waves, in order to provide
a better understanding for the behavior of the demodulation modules designed. Further
tests were performed using real time video signals as the modulating signals. The results
obtained from the different demodulation methods were compared.

D. ORGANIZATION OF THESIS

The tracking methods used in tracking antennas are presented in Chapter II.
Tracking techniques such as sequential lobing, conical scan and monopulse tracking are
also introduced. The theory of the RSNS is given and its application to direction finding
is briefly explained. Finally, the theoretical background for DBF and phase
interferometry is presented.

The theory behind the direct-conversion receiver architecture is covered in

Chapter III. The quadrature detection theory is introduced and the process of down-
conversion of signals to baseband is explained. Finally, the advantages and associated
problems of the direct conversion architecture are discussed.

The properties of analog and digital video signals are presented in Chapter IV.
The formats for the analog video formats are introduced with emphasis on the NTSC
format. The frequency modulation and demodulation techniques for baseband in-phase

7
( I ) and quadrature (Q) signals are discussed. The demodulation techniques discussed
include baseband demodulation, tangent-type demodulation and demodulation using
Kalman filtering.

The hardware and software components of the UAV digital tracking array system
are described in Chapter V. The parameters indicating the performance of the
components (e.g., antenna) are also given. The LabView programming environment is
introduced in order to provide familiarization with the software modules developed.
Finally, the details of the software modules are presented.

The validation and verification process for the software modules developed is
presented in Chapter VI. The simulations done in order to validate the applicability of the
baseband and tangent-type demodulation techniques to the system are presented. The
verification process, which includes bench-top testing of the software modules
developed, is covered and the results obtained are discussed.

A summary of the work done and the conclusions drawn are provided in
Chapter VII. Recommendations for future work are also made.

The Appendix contains MATLAB codes used for the simulation of the software
modules in order to validate FM demodulation techniques presented.

8
 II. ACQUISITION AND TRACKING TECHNIQUES

A tracking array must first find and acquire its target before starting to operate as
a tracker. In this section, target acquisition and tracking techniques, which were utilized
in the digital array antenna hardware, are covered. In Section A, basic principles of
tracking are introduced. In Section B, a brief overview of direction finding with the
RSNS is covered. Finally, in Section C, the key principles of DBF are discussed.

A. TARGET TRACKING

In this section, the target tracking concept is classified into two categories. In the
first part, the types of target tracking are discussed. In the second part, the techniques
used for target tracking are covered.

1. Tracking Types

There are four main types of target tracking that are described in the following
sub-sections.

a. Single-Target Tracking (STT)

Single-target tracking is used in applications that require a high data rate

(i.e., observations per second), such as guided-missile weapon control radar. An example
of this kind of radar is shown in Figure 4. As the beam of the antenna follows the target,
an angle-error signal is generated continuously. Then, a closed-loop servo system is
employed to keep the angle-error small in order to track the target with high accuracy
[14].

9
 Figure 4. AN/FPQ-6 Long-range precision tracking radar (From [15]).

b. Track While Scan (TWS)

This type of radar scans a limited angular sector to track more than one
target simultaneously within the coverage of the antenna. Common applications include
air defense radars and airborne intercept radars [14].

c. Automatic Detection and Track (ADT)

Different from the STT, the position of the antenna in an ADT is not
controlled by the processed tracking data. Instead, the antenna makes rotations
continuously and the rate at which observations are made depends on the time for the
antenna to make one rotation. This results in a lower data rate than the STT, but it
provides the capability of tracking a larger number of targets [14].

10
 d. Phased Array Radar Tracking

Phased array radars have the capability to steer their beams electronically
in nanoseconds. This provides many advantages, such as high data rates and the
capability to simultaneously track a large number of targets. With these features,
phased-array radars can be considered as the combination of the STT and the ADT radar
[14]. Most of the work in this thesis focuses on phased array tracking.

2. Tracking Techniques

In this section, three main techniques for angle tracking are discussed. These
techniques are: (1) sequential lobing, (2) conical scan and (3) monopulse tracking. Of
interest in this thesis is the monopulse tracking technique. However, the other two
techniques will be introduced as well.

a. Sequential Lobing

In this tracking technique, the antenna beam is time-shared between two

separate directions to obtain the angle measurement in a sequential manner. This process
is demonstrated in Figure 5. The difference in amplitude between the voltages obtained
in Position 1 and Position 2 is a measure of the angular displacement of the target from
the switching axis between the two beam positions. The beam of the antenna is then
turned to the direction of the position which has the larger signal level. If the difference
is zero, it means the target is on axis.

11
 Figure 5. Rectangular representation of the antenna beam
in two switched positions (From [14]).

b. Conical Scan

The concept of conical scan is illustrated in Figure 6. The antenna beam is

squinted and then rotated. The angle between the axis of rotation and the antenna beam
axis is called the squint angle. The axis that shows the direction of the axis relative to the
antenna is called the target axis. If the target axis and the rotation axis are aligned, the
echo responses received as the beam is rotated have equal amplitudes. If unequal
amplitude is detected, this means the target is not in the direction of the rotation axis.
Hence, the direction of the target can be determined relative to the rotation axis and
tracking can be achieved with continuous utilization of this scheme.

12
 Figure 6. Conical scan tracking (From [14]).

c. Monopulse Tracking

In monopulse tracking, two or more simultaneous beams are compared to

obtain the angular location of a target. The name monopulse comes from the fact that a
single pulse can be used for the measurement of the angle. Since the signals received
from the multiple antenna beams are simultaneous, the accuracy is improved compared
to conical scan and sequential lobing, which use time-shared, single-beam tracking
systems. Conical scan and sequential lobing techniques suffer degradation when the echo
signal amplitude changes with time. Conversely, the accuracy of the monopulse is not
affected by amplitude fluctuations of the target echo [14].

Monopulse angle measurements can be made by using two different

methods. These are called amplitude-comparison monopulse and phase-comparison
monopulse.

For amplitude-comparison monopulse, two antenna beams are generated

at small positive and negative squint angles. By the addition and subtraction of the two

13
beams, the sum and the difference beams are obtained, respectively. For transmission,
only the sum beam is used, while both sum and difference beams are used for reception.
In Figure 7, an example of the two beams generated is shown, along with the sum and
difference beams obtained. For tracking, the objective is to position the null of the
difference pattern in the direction of the target.

Figure 7. Sum and difference patterns of two squinted beams (From [16]).

The direction of angle error is obtained from the comparison of the

amplitude of the signal obtained from the sum beam and the amplitude of the signal
obtained from the difference beam. This information can be obtained by the
difference-to-sum voltage ratio, which is given as [16]
Δ difference voltage
= . (2.1)
Σ sum voltage

14
If this ratio is normalized and plotted versus the pattern angle in the vicinity of the null of
the pattern, a linear region is observed, depending on the antenna beamwidth and squint
angle. The plot of a typical Δ Σ ratio is given in Figure 8.

Figure 8. Typical plot of Δ Σ in the vicinity of the null (From [16]).

In the linear region shown in Figure 8, the Δ Σ ratio can be approximated

as [16]
Δ
≈ Kθ (2.2)
Σ
where K is called the monopulse slope constant. The monopulse slope constant can be
determined from measurement or simulation.

In phase-comparison monopulse, as in amplitude-comparison monopulse,

two beams are used to obtain an angle measurement. However, the two beams look in the
same direction and cover the same space rather than being squinted [14]. The amplitudes
of the received signals will be the same; however, the phases will be different. Hence, the
measurement of the phase difference of the signals received in the two antennas can
provide the angle θ to the target. This is the classical interferometer measurement.

15
 Monopulse tracking can suffer from some errors related to noise and
imperfections in the antennas used. Antenna errors include null shifting and null filling
due to antenna illumination errors. Another source of error is the target glint. Glint refers
to the distortion of the wavefront received from the UAV due to environmental and
interference effects. If a distorted wavefront is received, the target direction calculated
may differ from the actual target direction [16].

B. ROBUST SYMMETRICAL NUMBER SYSTEM (RSNS)

The tracking techniques discussed in the previous section use amplitude or phase
comparison in order to obtain the angle of a target. These techniques require the use of at
least two antennas separated by a known spacing, which is referred to as the baseline.
Another direction finding approach is the RSNS, which is a modular preprocessing
architecture. The RSNS has been used for both analog and digital signal processing.
When applied to direction finding, RSNS can provide reduced baselines and high
resolution [12].

1. RSNS Theory

The symmetric waveform that is coming out of an antenna can be mapped into a
sequence of numbers as in the following [11]:
xm = [ 0,1, 2,..., m − 1, m, m − 1,..., 2,1] (2.3)

where xm is a row vector and m is an integer (m > 0) called the modulus. The sequence

xm is called the RSNS sequence. In an N -channel RSNS sequence, each element in xm

is repeated N times. Hence the sequence xm takes the following form [11]:

⎡ ⎤
xm = ⎢0,..., 0,1,...,1,
123 {123 14 2,..., 2,..., m − 1,...,
4244 m −
311, m,...,
424 m ,
3 14m − 1,...,
4244 m −
3 {⎥
1,...,1,...,1, (2.4)
⎣⎢ N N N N N N N ⎦⎥

16
The RSNS sequence can be represented in a basic form as follows [13]:
⎧ ⎢ h − si ⎥
⎪⎢ N ⎥ , si ≤ h ≤ Nmi + si + 1
⎪⎣ ⎦
{ x} = ⎨ (2.5)
⎪ ⎢ 2 Nmi + N − h + si − 1 ⎥ , Nm + s + 2 ≤ h ≤ 2 Nm + s − 1
⎪⎩ ⎢⎣ N ⎥
⎦
i i i i

where h is an integer with a value 0 < h < m and si is the sequence shift value. A

number system is comprised of N row vectors of different pair wise relatively prime
moduli, m1 , m2 ,.., mN . When the row vectors are combined, they are shifted one cell per
row [18].

The dynamic range M̂ of the sequence is defined as the longest sequence of

unique vectors formed in the sequence. The dynamic range is mapped into the FOV of
the antenna and contributes to the angular accuracy of the sequence in tracking within
that FOV. For example, if the dynamic range of a sequence is Mˆ = 30 and the FOV is
180o , the angular resolution using the RSNS sequence will be 180o / 30 = 6o at broadside.
For a two-channel RSNS sequence with moduli m1 and m2 , the dynamic range M̂ can

be calculated for three different combinations of m1 and m2 given as follows [13]:

⎧3(m1 + m2 ) − 6, m1 > 3 and m2 = m1 + 1

⎪
Mˆ = ⎨3(m1 + m2 ) − 7, m1 > 5 and m2 = m1 + 2 (2.6)
⎪4m + 2m − 2, m1 > 5 and m2 = m1 + C (C > 3)
⎩ 1 2

The number of folds within each modulus can be found after the dynamic range
is known. This is given by

Mˆ
ni = , i = 1, 2,..., N . (2.7)
2 Nmi
In order to achieve n number of folds, the spacing between the channels needs to be
adjusted. The required spacing is found by
λ
di = ni (2.8)
2
17
where λ is the wavelength.

When ni folds occur in the visible region, there will be M̂ spatial resolution bins
of equal width in direction cosine space [17]. Threshold voltages have to be computed
for defining the different bins for the RSNS code. There are mi − 1 threshold values for
each channel. The values are found by the following equation [13]:
⎛ 1 ⎞
⎜ mi − j + 2 ⎟
V j ,mi = cos ⎜ π ⎟ , 1 ≤ j ≤ mi . (2.9)
⎜ mi ⎟
⎝ ⎠
The integer value within the RSNS sequence increments when the folding
waveform exceeds a threshold, and decrements when the folding waveform drops below
a threshold. This process is demonstrated for a two-channel RSNS sequence in Figure 9
for moduli m1 = 3 and m2 = 4 .

Figure 9. Folding waveforms mapped into RSNS threshold values (After [11]).

18
 The final step in developing an RSNS system is the computation of a phase
adjustment term ϕi for each channel to shift the center of the dynamic range to the

broadside angle of arrival. After the phase adjustment term ϕi is added, the output

voltage as a function of the zenith angle (θ ) , wavelength ( λ ) and antenna spacing

( d ) for each channel is found as [13]

Vi (θ ) = cos(kd sin(θ ) + ϕi ). (2.10)

2. RSNS Virtual Spacing

The concept of RSNS virtual spacing comes from the idea that an arbitrary
folding waveform can be generated in the processor from an unambiguous measured
phase difference Δφ [18]. The generated folding waveform is considered to arise from
an imaginary spacing rather than a real spacing, therefore leading to the description
virtual. Virtual spacing provides elimination of the limitations due to half-power
beamwidth of the antenna or physical space constraints for achieving a certain channel
spacing. A scale factor C is computed to translate the element spacing to the virtual
spacing for each channel [13]. Using the scale factor C and the real element spacing d
for a single channel, we find the RSNS virtual spacing as
di = Ci d . (2.11)
The channel output voltages can then be found for any spacing using [17]
⎛ Ckd ⎞ ⎛ kd ⎞ ⎛ C Δφ ⎞
Vi (θ ) = cos ⎜ sin(θ ) ⎟ = cos ⎜ i sin(θ ) ⎟ = cos ⎜ ⎟. (2.12)
⎝ 2 ⎠ ⎝ 2 ⎠ ⎝ 2 ⎠
Thus, the output voltage for an angle of arrival can be determined for any spacing di

provided the relationship between the phase difference Δφ and the AOA is
unambiguous. For amplitude combining, unambiguous phase difference occurs if the
physical spacing d is less than one λ .

19
 3. Direction Finding Using RSNS with Moduli [5, 9]

A two-channel RSNS design with moduli [5, 9] and virtual spacing proposed in
[12] was chosen by Tan and Pandya [13] for the implementation of tracking in the digital
tracking array. Using m1 = 5 and m2 = 9 (C = 4) in Equation (2.6), we calculate the

dynamic range of the system as 36. For a scanning range of −90o ≤ θ ≤ 90o , the angular
resolution of one bin at broadside will be 180o / 36 = 5o .

Using Equation (2.7), we calculate the number of folds for each of the two
channels as
36
n1 = = 1.8 (2.13)
( 2 )( 2 )( 5 )
and
36
n2 = = 1. (2.14)
( 2 )( 2 )( 9 )
The operating frequency of the proposed system is 2.4 GHz. This gives a wavelength of
λ = 0.125 m. Using n1 = 1.8 and n2 = 1 and λ = 0.125 in Equation (2.8), we calculate the
virtual channel spacings as d1 = 0.1125 m and d 2 = 0.0625 m , respectively.

The next step is the calculation of the scale factors for each channel. For the
physical spacing of 0.065 m used in the proposed system, the scale factors are
C1 = 1.73 and C2 = 0.96 [13]. The threshold values for the different bins can then be
computed using Equation (2.9). The calculated values for the thresholds are shown in
Table 1.

Table 1. Voltage threshold values for RSNS with moduli [5, 9] (From [13]).

m 0 1 2 3 4 5 6 7 8
j
5 -0.5878 0 0.5878 0.9511 1 - - - -
9 -0.866 -0.6428 -0.3420 0 0.3420 0.6428 0.866 0.866 1

20
 The phase adjustments ϕi for Channel 1 and Channel 2 are found as

ϕ1 = −1.2566 and ϕ2 = −1.5708 radians, respectively. These are used to align the folding
waveforms to the center of the dynamic range.

The voltage of the incoming wave can be found from

⎛ C Δφ ⎞
Vi (θ ) = cos ⎜ i + ϕi ⎟ , i = 1, 2 (2.15)
⎝ 2 ⎠
where Δφ is the measured phase difference mentioned earlier, Ci is the scale factor and

ϕi is the phase adjustment. For the system used in this study, the phase difference Δφ
can be found from the in-phase ( I ) and quadrature (Q) data, which is obtained from the
quadrature demodulator boards.

Finally, the voltages computed from Equation (2.15) are compared with the
thresholds that are given in a look-up table (i.e., Table 1) and the AOA can be found.

C. DIGITAL BEAMFORMING AND PHASE INTERFEROMETRY

For a simple linear antenna array, the wavefronts of the plane waves that arrive at
each of the array elements may not be parallel to the array axis. Therefore, there will be a
phase difference that occurs between each array element due to the time for the
wavefront to travel between two elements. This situation is shown in Figure 10.

The phase difference Δφ between the elements depends on the AOA ( θ ),

frequency and the element spacing ( d ) and can be calculated using
2π
Δφ = d cos(θ ). (2.16)
λ

21
 Figure 10. Digital array beamforming on reception (From [19]).

Also shown in Figure 10 is the process of the beam-forming by the application of

the weights calculated for each element and summation by the processor. The complex
weight added to each array element is given by wn = An e jα n . From the addition of the
weights, the sum beam array factor can be expressed by
N N
AFΣ (θ , φ ) = ∑ wn e jkdn cos(θ ) = ∑ An e jα n e jkdn cos(θ ) (2.17)
n =1 n =1

where N is the number of elements, d n is the location of element n relative to the center

of the array and k = 2π λ . The spacing d n is calculated from

(2n − ( N + 1))
dn = d. (2.18)
2

22
 In DBF, the beam is formed from the I and Q signals, which are obtained from
the down-conversion and sampling of the RF signals received by the array elements.
Quadrature demodulators are used to obtain I and Q signals during reception. The
output of the quadrature demodulators can be represented by [19]
I n = cos ( kd n cos(θ ) ) (2.19)
and
Qn = sin ( kd n cos(θ ) ) . (2.20)

Therefore, to obtain a response equivalent to a beam scanned in the direction θ s , the

weights must be given by
wn = e − jkdn cos(θs ) . (2.21)
Thus, the sum and difference beams are obtained using
N N
Σ ≡ AFΣ (θ , φ ) = ∑ wn e jkdn cos(θ ) = ∑ An e jα n e jkdn cos(θ ) (2.22)
n =1 n =1
and
N /2 N
Δ ≡ AFΔ (θ , φ ) = ∑ An e jα n e jkdn cos(θ ) −∑ An e jα n e jkdn cos(θ ) . (2.23)
n =1 N /2

Using the sum and difference beams given by Equations (2.22) and (2.23), we
obtain the plot given in Figure 8, and the phase error of the direction to the target can be
evaluated for the linear region using Equation (2.2).

D. SUMMARY

In this chapter, target tracking types and techniques were discussed. Monopulse
tracking and RSNS for direction finding were introduced since they are used in the
proposed system. Finally, digital beamforming for phased array antennas was described.

In Chapter III, the direct conversion receiver architecture will be discussed.

23
THIS PAGE INTENTIONALLY LEFT BLANK

24
 III. DIRECT CONVERSION RECEIVER ARCHITECTURE

The reception process for the FM modulated video signals involves signal
processing methods for the proper demodulation of the video data. The receiver
architecture used in this research employs direct conversion techniques in which
quadrature detection and frequency down-conversion takes place. In this section, the
direct conversion receiver architecture is described and problems associated with this
architecture are discussed.

A. QUADRATURE DETECTION THEORY

Quadrature detection is a common technique used for the demodulation of RF

signals. The signal is demodulated to form the in-phase ( I ) and quadrature (Q)
components of the input signal after frequency down-conversion.

In a quadrature detector, the input signal is divided into two equivalent channels
as seen in Figure 11. In one channel, the signal is mixed with a sinusoidal local oscillator
signal. If the local oscillator (LO) frequency is given as ωLO = 2π f LO and the input RF

signal frequency as ωIN = 2π f IN , the mixing process can be described mathematically as

Vout = cos(ωIN t ) cos(ωLO t )

1 (3.1)
= ⎡ cos ( (ωIN − ωLO )t ) + cos ( (ωIN + ωLO )t ) ⎤⎦
2⎣
where Vout is the output voltage obtained after the mixing process. The resulting signal is

passed through a low-pass filter (LPF) and the in-phase ( I ) component is obtained. The
reason for the LPF process is to eliminate one of the terms that is created after the mixing
process, usually the sum term on the right hand side of Equation (3.1). In the other
channel, the input signal is mixed with a 90o phase-shifted version of the local oscillator
signal. After low-pass filtering of the resulting signal, the quadrature (Q) component is
obtained [20].

25
 Figure 11. Quadrature detection process (After [20]).

For the current design, an LO signal equal to the input carrier frequency
(ωLO = ωIN = ω0 ) is used. In the mixing process, the I and Q signals are regarded as
being the real and imaginary components of a complex valued baseband signal. They can
be used to observe any changes in the magnitude and phase of the input signal originally
modulated onto a quadrature carrier and transmitted. This feature provides an advantage
for the phase information since the direct recovery of the phase of an RF carrier sine
wave is a difficult process and requires expensive hardware circuits. It is much simpler in
hardware to induce a 90o phase shift.

The mathematical derivation of the mixing process for I and Q signals is done as
follows. The in-phase baseband component can be described by [13]
j ( 2π f 0t +θ ( t ) )
X i (t ) = Re ⎡ Ac e ⎤ cos(2π f 0t ) (3.2)
⎣ ⎦
where θ (t ) is the frequency modulated input signal. Considering a complex quantity α ,
we find the real and imaginary parts of α from

26
 α +α∗
Re [α ] = (3.3)
2
and
α −α*
Im [α ] = . (3.4)
2
It is convenient to use the Euler’s identities for the cosine term in Equation (3.2)
since it is represented with exponentials. Euler’s identities are given as
e j 2π f 0 t + e − j 2π f 0 t
cos ( 2π f 0t ) = (3.5)
2
and
e − j 2π f 0 t − e j 2π f 0 t
sin ( 2π f 0t ) = j . (3.6)
2
Here e j ( ) represents the positive frequency component, and e− j ( ) represents the
negative frequency component.

Using Equation (3.3) and (3.5), we can rewrite Equation (3.2) as

X i (t ) =
( )
Ac e j (2π f0t +θ (t )) + e− j (2π f0t +θ ( t )) ⎛ e j 2π f0t + e − j 2π f0t ⎞
⎜ ⎟
2 ⎝ 2 ⎠
A
( )
= c e j (2π 2 f0t +θ ( t )) + e jθ (t ) + e − jθ ( t ) + e− j (2π 2 f0t +θ ( t ))
4
(3.7)

A
(
= c Re ⎡⎣e j (2π 2 f0t +θ (t )) ⎤⎦ + Re ⎡⎣e jθ ( t ) ⎤⎦ .
2
)
The leftmost component on the right-hand side of Equation (3.7) is the
high-frequency component. The rightmost component on the right-hand side is the
desired message signal. Hence, low-pass filtering the signal given in Equation (3.7)
provides the desired message component. After low-pass filtering, the following
expression is obtained:
Ac
I= Re ⎡⎣e jθ ( t ) ⎤⎦ . (3.8)
2
In Figure 12, a frequency domain representation of the process described above is
given. The real and imaginary components of the input signal are shown in the upper
section, while the real and imaginary components obtained after the mixing process are
27
shown in the lower section of the figure. The single arrow lines show the multiplication
with the positive frequency component, and the double dot-dashed lines show the
multiplication with the negative frequency component [13]. The shaded area represents
the low-pass filtered portion of the signal.

Figure 12. Frequency spectrum for the mixing and LPF process
for the I channel (From [13]).

The derivation of the quadrature component is similar to the process for the in-
phase component. The quadrature baseband component can be expressed by
j ( 2π f 0t +θ ( t ) )
X q (t ) = Im ⎡ Ac e ⎤ sin(2π f 0t ). (3.9)
⎣ ⎦
Using Equations (3.4) and (3.6), we can rewrite Equation (3.9) as

X q (t ) =
( )
Ac e j (2π f0t +θ ( t )) − e − j (2π f0t +θ (t )) ⎛ e − j 2π f0t − e j 2π f0t ⎞
⎜j ⎟
2 ⎝ 2 ⎠
A
( )
= j c e jθ ( t ) − e j (2π 2 f0t +θ ( t )) + e − jθ ( t ) − e− j (2π 2 f0t +θ ( t ))
4
(3.10)

= j
Ac
4
(
Im ⎡⎣e j (2π 2 f0t +θ (t )) ⎤⎦ − Im ⎡⎣e jθ ( t ) ⎤⎦ . )
28
 Similar to the previous situation, the left-most component inside of the
parenthesis in Equation (3.10) is the high-frequency component. The component on the
right side is the desired message signal. Hence, low-pass filtering the signal given in
Equation (3.10) provides the desired quadrature message component. After low-pass
filtering, the following expression is obtained:
Ac
Q= j Im ⎡⎣e jθ (t ) ⎤⎦ . (3.11)
2
In Figure 13, the frequency domain representation of this process is given. Again,
the real and imaginary components of the input signal are shown in the upper section,
while the real and imaginary components obtained after the mixing process are shown in
the lower section of the figure. The single arrow lines show the multiplication with the
positive frequency component, and the double dot-dashed lines show the multiplication
with the negative frequency component [13]. The shaded area represents the low-pass
filtered portion of the signal.

Figure 13. Frequency spectrum for the mixing and LPF process
for the Q channel (From [13]).

29
 One important point to notice in Figures 12 and 13 is that the filtered component
is the same for both I and Q channels. Hence, the amplitude of the summation of the
I and Q signals is two times that of the individual components. The final output signal
obtained from the summation of the I / Q signals after the down-conversion process is
shown in Figure 14.

Figure 14. Frequency domain representation for the filtered

output signal (From [13]).

B. DIRECT CONVERSION ADVANTAGES

Compared with other common architectures, such as the super-heterodyne

architecture, direct conversion provides some advantages. In the super-heterodyne
architecture, the carrier frequency is converted to an intermediate frequency before the
demodulation process. This means two or more stages of detection and filtering are
required. However, for the direct conversion receiver architecture, the carrier frequency
is directly down-converted to baseband, where the signal can be detected and
demodulated. This process requires only a single stage, which means a lower complexity
and power consumption. The elimination of the intermediate frequency stages provides
the ability to build these architectures with less circuitry and, hence, less cost.

Another advantage obtained from the elimination of the intermediate frequency

stages is that the problem of image rejection can be solved. There are two sidebands in
an RF carrier signal. These sidebands contain different information about the phase and

30
the frequency of the modulated signals and they need to be separated from each other. If
these signals overlap each other, irreversible corruption occurs and the phase and
frequency information are lost. Also, it is very expensive and difficult to design and
build circuits to directly recover the phase of an RF carrier wave [13]. The direct
conversion architecture eliminates these problems with the features described above and
provides cost efficiency.

C. DIRECT CONVERSION PROBLEMS

The direct conversion architecture has some problems and limitations due to
several reasons, including couplings and mismatches in the circuitry used. In this section,
four main problems associated with the direct conversion receiver architecture are
discussed.

The quadrature detection process was illustrated in Figure 11. During this
process, a finite amount of feedthrough exists from the local oscillator to the inputs of
the mixer. This effect is also called “LO leakage” or “LO feedthrough” and arises from
parasitic coupling of the LO and RF signals [21]. This coupling produces a dc
component in the mixing process which cannot be filtered out since the RF and LO
signals are in band. This problem can be solved by applying proper isolation between the
RF and LO signal paths on a circuit board to prevent the leakage. Another approach is to
apply a certain amount of shielding between these paths. The dc offsets may also arise
from the mismatches in the I / Q demodulator. Since in a direct conversion architecture,
the band of the down-converted signal extends to zero frequency, extraneous offset
voltages can corrupt the signal and saturate the following stage, therefore, causing
significant problems [21]. This problem can be resolved by applying a calibration routine
to find the offsets. After the offsets are found, necessary adjustments can be made to
compensate for the dc offsets. The calibration of the quadrature demodulators will be
discussed in more detail in Chapter V.

31
 Another problem for the direct conversion architecture is the mismatch between
the amplitudes of the I and Q signals. Referring back to Figure 11, we recall that it was

mentioned that the LO oscillator is shifted by 90o for the generation of the quadrature
signal. The errors that occur in this phase shift and additional mismatches between the
amplitudes and phases of the I and Q channels corrupt the down-converted signal and
cause errors in the information being sent. Inspection of the complex envelope
representations for the I and Q components reveals that any mismatches will create
differences in the amplitudes for the in-phase and quadrature signals. The phase
mismatches occur due to unequal path lengths in the two channels [13]. The problem of
I and Q mismatch can be eliminated by using higher levels of integration [21].

The third problem for the direct conversion architecture arises from the
nonlinearities experienced by the strong interferers in a low-noise amplifier (LNA). The
mixing process of two signals was described in Equation (3.1). For direct conversion
after the mixing process, the low-frequency term given on the left-hand side of Equation
(3.1) is produced. This term indicates that two high-frequency interferers generate a low-
frequency beat in the presence of second-order distortion. Another source of the second-
order nonlinearity is the second harmonic of the desired RF signal, which is down-
converted to the baseband if mixed with the second harmonic of the LO output. This may
occur for higher harmonics as well, but it is negligible in differential mixers since the
magnitude of the harmonics for both the RF signal and the LO signal is inversely
proportional to the frequency. Second-order nonlinearities can produce unwanted
components which fall within the desired band of frequencies. This situation is illustrated
in Figure 15. Second-order nonlinearity can be characterized using the “second intercept
point,” IP2 . To obtain IP2 , two equal amplitude interferers are applied at the input and

their low-frequency beat signal is observed at the output. Then, IP2 is obtained by the
extrapolation of the results after plotting the beat signal power versus the input power
[21].
32
 Figure 15. Down-conversion of signal harmonics (From [21]).

The last problem that will be discussed is the potential for flicker noise.
Considering a gain of roughly 30 decibels (dB) in the LNA/mixer combination in a direct
conversion architecture, the down-converted signal amplitude has a range on the order of
tens of microvolts. This situation makes the input noise of the following stages
(e.g., filters) critical. The 1/ f noise of devices has a profound effect on the signal
because of the down-converted signal spectrum being located around zero frequency.
The magnitude of flicker noise can be reduced by the incorporation of very large devices
(on the order of thousands of microns) by the stages following the mixer since they
operate at relatively low frequencies.

D. SUMMARY

In this chapter, the direct conversion receiver architecture was discussed. The
quadrature detection theory was introduced, and the I / Q signal down-conversion
process was described. Finally, the advantages and associated problems of the direct
conversion architecture were discussed.

In Chapter IV, the video signal formats will be described, followed by the
modulation and demodulation techniques used for the transmission and reception of
video data.

33
THIS PAGE INTENTIONALLY LEFT BLANK

34
 IV. UAV VIDEO SIGNAL PROCESSING

When the acquisition and tracking of the video signal coming from a UAV are
achieved, the next step is to retrieve the video data from the incoming signal. In order to
retrieve the video data successfully, the structure of the incoming signal must be known.
The video data, which is captured by the UAV’s onboard camera, is first encoded using a
particular analog video format. This encoded data is then modulated on a carrier
frequency and transmitted to the GCS. After the signal is received successfully by the
GCS, the signal is processed in a backwards order compared to the process before the
transmission. First the signal is demodulated using an appropriate method and then the
demodulated signal is decoded to retrieve the captured video data. In this chapter, the
structure and formats of video signals are described in detail and the modulation, and
demodulation techniques for video signals are summarized.

A. VIDEO SIGNALS

Video signals are essential parts of most applications in the world today. They are
used commonly in various fields and in both military and civilian applications. Military
applications include surveillance, communications and navigation. Some common
civilian applications are TV broadcasting, security surveillance and gaming. Video
signals can be classified into two main categories: analog and digital. Analog video is
continuous in form while digital video takes on the form of several discrete values that
represent the information content of the original data. Analog video is an older
technology compared to digital. Digital video offers advantages in compatibility with
digital equipment and signal processing methods. To date, most of the video signals
transmitted through free space have been analog video signals. The most commonly used
analog video format in the United States is the NTSC format. In addition to NTSC, there
are the Phase Alternating Line (PAL) and Systeme Electronic Pour Avec Memoire
(SECAM) formats. These formats differ in image size, line rate, frame rate and vertical

35
line resolution. The video data transmission to a GCS from a UAV, as considered in this
research, is in NTSC analog video format as well. For this reason, NTSC analog video
signals were examined in more detail.

1. NTSC Video Signals

Standard NTSC video signal is designed to be displayed on a television screen.

For this process, the incoming video signal needs to be converted to the individual pixel
values of the display. This conversion is done by a scanning scheme which is called a
left-to-right and top-to-bottom scan [22]. This scheme is demonstrated in Figure 16.

Figure 16. Left-to-right, top-to-bottom video scanning scheme (After [22]).

In most systems today, a method called interlaced scanning is used in order to

prevent the human eye from perceiving a flicker as the screen is updated at a rate of 30
frames per second. Here, the image frame is divided into two fields, one containing the
odd-numbered lines and the other containing the even-numbered lines. Odd-numbered
lines are scanned before the even-numbered lines, and the display is updated at a rate of

36
60 fields per second. The human eye cannot detect a flicker at this update rate, which is
the desired result. The interlaced scanning process is shown in Figure 17.

Figure 17. Interlaced scanning process (From [23]).

The structure of a single line of an NTSC analog video signal is given in Figure
18. The components of the signal are back porch, front porch, active pixel region and
horizontal sync. Back porch is used as a reference level to remove any dc components
from the video signal. In the active region, the intensity information for each of the lines
is represented by a low-voltage signal that is synchronized with the display device. This
signal is also called the luma (luminance) signal and provides a reference for the
brightness of the image. Another signal embedded within the active region of a single
video signal is the chroma signal, which contains the color information of the image. The
phase and amplitude reference for the color is provided by a color burst located in the
video signal.
37
 Figure 18. NTSC analog video signal for a single horizontal line (From [22]).

There are horizontal synchronization (sync) signals at the beginning of each new
video line, which indicate the next stream of pixels forming the next line. In the left-to-
right top-to-bottom scanning scheme mentioned earlier, this signal dictates when the
image should be scanned from left to right. The top-to-bottom scanning is dictated by the
vertical sync signal. The utilization of horizontal and vertical sync signals are
demonstrated in Figure 19.

As seen from Figure 19, horizontal sync signals are located between every line in
a field, while the vertical sync signals are located only between the fields. Also noticed is
that more than one pulse is used in vertical sync to indicate the start of a new field.

38
 Figure 19. Utilization of horizontal and vertical sync signals
in a video signal (From [22]).

2. Video Frame Format

In NTSC analog video format , there are 525 lines per frame. Of these, 485 lines
are used for display and the remaining 40 lines are used for vertical sync signals. The
485 display lines are separated into two fields with 263 lines in the first and 262 lines in
the other. There are 20 vertical sync pulses at the beginning of each field, therefore
making a total of 40 pulses. The line frequency for a color NTSC signal is 15,734
kilohertz (kHz). Therefore, the line duration is calculated as 63.556 microseconds (µsec).
The active horizontal signal duration for the same signal is 52.66 µsec. For a better
understanding of these time intervals, detailed timing of a video line is given in Figure
20.

39
 Figure 20. Timing of an NTSC video line (From [13])

Also seen in Figure 20 are the video levels and ranges for the different parts of
the video signal. The unit used to define video levels is the IRE (Institute of Radio
Engineers). The blanking level refers to 0 IRE and the white level refers to +100 IRE.
The blanking level is the reference level for the video signal but it differs from the black
level when a setup is applied to the signal. For NTSC, a setup of 7.5 IRE is usually
applied. This moves the black level to +7.5 IRE.

40
 3. Digital Video Signals

Digital video signals are not used as commonly as analog video signals.
However, they are chosen for applications which require high spatial resolution, high-
intensity resolution, high speed and noise immunity. Because the signal is already
digitized, the timing of the digital video is easier than the analog video. The timing
signals for digital video are pixel clock, line enable and frame enable. These are shown
briefly in Figure 21.

Figure 21. Timing signals for digital video (From [22]).

The pixel clock sets the timing of the data transfer. The line enable signal
indicates the beginning and end of each video data line. Finally, the frame enable signal
indicates the start and completion of each frame [22].

41
B. VIDEO SIGNAL MODULATION AND DEMODULATION

There are three properties of an analog signal that can be changed by a message
signal. These are frequency, amplitude and phase. Changing these properties of an
analog signal using a message signal is called modulation. When the frequency,
amplitude and phase of an analog signal are changed, the modulation is called frequency
modulation (FM), phase modulation (PM) and amplitude modulation (AM), respectively.
Frequency and phase modulation are also called angle modulation. Angle modulation has
some advantages over amplitude modulation but there are also disadvantages.
Advantages include noise reduction, improved system fidelity and more efficient use of
power, while disadvantages include a requirement for a wider bandwidth and utilization
of complex circuits in transmitters and receivers. Today, angle modulation is used
extensively for commercial radio broadcasting, television sound transmission, two-way
mobile radio, cellular radio, and microwave and satellite communication systems [24].

The video data transmission to the GCS from the UAV also utilizes FM. The
video data captured by the onboard camera is first encoded and then FM modulated on a
carrier before being transmitted to the GCS. When the transmitted signal is received by
the GCS, it is down-converted to baseband and demodulated. After the demodulation, the
resulting signal is decoded and the video is obtained. For the purpose of this thesis,
emphasis will be placed on FM and its associated demodulation procedures.

1. Frequency Modulation (FM)

As mentioned earlier, frequency modulation is a form of angle modulation. In

angle modulation, the instantaneous phase angle of a sinusoidal wave is varied with
respect to time. Mathematically, an angle modulated wave can be expressed as [24]
x(t ) = Vc cos[ωc t + θ (t )] (4.1)

where x(t ) = angle modulated waveform, Vc = peak carrier amplitude, ωc = 2π f c =

carrier radian frequency, and θ (t ) = instantaneous phase deviation (radians).

42
 In order to achieve angle modulation, it is required that θ (t ) be a function of the
modulating signal amplitude. In our particular application, the NTSC video signal is the
modulating signal. However, to illustrate the FM characteristics, we often consider the
modulating signal as a sinusoid with amplitude Vm and frequency ωm . It can be
represented as [24]
sm (t ) = Vm sin(ωmt ). (4.2)
FM and PM both occur at the same time and are related to each other. That is,
when FM occurs PM occurs as well. The difference between FM and PM lies in which
property of the carrier is directly varied by the modulating signal. If the frequency of the
carrier signal is varied directly, it is called FM. If the phase of the carrier signal is varied
directly, it is called PM [24].

In Figure 22, an FM modulated signal is shown in the frequency domain. The

magnitude and direction of the maximum frequency shift Δf which is also known as the
peak frequency deviation, is determined by the amplitude and polarity of the modulating
signal Vm which is shown in the lower section of Figure 22. As shown, the frequency

can vary about the carrier continuously assuming any value between + / − Δf . The rate
of change of the carrier frequency is equal to the frequency of the modulating signal. The
instantaneous frequency deviation is found by taking the first derivative of the
instantaneous phase deviation θ (t ) as
Δf = θ ′(t ) rad/s (4.3)
where the prime (′) is used to represent the first derivative with respect to time.

43
 Figure 22. Angle modulated waveform in the frequency domain (After [24]).

The frequency deviation Δf is proportional to the amplitude of the modulating

signal. Therefore, it can be represented as
θ ′(t ) = Ksm (t ) rad/s (4.4)
where K is a constant, which is called the deviation sensitivity of the frequency
modulator. The unit used for K is Hz/volt.

The ratio of the frequency deviation to the frequency of the message signal is
known as the modulation index. Modulation index is a unitless quantity and indicates the
depth of modulation achieved. Using Equation (4.4) for the frequency deviation, we find
the modulating index from
θ ′(t ) KVm Δf
m= = = (unitless) (4.5)
fm fm fm
where m = modulation index (unitless), K = deviation sensitivity (Hz/volt), Vm = peak

modulating signal amplitude, and f m = message signal frequency.

44
 Finally, using the relationships above, we can rewrite Equation (4.1) for an FM
signal as [24]
KVm
x(t ) = Vc cos[ωc t + sin(ωmt )]
fm
Δf
x(t ) = Vc cos[ωc t + sin(ωmt )] (4.6)
fm
x(t ) = Vc cos[ωc t + m sin(ωmt )].

2. FM Demodulation Techniques

With FM, the information is impressed onto the carrier in the form of frequency
variations. On the receiver side, this FM signal is demodulated using one of several FM
demodulation methods. FM demodulators produce an output voltage, which is
proportional to the instantaneous frequency at their outputs [24]. Of interest in this thesis
is the demodulation of FM modulated signals received from a UAV, with video
information encoded using the NTSC format prior to modulation.

Hui [25] studied three practical implementations of the FM demodulators. These

are: (1) digital phase locked loop (PLL) quadrature detector, (2) local oscillator with no
feedback and (3) Billotti’s quadrature demodulator.

In a PLL, a phase comparator is used to generate an error voltage signal

proportional to the difference between the voltage-controlled oscillator (VCO) and the
signal at the input. This error voltage is fed back to the input of the VCO. This process is
repeated continuously so that the VCO can keep track of the frequency changes in the
input signal. The demodulated information signal can be obtained from the output of the
phase comparator since it will be proportional to the frequency deviation.

The local oscillator with no feedback has a similar structure but it lacks the
feedback to the VCO. It is assumed that the FM signal can be demodulated if the local
oscillator frequency and the input signal frequency are sufficiently close.

45
 Bilotti’s quadrature demodulator applies a 90o phase shift to the FM signal. A
tuned circuit converts frequency variations to phase variations. The phase shift that
occurs is greater or smaller than 90o depending on the direction of the frequency
deviation. The output of the phase detector is applied to the input of a PM detector.

In [25], the three methods mentioned above were compared using simulations
that utilized digital beamforming. It was concluded that local oscillator with no feedback
showed similar performance as the other two detectors with less circuitry. Furthermore,
it was found to be tolerant to the expected frequency offset between the FM transmission
source and the LO. For this reason, this detector was selected for the FM demodulation
process in this thesis.

The demodulation of FM modulated video signals is performed in software by

the antenna processor, i.e., by digital signal processing (DSP). Three types of
demodulation schemes were studied. They are: (1) baseband demodulation, (2) tangent-
type demodulation and (3) demodulation using a Kalman filter. These demodulation
schemes are described in the following sections.

a. Baseband Demodulation

The array architecture that is employed in this research uses in-phase and
quadrature components for single-stage down-conversion from RF to baseband.
Assuming x(t ) is the received signal, we can obtain the in-phase and quadrature
components as
x(t ) = Vc cos [ωc t + θ (t ) ]
= Vc cos [ 2π f c t + 2πΔft + θ (t )] (4.7)
= Vc cos [ 2πΔft + θ (t ) ] cos(2π f c t ) − Vc sin [ 2πΔft + θ (t ) ] sin(2π f c t )

46
where
I (t ) = cos [ 2πΔft + θ (t )] (4.8)
Q(t ) = sin [ 2πΔft + θ (t ) ] (4.9)
t
and θ (t ) = K ∫s
−∞
m (t )dt .

Since the change in the frequency leads to a change in phase of the signal,
the original signal can be retrieved by taking the first derivative of the I (t ) and Q(t ) :

dI (t ) ⎡ t
⎤
= − sin ⎢ 2πΔft + K ∫ sm (t )dt ⎥ ( 2πΔf + Ksm (t ) )
dt ⎣ −∞ ⎦
(4.10)
dQ(t ) ⎡ t
⎤
= cos ⎢ 2πΔft + K ∫ sm (t )dt ⎥ ( 2πΔf + Ksm (t ) ) .
dt ⎣ −∞ ⎦
Finally, the message signal sm (t ) is retrieved by the expression I (t )Q′(t ) − Q(t ) I ′(t ) . This
process is shown with a block diagram in Figure 23.

Figure 23. Baseband Demodulation with I and Q signals.

Equation (4.10) can be obtained from a mathematical evaluation of the

block diagram given in Figure 23:

47
 dQ(t ) dI (t ) ⎡ t
⎤
I −Q = cos 2 ⎢ 2πΔft + K ∫ sm (t )dt ⎥ ( 2πΔf + Ksm (t ) )
dt dt ⎣ −∞ ⎦
⎡ t
⎤
+ sin 2 ⎢ 2πΔft + K ∫ sm (t ) dt ⎥ ( 2πΔf + Ksm (t ) )
⎣ −∞ ⎦
= ( 2πΔf + Ksm (t ) ) (4.11)
⎧⎪ 2 ⎡ t
⎤ ⎡ t
⎤ ⎫⎪
⎨cos ⎢ 2πΔft + K ∫ sm (t )dt ⎥ + sin ⎢ 2πΔft + K ∫ sm (t )dt ⎥ ⎬
2

⎩⎪14444444444
⎣ −∞
4 ⎦244444444444
⎣ −∞
3⎦ ⎭⎪
1

= 2πΔf + Ksm (t ).
The first component 2πΔf in the final expression is the dc offset, which occurs as a result
of the difference between the carrier frequency and the local oscillator frequency used
during the down-conversion of the signal to baseband.

b. Tangent-Type Demodulation

The tangent-type FM demodulator using DSP is shown in Figure 24. In

this figure, S1 (k ) is the detected output of a conventional tangent-type FM demodulator.
The circuit components that are used to construct this conventional tangent-type
demodulator are a 90o phase shifter, analog-to-digital (A/D) converter, memory element,
delay element, adder and subtractor [26]. The block that is shown inside the dashed lines
is used for improvement of the detection range and will be described in detail in later
paragraphs.

48
 Figure 24. Tangent-type FM demodulator using DSP (After [26]).

The input analog FM signal x(t ) and noise can be expressed as

x(t ) = 2 P cos [ 2π f c t + θ (t ) ] + n(t ) (4.12)

where θ (t ) = m sin [ 2π f mt ] , f m = message frequency, P = signal power, and n(t ) =

Gaussian noise. Here it is assumed that the sine wave used to represent θ (t ) is actually
the integral of a cosine function, which is the message signal described in Equation (4.2).

The signal y (t ) is obtained from passing x(t ) through the phase shifter.
Hence ,
y (t ) = 2 P sin [ 2π f c t + θ (t ) ] + n′(t ) (4.13)

where n′(t ) is the noise phase-shifted by 90o. After passing through the A/D converters
(i.e., sampling and quantization), the signals take the form
X (k ) = 2 P cos [ 2π f c kT + θ (k ) ] + n(k ) (4.14)
and
Y (k ) = 2 P sin [ 2π f c kT + θ (k ) ] + n′(k ). (4.15)

Here t (k ) = kT = k f s , θ (k ) = θ {t (k )} , n(k ) = n {t (k )} , n′(k ) = n′ {t (k )} , T is the period

of the signal and f s is the sampling frequency.

49
 The output of the phase detector is given by
⎡ Y (k ) ⎤
φ (k ) = tan −1 ⎢ ⎥. (4.16)
⎣ X (k ) ⎦
If we assume the noise n(k ) = 0 , this expression reduces to

⎡ Y (k ) ⎤ ⎡ 2 P sin [ 2π f c t + θ (t ) ] ⎤
φ (k ) = tan −1 ⎢ ⎥ = tan −1
⎢ ⎥
⎣ X (k ) ⎦ ⎢⎣ 2 P cos [ 2π f c t + θ (t ) ] ⎥⎦
= tan −1 ( tan [ 2π f c t + θ (t )]) (4.17)
= 2π f c t + θ (t ).
Including the noise in the signal, we can write this expression as
⎡ Y (k ) ⎤
φ (k ) = tan −1 ⎢ ⎥ = mod [ 2π f c kT + θ (k ) + η (k ) ] (4.18)
⎣ X (k ) ⎦
where η (k ) is the phase dispersion caused by the input noise, and mod [ ] expresses the

remainder when the argument is divided by 2π [26].

Finally, the demodulated output of the tangent-type demodulator S1 (k )

becomes
Δφ
S1 (k ) = = φ (k ) − φ (k − 1). (4.19)
Δt
The circuit that is shown between dashed lines in Figure 24 is used to
widen the detection range of the tangent-type FM demodulator. In this process, the
demodulated signal S1 (k ) is compensated with C1 (k ) , which is given as

⎧0, S1 (k ) ≥ 0 ⎫
C1 (k ) = ⎨ ⎬. (4.20)
⎩2π , S1 (k ) < 0 ⎭
Hence, the output S (k ) becomes
S (k ) = S1 (k ) + C1 (k )
= φ (k ) − φ (k − 1) + C1 (k ) (4.21)
= mod [ 2π f cT ] + Δθ (k ) + Δη (k )

50
where Δη (k ) = η (k ) − η (k − 1) . It is noticed that the first term obtained is a dc term as
was the case for baseband demodulation in the previous section. The second term is the
demodulated signal [26]. If the detection condition is taken as η (k ) = 0 , then the
following condition is satisfied:
0 ≤ mod [ 2π f cT ] + Δθ (k ) < 2π . (4.22)

Thus, the improved detection range becomes [ 0, 2π ] .

c. Demodulation Using a Kalman Filter

In [27] it is discussed that the instantaneous phase and frequency of the I

and Q signals can be estimated by using a Kalman filter. State space representation is
used for the estimation of the state variables. The current state is calculated as a weighted
mean of the last estimated state and the current state. This means that the estimation
depends on the uncertainty of the previous state and the measurement noise. For
example, when a noisy measurement is done for the current state, it is expected that the
influence of the previous estimated state will be higher than the measurements. For the
demodulation of I and Q baseband signals, the state variables are the phase φk and the

angular frequency ωk .

In [27] it is concluded that the Kalman filter shows a significant

improvement in noise performance. However, since the Kalman filter approach involves
complex computations such as calculation of state vector, state uncertainty vector and
covariance matrix, it is expected that it will perform slower compared to the other
methods. For this reason, demodulation of FM modulated baseband signals with Kalman
filtering was not implemented in the digital tracking array.

51
C. SUMMARY

In this chapter, the properties of analog and digital video signals were described
and analog video formats were introduced. The modulation and demodulation techniques
for baseband I and Q signals were discussed.

In the next chapter, the hardware and software components that are used in the
digital tracking array system will be discussed.

52
 V. ANTENNA SYSTEM STRUCTURE AND COMPONENTS

In this chapter, the general structure of the antenna system is given and the
hardware and software components of the system are described in detail. The
performance analyses of the commercial components are also provided.

A. SYSTEM STRUCTURE

The general block diagram of the system is given in Figure 25. As illustrated in
Figure 25, the system is a combination of hardware and software components. The
hardware components consist of the antenna, down-conversion block, A/D converter and
the LO. The receiver architecture used is the direct conversion architecture discussed in
Chapter III. The processor was implemented in the software and includes software
modules for tracking, FM demodulation and decoding of the video signals. In the
following sections, the functions and properties of each hardware and software
component are discussed in detail.

Figure 25. System generic block diagram.

53
B. HARDWARE COMPONENTS

The main function of the hardware portion of the system shown in Figure 25 is to
receive and down-convert the analog waveform coming from the UAV and provide it to
the processor as a digital signal after quantization. The incoming signal is demodulated
by mixing it with a local oscillator signal that has the same frequency as the carrier
signal. The mixing process provides the outputs, which were described in Equation (3.1).
The high-frequency component is rejected by the low-pass filters and the low-frequency
component provides the baseband envelope of the desired signal. This desired signal is
then provided to the A/D to be digitized. After digitization, the resulting signal is
provided to the processor block for additional signal processing, which will be discussed
in later sections. The descriptions of the hardware components are provided in the
following sub-sections.

1. Antenna

The array consists of six two-element subarrays over a ground plane. An image
of the array is shown in Figure 26.

Figure 26. Image of the phased array antenna.

54
 Each subarray was fabricated from a Rogers RO4003C substrate and consists of
two half-wave dipole antennas stacked in the elevation plane for increased directivity.
The two-element structure provides a higher signal level at the input of the demodulator,
therefore, allowing longer range operation. The dipole arms are fed by power splitters, as
shown in Figure 27.

Figure 27. PCB layout of the two element subarray (After [13])

In order to keep the current on the two arms of each dipole in phase with each
other, a 180o phase shift is applied to one of the arms of the splitter. This is done by
making one arm λ / 2 longer than the other. The width W of the microstrip lines is an
important parameter in terms of the operation of the dipoles. In [13], a MATLAB
program is provided for the calculation of the widths used in the printed circuit board
(PCB) of each subarray. The widths used in the subarrays are given in Table 2, along
with some other microstrip circuit parameters.
55
Table 2. Properties of the PCB layout of the two-element sub arrays (From [13]).

Dielectric constant (ε r ) 3.38

Thickness (d ) 60 mils (1.52 mm)

Trace width of the 180o splitter (W1 ) 35.7 mils (0.91 mm)

Line width of the rat-race hybrid feed lines (W2 ) 139 mils (3.53 mm)

Line width of the rat-race hybrid (W3 ) 76.3 mils (1.94 mm)

The approximate gain of a single subarray can be calculated by multiplying the

gain of a single dipole by a factor of 2 × 2 = 4, to include the fact that there are two
dipoles and the contribution from the ground plane image. Hence, the gain of each
subarray is found to be
Gsubarray = Gdipole × 2 × 2 = 6.56 = 8.17 dB (5.1)

where Gdipole = 1.64 = 2.15 dB [28]. The gain of the antenna subarrays was measured by

Tan and Pandya [13] in an anechoic chamber by using a Narda 645 standard gain horn
antenna as the reference antenna. The gain of the two-element subarray was estimated to
be around 10 dB, which is sufficiently close to the calculated value in Equation (5.1).
Tan and Pandya also provide simulations of the pattern of a subarray in CST Microwave
Studio.

The gain of the whole array can be obtained by multiplying the gain of a single
subarray by the total number of subarrays, which is six. Hence, using the result obtained
in (5.1), the gain of the array is found as
Garray = Gsubarray × 6 = 6.56 × 6 = 39.36 = 15.95 dB. (5.2)

56
 The gain pattern of the six-element array is shown in Figure 28. It is observed
that the measured gain is two decibels lower than the calculated value. This difference is
considered to occur because of the cable losses from the output of the antenna to the
network analyzer used to measure the received power [13]. The asymmetrical pattern is
due to the chamber layout, which is also asymmetrical.

Figure 28. Measured azimuth gain pattern of the six-element array (From [13]).

2. Down-conversion Block

The block diagram of the down-conversion process is shown in Figure 29. The
input signal coming from the antenna is down-converted into baseband I and Q signals
as the outputs. Three main components are used within the down-conversion process: (1)
the low-noise amplifier (LNA), (2) the quadrature demodulator and (3) a differential
amplifier. These components are discussed in the following sub-sections.

57
 Figure 29. Down-conversion block detailed block
diagram of an array channel.

a. Low-Noise Amplifier (LNA)

One of the key factors in achieving superior sensitivity in receiver

systems is to have a low noise figure. Devices such as amplifiers add noise to the signal.
When there are N devices in cascade, the noise figure of the overall system is calculated
as [29]
F2 − 1 F3 − 1 FN − 1
Foverall = F1 + + + ⋅⋅⋅ + (5.3)
G1 G1G2 G1G2 ⋅⋅⋅ GN

where Fn and Gn are the noise figure and gain of the nth (1 ≤ n ≤ N ) component,
respectively.

As is seen from Equation (5.3), the first element in the system has the
biggest contribution to the noise figure since it does not have a gain factor in the
denominator. The purpose of an LNA in a receiver system is to decrease the overall noise
figure. This can best be achieved by placing the LNA at the front end of the receiver so
that the noise contributions of the subsequent devices are much lower.

The LNA used in the system is the LNA-2700, which is produced by RF

Bay Inc. The LNA-2700 has a low noise figure (1.7 dB) and high gain (23 dB). The
operating frequency of the LNA-2700 ranges from 2.2 GHz to 3.2 GHz, which is
consistent with the receiver system used. The LNAs are located as close as possible to
the antenna elements in order to prevent cable losses and reduce mismatch.
58
 b. Quadrature Demodulator

The quadrature demodulator used in the system is the AD8347 Direct

Conversion Quadrature Demodulator integrated circuit (IC) by Analog Devices.
Quadrature demodulators are used in the receiver system in order to down-convert the
RF signal received by the antenna to baseband frequencies. A simplified diagram of this
process is illustrated in Figure 30. The baseband I and Q signals are then sent to the
A/D converters.

Figure 30. Quadrature demodulator operation block diagram.

As mentioned in Chapter III, the high-frequency components that are

created from the mixing process are rejected by the filters and the baseband signals are
passed. Baseband amplification is also performed.

From Figure 29, it can be seen that the AD8347 Direct Conversion
Quadrature Demodulator IC produces four differential outputs, which are indicated as I+,
I-, Q+ and Q-. This gives rise to the requirement that four ADCs are required for further
signal processing. However, this requirement is eliminated by using the differential
amplifiers described in the next sub-section.

59
 An important issue with regard to the quadrature demodulators is that they
have a dc voltage offset which affects the accuracy of the I and Q voltages. An
automated calibration process was developed in [30] in order to find this offset. The
calibration is done by a LabView program that uses the NI PXI-5112 digitizers on the NI
PXI-1044 chassis. The PXI-5112 digitizers also need to be internally calibrated using a
LabView calibration program distributed by National Instruments (NI).

In the demodulator calibration process, a loop in which the RF phase is

shifted by 10o within the range [0o ,360o ] is executed by the LabView program. The
phase shifts are introduced by using an AD8346 modulator board. It is required that the
AD8346 modulator and the AD8347 demodulator should be frequency synchronized
during the calibration process. For this purpose, the Vaunix Technology Lab Brick LSG-
402 signal generator is used. The signal out of the LSG-402 is provided to the LO inputs
of the AD8346 modulator and the AD8347 demodulator via a power splitter.

After the phase-stepping loop is executed, the I / Q plot, which is also

known as the I / Q circle, is generated using the values obtained at each 10o phase shift.
The dc offsets are calculated as the offset values that are required to center the I / Q
circle at zero.

In Figure 31, the user interface of the LabView calibration program is

shown. The user interface has an LO control section that is used to set the necessary
parameters (e.g., frequency, power) of the LSG-402 signal generator. As outputs, the
user interface displays information about the calibration process, such as the PXI-5112
digitizer calibration progress status, instantaneous I and Q signal plots, the calculated
offset values and plots of the I / Q circles. Both the uncalibrated and calibrated
(i.e., centered to zero) I / Q circles are displayed (on the upper and lower right of the
screen, respectively) in the program. Examples of these plots are also shown in
Figure 32.

60
Figure 31. LabView automated dc offset calibration program user interface.

Figure 32. Uncalibrated (left) and calibrated (right) I / Q circle plots.

61
 c. Differential Amplifier

The differential amplifier block is built on the operational amplifier

(Op-amp) chip LM622 from National Semiconductor. The differential amplifier is
composed of three main stages, which have different functions and provide different
advantages.

As mentioned previously, using the four differential outputs from the

AD8347 quadrature demodulators would require four ADCs for further signal
processing. The first stage of the differential amplifier block converts the four
differential outputs of the quadrature demodulators into two outputs by combining I+
with I- and Q+ with Q- using [10]
I = I+ − I−
(5.4)
Q = Q+ − Q− .
Thus, the required number of ADCs is reduced as well as the cost to build the circuit.

The second stage of the differential amplifier block provides a voltage

gain of 10 dB to the baseband signal. The third stage includes a low-pass filter that
contributes to the elimination of the high-frequency components produced during the
down-conversion process. The 3-dB cutoff frequency of the filter was chosen as 10
megahertz (MHz) in accordance with the bandwidth of the NTSC video signal, which is
6 MHz.

3. Local Oscillator (LO)

The LO provides a signal source to the quadrature demodulators during the

down-conversion process. The LO used in the antenna system is the Lab Brick signal
generator (LSG-402) by Vaunix Technology Corporation. In Figure 33, the connection
between the device and the quadrature demodulator board is shown. The Lab Brick
signal generator has a frequency range between 1000 and 4000 MHz, which makes it
suitable for use in the antenna system. The device also has the capability to provide an

62
output level up to 20 decibels per miliwatt (dBm). The power and control of the device is
provided by a Windows-based computer using a Universal Serial Bus (USB) port.

Figure 33. LO (Lab Brick signal generator) connected to the

quadrature demodulator.

4. A/D Converter

National Instruments PXI-5112 high-speed digitizers are used as A/D converters

in the receiver system. The PXI-5112 uses a high-speed Peripheral Component
Interconnect (PCI) bus, and therefore provides acquisition and processing of waveforms
at higher speeds compared to General Purpose Interface Bus (GPIB)-based devices.

Nyquist’s sampling theory states that the sampling frequency of a signal must be
at least twice that of the highest frequency component. The PXI-5112 A/D converters
provide a sampling rate up to 100 MS/s (mega samples per second), which means that
signals with frequencies up to 50 MHz can be acquired without aliasing.

63
 Another important feature of the PXI-5112 is that it has 8-bit resolution with an
adjustable voltage range (0 to 10 V). This means that 28 = 256 different levels can be
used for the quantization of an analog signal. Hence, it provides lower quantization noise
and high sensitivity compared to lower resolution A/D converters.

C. SOFTWARE COMPONENTS

The block diagram of the receiver system was shown in Figure 25. It was
mentioned that most of the signal processing is done with software. In order to receive
video data successfully from a UAV, the UAV signals should be tracked and the video
data within the received signals should be retrieved. The software portion of the system
shown in Figure 25 includes software modules, which have specific functions for this
purpose. The modular approach was used for the development of the software
components in order to make testing and further development of the system easier. Each
module can be upgraded separately or replaced by new versions in the future.

The software components consist of three software modules, which were

implemented according to the principles that were discussed in Chapters II, III and IV.
These modules are namely: (1) Tracking module, (2) FM demodulation module and (3)
NTSC decoding module. Since these modules were developed using NI LabView
software, familiarity with the LabView programming environment is necessary. In this
section, the LabView programming principles are briefly introduced and the
implementation of the software modules discussed. The testing of the modules is
discussed in Chapter VI.

1. NI LabView Software Overview

NI LabView is a graphical programming environment that is used to develop

sophisticated engineering applications. It is crucial to understand the basic structure and
principles in LabView since most of the signal processing in the receiver system was
created in LabView programming environment.

64
 Programs in LabView are developed using intuitive graphical icons and wires
that resemble a flowchart. Programs created using LabView are called virtual
instruments (VIs). This is because the appearance and operation of these programs
imitate physical instruments. The VIs in LabView contain two main components: the
front panel and the block diagram. The front panel is the user interface of the VI. The
controls and indicators are implemented in the front panel. After building the front panel,
graphical representations and functions are used to add code to control the objects in the
front panel. This graphical code is built in the block diagram. An example of a VI front
panel and the corresponding block diagram is shown in Figure 34.

Figure 34. VI front panel (left) and block diagram (right).

As the block diagrams get filled with several function blocks, it becomes hard to
follow the algorithm. This problem can be eliminated by creating VIs that are used
within other VIs. These kinds of VIs are called subVIs. SubVIs make the programs more
modular and more understandable. It is important to understand the terms “VI”, “subVI”,
“front panel” and “block diagram” since they will be used frequently in further sections.

2. Tracking Module

The principles of tracking in phased array antennas were discussed in Chapter II.
The tracking module used in the receiver system is an implementation of the RSNS and
DBF. The design of the tracking module was made by Tan and Pandya [13].
65
 Two separate DBF methods were implemented in the tracking module. The first
method utilizes RSNS for tracking. The RSNS system with moduli [5, 9], which was
described in Section B.3 of Chapter II, was implemented. This RSNS sequence provides
a dynamic range of 36 bins and a resolution of 5o at broadside. The second method uses
RSNS to obtain an initial scan angle and then utilizes monopulse DBF for target
tracking.

In order to start tracking, the signal that is of interest needs to be acquired. This
acquisition involves the calculation of the initial scan angle. It was mentioned in Chapter
II that for monopulse tracking, both sum and difference beams are used during reception.
If the target is not located at the boresight of the antenna, the difference beam output will
be non-zero. Hence, an error signal can be obtained by using the slope of the linear
region of the difference beam. In [13], the maximum initial scan angle that can be used
for efficient use of monopulse DBF was calculated by computing the angular range of
the linear region in MATLAB. It was observed that the linear region has a span of
approximately 20o . Hence, it was concluded that the initial scan angle must not be
greater than 20o off from the true direction of the signal. Since RSNS provides an
angular resolution of 5o , it can be used for finding the initial scan angle before
monopulse DBF is used for tracking.

The process flowchart of the tracking module is shown in Figure 35. The output
of the demodulator boards are sampled by the PXI-5112 A/D converters. Since there are
six array elements, there are six PXI-5112s and, hence, six array channels. Each PXI-
5112 board has two data channels (channel 0 for I data and channel 1 for Q data). One of
these boards is selected arbitrarily as the master, in order to act as a reference triggering.
The other boards are called slaves. This is done to ensure that the data sampled from
each channel is correlated. In Figure 36, a portion of the VI block diagram that is used
for this process is shown.

66
Figure 35. Tracking module process flow chart (After [13]).
67
 Figure 36. Data acquisition from demodulator boards by the PXI-5112 board.
(Only the master and a single slave channel shown.)

In the next step, the individual I and Q components are separated and each
board’s dc offsets are applied. This process removes any dc offset that is introduced by
the quadrature demodulators. The application of the dc offsets in the software is shown
in Figure 37 for two array channels. The board dc offsets are calculated with the AD8347
calibration software.

After the offsets are applied, a calibration process needs to be conducted in order
to account for the phase differences between the RF portions of the array channels.
Ideally, when a plane wave is incident normal to the array elements, it is expected that
each channel will have the same phase value. However, the test conducted in [13]
revealed that the channels did not have the same phase values. This was attributed to
variations in the phase of the connectors and the RF cables used. As a result, the phase
differences between the channels need to be adjusted by measuring each channel relative
to a common phase reference. The differences are removed in the digital beamforming.

68
Figure 37. Application of dc offsets for channel 0 and channel 1
of the six array channels.

69
 The next step of the algorithm is the calculation of the scan angle. The tracking
module allows the users to choose between two tracking methods: (1) RSNS and (2)
monopulse DBF. When RSNS is selected, the phase difference between adjacent
elements (1–2, 3–4, 5–6) is computed. The three values obtained from this process are
averaged, and the result is used to calculate the AOA. The reason for the averaging
process is that even though the RSNS requires only two elements to compute the AOA,
by averaging over the three pairs, the phase errors will average out.

When monopulse DBF is selected for tracking, the initial scan angle is calculated
with RSNS and then monopulse DBF takes over for the rest of the computation. The
array factors are computed based on Equation (2.16), which was given in Chapter II.
After the array factors are obtained, the sum and difference outputs are computed. Using
these values and the monopulse slope constant given in Equation (2.2), we compute the
correction angle for the direction of scan. The implementation of this process in software
is shown in Figure 38. The current scan angle is updated by using the correction angle
and the result obtained is used for the computation of the array factor in the next
iteration. The module continuously computes the pointing error in order to make the
correction angle approach to zero.

70
Figure 38. Computation of correction angle from array factor
and sum/difference beams.

71
 Finally, the front panel of the tracking algorithm is shown in Figure 39. It is
required that the necessary settings (e.g., delays) for the PXI-5112 A/D converters, board
offset values, phase adjustments and tracking method selection be made before running
the program.

Figure 39. Front panel of the tracking algorithm (From [13]).

3. NTSC Decoding Module

Of interest in this research, is the reception of video signals that are in NTSC
format. After the FM modulated signals are received and demodulated, they need to be
decoded in order to be displayed on a video screen. For this purpose, a VI for NTSC
video signal decoding for a single channel was developed by Professor David C. Jenn
(Naval Postgraduate School) and modified for use in the UAV digital tracking array
hardware. The VI was named as “NTSC Video Decode” and consists of three subVIs,

72
which are, namely: (1) Open and Fetch from 5112, (2) Fetch from 5112 and (3) NTSC
Processor. The functions of these subVIs are illustrated in the process flow diagram
shown in Figure 40.

Figure 40. Process flow diagram for the NTSC decoding module (After [13]).

The VI block diagram for NTSC Video Decode VI is given in Figure 41.
The program takes several inputs for the configuration of the settings for sampling of the
5112 boards and the decoding process. These include minimum sample rate, minimum
record length, vertical range, brightness level, horizontal sync adjust and trigger
coupling. Once the program is run, the data is obtained by the Open and Fetch from 5112
subVI, which is called within Fetch from 5112 subVI. After sampling, the vertical sync
level (Vsync) is computed in Fetch from 5112 subVI and sent to the NTSC Processor
subVI along with the sampled video data. The data obtained by this subVI is processed
and the video image is recovered and displayed on the front panel screen.

73
 Figure 41. NTSC decoding module VI block diagram.

In the following sections, the subVIs used in NTSC Video Decode VI are
described.

74
 a. Open and Fetch from 5112 SubVI

The VI block diagram of Open and Fetch from 5112 subVI is shown in
Figure 42. The main function of this subVI is to obtain the I and Q signals from the
corresponding channels of the A/D converter according to the configuration settings. The
trigger coupling is set to “dc” and the trigger source is set to “immediate” to acquire data
continuously. The minimum sampling rate used is 20 MS/s and the vertical range used is
2 volts.

Figure 42. Open and Fetch from 5112 subVI block diagram.

Since Open and Fetch from 5112 is used as a subVI, the front panel is not used as
a user interface and will not be shown in this section.

75
 b. Fetch From 5112 SubVI

Fetch from 5112 subVI uses Open and Fetch from 5112 subVI to acquire
data from the PXI-5112 A/D converter. This process is shown in Figure 43. Also it is
seen that some of the inputs that are entered to NTSC Video Decode VI are used in Fetch
from 5112 subVI.

Figure 43. Fetch from 5112 subVI block diagram data acquisition section.

Since Fetch from 5112 is used as a subVI, the front panel is not used as a user
interface and will not be shown in this section.
76
 Once the data is acquired, a calibration process is utilized for the calculation of
the vertical sync threshold value, Vsync. This threshold value is used for the
determination of a new synchronization pulse by the NTSC Processor VI. The details for
this process are discussed in the next section.

Figure 44. Fetch from 5112 subVI block diagram, calculation of vertical sync.

77
 c. NTSC Processor SubVI

NTSC Processor subVI is the most important and complex subVI within
the NTSC Video Decode VI. The NTSC-encoded video signal is retrieved and video data
is displayed by this subVI. It was mentioned that the video signals are acquired and the
threshold Vsync is calculated by the Fetch from 5112 subVI. NTSC Processor subVI
takes the video data along with the Vsync value and tries to find a valid sync pulse. The
location of this process in the video signal is illustrated in the timing diagram given in
Figure 45. The validation starts by comparing the amplitudes of each data point with
Vsync. After a data point with amplitude less than Vsync is found, another comparison
process starts to find a data point that has larger amplitude than Vsync. If this data point
is found as well, the time between the two data points is calculated and compared with
the time considered to be the minimum duration for a valid sync pulse. This duration is
represented by Tmin. If the calculated time is greater than Tmin, the pulse is validated as
a sync pulse. If it is less than Tmin, the pulse is considered as noise.

Figure 45. NTSC video signal timing diagram (After [13]).

78
 In Figure 46, the software implementation of the sync pulse validation
process within the NTSC Processor subVI is shown.

Figure 46. NTSC Processor subVI block diagram section for sync pulse detection.

The calculated duration of the sync pulse is also used for determining
whether the field found is odd or even. This process is done in the section of the NTSC
Processor subVI block diagram shown in Figure 47. Also implemented in this section is
the configuration of several parameters for the video decoding process that is done next.

79
Figure 47. Field type determination and parameter configuration in NTSC Processor
subVI.

In the next step, the data portion of the video signal is read. The odd and
even fields are read and the video image is constructed. There are 259 lines read until the
end of each field and 640 pixels until the end of each line. The odd field is read first and
the even field is read afterwards. When both fields are read successfully, the video image
is constructed and displayed in the video screen located on the front panel of the NTSC
Video Decode VI. The software implementation of the video data read process is shown
in Figure 48.

80
 Figure 48. Video data stream reading process in NTSC Processor subVI.

Since NTSC Processor is used as a subVI, the front panel is not used as a user
interface and is not shown in this section.

4. FM Demodulation Module

There are two versions of this module used in the digital tracking array hardware,
which are, namely, FM Demod Single Fetch VI and FM Demod Single Fetch Tangent
VI. The FM Demod Single Fetch VI was designed and built by Tan and Pandya [13] in
2009. It is based on the FM Demodulation function included in the Signal Processing
toolkit in LabView. The Fetch from 5112 subVI, introduced previously, is used within
this module in order to obtain data from the PXI-5112 channels. The settings that are
required for this subVI are the same as what was given previously in Section 3 of this
chapter and adjusted from the front panel of the FM Demod Single Fetch VI. The front
panel of this VI was modified from its original version to provide a better user interface.
The modified front panel is shown in Figure 49.
81
 Figure 49. FM Demod Single Fetch VI front panel.

There are five waveform charts in the front panel used for displaying the
following components: (1) modulated signal ( I + jQ , upper left), (2) demodulated signal
before filtering (upper middle), (3) demodulated signal after filtering (upper right),
(4) in-phase component of the incoming signal (lower left), (5) quadrature component of
the incoming signal (lower middle). The display located on the lower right of the front
panel is used for displaying the demodulated video data after being decoded.

The block diagram of FM Demod Single Fetch VI is shown in Figure 50. It is

noticed that the NTSC decoding module is embedded within the demodulation module
for displaying the video image after demodulation. When the program is run, the data is
fetched from the 5112 channel by the Fetch from 5112 subVI. This subVI provides
I and Q components of the signal at its output. These components are bundled together
along with the timing and sampling information and sent to the MT Demodulate FM
function. The location of this function is marked with a circle in Figure 50. The
demodulated data is then sent to the NTSC Processor subVI for decoding and displaying
the video data.

82
 Figure 50. FM Demod Single Fetch VI block diagram.

The second version of the FM demodulation module is the FM Demod Single

Fetch Tangent VI, which is built by the modification of FM Demod Single Fetch VI. The
difference in this VI is that the MT Demodulate FM function in the previous version was
replaced with a subVI called Tangent Demod. This subVI was designed to implement
tangent-type demodulation of the baseband I and Q signals. The implementation is
based on the theory that was given in Chapter IV Section B. The front panel of this VI is
the same as the one that was shown in Figure 49. The block diagram of this subVI is
shown in Figure 52. The inputs of the subVI are the I and Q components delivered by
the Fetch from 5112 subVI. The output of the subVI is connected to the Butterworth
filter, where the signal is filtered and displayed.

83
Figure 51. FM Demod Single Fetch Tangent VI block diagram.

Figure 52. Tangent Demod subVI block diagram.

84
D. SYSTEM POWER SPECIFICATIONS

In [13], the performance analysis of the antenna system is given. The minimum
power that can be received by each channel is given as
Pmin ch (dB) = Pmindemod (dB) − GLNA (dB) = −90 dBm= − 120 dB (5.5)

where Pmin demod = −65 dBm is the start of linear region of TPC30 on the AD8347 datasheet

[31] and GLNA = 25 dB is the gain of the LNA located after the antenna. Using the

Pmin ch value obtained, we find the maximum range of the array system as

t t Gr λ
2
PG
R= = 34.4 km (5.6)
(4π ) 2 Pminch

where Pt = 0.1 W and Gt = 3 dB are the UAV transmitted power and antenna gain,

respectively, and Gr = 10 dB is the array element gain. Using the same equation, we

calculate the maximum range without the LNA (i.e. Pminch = −95 dB) as 1705.8 m.

Comparing the maximum ranges with and without the LNA, we see that the LNA is an
essential part of the receiver in order to achieve practical maximum ranges. It also
eliminates the need to transmit higher power at the UAV to achieve longer ranges and
controls the receiver noise.

The minimum SNR is calculated by

Pmin ch
SNRch = N = 12.2 dB (5.7)
N sys

where N sys = kTsys B = −124.4 dB is the system noise calculated in [13] and

N = 6 = 7.8 dB is the number of array elements.

85
E. SUMMARY

In this chapter, the hardware and software components of the UAV digital
tracking array system were introduced. The performance parameters for some of the
hardware (e.g., antenna) were also given. LabView programming environment was
introduced and details of the software modules were presented.

In the next chapter, the results of the simulations done for the validation of the
demodulation techniques and bench-top tests conducted for the verification of the
software modules are presented.

86
 VI. DESIGN VALIDATION AND VERIFICATION

In this chapter, the validation and verification process of the software

components developed for the digital tracking array receiver processor are discussed. In
[32], the definitions of validation and verification are given. Validation is associated with
the question, “Are we building the right solution?” and verification is associated with the
question “Are we building the solution right?” The validation and verification processes
discussed in this chapter are based on these questions. In Section A, simulations, which
were done to validate the software components, are explained. In Section B, the tests that
were made for the verification of these software components on the hardware are
discussed.

A. VALIDATION OF FM DEMODULATION TECHNIQUES

The FM demodulation techniques that are used to demodulate video signals

transmitted from a UAV were discussed in Chapter IV Section B. These techniques
included demodulation at baseband, tangent- type demodulation and demodulation using
Kalman filters. It was mentioned that the Kalman filter approach is expected to be slow
and have a higher complexity compared to the other approaches. Hence, it was not
implemented for the proposed system.

A good method for simulation is to use simple inputs at the beginning and more
complex inputs, which are similar to real inputs, later. For this purpose, there were two
waveforms used for the simulation of each demodulation technique. One of these
waveforms was selected to be a simple sine wave with 50-kHz frequency. The second
waveform chosen was a video signal that was captured using a video camera with
LabView. The video signal captured was saved as a text file that included the sampled
video data. This text file was then fed to the simulation programs as the input (i.e.,
modulating signal). The simulations were done separately in MATLAB and LabView.
The results of the simulations are discussed in the following sub-sections.

87
 1. Baseband FM Demodulation Simulations

The baseband FM demodulation technique which was given theoretically in

Equation (4.11), was simulated using MATLAB and LabView. The demodulation
process was illustrated in Figure 23. The generic block diagram for the simulation of
baseband demodulation is given in Figure 53. The following sections include discussions
about the simulations based on this block diagram.

Figure 53. Block diagram for baseband demodulation simulation.

a. MATLAB Simulations

As mentioned previously, a 50 kHz sine wave and a data file including

real video data captured from a video camera were used as modulating signals for the
simulations.

The MATLAB code used to simulate the demodulation of the sine wave
is given in Appendix Section A. For the 50-kHz sine wave, the sampling frequency used
is 1.5 MHz and the frequency deviation applied is 20 kHz. The result of the simulation is
shown in Figure 54. It is seen that the frequency of the message signal was recovered
successfully. However, the demodulated signal had to be amplified to match the original
signal since the output of the demodulator produced very low voltages.

88
 Message signals
2
Original
1.5 Demodulated

1
Message Signal, volts

0.5

-0.5

-1

-1.5

-2
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Time, sec -4
x 10

Figure 54. Comparison of original and demodulated signals using baseband

demodulation in MATLAB.

In the second part of the simulation, real-time video data was used as the
modulating signal. A MATLAB program, which is given in Appendix Section B, was
used to display the video frames. The original video frame displayed using this program
is shown in Figure 55.

89
 Figure 55. Original video frame used as the modulating signal.

The MATLAB code for the simulation is given in Appendix

Section C. In the simulation, the input signal (i.e., video) needs to be integrated and
frequency modulated before sending it to the demodulator. The video signal that was
used as the modulating signal in the simulation is plotted in Figure 56. The output
obtained from the demodulation of the FM video signal is shown in Figure 57. Also, the
demodulated video frame is shown in Figure 58. The results show that FM video signal
was demodulated successfully using the baseband demodulation approach discussed in
Chapter IV. However, the amplitude of the demodulated signal was very low, which
reveals the need for amplification or rescaling in the processor.

90
 Modulating signal
1

0.8

0.6

0.4

0.2
Amplitude

-0.2

-0.4

-0.6

-0.8

-1
0 50 100 150
Time (sec)

Figure 56. Video signal used as the modulating signal.

-6
x 10 Output after demodulation
6

0
Amplitude

-2

-4

-6

-8
0 5 10 15 20 25 30 35 40 45 50
Time (sec)

Figure 57. Demodulated video signal.

91
 Figure 58. Demodulated video frame.

The phase and amplitude of the signal for every shifted RF phase degree
is determined by plotting the baseband I and Q values. A good phase response from the
RF signal fed to the demodulator is expected to give a circle with equally distributed RF
input signal samples for every phase degree [10]. The phase response obtained from the
video signal is shown in Figure 59. It is seen that the I and Q values are distributed on a
symmetrical circle as expected.

92
 I vs Q

0.8

0.6

0.4

0.2

0
I

-0.2

-0.4

-0.6

-0.8

-1

-1 -0.8 -0.6 -0.4 -0.2 0 0.2 0.4 0.6 0.8 1

Q

Figure 59. I vs. Q plot for the video signal.

b. LabView Simulations

The video signal demodulation simulation that was performed in the

previous section was repeated in LabView in order to verify the results obtained from
MATLAB. Moreover, LabView performs faster than MATLAB in most situations where
arrays of data are processed using loops. With this feature of LabView, long calculation
times encountered in MATLAB were avoided.

A LabView VI was developed for performing the baseband demodulation

simulation. The algorithm of the program was developed based on the block diagram
given in Figure 53. The construction of the video signal from a data file is illustrated in
the partial VI block diagram shown in Figure 60. The other portion of the VI block
diagram, which performs modulation, mixing, demodulation and amplification is shown
in Figure 61.
93
 Figure 60. Video signal construction from data file in LabView.

Figure 61. Modulation, mixing, demodulation and amplification process.

In the simulation, the video data file that was used previously is used to
create the modulating signal. The modulating signal obtained from the file is shown in
Figure 62. The resulting video signal after the demodulation process is given in Figure
63. Comparing the two figures, we see that the video signal was recovered successfully
using the baseband demodulation method.
94
Figure 62. Video signal used as the modulating signal.

Figure 63. Demodulated video signal.

95
 The only problem that was encountered with this simulation program was
that the frequency deviation had to be selected as a very low value [i.e., 100 Hz] since
high attenuation of the signals over time was encountered with high-frequency deviation.
An example of this attenuation is shown in Figure 64.

Figure 64. Attenuation of demodulated signal in the simulation

due to high-frequency deviation.

2. Tangent-type FM Demodulation Simulations

The tangent-type FM demodulation of baseband I and Q signals was discussed

in Chapter IV. The demodulation process, which was illustrated in Figure 24, is
simulated using MATLAB and LabView. The generic block diagram for the simulation
program of tangent-type baseband demodulation is given in Figure 65. Two simulations
for two different cases were performed, as was done for the baseband demodulation.
Also, the simulations were first done with MATLAB and then verified with LabView.
The following sections include discussions about the simulations based on the block
diagram given in Figure 65.

96
 Figure 65. Block diagram for tangent-type FM demodulation simulation.

a. MATLAB Simulations

The procedure given in Figure 65 was implemented in MATLAB to

simulate the two scenarios for demodulation. Similar to the method used for baseband
demodulation, a 50-kHz sine wave and a video signal were used as modulating signals.
The MATLAB code for the first scenario is given in Appendix Section E.
In this scenario, a 50-kHz sine wave was formed as the modulating signal. The carrier
frequency and the frequency deviation used were 1.5 MHz and 10 kHz, respectively. The
sampling frequency was set to be at least five times the sum of the message and carrier
frequencies (i.e., f s = 5 ⋅ ( f m + f c ) ). The 50-kHz sine wave was formed and impressed on

the carrier. It was then split into two channels ( I and Q) , which were then down-
converted by mixing and filtering. The baseband I and Q signals obtained after this
process were then input to the demodulator. The high-frequency components created by
the mixing process were suppressed by using a third-order Butterworth low-pass filter
with a cut-off frequency equal to 2/5 of the message frequency (i.e., f cut −off = 0.4 ⋅ f m ).

97
 The result of the demodulation process is given in Figure 66. It is seen
that the frequency of the modulating signal was successfully recovered. However, the
amplitude of the demodulated signal had to be amplified 32 dB to match the original
signal since the output of the tangent-type demodulator provided very low amplitudes.
Also seen is a small phase shift between the original signal and the demodulated signal.
This phase shift is created by the filter used. Additional tests performed with different
parameters showed that, as the cut-off frequency of the filter is increased, the phase shift
becomes larger.

Demodulated Signal vs. Modulating Signal

Demodulated
0.3 Original

0.2

0.1
Amplitude

-0.1

-0.2

-0.3

3 4 5 6 7 8 9 10 11 12
Time (sec) -5
x 10

Figure 66. Demodulated vs. original message signal

for tangent-type demodulation.

98
 In the second scenario, the video signal used previously was applied to the
demodulation procedure described in Figure 65. The MATLAB code for this process is
given in Appendix Section F. The video data read from a text file was impressed on a
carrier with a frequency of 20 kHz. The sampling rate and the frequency deviation used
were 100 kHz and 1 kHz, respectively. The modulation was done with MATLAB’s
fmmod function. The modulated signal was separated into two channels to form I and Q
signals. These channels were then down-converted to baseband by mixing and filtering.
The filter used was a third-order Butterworth low-pass filter with a cut-off frequency
equal to the carrier frequency (i.e., f cut −off = f c ). After this process, the baseband I and Q

signals were fed to the input of the arctangent block. The output of the arctangent block
was fed into a differentiator and the demodulated signal was obtained at the output. The
demodulated signal was then put through the improvement process. Finally, the resulting
signal was plotted. It was noticed that the demodulated video signal obtained from the
improvement section had a dc offset of 0.8375 volts. Also, the amplitude of the
demodulated video signal was lower than the original video signal. Hence, to
compensate for these conditions, a dc offset of -0.8375 was applied to the demodulated
video signal first and 12-dB amplification was applied to the resulting signal.

The comparison of the modulated (i.e., original) video signal and the
demodulated video signal is made in Figure 67 and Figure 68. It is seen that the original
video signal was successfully recovered. In addition to these figures, the demodulated
video frame, which was obtained from the MATLAB program given in Appendix
Section B, is shown in Figure 69.

99
 Video signal captured from camera
1

0.8

0.6

0.4

0.2
Amplitude

-0.2

-0.4

-0.6

-0.8

-1
0 50 100 150
Time (sec)

Figure 67. Original video signal.

Demodulated video signal

1

0.8

0.6

0.4

0.2
Amplitude

-0.2

-0.4

-0.6

-0.8

-1
0 5 10 15 20 25 30 35 40 45 50
Time (sec)

Figure 68. Demodulated video signal (tangent-type demodulator).

100
 Figure 69. Demodulated video frame.

The phase response obtained from the video signal is shown in Figure 70.
It is seen that the I and Q values are distributed on a symmetrical circle as expected.
However, the circle obtained is not a perfect circle. There are even four points where the
values are not on the circle. These points and the overall shape of the circle are attributed
to the distortion introduced by the low-pass filtering process.

101
Figure 70. I / Q circle obtained from tangent-type demodulation of the video signal.

b. LabView Simulations

The video signal demodulation simulation that was performed in the

previous section was repeated in LabView in order to verify the results obtained from
MATLAB. A LabView VI was developed for performing the simulation of the tangent-
type demodulator. The VI was developed based on the block diagram given in Figure 65.
Different than the previous simulations, modulated video data is used as an input in this
program. The modulation is done in MATLAB and the modulated video data is saved in
a text file. This text file is then fed to the LabView VI and the data is used as the
modulated video signal.

102
 In the program, the modulated video signal is down-converted and I and
Q channels are formed by the mixing and filtering processes. The implementation of file
reading and signal mixing process is illustrated in the first part of the VI block diagram
given in Figure 71. The VI also includes a single block which reads the original video
data from the unmodulated video data file to plot the original video signal in order to
compare with the demodulated signal. This block can be seen in the upper left section of
Figure 71.

Figure 71. Partial VI block diagram (Part 1) showing

the file read and signal mixing process.

The I and Q channels obtained from the down-conversion process are then
fed to the tangent-type demodulator. The output of the demodulator is differentiated and
sent to another block for the improvement process. Finally, the dc offset that occurs in
the demodulation process is removed and the resulting signal is amplified. The
implementation of these processes is shown in the partial VI block diagram given in
Figure 72.

103
Figure 72. Partial VI block diagram (Part 2) showing the filtering, demodulation,
improvement, dc offset and amplification process.

Similar to the MATLAB simulation, the carrier frequency, sampling rate

and the frequency deviation used in the LabView simulation were 20 kHz, 100 kHz and
1 kHz, respectively. The modulation was done with MATLAB’s fmmod function. The
original video signal is plotted in Figure 73 and the demodulated video signal obtained
from the tangent-type demodulator is shown in Figure 74. Comparing the two figures, we
see that the video signal was recovered successfully using the tangent-type demodulation
method.

104
 Figure 73. Original video signal before modulation and demodulation.

Figure 74. Demodulated video signal obtained from tangent-type demodulator.

105
B. SOFTWARE MODULES PERFORMANCE VERIFICATION TESTS AND
RESULTS

In Section A, the validation of the FM demodulation techniques were made. In

this section, the verification of the software modules developed using these
demodulation techniques is done by performing bench-top tests using the hardware. The
user interfaces of the NTSC decoding module and the baseband demodulation module
described in Chapter V were modified from their original version, which were given in
[13], in order to provide more user friendliness.

1. NTSC Decoding Module Tests

The NTSC Decoding Module test was performed by using a video camera to
acquire real-time video into the module. The test measures the ability of the module to
receive a video data stream and display it on the screen. The output of the video camera
was input into the PXI-5112 digitizers and the NTSC Video Decode VI introduced in
Chapter V was run. In this test only one channel (i.e., channel 0) of the PXI-5112
digitizer was used.

The modified front panel of the NTSC decoding module is shown in Figure 75.
The setup parameters needed for the sampling of the data from the PXI-5112 digitizers
and displaying the video signal are located on the upper left of the front panel. The
module has two displays on the right. The upper display shows the real-time video
stream captured by the video camera. Below the display are the two settings required for
the appropriate screening of the video stream. These are the “brightness level” and the
“horizontal sync adjust.” The horizontal sync adjust is an important setting, since an
inappropriate setting of this parameter causes tilting of the image. An example of this
tilting effect due to inadequate horizontal sync adjust is shown in Figure 76. The lower
display shows the amplitude plot of the streaming video. This plot can be enabled or
disabled by using the button located on the left of the front panel.

106
 Figure 75. Snapshot of the NTSC decoding module front panel while running.

Analyzing the video data plot verifies most of the features of the NTSC video
signals discussed in Chapter IV. A close-up view of one of the vertical blanks seen in the
video signal plot is shown in Figure 77. Here, the horizontal sync and the back porch
regions that indicate the beginning of a new line can easily be observed. The sync pulses
in the video signal region located to the right of the back porch region are also seen.

107
 Figure 76. Tilted video image due to inappropriate setting of
the horizontal sync adjust.

Figure 77. Video signal region showing the beginning of a new field.

108
 During the test process, we used a frame in which all objects were stationary for
demonstrational purposes. The output obtained from the module was shown in Figure 75.
For comparison, a still image of the same location was taken with a digital camera. This
image is shown in Figure 78. It is seen that the module displays the image with enough
accuracy. However, the quality of the image was traded off for frame-processing speed.
To achieve faster processing, only one sample per pixel is used during the decoding
process, which decreases the quality of the image.

Figure 78. Digital camera image of the location captured by the video camera.

109
 2. FM Demodulation Module Tests

In this section, the operational testing of the two demodulation modules

(i.e., baseband and tangent-type demodulation modules) introduced in Chapter V is
given. Both modules were tested using only one array channel. Two different test
procedures were performed. The antenna elements were not used in these tests. The setup
diagram for the tests is shown in Figure 79. As shown in the setup diagram, the only
difference between the two test procedures was the source of the modulating signal. In
the first test, a signal generator was used to generate simple waveforms (e.g., sine,
square, ramp waves). In the second test, a real-time video signal captured by a video
camera was input into the synthesizer for FM modulation. In both tests, the modulated
signal was up-converted to 2.4 GHz in order to simulate a transmission from a UAV. The
FM signal obtained at the output of the synthesizer was input into the quadrature
demodulator for frequency down-conversion to baseband. The quadrature demodulator
was frequency synchronized by using an LO. The four signals obtained at the output of
the quadrature demodulator were combined to construct I and Q signals and then
sampled by the PXI-5112 digitizers. The digitized signals were then processed by the
LabView VI for final demodulation. The details of the tests and the results obtained are
discussed in the following sections.

Figure 79. Demodulation test set up.

110
 a. Baseband Demodulation Module Tests

The baseband demodulation module introduced in Chapter V was tested

as described in Figure 79. In the first test, a sine wave with a frequency of 50 kHz and
amplitude 0.6 V peak-to-peak was generated using the signal generator. The amplitude
of the signal was adjusted to prevent over-modulation. This signal was then fed into the
signal synthesizer, which modulated the sine wave using FM, up-converting the signal to
RF at 2.4 GHz. The baseband I and Q signals obtained from the quadrature demodulator
were input into channels 0 and 1 of the PXI-5112 digitizers, respectively. The output
obtained from the FM Demod Single Fetch VI is shown in Figure 80. An expanded
version of the demodulated signal is shown in Figure 81.

Figure 80. Output obtained from the FM Demod Single Fetch VI.

It is seen that the sine wave was recovered successfully by the module.
The filter used in this test was a third-order Butterworth filter with 100-kHz cut-off
frequency.

111
 Figure 81. Demodulated sine wave with 50-kHz frequency
and 0.6-V amplitude.

Since forming different waveforms is easy by using a signal generator, a

square wave and a ramp wave were also tested. When 50 kHz was used as the frequency
for these signals, the demodulated signals were distorted and looked more like a sine
wave. For this reason a lower frequency (i.e., 20 kHz) was tried for the square and ramp
waves. The outputs obtained from the module are given in Figure 82 and Figure 83. It is
seen that the waves were not recovered accurately. However, for both waveforms, the
fundamental frequencies of the input signals could be retrieved successfully. For the
square wave, there were ripples with high magnitudes observed. However, these were
attributed to the attenuation of the high-frequency components by the filter used in the
module. Although the cut-off frequency was increased to several times the message
frequency, the module failed to provide an accurate square wave. Similarly, with the
ramp wave, there were distortions on the signal recovered. These distortions could not be
eliminated by different filter settings either.

112
Figure 82. Demodulated square wave obtained from
the baseband demodulation module.

Figure 83. Demodulated ramp wave obtained from

the baseband demodulation module.

113
 The next tests performed consisted of using a real-time streaming video
signal as the modulating signal. The video signal is known to be an NTSC-decoded video
signal with 6-MHz bandwidth. For this reason, the filter used in the module was adjusted
to have a low cut-off frequency of 9 MHz. It was also expected that the demodulated
signal could be displayed by the NTSC Processor subVI located in the module.

When the test described in Figure 79 was run, it was seen that the output
amplitude level of the video camera caused over-modulation in the signal synthesizer. In
order to overcome this issue, a variable signal attenuator was placed between the video
camera and the synthesizer. It was observed that 3-dB attenuation was sufficient to
prevent over-modulation. A snapshot of the module while running is shown in Figure 84.
An expanded view of the video signal obtained and a zoomed version, showing the
horizontal sync and back porch regions, are shown in Figure 85 and Figure 86,
respectively. It was seen that the video signal could not be recovered as expected from
the simulations. The demodulated signal obtained had distortions, which caused the
NTSC decoding module fail to decode and display the video stream.

Figure 84. Snapshot of the baseband demodulation module

running with video signal input.

114
 Figure 85. Demodulated video signal obtained from the baseband
demodulation module with fcut-off = 9 MHz.

Figure 86. Zoomed video signal obtained from the baseband demodulation
module with fcut-off = 9 MHz.

115
 The distortion observed in Figure 85 and Figure 86 was attributed to the
filter cut-off frequency being too high to filter the noise affecting the signal. In order to
resolve this issue, the cut-off frequency was then decreased to 6 MHz and the test was
repeated. The resulting video signal obtained is shown in Figure 87.

Figure 87. Demodulated video signal obtained from the baseband

demodulation module with fcut-off = 6 MHz.

From Figure 87 it was seen that lowering the filter cut-off frequency did
not have a significant effect. As a final test, it was decided to set the filter cut-off
frequency to a very low value. The cut-off frequency chosen for this was 100 kHz. The
resulting video signal obtained is shown in Figure 88.

116
 Figure 88. Demodulated video signal obtained from the baseband
demodulation module with fcut-off = 100 kHz.

A close-up view of the video signal given in Figure 88 is shown in Figure

89. Comparing this with the original video signal given previously in Figure 77, it is seen
that the demodulated signal is still too distorted to be decoded.

Figure 89. Zoomed video signal obtained from the baseband demodulation
module with fcut-off = 100 kHz zoomed.

117
 b. Tangent-type Demodulator Tests

The tangent-type demodulation module introduced in Chapter V was

tested as described in Figure 79. Similar to the previous module test, a sine wave with a
frequency of 50 kHz and amplitude 0.6 V peak-to-peak was generated using the signal
generator. The amplitude of the signal was adjusted to prevent over-modulation. This
signal was then fed to the signal synthesizer for modulation and up-conversion to RF at
2.4 GHz. The baseband I and Q signals obtained from the quadrature demodulator were
input into channels 0 and 1 of the PXI-5112 digitizer, respectively. The output obtained
from the FM Demod Single Fetch Tangent VI is shown in Figure 90. An expanded
version of the demodulated signal is shown in Figure 91.

Figure 90. Output obtained from the FM Demod Single Fetch Tangent VI.

It is seen that the sine wave was recovered by the module. The same third-
order Butterworth filter ( f cut −off = 100 kHz) used in the previous test was used to filter

the demodulated signal.

118
 Figure 91. Demodulated sine wave with 50-kHz frequency.

The same tests utilizing the 50-kHz square and ramp waves were repeated
for the tangent-type demodulation module. The outputs obtained from the module are
given in Figure 92 and Figure 93. It is seen that the waveforms were not recovered
accurately. However, in both situations, the fundamental frequencies of the input signals
could be retrieved successfully. As seen from Figure 92, the square wave has distortions
at the amplitude peaks. Initially, these were attributed to the attenuation of the high-
frequency components by the filter. Although the cut-off frequency was changed and
several experiments were done, the module failed to provide an accurate square wave.
Similarly with the ramp wave, there were distortions on the signal recovered as well. The
recovered waveform looks like a sine wave rather than a ramp wave at the peaks and
lows. This distortion is considered as caused by the filter.

119
Figure 92. Demodulated square wave obtained from the
tangent-type demodulation module.

Figure 93. Demodulated ramp wave obtained from the

tangent-type demodulation module.

120
 In the next test, the real-time streaming video signal was used as the
modulating signal. The filter cut-off frequency was adjusted to 9 MHz since the video
signal has a 6-MHz bandwidth. Similar to the previous test with video signals, a variable
signal attenuator was placed between the video camera and the synthesizer to prevent
over-modulation. Then attenuation of 3 dB was applied to the output of the video
camera. A snapshot of the module while running is shown in Figure 94. An expanded
version of the demodulated video signal obtained is also shown in Figure 95. It was seen
that the demodulated video signal obtained had excessive level of distortions. The
distortions corrupted the sync pulses and the NTSC decoding module failed to decode
and display the video.

Figure 94. Snapshot of the tangent-type demodulation module

with video input (fcut-off = 9 MHz).

121
 Figure 95. Demodulated video signal obtained from the tangent-type
demodulation module (fcut-off = 9 MHz).

Similar to the previous situation encountered with the baseband

demodulation module, the filter cut-off frequency was decreased down to 100 kHz to
reduce the distortions and the test was repeated. The resulting demodulated signal is
shown in Figure 96 and Figure 97 (zoomed).

Figure 96. Demodulated video signal obtained from the tangent-type

demodulation module (fcut-off = 100 kHz).
122
 Figure 97. Zoomed view of the video signal obtained from the tangent-type
demodulation module (fcut-off = 100 kHz).

It was observed that, once again, the noise problem affected the results
negatively even though the filter cut-off frequency was set very low. The demodulated
video signal could not be decoded and displayed by the NTSC decoding module.

C. SUMMARY OF RESULTS

In Section A of this chapter, the software simulations needed to validate

application of the FM demodulation modules introduced in Chapter V were performed.
The results showed that both the baseband and tangent-type demodulation techniques can
be used for the demodulation of FM video signals.

After the validation of the demodulation methods, in Section B the verification of

the functionality of the software modules developed based on the simulations was given.
It was seen that both modules showed similar performances. Both modules were able to
retrieve the sinusoidal signal used as the modulating signal. However, distortions were
observed when square and ramp waves were used. On the other hand, the fundamental
frequencies of the square and ramp waves were retrieved successfully. When a real-time
streaming video signal was used as the modulating signal, the modules were able to
123
provide the demodulated video signal back at the output. However, the signals obtained
could not be decoded by the NTSC decoding module. In Section C.3 of Chapter V, it was
mentioned that the video decoder measures the duration of the sync pulses in the video
signal and compares this duration with the minimum duration required for a valid sync
pulse. If the measured duration is not greater than this minimum pulse duration, the pulse
is considered as noise. Because of the excessive noise on the demodulated signal, the
video decoder failed to validate the pulses as sync pulses since it cannot measure a pulse
duration that is greater than the minimum duration for a valid sync pulse. Hence the
pulses were considered as noise and as a result the video could not be displayed.

A slight improvement was achieved by decreasing the low-cut-off frequency of

the filter used in the module; however the video stream could not still be displayed. The
fact that the baseband video signals can be frequency modulated and successfully
recovered indicates that the demodulation software is operating as expected. There are
several sources of error that could be causing the poor quality of the signal when
hardware is used:
1. The quality of the FM signal out of the signal generator is unknown.
Distortions may have occurred before demodulation.
2. The signal generator may be applying pre-emphasis increasing the
magnitude of the high-frequency components in order to improve the
overall SNR.
3. The demodulator noise is too high and more samples/pixels may need to
be collected.
4. Frequency offset between the incoming RF and LO may cause problems.
Simulations of this effect in MATLAB and LabView have indicated that
the demodulation process is very tolerant to a shift in frequency (up to
200 kHz).
In Chapter VII, a summary of the overall research presented in this thesis is given
and conclusions are made, along with recommendations for future work.

124
 VII. CONCLUSIONS AND RECOMMENDATIONS

In this chapter, a summary of the research presented in this thesis is given and
conclusions, along with future recommendations, are made.

A. SUMMARY AND CONCLUSIONS

UAVs play an important role in modern military warfare. They provide a high
degree of situational awareness by carrying out surveillance and reconnaissance missions
that lead to gaining air dominance against adversaries. The ability to gain air dominance
is crucial in the sense that it saves lives and affects the length of the conflict. In order for
a UAV to accomplish its surveillance and reconnaissance mission, the data link between
the UAV and the GCS must remain uninterrupted and provide a reliable information
transfer. An antenna with tracking capability is a good solution to sustain this datalink.

The ultimate objective of the ongoing project presented in this thesis is to develop
a digital tracking antenna array for use in communications with a UAV, using COTS
components. The research previously done in key areas included RSNS and DBF for
direction finding and tracking as well as demodulation and decoding techniques for
video data transmission. The research presented in this thesis focused on the
development of the demodulation techniques in order to allow the digital tracking array
to successfully retrieve and display the video data embedded in the FM signals received
by the antenna. An in-depth review of the properties and functionalities of the current
system and its components was also covered in order to provide a useful reference for
future studies.

The feasibility of utilizing new demodulation techniques for the demodulation of

the video signals was investigated. The tangent-type demodulation technique was studied
as a new approach for the demodulation process. The previous baseband FM
demodulation technique was enhanced to handle video signals. The applicability of these
two techniques in the receiver system was validated by computer simulations done in
125
MATLAB and LabView. It was seen that FM video signals received from a UAV can be
demodulated by using both the baseband and tangent-type demodulation techniques.

The modular approach used in the previous studies for the development of the
software modules for the receiver processor was maintained. Based on the results of the
simulations, an FM demodulation module utilizing the tangent-type demodulation
technique for the demodulation of FM video signals was developed. Additionally, the
previous FM demodulation module using the baseband demodulation technique with a
sinusoidal input waveform was modified to handle video signals. Both modules were
operated on a single channel. Subsequently, both modules were tested using a bench-top
test set up. Different waveforms were used as input modulating signals. These included
sine, square and ramp waves and real-time streaming video signal. The newly developed
tangent-type FM demodulation module showed similar performance compared to the
previous baseband demodulation module. Both modules were able to demodulate the
sine wave. The demodulated square and ramp waves by both modules had distortions
due to the attenuation of high-frequency components. However, the fundamental
frequencies of both waves were retrieved. The video signal used as the modulating signal
was recovered using both demodulators. However, the signals obtained were of poor
quality, which prevented decoding of the video data. The filter cut-off frequencies were
decreased to eliminate the high-frequency noise. A slight improvement was obtained;
however, the sync pulses were still distorted preventing the video data from being
decoded and displayed.

The development process also included the modification of the user interfaces of
the previously developed modules. The previous user interfaces of the NTSC decoding
module and the baseband FM demodulation module were examined and modified to
provide more efficient functionality and analysis. The functionality of the modified
NTSC decoding module was tested using a bench-top test setup. The functionality of the
module was verified using unmodulated streaming video obtained from a video camera

126
as the input. However, when integrated within the demodulation modules, the module
failed to decode and display the demodulated video obtained at the output of the
demodulators.

B. RECOMMENDATIONS FOR FUTURE WORK

In order to reach the ultimate goal of developing a digital tracking array that has
the capability to acquire, track and demodulate the FM video signals coming from a
UAV, research and studies in the following areas are recommended. The
recommendations presented cover not only improving the areas studied in this research,
but also enhancing overall system performance.

1. FM Demodulation Modules

Currently, the video signal obtained from the baseband and tangent-type FM
demodulation modules cannot be decoded by the NTSC decoding module. As a result,
the streaming video cannot be displayed on the screen. The noise problem was
considered to be the reason for this issue. Since the NTSC decoding module functionality
with unmodulated video signals was verified by bench-top tests, emphasis should be
given to improving the quality of the demodulated video signal by eliminating noise and
making it compatible with the NTSC decoding module. Concentration can be focused on
improving the quality of the vertical sync pulses included in the demodulated video
signal. Also the Vsync level computed by the Fetch from 5112 subVI should be the
proper value for the demodulated video signal. The computation is currently done
according to the input video signal. However, the Vsync computed is used in the NTSC
Processor subVI to display the demodulated video. It is possible that the amplitude level
of the demodulated video signal may not match the Vsync level delivered by the Fetch
from 5112 subVI. Future work can include developing a method to calculate the Vsync
value according to the demodulated video signal. Once this is done, the software
modules can easily be updated and integrated into the system.

127
 2. NTSC Decoding Module

The NTSC decoding module is capable of decoding and displaying video data in
black and white. Further studies could focus on modifying the module to have the
capability of processing color information.

3. Quality of the FM Signal

In the hardware tests, the frequency modulation was done using the HP 8341B
Frequency Synthesizer to simulate the data transmission from a UAV. The video signal
out of the camera was fed to 8341B and the frequency modulated signal obtained was fed
to the AD8347 quadrature demodulator. However, the quality of the FM signal provided
by the 8341B was unknown. Future work may include the assessment of the FM signal
quality obtained from the signal generators. Additional tests can be performed using a
different signal generator for frequency modulation in order to compare the quality of the
FM signal. The specifications of the signal generators used should be checked carefully
before performing the hardware tests. The input signal frequency range should be able to
include the video signal bandwidth (i.e., 6 MHz). Similarly, the input signal amplitude
range should cover + / − 2 volts.

4. Other System Enhancements

From an overall system perspective, there are several fields for research and
development that can be explored to extend the capabilities of the digital tracking array
antenna system. The values used for the moduli in the RSNS module can be increased in
order to decrease the bin size, which therefore provides the antenna array the capability
to track the incoming signal with higher accuracy. Another improvement for the tracking
aspect can be achieved by making the monopulse slope constant dynamically change
according to the current scan angle. This provides faster convergence compared to using
a constant value for the slope constant.

128
 As the new COTS equipment with better specifications become available for low
prices in time, the COTS equipment used in the digital tracking array system can be
replaced with new equipment with better performance. For example, the quadrature
demodulators used in the system can be replaced with new demodulators having higher
second-order input intercept point (IIP2) specifications. This can decrease the amount of
second-order distortion and improve the performance of the quadrature demodulators.
Similarly, new and faster A/D digitizers, which have higher resolution, can replace the
current PXI-5112 digitizers to obtain lower noise and better sensitivity. Also, an A/D
digitizer with a higher sampling rate (i.e. twice the carrier frequency) can provide anti-
aliasing of the acquired signals at high frequencies and eliminate the need for a separate
hardware demodulator.

129
THIS PAGE INTENTIONALLY LEFT BLANK

130
 APPENDIX: MATLAB CODES

A. MATLAB CODE FOR THE SIMULATION OF BASEBAND

DEMODULATION USING A 50 KHZ SINE WAVE

% fm_demod_BB_sine_wave.m
%
%**********************************************************************
%This m file is used to simulate Baseband FM demodulation. A sine *
%wave is created as the modulating signal and demodulated using *
%baseband demodulation *
% *
%Original version by: David C. Jenn (Professor ECE Department, NPS) *
%Edited by: Fevzi Aytaç KAYA / 1st LT. / Turkish Air Force / 2010 *
%**********************************************************************

clear
Fs = 1500000; % Sampling rate of signal
dt=1/Fs;
fm=50000; %modulating signal freq.
omegam=2*pi*fm;
dev = 20000; % Frequency deviation in modulated signal
disp(['deviation: ',num2str(dev)])
T=5*1/fm; % total observation time
t0=0;
t = [0:dt:T]'+t0;
Nt=length(t); % Number of time samples
disp(['time step (us): ',num2str(dt/1e-6)])
% single tone message signal
beta=dev/fm;
disp(['beta=',num2str(beta)])

m=cos(omegam*t); % message signal

mi=sin(omegam*t); % integral of message

%Modulated signal
theta=dev*mi.*t;
Noise=wgn(length(theta),2,-200);
I=cos(theta)+Noise(:,1);
Q=sin(theta)+Noise(:,2);

% plot FM modulated BB signal

figure(1), clf
clf
plot(t,I,t,Q)
xlabel('Time, sec')
ylabel('Modulated Signal, volts')

131
title('Modulated signals')
legend('I','Q')
axis([t0,t0+T,-1,1])
thr0=atan2(Q,I+1e-10)+pi; % 0 to 2*pi
thr=unwrap(thr0);

% recover using I and Q

for i=2:Nt
dQdt=(Q(i)-Q(i-1));
dIdt=(I(i)-I(i-1));
Ia=(I(i)+I(i-1))/2;
Qa=(Q(i)+Q(i-1))/2;

% Baseband demodulation formula

mr(i-1)=(Ia*dQdt-Qa*dIdt)/2/pi/t(i);
end

mr=mr/660; %Set this to adjust output amplitude

% plot demodulated signal

figure(2), clf
clf
plot(t,m,'b-',t(1:Nt-1),mr,'r-')
xlabel('Time, sec')
ylabel('Message Signal, volts')
title('Message signals')
legend('Original','Demodulated');
axis([t0,t0+T,-2,2])

%I vs. Q can be plotted by uncommenting the following

% figure(3)
% plot(I,Q,'.');
% title('I vs Q')

132
B. MATLAB CODE FOR DISPLAYING VIDEO FRAMES
% NTSC_video_display.m
% Written by David C. Jenn Professor, NPS (11/23/08)
% uses floating point raw data from 5112
% read NTSC data and process sequentially
% SIMULATING STREAMING DATA
% WORKS for all combinations of field
% time reset added
clear
% setup PXI-5112 data read
fs=20e6;
dt=1/fs/1e-6; % sample time in microseconds
dv=2/2^8; % voltage quantization interval
% load matlab file already read from text file
disp('reading text file ...');
V0=dlmread('TextFile7.txt');
Ns=length(V0); % number of samples read in from file
X=[1:640];
Y=[1:480];
Nlines=259; % starting at first FULL line of video (so not
260)
tpix=52.6/640; % pixel time in microseconds
tline=63.5; % complete line of data
Tmin=1.5; % pulses < Tmin are noise
% initialize counters and flags
nfld=0; % =0 searching for even field; =1 searching for
odd
frame=0; % counts frame number
isy=1; % set isy=0 if a start of field (end of long
pulse)
lnn=0; % counts lines in the current frame
npix=1; % counts pixels in the current line
iproc=1; % =0 processing, =1 searching
% use the first few lines to find sync and blanking level for this file
Vmin=V0(1);
for nn=1:floor(tline/dt)+10
if V0(nn)<Vmin, Vmin=V0(nn); end
end
Vsync=Vmin+10*dv;
disp(['Vsync=',num2str(Vsync)])
% -------------- PROCESS STREAMING DATA -------------------------------
n=1; % sample index of data from 5112 ADC
t=0; % time in microseconds
V=V0(1);
tb=0;
TS=0;
TE=0;
TSo=0;
TEo=0;

133
 Tmin=1;
isy=1;
V=V0(1);
Vmax=max(V0);
while n<Ns-1
n=n+1;
t=t+dt; % new time
tb=tb+dt; % blanking sample counter
Vo=V;
V=V0(n); % new voltage point
if floor(n/50000)==(n/50000), disp(['n=',num2str(n)]), end

if iproc==1
if V<=Vsync & Vo>Vsync % first point below threshold
TSo=TS;
TS=t; % samples below threshold
NS=n;
isy=1; % potential start point
end
if V>Vsync & Vo<=Vsync % step up
TEo=TE;
TE=t;
isy=0; % potential stop point
end
% determine if this is a valid pulse or just a noise spike
if isy==0
dsp=TS-TEo; % width of up spike or sync pulse spacing
if dsp<Tmin
TS=TSo;
isy=1;
end
end
if isy==0
if dsp>=Tmin % valid sync pulse
isy=1;
if TEo-TSo>3.5*tline
disp('new field')
isy=1;
lnn=1;
TEo=TEo-TS;
TSo=TSo-TS;
TE=TE-TS;
t=t-TS; % reset time reference
TS=0;
tb=0;
if TE-TEo<.75*tline
disp('odd')
itype=1;
end
if TE-TEo>=.75*tline
disp('even')
itype=0;
134
 end
if nfld==0 & itype==0, iproc=0; end
if nfld==1 & itype==1, iproc=0; end
end
end
end
end % end of iproc
% blanking level for this frame
if tb>tline/2 & tb<=tline/2+dt
Vb=V; disp(['Vb=',num2str(Vb)]);
Vwht=14/4*(Vb-Vsync)+Vsync;
end
if isy==1 & iproc==0
t1=2*4.7+1+(lnn-1)*63.42; % start time of video data for this
line
t2=t1+52.6; % end time of video data for this line
if (t>=t1 & t<=t2) % time window for line lnn
tp1=t1+(npix-1)*tpix; % start time of pixel number npix
tp2=tp1+tpix;
if t>=tp1 & t<=tp2 % in pixel npix
npix=npix+1;
Vp=V; % 0<=V<=1 black to white
if itype==1, P(2*lnn-1,npix)=Vp/Vmax*1.5; end
if itype==0, P(2*lnn,npix)=Vp/Vmax*1.5; end
end
if npix==640 % finished with this line
npix=1; % restart pixel counter
lnn=lnn+1;
end
end
if lnn==Nlines % finished with this field
lnn=1;
isy=1;
nfld=nfld+1;
iproc=1; % look for start of next field
end
% if an odd field was processed then a complete frame is collected so
plot
if nfld==2
disp(['completed frame number: ',num2str(frame)])
frame=frame+1;
figure(frame), clf
imagesc(X,Y,abs(P(12:516,1:640)));
colormap(gray);
axis off
title(['Frame number: ',num2str(frame)])
nfld=0;
end
end
end % end data loop
disp('OUT OF DATA BEFORE END OF FIELD PROCESSED -- INCOMPLETE FRAME')

135
THIS PAGE INTENTIONALLY LEFT BLANK

136
C. MATLAB CODE FOR THE SIMULATION OF BASEBAND
DEMODULATION OF A VIDEO SIGNAL

% fm_demod_BB_video_data.m
%
%**********************************************************************
%This m file is used to simulate Baseband FM demodulation. *
%A video data file (in txt format) is required for the input. *
% *
%Written by: Fevzi Aytaç KAYA / 1st LT. / Turkish Air Force / 2010
*%*********************************************************************
*

tic
clear
clc
%Amplification set here
volume_knob=1; % Range 1-30 (suggested)
%Read modulating signal data from text file
AA=dlmread('video_data.txt'); %File name is video_data.txt
n1=1; n2=length(AA);
V=AA(n1:n2);
clear AA
t=1e-4*[0:length(V)-1]';
Vmax=max(V); Vmin=min(V); V0=max([Vmax,abs(Vmin)]);
X=V/V0; %normalize
M=3; %interpolation factor
%Y=interp(X,M); % If doing interpolation, M times the sampling rate
Y=X; % If not doing interpolation

%Plot input message signal

figure(1)
plot(t,X)
xlabel('time, sec')
ylabel('Amplitude')
title('Modulating signal')

% FM modulation parameters at BB
Fs=10/1e-4; % Sampling rate of signal
dt=1/Fs;
Fdev = 10; % Frequency deviation in modulated signal
disp(['deviation: ',num2str(Fdev)])
t=(1e-4)/M*[0:length(Y)-1]';
t0=0;
Nt=length(t); % Number of time samples
disp(['time step (us): ',num2str(dt/1e-6)])
%Integration
YY=[];

137
YY=200*cumtrapz(t,Y);
%Modulation
theta=[];
theta=Fdev*YY.*t;
%Construct I and Q
I=[];
Q=[];
Noise=wgn(length(theta),2,-250);
I=cos(theta);+Noise(:,1);
Q=sin(theta);+Noise(:,2);
% plot FM modulated BB signal
figure(2), clf
clf
plot(t,I)
title('I')
figure(4)
plot(t,Q)
title('Q')

% recover using I and Q

mr=[];
for i=2:Nt
dQdt=(Q(i)-Q(i-1));
dIdt=(I(i)-I(i-1));
Ia=(I(i)+I(i-1))/2;
Qa=(Q(i)+Q(i-1))/2;
% Baseband demodulation formula
mr(i-1)=(Ia*dQdt-Qa*dIdt)/2/pi/t(i);
end
% plot demodulated signal
figure(3)
plot(t(1:Nt-1),mr,'r-')
title('Output after demodulation')
%uncomment below if amplification is required
%and set amplfication factor (volume knob)
% amplified=(mr)*volume_knob;
% for m=1:size(amplified,2)
% if(amplified(m)>1)
% amplified(m)=1;
% else if (amplified(m)<-1)
% amplified(m)=-1;
% end
% end
% end
%Plot I vs. Q
figure(4)
plot(I,Q,'.');
title('I vs Q')
axis([-1.1 1.1 -1.1 1.1])
grid on
toc

138
D. MATLAB CODE FOR THE SIMULATION OF TANGENT-TYPE
DEMODULATION OF A SINUSOIDAL WAVEFORM

% fm_demod_tangent_sine.m
%
%**********************************************************************
%This m file is used to simulate tangent type FM demodulation. *
%A sine wave is created as the modulating signal and demodulated *
%using baseband demodulation *
% *
%Written by: Fevzi Aytaç KAYA / 1st LT. / Turkish Air Force / 2010 *
%**********************************************************************

clear
clc
clf

m=pi/2e4; %Time interval

amp=1; %Amplitude
fc=1500000; %carrier freq
fm=50000; %modulating freq
df=10000; %freq. deviation

beta=df/fm; %Modulation Index

LPF1_order=3; %Order of LPF before atan2 block
fs=5*(fm+fc); %Sampling Freq.
s=1/fs; % Sampling period
t=0:s:m; % Time vector

%Construct message signal and integrate

message=beta*(cos(2*pi*fm*t));
theta=beta*(sin(2*pi*fm*t));

%Modulate the integrated signal

x=fmmod(theta,fc,fs,df);

%Uncomment below to plot I and Q

% figure(1)
% plot(t,x)
% title('FM Signal');

%Mix modulated signal with sinusoids I_mix and Q_mix

I_mix=sin(2*pi*fc*t);
Q_mix=cos(2*pi*fc*t);
I_lpf=x.*I_mix;
Q_lpf=x.*Q_mix;

139
%----------------------------LPF---------------------------------------
% Butterworth Lowpass filter designed using the BUTTER function.
% All frequency values are in Hz.
Fs = fs; % Sampling Frequency
N = LPF1_order; % Order
Fc = .4*fm; % Cutoff Frequency
% Calculate the zpk values using the BUTTER function.
[z, p, k] = butter(N, Fc/(Fs/2));
% To avoid round-off errors, do not use the transfer function. Instead
% get the zpk representation and convert it to second-order sections.
[sos_var,g] = zp2sos(z, p, k);
Hd = dfilt.df2sos(sos_var, g);
%----------------------------------------------------------------------

%Filter I and Q before arctangent block

I=filter(Hd,I_lpf);
Q=filter(Hd,Q_lpf);

%arctangent and differentiate

phi=atan2(Q,I);
S1=diff(phi);

%Set time vector (same number of elements)

t2=t;
t2(:,1)=[];

%Plot the output of the differentiator

% figure(3)
% plot(t2,S1)
% title('S1(k)');

%Set first 10 samples to zero to eliminate unwanted components

for i=1:10
S1(i)=0;
t2(:,i)=0;
end

%Set amplification to match modulating signal

amplitude_adjust=1930;

%Plot the amplified demodulated signal

figure(4)
plot(t2,amplitude_adjust*S1,t,message,'r')
legend('Demodulated','Original');
title('Demodulated Signal vs. Modulating Signal');
xlabel('Time (sec)');
ylabel('Amplitude');

140
E. MATLAB CODE FOR THE SIMULATION OF TANGENT-TYPE
DEMODULATION OF A VIDEO SIGNAL

% fm_demod_tangent_video.m
%
%**********************************************************************
%This m file is used to simulate tangent type FM demodulation. *
%A text file including video data captured by a video camera is *
%reuired. The video data is modulated and demodulated back by using *
%tangent type demodulation. *
% *
%Written by: Fevzi Aytaç KAYA / 1st LT. / Turkish Air Force / 2010 *
%**********************************************************************

clear
clc
tic

%Amplification at the output is set here

volume_knob=16.1; % Range 1-30 (suggested)
do_improvement=1; %Set to 1 if improvement is desired

AA=dlmread('TextFile7.txt'); %Read video data in TextFile.txt

n1=1; n2=length(AA);
V=AA(n1:n2);
clear AA

t=1e-4*[0:length(V)-1]';
Vmax=max(V); Vmin=min(V); V0=max([Vmax,abs(Vmin)]);
X=V/V0; %Normalization

% Y=X;
figure(1)
plot(t,X)
xlabel('time, sec')
ylabel('Amplitude')
title('Speech waveform from LabVIEW')

% FM signal
fc=20000; %carrier freq
fs=10/1e-4; %Sampling freq
Fdev=1000; %freq. deviation
t=(1e-4)/3*[0:length(X)-1]';

%Modulate
modulated=fmmod(X,fc,fs,Fdev);

141
%plot modulated signal
figure(2)
plot(t,modulated)
title('FM Signal');

%----------------Add-noise---------------------------------------------
%I_mod = awgn(x,30,'measured'); % Add white Gaussian noise.
%----------------------------------------------------------------------

%Mix with sinusoids

I_mix=sin(2*pi*fc*t);
Q_mix=cos(2*pi*fc*t);
I_lpf=modulated.*I_mix;
Q_lpf=modulated.*Q_mix;

%----------------------------LPF---------------------------------------
% Butterworth Lowpass filter designed using the BUTTER function.
% All frequency values are in Hz.
Fs = fs; % Sampling Frequency
N = 3; % Order
Fc = fc; % Cutoff Frequency
% Calculate the zpk values using the BUTTER function.
[z, p, k] = butter(N, Fc/(Fs/2));
% To avoid round-off errors, do not use the transfer function. Instead
% get the zpk representation and convert it to second-order sections.
[sos_var,g] = zp2sos(z, p, k);
Hd = dfilt.df2sos(sos_var, g);
%----------------------------------------------------------------------
%Filter I and Q signals
I=filter(Hd,I_lpf);
Q=filter(Hd,Q_lpf);

%Tangent type demodulation and differentiation

phi=atan2(Q,I);
S1=diff(phi);

%Start improved tan-type demodulator

%Add C1(k)
if do_improvement==1
C1=zeros(size(S1),1);
for j=2:size(S1)
if S1(j)<0
C1(j)=2*pi;
end
end
S2=S1+C1;
End
%-------------------End_of_Improvement--------------------------------

142
if do_improvement==1
demodulated=S2;
else
demodulated=S1;
end

t2=t;
t2(size(t,1))=[];

%Amplification is done here

amplified=(demodulated-0.83775)*volume_knob;

for m=1:size(amplified)
if(amplified(m)>1)
amplified(m)=1;
else if (amplified(m)<-1)
amplified(m)=-1;
end
end
end

amplified=amplified.*(-1);

figure(3)
%plot(t,X,t2,amplified,'r') %Compare with this
plot(t2,amplified,'r') %Output only
xlabel('Time (sec)');
ylabel('Amplitude');
title('Demodulated video signal');

%Plot I and Q circle

figure(4)
plot(I,Q,'.');
title('I vs Q')
xlabel('I');
ylabel('Q');
toc

143
THIS PAGE INTENTIONALLY LEFT BLANK

144
 LIST OF REFERENCES

[1] C. A. Hughes, “Achieving and enduring air dominance,” Research Report, Air
Command and Staff College, Air University, Maxwell Air Force Base, Alabama,
April 1998.

[2] C. Beal, J. Boatman, and M. Hewish, “UAV development: The art of

compromise,” Jane’s International Defense Review, vol. 26, issue 5, May 1,
1993.

[3] “Unmanned aerial vehicle,” July 5, 2010. [Online].

Available: http://en.wikipedia.org/wiki/Unmanned_aerial_vehicle. [Accessed July
11, 2010].

[4] P. E. Pace, Detecting and Classifying Low Probability of Intercept Radar. 2nd
Ed., Norwood, MA: Artech House, pp. 323-325, 2009.

[5] Photo by Photographer’s Mate 2nd Class Daniel J. McLain, “RQ-2B Pioneer
UAV,” December 22, 2006. [Online]. Available:
http://upload.wikimedia.org/wikipedia/commons/1/18/RQ-2B_pioneer_uav.jpg,
[Accessed July 11, 2010].

[6] D. C. Jenn, Y. Loke, T. C. H. Matthew, Y. E. Choon, O. S. Siang, and Y. S. Yam,

“Distributed phased arrays and wireless beamforming networks,” International
Journal of Distributed Sensor Networks, pp. 283–302, 2009.

[7] J. A. Bartee, “Genetic algorithms as a tool for phased array radar design,”
Master’s Thesis, Naval Postgraduate School, Monterey, California, June 2002.

[8] C. S. Ong, “Digital phased array architectures for radar and communications
based on off-the-shelf wireless technologies,” Master’s Thesis, Naval
Postgraduate School, Monterey, California, December 2004.

[9] G. M. Burgstaller, “Wirelessly networked digital phased array,” Master’s Thesis,

Naval Postgraduate School, Monterey, California, September 2006.

[10] B. L. Gezer, “Multi-beam digital antenna for radar, communications, and UAV
tracking based on off-the-shelf wireless technologies,” Master’s Thesis, Naval
Postgraduate School, Monterey, California, September 2006.

145
[11] A. Lee, “Variable resolution direction finding using the robust symmetrical
number system,” Master’s Thesis, Naval Postgraduate School, Monterey,
California, December 2006.

[12] J. Benveniste, “Design and development of a single channel RSNS direction

finder,” Master’s Thesis, Naval Postgraduate School, Monterey, California,
March 2009.

[13] I. L. Tan and D. J. Pandya, “UAV digital tracking array design, development and
testing,” Master’s Thesis, Naval Postgraduate School, Monterey, California,
December 2009.

[14] M. Skolnik, “Introduction to Radar Systems,” 3rd Ed., New York: McGraw-Hill,
New York, pp. 210–229, 2001.

[15] Photo by Hamish Lindsay, “The FPQ Skin Tracking Radar,” 22 October 2004.
[Online]. Available:
http://www.honeysucklecreek.net/images/other_stations/croq6dsh1.jpg,
[Accessed July 18, 2010].

[16] D. C. Jenn, “Microwave devices & radar,” vol. II, pp. 111–115, 2009
(unpublished lecture notes).

[17] K. W. Kwai, P. E. Pace, D. C. Jenn, and J. C. Chen, “Robust symmetrical number

system direction finding arrays with virtual spacing,” Naval Postgraduate School,
Monterey, California, (unpublished notes).

[18] D. C. Jenn, “RSNS processing using a single channel,” Naval Postgraduate

School, Monterey, California, (unpublished notes).

[19] D. C. Jenn, “Digital antennas,” Naval Postgraduate School, Monterey, California,

(unpublished notes).

[20] S. J. Roome, “Analysis of quadrature detectors using complex envelope

notation,” IEEE Proceedings, vol. 136, Pt. F, no. 2, April 1989.

[21] B. Razavi, “Design considerations for direct conversion receivers,” Circuits and
Systems II: Analog and Digital Signal Processing, IEEE Transactions on 44(6),
pp. 428–435, 1997.

[22] National Instruments, “Anatomy of a video signal,” December 6, 2006. [Online].

Available: http://zone.ni.com/devzone/cda/tut/p/id/3020, [Accessed July 14, 2010].

146
[23] “NTSC scanning, timing/sync, sync recovery, numbers.” [Online].
Available: http://www.ntsc-tv.com/. [Accessed July 14, 2010].

[24] W. Tomasi, Electronic Communication Systems: Fundamentals Through

Advanced, 5th Ed., pp. 254–282, Upper Saddle River, New Jersey: Pearson
Prentice Hall, 2004.

[25] K. M. E. Hui, “Digital tracking array for FM signals based on off-the-shelf

wireless technologies,” Master’s Thesis, Naval Postgraduate School, Monterey,
California, December 2007.

[26] M. Hagiwara and M. Nakagawa, “Performance of improved tangent-type FM

demodulator using digital signal processing,” Electronics and Communications in
Japan, vol. 67-B, no.10, 1984.

[27] F. Shadt, F. Mohr and M. Holzer, “FM demodulation of IQ baseband signals

using Kalman filters,” Radioelektronika, 18th International Conference, 2008.

[28] W. L. Stutzman and G. A. Thiele, “Antenna Theory and Design,” 2nd Ed.,
Hoboken, New Jersey: John Wiley & Sons, Inc., p. 173, 1998.

[29] D. C. Jenn, “Microwave devices & radar,” vol.II, pp. 9–10, 2009 (lecture notes).

[30] C. H. Eng, “Design and development of an automated demodulator calibration

station,” Master’s Thesis, Naval Postgraduate School, Monterey, California,
December 2009.

[31] Analog Devices, “Direct Conversion Quadrature Demodulator Datasheet,” 2005.

[Online]. Available:
http://www.analog.com/static/imported-files/data_sheets/AD8347.pdf.
[Accessed July 21, 2010].

[32] K. Forsberg, H. Mooz and H. Cotterman, “Visualizing Project Management,” 3rd

Ed., Hoboken, New Jersey: John Wiley & Sons, Inc., p. 361, 2005.

147
THIS PAGE INTENTIONALLY LEFT BLANK

148
 INITIAL DISTRIBUTION LIST

1. Defense Technical Information Center

Ft. Belvoir, Virginia

2. Dudley Knox Library

Naval Postgraduate School
Monterey, California

3. Prof. Dan Boger

Department of Information Sciences
Naval Postgraduate School
Monterey, California

4. Prof. Clark Robertson

Department of Electrical and Computer
Naval Postgraduate School
Monterey, California

5. Prof. David Jenn

Department of Electrical and Computer
Naval Postgraduate School
Monterey, California

6. Lt. Col. Terry Smith

Department of Information Sciences
Naval Postgraduate School
Monterey, California

7. Robert D. Broadston
Department of Electrical and Computer
Naval Postgraduate School
Monterey, California

8. Prof. Kevin Jones

Department of Aeronautics and Astronautics
Naval Postgraduate School
Monterey, California

9. 1st Lt. Fevzi Aytaç Kaya

Turkish Air Force Electronic Warfare Support Center Command, Ankara, Turkey
149