AD-A265 005

NSWCDD/MP-92/547

DEINTERLEAVER TECHNOLOGY FOR FUTUREELECTRONIC SUPPORT MEASURES (ESM) SYSTEMS

BY J. DARREN PARKER

SHIP DEFENSE SYSTEMS DEPARTMENT

L

DECEMBER 1992 V1 M2 31993

Apprcoved for public release; distribution is unlimited.

NAVAL SURFACE WARFARE CENTER

4 DAHLGREN DIVISIONDahlgren Uirainia 22448-5000

93-120479 3 5 27 10 4 IIII/IIII//II//IIIIIII!/II/II

SK 1 1

NSWCDD/MP-92/547

DEINTERLEAVER TECHNOLOGY FOR FUTUREELECTRONIC SUPPORT MEASURES (ESM) SYSTEMS

BY J. DARREN PARKER

SHIP DEFENSE SYSTEMS DEPARTMENT

DECEMBER 1992

Approved for public release; distribution is unlimited.

NAVAL SURFACE WARFARE CENTER

DAHLGREN DIVISIONDahigren, Virgin~a 22448-5000

NSWCDD/MP-92/547

FOREWORD

This report documents concepts and ongoing research related to the processingfunctions that future electronic support measures (ESM) systems will perform. Specifically,the focus of the report is on pre-filtering, sorting, and deinterleaving the pulse data generatedby the ESM receivers. The identification function is not discussed. The author acknowledgesthe significant contributions of Mike Garner (F23) and John Miniuk (F21) to this work.

This report was reviewed by Sam Stello, Electronic Warfare Systems IntegrationBranch, Tom W. Kimbrell, Head, Electronic Warfare Systems Integration Branch, andRichard Lee, Head, Electronic Warfare Systems Division.

Approved y:

T. C. Pendergraft, HeadShip Defense Systems Department

Acqoss1on For

SI)TII: 'J'?. Li

NTH -ý&T71;?

-, Av, .. t. I• O

' *'' Y C

i/u

NSWCDD/MP-92/547

ABSTRACT

The precision electronic support measures (ESM) systems of the future must operate indense environments of emitters that have agility in several parameters. A deinterleavingsystem that would be capable of processing the pulse data generated by a future precisionESM roceivor is described at the block diagram levcl. To be effective in a dense environmentof agile emitters, the deinterleaving system will need adaptable pulse data filters to eliminatethe data from friendly emitters, a processor dedicated to sorting the pulse data from new andhigh-priority emitters by angle-of-arrival and frequency, and sophisticated, multiparameterdeinterleaving ai-orithms. A sumimary of research related to ESM processing is given.

iii/iv

NSWCDD/MP-92/547

CONTENTS

Page

INTRODUCTION................... .................................. 1

PROPOSED DEINTERLEAVER STRUCTURE................................ I

SUMMARY OF RELA:TED RESEARCH....................................10SPECIFIC ESM PROCESSORS.....................................10DISTRIBUTED ARRAY PROCESSORS ............................... 12KNOWLEDGE-BASED ESM PROCESSING ............................ 13USING INTRAPULSE MEASUREMENTS AS A SORTING PARAMETER .... 14MISCELLANEOUS TOPICS.......................................14

CONCLUSIONS.....................................................14

REFERFNCES......................................................15

DISTRIBUTION.....................................................(1)

V

NSWCDD/MP-92/547

ILLUSTRATIONS

Figure Eage

1 DEINTERLEAVER FOR PRECISION ESM SYSTEM .................... 2

2 PROPOSED DEINTERLEAVING SYSTEM FOR A CHANNELIZED ESMRE C E IV E R ............................... .......... .... ..... 4

3 BLOCK DIAGRAM OF A PDW FILTER CHIP ......................... 5

4 PARAMETER WINDOW USING BOTH MAXIMUM AND MINIMUM ....... 7

5 PARAMETER WINDOW USING ONLY MEDIAN ...................... 8

6 AGILE RF DEINTERLEAVER STRUCTURE .......................... 9

vi

NSWCDD/MP.-92/547

INTRODUCTION

'Tile precision electronic support measures (ESM) systems of tile future must operate ina dense environment of agile emitters. Current emitters can stagger or jitter pulse repetitionintervals (PRIs). They can be agile in radio frequency (RF) on a pulse-to..pulse basis and agilein pulse amplitude (PA) and pulse width (PW) on a dwell-to-dwell basis. They call also usepulse compression techniques. Future emitters may extend their capabilities to include agilityin PA and PW on a pulse-to-punse basis. Consequently, the deinterleavers of the future must beable to reconstruct the pulse trains from emitters that can have random PRIs and can be agile inRF, PW, and PA on a pulse-to-pulse basis. This will require future deinterleavers to useintrapulse measurements and robust multiparamneter deinterleaving algorithms,

The ESM system's receivers generate 4- packet of digital data, which is called a 1 ilse.descriptor word (PDW), for every pulse detected. A PDW contains all of the measurements madeon a pulse, which includes azimuth (AZ), elevation (EL), RF, PW, PA, time-of-arrival (TOA),and possibly intrapulse modulations. In a dense environment, the ESM system's processors(deinterleavers, emittcr identification, emitter trackers, etc.) would be overloaded if every PDWwere processed. Therefore, the ESM system will need an adaptable PDW filter that can block thePDWs from friendly and low priority emitters (many of which will be agile in severalparameters), so that the PDWs from new emitters and threats can be processed quickly.

A block diagram for a deinterleaver that could provide the performance required by afuture ESM system is presented. Also, a summary of research related to ESM processing is given.

PROPOSED DEINTERLEAVER STRUCTURE

One approach to achieving the required ESM deinterleaver performance is to combine anadaptable PDW filter bank, an angle-of-arrival (AOA) clustering device, an RF histogrammingdevice, and robust multiparamneter deinterleaving algorithms. A block diagram of this type ofdeinterleaving system is shown in Figure 1. The inputs to the deinterleaving system are thePDWs generated by the ESM sensor. The outputs are the re-constructed pulse trains from everyemitter in the environment that are sent to the identification (ID) processor.

The first stage of the deirterleaving system (Figure I) is the PDW filter bank. Thepurpose of this bank is to block PDWs from friendly amd low-priority emitters, so PDWs fromhigh-priority emitters can be closely tracked and new emitters can be identified quickly. Each

NSWCDD/MP-92/547

Agile RIPDWs from PDWsESM Sensor 0 DW Filter Agile RF

Bank Deinterleaver Pulse: T'rains

Residue

I ,Pulse

Clustering and Stable RF i_'rains

Hisogramming -- - DeinterleaverDevices

~~I I " I . ..

Stable RPDWs I

FIGURE I DEINTERLEAVER FOR PRECISION ESM SYSTEM

fiter will have the capability to automatically allow a group of PDWs to pass periodically, sothat tracks on all emitters can be maintained. The period and the number of PDWs passed canbe controlled for each filter independently. When PI)Ws are passed, an emitter number tag canbe appended to the PDWs to eliminate redundant soiling. The filter bank can configure each ofits filters to block PDWs according to a combination of stable parameters. For instance,precision AOA alone could be used to block all of the PDWs from all of the emitters on a givenship. Similarly, a combination of AOA and RF could be used to block the PDWs from anemitter with a stable RF or one with agility in RF over a band (i.e., bandwidths tip to 500 MHz).Combinations of other parameters (including PW and PA) may also b'. used to filter PDWs fromagile ,niitters. When using AOA as a PDW filter parlamcter, both AZ and EL are considered.Therefore, a filter that is designated to block PDWs from a particular AZ and EL will allowPDWs from emitters at that AZ but with differ,•nt ELs to pass. Each filter's AOA window canbe adjusted as the emitter moves in angle by using information from the emitter's track file thatis maintained outside the deinterleaving system.

Tihe optimal implementation of the PDW filter will depend on the type of receivers usedin the ESM system. Most of the PDW filters that have been implemen.ted were designed for usewith instantaneous Frequency Measurement (IFM) receivers that typically have PDW output ratesslow enough for the PDW filter to process the PDWs one at a time. For a future ESM systemthat uses a channelized receiver, the PDW output rates will probably be so fast that the PDWfilter wvill not be able to process the data if it is input along a single path.

2

NS WCDD/MNl'-92/547

The receiver type that has traditionally been used in ESM systems is the 17M. A singleIFM can measure tile frequency of' signals over a wideband (i.e., 2 GIlz) by measuring thedifference between the input signal and a delaye(d version of the input signal. Its simplicity is itsmajor advantage, but it has great difficulty detecting simultaneous signals. The IFM receiver willtypicailly either ignore the weaker of the two simultaneous signlls or provide at corruptedfrequency measurement. The channelized receiver is much more complex and, consequently,expensive than the IFM receiver. It uses many narrow-band receivers that give the channelizedreceiver a high probability of intercept (POI) for simultaneous signals. The optimalimplementation of the PDW filter for a channelized ESM system would be Application SpecificIntegrated Circuit (ASIC) chips that are small and simple enough to be incorporated into eachchannel, so the inherent parallel structure of the chantelizer call be exploited. A single processorwould require that all of the PDWs generated in all channels be combined and sent down a singlepath to the procssor performing the PDW filter function. Therefore, if the PDW filter wereimplemented as a single processor, the througlhput of the system would be lower than the saimiesystem with PDW filters in each channel,

A simplified block diagram of a channelized ESM system that uses the proposeddeinterleaving system is given in Figure 2, which shows the receiver having N chatmels. The RF,PW, PA, and TOA ar-e measured in the receiver channels labeled "RECV CHAN" in the figure.

The AOA is measured in the phase interferometer channels labeled "AOA CHAN" in the figure.In Figure 2, the label "K" over the line leaving the phase interferometer antenna array is intendedto illustrate tmat tile outiputs of each of the K antenniae are inputs t)o each AOA channel. TheAOA is measured in each AOA channel from the phase differences between the antenna outputs.For this figure, AOA refers to AZ only. If the ESM system measures EL as wvell, another p)haseinterferometer must be included in the system. A POW assembler circuit to combine all themeasurements made on each pulse is labeled "PDW Assm" in the figure. A PDW filter chipfollows each PDW assembler, so the PDWs from friendly emitters can be removed immediately.This will require each PDW filter chip to have enough memory cells for the maximum numberof emitters to be filtered, but each must be simple enough to be placed in each channel. Twopulses arriving simultaneously within the same channel will cause a corrupted trequencymeasurement. Special frequency measurement techniques, which are beyond the scope of thisreport, will be required to handle this case.

A block diagram of a PDW filter chip is shown in Figure 3. At the top of the figure, thePDW from the PDW assembler is shifted into a buffer . ThFile parameters "hat are nceded forcomparison with the cells in memory are extracted and sent to each cll, Each cell is representedin this figure by a column of windows (a window being a range of acceptabie values for aparameter), each of which is denoted with the label "WIN," Cells arc formed for emitters thathave been identified by the ESM system as low priority emittes. Each cell will hIve windowsfor each parameter that can be used to characterize that pa;ticular emitter. 1lhe .:ontrol logicblock, shown at the bottom of the figure. p"-ormiis logic oporations On the results of, thecomparisons, so each cell can be configured to sc t on any combination of parameters. Note thatthe vertically aligned dots in the figure are intendedt to indicatc that each cell can be configuredwith windows fbr any number of parameters: however. only windows for AZ. EL, and "I' 'wreshown in Figure 3. The horizontally aligned dots in the figure indicate that the total number of

NSWCDDIMP-921/147

AQAHA1 PDW IPDWI

__ RECV AssmN1Fle

CHANN

AOA Cluster

RF Histogram

Stable RE Agile REPODWs PDWs

FIGURE 2. PROP'OSED IMEINTERLEAVING SYSTEM4 FOR A CIIANNELIZIEI ESM RECEIVER

NSWCDDiM, P-92/5,4"7

PIPW Input

- . AZEL __ RF

.* IL~NEL E.WIN1 I • -EWN N]

RIF WIN I o *0 * RFWINNW

•• ~Cont~rol-___

S~To Buffer

No MatchTo AOA ClusterCreate New Win

FIGU.I." 3. I1U.OCK DIAG(RI M O A A PI)\V IFiiTi R CIIIP

5

N SNVClDU/M 1P-92/5417

cells used tit tiny giveni time will (depend onl the total number olt emitterý thie systeml desires tofilter. For example, it cell thalt filters PDWs from an RIF stable emitter couild use AZ. EL. RE.mind PW and a coll - t filtcrs I'DWs from in RI: a niHe emitter coutld iise just AZ and EL orpe-rimps, AZ, 1-71 aiiu !,1W. It is noted that PW is not agood sorting parameter when uised alonebecause inultipath can cause corrupted 1PW nleasuremonts. However, PW may provide bettersor'ting Whenl 1us0( inl Con~julction with other paui~eters. If there tire eniltter-, in thle environmnentwith stable FRis, expected TOA could he tisedl along with thle other soring paramneters. Thecontsol logic block. at tie bottomi of Figure 3. determines if thle 11ptipt PIDW niatclies any of thecells if a mnatch occurs, (lhe P13W is sent to a bulifre -since it corresponds~l to it low~-priorityeinitter. If no match occuirs, (lie PDXV is sent dfirectly to thle AOA clustering dlevice. The emlitteridentifictition fuinction of tile ESM system wvill cuec the P13W filter to create a new cell if anemitter is identified as at low-priority emitter. Thie emitter tracking function of tile ESM4 systemi"will initiahizc. tiners th~at arc incluided in thie control logic block (Figure 3). A timier for each cellwill he set to allow PDWs to he passed to the emitter tracking I'mnction of thie ESM4 system ata sufficient rate to maintain an accurate track of the emitter.

The wvindows shown inl Figure 3 can he implemented in two ways. One implementation.which is illustrated in Figuire 4, rcquires storing both ta max1imum111 Mnd a inininuimi valtic to definethle acceptable region for thle pi-arameter.'iheut, thie measurfed paramecter miust be compared to boththle nmaximum and minimutm values and !ogic opertitions mutst be lperrorImed to determine if thleIlleasuredI value is less thant thle maxinium and greater than thc miinimumi. The other

the maximumin and minimutm. Thenl, thle ioen1st red Valuew is comiparedl to thle mcldianl vailue to therequired number of miost significant bits to obtain thie desired resolution.'

'l'hc second and third stages of the deinterleaving systom (Figuire 1) consist of on AQAehisteting device andI an RF histograniming device, res;pectively. Thle AOA clustering deviceclusters all PDWs by AOA. This is the lo~gical first step ill tw. piulse Sorting process since this

paranine.LM is thle only one that w~ill he stable onl a ptilse-to. puke basis l'or all emitters inl tileenvironment. A state-of-the-art F.SM sensor can mecasuire AZ and EL to accuracies onl thle orderof tenthis of a degree which will enable fihe AOA clusterine device to separate mlost p~latforms.Aftor thle PDW\.i. have been clustered by AOA, the RF histo-ounnuin:- device clusters the PDWsfrom -,table RF emitters. These stable RIF IPD\s are then sent to a stable RF deiinterlcavimugalgorithm). Thie PDWs from agile RE emitters will be scattered by the histogranmm lg device, anidthese 11"Ws will hle sent directly to anl ag1ile R F (leinterleavine, algorli inm.

The optimial implementation "or both the AOi\ clustering device and thle REhimolsogamminuig device is 1)1ol)xbl' at Content Addr11essable N'leinory (CAM4), which is a nmemlorythat simuitltanleouisly compares thle inp)ut to all mlemlory cells and ilenerates pointer,; to all miatching,cells. A CAM that is commnercially availiable is the Cohierent Proce.s-sor (CP)' wvhich has 41096cell of 32-bit CAM1. Each cell also ha',s at pr-OcSSifll" element thait COUld be useWd to ('ount1 thetitimtber of PD\Vs falling into each cell. Prog'rams tha1.t call Completely simlailte tile CIP arenvailable, so bench-marks cani be runl to dletermine if implementing" one's processing a1-lgoithmlison thle C11 w~ill he practical. The Associative Processor, which Wat' d1Cevehlod bV I 13M, l)L'rl0)InS

thie samle function a,, a CAM anMd Wats developed sp)ecifically fur ESMI puilse data~ soiun111',- This1

6

NSWCDD/MP-92/547

MeasuredParameter

10110

Minimum 10100Comparator

Not LessThan *

Maximum1011 1 Comparator

Not GreaterThan

AND

WITHINWINDOW

FIGURE 4. PARAMETER WINDOW USING B3TH MAXIMUM AND MINIMUM

7

NSWCDDM•P-92/547

MeasuredParameter

10110

Median 1 iComparator 101XX

Equal To

WITHINWINDOW

FIGURE 5. PARAMETER WINDOW USING ONLY MEDIAN

8

NSWCDD/MP-92/547

processor is discussed in greater detail in the next section (SUMMARY OF RELATEDRESEARCH).

The deinterleaving function of the system, as shown in Figure 1, is performed in twosepaiate algorithms operating in parallel; one deinterleaver operates on stable RF PDWs only,and the other operates on agile RF PDWs. This arrangement makes it possible to deinterleavethe PDWs from simple emitters quickly. The stable RF deinterleaver forms a two-dimensionalarray from the stable RF PDWs and uses all of the available parameters to deinterleave them.The array will contain a one-dimensional array of vectors; each vector will contain all themeasured parameters for a received pulse at a particular TOA. Intrapulse parameters will be usedonly if necessary due to their pulse train complexity. Any residue will be sent to the agile RFdeinterleaver. The agile RF deinterleaver forms a three-dimensional array from 'he agile RFPDWs and the residue from the stable RF deinterleaver. The array will contain a two-dimensional array of vectors; one vector for each discrete TOA and each RF, as illustrated inFigure 6. Like the stable RF deinterleaver, the agile RF deinterleaver will use all of the availableparameters to deinterleave the PrWs, and it will use the intrapulse parameters only if necessary.In Figure 6, the intrapulse measurements., which are shown as secondary sort parameters, arefrequency modulation (FM), phase mood ,tion (PM), and unintentional modulation (UM). Theresidue from the agile RF deinterleaver will be maintained in memory for several seconds, soemitters with very slow FRIs can be detected. The performance ef both deinterleaving algorithms(especially the agile RF deinterleaving algorithm) will depend on the ESM system's ability tomaintain accurate TOA tags over a span of several se, .*s.

TIMES_/] PRIMARY SORT

S- PARAMETERS

TOARFR

AMP -

-- SECONDARY SORT2 PARAMETERS3

AGILE RF DEINTERLEAVER INPUT PDW MATRIX

FIGURE 6. ACLE RF DEINTERLEAVER r- ,UCTURE

9

NSWCDDiMP-92/547

SUMMARY OF RELATED RESEARCH

A summary of research published in the open literature concerned with improving theperformance of ESM processors is given. This summnary is provided because much of thisresearch may be practical for future ESM deinterleaving systems.

SPECIFIC ESM PROCESSORS

Four ESM processors are described:

SADIE ESM Signal Processor---Kellet describes the ESM signal processor developed byRacal Defence called SADIE.3 This processor was designed based on the assumption thatmost of received pulses have been previously characterized, so extensive processingshould be reserved for complex emitters. The SADIE consists of a segregator, ananalyzer, a monitor, and a control processor; it has already been used in ship, air, andground ESM systems.

The segregator is the most important part of the system. It can use all of themeasured pulse. parameters; e.g., RF, PW, AOA, Modulation Flags, TOA, and PA. Thekey component of the segregator is the extended WAM (EWAM), which is implementedon ASIC chips. It stores mean and tolerance for comparisons. Many EWAMs operate inparallel on each parameter. Control logic determines the combinations of matches that areacceptable. The EWAM chip is based on 2.5-micron CMOS technology and performsseveral billion bit comparisons per second. The segregator can also use expected TOA inthe matching process. It allows for 0,1, or 2 missing pulses and accounts for scan burstsas well. Each EWAM can operate on its own after being initialized.

The analyzer is the deinterleaver and operates on pulses that do not match theparameters stored in the segregator. The monitor detects changes in the pulseenvironment. The control processor takes outputs from the analyzer and the monitor toupdate an emitter track file. It configures the segregator so that a scan analysis can beperformed.

IBM's Associative Comparator (AC) Chip---Hanna describes the AC chip developed byIBM Federal Systems Division.2 Each AC chip has 32 comparison cells; each cellconsists of a pair of upper and lower limits. It can perform 32 two-parametercomparisons in 1.6 microseconds, 16 two-parameter comparisons in 1 microsecond, and8 two-parameter comparisons in 0.5 microseconds where each parameter is 16 bits. Thisgives the AC chip a minimum throughput of 625,000 pulses per second. Hanna states thatthis throughput is adequate for IFM type ESM receivers, but channelized, micro-scan, oracoustic-optical type ESM receivers will require compare times less than 100

10

NSWCDD/MP-92/547

nanoseconds. Any cell can be independently enabled/disabled, or all cells can be globallyenabled/disabled. Multiple AC chips can be cascaded to compare more parameters and/orto have more cells.

IBM developed an ESM processor that uses AC chips to sort and filter incomingpulses. It uses three AC chips, which gives the processor a total of 96 comparison cells.Each comparison cell can either accept or reject matching pulses. The pulses that areaccepted are sent to the main memory of the ESM processor along with a number thatidentifies the matching cell. New cells are created for pulses that do not match any of thecurrent cells.

IBM made improvements to the ESM processor by improving the AC chip.' Thenext generation chip was called the Advanced Associative Comparator (AAC), and it wasincorporated or, a module called the Pulse Sort Module (PSM). The AAC chip cancompare an input to 128 cells 'a 2 microseconds. The input can be two 16-bit, one 16-bitand two 8-bit, or four 8-bit parameters. The PSM consists of a pulse processor and a cellprocessor. The pulse processor uses four AACs to compare incoming pulse data to 512cells in memory. When a match occurs, the data and an ID number are passed to the cellprocessor. If no match occurs, a new ID number is formed, and the data, the new IDnumber, and a new pulse flag are sent to the cell processor. When the cell processorreceives a new ID flag, it performs a 2D histogram in AOA and RF on the data. The cellprocessor maintains a 16 pulse history for each of the 512 compare cells. The cellprocessor has 8 analysis channels to perform PRI analysis and parameter averaging. Eachanalysis channel has enough memory for ':56 pulses.

Anaren's ESM Processor---Anaren developed an ESM system capable of operating insignal densities of one million pulses per second without any frequency, amplitude, orspacial filtering.6 I he receiver generates 55-bit PDWs that are sent to a Pulse Controller(PC). The PC forms a 2D histogram using bearing and frequency and has 4 Mbytes ofmemory. It takes less than one microsecond to map a pulse's bearing and frequency toa cell on the histogram. The Emitter Processor (EP) is linked to the PC through abidirectional bus. The EP searches the memory map of the PC looking for clusters andperforms PRI and scan analysis on the clusters it finds. If the EP obtains poor results, itretrieves more pulse data from the 1xemory map and repeats the analysis. The algorithmsused by the EP are written in Fortran 77, but the author states that plans for futureimprovements include converting the algorithms to Ada. This ESM processor wassimulated during development using the techniques developed by Hollands."

AN/SLQ-32(v) Inner Processor (IP)---The IP of the AN/SLQ-32(v) performs the initialdeinterleaving functions of sorting the incoming PDWs into pulse trains.8 This documentis the only one referenced that is not published in the open literature. It uses AOA(bearing only) and RF as sorting parameters. When a PDW is received, the IP tries tomatch its AOA and RF with cells stored in memory. If it finds a match, the IP incrementsthe counter tor that cell. When the count reaches a threshold, the IP informs the CPU ofa new emitter so the PDWs corresponding to this cell can be analyzed for identification

11

NSWCDD/MP-92/547

of the emitter. However, before reporting this new cell as a new emitter, the IP firstchecks the adjacent AOA bins. If the same RF bin is found in an adjacent AOA bin, theIP assumes it is the same emitter and removes the older cell from memory. The IPattempts this same correlation using the next two adjacent bins if the amplitude of thepulse is below a threshold or if the AOA bin is at the bow or stern of the ship.When theIP receives a PDW that does not match any cells in memory, it forms a new cell. Noreceived PDW is reported to the CPU until the count threshold for that cell is reached.Also, a timer is used to remove cells for which no PDWs have been received in a givenamount of time.

DISTRIBUTED ARRAY PROCESSORS

Parallel processors have been applied to a variety of fields, and they will be required orfuture ESM systems that will operate in very dense environments. Transputers are a SingleInstruction Multiple Data (SIMD) type of parallel processor architecture, and they seem to beparticularly well suited to ESM processing. The research of Roberts, Merrifield, and Beton inapplying transputers to ESM processing is discussed.

Roberts discusses the use of the Mil-DAP as a processor for an airborne radar anddescribes the next-generation processor called the MSI DAP.9 Both consist of inany l-bitProcessing Elements (PEs) with each PE connected to its 4 neighbors. MSI DAP has 4096 PEsin 64 by 64 lattice with each PE containing 16 Kbytes of memory. Mil-DAP has an array of 32by 32 PEs with each having 8 or 16 Kbytes of memory. Each PE has a 1-bit adder and 3registers. Any word length can be used because all operations are built up bit wise. The Mil-DAPwas used as the processor on an airborne radar with several operating modes. The measures ofperformance were the data input time, the times to compute FFTs and CFARs, and time toresolve ambiguities in range and doppler. It was determined that the Mil-DAP processed data fastenough to have 20- to 25-percent dead time between data inputs.

Merrifield investigated the use of the Mil-DAP for ESM applications."0 He reports thatESM-related benchmarks demonstrate that the Mil-DAP can perform real-time processing. TheMil-DAP can be programmed either at the bit level with an assembler language or by using anextended version of Fortran developed for this processor. The Mil-DAP was programmed to sortESM pulse data into chains and to compare the chains. The processing functions were dividedinto association and recognition functions. The association process forms pulse chains andmaintains a database of received pulses. The recognition process performs the ID and maintainsa database of active emitters. The parameters used by the association function include RF, PW,AOA, and TOA. The recognition function uses all of these to compute PRI; the researchers planto add scan analysis at a later time. The rate of change of the number of emitters in theenvironment affects the pe:formance because reformatting of the databases contained in each PEis time consuming. Merrifield states that special-purpose hardware is needed to perform

12

NSWCDD/MP-92/547

associations of simple emitters before the input to the Mil-DAP associator. He also states thatperforming PRI analysis using histograms of TOA differences is slow on this type of SIMDmachine.

Beton discusses the perfonnance achieved by ESM processing algorithms implementedon a array of transputers." The ESM processing algorithms were originally written in PASCALon a Vax computer and then, were translated to OCCAM, which is a language designed forimplementing parallelized algorithms. The ESM processor functions were broken into three parts:a deinterleaver, a merger, and an emitter identification. Beton did not implement a pulse filterprior to the deinterleaver, but stated that the deinterleaver would have a maximum input rate ofto, pulses per second after front-end filtering. The deinterleaver function required the highestthroughput and was implemented on 40 processors. Beton stated that the transputerimplementation of the deinterleaver performed 350 times faster than the Vax version. He did notdetermine if the transputer implemented ESM processor could achieve real-time performance.

KNOWLEDGE-BASED ESM PROCESSING

A area of research that seems to promise major improvements over current ESMprocessors is the application of knowledge-based (also called expert systems) algorithms todeinterleaving. The work of Cussons, Feltman, and Self is discussed.

Cussons is researching knowledge-based processing for deinterleaving, merging, and ID.' 2

Admiralty Research Establishment (ARE) contributed to the SEAVIEW knowledge-based emitterID processor. Its most important characteristic is its ability to reverse a hypothesis that is provenfalse. The information released by the reversed hypothesis is available for the creation of newhypotheses. Cussons plans to use this same knowledge-based architecture for deinterleaving andpulse chain merging. Also, he plans to develop real-time implementations of all three processors.Initial research indicates that a transputer implementation is most likeiy one for real-timeapplications.

Feltman is researching transputer implementations of knowledge-based deinterleaving andmerging. 3 lHe states that conventional, algorithmic type deinterleavers can already beimplemented on transputers for real-time applications, but knowledge-based deinterleaversimplemented on transputers are a little slow for real-time. However, he expects the next-generation transputers to be fast enough for real-time applications of knowledge-baseddeinterleavers. Feltman predicts that the processors of ESM systems developed in the near futurewill consist of a prefilter implemented in hardware, an algorithmic deinterleaver implemented ona Multiple Instruction Multiple Data (MIMD)-type parallel processor, and a knowledge-basedmerger and ID implemented on an MIMD-type parallel processor. Then, in the future, replacingthe algorithmic deinterleaver with a knowledge-based deinterleaver will be practical.

The work of Self has resulted in an ESM processor implemented in ADA.'4 Self iscontinuing research to incorporate expert system techniques, to develop a multiprocessorimplementation, and to develop an ASIC implementation of the preprocessor.

13

NSWCDDIMP-92/547

USING INTRAPULSE MEASUREMENTS AS A SORTING PARAMETER

Another area of research that holds promise for ESM processor improvements is the useof intrapulse measurement data for deinterleaving pulse trains. Danielsen has investigated adeinterleaver that operates on a pulse-by-pulse basis and ielies exclusively on intrapulsemeasurements 1 He reports that using this technique alone does not provide adequatedeinterleaver performance. Agg continued this research by adding TOA analysis after theintrapulse measurement sorter."6 Agg reported a significant improvement in performance, but thisdeinterleaver could not handle the case of two identical emitters along the same line of bearing.He speculates that a deinterleaver using intrapulse measurements as a sorting parameter will beeasier to implement in real-time than a knowledge-based approach.

MISCELLANEOUS TOPICS

A general discussion of microprocessor applications to electronic warfare (EW) andmethods to evaluate the performance of a particular processor are given by Lemley.17 Hollandsdiscussed a software package he developed in Fortran 77 that can simulate and evaluate theperformance of an ESM processor.' It models the interfaces between hardware and softwarecomponents of the processor to evaluate the real-time performance.

CONCLUSIONS

Future ESM systems will need highly sensitive receivers in order to maintain a highprobability of detection for threat emitters. This will make these ESM systems vulnerable toprocessor overload due to the large number of nonthreatening emitters in the environment. Ablock diagram of an deinterleaver that would allow a sensitive ESM system to operate in a denseenvironment was presented. Also, a suirrnary of recent research that has potential applicationsto ESM data processing was given.

14

NSWCDD/MP-92/547

REFERENCES

1. Coherent Research, Inc., Sales Literature for Coherent Processor Products. I Adler Drive,East Syracuse, NY 13057, 315-433-1010.

2. Craig A. Hanna, "The Associative Comparator: Adds New Capabilities to ESM SignalProcessing," Defense Electronics, February 1984, pp. 51-62.

3. A.G. Kellet, "SADIE - A High Performance ESM Signal Analysis Processor," lEEColloqutim on Signal Processing for ESM Systems, April 1988, London UK, pp. 7/1 - 7/3.

4. Craig A. Hanna, Joseph A. Cirucci, and David R. Prunty, "VLSI Application to ESMProcessing," IBM Technical Directions, Vol. 13 No. 1 1987, IBM Federal Systems Division,Owego, NY, PP 32-36.

5. Stephen C. Smith and Michael P. Beakes, "The Pulse Sort Module: VLSI Technology inthe ESM System," IBM Technical Directions, Vol. 13 No. 1 1987, IBM Federal SystemsDivision, Owego, NY. pp. 37-42.

6. R.S. Andrews, "ESM Processing Using 3-Dimensional Memory Mapping and AdaptivePattern Formation Algorithms," Conference Proceedings - Military Microwaves '84, Oct 1984,London England, pp. 27-36.

7. P. Hollands, M.A., "Us:e of Simulation Methods as a Design Tool in the Development ofan ESM Processing System," !"-E Procs, Vol. 132, Pt F, No. 4, July 1985, pp. 292-297.

8. AN/SLQ-32(v) DTU/DFC/Presorter Inner Processor Functional Description.

9. J.B.G. Roberts, P. Simpson, B.C. Merrifield, and J.F. Cross, "Signal ProcessingApplications of a Distributed Array Processor," lEE Proceedings, Vol. 131, Pt F, No. 6, October1984, pp. 603-609.

10. B.C. Merrifield, J.B.G. Roberts, P. Simpson, A. Stanley, "Real Time Applications ofDAP," ICS 88, International Conference on Supercomputing Proceedings, Supercomputing '88,pp. 54-62, May 1988, Boston MA.

11. R.D. Beton, S.P. Turner, C. Upstill, "Hybrid Architecture Paradigms in a Radar ESM DataP'rocessing Application," Microprocessors and Microsystems, Vol 13 No. 3, April 1989, pp. 160-164.

12. S. Cussons, J. Roe, and A. Felthamn,"Knowledge Based Signal Processing for Radar ESMSystems," ESM Division, Admiralty Research Establishment, Portsdown, Cosham., Hants.

15

NSWCDD/MP-92/547

13. AM Feltman, Admiralty Research Establishment, Portsdown, Cosham, UK, "Real-TimeESM Signal Processing with Transputers," Conference Title: lEE Colloquium on 'The Transputerand Signal Processing,' Digest No. 1990/042, p.5/1-3, Conference Date: 05 Mar 1990, ConferenceLoc: London, UK.

14. A.G. Self, MEL Defense Systems Ltd., Stittsville, Ontario, Canada, "A New Ada-basedESM Processor," Journal of Electronic Defense, Feb. 1991, Vol.14, No.2, p.49-51,55.

15. P.L. Danielsen, D.A. Agg, N.R. Burke, Smith Associates Limited, Guildford, UK; "TheApllication of Pattern Recognition Techniques tc ESM Data Processing," Conference Title: IEEColloquium on 'Signal Processing for ESM Systems' Digest No. 1988/62 p.6/1-4, ConferenceDate: 26 Apr 1988, Conference Loc: London, UK.

16. D.A. Agg, Smith Associates Ltd., Guildford, UK, "The Development of an ESM DataProcessing Scheme Based on Pattern Recognition Techniques," Conference Title: IEE Colloquiumon 'Electronic Warfare Systems' Digest No. 1991/009, p.5/1-4, Conxrence Date: 14 Jan 1991,Conference Loc: London, UK.

17. L.W. Lemley, "Computers/Processors (for Electronic Warfare)," NRL Report 8247, August15, 1978, Naval Research Laboratory, Washington, D.C. 20375-5000.

16

NSWCDD/MP-92/547

DISTRIBUTION

DEFENSE TECHNICAL INFORMATION ATITN CODE 06D21CENTER 12 CODE 06D23CAMERON STATION CODE 06D24ALEXANDRIA, VA 22304-6145 PMS 400B

COMMANDERATrN TECHNICAL LIBRARY I NAVAL SEA SYSTEMS COMMAND

ONT-0215 CDR WALLACE 2531 NATIONAL CITYDIRECTOR BLDG 3OFFICE OF NAVAL TECHNOLOGY WASHINGTON DC 20362-5160800 N QUINCY STARLINGTON, VA 22217-5000 ATTN MICHAEL CALA

COMMANDERATTN DIRECTOR OF RESEARCH I CENTER FOR NAVAL ANALYSESUS NAVAL ACADEMY PO BOX 16268ANNAPOLIS, MD 21402 ALEXANDRIA, VA 22302

ATTN TECHNICAL LIBRARY I ATrN JEWC/DVPJOHN HOPKINS UNIVERSITY COMMANDERAPPLIED PHYSICS LABORATORY JOINT ELECTRONIC WARFARE CENTER11000 JOHNS HOPKINS RD SAN ANTONIO, TX 78243-5000LAURLL, MD 20707

ATI'N CODE 5023ATTN TECHNICAL LIBRARY I COMMANDERSUPERINTENDENT NAVAL AIR DEVELOPMENT CENTERNAVAL POST GRADUATE SCHOOL JOIINSVILLEMONTEREY, CA 93940 WARMINSTER, PA 18974

ATTN CODE 804 1 ATTiq JOHN MONTGOMERYCOMMANDING OFFICER COMMANDERNAVAL WEAPONS SUPPORT CENTER NAVAL RESEARCH LABORATORYCRANE, IN 47522-5030 TACTICAL EW DIVISION

WASHINGTON DC 20375-5000ATrN CODE 7500 1COMMANDING OFFICER ATTN TECHNICAL LIBRARYNAVAl. OCEAN SYSTEMS CENTER COMMANDERSAN DIEGO, CA 02152 NAVAL AIR WEAPON CENTER

CHINA LAKE. CA 93555-6001

(I)

NSWCDD/MP-92/547

ATTN TECHNICAL LIBRARY I INTERNAL DISTRIBUTIONCOMMANDERNAVAL SECURITY GROUP COMMAND A it3801 NEBRASKA AVE NW B 4WASHINGTON DC 20393 B 32/Bhlair

CATTN TECHNICAL LIBRARY DCOMMANDER D25NAVAL AIR WARFARE CENTER D4WEAPONS DIVISION EPOINT MUGU, CA 93042-5000 E231 10

E232 2ATTN SIONT/EW I E28 ICOMMANDING GENERAL FMARINE CORPS RESEARCH F0! IDEVELOPMENT AND ACQUISITION F02COMMAND F07/DORSEY IQUANTICO, VA 22134-5080 FOX I

FI0 4ATTN TECHNICAL LIBRARY 1 -P20 5DIRECTOR F21 20DEFENSE ADVANCED RESEARCH F22 IPROJECTS AGENCY N23 I3701 N FAIRFAX DR 1021, IARLINGTON, VA 22203-1714 F30 3

!F40 2ATrN W44 1741 2DIRECTOR F42 2NATIONAL SECURITY AGENCY 1743 19800 SAVAGE RD F50 2FORT' MEADE, MD 20755-6(X) G I

G52ATTN TECHNICAL SUPPORT GROUP J IDIRECTOR K INAVAL TECHNICAL INTELIAGENCE NCOMMAND N05/GASTON I4301 SUITLAND RD N20/I!ENDIERSON IWASHINGTON DC 20390 N22 I

N24 IATI'N GIFT AND EXCHANGE DIVISION i N25 ILIBRARY OF CONGRESS N70 2WASHINGTON DC 20548 N74/GIDEP I

R IATTN HUGH VANLANDINGHAM IVIRGINIA POLYTECHNIC INSTITUTE ANDSTATE UNIVERSITYDEPARTMENT OF EILECTRICALENGINEERINGBLACKSBIJRG, VIRGINIA 24061-011!

(2)

REPORT DOCUMENTATION PAGE FormlTAPrpoved

I'ubI1C 101)311It U VII bu fl~ lt hij CU1Lktttnn Of intoirn at at' is ntioaf to avroq I ho j. ur P1. r lclnc "''Cliu t,'; tha e ",c lt~sc.' *flluktso ica~h- dI* a.r t a %~13otu'm~

callt ci."thltnifortta too n. ondludtrt( sutqv Wart for wedwiuar that burdien, to WAtt' nttOn H eaaluqatIt~ea,(I srie. oilcelo'aaC Itt, tnfoto-aataa, Oialmno Soo Report taa. 121 111 e0l'.oDyavs H w1b ay.Suit 1204.Ata~loaV 22202 41302. anti to th,1 Offieo No) -,moa-t aanf i C1a t.h , iJ tp 5to or;,tjt otrojtfl t (070.1 10I0), VV-h~m) %oro DC.0%

1, AGENCY USE ONLY (Leave blank) 2.REPORT DATE 3. REPORT TYPE AND DATES COVERED

I 1)eceniber 1992

4. TITLE AND SUBTITLE S. FU)NDING NUMBERS

De interleaver Technology for FutureElectronic Support. Moaskires tESM) Systemns

6. AUTHOR(S)J. Darren Parker

7, PERFORMING ORGANIZATION NAME(S) AND ADDRESS(ES) 0. PERFORMING ORGANIZATIONREPORT NUMBER

Ntival Sur-face Wiati-are Centel- LNWClDI)/MP-92o,147Dahigren Division WF21)Dahigren, VA 22448-5000

9. SPONSORINGIMONITORING AGENCY NAME(S) AND 10. SPONSORINGorMONITORING

it. SUPPLEMENTARY NOTES

12a. OISTRIBUTIONIAVAILABILflY 12b, DISTRIBUTION CODE

Approved foor public release; distribution is unlitnited,

13. ABSTRACT N-a'ximu'n) 200 %vords)

The precisioll electronic support ineasures (E.SM) systems of tile futuire mlust Operate inl (elene enlv' ,o: nentts Ofemitters that have agility in several paramieters. A deinterleavingr systemn that would be capable otf procers-ingthe pulse data generated by a future precision EISNM receiver is do-cribed at t)-- blocl diag, %il level. TfO beeffoctive in a dense environnment of aigile emnitters, thle deinterleaving systell ~vill need tadaptiabie puilse datafilters to eliminate the data fromi friendly einitters, A processor dedicated to sorting the pokse data fromn inevand high-priority emritters by angie-of-arrival and frequency, and sophikticated. tnultipnu-amleter deiitter-leaving algoritluna. A. summ11ary of' research reae to ESIM poa;suisgivenl.

14. SUBJECT TERMS 15. NUMBER Of PAGES

F~loetroili Warfar~e IFWV) 2 7D~e interlea ving R.&It d Stit lI P~rocessing 16 ll 711(.7 -C(),)

17. SECURITY CLASSIFICATION 18. SE(.' lilY CLASSIFICA1ION V). SECURITY CLASSIFICATION 20. LIMITATION Of AIISTRACTOF REPORT OF THIS PAGE Of ABSTPACT

LINCI.ASSIIIE1I) U NC \b5'lIPI ' UNCLAS3il"l FI ED

NSN 75a10,01-280-5500 Standard F of i 298 (Rev 2 89)*~~~,*.,, ~ 1 *0, wA~ ta. ~ iR

GENERAL INSIRUCTIONS FOR COMPLETINq 5F

The Report Documentation Page (11)0) is used it announcing and cataloging reports. it is important thatthis information be consistent with the rest of the r epor t, particularly the cover and its title page.instructions for filling io each block of the form follow. It is important to stay withir~ the lines to meetoptical scanning requkrements.

Block 1. Auencv Use OLvjIy.eavebkiank). Block 12a. Distribuition/Availability Statemienit.Denotes public availability or limitations. Cite any

Block 2. agprj eFllpbitindticuin availability to the public. Enter additional imitationsday, month, and year, if available (e.g. 1 )an 88). Must or special markings in all capitals (e.g. NOFORN, REL,c~te at least the year. ITAR).

Block 3. ly of Report and-Dates Covered. Statewhether report is interim, final, etc. If applicable, enter DOD See DoDD 5230.24, "Distributioninclusive report dates (e.g. 10 Jun 87 -30 Jun 88). State ments on Technical Documents."

DOE - See authorities.block 4. Tilendbtteitle tlkstknfo h NASA -See Handbook NHI3 2200.2part of the report that provides the most meaningful NTIS -Leave blankand complete information. When a report is preparedin more than one volumne, repeat the primary title, addvolumne number, and include subtitle for the specific Block 12b. Distribution) Code.volume. On classified documents enter the titleclassification in) parentheses.

DOD - Leave blank.Block S. Funding Numbers. To include contract and DOE - Entter DOE distribution categories fromgrant numbers; may include program element the Standard Distribution fornurnbei (s), project number(s). task number(s), anid Unclassified Scientific and Technicalwo rk u nit n u nb er(s). 11e s 4 %- iwn I Ne!sRC-)Or45

NASA -Leave blank.C - Contract PR Project NTIS -Leave blink.G Grant TA TaskPE - Prorgram WU Work Unit Block 13. Abstract. Iniclude abi-ief (Ma)ximium )200

Element Accession No. words) factual summary of the most significantinformation contained in the report.

BLOCK 6. AjdioJ~.( Name(s) of p~erson(s) responsiblefor writing the report, performing the research, or Block 14. Sub'ect Trmis. Keywords or phrasescredited with the content of the report. If editor or identifying major sub'jects in the report.compiler, this should follow the name(s).

Block 15. Number of Pages. Enter the total numberBc 7. Performinq OrcianizationNameo(s) and of pages.

s -qps(L) Self-explanatory.Block 16, Price Code. Enter appropriate price code

Block S. Performing Organization Repo(rt Nutmber. (iMSl only)Enter the unique alphanrnieiic report numiber(s)assigned by the organization perfornni:.-. the report. Block 17.-19. SecrityCIlissificationio, Self.

explanatory. Enter U.S. Securtity Classification in)Block 9, ý.p~nqj)ApjýýMonior I Aqency iMenw-) and accordance with U.S. Security Regulations (i~e..AddressL"e4 Self-explanatory. UNCLASSIFIED). if form contains classified

information, stamip classification on) the top andBlock 10. Spqnso ringlMonitoring gnqnr bottom of this page.Number. (if Known)

Block 126. Limitation of Abstract. This block must beBlock 11. S vlemeintary Notes. Enter information not completed to assign a limitation to the aistract.indclued elsewhere such as: Prepared in cooperation Enter either UL (unlimnitecd or SAR (same as report).with...; Trans. of.. ; ro be published in... . When a An entry in this block is nccessa~ry if the abstract is toreport is revised, include a statement whether the new be limited. If blank, the abSti a-t N aSSuimed to bereport SUpersedles or supplements the older repor t. Unhi-noted~

Sta ulatd Frmrr 298 Back (Rev 2-89)