A Multiple Beam Synthetic Aperture Radar Concept...

7
IEEE TRANSACTIONS GEOSCIENCE AND REMOTE SENSING, VOL. GE-21, NO. 2, APRIL 1983 A Multiple Beam Synthetic Aperture Radar Design Concept for Geoscience Applications BUFORD R. JEAN, MEMBER, IEEE, AND JOHN W. ROUSE, JR., SENIOR MEMBER, IEEE Abstract-The multiple beam synthetic aperture radar (SAR) has been presented as a superior alternative to conventional side-looking SAR designs for mapping earth resources from space. The multiple beam approach provides wide swath coverage at nearly constant incident angle while conserving peak power. These desirable system character- istics, however, are realized at the expense of additional signal process- ing. A design example illustrates the attributes of the multiple beam SAR in comparison to a conventional side-looking system. A novel beam formation processing architecture is shown which can provide simultaneous range compression while forming the beams. A con- figuration for the image formation processor is proposed and analyzed in terms of an analog implementation which can perform Doppler com- pression at speeds faster than real time. These implementation schemes indicate that on-board processing for the multiple beam SAR is possible. I. INTRODUCTION IN RECENT YEARS, the role of synthetic aperture radar (SAR) systems has been extended beyond gathering mili- tary intelligence to providing all weather, day or night, maps of the earth [I ]. There is a growing demand among scientists in a wide range of disciplines to obtain timely, high resolu- tion, wide swath active microwave imagery of the earth from a spacecraft platform in much the same way that optical spec- tra data are provided by the LANDSAT series of satellites [2]. The design of SAR systems for earth observation applications represents a different set of performance objectives from those identified for military systems, particularly when quantitative analysis of the image data is required. Conventional SAR systems, which employ a single side- looking antenna beam, fail to satisfy many important criteria imposed when monitoring earth resources. The criteria in- clude both image interpretation and system implementation requirements. A wide swath image formed with a single side- looking beam can be difficult to interpret because of the vari- ation in incident angle across the scene. The radar scattering coefficient uo is highly dependent upon incident angle. A large incident angle variation across the image can obscure the effects of other scene parameters on uo [3] . There is also geo- metric distortion introduced across the swath because spatial resolution in this dimension is also angle dependent. A side-looking SAR requires a high peak power transmitter Manuscript received June 17, 1982; revised October 15, 1982. B. R. Jean is with the Remote Sensing Center and Department of Electrical Engineering, Texas A&M University, College Station, TX 77843. J. W. Rouse, Jr. is with the Department of Engineering, The Univer- sity of Texas at Arlington, Arlington, TX 76019. for adequate signal-to-noise ratio and a long antenna in the along-track dimension for azimuth ambiguity reduction. These requirements compensate for the low pulse repetition fre- quency (PRF) needed for range ambiguity avoidance. In addi- tion, the complexity of the radar return signal makes the de- sign of signal processors difficult and expensive. An interesting system, called SCANSAR, has been suggested for achieving wide swath coverage that circumvents the ambi- guity problems of broad fan beam antennas and also conserves transmitter power when the antenna length is constrained. The approach is that of step scanning a narrow antenna beam over the desired swath and synthesizing a synthetic aperture image within the beam at each step in the scan [4]. With careful control of the dwell time and scanning rate, continuous coverage over a wide swath can be obtained. The scanning synthetic aperture radar, however, is limited in its resolution capability and in the amount of noncoherent averaging that can be done to reduce coherent speckle in the radar image. Also the currently proposed SCANSAR does not provide a constant incident angle image. The problems and limitations imposed by conventional side- looking SAR and the SCANSAR have led to an alternative sys- tem design concept first proposed by Claassen and Eckerman [5]. The concept is that of forming multiple real antenna beams by appropriately processing the signals from the individ- ual elements of a single receiving array. The beams are pro- jected along the surface of a cone as shown in Fig. 1, so as to intersect the earth at a nearly constant angle of incidence. The returns from the independent beams are further pro- cessed to provide a high resolution synthetic aperture image. This paper presents the basic architecture of a multiple beam SAR and suggests possible implementation techniques for the various subsystems. Antenna beam requirements are related to the image resolution and swath coverage requirements. A novel beam forming technique which provides simultaneous beam formation and range compression is identified. A pro- cessing technique for SAR image formation which lends itself to real-time analog electronic implementation is also presented and analyzed. The advantages of the multiple beam SAR concept, in addi- tion to providing wide swath coverage at a nearly constant incident angle, are that it reduces the peak, and often the average, transmitter power requirements, it gives considerable relief from the range and azimuth ambiguity suppression con- straints, and it employs an antenna which may be more easily deployed in space. The penalty associated with the multiple 0196-2892/83/0400-0201$01.00 © 1983 IEEE 201

Transcript of A Multiple Beam Synthetic Aperture Radar Concept...

Page 1: A Multiple Beam Synthetic Aperture Radar Concept ...web.ecs.baylor.edu/faculty/jean/tenure/04157387.pdf · ratio improvement over a single beam system giving com-parable swath coverage.

IEEE TRANSACTIONS GEOSCIENCE AND REMOTE SENSING, VOL. GE-21, NO. 2, APRIL 1983

A Multiple Beam Synthetic Aperture Radar DesignConcept for Geoscience Applications

BUFORD R. JEAN, MEMBER, IEEE, AND JOHN W. ROUSE, JR., SENIOR MEMBER, IEEE

Abstract-The multiple beam synthetic aperture radar (SAR) has beenpresented as a superior alternative to conventional side-looking SARdesigns for mapping earth resources from space. The multiple beamapproach provides wide swath coverage at nearly constant incidentangle while conserving peak power. These desirable system character-istics, however, are realized at the expense of additional signal process-ing. A design example illustrates the attributes of the multiple beamSAR in comparison to a conventional side-looking system. A novelbeam formation processing architecture is shown which can providesimultaneous range compression while forming the beams. A con-figuration for the image formation processor is proposed and analyzedin terms of an analog implementation which can perform Doppler com-pression at speeds faster than real time. These implementation schemesindicate that on-board processing for the multiple beam SAR is possible.

I. INTRODUCTIONIN RECENT YEARS, the role of synthetic aperture radar(SAR) systems has been extended beyond gathering mili-

tary intelligence to providing all weather, day or night, mapsof the earth [I ]. There is a growing demand among scientistsin a wide range of disciplines to obtain timely, high resolu-tion, wide swath active microwave imagery of the earth froma spacecraft platform in much the same way that optical spec-tra data are provided by the LANDSAT series of satellites [2].The design of SAR systems for earth observation applicationsrepresents a different set of performance objectives from thoseidentified for military systems, particularly when quantitativeanalysis of the image data is required.Conventional SAR systems, which employ a single side-

looking antenna beam, fail to satisfy many important criteriaimposed when monitoring earth resources. The criteria in-clude both image interpretation and system implementationrequirements. A wide swath image formed with a single side-looking beam can be difficult to interpret because of the vari-ation in incident angle across the scene. The radar scatteringcoefficient uo is highly dependent upon incident angle. Alarge incident angle variation across the image can obscure theeffects of other scene parameters on uo [3] . There is also geo-metric distortion introduced across the swath because spatialresolution in this dimension is also angle dependent.A side-looking SAR requires a high peak power transmitter

Manuscript received June 17, 1982; revised October 15, 1982.B. R. Jean is with the Remote Sensing Center and Department of

Electrical Engineering, Texas A&M University, College Station, TX77843.

J. W. Rouse, Jr. is with the Department of Engineering, The Univer-sity of Texas at Arlington, Arlington, TX 76019.

for adequate signal-to-noise ratio and a long antenna in thealong-track dimension for azimuth ambiguity reduction. Theserequirements compensate for the low pulse repetition fre-quency (PRF) needed for range ambiguity avoidance. In addi-tion, the complexity of the radar return signal makes the de-sign of signal processors difficult and expensive.An interesting system, called SCANSAR, has been suggested

for achieving wide swath coverage that circumvents the ambi-guity problems of broad fan beam antennas and also conservestransmitter power when the antenna length is constrained.The approach is that of step scanning a narrow antenna beamover the desired swath and synthesizing a synthetic apertureimage within the beam at each step in the scan [4]. Withcareful control of the dwell time and scanning rate, continuouscoverage over a wide swath can be obtained. The scanningsynthetic aperture radar, however, is limited in its resolutioncapability and in the amount of noncoherent averaging thatcan be done to reduce coherent speckle in the radar image.Also the currently proposed SCANSAR does not provide aconstant incident angle image.The problems and limitations imposed by conventional side-

looking SAR and the SCANSAR have led to an alternative sys-tem design concept first proposed by Claassen and Eckerman[5]. The concept is that of forming multiple real antennabeams by appropriately processing the signals from the individ-ual elements of a single receiving array. The beams are pro-jected along the surface of a cone as shown in Fig. 1, so as tointersect the earth at a nearly constant angle of incidence.The returns from the independent beams are further pro-cessed to provide a high resolution synthetic aperture image.This paper presents the basic architecture of a multiple beam

SAR and suggests possible implementation techniques for thevarious subsystems. Antenna beam requirements are relatedto the image resolution and swath coverage requirements. Anovel beam forming technique which provides simultaneousbeam formation and range compression is identified. A pro-cessing technique for SAR image formation which lends itselfto real-time analog electronic implementation is also presentedand analyzed.The advantages of the multiple beam SAR concept, in addi-

tion to providing wide swath coverage at a nearly constantincident angle, are that it reduces the peak, and often theaverage, transmitter power requirements, it gives considerablerelief from the range and azimuth ambiguity suppression con-straints, and it employs an antenna which may be more easilydeployed in space. The penalty associated with the multiple

0196-2892/83/0400-0201$01.00 © 1983 IEEE

201

Page 2: A Multiple Beam Synthetic Aperture Radar Concept ...web.ecs.baylor.edu/faculty/jean/tenure/04157387.pdf · ratio improvement over a single beam system giving com-parable swath coverage.

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. GE-21, NO. 2, APRIL 1983

Vertically mounted' antenna array

Fig. 1. A multiple beam configuration for a wide swath, constantincident angle synthetic aperture radar.

antenna beam configuration is the additional signal processingrequired to form the beams. For the modest resolutions oftenassociated with orbital mapping of earth resources, the in-creased signal processing burden required for beam formationcan be somewhat offset by the reduced complexity in the SARimage formation processor.

II. THE MULTIPLE BEAM CONFIGURATION

The antenna for the multiple beam SAR is mounted verti-cally. It consists of a rectangular array which is oriented withrespect to the vehicle velocity vector as depicted in Fig. 2.The array is rotated about its vertical axis (the z axis in Fig. 2)to point the direction of t/ . Multiple beams are formed about¢0 to give angular coverage from tmin to Omax- The verticaldimension of the array is designed to produce a narrow eleva-tion beamwidth. This contrasts with the side-looking config-uration which requires a broad elevation beam.

In the horizontal dimension, signals from the individual ele-ments are combined by a beam formation processor to yieldoutput signals corresponding to multiple, independent antennabeams, each having the gain of the full array. These signals arefurther processed to form a synthetic aperture image withineach of the real antenna beams. A functional block diagram ofthe multiple beam SAR is shown in Fig. 3.The incident angle, swath width, and resolution require-

ments for the SAR system constrain the design of the an-tenna array. Consider first the inside portion of the swath(t near kmin). The available synthetic aperture length is set bythe elevation beamwidth, ,Be. The relationship may be writtenapproximately as

vTa-Rop e

Cos cos 00where

vTa is the aperture integration length (velocity times aper-ture integration time),

Ro is the range to the beam center,0 is the azimuth angle variable, and00 is the incident angle (elevation angle variable)

If N1 independent looks are required, with an along-trackresolution px, the aperture integration length must be such

Fig. 2. Antenna orientation and imaging geometry for the multiplebeam SAR.

MITTING RAOMNTRTROLLTANTENNAA

RECEIVING > N ARASR IMAGEPRENDEBEAM NB IEEEENT MAGE IMAGEANTENNA > _ELEMENT FORMATION SAR IMAGE COMBINING PRODUCTARRAY RECEIVER PROCESSOR PROCESSING PROCESSORO

> CHANNELS CHANNELS

Fig. 3. Block diagram of a multiple beam SAR system.

that

vTa, > N, X Ro2p, (1 - sin2 00 COS2 t)

where X is the wavelength [6]. Note that px is the "along-track" resolution, not the azimuth resolution. The distancePx represents the projection of the true azimuth resolution inthe along-track direction. The use of px here makes sense,since a target passes through the beam in the along-track or xdirection. A minimum value for,e is

NiXcos oCos.min2 PX (1 - sin2 00 cos2 tkmin)

The maximum elevation beamwidth is determined by rangeambiguity constraints. Expressed in terms of AR, the unam-biguous slant range interval, the bound on 13e is

AR cot 00ge < Ro

Slant range interval and the along-track resolution must betraded via limits placed on the PRF in sampling the Dopplerspectrum. The Nyquist sampling rate for range offset process-ing may be expressed in terms of the vehicle velocity, along-

202

Page 3: A Multiple Beam Synthetic Aperture Radar Concept ...web.ecs.baylor.edu/faculty/jean/tenure/04157387.pdf · ratio improvement over a single beam system giving com-parable swath coverage.

JEAN AND ROUSE: SAR RADAR DESIGN CONCEPT FOR GEOSCIENCE APPLICATIONS

track resolution and number of looks

PRF> vNjPx

To avoid range ambiguities, the slant range interval and thePRF are related as

CAR < 2R - cr (2)

2 PRF

where c is the electromagnetic wave velocity of propagation,and r is the transmitted pulse width. The Doppler spectrumand range ambiguity limits may be combined in a single ex-pression as

ARK Ccp .2vN1

Thus the maximum elevation beamwidth may be written interms of the along-track resolution, the number of multiplelooks, the incident angle, the vehicle velocity, and the trans-mitted pulse width. The resulting expression is

cc< C cot o0 Px 7 (3)fe R0 2vN1- 1 3

The limits imposed by (1) and (3) are rather broad and allowconsiderable flexibility in specifying fe. The upper limit onO3e is a significant constraint for a side-looking system sincewide swath coverage requires a large elevation beamwidth.This limit is seldom of concern for the multiple beam config-uration, since wide swath coverage is obtained in the azimuthdimension, not in range. In general, the elevation beamwidthwill be chosen to be as small as possible to take advantage ofthe corresponding large antenna gain.The considerations for determining the horizontal dimension

specifications are similar to those for the vertical dimension.An additional requirement is that the beamwidth in 0 mustbe wide enough at each beam position to assure overlappingcoverage across the entire swath. For large 0, it is the hori-zontal beamwidth which determines the synthetic aperturelength. This forces a lower limit on gh which is similar to therequirement on f3e as expressed in (1). That is, Oh is boundedby the approximate inequality

N1 X/sin 'kmaxh

p (1 - sin2 So Cos2 4max)Again it is desirable to achieve as small a beamwidth as pos-sible to reduce power requirements.The familiar expression for the signal-to-noise ratio of a syn-

thetic aperture radar may be applied to the multiple beamSAR with slight modification [7]. The signal-to-noise ratiomay be written as

PavgDhtDhrDe PrSINT a hre3 P (4)

NB 817k ToFhR3 XV sin 0

where

Pavg is the average transmitted power,ao is the cross section per unit area,

Pr is the slant range resolution,k is the Boltzman's constant,To is the reference temperature,Fn is the receiver noise figure,Dht is the horizontal dimension of the transmitting antenna,Dhr is the horizontal dimension of the receiving antenna,De is the receiving and transmitting antenna elevation

dimension, andNB is the number of beams.

The transmitting antenna beamwidth in the horizontal di-mension is roughly NB times the beamwidth of the receivingantenna, where NB is the number of beams. The signal-to-noise ratio thus appears to be reduced by the factor I/NBfrom that of a competing single beam system. However, forcomparable swath coverage from a single beam system, bothtransmitting and receiving beamwidths would be larger by afactor of NB for an overall 1/NB signal-to-noise ratio reduc-tion. Thus in terms of antenna gain alone, the multiple beamsystem appears to offer roughly a factor ofNB signal-to-noiseratio improvement over a single beam system giving com-parable swath coverage. However, this signal-to-noise ratioadvantage is lost in the beam formation network since theoverall processing bandwidth must be NB times that of asingle beam system.A signal-to-noise ratio advantage that is not lost however, is

that due to the allowable increase in the PRF. The PRF canbe increased by the same amount that the elevation beam-width is reduced. Typically the repetition rate can be in-creased by a factor of at least NB over a conventional singlebeam side-looking system. Because of the accompanying in-crease in the PRF, increasing the length of the antenna in theelevation dimension has a two-fold impact on reducing thepeak power requirement. For wide swath designs this savingsin peak power can be significant.In addition to reducing the peak power requirement, the

multiple beam configuration can significantly reduce the dy-namic range requirement of the receiver. Because the imageis formed at a nearly constant incident angle, the large varia-tion in ao as a function of angle no longer becomes a factor inspecifying the dynamic range of the system. As a practicalmatter, this also leads to an improvement in signal-to-noiseratio for systems employing an analog to digital (A/D) con-verter. A reduction in dynamic range for a fixed number ofbits in the A/D leads to a reduction in quantization noise. Ifquantization noise is not a limiting factor then the number ofbits can be reduced.

III. A DESIGN EXAMPLE

To illustrate the potential advantages of the multiple beamconfiguration over conventional side-looking designs, a roughset of calculations have been made for a system which wouldbe somewhat comparable to the SEASAT SAR [8]. The rel-evant specifications for the SEASAT SAR are listed in Table I.The design procedure for specifying the multiple beam SAR

starts with determining the elevation beamwidth. From (1)the minimum value of elevation beamwidth is calculated to be1.120. It is not advisable to design for this minimum beam-

203

Page 4: A Multiple Beam Synthetic Aperture Radar Concept ...web.ecs.baylor.edu/faculty/jean/tenure/04157387.pdf · ratio improvement over a single beam system giving com-parable swath coverage.

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. GE-21, NO. 2, APRIL 1983

TABLE ISEASAT-A SAR SYSTEM CHARACTERISTICS (AFTER JORDAN [8]).

Satellite Altitude 800 kmWavelength 0.235 mRF Bandwidth 19 MHzPulse Repetition Frequency 1463 to 1640 HzTime Bandwidth Product 634Radar Transmitter Peak Power 1000 wattsRadar Transmnitter Average Power 55 wattsRadar Antenna Dimnensions 11 by 2.3 in

width although it would be desirable to do so. If the antennalength in the elevation dimension can be as large as the SEASATantenna length in a horizontal dimension, that is 11 m, then an

elevation beamwidth of 3.160 can be achieved. The antennabeam must be steered downward by approximately 700 fromthe broadside direction to provide a nominal 200 incidentangle.The minimum spacecraft altitude is dictated by the swath

width specification and the angular extent over which beamsare formed. If beams are formed within the angular regionfrom 0 = 100 to = 500, then the minimum altitude is com-

puted to be about 464 km. For this design, a value of 500 kmwill be taken as nominal. With the specification of the beam-width and altitude, the unambiguous slant range interval can

be computed and is found to be approximately 10.7 km.From (2), an upper limit on the PRF can be calculated. Themaximum PRF is found to be about 14 kHz.The horizontal dimension design requires computation of

the element spacing for the horizontal array. A maximum ele-ment spacing of about 1 m or roughly 4.3 X is required to scan

an angular region of ±200 in the dimension. Recall fromFig. 2 that is measured from the x axis in the x-y plane.With respect to the horizontal array, the scan angle is mea-

sured from the z axis in the plane of the antenna array. Toscan ±200 in 0 requires an actual scan angle measured withrespect to the antenna of ±6.720.A number of tradeoffs must be considered before an op-

timum selection of the number of beams and the number ofelements in the horizontal array can be made. For the sakeof brevity, we assume that four elements will do an adequatejob. The element width must provide an elemental patternwhich will control the grating lobes. An element width of ap-

proximately 0.9 m should be adequate. The overall antennasize turns out to be about 4 X m. This is somewhat largerin the narrow dimension than the SEASAT antenna but is notan unreasonable size.

If beams are formed symmetrically about the antenna broad-side then the beam pointing angles, measured in X, occur at= 18, 26, 34, and 42 degrees. The width of each beam will

be on the order of 200. This beamwidth figure takes into ac-

count aperture weighting to reduce side lobes levels to accept-able values. A 200 beamwidth provides adequate overlap to

insure contiguous coverage across the entire swath.The antenna beamwidths define the Doppler ambiguity con-

straints. A maximum Doppler bandwidth of about 5 kHz can

be expected. The PRF can therefore be selected from therange of about 5 to 14 kHz. A nominal value of 10 kHz seems

reasonable for the calculations which follow.The required average power can be computed from (4) using

parameter values comparable to those for the SEASAT radar.

The minimum signal-to-noise ratio is chosen to be 10 dB, thereceiver noise figure is estimated at 6 dB plus an additional6 dB to account for the wider bandwidth required in the beamformation processor and the minimum average value of ao at200 is - 13 dB. A nominal value for velocity is 7.5 km/s. Theslant range resolution corresponding to 19 MHz is 15.79 m.From these values the total average transmitter power is

found to be 16.96 W while the peak power is 50.76 W. Thepeak power reduction achieved over the SEASAT design is ap-proximately 13 dB. If these numbers are adjusted for the dif-ferences in altitude, the average powers become comparablewhile the peak power is still reduced by 6.8 dB.While many other factors would enter into play in a detailed

system design study, the power reduction indicated here issignificant and demonstrates one principal advantage of themultiple beam configuration.

IV. BEAM FORMATION TECHNIQUES

There are a number of ways to implement multiple beamformation. Both digital and analog techniques for formingmultiple beams following coherent reception of the returnsignal by the individual horizontal element receiver channelsare available [9], [10], [11], [12].Analog beam forming requires time or frequency multiplex-

ing of the individual element receiver outputs. Johnson [9]has shown that if the antenna element signals are time multi-plexed, then the antenna beam signals are frequency multi-plexed. Conversely, frequency multiplexed elements produceoutput signals that are multiplexed in time. A novel imple-mentation utilizing the chirp Z-transform (CZT) algorithm isshown in Fig. 4. The CZT implementation incorporates timemultiplexing of the element signals precisely as described byJohnson. However, by virtue of the Fourier transform opera-tion, the resulting frequency multiplexed beam signals aretransformed into time multiplexed beam signals. An addi-tional outcome of utilizing the CZT algorithm is that rangecompression is accomplished within the same processingoperation.The CZT technique requires that the input signal be premul-

tiplied by a chirp signal of the form ej1I12pt . The multipliedsignal is then passed through a dispersive filter whose impulseresponse is a similar chirp waveform of the opposite sense[13]. A second multiplication by e_12't2 following the chirpfilter removes a residual quadratic phase factor, if required.

In the beam formation processor of Fig. 4, the premulti-plication function occurs in the transmitter. The transmitterwaveform is a linear FM pulse of duration T seconds with achirp rate of - p rad/s2.The receiving array element outputs are multiplexed at a

sampling rate of ws. The sampling frequency must be greaterthan the Nyquist rate in order to provide adequate time sep-aration between the beams. The time separation betweenbeams must be large enough to allow the entire range intervalto be processed for one beam direction before the next beamsignal appears.The bandwidth of the multiplexed signal is determined by

"s and the number of beams formed. It is roughly equal toNB,s/27r Hz. Since the processing filter must be matched to

204

Page 5: A Multiple Beam Synthetic Aperture Radar Concept ...web.ecs.baylor.edu/faculty/jean/tenure/04157387.pdf · ratio improvement over a single beam system giving com-parable swath coverage.

JEAN AND ROUSE: SAR RADAR DESIGN CONCEPT FOR GEOSCIENCE APPLICATIONS

Tapped Delay Line =RO f5

Fig. 4. An analog beam formation processor based upon the CZTalgorithm.

the transmitter linear FM rate, the impulse response durationTf of the filter can be computed as

Tf NB&S Tf27rB

where B is the RF bandwidth in Hertz.Jean has shown that the output signal from the processing

filter is of the form [6]

ET2iAt2 E sin (gT/2) (t - mcs/p)Eout m (UT/2) (t - mslM)

E,el' [kdh sin 00 sin- N2 ],wheeB

where

Enkdh~

is the amplitude of the signal from the nth element,is the wave number (2ir/X), andis the horizontal element spacing.

The sin x/x term in the first summation appears due to thepulse compression operation. It is noted that the output foreach beam occurs at time

mwSt-.

The second summation indicates the formation of the individ-ual antenna beams.A potential disadvantage of the CZT beam forming proces-

sor is that it imposes another limit on the PRF of the trans-mitter. If the full unambiguous range interval is to be pro-

cessed, the PRF must be reduced so that all NB range intervalscan be processed before the return from the next succeedingpulse arrives at the receiver. This need not be a serious limita-tion for the multiple beam SAR, however, provided that theantenna elevation beamwidth can be made sufficiently narrow.

The problem may also be circumvented by employing a range

gating operation prior to beam formation to allow only the

range interval of interest to enter the CZT processing filter.

V. IMAGE FORMATION PROCESSING FOR THE

MULTIPLE BEAM SAR

Aside from the beam formation operation, the processingsteps required to form the synthetic aperture image are essen-

tially the same for both the multiple beam and conventionalside-looking radars. A processing scheme is presented below

' "SOMRSIN IMAGEA, FI|LTER STORAGE

DOPPLER SAMPLE STORAGE BUFFERS

Fig. 5. Architecture of a Doppler processor for the multiple beamSAR.

which is particularly well suited to the multiple beam con-figuration. The processor architecture takes advantage of thefact that the imaging geometry naturally conforms to therange-Doppler coordinates of the radar.Swath coverage is obtained in the Doppler dimension, along

contours of constant range. Side-looking systems achieve wideswath coverage in the range dimension along a contour of con-stant Doppler, usually along the zero Doppler line. For side-looking systems the Doppler compression operation consti-tutes the major processing burden. Because the Doppler chirprate is a function of range, many Doppler filters are requiredto process a wide swath image. A separate Doppler compres-sion operation must be done for each range resolution interval.The approach taken in processing the multiple beam image

results in having to process only a few range bins. More signifi-cant than the reduction in the number of range bins is the factthat Doppler compression operation can now be accomplishedwith a single Doppler filter. This comes about because theimaging geometry provides a nearly constant Doppler chirprate across the entire swath. The filter bandwidth is deter-mined by the Doppler bandwidth within a single antennabeam. The Doppler dimension resolution is determined bythe number of Doppler samples that are integrated by thefilter.An architecture for a Doppler dimension processor is shown

in Fig. 5. The configuration is based upon a processing tech-nique suggested by Claassen [5]. The processor has a mastercomputer which provides control of the range gating operationand calculates the adjustments required for motion compensa-tion. The function of the range gate controller is to assurethat the signal samples are collected in a precisely controlledsequence. The Doppler signal history from which a single setof azimuth resolution elements are formed is collected from anarrow strip of terrain through the center of the beam. Acomplete sample set corresponds to the return signal collectedfrom a strip roughly equivalent to a single range bin. Therange gate for a given strip is continuously advanced so as tocapture successive pulses from the same patch of terrain as thevehicle advances toward the patch.To obtain continuous imaging along the swath several range

intervals must be tracked simultaneously. The number ofrange bins required is roughly given by

N, vTaNrb =px

205

Page 6: A Multiple Beam Synthetic Aperture Radar Concept ...web.ecs.baylor.edu/faculty/jean/tenure/04157387.pdf · ratio improvement over a single beam system giving com-parable swath coverage.

IEEE TRANSACTIONS ON GEOSCIENCE AND REMOTE SENSING, VOL. GE-21, NO. 2, APRIL 1983

Direction

Fig. 6. Range tracking geometry required for obtaining the Dopplersignal history from a given azimuth strip within a single beam. (Beamsize and aperture length are greatly exaggerated.)

Fig. 6 provides an illustration of the range tracking opera-

tion. Consider the strip of terrain through the center of thebeam that corresponds to the range bin centered on range R 1 .

At time t1 = 2R 1 Ic, after the first pulse is emitted, the samplecorresponding to this constant range contour is acquired. Thesecond pulse emitted by the transmitter occurs at t = 1/PRF.The range to the center of the strip being sampled has de-creased from R I 'to R2, where R'2 is approximately

vR2 R1 - PRF sin O cosks.

The range gate for the second sample in the Doppler signalhistory of this strip of terrain must occur earlier in time byan amount Ati given by

2v sin 00 cos /iA tj =

cPRF

This range gate adjustment is repeated for each successivepulse until the required number of samples have been collected.

It must be noted that while the sample time can be adjustedto account for the decrease in range to the center of the strip,the range bins rotate about this center point as the radar ad-vances along the flight path. There is also an increase in thecurvature of the range bins as the range to the strip decreases.This rotation and change in curvature introduces a range track-ing error at the ends of the strip which depends upon thevelocity, the integration time, the range to the target, and theantenna beamwidth. The ground range error can be com-

puted by approximating the strip as a straight line segment oflength

1 = Rghwhere R is the nominal range to the center of the strip and ghis the azimuth beamwidth. The line segment rotates by an

amount

vTa sin ¢kR sin 0O

The ground range error is approximately

ARg = 1/2 lAb.

If this error is to be less than pg, the ground range resolu-tion, then the following inequality must hold

< 2pg sinOoOh sin6i

A. The Doppler Compression OperationTo form the individual image resolution elements, in the

azimuth dimension the collected samples must be processedby the Doppler compression filter. As noted before, the band-width of this filter is governed by the Doppler signal band-width across the antenna beam, whereas the frequency resolu-tion is governed by the integration time, i.e., by the number ofsamples that are collected.The identity of each resolution element appearing at the

filter output is expressed in terms of the time delay of thesignal through the filter. If cdj is the offset frequency of thejth Doppler resolution element, referenced to the beam center,and Ad is the linear FM rate, then this time delay may bewritten as

T i::Jdj

Ad

It is not necessary to perform the Doppler compressionfunction at the same rate that data are collected. Since all ofthe data are available in storage, the data may be processed ata rate much faster than the real-time acquisition rate [14].The adjustments necessary for motion compensation are also

easily incorporated into the Doppler filtering operation. Vari-ation in vehicle velocity create corresponding variations in theDoppler chirp rate. The chirp rate of the stored data can bedynamically corrected to compensate for nonuniform velocityby simply varying the rate at which the data are extractedfrom the sample storage buffer [13] Since the Doppler chirprate is very nearly constant over the entire imaged swath, thisform of motion compensation is applied to all the data atonce.This chirp rate adjustment technique implies the use of an

analog filter. Other filter implementations are possible withappropriate resampling of the data after the signal has beenconverted to an analog form. Of course an all digital pro-cessor is possible, but such would lack the simplicity of thevariable clock-rate motion compensation scheme.

B. Continuous Coverage and Multiple Look ProcessingAs noted previously, several range bins must be tracked con-

currently to assure continuous coverage along the swath. Theprocessor must begin tracking a new range interval each timethe vehicle advances by a distance equal to one ground rangeresolution element. This interval must be tracked until thevehicle has advanced by one aperture integration length. If amultiple look image is desired this interval must be tracked fora number of aperture lengths equal to the number of multiplelooks N1.The multiple look images from a given strip within the beam

will not overlay precisely due to the previously identifiedrange error which occurs at the ends of the strip. If the samecriterion for the maximum allowable range error is applied for

206

Page 7: A Multiple Beam Synthetic Aperture Radar Concept ...web.ecs.baylor.edu/faculty/jean/tenure/04157387.pdf · ratio improvement over a single beam system giving com-parable swath coverage.

JEAN AND ROUSE: SAR RADAR DESIGN CONCEPT FOR GEOSCIENCE APPLICATIONS

multiple look registration as was specified for the range errorin sampling the Doppler signal hstory, then the number oflooks which can be combined without range error correctionis limited by

2pg sin O6N h sin 0 vTa

VI. CONCLUSIONSThe multiple beam synthetic aperture design concept has

been presented and compared with conventional side-lookingarchitectures. Calculations have shown that the multiple beamdesign offers considerable savings in transmitter power require-ments at the expense of additional processing requirements.Other attributes of the design include the ability to image ata nearly constant incident angle while achieving wide swathcoverage.A processing scheme has been presented for the multiple

beam SAR which offers the potential of on-board, real-timeimage formation. The processor is relatively simple in con-cept and lends itself to implementation with CCD and SAWtechnologies.

REFERENCES[1] R. E. Mathews, Ed., Active Microwave Users Workshop Report,

NASA Conference Publication 2030, Johnson Space Center(Houston, TX), Aug. 1976.

[2] R. E. Mathews, Ed., Active Microwave Workshop Report, NASASP-376, Scientific and Technical Information Office, NationalAeronautics and Space Administration, Washington, D.C., 1975.

[3] B. J. Blanchard, "Analysis of synthetic aperture radar imaging,"Final Report, 3359-2, Remote Sensing Center, Texas A&MUniversity, June 1977.

[4] R. K. Moore, J. P. Claassen, and Y. H. Lin, "A scanning space-borne synthetic-aperture radar (SCANSAR) and a combinedmicrowave radiometer sensor," University of Kansas RemoteSensing Laboratory, Submitted to IEEE Transaction of AES,January 1980.

[5] J. P. Claassen, and J. Eckerman, "A system concept for wideswath constant incident angle coverage," in Proc. SyntheticAperture Radar Technology Conf., New Mexico State Univer-sity (Las Cruces, NM), Mar. 1978.

[6] B. R. Jean, "Multiple antenna beam formation techniques forsynthetic aperture radar," Tech. Rep. RSC-99, Remote SensingCenter, Texas A&M University, Dec. 1978.

[7] L. J. Cutrona, "Synthetic aperture radar," in Radar Handbook,M. I. Skolnik, Ed. New York: McGraw-Hill, 1970, ch. 23, pp.23-1 to 23-25.

[8] R. Jordan, "The Seasat-A synthetic aperture radar design andimplementation," in Proc. Synthetic Aperture Radar TechnologyConf., New Mexico State University (Las Cruces, NM), Mar.1978.

[91 M. A. Johnson, "Phased array beam steering by multiplex sam-pling," Proc. IEEE, vol. 56, no. 11, pp. 1801-1811, Nov. 1968.

[10] S. Haykin, "Efficient implementation of multiple beam samplesfor continuously scanned array antennas,"Electron. Lett., vol. 12,no. 2, p. 48, Jan. 22, 1976.

[11] D. E. N. Davies, "A fast electronically scanned radar receivingsystem," Journal Brit. l.R.E., pp. 305-321, Apr. 1961.

[12] W. P. Delaney, "An RF multiple beam-forming technique,"IRE Trans. Mil. Electron., pp. 179-186, Apr. 1962.

[13] R. C. Williamson, "Properties and applications of reflective-array devices," Proc. IEEE, vol. 64, no. 5, pp. 702-710, May1976.

[14] J. P. Claassen, "The potential of incorporating a synthetic aper-ture radar capability into the meteorological radar facility,"Tech. Rep. RSC-87, Remote Sensing Center, Texas A&M Uni-versity, May 1977.

Buford Randall Jean (S'67-M'71) was born inHilisboro, TX. on May 14, 1948. He receivedthe B.S., M.S., and Ph.D. degrees all in electri-cal engineering from Texas A&M University,College Station, in 1970, 1971, and 1978,respectively.From 1971 to 1974, he worked at the Bendix

| 2Research Labs in Southfield, MI, where he wasan engineer in the Electromagnetic Systems andTechnology Department. In 1974 he returnedto Texas A&M University, College Station, as

an Engineering Research Associate at the Remote Sensing Center. Hecurrently holds a joint appointment between the Remote Sensing Cen-ter and the Department of Electrical Engineering. His current interestsinclude microwave sensor development and signal processing.Dr. Jean is a member of Phi Eta Sigma, Phi Kappa, Tau Beta Pi, and

Eta Kappa Nu. He is a Registered Professional Engineer in the State ofTexas.

JJohn W. Rouse, Jr. (M'62-SM'71) was born inKansas City, MI. He received the B.S.E.E. de-gree from Purdue University, West Lafayette,

> Indiana, in 1959, and the M.S.E.E. and Ph.D.degrees from the University of Kansas, Law-rence, in 1965 and 1968, respectively.From 1968 to 1978 he was Professor of

Electrical Engineering and Director of theRemote Sensing Center at Texas A&M Uni-versity. He was Program Manager of micro-wave remote sensing.research at NASA Head-

quarters for two years (1975-1977) on leave from Texas A&M. Hewas the E.A. Logan Professor of Engineering and Chairman of theElectrical Engineering Department at the University of Missouri-Columbia from 1978 to 1981. He is presently Dean of Engineeringat The University of Texas at Arlington. He has been active in micro-wave remote sensing and electromagnetic scattering research since1964.Dr. Rouse is a Registered Professional Engineer in the states of

Missouri and Texas and is a member of Tau Beta Pi, Sigma Xi, andEta Kappa Nu.

207