+ All Categories
Home > Documents > Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue...

Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue...

Date post: 23-Mar-2020
Category:
Upload: others
View: 2 times
Download: 0 times
Share this document with a friend
128
Acousto-Optic Processors for the Detection of Spread Spectrum Radar Signals Andrew M. Kiruluta A thesis submitted in conforrnity with the requirements for the degree of Doctor of Philosopliy Gradiiate Department of Electrical Sr. Conipiiter Engineering. University of Toronto. @Copyright by Andrew M. Kimhta 199%
Transcript
Page 1: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Acousto-Optic Processors for the Detection of Spread

Spectrum Radar Signals

Andrew M. Kiruluta

A thesis submitted in conforrnity with the requirements

for the degree of Doctor of Philosopliy

Gradiiate Department of Electrical Sr. Conipiiter Engineering.

University of Toronto.

@Copyright by Andrew M. Kimhta 199%

Page 2: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

National Library I * m of Canada Bibliothèque nationale du Canada

Acquisitions and Acquisitions et Bibliogmphic Services services bibliographiques

395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada

The author has granted a non- exclusive licence dowing the National Library of Canada to reproduce, loan, distribute or seil copies of this thesis in microform, paper or electronic formats.

The author retains ownership of the copyright in this thesis. Neither the thesis nor substantial extracts fiom it may be printed or otheMrise reproduced without the author' s permission.

L'auteur a accordé une licence non exclusive permettant a la Bibliothèque nationale du Canada de reproduire, prêter, distribuer ou vendre des copies de cette thèse sous la fome de microfiche/film, de reproduction sur papier ou sur format électronique.

L'auteur conserve la propriété du droit d'auteur qui protège cette thèse. Ni la thèse ni des extraits substantiels de celle-ci ne doivent être imprimés ou autrement reproduits sans son autorisation.

Page 3: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

ACKNOWLEDGMENTS 1 woulcl like to express my heartfelt gratitude to my supervisor Dr. Y. Ristic for

his inspiring guidance arid n-hose advice that - o u cannot do engineering op tics without

knowing electronics ivell- provecl to be in~altiable. The sound advice from Dr. J . P. E-. Lee

of Defense Research Establishment Ottawa is very much appreciated. -4s well. 1 am deeply

indebtecl t O 111~- CO-super-isor Dr. -4. Iéne t sanopoulos for his encouragement ancl support.

Special thanks to my friends and colleagues ivho macle my gracluate esperience at the

L-niversity of Toronto a mernorable one. The early inspiration of m - mentors Dr. R. Don-

nelly. Dr. J . Quaicoe ancl Dr. R. Langford. al1 of Mernorial University of Xewfoundlancl.

who t aiight me. in t heir own unique way t hat the long and seeniingly chaotic pa t h to learn-

ing is through critical probing and a keen sense of obser\ation. is immenselu acknowledged.

This research effort {vas jointly sponsorcd by the Defense Research Establishment

Ottawa (DREO ). The Satural Sciences ancl Engineering Research Council of Canada

( SSERC ) . the Ontario Informat ion Technolog- Research Ceuter (ITRC ) ancl the .~coiisto-

Optic Research Laboratory (.AORL) at the University of Toronto.

Page 4: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Acousto-Optic Processors for the Detection of Spread Spectrum Radar Signals

Andrew M. Kiruluta, PhD 1997

Department of Electrical & Computer Engineering University of Toronto

Abstract Xewer radar systems using such techniques as direct-seqiience phase modulation. fre-

quency hopping and linear FM (chirp) require large transmission bandwidth and increas-

ingly sophis t icatecl receivers t O int ercept and classify t heir ret urns. These signals are

ctifficult to distinguish from background noise and are hence said to have low probability

of intercept (LPI). Current electronic receiver systems work ireil at cletecting and cbarac-

terizing narrowband pulse signals but are not very effective wit h spread spectrum signals.

.-\cousto-optic (.\O) processors have shon-n great potential for dealing with tvideband sig-

n a l ~ ancl offer the ability to detect and analyze \-aarioiis LPI wavefornis. In this thesis ive

present a theoretical. numerical and an esperimental stody of a t ime integrat ing acouste

op t ic processor. with an electronically inserted reference tone. to detect and characterize

linear FM and freqiiency hopped spread spectriini LPI signals corrupted hy additive noise

and narrowband interferers. Electronic and opticaI components w r e designed ancl the

processor assembled. -4nalytical espressions for chirp and freqiiency hopped signal corre-

lat ions wit h the processor were t hen derit-ed and niimerically simula ted. The processing

gain of the processor iras subseqiiently tlerived using a stochastic analysis approach for a

niirnber of SXR scenarios. It is shown that for relativelu low SSR intercept signals. the

laser noise is de-emphasized leading to a processing gain that is proportional to the square

of the detector dynamic range. .At high S'iR levels. the laser intensity noise becomes the

predominant factor. Experiment al e~daluat ion of the noise loading and effec t of the narrotv-

band interferers on the processor output w r e then carried out. -4 near real-time niethocl

based on digitally tunable not ch filters nas developed to excise the narrowbaod interferer

energy prior to correlation. The method uses a unique space-integrating electronically in-

serted reference tone arrangement. Finally. a novel way of estimating the overall optical

misalignment in the processor hardware \vas proposed and dernonstratecl. It represents a

simple and robust alternat ive t O many purely opt ical procedures. The resolution accuracy

of the rnethod is shown to be limited by the CCD pixel dimensions. .As a side result. the

scaling law that applies to the output signal when using CCD detection \ras also derived.

Page 5: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Contents

1 Introduction 1

1 . I Transmit tance Function of an A0 ce11 . . . . . . . . . . . . . . . . . . . . . 6

1 ThesisOutline . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 10

2 Theoretical Studies 12

2.1 Frecpency Selection Constraints for the Ilultiplicative ( Iu-Line) .Architecture 13

2.2 A0 Correlator with Linear FU Inputs . . . . . . . . . . . . . . . . . . . . 20

. . . . . . . . . . . . . . . . . . 2.3 A0 Correlator n-ith Frequency Hop Inputs 25

2.4 Xiinierical Xnalysis for the Two Tjwpes of Inputs . . . . . . . . . . . . . . . 26

2.5 Spacc-Integrarecl Output of the Correlator . . . . . . . . . . . . . . . . . . 2S

2.6 Effect of Soise on the Intensity and Correlation F~inctions . . . . . . . . . 35

2.7 Output SSR and Processing Gain in the Presence of Laser Soise . . . . . . 39

3 Experimental Investigation 50

3.1 Detection of Signals at the Spatially Integrated Output . . . . . . . . . . . 52

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.3 CCD Scaling Law 53

3.3 Correlation of Sarrowband Signals in White Gaussian Soise . . . . . . . . 54

Page 6: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

3.4 Optical Misalignment Estimation . . - . . . . . . . . - . . . . . . . . . . . 5s

3.5 The 50 MHz Signal Bandwidth A 0 Processor . . . . . . . . . . . . . . . . 63

3.6 Methods for SIS Detection and Remot-al . . . . . . . . . . . . . . . . . . . 64

3.7 Simulation of the LPI Signal . . . . . . . . . . . . . . . . . . . . . . . . . 6s

3.7.1 Linear F M Chirp . . . . . . . . . . . . . . . . . . . . . . . . . . . . 6s

3.7.2 Frequency Hop . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 71

3.8 LPI Correla t ion De tect ion Circiii t Recluirenients . . . . . . . . . . . . . . . Y3

3.9 Processing Gain and Detecror Integration . . . . . . . . . . . . . . . . . . . 74

-" 3.10 Correlation of LPI S ipa l s in Soise . . . . . . . . . . . . . . . . . . . . . . J .>

3.11 SIS Escision Csing Tunable Filters . . . . . . . . . . . . . . . . . . . . . . YS

3.12 Impact of the SIS Sotching Feedback Loop . . . . . . . . . . . . . . . . . S3

4 Processor and Measurement Hardware 88

5 Conclusions 106

A Imaging Optics 109

Page 7: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

List of Figures

1.1 Scheniatic of radar LPI correlator . . . . . . . . . . . . . . . . . . . . . . . -4

. 1.2 Bragg ce11 . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 1

2.1 1-D multiplicative architecture . . . . . . . . . . . . . . . . . . . . . . . . . 15

2.2 Spectra grouping of the lariotis processor outputs . . . . . . . . . . . . . . . 1s

2.3 Frequency selections and spectriirn placement of LPI signals . . . . . . . . . 19

2.4 LPI signatures S ( t ) and F ( t ) are detected by two widely separatecl antennas . 21

2.5 Soise loading on the correlation for ~ar ious SSR ~al i ies . . . . . . . . . . . . 27

2.6 Effect of varying the chirp duration (Tc ) on the correlation . . . . . . . . . . 2S

2 . 7 Space integrating pin detector arrangement . . . . . . . . . . . . . . . . . . 29

2.S Simulation outputs for terms ( 1)-(4) respective15 . . . . . . . . . . . . . . 33

2.9 Overall simulation output of the space-integrating processor . . . . . . . . 34

2.10 Time integratiug A 0 intcrcept receiver . . . . . . . . . . . . . . . . . . . . . 40

2.11 For the case n-hen S-W2 + x . . . . . . . . . . . . . . . . . . . . . . . . . 4s

2.12 Output SSR for S.VR, . SXR2 >> 1 . . . . . . . . . . . . . . . . . . . . . 49

3.1 Reference tone .\O processor for radar signai processing . . . . . . . . . . . 51

3.2 A correlation peak cnvelope detected by the CCD array . . . . . . . . . . . 53

Page 8: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

3.3 Correlatoroutpiit f o r r=1 .2p . i . P R F = O . l . V H z . . . . . . . . . . . . . . 55

3.4 Sliiltiple Correlation for P R F = 1 .\TH= . . . . . . . . . . . . . . . . . . . . 56

- - 3.5 I n a l l c a s e s t k e P R F = S O I<Hz.O.5l~ ' /DIl~and t in iebaseof20/ i s /DII ' .

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 3.6 Soise only figure 62

3.7 Effect of SIS on correlation . . . . . . . . . . . . . . . . . . . . . . . . . . 65

3.S Overall optic and electronic hardware la>.oiit . . . . . . . . . . . . . . . . . . 67

3 -9 Chirp generating circuit configuration . . . . . . . . . . . . . . . . . . . . . 69

3.10 Prac tical relationship of frequencies . . . . . . . . . . . . . . . . . . . . . . . 70

3.11 Frequency Hopped LPI Generation . . . . . . . . . . . . . . . . . . . . . . 12

3.12 Processor output for a linear FM signal . . . . . . . . . . . . . . . . . . . . 76

-- 3.13 Effect of varying the chirp duration (Tc) . . . . . . . . . . . . . . . . . . . . t I

-- 3.14 Oscilloscope oittpiits shon-ing the iclentification of a chirp . . . . . . . . . . . t I

3.15 Corresponcling oscilloscope outpiits for the frccluency hopper . . . . . . . . . ï S

3.16 ( a ) in-band ancl ( b ) out-of-band SIS locations relative to the LPI banclidth . 79

3.17 Real Tinie SIS Excision Module . . . . . . . . . . . . . . . . . . . . . . . . S1

3.18 Correlation wi t h in-band interferers . . . . . . . . . . . . . . . . . . . . . . . S4

3.19 Effect of notchiog in-band interferers . . . . . . . . . . . . . . . . . . . . . . S4

3.20 Correlat ion with interferers . . . . . . . . . . . . . . . . . . . . . . . . . . . S5

3.21 Sotching of interferers . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . S5

3.22 Correlation with o.itt.of .band interferers . . . . . . . . . . . . . . . . . . . . . S6

3.23 'iotching of ozrt-of-band interferers . . . . . . . . . . . . . . . . . . . . . . . S6

3.24 RAIS error in approximating the peak of the correlation envelope . . . . . . S7

4.1 Laser diode source SDL-2432-Hl from Spectra Labs . . . . . . . . . . . . . . S9

Page 9: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

. . . . . . . 4.9 Another \.iew of the laser heat sink and anarnorphic prisni pair 90

. . . . . . . . . . . . . . . . . . . . . . . 4.3 i ï e w of the fiber couplecl detector 91

. . . . . . . . . . . . . . . . . . . 4.1 Another vie\\- of the fiber coupleri detector 92

. . . . . . . . . 4 \ ïew of the spatial filter precedecl hy a Sewport Fourier lens 93

. . . . . . . . . . . 4.6 Expancled view of spatial filtering and detection sections 94

. . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . . 4.7 .-\ iisiliary elec t ronics 9.5

4.8 -4nother vietv of the tunable filter/DR controller modtile . . . . . . . . . . . 96

. . . . . . . . . . . . . . . . . . . . . . 4.9 Another vie\\* of the overall processor 97

. . . . . . . . . . . 4.10 A magnified ïiew of the single diode tap-off arrangement 98

. . . . . . . 4.11 Cornputer connection to the feedback arrangement via a GPIB 99

. . . . . . . . . . . . . . . . . . . . . 4.12 Test ancl ?ileasiirement eqiiipment rack 100

. . . . . . . . . . . . . . . . . . . . . . . . . 4.13 Lïen* of the 50 J I H z processor 101

. . . . . . . . . . . . . . . 4.14 Another estendecl view of the 30 .\[Hz processor 102

. . . . . . . . . . . . . 4.15 Side perspective of the entire 50 .1 [ H z LPI processor 103

. . . . . . . . . . . . . . . . . 4.16 Slagnified side view of 50 .\ I H z LPI processor 104

. . . . . . . . . . . 4.1 T i ï e m of the 20 .\ I H 2 processor from the laser diode end 105

. . . . . . . 1 Colliniating and pe-detection -40 cell aperture matching optics 110

vii

Page 10: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

ABBREVIATIONS

-40

ES11

-AR11

LPI

SIS

SSR

FM

FH

DSP

BS

SS

TDOA

CCD

RUS

AWG S

YB T

CRR

P S

s -4

GFLOPS

VCO

RIS

DR

P R F

I M P

Acoiis to-Opt ic

Electronic Siirveillance Ueasiire

Ant i-Racliat ion 'cIissiles

Lon- Probability of Intercep t

Sarrowband Interference Signals

Signal-t o-Soise Ratio

Frequency Slodulat ion

Frecluency Hopped

Digital Signal Processing

Beam Split ter

Spread Spectrum

Time Difference of Arriva1

Charge Coupled Devices

Root Mean Square

Adcli t ive llrhite Gaussian Soise

l'ariable Bias Term

Chirp Repetition Rate

Pseudo-Soise

'l'umerical Apert rire

Giga-Floating Point Operations per Second

Voltage Controlled Oscillator

Relative Intensity Soise

Dynamic Range

Pulse Repet it ion Frequency

Int ermoclulat ion Prochic t

Page 11: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

SYMBOLS

difference freqiiency

electronic reference frecpency

static index of refraction of A0 ce11 medium

transit time of -40 ce11

ampli t ticle transmit t ance func t ion of --'O ce11

effective t ransnii t tance of A0 ce11

banclpass LPI appliecl to the laser

banclpass LPL applietl to -40 ce11

modnlat ion tlept h of laser

ratio of . s 2 ( t ) amplitude to the reference tone amplitude

Laser DC bias amplittide

reference tone amplitude

cent er frequenq- of .s2 ( t )

A0 ce11 cliffrac t ion efficiency

pliase of detectetl spatially varying correlation

iipper cittoff freciuency of -40 ceII banctwidtli

loiver ciitoff freqiiency of .\O ce11 bandwiclt h

-40 banclwiclt h

LPI bandwictth

intcnsit y of 'noiseless ' modiilated laser

intensity of noisy modulatecl laser

detected intensity at the CCD plane

time-al-eraged noise of laser source

processing gain of processor

Page 12: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Tt integration time of CCD array

-Y, interna1 A 0 ce11 acoustic/optic- interaction noise

1 CCD saturation voltage

- 1 (x ) -40 noise poiver contribution from mu1 tifrequenc?- diffraction

.lÉ(x) . A 0 noise poiver contribution from nonlinear acoustic interaction

(2 Bandpass filter qualit- factor CS

J u itpper cut-off frequency of the upshifted LPI applicd to -40 ceil

f; lowcr ciit-off freqiielicy of the iipshifted LPI appliecl to .-\O ce11

L' acoust ic veloci ty in -40 ce11 interaction medium

T c chirp durat ion

n chirp rate

f.v[s frequency of SIS

Page 13: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Chapter 1

Introduction

In the past few decacles the increased amoiint of informat ion conveyecl hj* cornmitnication

systerns has been follon-ecl iip b>- a corresponding increase in the amoiint of p r i ~ ~ t e infor-

mation [Il. [2]. In many cases. it may be usefiil to hicle the presence or the esistence of

the comniiinicator's signal. Thits. an unaut horizcd receiver cannot dis t inguis h t lie coni-

municatecl signal from the receiver noise pliis interference. Since detection is a process

of choosing between noise alone and signal plus noise. hiclclen signals chat are clifficult to

clctect hy iinatithorized receivers are said to have loiv probability of cletectiun ( L P D ) [2].

Froni an intercepter's point-of-view. LPD signals appear as backgroiincl noise.

111 ot her applications. it ma>- not be possible. using LPD techniques. to a\-oicl ha\-ing the

signals detected hy unaut horized receivers. In those circumstances. it is possible to cleny an

iiriauthorizecl intercept receiver the features of the signal that coiild be ernployecl to derive

its form. Siich signals are said to ha1.e a low probability of intercept (LPI). Therefore. the

objective of LPI signal design is to produce signal iincertainty at the iinaiithorizecl receiwr

ivhich results in a signal-to-noise ratio (SXR) much lower than coulci occiir in the absence

Page 14: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

of LPI signal characteris tics.

Modern radars increasingly use spreacl-spectrurn techniques to arhieve the LPI contli-

t ion. Radar designers are ever considering wawfornis which are more difficult t O intercep t

to protect against anti-radiation missiles (-ARMs) and to recliice the detecrion range of other

hostile act il-i ties. Se\-eral of these radar systems are ciment 1'- operat ional tri t h man- niore

in the del-elopment stages. employiug tlirec t-sequence. frecluency-hoppect. p hase-mocliilated

or linear FM (chirp) signals. These signals have insufficient potver to be ahoïe the noise

ffoor and to be easily processecl. -4s a result. the receiver in processing these signals.

miist achieve sufficient SSR gain. From the interceptor's view point. a tliffereiit sort of

siirveillance measiire receiver is needed for these LPI radar 11-a\-efornis.

In using sitrveillance measure receivers to detect LPI radar signals. it is necessary to

estract large bandividtli signals. froni the electromagnet ic spectruni. having low SSR. One

approach to solring this problem is to use a matched filter [3]. [-LI. which produces at irs

output a signal with S S R , = ( T S B ) S T R t where TqB is the tirne-bandwidtl procliict of

the receiver input signal (of cliiration T, and bandividth B) antl. S-1-R, ancf S I R , are

the output and input SSR's of the matched filter. respectively. The procluct TsB (for

radar LPI signal. typically T, = 1 nis. B = 0.5 GHz) is often referred to as the processing

gain [5 ] of the recei\*er (in this case the filter) since its magnitiide increases S.l-Ro and tkos

facilitates signal detcction. By definition. however. the LPI radar signal form is not knoivu

a priori and. therefore. the matched filtering technique is difficult to implenient. lioreover.

the detecrion system must ha\-e flesibility to alter the matchcd filter in order to cletect a

large varicty of the LPI radar signals that are -hiclden" in the loir- SSR environment. This

rules out "ordinary- analog mat ched filters.

Page 15: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

By using a digital signal proccssing (DSP) search technique. such as masi~iiuni-likelihood

detection [6]. it is possible. in principle. to determine the type of signal iisecl bu the LPI

raclar. In practice. the more samples of the LPI signal we collect. the larger the prohahility

of interception. Cnder tkese circunistances it niakes sense to deploy a niimber of receiving

antennas in order to capture man- signals. These antennas. a niininiiim of two. are con-

nectecl to a central processor where the LPI signals are then processecl. However. this DSP

approach is estremel- slow and for typical LPI raclar signals. inapplicable. For esample [Tl.

a 500 MHz signal. with 1s time duration. has a time-bandwidth procliict of 5 x 10" and

requires on the orcler of 3 x 10'' digital computations to produce the conipressed pulse.

If piilse compression is performed every miilisecond. 3 x 101%peertions per second are

required: well beyoncl current D SP capabili t ies.

Correlation techniques have heen clernonstrated in the past as a riahle means of cle-

tecting radar signals with LPI [SI. Csing a correlator architecture as sliown in Fiy. 1.1.

we consider t lie interceptecl wa~eforrn as i ts own reference resiilt ing iii an autocorrclator

implementation of an 'automatic' mat ched filter. The two receivers capture the wa\-efor~n

wi t h some relative time-difference-of-arri\al ( TDO A ) related to the t arget angle 8. Because

t lie two interceptecl noise signals are statis tically uncorrelatecl. only the desirecl n-iclebanci

LPI correlation is available at the output for further processing to clerire time-of-arri~al.

TD OA. cent er frequency. bandwiclt h. signal type and ampli tucle.

Complicating the detection process is the ptesence of strong narrowbancl interference

signals (SIS) mithin the LPI bandwidth [II. The SIS may be unintentional as t h e to

a nearby trammitter or intentionally inserted by a hostile jamnier. Often. the SIS has

an amplitude much greater than the wideband signal and autocorrelates to dominate the

Page 16: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

A 1 Translater 1 1

A n t e n n a '1

Figure 1.1 : Schematic of raclar LPI correlator

A n t e n n a 1 Ampli tutle

conipressed LPI raclar signal. Thiis. one of the key technical issues is how to rernoi-e the

r Correlator

SIS hefore processing the wiclehand signal. -4fter rke SIS has heen iclentified autl ~~~~~~~~ed.

--c

a correlation rcceiver is t heu iised to characterize the u-iclebaud signals.

Correlation receivers can be classifiecl into two categories: digital and optical (91. (101.

- O Delai

4 1 )

Digit al spreacl-spec t rum correlat ioii receivers are well known as t hey are iisecl in satellite

,

communications [Il]. [12]. Hoivever. they lack the speecl and the bancl~vitlth to hanclle the

Frequency Trans la tor

demands of newer spread spectruni radar signals which are typically in the freqiiency range

\

0.5 to 26.5 GHz with a minimum bandwidth of 0.3 GHz. Recently. the most proniising

correlation receiwr architectures for carrying out detect ion and classification of LPI raclar

signals. have been t hose based on acousto-op tic (-40 ) t echnology [13]. [14]. Thcir op t ical

Page 17: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

engines have the potential to provicle S-10 GFLOPS of processing power which is a 1000-

folcl irnprovernent over current electronic systems [9].

A 0 technology has recently played a pivotal role in no\-el c1ti.elopments applicable to

op tical comput ing and holograp hy [l5]. optical cornmitnications [l6]. (1 71. op tical sens-

ing [ l G ] . [lS]. [19] as well as to medical image processing [201. Tirne-integrating -40 based

processors have generated great interest as potential LPI receiven. These processors are

attractive for LPI receiver design from three points of v i e x ( i ) t h e - can opcrate in such

a way as to --automat ically" penerate the required mat checl filter for nitilt i tude of simult a-

neous LPI radar signals. ( i i) they can often provicie the requirecl information in real-time.

and (iii) they can provide in escess of 1 GHz real-time banclwidths and processing gain.

clefined as S.\-&/S.SRi. on the order of 10Iog(T, B) r;: 60 dB. The latter allows for the

design of LPI recei\.ers n-itk input S-YR, as low as -50 dB.

;\O based receivers for the cletection of LPI signals bal-e been investigatecl recently by

a number of researchcrs [21]. [SI. [Z]. ['BI. In the last few years. the tiw A0 processors

niost coninionly studied have been the adc1itix.e ( Slach-Zehntler ) arclii tecture [l-k] . and t lie

niiiltiplicative (in-line) architecture [l-L]. 1161. [21]. [10j. In the first approach. two A0 cells

are usecl. one for each intercept LPI signal (or one with a reference waveform). and the

signal bandwidth B is eclual to the full A0 baridwidth B.4o. Howewr. this design suffers

froni mechanical instability and requires high precision optical components ancl cletect ion

system. In the second approach [IO]. an electronic reference tone is used. The use of an

electronic reference tone offers mechanical s tabili ty and reduces clemand on t lie op t ical

and cletection cornponents. hoivever. as shown in the next chapter. the axailable signal

banclnidth is reduced hy half to B = B,lo/2. In addition. this approach reqiiires much

Page 18: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

more ela.borate atisiliary elec tronic hardware as described in Chapter 3. Time-intcgrat ing

.-\O receit-ers based on the concept of an electronically inserted reference tone. ha\-r sho~vn

consiclerable proniise in the area of LPI cletection and are the basis of the theoretical and

esperimental investigation of t his t hesis.

In a section that folloivs. a simplifiecl analysis of the operation of an -40 (or Bragg) ceIl

is presentetl. The theory leads to the clefinition of the transmittance function of the ce11

that is esploitecl in the analysis gil-en in Chapter 2.

1.1 Transmittance Function of an A 0 ce11

An acousto-optic ce11 is a spatial and temporal ligkt niocliilating device that ericotles light

with information containeci in an electrical clri\-e signal. It consists of an interaction ma-

terial. silcli as g l a s or an esotic crystal (for esample LiTnOÎ. Li.1-bO3 J; Ti@) . to ivhich

a piezoelectric transdticer is honcled as shoisn in Fig. 1.2. The drive electrical signal f ( t )

causes the transdiicer to l-ihrate lnunching either a compression or a shear acousric wave

into the mediilni which in turn sets up strain waves in the crystal. Thesc strain waves

lead to density changes in the interaction medium and consequentl~: to incles of refractiou

changes. The net result is that light passing through the -40 ce11 in the --direction is mod-

illatecl in phase according to changes in the optical path inditcecl by message signal f ( t ) .

The diffracted light enwrging from the nieclium is t hus modulated \vit h the iuforniation

containecl in the original electrical signal.

The nature of the acousto-optic interaction is influenced by the thickness. in the di-

rection of the light wave propagation. of the interaction region. A0 clevices irith large

interaction lengt h compared to the distance between neighboring acoits tic ~va\.efronts in

Page 19: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

incident

light

. . .

l r Acoustic Absorber

Transit time T

Figure 1.2: Bragg cell: For a given drive signal f ( t ) with center freqiienq- f. n-e see chat the acoiistic wavelength is .\ = 7. The spatial frecliiency is therefore n = +. Hence. tlie

Bragg angle of the ce11 OB = Xci = y. .An ncoustic absorber at the other end is used to prel-ent acoiistic reflections wit hin the ccll.

the cell. are said to be operating in the Bragg regime ancl are referred to as Bragg cells.

Uï th shorter interaction length. al1 cliffractiori orders are present. and the clerice is saicl

to be operating in the Raman-Sath reginie [25]. This n i ~ ~ l t i tlicle of diffracted orclers leacls

to a waste of potver. Hence. most signal processing functious are carried out in the Bragg

regime [14].

Assuming a linear isotropic interaction medium in the .-\O ceIl columri. the modulated

Page 20: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

optical incles of refractiou is thus given bu [13]. [Xi]

1 for I z l s f r e d ( z ) =

( O otherwise

where no is the static indes of refraction of the material. L' is the acoustic velocity in the

interaction niecliiim. and ni is a constant of proportionality. nh choosc tlie niid point of

the .-\O ce11 as the origin of the CO-ordinate system so that the dri\-e signal a t the estrcme

end of the ceIl ( s = L 12) is given by f ( t - T ) 11-here T = L / L* is t ke a dela- ecliial to the

transit t ime of the cell.

At the incident plane of the ce11 (; = O ) . the optical aniplitucle of the light beam is

-4 0

11-liicli is a plane mave wit h angular frrcpenc~- dl. At the esit face of the cell. z = 2. the

aniplitucle of tlie light beam is

where the transniittance or modulation transfer function of the ce11 is clefinecl as

Page 21: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

which is derived from a o that is a function of al1 spatial ceordinates as follows

wliere do is the iucreriiental change in the phase of the light beam as it propagates across the

acoiis to-op tic interaction mediiini wi t h \va\-elengt h A. The total phase accriiecl is oh t ainecl

hy integrating Eq. (1.5) to yielcL

- constant tcrm

By disregarding the constant phasc factor ancl keeping only the first two terms in a Taylor

series expansion of the esponential terni and with the assiiniption of a Bragg riiode of

operation. we have

where m = '":"L is defined as the rnocliilation i d e s of the cell. The transrni t t ance fwiction

consists of a constant term mhich passes the oth uncliffractecl orcler plus a linear spatially

and temporally niodulated image of the drive signal f ( t ) corresponcling to eit her a positive

or negat ive diffraction order.

Hence. the drive function f ( t ) causes the cell to behave as a phase diffraction grati~ig

Page 22: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

in the s-direction. The light leaving the acoiisto-optic ce11 in the first positive diffraction

orcler may thiis be espressecl in phasor form hy the equivalent f~inction [14]

d w r e (s ) is the aperture function of the system xhich accoiints for nonlinear distort ions

due to the illumination function of the source. attenuation factors. laser powr le\-el. and

truncation effects from the acousto-optic ceil or ot her optical elenients in the systern. IL-hile

r i t is the mocliilation indes of the ce11 representing the power distribution in the cliffractecl

orcler relative to the uncliffractecl mode. The final term implies that f ( t ) has heen miiltiplied

by c o s ( k f , t ) to translate it to the center frecliiency f, of the acousto-optic cell. \\é are now

in a position to stuclj. the in-line optical correlator architecture based on this acousto-opt ic

interaction phenoniena.

1.2 Thesis Outline

This thesis consists of a theoretical stiicly. numerical simiilation ancl an esperimental in-

vestigation of a time-integrat ing .\O processor architecture wit h a goal of detecring and

classi-ing spreacl spectrum LPI radar signals in presence of noise and SIS.

Chapter 2 esplores the t heoret ical feat ures of the mult iplicath-e in-line architecture.

The criterion for select ing the various frecpencies and bandwidt h cons traints of the receiver

is de~doped . In addition. the overall processing gain of the receiver. assuming noisy input

signals and a noisy laser intensity. is developed using a stochastic modeling approach.

-4nalytica.l espressions for the processor output IL-hen fed wi th either chirp or frequency

Page 23: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

hoppecl signals are also derivecl and niimericaliy ecaliiated.

Chap ter 3 iutrodtices the esperiment al setiip and equipment interconnect ion usecl in

this work. The \-arious signal and noise gcnerating siibs>-stems are first disciissed. Direct

mocliilat ion of the laser as reqiiired by the the-integrat ing archi tec t lire necessi t atecl t ke

design of aclcli t ional circuitry. The performance of a 20 JI H 2 signal banclwicl t li processor

in correlating two ptilses embedcled in noise of ~-arying SSR is presented. The influence of

additive Gaiissian noise on the correlation peak and width was nieastirecl ancl analyzed for

\arying SSR \-dues. -4 no\-el approach for estimating the overall optical misalignment in

the processor is then de\-isecl and demonstratecl. The observecl clilation in the correlation

width is esplainecl in terrns of the sampling characteristics of the CCD arra-.

-4 50 M H z signal handn-idth processor that is a scalecl i-erçion of a typical radar

receiver. ivas designed and assemblecl to correlate chirp and frequency tiopped spreacl spec-

t riini signals. An esperiment al analysis of the clet ect ion ancl categorizat ion of LPI signals

corruptecl b>. noise \vas then carried out. In a section that fol lo~s. the impact of the SIS

on the LPI is esperimentally demonstratecl. -4 novel approach for rcal-tinie detection and

removal of the XIS from the LPI signals prior to correlatioo is proposecl and denionstratecl.

C hap ter 4 illus t ra tn the hardware aspects of the various processor niocliiles. measure-

nient ancl test equipnient from a nurnber of mutage points. The O\-erall processor 1q.oiit

is also included

Finally. a summary of the results and contributions of this work is presented in Chapter

- 3.

Page 24: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Chapter 2

Theoretical St udies

In t his chapter. the theory of a one-dimensional .\O correlator with an electronic refereuce

tone is reviewed with the aini of deriving the frequenc?- selection constraints that niiist

be followetl in its design. The output of such a correlaror is then analyzecl for linear FI1

chirp and frecpency hop (FH) input spread spectriini signals ancl t hereafter nieasiirzcl in

esperiment al cases considered in Chapter 3. Sirnidation of t lie correlator out put II-as also

done for input signals of ~ary ing STR levels wi th the assumption of additive n-hite Gaitssian

noise (.\it'GS).

In orcler to speecl up the fectlhack loop necessary for identification and excision of the

SIS. a space-integratetl output is considered in adclition to the conventional time-integratetl

output. The idea of using a space-integrated output. witli this processor architecture. is

an original research contribution of this work.

The behavior of the correlator in the presence of noise is then analyzed ancl the detected

"noisy" intensity is espressed in tcrms of the --noiseless" intensity and an ascending order

of noise cross terms.

Page 25: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

The correlator noise theory. first piihlished by Iiellman [26] ancl also consiclered hy other

researchers [XI. [->SI. [09] rvas generalizeti to include laser intensity noise in this work. The

correlator output SSR was then clesivecl and cliscussecl for a nurnber of scenarios. The

inclusion of the effect of the laser intensity noise to the overall processin:, gain expression.

is another original research contribution of this work.

2.1 F'requency Selection Constraints for the Multi-

plicat ive (In-Line) Architecture

In a t inie-integrating -40 correlation receiver. shown in Fig. 2.1. the laser is directly niocl-

ulated with one of the intercepted LPI dong with an appropriate DC bias l e d The laser

iight is then matched to the active aperture of the -40 cell by colliniation ancl ananiorpliic

optics.

-4t the other encl. another interceptecl LPI (or reference wal-eforni) is iisecl to phase

nioclulate the .-\O ce11 transmit tance fiinction t h e r e - converting the LPI inforniatiou to

optical form. The optical intensity incident on the detectors is the procliict of the laser

intensi t ~ - ancl the intensi ty diffraction pattern from the illiiminat ed A0 cell. Detector

integration then yields the reqiiired time averaged correlation of ttvo LPI signals. The

position ( h g ) of the correlation peak on the charge couplcd detectors ( CCD ) is proportional

to the tinie-differencc-of-arriva1 (TD0.A) between the two interceptecl waveforms.

In what follo~vs. al1 the terms in the output expression for the processor as they relate

to frecpency and LPI handwidth selection requirenients. are consiclered. Tliese selection

constraints are derived t hroiigh the following analysis. IVit h refereiice ro Fig. 2.1. the

intensity of the niodulated laser diode is giveii by

Page 26: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

where .-li is the DC hias aniplitiide. ni1 is the niotliilation depth. s l ( t ) is the banclpass LPI

signal of intercst and do is the modulation carrier frequenc.

For the correlat ion. ire i d 1 use the negat il-e diffraction order as it natiirally ?-ieltls

the cornples conjugate in the correlation integral. The -40 cellas amplitiide t ransniit tance

function for this negative diffraction orcler (when the oth order is ornittecl). as given

in (1-11 and modifiecl to apply to the architecture in Fig. 2.1. is

where r = T/2 + .r/c*. rl is the A0 ce11 diffraction efficiencj-. A; is the electronic reference

frequency. d l ~ = dc + J, is the crnter frecluenc~. of the spread spectrurn LPI signal s 2 ( t )

niodulating the acousto-optic cell: and rnz is tlie ratio of tlie signal amplitude to the

reference frequency oscillator arnplitiicle.

The correspondiiig effective transrnittance that applies to the intensities is giwn by the

product of the mot mean square (RLIS) \allie' of t.40

'LC'ith a small modulation depth (m2 << 1 ) needcd to mairitain linearity iri intensity. Eq. (2.2) is approsimately equal to a constant amplitude pliasor.

Page 27: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure 2.1: l-D nlultiplicative architecture: do = . i . l ~ - dc: do = difference (offset) fre- qwncy: xc = refereuce freqiiency: the center frequency of the s 2 ( t ) signal.

Page 28: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

The intensity iniaged onto the CCD a r r q is

Espanding Eq. (2.4) r-e have

ivhere we have used the itlentity

The first terni contains the DC hias q.4i-42/2 and a signal depenclent bias terni rlr ic: . -L1 .A2/' 1

s 2 ( t - r ) I q h a t integrates to a constant for sufficiently large integration tinie T,. Terms

( 2 ) . ( 3 ) and (4 ) are bandpass terms that integrate to zero for T, >> k. Of interest is the

last terni in Ecl. (2.5) which can be sliomn to eclual

ivhere the tinie integration of (Sb) term leads to the desired correlation.

Page 29: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

In Eq. (2.7). (5,) is a bandpass terni at frequency L,. outside the acoiisto-optic band-

width. that also integrates to zero for T, >> 112 f,. Terms (1) and ( a b ) after integration.

becorne

- \ciriahle bias term (L-BT)

Hence. the amplitude rather than intensity modulation of the A0 cell. bas producecl R ( r )

where t lie elcctronic refereuce frecluency f, allows one to separate the correlation terni from

the \-BT. This correlation oiitpiit can be written as

where o12(r) is the phase of the tirne averaged terni JO s l ( t ) . s i ( t - r)dt such that

Therefore. we can find R ( i ) by recovering the envelope .412(r) and the phase 0 1 2 ( r )

of the spatial carrier by coherent detection. Xote that since T = T / 2 + .c/c. the peak of

- the correlation i d 1 occur at s = -2 2 - -L 2 ' which corresponds to the bottom edge of the

out put plane. leading to a one-sided correlat ion func tion. This correiat ion funct ion can

thiis be centered in the output plane by introducing a relative ciel- eciiial to T l 2 in either

Page 30: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Tetriporal Freqtrency Spectruni

-B O B 10 2 J O

Figure 2.2: Spectra grouping of the varioiis processor outptits from Eq. (2.5)

r - - - - - - I I I I I I t

s ( t ) or s 2 ( t ). né choose to dela? si ( t ). the LPI signal mocliilating the laser source.

Finally. it is instructive to consider the \arious terms given in Eq. ( 2 . 5 ) in the temporal

freclueuc~ do~iiain as shon-n in Fig. 2.2. In order for the VBT not to overlap with the

correlation signal at fo. we require that f, - 8 1 2 3 B or f, 2 3B/2 . Similarl~: to avoicl

overlap with the signal at 2 f o carrier (terni ( j a in Eq. (2.7)). ive must have f, + BI2 5

2fo - BI'. or f, 2 B. Both these constraints are siniiiltaneously saiisfiecl if 1 f, 12 3 BI?.

where B is the handwidtk of spread spectriim LPI signals s l ( t ) and s 2 ( t ) . Biit the center

frequeuc- of the signal s z ( t ) niodulating the cd1 is f iF = f a + f,. Hence. two solutions are

possible with the constraint 1 f r F - f, 12 3Bl': fi = f f r F - l.5B or f:' = f I F f l.5B. In

I I * frequency

- . . - . . . . . . - . . . - . . - . . . -- : - - - . . . - . . - . . . - * - . . . . .- *- : . . . - . - . : : : . - - . . .. . . - S . . - . -

the first soiution. the electronic rcference f, is placed nt the lower cutoff frecluenc- of the

.-\O passbancl while the upper cutoff frequency of the radar LPI signal s 2 ( t ) is p ~ t at the

upper cutoff frequency of the -40 passbancl. This solution is shown in Fig. 2.3.

Alternat ively. the frecpency selection const raints can also be sat isfiecl by placing the

elcctronic reference f, at the upper cutoff frecpency of the A0 passband. The lower cutoff

frequency of the radar LPI signal is then placecl at the lower cutoff frequency of the -40

(2).(3).(4) and (.5b) . . . . - . . . .

0 . . . . . .+ ..- :- B.-- - . . . . . . . - . * . . - .

terni (.?a) . . . . . .

+.a

Page 31: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

10 (a) laser diode input

(b) :\O ceIl input

Figure 2.3: Frequency selections and spectriini placement of LPI signals modiilating (a ) the laser and ( b ) the acoiisteoptic cell. The electronic reference f: is at the lon-er ciitoff ( flo) of .-\O passband. The upper cutoff freqiiency ( f r F + B / 9 ) of the LPI is at the iipper ciitoff ( fko ) of the cell. Alternative solution with f: at fito and LPI downshiftcd to f iF - B/3 = f)' is also possible. Sote that in the figure shown. the modulation carrier frequency of s z ( t ) is f r F = f, + fo. and = f i o - ftO.

Page 32: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

passhand. Froni a theoretical point of view. both solutions are iclcntical. In this thesis. the

first sol~ition iras useci througlioiit .

Before we leave this section. let lis cliscuss one last constraint iniposecl by the -40 ce11

bandwidth ( B.4o ). In the soliition aclopted in this thesis. n-e have that fi;'' = f r F + BJZ.

where f:' is the iipper cutoff frequency of the -10 ce11 handi\-idth and B is the LPI

handwiclth. Ré also reqiiire that the lower cutoff frequency of the A 0 ce11 bancl~viclth

fiLo = f:. But Iiellrnan's frequency constraints require t hat

f l F - f: = l 3 B . hence

Thus. the A0 ce11 bandwidth m u t therefore be twice the LPI banclwiclth. -4lthough the

insertion of an electronic reference tone tielps to make the design practical stahilizing

the output ancl provitling a means for controlling the bias ternis. the pena1t)- ire pay is a

net reduction in the overall availahle -40 bandn-iclt h.

These constraints were theu iised to select the various frecpeucies necdecl to setup the

processor. The recpired spectrurn placenient for the signals motlulating the laser diode

and driving r he -10 ce11 are as shown in Fig. -2.3 respectively.

2.2 A 0 Correlator with Linear FM Inputs

Consicler the -40 processor nioclel for LPI detection shomn in Fig. 2.4. LI-e shou-ed in the

last section how the various freqiiencies and banclmiclth. shown in insets. are clerivecl in

terms of placement and specifications. Let us assume that linear FM chirp signals S ( t )

Page 33: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Seu t ral Density Filter

1 L-- - R B P F

Figure 2.4: LPI signatures S(t ) and F ( t ) are detected by tmo widely separated antennas: boosted bp an LX -4 t hen subsec~uently clown comrerted to the frecpency band specifications of the optical processor. The spectrum at each critical stage of the processor are shoim in the insets. The ciel- line in the circuit for input S(t ) is not shown. Further details can be found in the text.

Page 34: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

and F ( t ) are received frorn tivo wiclely separated antennas. This is characteristic of a real

situation in which correlation of these two signals provicles a good indication of n-hethcr

or uot a transniitting source esists in the vicinit- This is the classical detection problenl.

.-\ltcrnati~ely. the second signal F(t ) is used as a reference signal ishich is comparecl with

the interceptecl signal S ( t ) . The problem then is one of identification by correlatirig S ( t )

isith a series of sample reference signals until a strong correlation is cletectecl signifying

that a niatch has been founcl.

\Ié i d 1 restrict oiirselves to the matchcd filtering of the former case. Il-e ivill motlel

the LPI chirp signals detectecl by the processor after passing through a loi\- noise amplifier

( LX;\). clown-con\-ersion and bandpass filtering as

and

F ( t ) = cos(i?af,t + ant' + < P t ) + I I > ( f )

where the phase terms <DI. are uriiformly distributecl ranclom Lariables over the intcrval

O to 'a and a is the chirp rate in Hzlscc . The addition of tkis rancloni phase recoguizes

the fact that the timc origin is arhitrarily chosen for the case when the signals corne froni

tn-O widely separated antennas. Since the correlator shifts one of the inputs h - a tinie slot.

before each integration until a correlation peak occurs. this phase dela. is thits inherent in

the design and will be ignored in the subsequent analysis. \iè mil1 also represent the noise

corruption in the signal paths by the additive white Gaussian noise processes Ilel ( t ) and

I4>(t) with spectral densitp .\o/2.

The classical time-integrating correlator with an electronically inserted reference tone

Page 35: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

leacis to an iutegrating action by the photocletector given 11'- Eq. ( 2 . 5 ) and repeated Lere

for convenience. ;'

where I,(t) ancl 12(x. t ) are the modulatecl intensities of the laser source and light esiting

the -40 ce11 respectively. Rt( T,. r ) = J: S( t ) F ( t + r )dt is the desired correlation. T = + f is the displacement kariable and Ti is the integration time of the detectors. \Ié non- proceed

to calculate this correlation iising tlie mode1 of LPI i~aveforms that are pickecl up by the

ant ennas.

In man? practical situations. the only thing ive usiiallj- have a\ailable is the recorcling

of one or a sniall nuniber of saniple fiinctions of the randorn process. né then consider

the time averages of indi\-iclual sarnple functious of the processes. Let us clenote tlie

saniple functions of each of the processes hy f ( t ) = ci cos(?;; fol + a n t L + BI ) + c r i ( t ) and

~ ( t ) = (i2 c o s ( 2 ~ f ~ t + ;id2 + 0 2 ) + ~ ( t ) . In anticipation of the limitation of having a single

noise source for the esperirnental setiip in Fig. 3.8. ive will cousider the specific case i h e n

0 , = Br ancl tri ( t ) = w2(t ). The output of the noise generator is ecluallj- clividetl iising a

power splitter and added to each LPI signal path.

Of interest here is the last term in Eq. (2.14) representing the spatially varying cor-

relation function. Let c i l = ci2 = rect{$) for brevity and the instantaneous frcquency of

the chirp f; = fo + a t . The linear FM correlation term V(T,. r ). assuniing a single chirp

'.-Ifter integration when al1 bandpasa terrns with carrier frequcncies at JO and ?Io have vanished.

Page 36: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

wi thin the integration n-inrlow Tl . ma- thtis be expressecl as

For an integration tinie Tl which is much greater than tlie ;\O ce11 dela- T . Eq. (2.15) can

be approsiniatecl as

This resiilt shows t h the correlator oiitpiit is thiis a spatial carrier niocliilating a sinc

fiinrtion 11-ith a niain lobe of diiration -& meastirecl froni a zero to zero crossing. But the

R CI' 2 Te chirp rate is given hy o = 7. hencc & = m. This expression re\-eals t hat the wiclth of

the niain lobe of the correlation is inversel- proportional to the protliict of the integration

tinie ( Tl ) and the chirp hanclwidt h ( BI.1.-): and direct ly proportional to the chirp durat ion

Tc.

Son+ the acltlit ive white Gaiissian noise component is ergoclic in tlie autocorrelation and

equals to ) 6 ( ~ ) . Hence. the total espression for the correlator otitpiit can be su~nrnarizecl

.-&-A2n2 1m2T( -Yo a ( - 4 [ r l => cos(2n for)s inc(aTir) +,d(r) + VBT]

L'(Ti *T)

where <i ( r ) is the aperture function of the processor as derived I>y \7anderliigt [l-l] and

Page 37: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

repeated herc for convenience

and .r, = -&Y s .4 is the shifted central \ d u e of the Ciaussian illumination profile of the

laser. c is the atteniiation constant of the A 0 niaterial. L is tlie lengtk of the ce11 and -4

defines the light amplitude at the edges of the cell. The Gaussian ilhimination is modifiecl

hy a frequency dependent (f k i n g the center freqiiency of the -40 cell). esponentially

at teniiating factor of the for* (1, = E - ( ' ~ ! ~ / . ' ) ( L-cLf2/8-4).

2.3 A 0 Correlator with Frequency Hop Inputs

-4nother popular spread spectriini signal commonly iised in radar applications tocla>- is the

frequency hopper schenie. Consider the same setup as sholvn in Fig. 2.4 trith signals

n B 1.1,-t F ( t ) = (1.~ COS(^ fOf + - + (a2) + lV2(t)

&V + 1

d e r e again ive set (11 = « 2 = 1. Bit' is the total hop hanclivitlth. -1- + 1 is tlie to-

tal numher of discrete frequeiicies hoppecl ciiiring integration periods such that If,) =

nB\\' - +

= f l . . . . fn . - - . f-y. does not imply a monotonie secpience of freqiiencies and

3 ~ s e d to describe the amplitude aeigliting function of the laser illiiniination and truncation effects due to lenses or other elements in an optical systern.

Page 38: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

(ri } = --\'/2. -( .Y/2) + 1. . . : SI2 in rancloni order. The rvicleband LPI correlation yielcls

Hence. for Ti >> T . where T is the A0 ce11 clelaj- aperture. Eq. (2.21) can be approsi-

niated as TB\\' --11=12m1rn2T; stn(-)

\-(Tl . Î ) z q c o s ( 2 ~ f o r ) L 9 - B C I ' - (3- + l).sirz(-)

rB \ lW r B LI' Sote that in the limiting casc. for S > > B1.l'. sin(-) z - ancl the freqiiency

hopper approaclies the linear FI1 ckirp case as Eq. (2.22) ceduces to

The nest section investigates the under1:-ing features of Ecp. (2.17) and (2.22) iising a

numerical simiilat ion approach.

2.4 Numerical Analysis for the Two Types of Inputs

The niodels of tlie LPI correlations rvi th the -40 processor tleveloped in the previous section

w r e implemented in the MAT LAB numerical simulation pachge and the resul t s coniparecl

to the approximations in Eqs. ( 2 . ) a n ( 2 . 2 ) Figure 2.5 shows the processor oiitput

simulations for \-arying S N 3 levels mith Gaussian noise. The most noticeable effect is the

scnsitivity of the processor in detccting a correlation ewn under severe noise conditions

( < -20 dB). -4s ivell. as predicted. the base of the correlation is broadened 1>y tlic additive

Page 39: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

noise. Similar resiilts werc obtainecl with the freqiiency hopper case iising 300 hopper

frequencies in the signal ba~iclividt h. 2 : 2 :

Cc) Cd >

Figure 2.5: Soise loaciing on the correlation for i+arioiis S S R tdties: (a ) O dB (b ) -20 dB ( c ) -40 dB and (cl) -50 dB. Horizontal scale: lag position in riuniber of points: Xértical scale: arbitraq- units.

Consistent !vit h the t keoret ical predict ions. the envelope of t lie correlation was found to

be a jinc function tri th a duration of the niain lobe giwn by &. -4s well. the predicted

frecliiency of the spatially iar3-ing carrier on which the correlatiou is riding was in direct

agreement with the theory. Figure 2.6 shows the hehavior of the correlation as a fiinction

of the ratio of the chirp rate to the signal bandwiclth ( &= ).

Page 40: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure 2.6: Effect of Larying the chirp cliiration (Tc) on tlie correlation. ( a ) 10 paec. ( 1 1 ) 5 pscc. ( c ) 2.5 pet. ( d ) 1.5 psec. Horizontal scale: tirne ( p s e c ) : kértical scale: arbitrary

2.5 Space-Integrated Output of the Correlator

In this section. Ive consider the case where a beani splitter (BS) is inserted as as shoivn in

Fig. 2.7 follon-cd by a Fourier trausforni lens that integrates the spatial light clistribiition

and focuses it on a single PIS photodetector element. The goal of siicli a design is to pick

out tlie freqiiencies of the SIS signals with high speed. as selecting of the SIS signals froni

the CCD detectors is de1-d by the integration tirne Ti. This information is then used to

identify and excise tlie interferer energy from the LPI signal spectriini with near real-tinie

speed.

Our starting point. in applying this metlod. is the temporal aiid spatial light interisity

distribution at the photocletector plane giren by Eq. ( 2 . 5 ) Let us cousider. term by term.

the effect of the wide angle Fourier transform leas followed by spatial f rcc~uenc~ integration

Page 41: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Digital Sarnplirig Osci Ilosco pe

k Spectruni Analyzer

modules

Figure 2.7: S p c e integrating pin detector arrangement. -4 beam split ter (BS) is iised to channel part of the optical beam into a Fourier lem ( L 3 ) which focuses it ont0 a photodc- tector (PD) connected to a sampling oscilloscope for analysis.

Page 42: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

the pin cletector.

The action of the Fourier lens is to integrate the temporal and spatiall- \ar>-iug intensity

profile I& ( t . r j gii-en in Eq. (2.5) leading to

n-here T = T/2 + x/c and with the asstimption that the lem dianieter is sufficiently large

relative to the optical heam widt h and hence the choice of limits. The transforni idet ( t . ci )

is centerecl at no. the spatial carrier frequency at the photodetector plane and spans an

fo+Bl'L interval 2ûd. where ad = - corresponds to the spatial cut-off frequency of LPI signals

a , ( t ) and . s 2 ( t ) n-ith bandwidth B. The Fourier transform of the aperture function of a

single element p hotodetec t or is modeled as [l-t]

and clo = l / h m is the Syquist sanipling interval. The photocurrent obtainccl bj- integrat-

ing the iutensity over the photodetector surface is proportional to

In this analusis. \se are going to assunie P(a) is uniform oïer the surface of the cletector

and with iinity magnitiicle so that

Page 43: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

This then is oiir kernel function operating on al1 the terms in Eq. ( 2 . 5 ) . Siibstituring in

Eq. (2.26). ive get

with r = Tl2 + s / ~ * : hence for the 13' terni of Eq. (2 .5) . LI-e have"

where i t is assiimecl that the system's aperture function n ( r ) = 1. Invoking the sifting

theorem of delta fiinctions, we have

ive apply the same analysis to the remaining ternis to yield

n-here to th .s 1 ( t ) and s z ( t ) were assurned real.

" Where we have ignored the multiplicative constant 7.4 i.-lz/2.

Page 44: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

-- 7-

out-of-band

I lé make the argiinient that the last terxn in Eq. (2.33) ~mishes . Since the detector is

centered nt n o = fo / r spanning an in t end of spatial freqiiencies 2nd. this last terni ivill

1~ out of band.

The overall temporal signal output from the detector is thus sunimarizrd as

Page 45: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

1

58 iee ISE!

c2 >

I I 8.82 ' sa iee ise se iee 158

(3 1 (4) Figure 2.S: Simulation oiitput s for ternis ( 1 ) - ( A ) respect ively. Sote t hat because both the modulation clepth ( n z ) of the laser ancl that of the -40 cell ( r n z } are sniall ( n2 1. rnz < 1) for improved processor gain. both terms ( 2 ) and (4) ivill be negligible in the overall output.

lié can t hiis directly recover ail LPI frequencies incliiding t hose of the nt tendant SIS hy

connecting the output to a spectriirii analyzer. Eqrtation (2.36) was siniulatcd in 1I.XTL.AB

for each of the component terms as sho~n i in Fig. S.S.

Ternis ( 1) and (3 ) are sirnply double sidehancl .AM with one being a delayecl hy T/i?

relative to the other. The second term is nonlinear -411 while the last terni is analogous

to envelope detection of the prodiict sl ( t ) s 2 ( t - T / 2 ) . Because in general the modulation

tlept hs rn 1 . m < < 1. terms cont aining produc ts of t hese dept hs are thiis negligible relative

to term ( 3 ) . As a result. the overall output is a linearly chirped AS1 ivave riding on a

Page 46: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure '7.9: Overall sinidated oiitput of the Eq. (2.36 ) showing the li~iearly chirpetl signal.

hias as shown in Fig. 2.9. Heuce. ive can recover the freqitency content of the ititercepted

signals si ( t ) ancl sz(t). bj- Fourier transforrning Ecl (2.36). This result is esplorecl furrher

in the nest chapter.

Page 47: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

2.6 Effect of Noise on the Intensity and Correlation

Funct ions

Consider the schematic shoim in Fig. 2.1. the iritensity of a noiseless laser diode is

where .Ai is the DC bias amplitude. m l is the modulation depth. .sl ( t ) is the bandpass

signal of interest. and do is the modulation carrier frequency. \Ve can espancl this eqiiation

as

and ru-Al = .4,, is the magnitude of .si ( t ) . In the presence of laser intensity noise n r D ( t ) .

insteacl of the Eq. (2.3s). ive haïe tkat for the intensity Id( t ) of the nais>- laser diode.

In the presence of both laser intensity noise and the signal .-voltage-- noise .\;, n 1 ( t ).

where .Ys, is the magnitude (assuinecl real) and where the noise signal has been representecl

as a bandpass signal n L D ( t ) : is the tinie-averaged quantum AM noise intensity for a directly

modulated semiconductor laser. LVe will rcstrict ourselves to ALI noise as it directly affects

the shape of the correlation envelope. However. any directty modtilatecl laser cliode will

exkibit both AM and FhI noise characteristics [30]. [31]. The FM noise will impact the

Page 48: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

sampling requirements of the CCD array and the stabiliry of the correlation in the detection

plane as it affects the spatial carrier frecpency on which the correlation is riding. LI-e will

assume that this effect is negligible since the spatially \ar>-ing carrier does not pl- a

role in the analysis that folloi~s. Denoting by S L D ( t ) = ~ ~ ~ ( t ) / - 4 ~ . -Y3, /-A,, = .\Tl and

m l = .4,,/--Li. ive can then w i t e as

Ecliiation (2.11 ) is the expression to be iised for a noisy laser cliode intensity. Coinparing

E c p (2.373 ancl (2.41) we have

where I d ( t . -ho. n ) is the noisy part of IA(t) dependent on the laser intensity uoise .j-LD ( t ).

ancl the signal noise n l ( t ) . né aclopt the same approach to cleriïe the noisy ecl~iivalcnt of

the -40 ce11 transmit t ance funct ion.

The noiseless -40 ce11 transmit tance. tmlo of the -ve diffraction order . IL-hen the O l h

order diffraction light is ornitted is gii-en bj-

where r = T/9 + s/r. A; is the electronic reference frequency. is the center frecpency

of . s 2 ( t ) such that do = - 4; and rn2 is the ratio of the signal magnitude .A,, to the

Page 49: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

reference freqiiency oscillator magnit iicle fi. Eqoation (2.43) can be writ ten as

-\oltage" noise .\;, n 2 ( t ) is real. ive can mite the noisy -40 transmit tance

that the signal

t;,(t - r ) as

n-here ive have neglectecl the reference oscillator noise. Equation (2.45 ) can be re-wri t t en

as

-4:- .-L" .y:., ',Y2-, and where nt- = = x. -11, = - = - .a:* .4a2

The effective noisy transniitrance of the A0 ce11 can be m i t ten as

Page 50: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

The noisy intensity IA,(t. r ) . iniagecl onto the CCD is thns given by

Combiriing Ecls. (3.41). (2.48) and ('7.49). ive Lave

1" order noise temi

+ &(t )TEj f (n2) + Id(t.-l'~~.rri)Rf{t.40(f - ~ ) f ; ~ ( f - T - h ) } }

v

3rd order noise term

therefore. Eq. (2.50) can be espressed as

a n )

For our purposes. we are going to assume that the noisy intensity is sufficiently accuratel-

descril~ed 11y the first order approximation. narnely that

Page 51: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Cpon detector integration. we haw

where R( r ) is the correlation function of the noiseless signals.

2.7 Output SNR and Processing Gain in the Presence

of Laser Noise

-4s pointeci out earlier in this chapter. the first noise analysis of the time-integrating A 0

correlator was clone by Iïellman [XI. In this thesis. ive generalize these results to accourit

for contributions froni the laser intensitl- niodtilation noise.

For two comples signals s , ( t ) and .s?(f). the cross-correlation function is defined as

where E { - } represents an ensemble average. The tirne-integrating est iniate ( clenotetl liere

by R ) of the function is

ivhere Ti is tlie integration tirne. Uiith reference to Fig. 2.10. the output STR for tlie

correlator is then definecl as

Again. with reference to Fig. 2.10. the input S'YR's. S.VRI and S.VRr corresponding

39

Page 52: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure 2.10: Tinie int egrat ing .-\O intercep t receiver. -4tlapted from [5] .

ro signals si ( t ) and s r ( t ) respect il-ely. are defined at the peak of the correlation as

where RI (0 ) and Rz(0) are input noise powers [A] and R,(O) is the input power of the

signals

s i ( t ) = s ( t - ta) + r l ~ ( t - ta)

40

Page 53: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

and

wherc for ivhi te Gaiissian noise R 1 ( t - f ) = n 1 ( f ) .

If B is the signal bandxidths ancl S.VRi < < 1. S-YRz < < 1. with the assuniption of

zero mean .\\I'GS in Eq. (2.56). it can be shoim that [-61. [29] if the peak of t ke correlation

occurs a t t = tu. we have

This allows us to define the processing gain Gp as

Thus for B = 50 l IHz and T, = 0.2 nis. the processing gain can be at niost 40 dB. If the

objective. for esaniple. is to actiicl-e S.\-Rn = 1 ( O dB) . i r e rnay choose to process to signals

each ivith S.YR = S S R i = S.\;R2 = -20 dB. or tivo signals. one with S'.\-RI = -30 dB

and the otker n-ith S.l-RZ = - 10 dB. This is an important conclusion tkat is of reh-ancc

to the practical use of the receiwr. Rewriting Eqs. (2.59) and (2.60) with ta = T/2 . ive

Page 54: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

i~liere s{ ( f ) is the noisy signal modiilating the laser. clelayecl by half the active A0 ce11

aperture T to ensure that the correlation is centered in the niiddle of the outpiit plane.

i d d e s!,( t ) clrives the A0 ce11 transclucer. The t ~ o sigrials cliffer only in their respective

adcli tive noise components.

The action of the CCD tletectors is to integrate the intemit' in Eq. (2.52) leacling to

Terms containing factors of cos(2;; fot J and cos(4;; for) are bantlpass ternis that intrgrate

to zero for integration tinie T, > > l/fu ancl have thiis heen neglectecl in Eq. (2.66). Soi\-.

terrn ( 1) contains a DC hias proportional to Ti. ancl a baselxmcl terni r u : 1 s;(t - r ) 1'. thnt

iutegrate to a VBT for large T,. while term ('7) contains the required correlation fuiiction.

Terxns ( 3 ) ancl (1) are the laser noise component and intermodulation proclucts respcctively.

The last term is noise power interna1 to the .-\O ce11 arising froni thc interaction of the

acoustic heam ivith the optical intensity [SI.

Page 55: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Applying Eq. (2.56) to the RHS of Eq. (4.66). it follows that

.1 .4?nii ni-, 3 ' 1 ' 2 S-YR3 = , Y 1 ~ { k , & } l 2 (2.61)

=Iqîn1 n l - y { ~ 1 R,,., ( T ) l 2 - 1 E{R, , y , ( . i ) } 12) + O; + O: + U; + O;-@ - 2

where the \-ariance of the correlation estimate r n r { ~ , , , , ( r ) } = E 1 R , , , , ( T ) 1" 1 -41 A2n1: E { R ~ , , ~ ( T ) } 1'. 0; =</ r l y E { - i i D ( f ) } j Z > and a: =<I r l ~ E{- l -LD( t ) 1 . s i ( t - Ï ) l2

} 1 2 > : R,, ,? ( 7 ) is the tirne average of the product . s ; ( t ) . s l , ( t - 7)' with < - > clenoting

time-averaging by the cletector array. Fiirtliermore. 02 is the niean of 1 I 1'. the electrical

noise voltage. times the number of cletector elements per correlation wiclt h wi-hile o& is cliie

to the niiiltifreqtiency diffraction component noise pan-er and the nonlinear acoiistic

component noise power .\É(s) as disc~ssed in (51.

Son-. with the assumption that n , ( t ) and n z ( t ) are zero niean independent 115th white

Ganssian statistics. Ise have that at the peak of the correlation function

Similarl~: the variance of the correlation peak estirnate is

Page 56: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

n-here it is assumed that the integrartion time T, > > 1/ B and u-here the effective noise

N o ) banclwiclt h is definecl as B = = JXX S ( j ) d f '>s'(O) : S( f ) is the noise pon-cr spectnirn [32].

- -G

Iiellman [36] o b s e r t d that the first term in Eq. (2.69) refers to the error in the correlation

esti~riation clue to finite integratiou while the last term is the error due to the additive

Gaussian noise terms n ( t ) and n 2 ( f ). For the correlator sholvn in Fig. 2.10. B a 12

where Blo is the acousto-optic 1)andwidth clefined in Fig. 2.3.

To include the laser iritensity noise. consider first t he noise cross term with mriance 4 in Eq. (2.6;). If -Y and 1- are an- ranclom variables. real or comples. n-e have the follov-ing

ineqiiali ty [33]

1 E { - Y I - } 1'4 E 1 -Y l 2 E 1 1- 1' (2.70)

Hence. rve set the iipper liniit on the fourth term as

since E { s 4 ( t ) } = 30.' = 3E2{ .5 ' ( t ) } = 3Ri(O) O for S-YRL << 1 and S.YR2 << 1. the

latter being a typical case in an intercept receit-er requiring an increase in the processing

gain of the signais.' i i . \ é can thus neglect the cross noise power contribution of terni ( 4 ) in

Eq. (2.67) for low SSR cases.

Furtherniore. for term (3 ) . a3 = < 1 qy E {-Go ( t ) } 1 2 > can be nornializecl with respect

to the correlation signal power of term ( 2 ) resulting in

?.-\ssurning Ciaussian noise statistics.

Page 57: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

wliich ive recognize bu definition. as the relative irrtemitp noise ( R I S ) for a laser cliocle [30].

11-e assume that riiode partition noise is negligil~le and that al1 lasirig modes are eqiially

tletcctable ivitk none of the emitted light reflected back into the laser. n l e n niode partition

noise is prescrit. each lasing mode generates a photon nuniber S, witk its respective R I S

refereucecl to the control mode So as

ivliicli is siniply an additive component to ECL (2.72). llethocls for rneasuring the R I S for

clirectl?- rnodiilated laser diodes have been dernonstratecl by a nitniber of researchers [3l]. [34].

R I 5 values are often providecl by laser diode manufacttirers and are generally specifiecl as

a fiinction of system bandt~-idth. ivhich in oitr case is B.

Conihining al1 the results. the output SSR. after integration. is thiis gi\-en hy

where 3 = 20 BT* wit h ci the nuniber of detector elements per correlation ivitlt h and where

parameter X represents the nornialized error due to finite integration and is giwn by [26]

witL typical values in the range O 5 X < 1. In deriving Eq. (2.75). it is iinderstood that

the signal bandwidth B is eclual to the noise banclmidth throughout the system.

Page 58: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

i\é identifj- three specific cases for analysis here: ( a ) With our original assuniption of

moderately low STR intercep t signals (S.VRi. ST Rr < < 1 ) . ive have the further approsi-

mat ion that

The dynamic range (DR) of the processor. definecl as the ratio of the peak correlation

output to RAIS noise. is given bl-

trkere 1.; = r , ~ . - I ~ . - L ~ r n m.,/?. is the peak voltage of the correlation signal. The processing

gain of the processor. for these sniall input SSR's. can thils be summarized as

wliich is ec~iiivaleut to the simple case of a noiseless laser source with the adclitional as-

sumption of a noiseless acoiistic and optical \va\-e interaction in the A0 cell. Hence. the

1011- SSR condit iou tends to de-ernphasize t ke effect of laser intensi ty niotlulation noise

leacling to the faniiliar forni @\-en I>y B a d a et al. [PJ]. This ecluatioli sets the uppcr botincl

on the processing gain of the recei\-er as dictatecl by r he finite lalue of the DR. Horvel-er.

the overall DR and processing gain of the receiver are further restricted bu the saturation

voltage of the detectors. A high attendant L-BT tends to saturate the detectors resirlting in

a reduced gain. For esample. we sliowed in Eq. (2.63 ) t hat for B = 50 MHz. Tt = 205 p rris.

and DR = 500 and with the assumption of 10 detectors per the length of the correlation.

the resultant processing gain has a limiting value of about 40 dB. This limit is a ftmction

Page 59: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

of the hanclwidth of the signals (B ) and the integration time (Ti). hoth of which can be

mrietl. In practice. the CCD array imposes the ultimate ceiling on the processing gain of

the receiver due its own DR limitations. clefined as the ratio of the saturation 1-oltage of

the array to the detector peak-to-peak noise (b>;LT/\'b-P ). The CCD array usecl in this

work had a liniiting DR = 2600. ancl theoreticallu. GS dB of processing gain \vas achielxble

n-i t h t his receit-er.

This gain ma!- not be sufficient for \+ery low signals in a high noise environnient and is

further reduced by other interfering signal contributions in the bandn-iclth that tend to acld

to the bias of the output. These aclclitional contribiitions inclticle .\O ce11 intermoclulation

prodiicts as well as laser niodulation artifacts as in the general case 11-ith R I .Y # 0.

( b ) For a noiseless refercnce case (S-V R2 + x ). as in in a search ancl itlentify receircr.

In such a scenario. the intensity mocliilation noise of the laser conies into play. Both the

interisi ty niodiila tion noise and the fini te integration paraniet er X iriversely affect the 01-erall

processing gain. -4 plot of the output S.1-R3 versiis the intercepted signal poiwr is given

in Fig. 2.11.

-4s the input S.VRl increases. the output SSR.3 increases asyniptotically to its linii t ing

value approximated by 1 / ( A + R1.Y). Howewr. for very large input SSR \dhliies. the effect

of the laser intensity noise becomes niore dominant as terni (4) in Eq. (2.66) is no longer

negligible. This gives rise to the nest case.

( c ) For S S R i . S:VR2 >> 1. Ecb (2.71) no longer ~xnishes and so we may espress the

Page 60: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

1.99 la al 3a 40 !5a M m

S ( R 1

Figure 2.11: For the case when S-\-R2 -+ x. The output SXR is asymptotically I~ounclecl by 1 / ( A + RI.)-) n-here tj-pical values usecl in the siniiilation were DR=5OO. n = 10 tletec- tors/correlation width. T, = 205ps. B=5O MHz. R I T = je-"/B and X is e\-alii.îtecl from Eq. ('2.75).

output SSR as

for sufficient ly large signal- to-noise ratios. i t is reasonable to approsiniate t ka t % ( O ) 5

SSR: where S-YR: = R:(O)/S:(O) and from Eq. ('2.2). ive set the upper boiind on

acliie~-ahle output SSR in this case as

where we have incorporateci the additional assurnption t bat the two autennas are located

in a statistically similar noise environment such that S-\'Ri % S S R 2 . The laser iritensity

niodulation noise becomes a significant limitation on the processing gain of the system.

From Eq. ( 2-81 ). we see t hat t Le laser intensity noise is aniplified by a factor corrcsponding

to the SSR. The higher the SXR, the more the influence of the intensity nioclulation noise

Page 61: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure 2.12: Output S3'R for S.l'Ri. S S R 2 >> 1

is reflected in the output throiigli a rechction in the processing gain. Figure 2.12 shows

the degradation in the processing gain as the input SXR is increased.

Returning to the low input signal-to-noise ratio case in ( a ) given in Ecl. ('2.7s). we can

proceed to iclentif>- three suh-cases similar ro Ristic and Lee's analysis in [j]. For very

videb band and short A0 t ime aperture. the noise power cout ribut ion froni t lie nonliuear

acoustic cornponent .VE(r) is negligihle resultirig in a larger processing gain.

Similarlc for long time apertures and wide bandwiclt h. the niultifrequency diffraction

component is sniall relative to .\É(.r ) and can tlitis be ueglected in Eq. (2.73) leading

to irnproved processing gain performance due to a reduction in the overall hias. Since

.YD is approsimately equal to the poiver of the cliffracted beam (-1'~ zz I';l. wtiere Id is

the intensity of the diffracted light beam with cross-section area -4). it is evident that a

strong .\O ce11 modulating signal will inadvertently reduce the overall processing gain of

the receiver. Tlius. the dynamic range limitation is aggra\-ated bu the presence of these

noise terms which add to the signal dependent bias term of the correlation output. Hence.

the overall processing gain can be increased significantly by ~ising cletectors with higher

saturation values and relatively loir SSR intercept signals.

Page 62: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Chapter 3

Experimental Investigation

In general. a correlator serves two piirposes: ( a ) ctetection of the presence of a signal ancl ( b )

spectrtim analysis of the cletected signal. Each of these tivo applications usually reqiiires

a difÏerent architecture [l-l]. [?Il. [Z!. [?-LI. The objective of this thesis. as esplained in

Chapter 1. \vas the detection of the presence of a signal.

In carrying out the esperimental investigation. two -40 deflectors were tisecl. The first .

nianufact iireti b!- SEOS. was used because i t was iriimediately aiailable in the laboratory.

It had a handu-idth = 40 M H z centerert at feqo = SO .\IH- n-ith a delay tinie

of T = 10 pa. The resiilting in-line A0 design. clisctissed in Cliapter 2. and shorr-n in

Fig. 3.1. protluced a correlator witli signal bandrvidth B = 20 .\IH= and integration tinie

of Ti 0.2 ms.

ii.\'ith reference to Fig. 2.3. the system freqitencies were: fu = 30 -11 H z . f = 90 JIHi

and f, = 60 J I H - . This correlator mil1 be referreci to as the 20 JIHi signal baridwiclth

processor. -4s shoivn in Insets A. B and E. the tivo input signals s, (t)cos(2a fot) and

s2(t)cos(2~ fot ) mere used to simulate pulsed C W radar signals in the subsequent esperi-

Page 63: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

DC Bias CoILiniat

Seri t ral Densitv Filter

k

CSSB 4) Slodulator

cos(2r foi) cos (2 nfct )

Figure 3.1: Reference tone .IO processor for radar signal processing. Simiilated radar signals si ( t ) cos(2;i f,t ) and sz ( t ) cos (2r f0 t ) are introditced to the processor as shown. CSSB is an upper single sideband modulator that is used to translate the input signal to f, + fo. and to remove signal bandividth redundancy by passing only the iipper band as shown in Inset C. B = 20 M H z is the signal bandividth. and = 40 M H z is the bandwiclrli of the - A 0 cell.

Page 64: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

mental work carriecl out in t bis section.

In the arcliitectiire shown in Fig. 3.1. the focal length of hoth lenses L 1 and L2 !vas

3s. 10 nini providing for an overall magnification of F2/ Fi = 1. The effective Lmgt h of the

A0 ce11 was 6 nim. hence the niinimum reclitirccl CCD array length to niatch this aperture

Iras F2 x 6 rnm/FI. The CCD arr- used in this esperinient hacl 20-1s pixels each of

dinielision 13 ltnl x 13 /<ni a d a reatlout rate of 10 .\IHz. This translates into a total

length of 13pni x 20% 2 26.6 niiii IL-hich is weil above the niinimum reqiiired to meet the

SJ-quist smpling constraint .

3.1 Detection of Signals at the Spatially Integrated

Output

At this point it ivas instructif-e ro look at another wa?. of detecting the signals that cloes

not in\-olve CCD arrays. In this approach. a beani splitter \vas placecl nt plane P? as shoivu

in Fig. 2.7. The beam propagating downwarcls was focussed on a large arca photocietector

by a ivide angle Fourier transforni lem. Tlie oiitput of the detector !vas fecl into a digital

sampling oscilloscope for analysis. -4 mathemat ical de r in t ion of t ke out put of the p h -

todetector in siich an arrangement was given Section 2.5 . Here 11-e \vil1 coucentrate on the

esperirriental mlidation of this resiilt. Figures 3.2 (a) and (b ) illustrate the same signal

detected by (a ) a CCD arra!.. after tinie-integrat ion. and follo~vecl bu coherent detection

which results in the envelope of the correlation and ( b ) a single photodetector. after spatial

integration. that is connected to digital sampling oscilloscope.

This technique niakes fast identification of high power interferers possible by removing

the waiting period associated with the CCD integration time T, zz 0.2 nis. Tkese narroiv

Page 65: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure 3.2: ( a ) .A correlation peak envelope clctected by the CCD array. ( b ) tlie sanie signal with a tinie varj.ing carrier detected at the spatially integrated outpiit .

band interference signais. taken from the spat ially integratetl out put. are Fourier t ranc

fornied. tlieir frequencies identified. and a cornniancl issiied to a bank of cligirally tiinable

notch filters to escise tkeir respective enegies. Details of this SIS notctiing tecli~iic~iie are

given in Sections 3.11 and 3.12 of this thesis.

3.2 CCD Scaling Law

Consider a sample output of a typical correlation of tivo pulses each of cliiration t'. The

total time duration of the correlation triangle iri the tletector plane is W . Howewr. if we

plot the correlation as a function of r. its cliiration .cf ~vill be clilated by a factor of r : tlie

velocity of the acoustic wave in the A 0 cell. As well. other factors can contribute to the

change in spatial dimension of the correlation: na~riely the niagnification of the irnaging

Page 66: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

optics. eqiial to one in oiir case. and proper alignment of the ce11 aperture to the CCD

cletector array. This spatial correlation is sampled by the CCD pisels at a rend-out rate

of 10 -11 H z . LVith the correlation width of x'. the number of pixels to (.ol-er this ividth is

2' - 4 - where O is the length of each pixel in the a r r . The resiilting ividth of the correlation

z' in the time domain (as viewed on the oscilloscope) d l thiis he given hj- t" = l , o I H z 1 6 .

Since the CCD pixel dimension \vas 13 prn x 13 prn it follows thar t" = &. The original

tinie correlation rvas rnapped into the spatial domain t hrough 2t' = 5. hence. the observer1

correlation 11-iclt h tu is related to the original tirne correlation hy

Therefore. the ohserved correlat ion d l have i ts base dilat ed by a factor of = 5 relative

to the original [%]. [36].

3.3 Correlation of Narrowband Signals in White Gaus-

sian Noise

The output of the processor for the simulated narrom band pulse radar signals is sholvn

in Fig. 3.3. The farniliar triangular shape of the waveform is consistent with the elcpected

shape of correlating tivo square pulses. The base of the maleform should be twice the size

of the pulse duration. However. due to imaging considerations and rhe nature of the spatial

correlation when CCD detection is used. it appears dilared as explained in the previous

section.

The simulaied radar pulse used had a pulse duration r = 1.2 ps and a pulse repetition

Page 67: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure 3.3: Correlator output for r = 1.2 ps. PRF = 0.1 -1lHz. The optical center rvas establishect ancl maintainecl at the 22 11s time location for al1 siibsecluent esperinients in tbis section.

frequency (PRF = 0.1 -11 H z ) . Since the A0 ce11 transit tirne T eqiials the pulse repetition

periocl. only one correlation can fit into the active ce11 aperture during each PRF cycle as

shown in Fig. 3.3.

In Fig. 3.4. the PRF iras increased to 1 M H z while keeping the pulse duration at

1.2 p s . In this case. ive have a number of correlation peaks fit ting witliin the 10 p s -40 ce11

apertiire rvinclow since the period of the sirnuiated radar signal is now 1 p. Theoretically.

tve should be abie to fit about ,.,s+ , z 4 correlation peaks. Sote the roll off in the P H F

peak amplitude away from the center of the .\O cell. This ma!. be attributcd to a drop

in the cell' s diffraction efficiency close to its eclges reducing the power of the correlation

peaks at t hese locations.

To e d u a t e the processor' s performance in presence of noise. r~liite Gaussian noise ivas

added to both signals being correlated. Fig. 3.5 shows the processor output for varying

SSR values.

Page 68: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure 3.4: hliiltiple Correlation for PRF = 1 -1IHz.

The first noticeable effect. as the noise was introduced. was the hroatlening of the corre-

lat ion base. This broadening was inversely proportional to the SSR. This effec t iras due to

the fact that the noise was not time-limitecl to the pulse duration Ï of either sl ( t ) or h ( t )

and hence the correlation of the composite signal vanished outside of tirne period 27. In

addition. the noise acted as a DC bias on the output. progressively pushing the correlation

peak into the saturation region of the CCD array as shown in Fig. 3.5(tl) [35].[36].

Page 69: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Cc)

Figure 3.5: In al1 cases the PRF = SO I iHz . 0.5V/DIV and tirne base of 20 p s / D I I - (a)

S-VR = 20 (b ) S-YR = 6.4 ( c ) S.\-R = 0.7 (d ) S-VR = 0.4 The pectiliar shapes are due the noise component as i t drives the correlation envelope into the saturation regions of the CCD array.

Page 70: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

3.4 Opt ical Misalignment Estimation

Csing the resiilts of the previous sections ancl a white Gatissian noise source. a procediire

was devisecl to estimate the overall optical misalignment in the processor. Because white

noise contaius all freqiiencies in eqiial proportion. it is a convenient signal for systern

rneasiirement s since i t covers any chosen signal bandwiclt h full>-.

Suppose that the signals s l ( t ) and sG(t - T ) are replaced by a white Gaiissian noise

sarnple function n ( t ) and its delayed version n R ( t - r ). The tirneaverage correlation integral

in Eq. (9.10) can be ivritten as

1 " n ( t ) n œ ( t - r ) d t = - Ti 1% 1 -Y( f. T,) l 2 &TfTi ( I f

where a truncatecl sample segment of the noise process n ( t ) . clefined over the integration

range O 5 t 5 Ti with short-time Fourier transform .Y( f. Ti ) given by

If this white noise process is then passed though a band-pass filter of center frecioency k

ancl qualit- factor Q. the spectral density Ss. for Q > > 1. is given by [32]

*Y, / 2 l + _ i Q 2 ( f - I C ) ~ / / ?

for f > O S d f ) = .y, / 2

~ + ~ Q 2 ( f + f c ) ' / f ~ ot herwise

and .\o/2 is the peak power spectral clensity.

Page 71: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

The time averaged correlation is thus given by

where 3-' {. } clenotes the inverse Fourier transform operator. Subst ituting the correlation

function given by Eq. (3.5) in Eq. (2.9) yields

n-here it is understood that the CCD discards the carrier information and hence R , ( Ï )

represents the envelope of the oittput plus bias.

A ivicleband white Gaussian noise generator was used to protide the input signals.

.A pair of banclpass filters ivith center freqiienc- equal to the clifference frequency fo =

30 M H z and bandwidth B = 50 I ï H z for a Q = fo /B = 600. provided inputs shown

in Inset .A and B of Fig. 3.1. Such a high Q-factor was chosen in order to 'sharpen' the

correlation response peak and provide a niore accurate measurement of the misalignnient

.Yo / 2 error. Furt hermore. for an arbitrary Q-factor. Say( f ) = [ f c l l 1 2 and its inverse

Fourier transform ( R';(r) = F1 {Sn( f ) } ) cannot be solved in closed form ivit hout a large

Q assumption as in Eq. (3.5). This is an important consideration because the methoci

proposed here requires the existence of a closed form espression for R X ( r ) which is then

used to fit the experimental data and subsequently estimate the alignment error.

The resul ting noise- only correlation data. after bias r e m o d , ivas then fit ted niimer-

ically with the correlation mode1 given in Eq. (3.6) iising a least squares fit ivitli Gauss-

Page 72: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Sewton optimization of the cost function. The residtials of the fit ting function w r e then

used to calculate the \ariance and confidence in tends of the fit. This error ivas partly

at tributecl to the optical misalignrnent in the system which introciuced clistortions in the

correlation envelope and was estimated by ~napping it to the spatial domain using the CCD

scaling latv given in Eq. (3 .1) .

Howeïer. additional factors. predominantly CCD t rapezoicial spatial responsc effects.

system aperture function and detector nonlinearities. ais0 play a part in the size of this

error. The apertiire funcrion component is intrinsically tied to the overall processor mis-

alignment problem and hence is assumed to be accounterl for in the estimated displacement.

CCD trapezoidal spatial response effects d l introduce errors in estimating the location

of the correlation peak. This error mas Iumpecl together with other anomalies by fitting

the data of a 'perfectly' aligned system. The resulting fitting error was then assurnecl to

inclilde the spatial response effects and becornes the limiting kalue of the accurac- of this

approach. More on this later.

For situations where the operating range of the processor includes the non-linear regions

of the CCD response cun-e. the output data m u t first be compensated prior to the fit ting.

If R ( r ) represents the correlation of signals sl ( t ) and s2(t). the CCD detector response

is then giveu by g [ R ( r ) ] . where g[ - ] is the detector nonlinearity [37]. This amounts to a

dynamic range compression of R ( r ) . In theory. if the nonlinearity is known and nionotonic

then we ma' find its inverse g-l [el' and the inverse can then be used to recover R( r ). CCD

dynamic transfer plots are usually provided by the manufacturers and the inverse g-l[-]

can readily be obtained from them to linearize the output.' Once the non-liuearity has

'Sote that in niost cases. this step is unnecessary as the CCD array is normally operated in its linear range thus avoiding over-drive and sat utation.

Page 73: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

been removecl. the correlation data is then fitted with the second term of Eq. (3.6) as

shown in Fig. 3.6.

The error in estiniating the displacement between the data and the fit dong the r -

asis was used to calculate the overall misalignment by mapping. using Eq. (3 .1) . its tinie

domain represent ation. as sampled by the oscilloscope. to i ts corresponding spatial clornain

image while also compensating for the overall dimensional dilation relative to the original.

In the esperinient. the laser-beani optical asis n-as ixried with respect to the .r'-asis

of the CCD plane Pl shown in Fig. 3.1. The amount of misalignment so introduced n-as

foiind to be approsimately equai to that from the fit estimates. This then beconies an

independent check on the accuracy of the method. The correlation data dong with the

corresponding fits for mrying degrees of misalignments relative to the optical asis is shown

in Fig. 3.6. The various estimates of the misalignment errors. for ~xrying laser-beam offsets

from an established optical center. ranged from 0.05 nim to 1.2 min.

In theorc the accuracy of this method is boundetl b). tLc degree of confidence in the

fit ting algorithm usecl and bl- the size of the pixels in the CCD array which determines the

spatial resoliition. The array used in this work consisted of 13 /lm length pisels. Hence.

we cannot resolve a rnisalignment less than this magnitude. The trapezoiclal aperture

response of the array sets this limitation. Figure 3.6(a) was usecl to estimate the acciiracy

of this method. In this case. the residual error was obtainecl by fitting the data for the hest

possible alignment of the laser beam. Hence. 50 prn is the best accuracy of the alignment

error that can be obtained mit h this particular esperi~nental setup. -4s such. this beconies

the uncertainty associated will al1 subsequent rneasurernents.

Page 74: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure 3.6: Fit of the noise only correlation data (represented b?- t x x x 1 . ) the secoud term of Eq. (3.6). The estimatecl alignmerit error \\-as: ( a ) 0.05 mm ( h ) 0.3 k 0.05 mm ( c )

1.0 f 0.05 mni (d ) 1.2 k 0.05 mm. Horizonral asis: lag position in seconds. ièrtical scale: arbitrary.

Page 75: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

3.5 The 50 MHz Signal Bandwidth A 0 Processor

In practice. the minimuni instantaneous buidwiclth that is acceptable in radar signal pro-

cessing is B = 0.5 GHz. Radar signals of interest are usually in the frecluency range

from 0.5 to 26.5 GHz. Tuners that c m translate any banchidth B 5 0.5 G H z in this

frequency range to an IF output of 0.75 to 1-25 GHz have recently become amilable.'

In terms of the ;\O ce11 of the in-line architecture. the reciuired ;\O bandwidth m u t be

3 2B = 1 GHz as discussed in Chapter 2. Furthermore. in order to al-oid two-tone

interniodulat ion prodiicts ( IMPs) . the center frecpency fOo of the ;\O ce11 must sat isfy

fOo 2 l.jB,io = 1.5 G H z .

The design of the multiplicative radar signal optical processor also requires. in orcler to

al-oid spectral overlapping in the detector plane. that the clifference frequency f, 2 1.5 B =

0.75 GHz. If me choose to place the reference tone frequency at j:*. as sho~vu in Fig. 2.3.

then we have that f, = Po = 1 GHz. Before being applied to the A0 cell. the signal

bandwidtli B is upshifted to the center frecluelicy f r F = f, + fo = 1.7s GHz in orcler for the

signal spectrum to be situated between f; = 1.5 GHz anci f: = fto = 2 GHz. Thus the

upper cut-off frecpency of the upshifted signal fi coincides. nsing the solution in Fig. 2.3.

n-ith the upper cut-off frequency of the .-\O cell. To sumrnarize. the characteristics of the

A0 ce11 that are required for the prototype construction of this processor. are:

-10 bandwidth: 1 GHz

A 0 center frequency: 1.5 GHz

Present A0 technology allows dwdl time for suck an XO ce11 of about T = 0.4 p. This

results in time-bandnidth product of TB-4o = 400 that is sufficiently high for envelope

' For instance, LVatkins-Johnson 1V.J-TX-170 tuner, or Condor T3-6 18.

Page 76: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

signal processing in the detector plane.3

The real life mode1 just described in the prececling paragraphs. iras freqiiency scaled

clown 10 times in order ro arrive at design characteristics of a custom tlesigned A 0 deflector.

nanielu

-10 bandwidt h: 0.1 GHz

A0 center frequency: 0.15 GHz

The Brimrose Company was cont acted to procluce t his cus tom -40 deflector designatecl

as TED- 150-100. wit h drive RF power amplifier designatecl as Pal- 150-100-2-2211- O t hcr

relelant design parameters of the deflector were the transit time T = 5 { r s . optical aperture

0.75 ni m x 21 n2 rn ancl an acoustic \.elocity of I* = 4200 m/s . The system freqitencies were

fo = 75 M H z . f, = 100 .\IHz ancl f r F = 175 M H z . This 30 J I H z signal bancln-idth

processor was then used in the LPI esperiments that follow.

3.6 Methods for NIS Detection and Removal

.in important consideration in the LPI signal cletection problem is the rejection of high

power SIS. The 371s autocorrelates and dominates the compressed LPI radar pulse wave-

form at the detector plane. This is illustratecl in Fig. 3.7. shotving the effect of an attendant

SIS on the correlation. In (a ) . the SIS is locatecl outside the LPI banclrvidth leatling to a

correlation peak that dominates the clesired signal. The case of an in-band SIS is shown ii i

Fig. 3.7(b). In this case. the desirecl correlation envelope is distortecl by the autocorrelation

of the XIS energy. In a typical intercept radar receiver situation. the high TIS p o w r can

Tp to 1 p s is available using acoustic longitudinal niodc. Ho~vever. severe attcniiatian at CiHz fre- quencies lirnits the usable tinie delay t o about 0.5 p.

Page 77: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

lead to fdse cletection by owrpotvering the desired correlation signal. Shus. the r e m o ~ d of

the SIS before processing the wicleband LPI radar signal is of critical importance [37]. (381.

- . .. #

. . l . , , . , , . - . . . . , . . . . . ; - . j.. ..; . ; . . . j . . . ' . . . . a . .

.: . .,-. * : r

't

...

Figure 3.7: Oscilloscope pictures of (a) XIS outside and abo1.e the LPI correlation ivaveforrn and ( b ) inside the LPI corrclation envelope. Oscilloscope set tings: SOOmVJDIV. 5Olis/DI\-.

Two methods were used for detection of the SIS in this thesis. One methocl used

a feedback loop originating. after integration. at the post-detector circuit. as shotvn in

Fig. 3.S. This methocl is slow and t ~ a s first described in [39].

More recently. a real-tirne feedback scheme was proposed and demonstrated by Ward

and Vanderlugt [3S]. In this methocl. an optical Fourier transformer is iised to detect the

frequencies of the narrowhand interferers. The energy of the interferer is then no tchecl in a

feedback arrangement using a spatial light modulator (SLII). The methocl has a relatively

slow response time of about 5 m s due to the CCD readout rate and the speed of the

Page 78: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

post -processing elec troriics.

In this thesis. ive use a feeclhack loop frorn the spatially integrated output of the pro-

cessors described in Section 2.5. followed by an electronic Fourier transformer to itlentify

XIS frequencies. The resulting feedback notcking scheme promises to be fast ( << 100 p.<

response tirne)" and robiist. This method is one of the original contributions of this thesis.

A schematic block diagram of the processor in ternis of its op t ical anci electronic modules

is shon-n in Fig. 3.S. The major suhcomponents of the layout in Fie. 3.8. designecl and

assembled. in t his work were: ( a ) spread spect riim LPI generat ion circiii t ry ( b) correlat ion

cletec-tion and cligitizing mocliile (c) XIS cletection ancl excision subsysteni (c i ) laser and

colliniation optics and (e ) imaging optics. Item (cl) is cliscussecl in -4ppendis A. while the

ïes t are presented in the sections that follow. These modules are further ill~istrated in the

photographs given Chapter 4.

Furthermore. two types of signals were used to siniulate the LPI signal: ( a ) the lintar

FlI chirp and ( b ) the freqiiency hop signal.

" W t h a custom designed chip for FFT operations and SIS identification.

Page 79: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

RS-232 CO JI Jl Cornputer GPIB interface (IE'EE -188) c Control

t rigger

TEK.506

VCO

SIS escision filter barik

sptitter l I\

pu Ise generator

DC bias .\O ceIl L 1

Laser

Seutral Density Filter

i(

dela? line

diode

Collimat optics T

C'CD array

band pass cont rol Bri tnrose

P.\- 1.50- 100-22 IO

generator splittcr filter spli tter \C'JH'>

l A to oscilloscope

digi tizer

TIi:I!lO.-\ D t

Figure 3.8: Ovcrall optic and electronic hardware layoiit of an acous to-optic processor tor LPI detection ancl classification. L1 and L2 are part of the iniaging optics discussed in -4ppendix A.

Page 80: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

3.7 Simulation of the LPI Signal

3.7.1 Linear FM Chirp

In general. a linear FM chirp signal can be characterizecL by

wkere a is the chirp rate in H z / s e c and Tc is the ckirp ditration. The instantaneoiis

frequency f, within the chirp hancln-idth is given hy

B This instautaiiecus freqiiency stveeps over the signal banclwitlth B at rate n =

The chirp signals tvere generatecl froni a fast IIini-Circuit POS-100 voltage cont rollecl

oscillator (l'CO 1. The \-CO sweep cliiration and estent were deterniinecl hy a ranip signal

suppliecl 1,~- an esternally triggerecl Il%\-etek 1% generator as shown in Fig. 3.9. The time

duration of the ramp determines the chirp cluration Tc while its voltage dynamic range

(escursion) sets the lowest and highest freqiiencies of the chirp banclwidtli. The \-CO hacl

a timing voltage of 1 to 16 C.* requirecl to cover the 50 JI Hz LPI bandwidtli ( a frequeucy

sweep of JO - 100 3 1 H z ) . a tuning sensitivity of abolit 9.4 .IIHz/L.- ancl a 3 dB mocliilation

bancln-idth of 100 IiHz. The chirp repetition rate (CRR) \vas determinecl by the oiitpiit

freqiiency of a Tektronir FG5Ol fiinction generator that provided an esternal trigger to the

HP pulse generator (HPSOlPB). Operated in this mode. the pulse generator output coiilcl

he varied independent of the pulse train frecluency. leacling to long dwell tirries betwen

t rigger pulses. Instrumentation linii t ations led to the relocat ion of the difference frequency

Page 81: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

FG.50 1

TEK Gerierator

TRG W R R Ranip HP Pulse ni C;eiierator

HP 1053A double balanced miser

HP Function Cienerator ZHL-3. l l$!JS* 'O ' O ceil

h

Generator

llirii-circuit Z >[SC-2- 1 B Conibi ner

A ZHL-3:l fVV\ = I I P w

Figure 3.9: Chirp generating circuit configuration.

SIini-Circuit POS- 100

\-CO HP80 1% iiavetrk 185

Page 82: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

f o fc f.40 = IF Figure 3.10: (a) The proper rclationsliip of processor frequencies and ( b ) the processor frequencies used due to instrumental limitations.

Page 83: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

fo to the loivest freqiiency (50 11 HI ) of the LPI 11-weforni which led to an frF = 150 JI H z .

as shorm in Fig. 3.10. This re-arrangement of frequencies allorvs for a uat ural sweep of the

spread spectrum bancl from loiv to high rvith the VCO.

3.7.2 Frequency Hop

-4nother popiilar LPI spread spectrum signal is obtained by randomly kopping a chta mod-

ulated carrier froni one frequency to the nest. In effect. the spectrum of t ke transmit ted

signal is spreacl sequent ially rat her t han instant aneouslÿ. This leacls t O a pse~itlo-randoni

ortlerecl sequcnce of frecluency hop .

.Analytically. the transmitted signal is giveu by [Z]

w herc

t - a 1 for « - 612 5 t 5 a + b / 9

O otherwise

rvhich describes a srclilence of .\I frequencies {f,} = f i . . . . . f,, . f-\[ each of chirp cltiratiou

t u correspontling to mto 5 t 5 ( m + l ) to . centerecl at t = ( m + f )t,. This sunimation

describes a secluence of frequencies used in the clifferent tu time dots.

-4 cornputer interface to a freqiiency synthesizer shown in Fig. 3.11 was osecl to randomly

generate freqitencies nithin a pre-determinecl banclwidth of 50 .\IHi resol\-ecl iuto 300

frequency bins. The LPI signal \ras then itseci to characterize the processor' s cletection

performance as cLscussed in a subsecliient section.

Page 84: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

data Signetics

t I

I

I I

F S 1' / r wave I

I I

FSIi Barid p a s I 2-

Filter t

t I C'-50- I --I B B I

I I

Freqiiency r-- 1 Signet ics

HEF-li.50\'

Corn pu ter generat cd random P S code

Figure 3.11: llodificatious of Fig. 3.8 to acconimodate freqiiency hop generation. The frecluency synthesizer is clriyen by a pseudo-noise (PT) code generatecl from a cornputer. The oiitput frecpenc- is hopped within the sigrial banclwidth which is then niised with tlie baseband data

Page 85: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

3.8 LPI Correlat ion Detection Circuit Requirements

The correlation signal was detected iising a CCD array. Each eiement of the detertor array

beliaves as a photocapaci tor on ivhich an electric charge accuniiilates proport ioual to the

square of the incident light and to the integration time. The array is reacl out seriall- by

transferring the charge frotn a CCD structure. Invoking Sycpist sampling tkeorern. the

ininimum number of detectors needed to adequately sample the correlation signal is ttvo

for each period of the highest spatial fretluenc-

The correlation signal at the detection plane P3 ivas shown in Eq. (2.10). to he given

by :

n-here ive have neglected the bias terni. ancl f ' = fo + BI2 is the highest frequency in

the correlated LPI signal. One cycle of tLis freqiienc- lias cluration T' = +. To record

this signal. according to the sampling tlieorem. Ive neecl pisel spacing eqiial to or less than

O.5T'. Thus. for an -40 ce11 tr-ith total aperture T = 5p.q. the number of pixels iri the

. ~ p y e c -- cletector plane must be at least m. For fo = ta J IH2 . the spatially ~arying carrier at

f f = 100 -1IHz recpires a minimum niimber of pisels eclual to 1000 which is i\-el1 rvithin

the arrq- capability used in this work.

As shotm earlier. to separate the bias rerms from the rcqtrired correlation. the integra-

tion time must be much larger than the difference period i. The integration time ( T t ) of

the CCD array used in this mork mas 2048 + & 0.2 rnsec. Hence. al1 ternis containing

-- co.i;(%r fot) or C O S ( ~ ; ; fat). fo = , a M H z . integrated to zero since Ti is substantially larger

s T than - = 13.3 nsec. The peak of the correlation occurs when r = O = + 7 or at

Page 86: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

.r = - 7. To mow the correlation peak to the center of the c d . a delay of approsimately

= 2.5 ps was requirecl between the tira signals being correlatecl. This dela>- i.as realizecl

electronically iising the piilse generators ' biiilt-in delay network.

3.9 Processing Gain and Detector Integration

In Section 2.7. we bat-e defined the processing gain in Eq. (2.62) without coherent i n t c

gration of niiiltiple radar returns within time span Ti. In the prcsence of coherent detector

integration [SI. the processing gain can be written as

ivhere Tp = 1IPRF is the interval between the radar pulses and KIT, = .\- is the nimber

of radar pidses integratecl. In terms of clecibels. this can be writ tcn as

Since Ive are stuclying autocorrelation. S S R i z S.\-R2 and ive choose to define the inpiit

SSR in dB as

For the 20 .II Hz signal handmidth processor with B = 20 -\{Hz. T, = 0.2 rns and

Tp = 10 11s. the maxinium processing gain. as giwn in Eq. (3.1 1) is 49 dB. \l'hile for

the 50 WH: signal bandwidth processor witù Ti = 0.2 rns and Tp = 10 p s . the masinium

processing gain is 53 dB.

Page 87: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

3.10 Correlation of LPI Signals in Noise

The numerical results of Sections 2.2 and 2.3 were further validatecl esperimenrallc -4

50 M H z bandwidth optical LPI correlator 11-as assemblecl and usecl to verify the princi-

ple of operation as cliscussed in Section 3.5. .A widebancl white Gaiissian noise generator

was also used to corrupt the system commensitrate with the loacling anticipated in wire-

less communication channel. The bandwidth of the noise \vas bandlirni ted to the same

bandwidth as that of the LPI.

As predicted and esperimentally verified. the correlation was a sinc envelope riding on

a spatially ~xry ing carrier for both chirp and frecliiency hoppetl LPI signals. -A niimber of

factors were foiind to pl- a role in distorting the shape of this en\-elope. These incltided sta-

bilit y of the frequency sweep hetween chirps. detector non-lineari t ics. phase nou-linearit ics

of correlating signals. and stahilitj- of the optical setiip. Detcctor non-linearities were coni-

pensated for by post-processing the correlat ion clat a. while optical art ifacts were foiincl to

he minirnized by proper alignnicut procedures using the approacli developecl in this work

and discussed in Section 3.4. The setup yielclecl reasonable processing sain and scrisiti\-ity

to low STR signals with goocl noise immiinity characteristics commemurate with practical

radar operating environments. Even for a very low input S S R of -50 dB. the correla-

tion peak was still detectable for the case of a reference signal correlation as shown in

Fig. 3.13(a) with a processor gain of about 50 dB. This result clearly denionstratecl tlie

capahility of the processor in tletecting very weak sig~ials.

Similady. the theoretically predicted relationship between the main lobe of t Le corre-

lation envelope and the chirp duration \vas es~er iment ally demonst rated by the results

shown in Fig. 3.13. Compared to Fig. 3.12(b). it is clear that tlie correlation wiclth is cli-

Page 88: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure 3.12: (a ) Processor output for a linear FI1 signal (SSR of -30 dB) and ( h ) the reference cliirp: vertical scale. 0.5 V/DIJW. horizontal scale. 5Op.s/DI\'.

rectly proportional to the chirp diiration as pre~ioiisly establishecl bu nunierical sinidation

and siimrnarizecl in Fig. 2.6 In this case. the chirp diiration \vas 1.5 tinies the cluration

usecl in Fig. 3.li?(h). The sinc envelope in Fig. 3.13 shoivs a proportional incrcase in the

width of the correlation. [ X higher chirp dtiration results in a wide correlation i d t h : cf

Figs. 3.12 and 3-13.]

Figures 3-14 and 3.13 show the progressi\-e effect of ~xry ing S S R levels on the cor-

relation for chirp and frequency hopped signals respectivel- It is clear from the results

that the processor is capable of operating in relatively noisy environment and still pick

out hidden radiating LPI sources with iip to -50 dB of input STR. f i r y often hoivever.

tliese LPI signals are not only masked by noise but are also overwhelmecl by s trong nearby

interferers. This additional energy greatly climinishes the capacity of the processor to ideri-

Page 89: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure 3.13: Effect of \ar!-ing the chirp duration (Tc) on the correlation. -4 siniilar resiilt ivas obtained with a frequency hopper signal rvith 300 hopper freqiiencies: x-ertical scale. 0.5 L-/DI\-: horizontal scale. 5O/ts/DIV.

Figure 3.14: Oscilloscope outputs shoiving the identification of a chirp gelierating source for \-arying SXR levels: vertical scale. 0.5 V/DI17: horizontal scale 5Ops/DIV. ( a ) correlat ion of chirp signals with a ratio of peak to the sidelobes 20 d B . (b ) SSR=lO dB (c) SSR=5 dB (d ) SSR=-5 dB

Page 90: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure 3.15: Corresponding oscilloscope oiitputs for the freqtlency hopper LPI case under varying SSR levels: vertical scale. 0.5 i-IDIL': horizontal scale 5Op/DIL'. ( a ) correlation of chirp signals n-ith a ratio of peak to the sidelobes is % 20 dB. ( b ) SSR=10 dB ( c )

S S R S dB (cl) SSR=-5 dB

tif? weak radiating sources due to the CCD dynamic range limitation. Hence. the renionl

of these interferers becomes a key design issue. The nes t section esanlines the impact of

these interferers on the correlatiori and ivays to O\-erconie them.

3.11 NIS Excision Using Tunable Filters

Two scenarios of corriiptive SIS on the correlation are investignted: in-band and o-ut-

of-bartd. For the in-band case. the SIS lies within the signal ba~icltriclth (B) siich thnt

T IF - BIS 5 ffvlS 5 f i F + BI2 while in the O-ut-of-band case. the SIS lies outside the

signal bandividth but s tays within the -40 ce11 bandu.iclth (B.40). jc <_ f.vtr < IF - B/?.

as shown in Fig. 3.16.

Page 91: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

4 4 4 I I I I I I

' l ... l 1 I I I I I I I I 1 I I

lc 11 F Figure 3.16: (a) in-band and (b) out-of-band SIS locations relative to the LPI haridwidtk.

Page 92: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Consider an LPI chirp signal corritptecl hl? 'i interferers sitch that

with amplirucles -4,. frecliiencies fj and random phases O,. for j = 1. - - . . -\-. which is

re-IL-ri t ten as

Because -4, >> -4 V j . the SIS autocorrelates and dominates the conipresseci LPI

n-a~eform. Hence. a niethocl is recliiired to notch the XIS energy prior to the correlatiou

operation.

In this section. n-e descrihe an aclaptive system in rvhick the responsc time of tlie

feedback loop is fast enough to track changes in the LPI bandwiclth for niost applications.

Figure 3.17 shows the sub-modiile of a time-intgrating recei\*er usecl to escise the SIS

energ? from the spatially integrated outpiit signal. -4 heam splitter was used to tap off

some of the optical energy incident on the CCD arrac This tappetl off light beam \vas then

focussecl ont O a single photodiode t O recover the correlat ing signals dong wi t h the attendant

tcniporally \ q - i n g carrier and any present interferers. -4 cletailecl mathe~iiatical treatnient

of the signal plus SIS detected bj- this arrangement was del-eloped in Section 2.5. The

resulting analog signal \vas then sampled bj* a 10-bit digitizer and the output digital streani

fcd into a cotnputer for analysis via a General Piirpose Interface bus (GPIB) connectiou.

The computer performed a Fourier transform on the stream wkich decomposecl the

signal into the center frequency of the LPI. fo. SIS frecpncies f,. j = 1. . . . . .Y aucl

Page 93: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Filter Bank

ptiotodetector 7;)

C'CD

Con1 piiter PC-S-lSfi/(ifj 11 Hz

10-bit TEK digitizer

GPIB

RS-232 Iriterface

Figure 3.17: Real time detection and escision of NIS using digitally titnable filters and a single photodetect or arrangement.

Page 94: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

their respectil-e higher order terms fu k 2 f j . fo f 3f, ancl so on. Al1 higher orclcr terms

are oiitside the A0 ce11 banclwiclth and hence do not iuterfere with the correlation otitpiit.

Ouly the frequencies at f, were deerned as NIS and hence the computcr comparecl their

respective energies to a threshold that can he fised or aclaptix-el- set in the prograni. If

a givcu energy esceecled this threshold. a cornmancl \vas issiiecl to the SIS suppressiou

niodule. via an RS-232 serial interface". to a tunable notch filter" to escise the SIS cnergy

at that offencling frequency. The freqiiency to notch was translatecl linearly into a coclccl

digital word corresponding to the center freqiiency of the notch.

For multiple NIS notching. the tunable filters were setup in a parallel hank arrangement.

The first tirne the specrriini !vas compiitecl. the system aiitornatically set the threshold to

i ts largest value. The nuniber of interferers esceecling t his t hreshold w r e t hen comparecl

to the preset nuniber of interferers. In subsecLiient processing. the nimber of interferers

esceeding this threshold were again counted and conipared to the preset nurnber so that

the appropriate filters can he tiined to reniol-e the SIS that esceedecl the threshold. The

thresholcl remainecl fisecl until sonie SIS disappearecl or appearctl in the spectrum. or a new

preset nuniber $sas then chosen. The electronics design inclucled operator control circiiitr?-

to d u c e or increase the threshold depending on hors man- SIS n-ere to be escisecl.

\Vithout taking into account the speed of the electronic Fourier transformer ancl the

PC notch decision circuit. the folloiving can he statecl: when the notching feedback loop

was connectecl to the post-detection electronics. as shown in Fig. 3.8. the speed of notching

"or faster throiighput, digitally turiabie filters with a parallel interface are also available. 'POLE/ZERO Corporation's tunable notch filtcrs are liigh Q. narrowband filters whose center frequency

is controlkd digitally. Ttiey have a tuning range from 1.5 . l lHz to 1 G H z . Filter tune tinies are typically I O ps to any freqiiency within the barid after receipt of tunc command. 3 dB bandwidth of about 1.0 .\IN: and an insertion loss of 20 dB at the center frequency.

Page 95: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

n-as about 0.2 rns. Hotvever. when the notching feedback was connectecl to the spatially

integratecl output. as shomn in F i 0 3.17. the speetl of the notching was abolit 10 p s

The electronic P C basecl Fourier t ransform operation slows dorrn the feedbark loop

to about 1 nis thiis partially masking the benefits of the spatially integratecl oiitpiit.

.A specially designeci controller with a declicated DSP chip coulcl perform the reqiiirecl

elec tronic Fourier t ransform and not ch select ion decision in consitlerabl- fast er t ime t hus

benefiting of the novel architecture.

3.12 Impact of the NIS Notching Feedback Loop

To illustrate the SIS escision techniqtie developed in this thesis. wc corriipted the LPI

wa\-eforms n-i t h in-batid and out-O f-band interferers respectil-ely Figure 3.1 S shows corre-

lations with a number of in-bancl interferers located as shown in (a)-(cl). The interferers

were added to rhe LPI to produce a composite signal. The LPI signal \vas thiis cornplere1~-

ol~scured in the composite signal. Clearly in al1 tlicse cases the desirecl correlation signal

is dominated 11'- the autocorrelation of the in-band 31s. In case of an intercept rcceiwr. a

false alarm can be raised when the SIS is confused with the desired nideband LPI activitj:

After notching out the interferers. the output of the correlator is s h o ~ n in Fig. 3.19

for the same varying number of in-band interferers. This esample illustrates that the

notching technique is capable of recovering the LPI correlation even in multiple in-band

SIS en\-ironments. Siniilar results were obtained for the freqiiency hopped case. Hoivever.

it was founcl that the uotching technique excises a portion of the desired correlation energ-

leading to a dis tort ion in the correla t ion envelope.

Page 96: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure 3.1s: Correlation with in-band interferers at ( a ) 55 LIHz ( b ) 55 MHz and 7.5 1IHz -- (c) 55 UHz. 75 MHz and S5 AlHz (cl) 55 > ~ H Z . ~ C I MHz. S5 MHz ancl 100 MHz on the correlat ion en\-elope. Oscilloscope set tings: 1 .O i-/DI\.'. 50 ps/DI\-.

Figure 3.19: Effect of notching in-band interferers at (a) 55 MHz ( b ) 55 MHz and 75 MHz (c ) 55 MHz. 75 SIHz and S5 MHz (cl) 55 UHz. 75 11Hz. S5 MHz and 100 MHz on the correlation envelope. Oscilloscope settings: S00 niV/DIV. 50 ps/DIV.

Page 97: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

-- Figure 3.20: Correlation with interferers nt ( a ) 55 SIHz and (b) a I I Hz in the in-band case wit h SIS to LPI S S R ratio of 5 dB. Oscilloscope set tings: 1 .O \-/DI\' and 50 ps/DI\-.

Figure 3.21: Sotching of interferers at (a) 55 MHz and ( b ) 75 SIHz in the ai-band case

with SIS to LPI SSR ratio of 5 dB

Page 98: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure 3.2'2: Correlatiou with out-of-band interferers at (a) 30 MHz and ( b ) 100 MHz for a SIS to LPI SXR ratio of 5 dB. Oscilloscope settings: 500 mi-/DI\.-. JO ps/DIV.

. ... A.. ............. ........................................................ ......................-.r...........................-............... 4

i tr r

Figure 3.23: Xotching of out-of-band interferers at (a) 30 MHz and (b ) 100 MHz for a SIS to LPI SXR ratio of 5 dB.

Page 99: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure 3.24: RSIS error in approsimating the peak of the correlation ewdope as a function of the location of the peak of the narrowband interferer.

This penalty [cf Fig. 3.20 and Fig. 3.21)] is more severe when comparecl to the out-of-

band SIS case shown in Fig. 3.22 and after notching in Fig. 3.23. These Figures show the

effect of notching SIS energies for the same NIS power-to-SSR of the LPI. Clearlc the

envelope and niagni t i d e of the correla t ion are niore aclversely affect ecl in t lie i7t- band case.

Since niost of the LPI signal energy is in the banclwiclth B = 50 M H z . Ive espect the

greatest distortion of the correlat ion to occur when notching in-band XIS. Figure 3.24 shows

the root-mean-square (R l IS ) error as a fiinction of the location of the interferer. II'here

the RAIS error is defined as the root mean square of the clifference between the location of

the peak of the correlation output in absence of interferers and its corrrsponding location

after the escision (381.

The RAIS error is small whcn the interferer frecliiency is locaterl far awa- froni the

center of the LPI bandwidth: as in the out-of-band case. The RUS progressi~ely increases

until it peaks at the point whcre the interfcrer is at the center of the LPI banclwidth.

This can be attribiited to the fact that in notching a centrally locatetl SIS. a considerable

amount of signal energ'. in addition to escising the interferer. is removecl. The results are

again shomn to h e consistent for the frequency lioppecl case.

Page 100: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Chapter 4

Processor and Measurement

Hardware

The \arious processor moclules. nieasiirement and sigiial condi tioning elec troniçs disciissecl

in the previoiis chapters. are s h o ~ m in the photographs that folloiv ivi th the riames of the

manufac t iirers and coniponent specificat ions

ji-itli referruce to Fig. 3 3 . n-e consider the following stzbs~-steriis:

a (a ) laser n-ith DC bias. direct mocliilation network and colliniating optics.

O ( b ) .\O deflector with RF driver aniplifier.

a ( c ) pre-detection optics.

(d ) CCD array and post-detection electronics.

O (e) beam-split ter for spatially integrated output and the feedback loop ivith SIS

excision notch filters.

a ( f ) input radar LPI signal generation and XIS sources.

SS

Page 101: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure 4.1: Laser diode source SDL-2432-Hl from Spectra Labs operating at X = '791 nm with Pm,, = 500 mW and a horizontally polarized beam. It is mounted in an SDL-SOOH heat sink. The output laser bearn is expanding at half angles OL = 40' and = 10'. A collimating lens, LiCONiX L80, with numerical aperture (NA=0.65) collects this light and outputs a parallel beam. This parallel beam has an elliptical cross section. In order to evenly illuminate the A 0 cell, the beam was expanded into a circular cross section using a llelles Griot anamorphic prism pair (06-GPA-003).

Page 102: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure 4.2: Another view of the laser heat sink and anamorphic prism pair. In an anamor- phic prism pair' two identical prisms are mounted at an incident angle which is the Brew- ster's angle. The prism pair introduces an anarmophic expansion of the beam while shifting the beam a small amount literally. The particular pair used had a magnification factor of 6x. wavelength range 780 nm-1550 nm and a clear aperture=& 0.5 mm.

Page 103: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure 4.3: View of the fiber coupled detector connected to an oscilloscope via a Tektronix optical-toelectrical converter. This module was used to characterize the modulation depth and sensitivity of the laser diode to direct high frequency modulation. The OR503 opti- cal/electrical (OIE) converter is connected to a Tektronix TM500 power module mainframe for its power supply. It has a wavelength sensitivity of 700 - 1500 nm delivered through a 12.5 prn multi-mode fiber, maximum optical tolerance of +20 dBm and a modulation response of 10 K H z - 1 G H z .

Page 104: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure 4.4: r\nother view of the fiber coupled detector along with the laser diode assembly, collimating optics and an anamorphic prism pair. The laser diode's bias current is provided by an SDL-800 Spectra diodes labs driver. The bias range was 10 to 1000 mA which translated into O to 0.5W CW optical output from the laser. A current limit control network provided overdrive protection for the diode. The driver allowed a modulation bandwidth of 100 KHz. This bandwidth Ras too low for LPI requirements hence a custom modulating network was designed and interfaced to the driver. The purpose of the driver was thu î reduced to simply providing a DC bias to the laser

Page 105: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure 4.5: View of the spatial filter preceded by a Newport Fourier transform lens of focal length f = 38.1 mm and an equivalent lens at the other side of the filter for the inverse operation. Detection is achieved with a CCD EG & G Reticon RL2048D.4G-011 linear array with pixel dimension of 13 pm x 13 p m and 2048 pixels. The arraj* was pre-mounted on the RCOï30LNN evaluation board that supplied the bias and sampling requirements of the device. Maximum throughput from the array is 10 M H z with a typical peak-to-peak pixel noise of 3 mV and a saturation voltage of 2.4 V. The data rate can be increased t o 20 :\IV: hy ronfiguring the board to read "odd" or "p-2'' ~ I x ? ! s only.

Page 106: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure 4.6: Expanded view of spatial filtering and detection sections of the processor. The A0 ce11 is attached to a rotating platform with 4 degrees of freedom (x, y, z and rotation about the z-axis). The cell, a Brimrose TED5-150-100 deflector, had center frequency f.40 = 150 M H z , bandwidth = 100 M H z I acoustic velocity v = 4200 m/s. active aperture 0.75 mm x 21 mm, efficiency = 5576, a Bragg angle es = 14 mrad and time aperture T = 5 ps. The ce11 had a TeOz crystal and could accommodate a maximum RF drive power of 2 W terminated into a 50 R irnpedance. The driver that provided power to the ce11 was a n RF amplifier with a frequency range 40 - 250 M H z . gain="- dB (at j-40 = i50 J l f i z ) and a noise figure=9.5 dl3 maximum.

Page 107: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

4.7: Auxiliary elect ronics showing ampli fiers, mixers, phase shifters, power combin- Figure ersfsplitters used for LPI signal generation and conditioning. On top of the power supply are the POLE/ZERO digitally tunable notch filters (Maxi-Notch 90-200-20) with accom- panying controller. The filters have a tuning range of 1.5 MHz to 1 GHz, a notch depth of 20 dB at t he center frequency, 3 dB bandwidth=3.5% of selected band, and tune times of about 10 ps to any frequency within the band after receipt of tune command from the D i l controller. The filter can handle up to 10 W input power and has a maximum passband insertion loss of 1 dB. The controller in turn is connected to a cornputer via an RS-232 int enace for programmed operation.

Page 108: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure 4.8: Xnother view of the tunable filter/DR controller module sitting on top of a white DC power supply, linear chirpfhop LPI generating circuitry and measurement equipment in the foreground. The fiber coupled detector module is located next to the laser diode assembly.

Page 109: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure 4.9: Another view of the overall processor from the A 0 ce11 plane to the CCD detection board. At the detection plane is a Melles Griot beam splitter (03BSC027 with near-infrared spectral response 700- 1100 nrn that channels some of the intensity downward to a single photodetector arrangement. This ilelles Griot 13DAH001 photodiode had a spectral response of 350-1100 nm, a response at 830 nm of 0.35 A/W, active area of 0.04 mm2, and a typical rise time of 0.35 nsat -9 VDC.

Page 110: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure 4.10: A magnified view of the single diode tap-off arrangement used to detect and excise XIS energies in a feedback loop. The long black rod holds the beam splitter that directs part of the intensity downwards to the diode. The black coax cable is connected to the diode. then to a Mini-Circuit ZHL-SA amplifier. The output of the amplifier is connected to channel A of a Tektronix TK39OAD digitizer which directs the bitstream to a computer via a GPIB (IEEE-488) interface.

Page 111: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure 4.11: Computer connection to the feedback arrangement via a GPIB interface. Beloiv the monitor is the Tektronix TK39OAD 10-bit digitizer that samples the processor output. This digitizer has the full functionality of an oscilloscope with the added benefit of remote or programmed control via a computer interface. The computer performs Fourier analysis and peak detection on the input and sends control signals to excise the NIS energies through an RS-232 connection to the DR notch filter controller. This overall real-time operation is controlled by a C program.

Page 112: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure 4.12: Equipment rack with oscilloscope, HP 8590 spectrum analyzer, HP 8012 pulse generators, Noise/Com NC6107 generator, HP8690B sweep oscillator, power supply, optical power meters, HP 5301A 10 AlHz counter and SDL-800 laser diode driver.

Page 113: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure 4.13: View of the 50 M H z processor along with the test and measurement equip- ment with specifications: difference frequency fo = 73 M H i , LPI bandwidth B = 50 M H z , -40 bandwidth BAO = 100 AlHz and reference tone frequency f, = 100 M H z .

Page 114: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure 4.14: Another evtended view of the 50 !CI H z processor.

Page 115: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure 4.15: Side perspective of the entire 50 M H z LPI processor.

Page 116: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure 4.16: Magnified side view of 50 M H z LPI processor. Note the infrared viewer (FJW) used for alignment and beam find purposes. The viewer has a focusing range of 15 cm to infini ty allowing for close monitoring support.

Page 117: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Figure 4.17: View of the 20 M H z processor from the laser diode end with system specifi- cations: ciifference frequency fo = 30 M H z , LPI bandwidth B = 20 M H z , A 0 bandwidth B.4o = 40 M H z and reference tone frequency f, = 60 M H z .

Page 118: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Chapter 5

Conclusions

The objective of t his work iras to design and irnplemenr tinie integrating optical processors

for the purpose of detecting and iclentifying spread spectriini LPI radar signals in Ion- SSR

ancl SIS corrupted environnients. To this end. the folloiving original research contribiitious

have been macle:

1. A stochastic clerilation of the processing gain of a tirne-inregrathg -40 processor

2. .A novel nier Lod for real- t inie escision of corritp th-e IiIS energ:. iising digit ally t iiuat~le

notch filters in a ttniciiie space integrating refereiice tone detection and feedback

arrangement has been devised and integrated in the processor.

The professional coutriliitions of this work can be suniniarized as:

1. -4 novel optical alignment niethod that is a cheap ancl rohitst alternative to niauy

piirely optical procediircs. has been devised and denioustratcd in this work.

Page 119: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

2 .Anal-t ical espressions for the noisy inteusi ty A 0 transmit tance ancl correlation func-

t ions w r e clerivcd.

3. .A CCD scaling law \vas postulatecl and demonstrateci.

The designed processor consisted of an intensity modulatecl laser diode followed by col-

liniating optics to produce the reciuired beam cross sectional match for optimum escitation

of the A 0 cell. The necessarl- design constraints in meeting this niatching were presentetl

dong n-it h the lem rnagnification factor. eqiial to F2/ Fi for sniall angles.

B y an appropriate choice of op t ical. high-freqiiency elec tronic component s. precision

alignment and detection circuit ry. the processor vas broiight close to an op tirnuni operat ing

point. This operating point mis shown to be very sensitive to the correlation signal-to-bias

ratio as the overall d!*naniic range of the processor is cleterminecl 11)- the dj-naniic range

of the CCD array. Hoivever. operating a laser nt a high average potver leads to a high

nioclulation deptli at the espense of an increased signal-to-bins ratio. To o\*ercome these

conflicting requirements hetween the CCD ancl the laser' s opti~iiiini operation. a neiitral

density filter ivas used to liriearl?- a t teniiate the overall optical poil-er prior to detectiori.

The sigrial-to-bias problem was also shown ro he controllable 1ij. an appropriate choice of

reference tone poiwr. The scaling laws that apply when CCD detection is used were also

de\doped and esperimentally \aliclated.

The optimum operating point of the processor was also fount1 to he liighly dependent

on the overall optical alignment of the \arious components. -4 niechoc! was thus requirecl

to acciirately quanti- the overall optical niisalignnient relative to the oprical center ancl be

able to iteratively correct it throiigh a feedback approach. -4 totally new and robiist schenie

for rneasuring the overall optical misalignment error mas developecl and denions trated.

Page 120: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

.-\nalyt ical expressions for the correlat ion of eit her linearly chirped or frecpency hoppetl

signals with the processor were also derivecl. It was shon-n that in both cases. the correlation

is a sinc function envelope moddating a spatially var>-ing carrier. In the special case of the

frecpency hupper scheme. for a sufficientl~r large nuilmer of hopper frequencies S. it 1r.a~

shown that the output is siniilar to the chirp case with the signal hop handwidth ( BIfI.')

replacing the product of the chirp rate (a) and the integration time (T,,). The effect on

the correlation output. of SIS energy in the bandwitlth of the LPI. was also cliscussed.

-4 novel approach for notching this XIS energ'- usin:, digitally tunable filters and a

space integrat ing reference tone arrangement was shown to be very effec t i\.e part iciilariy

for out-of-band cases. The processor's performance in the presence of both in-band ancl

out-of-band narrowband interferers. in adcli t ion to broadband noise. mas studied and the

esperiment al results were found to be consistent n-i t h t heoretical predictions. -4 technique

for aclaptil-el? identifying the SIS based on an alarrn nwchanism and notching it 11-as

proposed ancl denionstrated. It rvas shown that the error in notching SIS energy peaks at

the center of the LPI frequency span and falls off siil~stantially for SIS freclucncies locatecl

away from the center. This discrepancy was attributcd to the practical limitation of the

filters to restrict the notch only to the SIS energy.

Finall!: the processing gain of the processor was derivecl using a stochastic analysis

approach for a number of input SSR scenarios. This deri~xtion represents a unique new

approach to evaluating a tirne-integrat ing receivcr sensitivity.

Page 121: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

Appendix A

Imaging Optics

The light soiirce was a GaA1.k semiconductor Laser operating at a wavelength of 791 nrn

and a niasirnurn lasing potver of 500 rnIV at a 1 -4 drive curent . The laser is modulatecl by

a reference spread spectrum signal si ( t ). In addition to the laser mocltdating signal. a bias

is added to prevent the laser frorn being pushed belom its t hresholcl and allow operat ion

iri the linear region of its power transfer curve.

The light. as i t leaves the laser diode. is espanding at lialf angles = -10' and = 10'.

This leads to a requirenient for a narnerical aperture (T..i.) of 0.65 to coilimate the beam.

-4 colliniating lem with a S..\. of 0.65 was then mecl to collect this light and output a

paraIlel heam. This output parallel beam. hoirever. has an elliptical cross section. In orcler

to evenh illuminate the .A0 cell. the elliptical cross-section is espanded into a circular

one using an anamorphic prism pair. To efficiently focus the beam energy to the A0 ce11

aperture. a cylindrical lens was then used to provide the line illumination that fits the ce11

aperture windom.

LVith an ohject ?;.-A. of 0.65 and an object distance of 10 mm. the recpired dianieter

Page 122: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

collimating optics Cl

Figure -4. 1: Collimat ing and pre-detect ion .-\O ce11 aperture mat ching op t ics.

(O) of the lens to collect al1 the Laser light \vas giren bu:

O 0.65 = O = 13rnrn

2 * 10 mnz

- 1.6. Thus. the The requirecl magnification ( rn ) to match the -40 aperture i ras - - focal length of the reqiiired cylindrical lens was determinecl to be:

Accomplishing t his imaging t ask wi t h a single cylindrical lens would t herefore reqiiire

an optic with a 5.92 mm focal length. ancl a 13 mm diameter. the optimum clear aperture

of the lem.

Page 123: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

References

[l] R. C. Dison. Sprcad Spectrum S y s t e m . 4ew kork. Sen* Eork: .John IViley aucl Sons.

1994.

[%] D. .J. Torrieri. PriticipZes of Secure Commn.icat ions Systerns. 2nd ed. Boston. CS;\:

Artech House. 1992.

131 D. Torth. -.-\nalysis of the factors which determine signal/noise discrimination in

radar." Proc. IRE. vol. 51. pp. 1016-102s. July 1963.

[1] P. Peebles. Probabilitg. Random Variables and R a n d o m Signal Principles. Xew Eork.

Yen- kbrk: 11cGraw-Hill. 1987.

.- [j] V. 11. Rist ic and J . P. Y. Lee. %odeling the noise figiire of an acousto-op tic receivzr.

Applied Optiw. vol. 36. pp. S32-S35. Feh 1996.

[6] C. Helstrorn. Elements of Signal Detection a n d Es tirnation. Englewood Cliffs:

Pretince-Hall. 1995.

[7] C. S. .Anderson. II. W. Haney. and J . 11. Pallegrino. -Signal delay and cornpression.-*

in Acousto-Optic Signal Processing ( S . J . Berg and J. H. Pellegrino. eds.). pp. 100-1 15.

Xew York: Marcel Dekker. 1996.

Page 124: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

[SI M. 1. Skolnik. Introduction to Radar Systems. Sem York. Sew York: SfcGraw-Hill.

1962.

[9] R. Griffin and J . 3. Lee. -.Acoust~optic tvideband correlation SJ-stem: Design . im-

pleniantation and elialuation." A p p l . Optics. vol. 33. pp. 677-1-6787. Oct 1994.

[IO] C. Anderson. .A. Dorset. R. Berinato. -4. Filipov. E. Adler. and J. Pellegrino. -.in

acousteoptic ESM receiver iipgrac1e.- Proc. SPIE: Tram. of Optical Processors into

System. vol. 2336. pp. 31-38. 1994.

[ I l ] J . B. Y. Tsui. Micm.wave Receiiler with Electro~iic Warfare Applications. Xen- York.

Sen- York: John ii'iley and Sons. 1986.

[12] G. i\'. .Inderson. -;\dvanced channelization technology for rf. rnicroivave ancl millinie-

- ter ivave applications.- Proceedings of the IEEE. vol. l9. no. 3. p. 355- 1991.

[13] l\'. T. Rhodes. -Acousto-optic signal processing:corivoliition ancl correlation." Proc.

IEEEE. vol. 69. p. 65. 1951.

[14] A. \-anclerlugt. Optical S i p a l Processing. Yeiv lbrk: John Il7ile>-. 1992.

[ l5] iV. T . Cathey. Optical Information Processïrrg and Holography. Sew York: l1,ïley.

1973.

[16] Y. J. Berg ancl J. AI. Pelligrino. Acousto-Optic Signal Processing. Yew York. Sew

York: Uarcel Dekker. 1996.

[17] .J. Gower. Optical Commsnication Sys tem. Englervood Cliffs: Prentice-Hall. l9Sl.

Page 125: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

[lS] H. Zniiida ancl E. N. Toughlian. Photonic Aspects of M o d e m Radar. Boston: H.

Zmiida and E. S. Toiighlian. edts. Xrtech Hoiise. 1994.

[19] -1. D. Goutzoulis and E. D. R. Pape. Design and Fablication of Acousto-Optic Devices.

Sen- York: llarcel Dekker. 1994.

[-O] A. Kiridut a and Y. Ristic. w.~cousto-op tic tlevices: Application to the reconstr~ictiou

of spatially samplecl data in Slagnet ic Resonance Imaging ( LIRI) ." Can. J. Elect. kY

Comp. Eng.. vol. 21. no. 4. pp. 145-14s. 1996.

1-11 B. Tasic and \-. SI. Ristic. ~~Triple-procluct optical processor for analysis of pulse

repeti tion and carrier freqiiencies of raclar signais: Applied Optics. vol. 34. pp. S14-I-

8147. Dec. 1995.

1221 D. Psaltis and D. Casasent. -5pread spectruni time- and space- integrating optical

processors ... Appl. Optics. vol. 19. no. 9. pp. 1546--1349. 19SO.

[23] D. Hart up ancl i\'. Rhodes. --Acousto-op t ic processor for carrier frequency ancl enve-

lope moclula tion analysis.*' Proc. SPIE. vol. 1104. pp. 98-104. 1992.

[?-LI P. Iiellmau. --Integrating optical signal processing." Optical Engineering. vol. 39. no. 3.

pp. 370-375. 1980.

[25] A. Iiorpel. -.Acous to-opt ics." Applied Solid State Science. vol. 3. pp. 73- 179. 1972.

[26] P. Iiellman. Tirne Integrat ing Opt icai Processing .'. Ph. D. Dissertation. Stanford Uni-

versitg. 1979.

Page 126: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

[27] P. Iiellman. H. S. Saver. and J . W. Murray. -1ntegrating .-\coiisto-Optic Channelized

Receivers." Proc. IEEE. vol. 69. pp. 93-100. 1951.

[-SI D. Psaltis and B. V. Iiiimar. -Acoust~Optic Spectral Estimation: a Stastical Analy-

.- sis. Appl. Optics. vol. 20. pp. 601 -608. 1951.

[29] T. Bader. P. Iiellman. and H. Shaver. T i r n e integratirig optical signal processing."

Technicul report Airforce W c e of Scientijic Reseach. vol. F49260-7s-C-0102. 19S1.

[30] E-. I'arnamoto. -;\SI and F l I Qtiantiim noise in semiconductor lasers -Part 1: Theo-

retical anal>-sis." IEEE J. Quant. Electron.. vol. 19. pp. 34-46. Jan. 19S3.

[31] T. lliikai and Y. E'anianioto. -.UI Quantum noise in 1.3prn InGa-4sP lasers." Elec-

tron. Lett.. vol. 20. pp. 29-30. .Jan. 19S4.

[32] S. Haykin. Communication System. Toronto. Canada: John IViley. 1983.

[33] -A. Papoulis. Probabilitg. Random Variables. and Stochastic Processes. Sen- Ebrk:

llcGraw-Hill. 1965.

13-11 P. Hill. R. 0lshansk)-. and MF. Powazinik. -Reduction of Relative Intensity Soise in

1.3prn InGa-4sP semiconductor lasers." AppI. Php. Lett.. vol. 50. pp. 1400-1402. SI-

19s';.

[33] .A. Iiiruluta. V. SI. Ristic. aud J. P. Lee. -Tirne-integrating acousto-optic correla-

tion in Gaussian noise." in IEEE Canadiun Conference on Electrical and Cornputer

Engineering. (St. John3 SF). pp. 97--104. May 25-2s. 1997.

Page 127: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

[36] -4. Iiiruliita. 1-. SI. Ristic. and .J. P. Lee. -Optical misalignment estimation in a timc-

integrating acoiisto-optic processor. .- Accepted for publication in Opttcal Engineering.

no. OE110-7. p. 13 pages. M a - 1997.

" . [37] B. SI. Sadler and A. 3. Filipov. -1ntegration of acousto-optic spectrum analyzers. in

Acoasto-Optic Signal Processing ( S . J . Berg and J . H. Pellegrino. ecls.). pp. 131-175.

Sel- lb rk : Marcel Dekker. 1996.

[3S] R. IVard and -4. ianderlugt. -Signal distortion in an adaptive excision system.- Optical

Engineeririg. vol. 31. no. 3. pp. 606-613. 1992.

[39] E. \.'iwiros. E. Adler. ancl 11. Pat terson. -Dernonstration of witleband acorist<~optic

correlator in an ESAI testbed." AMSRL-SS-TB. pp. 16-21. 1995 (tinp~ihlishcd).

Page 128: Processors the Detection of - University of Toronto T-Space · 395 Wellington Street 395. rue Wellington OttawaON K1AON4 OttawaON K1AON4 Canada Canada The author has granted a non-

IMAGE NALUATION TEST TARGET (QA-3)

APPLIED - 4 IMAGE. lnc = 1 653 East Main Street - -- . , Rochester, NY 14609 USA 7- ,== Phone: 71 61482-0300 -- -- - - Fax: 716/288-5989

O 1993. Applied Image. Inc.. Ail Aights Aesewed


Recommended