Chapter 1

Introduction

1.1 Preliminary remarks

Sensor systems such as radar or sonar receivers are tools for the interpretation of waves.A wave is by definition a function of space and time. Therefore, wave processingtechniques basically include the spatial or temporal dimension or both. Synthesis ofan antenna directivity pattern can be interpreted as spatial processing. Operations suchas demodulation, filtering or Fourier analysis of the antenna output signal are temporalprocessing techniques.

Space-time signal processing is required whenever there is a functional dependencybetween the spatial and temporal variable. This is fulfilled in several applications, forexample:

moving pulse Doppler radar: dependency of the clutter Doppler frequency onthe direction of arrival;

frequency dependency of the directional response of an antenna array withnarrowband beamforming;

ambient noise in sonar (different frequencies arriving from different directions).

Space-time processing techniques can typically be applied in areas such as

airborne MTI radar (B RENNAN et al. [54]);

synthetic aperture radar (SAR), see ENDER [108], BARBAROSSA and FARINA[31], DONG et al. [98];

spaceborne MTI radar (SEDWICK et al. [456], MAHER and LYNCH [330]);

clutter cancellation for SAR/ISAR (ENDER [108], MERIGEAULT et al. [354],GENELLO^a/. [146]);

interference suppression in narrowband radar through artificial array motion(LEWIS and EVINS [304]);

interference suppression for broadband radar (COMPTON [80, 79]);

wideband interference rejection in GPS receive arrays (FANTE and VACCARO[121]);

terrain scattered jamming (GABEL et al [139], KOGON et al [282, 283],ABRAMOVICH^a/. [1,2]);

cancellation of mainbeam interference in the presence of multipath (KoGON etal [284]);

combinations: suppression of terrain scattered jamming and clutter;1 (RA-BlDEAU [416], TECHAU et al. [488]), clutter rejection with wideband radar(FANTE etal. [116], RABIDEAU [419]);

separation of closely spaced sources (SYCHEV [483, 484]);

space-time coding of an array for simplification of beamforming (GUYVARCH[188]);

clutter mitigation in over-the-horizon (OTH) radar; (ABRAMOVICH et al [1, 2],KRAUT et al [287]);

suppression of reverberation for active sonar;

simultaneous localisation and Doppler estimation for passive sonar; (matchedfield processing);

signal processing for communication networks (PAULRAY and LlNDSKOG[407], SEE etal [458], HOCHWALD etal [212]);

simultaneous frequency and DOA estimation in a multiple source environment(ROBINSON [437], LUKIN and UDALOV [325]).

An overview of some of the techniques listed above is given by WARD et al [535].Further detailed information can be gathered in the special issues of ECEJ (KLEMMed., [262]) and IEEE Trans. AES (MELVIN [353]).

Applications in other areas may be possible. As the author's expertise is mainly inthe field of airborne MTI radar the major part of this book is focused on this subject.Some of the other aspects, however, will be touched on in later chapters.

1.1.1 Basics of MTI radar

MTI radar (Moving Target Indication) has by definition the capability of detectingmoving targets before an interfering background (usually called clutter). Morespecifically, by MTI radar, pulse Doppler radar (PDR) is understood which usesthe Doppler effect to detect moving targets before a clutter background (Dopplerdiscrimination). The difference between target and clutter velocities is exploited

1 Frequently referred to as 'hot and cold clutter'

for target detection. The pulse Doppler radar transmits phase-coherent pulses andmeasures the phases of the backscattered echoes. The radial velocity of any movingobject results in phase advances of successive echo pulses. By spectral analysis of thephase history of an echo sequence the Doppler frequency and, hence, the radial velocityof the reflecting object can be found. For spectral analysis either DFT, FFT or a bankof Doppler filters may be used.

MTI techniques are based on the temporal coherence and, hence, of the phases ofecho signals. This is in contrast to change detection methods as known from imagingradar (e.g., KlRSCHT [227]) which compare successive amplitude images.

KOCH [279] and more recently KOCH and VAN KEUK [280] demonstratethat tracking of air targets in a densely cluttered environment can be successfullyaccomplished by smoothing via retrodiction without using any anti-clutter devicebefore tracking. It can be expected, however, that MTI pre-filtering will alleviate thetracking workload which may result in reduced time delays due to multiple hypothesistesting.

There are several types of clutter which differ in their spectral parameters. The mostimportant kind of clutter is ground clutter caused by echoes scattered from the ground.Ground clutter received by a stationary radar exhibits a symmetric Doppler spectrumcentred at zero Doppler. The clutter power and the Doppler bandwidth depend on thetype of background. Hard objects (buildings, urban areas) will produce high-powernarrowband Doppler spectra while areas with a high degree of roughness or internalmotion (agriculture, vegetation, forests) cause less clutter power at larger bandwidth.The bandwidth increases with wind speed and radar frequency, see NATHANS ON [365,p. 274]. Moreover, antenna rotation causes additional broadening through spatialdecorrelation of clutter echoes.

Weather clutter may show a shift of the Doppler spectra towards non-zerofrequencies. This frequency shift reflects the average radial motion of weather (clouds,rain, snow, hale, chaff) due to the wind speed. Positive Doppler frequencies arean indication for approaching, negative for receding weather. Usually the Dopplerbandwidth of weather clutter is larger than in case of ground clutter. There are fourmechanisms that are responsible for the shape of the weather clutter spectrum: windshear, Doppler spread in the cross-wind direction, fluctuating currents, and a fallvelocity distribution of reflecting particles, see NATHANSON [365, p. 205]. For seaclutter a shift of the Doppler spectrum occurs due to the velocity of the ocean wavesrelative to the radar. Moreover, there are effects such as individual wavelets, foam andspray which broaden the clutter spectrum, see NATHANSON [365, p. 241].

1.1.2 One-dimensional clutter cancellation

Ground clutter echoes can be cancelled by use of a discrete filter (FIR or HR) operatingon a pulse-to-pulse basis, with a notch at zero Doppler frequency. The clutter notchmay be formed adaptively so as to match the individual shape of the clutter Dopplerspectrum. Normally, the variations of the clutter bandwidth are taken care of by afixed filter centred at zero frequency whose clutter notch can be matched to the actualclutter bandwidth. Of course, low Doppler targets, i.e., targets whose radial velocity

component relative to the radar is small, may be buried in the clutter bandwidth and,hence, are hard to detect.

For sea and weather clutter the Doppler shift of the spectrum due to the radialvelocity components of ocean waves or clouds may result in an offset between themaximum of the Doppler spectrum and the clutter notch of a fixed zero Doppler clutterfilter. There are techniques for aligning the centre frequency of the clutter spectrumwith the clutter notch of the pre-filter. This may be done by adjusting the Cohofrequency in such a way that the clutter energy maximum falls into the filter notch(TACCAR, see SKOLNIK [467, p. 17-32]). Alternatively, the motion induced phaseprogression of clutter returns may be used to compensate for the clutter motion in thebaseband (clutter locking, see VOLES [509]).

Since the bandwidth and the centre frequency of sea and weather clutter returnsvaries within a wide range, adaptive clutter filtering is the appropriate way of solvingthis problem. A lot of solutions have been reported in the literature. Basically optimumclutter cancellation can be formulated as a binary hypothesis test ('target present' or'no target') which is given by the Neyman-Pearson test. It requires knowledge of thecovariance matrix of temporal clutter samples which has a Toeplitz form if the pulserepetition frequency (PRF) is constant. In [64] (B UHRING and KLEMM) an adaptivefilter operating with a surveillance radar with rotating antenna has been described (fordetails see Sections 1.2 A A and 1.2.5.2). In this system the Levinson algorithm as usedby Burg in his papers on maximum entropy spectral analysis (BURG [67, 68]) has beenimplemented to adaptively estimate the clutter correlation function and to calculate theassociated least squares FIR filter which minimises the clutter power.

A lot of literature on clutter and clutter suppression for stationary radar has beenpublished. For the sake of brevity, only a few are quoted here. In SKOLNIK'S RadarHandbook [467, Chapter 17] and [468, Chapter 15] an overview of the problems andtechniques for clutter suppression is given. An even more detailed and comprehensivedescription is given by SPAFFORD [470] and in the book by SCHLEHER [451]. In thebook by NATHANSON [365] a collection of measured clutter backscatter data can befound, with conclusions on detection of signals in clutter.

1.1.3 Aspects of air- and spaceborne radar

Observation of the earth's surface by air- and spaceborne radar has several advantagescompared with ground-based radar which are caused by the elevated position of theradar and the radar platform motion:

The horizon, i.e., the visible range, is extended through the elevated position.

The elevated position leads to a reduction of terrain masking effects. The effectof shadowing due to hilly terrain is mitigated for a radar looking from above.

For air- and spaceborne radar the wave propagation conditions are morefavourable than for ground-based radar. The beam of a high-flying radar hasto cross the lower atmosphere layer while the beam of a ground-based radaris entirely buried in the lower atmosphere. Moreover, the interaction of the

transmitted wave with the ground is reduced. This interaction causes additionalsidelobes in the directivity pattern.

The platform motion offers the potential of high-resolution imaging throughsynthetic aperture processing, including interferometric imaging and changedetection.

There are some specific problems associated with a flying radar platform as far asclutter is concerned. The extended horizon means an extended visible range but alsoan extended clutter area. For ground-based radar ground clutter ends at about 50 km,but for spaceborne radar clutter will be present in the whole visible range. Airborneradar clutter will cover some distance in between, depending on the platform altitude.Furthermore, since the depression angle of the radar beam to the ground is larger for air-and spaceborne than for ground-based radar higher clutter power can be expected. Inparticular, from underneath the platform high-power clutter returns with zero Dopplerfrequency are received which are commonly called the spectral altitude line. Sincethe altitude line is generated through specular reflection the associated clutter power isparticularly high.

1.1.4 Impact of platform motion

1.1.4.1 From DPCA to STAP

Another, even more important property of clutter echoes received by a moving radar isthe motion-induced Doppler spread. A radar mounted on a moving platform receivesclutter echoes that are Doppler shifted. The Doppler shift depends on the radial velocityof the individual scatterer relative to the radar which in turn is a function of thedirection, i.e., azimuth and depression angle, see the formula at the beginning of thepreface. The maximum clutter Doppler frequency occurs in flight direction while in thecross-flight direction the Doppler frequency is zero. The total of clutter arrivals from allpossible directions sum up in a Doppler broadband clutter signal whose bandwidth isdetermined by the platform speed and the wavelength. Any kind of Doppler spread ofthe clutter spectrum leads to a degradation of the detectability of low Doppler targets.Conventional temporal clutter filters operating on echo data sequences can be designedin such a way that the clutter is suppressed optimally. Low Doppler targets, however,will be suppressed as well.

These problems can in principle be circumvented by appropriate radar antennaand signal processing design. This is illustrated by the following consideration. Asmentioned before for clutter echoes there is an equivalence between the direction ofarrival and the Doppler frequency, whereas the Doppler frequency due to a movingtarget is independent of the target direction. This fact can be exploited for targetdetection in a motion-induced Doppler coloured clutter environment.

Suppose, for example, a beam is steered in a certain direction. Then the clutterbandwidth at the radar output is determined by the antenna beamwidth. If this clutterpart of the spectrum is suppressed any moving target whose Doppler frequency fallsoutside the received clutter spectrum can be detected. Assuming an infinitely narrowbeam then the received clutter spectrum is just a single frequency line. In this ideal case

any target Doppler frequency different from the clutter line can be detected. There isobviously no limitation due to the platform motion. Target detection is limited only bythe spatial resolution of the beam and the spectral resolution of the Doppler analysis.

It should be emphasised that the processing described in the previous paragraphconsists of two parts, a spatial and a temporal part. Likewise, in the Fourier domainone may speak of directional and spectral processing. In the example given above(beamformer + Doppler analysis) the temporal part depends on the spatial part becausethe beamformer determines the clutter spectrum.

Since beamforming and Doppler filtering are linear operations the succession ofspatial and temporal processing can basically be interchanged. This might not berelevant for realistic radar operation but helps for further clarification of the two-dimensional processing principle. Consider an omnidirectional transmitter and an arrayantenna with individual sensor outputs. First the echo signals of all antenna channelspass narrowband filters in order to select one specific Doppler frequency (temporalfiltering). The second stage is a multibeamformer with one beam pointing in the clutterdirection. The clutter beam will be discarded (spatial filtering) while all the others willfind those moving targets which have the same Doppler frequency as the narrowbandfilters.

It should be noted that in the above discussion on the relations between Dopplerfrequency and angular direction no assumptions on the orientation of the antenna weremade. It follows that basically any kind of array configuration may be used 2. This willplay a major role in the discussion of sideways and forward-looking antenna arrays,that means, array antennas aligned with the flight or cross-flight direction.

From these considerations one can see that basically the platform motion does notlimit the detection of slow targets, provided that appropriate space-time processing isapplied. The first approach to space-time processing of clutter echoes was the DPCAtechnique (displaced phase centre antenna), see ANDREWS [15, 16, 17], SKOLNIK[467, pp. 18-1], [468, pp. 16-1] and TSANDOULAS [501]. The DPCA technique isbased on a sidelooking antenna arrangement with two phase centres displaced alongthe flight axis. This can be realised by use of two identical antennas or a monopulseantenna (ANDERSON [14]). The PRF, the displacement of the phase centres and theplatform speed are adjusted in such a way that the second phase centre assumes theposition of the first one after one PRI (pulse repetition interval). Therefore, any twosuccessive pulses received by the two different antenna parts appear to come fromone phase centre fixed in space so as to compensate for the influence of the platformmotion3. A subsequent two-pulse canceller subtracts the second echo received by thefirst antenna from the first echo of the second antenna. In essence this is a kind ofspace-time processing.

As pointed out above conventional clutter filters as used in stationary radar are bynature temporal, that is, they operate on a echo-to-echo basis. By talking about space-time processing we enter the area of signal processing for array antennas. Fortunately

2However, as will be shown below, the antenna geometry has a dramatic influence on the clutter spectraand the kind of signal processing.

3 Actually, it is not the positions of the two sensors that have to coincide but the phases of two subsequentclutter returns. Due to the two-way propagation the distance the antenna has to move forward is only halfthe sensor spacing.

here we can refer to a large choice of literature on adaptive arrays. Since the earlypublications by BRYN [62], MERMOZ [355], SHOR [466] and WIDROW et al [547] alot of authors have contributed to this subject. There are several textbooks, for examplethose by MONZINGO and MILLER [360], SCHARF [450], HUDSON [216], NICOLAUand ZAHARIA [383], HAYKIN [203], and FARINA [122]. Fundamentals of statisticalsignal processing can be found in SCHARF [450].

There are two principal applications of adaptive array processing which are closelyrelated to each other: 1. cancellation of directional noise (adaptive nulling), 2.superresolution of targets (adaptive beamforming). First applications of adaptivearrays were in the field of underwater acoustics, mainly with application to submarinelocalisation, and in geophysics, to locate earthquakes and subterraneous nuclearexplosions. With the advent of phased array technology the theory of adaptive arrayshas been adopted by the radar community, see BRENNAN and REED [53]. Predominanttasks are cancellation of unwanted radiation, such as jamming or directional clutter, andresolution of close targets.

After a few years of initial enthusiasm over adaptive systems, it was recognisedthat adaptive array processing may be seriously limited by several effects. In sonarapplications, various stochastic irregularities of the propagation medium lead to a lackin spatial and temporal correlation of received waves. This lack of correlation limitsthe capability of noise cancellation. Moreover, the environmental conditions of thepropagation path may cause a mismatch between the expected and the received signalwhich leads to degradation of the performance of superresolution techniques. Sinceambient noise is not normally highly directive, spatial noise cancellation requires a highnumber of degrees of freedom which in turn means high computational complexity andcost.

In the radar world the propagation medium is much more predictable than inunderwater acoustics. There are, however, 'home-made' problems in the radar receiveprocess. Receiver noise, instabilities of the amplification and down-converting chainas well as tolerances in the transmission characteristics of individual array channelsmay limit the performance of adaptive radar array antennas. The impact of channelerrors on the performance of adaptive radar arrays has been analysed in some detail byNICKEL [374].

DPCA techniques are basically non-adaptive. As pointed out earlier adaptiveprocessing is necessary whenever parameters of the received clutter spectrum areunknown, like for instance the centre Doppler frequency and bandwidth of weatheror sea clutter. For such application the TACCAR loop (SKOLNIK [468, pp. 17-32])provides an adaptive shift of the centre frequency of the clutter spectrum towards theMTI clutter notch.

Even for ground clutter adaptive processing may offer some advantages. On theone hand the clutter characteristics may change while being passed by the radar beam.On the other hand airborne platforms are subject to irregular motion due to atmosphericturbulences, wind drift, and vibrations. An adaptive filter will be able to compensatefor such perturbations provided that the adaptive algorithm converges fast enough.

Space-time processing has already been recognised as a technique for broadbandjammer cancellation. In 1976 BRENNAN et al. [54] proposed the principle of space-time adaptive processing (referred to by several authors as STAP). for suppression

Figure 1.1: Functional block diagram of STAP radar

of airborne radar clutter. They analysed the performance of the optimum processor,see Section 4. More details on the clutter suppression performance of the optimumprocessor can be found in a paper by KLEMM [238].

Since then a lot of papers on suboptimum techniques for clutter suppression havebeen published by ENDER [99], ENDER and KLEMM[109], KLEMM [240, 241, 242,244, 246, 253, 274], LiAO et al [308, 309, 310], NOHARA [384], Su et al [473],WANG H. et al [511, 512, 516], WANG Y. and PENG [524], WICKS et al [544]

and ZEGER and BURGESS [578], and others. The report by WARD [530] and thepapers by KLEMM [256, 257] can be used as an introduction to the area of space-time adaptive processing for airborne radar. A recent overview paper on space-timeadaptive processing has been published by WARD et al [535]. The digest of the IEESTAP colloquium (1998) (KLEMM [276]) gives an overview of the state of the art inspace-time adaptive processing. It contains contributions by KLEMM, WARD, FARINAand LOMBARDO, ENDER, RICHARDSON, GABEL, and WRIGHT and WELLS.

antenna array

transmit/receive channels, including amplifiers,phase shifters, down-mixers, matched filters,AD-converters

e.g. estimation and inversion ofspace-time covariance matrix;other adaptive algorithms

denotes multi-channel)

filtering the echo data of all range binsat range sample rate

adaption ofspace-timeclutter filter

space-timeclutter

suppression(whitening)

space-timematch of target

signal

beamformer in look directionDoppler filters for all possible target velocities

targetsignal

detection

calculate test function (e.g. maximum modulusof Doppler filter output signals), compare withdetection threshold

The optimum detector (maximum likelihood) concept leads to a fully adaptivespace-time processor which in practice can be difficult to implement for realisticantenna dimensions (possibly several thousands of sensor elements, hundreds of echopulses per CPI). Therefore suboptimum techniques which promise near-optimumclutter suppression at reduced computational expense are of great interest. Thediscussion of suboptimum techniques for air- and spaceborne clutter suppression willform a significant part of this book.

An alternative use of space-time processing has been proposed by LEWIS andEViNS [304]. The authors propose to generate an artificial Doppler by shiftingthe illumination of a subarray of a linear (or rectangular) transmit array during thetransmitted pulse length. Therefore, sidelobe clutter appears at frequencies outsidethe radar bandwidth and will be suppressed. A similar technique involving sequentialtransmission of individual transmit elements leads to improved beamforming and targetsearch, and simplified processing (MAHAFZA and HEIFNER [328]).

In most of the published papers on space-time clutter suppression the authorsassume a sidelooking antenna in which case the function of the motion compensatedMTI can be explained by the DPCA principle (see above and Chapter 3).

Besides sideways looking radar, forward-looking radar plays an important role. Areview of the references shows that so far very little information is available on forward-looking MTI. In SKOLNIK'S Radar Handbook [468, Chapter 16] one finds a plotwhich indicates that forward-looking MTI can function for an antenna arranged in thecross-flight direction. Two or more phase centres of the antenna have to be generatedalong the flight axis by appropriate illumination so as to create the DPCA effect. In[253, 259] (KLEMM) and [427] (RICHARDSON and HAYWARD) some properties offorward-looking MTI radar as compared with sidelooking radar are discussed. In thesubsequent discussions a major part of our attention will be devoted to forward-lookingradar. Forward-looking radar plays a significant role in aircraft nose radar as wellas spaceborne radar. Even synthetic aperture imaging may be accomplished with aforward-looking antenna (MAHAFZA et al. [327]). MTI for forward-looking radar willplay a major role in this book.4

The content of this book consists mainly of numerical examples which illustrate thefunction of the various processors discussed. Although a large number of numericalresults have been presented it should be emphasised that the material presented is notexhaustive. It is merely thought that the examples show the radar system designer away to analyse his specific problem under his actual system requirements.

1.1.4.2 The principle of space-time adaptive processing

In this subsection the principle of space-time processing and the role of this bookin this context is briefly explained. Figure 1.1 shows a coarse block diagram of aMTI radar including space-time adaptive processing for cancellation of clutter withmotion-induced Doppler colouring. A sequence of coherent pulses is transmittedvia the array antenna. The associated echo sequence enters the antenna, eventually

4Many papers on space-time processing do not indicate whether they deal with sideways or forward-looking array geometry. Presumably, since most of the ongoing experiments are based on sidelooking arrays,these papers address the sidelooking geometry.

through a network of pre-beamformers.5 Echo signals are amplified, down-mixed anddigitised. The next step includes adaptation of the clutter filter to the actual clutterechoes received. Adaptation is done by estimating the space-time covariance matrixfrom secondary training data or other covariance matrix-free algorithms (TUFTS et al[503, 504, 505], HUNG and TURNER [217]).

The received echo signals are filtered so as to cancel the clutter component in theradar returns. The remaining signal + noise mixture is then matched with a space-timematched filter which in essence is a beamformer cascaded with a Doppler filter banksince the target Doppler frequency is unknown. The output signals of the Doppler filterbank are used to design a test function which has to be compared with a detectionthreshold. Usually the maximum of the squared magnitudes of the output signals ofthe Doppler filter bank is chosen for threshold comparison.

Notice that the STAP techniques are contained in the two blocks 'adaptation' and'clutter suppression'. The design of suitable architectures of adaptive space-time filtersis the main subject of this book.

1.1.5 Some notes on phased array radar

Space-time processing requires a phased array radar with an array antenna which hasseveral output channels so as to provide a spatial processing dimension. In future yearsphased array radar will play an increasing role, especially for military applications.Phased array radar has a number of unique capabilities and properties:

Inertialess beamsteering;

Multifunction operation (e.g., search, track, terrain following, guidance,mapping, SAR, ISAR);

potential of energy management strategies;

advanced search techniques such as the sequential test (WiRTH [554, 551, 557,559]);

multiple target tracking (VAN KEUK and BLACKMAN [507]);

potential of multiple beamforming (WiRTH [558, 556]);

potential of spotlight SAR;

correction for motion-induced azimuth errors of moving targets in SAR images(ENDER [101, 102, 103]);

high reliability of active arrays;

high efficiency of active arrays;

design of antennas with 360 coverage (WILDEN and ENDER [549]);

5 This depends on the actual STAP algorithm used.

potential of spatial signal processing on receive (anti-jamming, superresolution),see for instance WIRTH and NICKEL [561], WIRTH [552];

potential of spatial filtering on transmit;

space-time, space-TIME, space-time-TIME, space-frequency processing forsuppression of various kinds of interference.6

As one can see from the above listing the reasons for using phased arrays in future radarsystems are manifold. The development is not necessarily driven by the requirementfor space-time processing.

1.1.6 Systems and experiments

Several experimental or operational radar systems with multichannel antenna for space-time processing currently exist or are in the planning phase:

1. The operation of a phased array based DPCA antenna has been testedexperimentally by TSANDOULAS [501]. The PRF was automatically adjustedto variations of the aircraft attitude.

2. The AN/APG-76 (see TOBIN [496], NORDWALL [390]) by WestinghouseNorden (now Northrop Grumman Aerospace) is a fielded and operationalmultimode radar system with SAR and GMTI (ground moving target indication)capability.

3. The JOINT-STARS radar has space-time MTI capability based on an arrayantenna with three subarrays, see HAYSTEAD [205], SHNITKIN [464, 465]. Theclutter suppression achieved is of the order of 20 dB. The MTI function hasproven to be useful during the NATO operations in Kuwait and Bosnia, seeCOVAULT [85],ENTZMINGER^a/. [112].

4. The Multichannel Airborne Radar Measurements program MCARM by RomeLaboratories (see SURESH BABU et al. [478, 479] and FENNER and HOOVER[132]) uses an L-band airborne phased array radar testbed. Studies on real-timeprocessing architectures are part of this program (see BROWN and LINDERMAN[59], LINDERMAN and LINDERMAN [315], LITTLE and BERRY [316]).

5. NRL used an eight-element UHF linear array under an EP-3A aircraft(BRENNAN et al [57] and LEE and STAUDAHER [297]).

6. The experimental C-band SAR by DRA (UK), see COE and WHITE [77, 78],uses three antennas in an along-track arrangement on an aircraft.

7. The MOUNTAINTOP program is an ARPA/NAVY sponsored initiative startedin 1990 to study advanced processing techniques and technologies required tosupport the mission requirements of next generation airborne early warning(AEW) platforms (Tm [494], T m and MARSHALL [495]).

6By 'time' we mean the slow (pulse-to-pulse) time, while 'TIME' denotes the fast (range) time.

8. The motion of an airborne radar can be emulated by use of a linear array oftransmit antennas which transmit radar pulses successively. Then the clutterreturns received by a stationary radar are equivalent to clutter echoes receivedby a moving radar. This technique has been described by AUMANN et al [21],and has been used in the framework of the MOUNTAINTOP program.

To emulate a sidelooking array the transmit array has to be aligned with thereceive array. For forward-looking operation the transmit array has to be rotatedby 90.

9. The NFMRAD by GOGGiNS and SLETTEN [157] (Air Force CambridgeResearch Labs., USA) has been designed to analyse experimentally the principleof space-frequency clutter nulling. First experiments used a truck as carrier.

10. The ADS-18S antenna (Loral-Randtron), operating at UHF (DAY [92]).

11. The DOSAR by Dornier (Germany). HIPPLER and FRITSCH [211] use twoantennas in the along-track configuration for space-time MTI operation.

12. The AER II developed at FGAN-FFM (Germany) by ENDER [100, 104, 106];ENDER and SAALMANN [111], ROSSING and ENDER [439], is a four-channelexperimental airborne X-band SAR. It has the capability of sideways andforward-looking operation. For testing the aircraft was replaced by a truck. Bydriving over highway bridges and looking down into the valley a flight geometrywas simulated. Flight experiments were conducted later.

13. The APY-6 by Northrop-Grumman (GROSS and HOLT [177]), an airborne X-band radar with SAR and GMTI (ground moving target indication) capability.The antenna is subdivided into a large part for transmit and SAR receive, andthree smaller panels for MTI receive. The radar can be operated in sidelookingand forward-looking configuration. APY-6 is a low-cost approach using COTScomponents.

14. There are deliberations going on to install surveillance functions such as AWACSor JointSTARS in space. The Ground MTI is a key function in these concepts(COVAULT [86]).

15. The Airborne Stand-Off Radar (ASTOR) developed in the UK will have a GMTImode in conjunction with high-resolution SAR imagery in such a way thatmoving target scenes can be overlaid on SAR images (NORDWALL [391]).

16. AMSAR (Airborne Multirole Solid-state Active array Radar) is a currentEuropean project (Thomson-CSF, France; GEC Marconi, UK; DASA, Germany)concerning a nose radar for future fighter aircraft (ALBAREL et al [12],GRUENER et al [178]). The antenna is a circular planar array subdividedinto two-dimensional subarrays. The concept includes anti-jamming and STAPcapabilities.

17. SOSTAR (Stand Off Surveillance and Target Aquisition Radar) is a plannedEuropean project (HOOGEBOOM et al [214]). A sidelooking linear X-band

array will be mounted under the fuselage of an aircraft. The system will includeboth SAR imaging and moving target indication. Participating companies:Dornier (Germany), Thomson-CSF Detexis (France), FIAR (Italy), TNO-FEL(The Netherlands).

18. The UESA program (UHF Electronically Scanned Array) is an experimentalradar system including a circular ring array being fabricated by Raytheon, USA.Application might be Airborne Early Warning (AEW), possibly as a successorto systems such as AWACS (ZATMAN [577]).

19. RADARSAT-2 (Canada, to be launched in 2003) will be the first earthobservation satellite with MTI capability (MEISL et al. [346], THOMPSON andLIVINGSTONE [492]).

1.1.7 Validity of models

This textbook is meant to be a primary tutorial and reports on the status quo in researchin space-time processing. In order to make the various effects associated with thesuppression of moving clutter as clear as possible I have tried to simplify our models asmuch as possible. This may have the problem of losing contact with the 'real world'.I believe, however, that the underlying principles should be isolated and consideredseparately as a superposition of various effects causes problems in understanding. Itis certainly up to the system designer to refine such models up to a level which issufficient for his requirements.

Most of the considerations presented in this book have been carried out for onesingle range increment. This is justified by the fact that space-time clutter filtering isbasically carried out along each individual range ring. In practice, however, clutterstatistics of an individual range ring depend on other range increments. Such effectsare listed below.

1. Adaptation of the space-time covariance matrix has not been considered inthis book. It is assumed that the clutter covariance matrix is known for anyindividual range increment. In practice the clutter covariance matrix has to beestimated, e.g., by averaging clutter dyadics over various range rings. This maycause problems induced by range dependency of the clutter Doppler and non-homogeneity of clutter returns. These aspects exceed the scope of this book. InChapter 15 related literature has been quoted.

2. Mutual coupling effects (GUPTA and KSIENSKI [189]) between array elementson the performance of the array have been neglected.

3. It is assumed that the clutter returns are gaussian. Whenever we talk about'optimality' this refers to gaussian statistics. The space-time technique and itsapplication to non-gaussian clutter has been investigated by RANGASWAMY andMlCHELS [420]).

4. Multiple-time around clutter occurs whenever the PRF is chosen such that theradar is range ambiguous within the visible radar range. In most of the examples

presented in this book the multiple clutter returns have been neglected. Thisis in accordance with almost all available references on space-time processing.There are currently only a very few papers in the open literature that include theeffect of ambiguous clutter returns. The reason is that most papers on space-time processing focus on sidelooking radar. As will be shown in Chapter 3sidelooking radar is insensitive to ambiguous clutter. In Chapter 10 the effectof ambiguous clutter for forward-looking radar is specifically addressed. It isshown that an array antenna which is adaptive in two dimensions can suppressthe multiple returns.

5. I have assumed that the reflectivity of the ground is independent of thedepression angle. In practice there is a strong dependence which is in turnassociated with the kind of clutter background (roughness). This assumptionimplies also that the effect of altitude return is not specifically emphasised.

6. Most examples given in this book are based on linear arrays. As will turnout in Chapter 6 linear arrays have favourable properties concerning space-time processing. They are also most useful for illustrating the fundamentals.Therefore, most of the existing literature is focused on linear arrays. Moreover,linear arrays include all kinds of cylindrical array configurations whose axes arehorizontal, and rectangular planar arrays. In Chapter 8 some considerations oncircular arrays have been included.

1.1.8 Historical overview

A brief historical overview of the evolution of STAP is presented in Table 1.1. Thistable includes a number of major publications, programs or events which contributedto the development of STAR This listing is certainly not complete but includes to ourunderstanding the major steps in the development of STAR

1.2 Radar signal processing tools

In the following Sections of this chapter a few well-known signal processing toolswhich will be used in this book are summarised.

1.2.1 The optimum processor

The principle of detecting a signal vector s before a noisy background given by q isbriefly summarised. Let us define the following complex vector quantities:

(1.1)

Next Page

Front MatterTable of Contents1. Introduction1.1 Preliminary Remarks1.1.1 Basics of MTI Radar1.1.2 One-dimensional Clutter Cancellation1.1.3 Aspects of Air- and Spaceborne Radar1.1.4 Impact of Platform Motion1.1.5 Some Notes on Phased Array Radar1.1.6 Systems and Experiments1.1.7 Validity of Models1.1.8 Historical Overview

1.2 Radar Signal Processing Tools1.2.1 The Optimum Processor1.2.2 Orthogonal Projection1.2.3 Linear Subspace Transforms1.2.4 Clutter Suppression with Digital Filters1.2.5 Examples1.2.6 Angle or Frequency Domain Processing

1.3 Spectral Estimation1.3.1 Signal Match (SM)1.3.2 Minimum Variance Estimator, MVE1.3.3 Maximum Entropy Method, MEM1.3.4 Orthogonal Projection, MUSIC1.3.5 Comparison of Spectral Estimators

1.4 Summary

Index