Feaaeeeee
`
`ANTENNAS
`
`
`
`san
`
`ary1
`
`Next-generation platforms will demand next-generation antenna performance. (TRW
`photo)
`
`no
`roa
`we
`wey,
`
`is
`
`Hans Steyskal
`
`| Digital Beamforming Basics
`Aes as evermore sophis-
`
`ntenna performance is steadily
`
`ticated information is demanded
`from radar and communications sys-
`tems. Digital beamforming (DBF) is a
`powerful technique for boosting anten-
`na performance. Beamforming in the
`strict sense corresponds to weighting
`and summing elemental signals, where-
`as digital beamforming includes almost
`any spatial processing ofdigitized sig-
`nals from a sensor array. State-of-the-art
`microwave integrated circuits, signal
`processing and high-speed digital elec-
`tronics are now beginning to make DBF
`feasible for microwave radar and com-
`munications. The importance of this
`technology will continue to grow. This
`paper describes general DBF concepts
`and early implementation.
`Each element of an array antenna re-
`ceives or transmits a signal. In the re-
`ceive mode, analog beamformers output
`the weighted sum of the sensorsignals,
`reducing the signal dimensionality from
`the number of elements N to 1. Unlike
`analog beamformers, DBF arrays digi-
`tize received signals at the element
`level, then process these signals in a
`special-purpose digital processor to
`form the desired beam. The total infor-
`mation available at the aperture is pre-
`served and is represented by the N indi-
`vidual elementsignals.
`Once the physical input signals have
`been properly digitized, they can be ma-
`nipulated indefinitely, without incurring
`any further error since a digital repre-
`
`sentation of the signal is used rather
`than the real received signal power. As
`a result, any number of beams can be
`formed orthe signal can be subjected to
`multiple hypothesestesting.
`
`DBF PROMISES
`Digital beamforming is applicable
`both on transmit and receive, although
`mostofits advantages are realized in the
`receive mode. The most important ad-
`vantages are fast adaptive pattern
`nulling, super resolution and direction
`finding, antenna self-calibration and ul-
`tralow sidelobes, array element failure
`and pattern correction, closely spaced
`multiple beams, adaptive space-time
`processing and flexible radar power and
`time management. Many of these fea-
`tures have been addressed previously.'!4
`Adaptive space-time processing is a re-
`cently proposed technique>© for air-
`borne surveillance radars to suppress
`groundclutter that is spread over a large
`Doppler frequency band. DBF in the
`transmit mode mayfind its first applica-
`tion in cellular telephone networks.
`
`DBF IMPLEMENTATION
`A generic DBF array consists of an-
`tenna elements, receiver modules, ana-
`log-to-digital (A/D) converters and a
`digital beamformer and controller, as
`shown in Figure 1. The simple ap-
`proach of directly sampling and digitiz-
`ing the incoming microwave signal is
`not yet practical because of digital hard-
`ware speed limitations. Therefore, the
`
`50
`
`Page 1 of 6
`
`JOURNAL of ELECTRONIC DEFENSE « JULY '96
`
`SAMSUNG EXHIBIT 1036
`
`Page 1 of 6
`
`SAMSUNG EXHIBIT 1036
`
`

`

`DIGITAL
`BEAMFORMER
`
`Wu Xn
`
`bi
`Fig. 1 The generic DBFarray.
`
`
`_modules comprise complete heterodyne
`
`_ receivers, performing frequency down-
`_conversion, filtering and amplifying to a
`
`_ power level commensurate with the
`
`__A/D converter, as shown in Figure 2.
`_ Separate A/D conversion of baseband I
`
`|and Q signals allows a low sampling
`rate but requires channel matching. IF
`sampling anddigital 1/Q generation
`| avoids this problem at the expense of a
`
`_ higher sampling rate. The equalizer
`_compensates for dispersion differences
`
`_between individual receivers. The
`_downconversion can be done in one or
`
`|more steps, which are governed by
`
`_ hardware considerations.
`___
`After A/D conversion, the signals
`_ enter the actual digital beamformer, a
`_fast parallel processor that forms the
`_inner product beams at a rate commen-
`ate with the signal bandwidth, typi-
`y in the MHz range for radar. The
`eight vector(s) are derived from a sep-
`e beam controller. For open-loop
`ptive pattern control, the element sig-
`
`
`Page 2 of 6
`
`
`
`| || |5 a
`
`nals are also input to the beam controller
`that determines the proper weights via a
`suitable algorithm. This operation is per-
`formed at the much slowerrate ofthe
`external scenario change, typically on
`the order of a KHzorless.
`
`FUNDAMENTAL SYSTEM
`CHARACTERISTICS
`Dynamic range and signal band-
`width are fundamental characteristics
`of microwave systems. In most cases,
`A/D converters set these parameters
`for DBF arrays. For adaptive pattern
`nulling the beam controller character-
`istics are also important.
`Dynamic Range: The system's in-
`stantaneous dynamic range is deter-
`mined by the numberofbits Nj, of the
`A/D converters and the number of
`parallel channels N. For a Gaussian
`signal, the upper limit is set by Nj, and
`the lower limit by the quantization
`noise, leading to a dynamic range of
`approximately (6 N, + 10 log N) dB.?
`Thus the range increases by 6 dB per
`bit, and the factor N expresses the
`gain arising from the coherent integra-
`tion of N element signals. A/D con-
`verter nonlinearities may further re-
`strict the usable dynamic range.
`Signal Bandwidth: The signal
`bandwidth of the DBF system is de-
`termined by the A/D converter sam-
`pling rate and by the beamformerpro-
`cessing. To digitize IF signals with
`components up to f,,,,, the Nyquist
`criterion requires an ideal sampling
`rate of f, 2 2f,,,,, whereas the same
`signals at baseband with analog I and
`Q channels can be sampled with two
`A/D converters at a rate f, 2 f,,,,. In
`
`practice the signals must be slightly
`over sampled to provide a margin for
`realistic, finite slope filters.
`Despite modern computers’ impres-
`sive performance, the processing re-
`quired by digital beamforming may
`yet be challenging. With N elements,
`a beam requires N complex multiplies
`at the sampling rate (approximately
`B) that brings the total count to ap-
`proximately NB complex operations
`per sec (COPS), or about 10° COPS,
`with N = 100 and signal bandwidth B
`= | MHz, which poses nodifficulties.
`However, this hypothetical set does
`not representa full-scale radar. A full-
`scale radar may have several thousand
`elements and many independent
`beams, and this load is still challeng-
`ing. The processing cost becomes an
`implementation issue.
`As an alternative to forming custom
`beams with individual weight vectors,
`a fast Fourier Transform (FFT) can be
`utilized to calculate the entire set of N
`orthogonal beams simultaneously, as-
`suming the array elements are uni-
`formly spaced. This approach is ex-
`tremely efficient and uses only (N/2)
`logs (N)-N complex multiplies. How-
`ever, in application, the FFT and the
`custom beams are not necessarily
`equivalent. The custom beams have
`arbitrary patterns and directions, and
`the FFT beams have identical patterns
`and fixed angular spacing.
`The beam controller processing ca-
`pacity depends on the desired control
`algorithm. As a representative exam-
`ple, consider adaptive nulling via the
`sample matrix inversion (SMI)
`method. The formation of the sample
`matrix and its inversion take approxi-
`mately (7/6) N3 complex operations.
`With N equal to approximately 100
`elements and an update rate of 100
`times/sec, this formation leads to ap-
`proximately 5 x 105 floating point op-
`erations (FLOPs), which is currently
`reasonable. However, for the more
`ambitious adaptive space-time pro-
`cessing envisioned for airborne sur-
`veillance radar, the number ofspatial
`degrees of freedom is multiplied by
`the number of temporal degrees of
`freedom,a factor of three. In addition,
`the update rate needs to be increased.
`A factor of 10 is reasonable. This in-
`crease leads to processing loads on
`the order of 100 gigaFLOPs, which
`pose serious difficulties. Due to the
`N} dependence, the number of adap-
`tive degrees of freedom clearly must
`be kept to an absolute minimum.
`JOURNAL of ELECTRONIC DEFENSE « JULY"96
`
`Page 2 of 6
`
`

`

`OTHER IMPLEMENTATION
`ISSUES
`
`A multitude of system concepts and
`issues are of interest aside from the ba-
`sics discussed previously. For instance,
`as often done in radar, the dynamic
`range may be extended by coherent in-
`tegration in time. In effect, this exten-
`sion allows detection of signals smaller
`than the least significant bit because of
`the offset of the zero-mean noise volt-
`age. Based on this concept, an interest-
`ing radar has been proposed® with 4,000
`elements, 32-bit compression code and
`32 Doppler output cells, leading to 66-
`dB processing gain. With an operational
`dynamic range of 55 dB, the element
`signals are so far below the noise that a
`multibit A/D converter seems unneces-
`sary. A 1-bit A/D converter, which sim-
`ply outputs the plus/minus sign of the
`input signal, may be sufficient and
`would result in enormous A/D convert-
`er and IF channel savings.
`Other computational simplifications
`may be gained by going from element
`space to beam space via a spatial
`Fourier transform. Assuming a scenar-
`io where there are signals only in a rel-
`atively small number of M directions,
`then only a correspondingly small
`number of beams needs to be consid-
`ered, rather than the large number of N
`elemental signals. In this case, process-
`ing these beams by an SMI orsimilar
`adaptive algorithm requires on the
`order of M3 rather than N? mathemati-
`cal operations, which usually dramati-
`cally reduces complexity. Sophisticat-
`ed adaptive systems based on this con-
`cept have been proposed.?.!0
`
`DIGITAL BEAMFORMINGAT
`ROME LABORATORY
`
`Early Activities
`Digital beamforming has long been
`used in sonar systems and also for low-
`frequency over-the-horizon radars,
`such as the Air Force East Coast Radar
`System (OTH-B). However, the data
`rates are much lower than required by
`modern microwave radars, Serious in-
`terest in employing DBF for such sys-
`tems was first stimulated in the US
`about
`10 years ago by a joint
`DARPA/US Army MICOM study. Its
`objectives were to chart potential tech-
`nical options offered by DBF tech-
`niques and to quantify the benefits de-
`rived from them in selected applica-
`tions. Rome Laboratory consulted on
`this study and at the same time started
`
`its OWN program.
`
`Hardware Developments and
`Experiments
`An early US Air Force program ad-
`dressed array calibration, a problem
`common to all DBF arrays. For this re-
`search, a 32-element linear array operat-
`ing at C-band was built with high-per-
`formance commercial components.!!
`Triple conversion receivers provided
`analog I and Q signals that were digi-
`tized by a pair of 10-bit/0.5-MHz A/D
`converters. The array featured a novel
`self-calibration system that injected an
`accurate pilot tone immediately behind
`each antenna element and thus moni-
`tored the 32 receive channels. After digi-
`tal correction of channel imbalances
`(+2-dB amplitude, random phase) and
`1/Q and DC-offset errors, the RMS
`phase error was 2°, consistent with a -
`45-dB sidelobe level.
`Another, separate project was the
`technology demonstration of a fast digi-
`tal beamformer!? with performance
`compatible with an actual radar system.
`This successful design has several
`unique features. Four parallel processors
`accept 64 complexdigital input channels
`and form four independent inner product
`beamsat a 20-MHz clockrate. Since the
`bandwidth is excessive for many appli-
`cations, the design allows flexibility
`with respect to the number of beams
`formed. Thus, 4, 8, 16 or 32 beams can
`be multiplexed with a proportionally re-
`duced bandwidth. The high computa-
`tional capacity, 5 x 10? complex multi-
`plies per sec, is achieved with a systolic
`processor architecture based on the qua-
`dratic residue number system. This
`number theoretic approachis highly ef-
`fective for multiplications and additions
`and uses integer arithmetic so that
`round-off errors are avoided. The finite
`dynamic range is tailored to correspond
`to the limited range of the quantized
`input signals. Compared to a conven-
`tional approach,this processor has about
`40% less complexity (defined as the
`productof gate count and gate delays).
`These two systems coupled with a
`beam controller (a general-purpose
`computer) form the major components
`of the RL/Ipswich antennasite real-
`time digital beamforming testbed. This
`testbed was used to study adaptive pat-
`tern nulling and super resolution.
`Fast adaptive jammer suppressionis
`presently the strongest drive for DBF.
`Analog systems are usually based on
`feedback loops and their convergence
`is scenario-dependent — and often
`
` (
`
`ton Drive
`
`Monrovia, CA 91016
`Tel: (818) 358-9500, ext. 771
`Fax: (818) 358-9100
`
`JOURNAL of ELECTRONIC DEFENSE « JULY '96
`
`CIRCLE 15
`
`Page 3 of 6
`
`Page 3 of 6
`
`

`

`slow. Digital open-loop systems need
`no feedback and are scenario-indepen-
`dent and significantly faster. The SMI
`algorithm was used to demonstrate
`open-loop adaptive nulling.
`A jammerwasplacedat -10°, the in-
`terference-plus-noise covariance ma-
`trix was measured and the adapted
`weights were computed using a steer-
`ing vector for a desired-look direction
`at broadside with a 20-dB Taylor taper
`imposed. The array was then phase
`scanned and the received power was
`measured as a function of scan angle,
`as shown in Figure 3. This pattern rep-
`resents jammer poweronly, and ob-
`taining the adapted response in this
`way provides a direct measure of jam-
`mer cancellation. The main beam cor-
`responds to maximum received jam-
`mer power and the broadside null to
`
`jammer cancellation of 59 dB, whichis
`the full dynamicrange ofthe system.
`The resultant adapted pattern does
`not yield 20-dB sidelobes. which can
`be explained by an eigen-decomposi-
`tion of the covariance matrix.
`It ulti-
`mately depends upon the array errors
`— that is, the DC offset, I/Q, third har-
`monicand even the differences be-
`tween the noise figures of the element
`receivers. !3
`Super-resolution and direction-find-
`ing techniques aim to resolve sources
`closer than the Rayleigh limit. The key
`lies on the a priori assumption that the
`received arraysignal vectoris generated
`by a few point sources only, and then
`the corresponding spatial frequencies in
`this signal are determined with moder
`spectrum estimation methods. These
`methods involve nonlinear signal pro-
`
`cessing, and the algorithms tend to be
`highly complex, making a DBF array
`requisite for their practical implementa-
`tion. An experiment using the well
`known multiple signal classification
`(MUSIC) algorithm was performedin
`which two uncorrelated sources illumi-
`nated the array from directions that
`were only 1.7° (approximately 0.4
`beamwidth) apart. The received signals
`were analyzed with this algorithm. The
`response is plotted in Figure 4. The
`sharp peaks correspondexactly to the
`directions of the two signals. For com-
`parison, the response is shown for the
`sources when scanned by a convention-
`al beam and a monopulse beam, respec-
`tively. Clearly, only the super-resolu-
`tion algorithm is able to resolve such
`closely spaced sources, and a sum beam
`or a monopulse beam fails completely.
`
`And Elsewherein the World of
`Antenna Technology
`
`The theory associated withDBF’s ability to manipulate
`the appearance and response ofan antenna pattern seems
`almost mystical, but progress is being made in translating
`the theory into practice. Several DBF applications and
`hardware advances in several adjunctfields deserve spe-
`cial recognition.
`Andrew SciComm (Garland, TX) has developed vari-
`ous beamforming technologies for both military and
`commercial applications. An HF Digitally Adaptive
`Beamformingsubsystem is a case in point. This M an-
`tenna system isable to null up to M-1 undesired (jam-
`ming) signals and formabeam directed at a desired sig-
`nal location. The typicalsystem consists of four to as
`many as eightantennas. Current activity is aimed at im-
`proving thetequiredalgorithms and transitioning to a
`VXI form factor. VHFand UHF versions of the system
`are under consideration.
`Scientists at Roke Manor (Romsey, Hampshire, Eng-
`land), the research arm of Siemens Plessey, have been
`involved in advanced antenna beamformingsince the
`early 1970s. Digitally adaptive beamforming has been
`applied to the Multifunction Electronically Scanned
`Adaptive Radar (MESAR) developed in conjunction
`with the UK Defence Research Agency. The MESAR
`consists of some 1,000 S-band active phased-array ele-
`ments. While standard phase shifters provide electroni-
`cally controlled beam steering, each outputis then com-
`bined into oneof 16 subarrays for adaptive beamform-
`ing. Each subarray feeds its own receiver channel for
`conversion to digital I and Q basebands. The 16 1/Q
`channels are adaptively beamformed using complex
`weight multipliersto provide protection against up to 15
`sidelobeor mainbeamjammers.
`
`54
`
`Page 4 of 6
`
`Siemens Plessey has also entered into a strategic alli-
`ance with the Watkins-Johnson Co. (Palo Alto, CA) to
`interface advanced communications digital receiver tech-
`nology into Roke Manor adaptive array processing and
`DBFalgorithms. In this HF communications application,
`super-resolution DF processing provides improved read-
`ability in the direction of signal interest, while nulling
`other unwanted co-channel signals.
`The ability to generate numerous antenna beamsin ar-
`bitrary directions poses a challenge to measurement in-
`strumentation. Antenna patterns are generally measured
`with instrumentation receivers that measure the response
`of the antenna as a function of angle off boresight. Con-
`ventional antenna pattern receivers perform a few thou-
`sand measurements per sec. For an antenna system capa-
`ble of synthesizing hundreds of thousands of different pat-
`terns, however, the measurement time per pattern must be
`greatly reduced. To fill this need, the Aeroflex Lintek
`Corp. (Powell, OH) has produced the élan antenna-meas-
`urement system. The élan system is capable of collecting
`four million samples/see at frequencies from 100 MHz to
`100 GHz. This increased measurement speed not only re-
`duces test time but in some cases can Jead to new insights,
`such as a study of the time-varying response of an adap-
`tive DBF antennaas its pattern evolves.
`A portable kit of antennas intended for cellular tele-
`phone applications in the Nordic 450, AMPS and TACS
`bands has been introduced by Electro-Metrics (John-
`stown, NY). Consisting of directional Yagis and omnidi-
`rectional antennas covering the 430- to 470-MHz and
`§24- to 960-MHz range, along with required preampli-
`fiers, switch, filters and cables, the kit allows antennas to
`be assembled in the field withouttools.
`
`—Don Herskovitz
`
`JOURNAL of ELECTRONIC DEFENSE « JULY '96
`
`Page 4 of 6
`
`

`

`well with measurements. The feeding
`network was simulated by sequentially
`switching one digital receiver to the
`array elements, recording patterns and
`forming the composite array patterns off
`line. Figure 5 shows three 30-dB
`Chebyshev pattems taken at 4, 5 and 6
`GHz, demonstrating stable nulls and no
`beam squint over a 40% bandwidth.
`This study showed that circular ar-
`rays with DBF can generate high-qual-
`ity patterns, with narrow beams and
`deep pattem nulls, that are stable over
`large bandwidth. These features are at-
`tractive for many applications, includ-
`ing adaptive pattern nulling.
`
`Array ElementPattern
`Correction
`In a small array, all elements have
`different radiation patterns because of
`mutual coupling, which may preclude
`precise pattern control. However, with
`DBF these adverse effects can be cor-
`rected.!? Rome Laboratory's method is
`based on the following observation: in
`the receive mode, the individual antenna
`element signal has several constituents,
`one dominant constituent resulting from
`the direct incident plane wave and sever-
`al lesser constituents resulting from scat-
`tering of the incident wave at neighbor-
`ing elements. These constituents can be
`resolved and scattering compensated for
`by linear transformation, which is ac-
`complished by a matrix multiplication
`performed on the element output sig-
`nals. The compensation is scan-indepen-
`dent; the matrix is fixed and applies for
`all desired pattems and scan directions.
`The theoretical technique has been veri-
`fied in a demonstration with an eight-
`element array, where the initial sidelobe
`level was reduced from 20 dB to 30 dB,
`as shown in Figure 6.
`
`— DIFFERENCE— MUSIC — SUM
`BEAM
`BEAM
`
`AMPLITUDE
`
`Circular Array with Frequency-
`Invariant Pattern
`
`Circular array antennas have a
`unique capability. They can generate a
`main beam and sidelobes that are es-
`sentially independent of frequency be-
`cause the far field pattern can be repre-
`sented in terms of orthogonal phase
`modes (mode of unit amplitude/linear
`phase azimuth variation). So long as
`the relative amplitudes of these modes
`are constant, the entire pattern is con-
`stant with frequency. However, to gen-
`erate constant amplitudes, the required
`beamforming network is complicated.
`For arrays with more than 32 elements,
`it becomesfeasible only with DBF.'4
`The basic pattern synthesis tech-
`nique has been given previously.!5 On
`receive, the N element signals are
`transformed into phase modes by an N
`X N Fourier transformer, whichis fol-
`lowed by in-line filters that remove the
`frequency dependence of the individ-
`ual phase modes. Finally, the ampli-
`tude and phase taper corresponding to
`the desired beam shape and look direc-
`tion are imposed, and the phase modes
`are summed to produce a frequency-in-
`variant pattern.
`The beamwidth is determined by the
`number of phase modes used in the
`pattern. For given numbers of elements
`N and phase modes M,the usable pat-
`tern bandwidth is roughly given by M
`<2ka <N, where ka is the cylinder cir-
`cumference measured in wavelength.
`At the lower end, the pattern becomes
`super directive, and at the upper end
`the element spacing exceeds half a
`wavelength, leading to pattern pertur-
`bations. The filter responses for the
`phase modes have to be determined
`corresponding to the actual array ele-
`mentpattern.
`To demonstrate the features of fre-
`quency-invariant patterns, a circular
`array was built with 64 elements com-
`posed of monopoles in a parallel-plate
`region. The electrical design was based
`on a theoretical analysis!® that agreed
`
`
`
`POWER(dB)
`
`egeeee.
`
`80 -60 40 -20
`
`20 40 60 680
`
`0
`ac)
`Fig. 3 Adaptive pattern nulling with the SMI
`algorithm on the RL 32-element array.
`
`JOURNAL of ELECTRONIC DEFENSE * JULY ‘96
`
`Page 5 of 6
`
`30 -20
`
`-10
`
`(dB) 10
`
`O
`9(")
`Fig. 4 Super resolution with the MUSIC
`algorithm on the RL 32-elementarray.
`
`20
`
`30
`
`
`
`Page 5 of 6
`
`

`

`
`
`POWER(dB)
`
` Q
`
`Fig. 5 Measured 30-db Chebyshev
`patterns at 4, 5 and 6 GHz, overlaid.
`
`— BEFORE CORRECTION
`—AFTER CORRECTION
`—IDEAL PATTERN
`
`ital systems will realize more sophisti-
`cated control algorithms. With the
`spectacular advances in high-speed
`digital electronics, these systems may
`be here sooner thananticipated.
`
`This article was adapted from “Digital Beam-
`forming at Rome Laboratory,” Microwave
`Journal, February 1996, pp. 100-126.
`
`
`
`30—S
`
`(«GO
`
`tenna, RADC-TR-88-83, June 1988.
`12. L. Langston, S. Sanzgiri and K.
`Hinman, et al., Design Definition for a
`Digital Beamforming Processor, RADC-
`TR-88-86, June 1988.
`13. W. Humbert and H. Steyskal, “A Re-
`cent Evaluation of the Digital Beamform-
`ing Testbed at Rome Laboratory,” Rome
`Lab., TR-93-198, September 1993.
`14, H. Steyskal, “Circular Array with
`Frequency-invariant Pattern,” IEEE AP-
`S International Symposium, San Jose,
`CA, June 1989.
`15. D.E.N. Davies, “Circular Arrays,” in
`Rudge et al eds., The Handbook of An-
`tenna Design, Vol. 2, Peregrinus Ltd.,
`UK, 1983.
`16. B. Tomasic, “Circular Array of
`Coaxially-fed Monopole Elements in a
`Parallel Plate Waveguide Experiment,
`Rome Lab, TR-85-243, December 1985.
`17. H. Steyskal and J. Herd, “Mutual
`Coupling Compensation in Small Array
`Antennas,” IEEE Transactions, AP, De-
`cember 1990.
`
`Hans Steyskal studied electrical engineer-
`ing at the Royal Institute of Technology,
`Stockholm, Sweden, from which he received
`the degrees Civ. Ing. in 1963, Tekn. Lic. in
`1970 and Tekn. Dr.
`in 1973. In 1962, he
`joined the Swedish Defense Research Insti-
`tute (FOA), where he worked on microwave
`radiation and scattering problems. In 1980,
`he gave uphis position as chief section for
`field and circuit theory and moved to the US.
`He now pursues his interests in electromag-
`netics and applied mathematics at the Rome
`Laboratory, Hanscom Air Force Base, MA.
`Dr. Steyskalis a fellow of the IEEE.
`
`NOW YOU BE THE EDITOR
`Use our reader service card to tell us if
`this article was ofinterestto you.
`
`Yes: 154 No: 155
`
`
`
`
`r
`
`60
`
`-30
`
`O
`ANGLE (°)
`
`Fig. 6 Correction of mutual coupling
`effects in an eight-element array.
`
`CONCLUSION
`
`REFERENCES
`1.
`P. Barton, “Digital Beamforming for
`Radar,” IEEE Proceedings, Vol. 127, Pt.
`F, August 4, 1980.
`2.
`4H. Steyskal, “Digital Beamforming
`Antennas: An Introduction,” Microwave
`Journal, January 1987.
`3.
`H. Steyskal and J. Rose, “Digital
`Beamforming for Radar Systems,” Mi-
`crowave Journal, January 1989.
`4.
`A. Farina, Antenna Based Signal
`Processing for Radar Systems, Artech
`House, Norwood, MA, 1992.
`5. . Barile, R. Fante, T. Guella and J.
`Torres, “Performance of Space-time
`Adaptive Airborne Radar,” IEEE National
`Radar Conference, MA, USA, April 1993.
`6.
`R. Klemm, “Adaptive Air- and
`Space-borne MT! Under Jamming Con-
`ditions,” IEEE National Radar Confer-
`ence, MA, USA, April 1993.
`7.
`J. Proakis and D. Manolakis, Digi-
`tal Signal Processing: Principles, Algo-
`The technology exists to make rea-
`rithms, and Applications, Macmillan
`sonably sized DBF arrays feasible
`Publishing Co., New York, 1992, p. 419.
`today. For the near future, the high cost
`8
`A.C.C. Wong, “Radar Digital
`Beamforming,” Military Microwave Con-
`will necessitate strong economizing
`ference Proceedings, UK, 1982.
`with the number of digital channels.
`9, W. Gabriel, “Using Spectral Esti-
`For the more distant future, array an-
`mation Techniques in Adaptive Process-
`tenna systems that incorporate mixed
`ing Antenna Systems,” IEEE Transac-
`tions, AP-34, March 1986.
`analog and digital circuitry are envi-
`10. E. Brookner and J. Howell, “Adap-
`sioned. Reconfigurable, switched ana-
`tive-Adaptive Array Processing,” Pro-
`log subsystems will be used for their
`ceedings IEEE, Vol. 74, April 1986.
`tremendous bandwidth advantage. Dig-
`11.
`L. Eber, Digital Beam Steering An-
`
`
`
`
`
`
`
`
`
`
`
`
`
`The Triton name is newto the EW
`field. But nowherein the world wil
`you find people with more hands-on
`experience with microwave devices
`By combining the technical talents
`and facilities of the former ITT Electron
`Technology Division with visionary
`newleadership, Triton has created a
`resourceiul, responsive newteam.
`
`FOr QUICK, economical solutions—
`using TWTs, TWTAs, Klystrons and
`integrated sub-systems—
`talk to Triton.
`
`pyseteen Technology Division
`;
`
`
`A Visionary Approzch To Technical Solutions
`
`
`
` Solving EW Problems
`
`FOR MORE INFORMATION CIRCLE 38 ON READER SERVICE CARD
`
`JOURNAL of ELECTRONIC DEFENSE + JULY '96
`
`Page6 of 6
`
`
`
`Page 6 of 6
`
`