Molcular and Quantum Acoustics vol. 6, (005) 175 RADAR SIGNAL FEATURE EXTRACTION SYSTEM BASED ON A SAW DISPERSIVE DELAY LINES Adam KAWALEC 1, Andrzj PIENIĘŻNY 1, Th Institut of Radar Tchnology, Faculty of Elctronics, Military Univrsity of Tchnology, S. Kaliski St., 00-908 Warsaw, Poland 1 Adam.Kawalc@wat.du.pl, Apinizny@wl.wat.du.pl 1. INTRODUCTION Spctral analysis of signals using frquncy-to-tim transformrs (transducrs) and tim-to-frquncy digital FFT procdurs is thortically rcognizd in gnral and ffctivly usd in practic. Such analysis rfrrd to CW harmonic signals or pulsd signals with known and rlativly long tim of apparanc (larg puls width) is widly usd bcaus of its thortically unlimitd rsolution and du to many practical bnfits. Today howvr, application of ths tchniqus and mthods to th analysis of CW signals with mor sophisticatd wavforms or pulsd signals with short and unknown tim of apparanc causs significant difficultis in sufficintly accurat and fast procssing of such signals transmittd by modrn radars using ECCM tchniqus. Hnc, thr is a nd of analysis of such signals for EW applications (thrat valuation, countrmasurs oprations against nmy radars including ffctiv jamming and ARM dlivry). Th thrat valuation bcoms mor and mor difficult du to incrasing complxity of radar wavforms. Som wavforms ar dvlopd intntionally to mak thir intrcpt almost impossibl. Th distinctiv faturs of modrn radar signal ar hiddn in its tim-frquncy structur. In th nar past th problm of radar signal fatur xtraction was considrd in tim or frquncy domain sparatly, bcaus radar wavforms wr rlativly simpl. Today, howvr, th signals should b obsrvd simultanously in both domains. Modrn radar signals hav complx tmporal and spctral structur with short tim duration and larg frquncy bandwidth. It implis dirctly an urgnt nd to dvlop an xtrmly fast mthod/tchniqu providing rproduction of signal tim-frquncy structur without loss of its rprsntation fidlity du to ralizd transform oprations. Particular xampl of practically ffctiv combining (link) of analogu and digital signal procssing tchniqus from minimum procssing tim and sufficint procssing accuracy points of viw is prsntd. Th papr prsnts th masurmnt suit that utilizs surfac acoustic wav [1,,4,6,1,19] (SAW) dvics to raliz quasi-ral-tim
176 Kawalc A., Piniężny A. analogu signal procssing and FFT procdurs prcdd by th puls digitizr to prform nar-ral-tim DSP. Som xprimntal masurmnt and analysis rsults ar prsntd.. FOURIER PROCESSOR ON SURFACE ACOUSTIC WAVE DEVICES Th procss of signal filtration is prformd by th surfac acoustic wav filtr mad in tchnology of intrdigital transducrs (Fig.1). Output signal of such a filtr is convolution rsult of input signal and filtr puls rspons i.. S wy ( = S ( h( (1) whr dnots convolution opration and h( is a filtr puls rspons. From th convolution thorm rsults that th quation 1 in frquncy domain b can xprssd in th form S w w ( h( S ( ω) H ( ω) () whr convolution in tim domain is quivalnt to spctra multiplication of signal and filtr puls rspons in frquncy domain. Puls rspons of a filtr in tim domain is quivalnt to its frquncy charactristic in a frquncy domain, i..: In a cas whn puls rspons is givn by w h( H ( ω) (3) h( than th output signal of a filtr is givn by S wy jµ t = (4) jµ ( t τ ) jµ t jµτ jµ tτ = Sw( τ ) dτ = Sw( τ ) dτ ( ω) (5) whr µ=b/t is th frquncy-to-tim convrsion factor. Similarity btwn last transform and Fourir transform of signal r(, i.. jω t = r( t dt can b proofd in th following way. Lt us assum ω=µτ than: R( ω ) ) (6) jµ tτ = r( dt R( µτ ) (7) Thus, to rciv th last quality on hav to multiply in quation, output signal s wy ( and input signal s w ( rspctivly Finally R( µτ ) = S wy ( τ ) jµτ, S w ( = r( jµ t (8) R S r t dt ( ) ( ) j µτ ( ) j µ t τ µτ = wy τ = (9)
Molcular and Quantum Acoustics vol. 6, (005) 177 Th last quation rprsnts th Fourir procssor that consists of thr lmnts oprating on surfac acoustic wav concpt. s w ( t ) s ( t ) wy h ( t ) r( s t w ( ) s wy ( τ ) R( µτ ) M1 jµ t T, B c c M T, B T, B jµ t jµ t Fig. 1. SAW filtr structur Fig.. Fourir procssor arrangmnt It has lctrical implmntation, prsntd in Fig.. Th input signal r( is multiplid by a rfrnc signal. Thn th rsultant product is convolvd in puls comprssion filtr. Finally, a multiplication by a scond rfrnc signal is prformd. Last opration rmovs th phas distortion introducd by rfrnc signal. In spctral analysis application, th scond multiplication is usually omittd sinc th dsird information is includd in th nvlop of th Fourir transform. Such an approach is calld as a comprssiv rcivr concpt, usd for intrcption and instantanous analysis of radar signals [3,13,0,1]. 3. MEASUREMENT MODEL Th masurmnt modl includs comprssiv rcivr (CR) modl and procssing unit (PU). Th first on provids instantanous Fourir transform and th procssing unit (PU) prforms procssing of th transformd signal. Th CR oprats on two idntical DDL with following paramtrs: cntr frquncy f o =70MHz, bandwidth B=0MHz, tim dlay T=10µs. Bcaus th DDL ar idntical, th frquncy bandwidth limiting of th rfrnc signal is ncssary to provid an optimum analysis of an input signal [14-18]. Frquncy valus f h and f o1 satisfy th following rlationships: f h =f s1o, f o1 =f s1o +f o, whr f s1o =30MHz, f o =70MHz, f s1 =f s1o ±B 1 /. Th PU is digitising CR output signals and procsss thm to rproduc instantanous spctrum structur of a signal. Svral xprimnts has bn mad for continuous (CW) and pulsd narrowband signals and for widband linar frquncy modulatd (LFM) signals. Som rsults of masurmnts ar prsntd in Fig.6,7. Disprsiv dlay lins usd in modl hav filtr structur with cntr frquncy f o =70MHz [5,7-11]. Filtr has puls rspons duration tim T=10µs, which amplitud charactristic is flat in B=0MHz bandwidth. Disprsiv priodical transducr oprats in two frquncy bandwidths situatd symmtrically around frquncy of lctrods structur that is f oo =87,75 MHz. It has bn shown on Fig.4. Thus th lctrods structurs priod is λ=18µm. Additional bandwidth is not usd and it is attnuatd by xciting transducr, which spctrum is distributd around
178 Kawalc A., Piniężny A. cntral frquncy of a filtr. Disprsiv filtr is mad on quartz substrat having (X,Y) orintation. Exciting transducr consist of 11 lctrods having width 11µm, whras disprsiv on consists of 191 lctrods having 9µm width, what rsults from assumd frquncy of lctrods structur. Insrtion losss (65dB) of disprsiv filtr ar considrably high and rsult from low lctromchanical coupling factor of a quartz substrat. INPUT SIGNAL f s = f s1 n( f o1 f o COMPRESSION FILTER RF AMPLIFIER AND VIDEO DETECTOR f h T 1 TIME GATE f o CHIRP GENERATOR TIMING PROCESSING UNIT Fig. 3. Comprssiv rcivr modl 0.00-60 -0.00-40.00-60.00 Insrtion losss [db] -80-100 -10 Thory Exprimnt -80.00-140 50 60 70 80 90 100 Frquncy [MHz] Fig. 4. Amplitud charactristic of a disprsiv filtr Fig. 5. Puls rspons of disprsiv filtr (tim duration 10µs) Calculatd frquncy charactristic of a filtr diffrs a littl bit only from thos rcivd by practic. It has bn shown of Fig.4. Th filtr rspons has bn prsntd on Fig.5. Rippls rsult from th rquirmnt of flat charactristic of disprsiv filtr. 4. SIGNAL PROCESSING As rsultd from analysis output of a CR, is followd by many short pulss with positions dtrmind by th input signal frquncy. Thr xist also sid lobs associatd with ths pulss. Thus to masur th frquncy of th input signal th masuring of th main puls cntr with simultanous nglcting of sid lobs is rquird. To dtrmin th
Molcular and Quantum Acoustics vol. 6, (005) 179 frquncy paramtrs of th input signal, th tim paramtrs masurmnts should hav to b prformd. Th simplst approach to solv that problm is to compar th CR output signal with fixd thrsholds. Whn th procssd puls braks ths thrsholds, it will b dclard a lgibl output. This approach includs two main shortcomings. Firstly, th amplitud of th output puls changs with th input signal lvl. This rsults with puls bas widning. Scondly th fixd thrsholds dtction schm dosn't distinguish btwn main and sid lobs in gnral. All this disadvantags caus th digitising tchniqu to b th most suitabl to th CR output signal procssing. Fig. 6. CR output signals, pulsd unmodulatd signal (puls width 5µs) Fig. 7. CR output signals, pulsd LFM signal (puls width 10µs) Fig. 8. CR output signals, pulsd LFM signal (puls width 10µs) Fig. 9. CR output signals, pulsd LFM signal (puls width 10µs), rcord numbr dnots tim, sampl numbr dnots frquncy To procd xaminations th masurmnt systm practical modl has bn usd (Fig.3). Output intrmdiat frquncy signal is applid to that systm. It transforms xamind signal into vido pulss rprsnting instantanous spctrum of signal. Th systm prformanc is controlld by synchronization pulss. Usful information in th viw of signal sampls is stord on digitizr and than passd to procssing unit that maks final stag of procssing. In gnral it groups signals in mannr showing modulation within th puls. Som
180 Kawalc A., Piniężny A. rsults of th xaminations ar prsntd on Fig. 6-9. Ths figurs rprsnting instantanous spctrum of pulsd radar signals unmodulatd and linarly frquncy modulatd ar showing th spctral fatur of radar signal. Numbr of pulss undr analysis dpnds on digitizr thrshold. As it is sn from xprimnt, in th cas of puls train it is possibl to stimat frquncy dviation within th puls (on th bas of puls train), and what is vry important th stimation procss is prformd in th nar-ral tim. It is not possibl to dtrmin modulation typ howvr systm can distinguish ovrlapping signals what is its uniqu advantag. 5. CONCLUSIONS Th masurmnt systm analyzd in th papr is a vry fficint tool for th timfrquncy signal analysis. Th most important part of a systm is a comprssiv rcivr mad on surfac acoustic wav tchnology. Its advantag is high spd prformanc, that rcommnds to apply it in radar signal nvironmnt, for masuring thir frquncy paramtrs. This rcommndation is spcially dirctd to th analysis of th short-tim narrowband and widband linar frquncy modulatd signals. Apart from its high spd prformanc th CR absolutly rquirs sophisticatd hardwar and softwar for output signal procssing to provid its practical usfulnss. Bcaus th CR output signal rprsnts th instantanous spctrum of th input signal only, th puls tim of arrival masurmnts ar strongly rquird. Th masurmnts should b compltd for ach duty cycl of th CR to valuat ral tim-frquncy structur of a signal. It should b notd that procssd output signals ar xtrmly short (narrow) and thir tim of apparanc at th output has stochastic natur implid by random tmporal rlations btwn th rfrnc and input signals. This will caus losss at CR output signal. Finally, as far as th CR concpt is discussd, on can admit that it may b implmntd to th analysis of instantanous frquncy structur of th signal. Th CR offrs possibility of high-prcision intra-puls and intr-puls analysis particularly usful for idntification of pulsd linar modulatd signals. Th main conclusion is that th analyzd systm is high spd Fourir transformr, that aidd by task orintd softwar ralizing functions of th virtual masurmnt instrumnt provids ffctiv possibility to xtract frquncy paramtrs of radar signals. Th systm most significant fatur is ability to analys signals that ar simultanously prsnt at th input. This rcommnds it to us in dns signal nvironmnt typical for ELINT/ESM systms.
Molcular and Quantum Acoustics vol. 6, (005) 181 REFERENCES 1. Atzni C., SAW signal transform tchniqus; Wav Elctronics, vol., no.1/, (1976), pp. 38-65.. Atzni C., Mans G.F., Masotti L., Ral-tim ntwork analyzr using dual analogu chirp transform; Elctronics Lttrs, vol.1, no.10, (13th May 1976), pp. 48-49. 3. Brur K.D., Lvy J.S., Paczkowski H.C., Th comprssiv rcivr, A vrsatil tool for EW systms; Microwav Journal, vol.10, (Octobr 1989), pp. 81-98. 4. Campbll C.K., Application of surfac acoustic wav and shallow bulk acoustic wav dvics; Procdings of th IEEE, vol.77, no.10, (Octobr 1989), pp. 1453-1484. 5. Danicki E., Filipiak J., Kawalc A.,: SAW disprsiv dlay lin utilising apodisd IDT with priodic lctrods, Elctron. Ltt., 19 (1986) 976-977 6. Jack M.A., Grant P.M., Collins J.H., Th thory, dsign, and applications of surfac acoustic wav Fourir-transform procssors; Procdings of th IEEE, vol. 68, no.4, (April 1980), pp. 450-468. 7. Kawalc A., Procsor Fourira z podzspołami na AFP, Elktronika 3 (1994) 13-16. 8. Kawalc A., Synthsis and xprimntal rsults of SAW disprsiv filtrs for spctrum analysrs, J. of Tchn. Phys 37 (1996) 39-49. 9. Kawalc A., Low BT product SAW wid-band disprsiv dlay lins for radar applications. MIKON 96, (1996) 45-49 10. Kawalc A., Progrss in P-disprsiv filtrs for radar application, Biul. WAT 6 (1999) 15-131. 11. Kawalc A, :Wybran zagadninia analizy i syntzy szrokopasmowych linii dysprsyjnych z akustyczna falą powirzchniową. Wojskowa Akadmia Tchniczna. Warszawa 000r 1. Kočmasov V.N., Dolbnja E.V., Sobol N.V., Akustolktronny fur'-procssory; Radio i Svjaz', Moskva, 1987. 13. Moul G.L., SAW comprssiv rcivr for radar intrcpt; IEE Procdings, pt. F, Communications, Radar and Signal Procssing, vol. 19, no. 3, (198), pp. 180-186. 14. Paradowski L., About som vrsion of SAW chirp spctrum analyzr - analysis and prformanc; Procdings of th 5th Confrnc on ACOUSTOELECTRONICS'91, 10-13 Sptmbr, 1991, Varna, Bulgaria, World Scintific Publishing Co. Pt. Ltd., Singapor, 1991, pp. 13-131. 15. Paradowski L., Piniężny A., An xprimntal study of som comprssiv spctrum analyzr; Procdings of th XV-th National Confrnc CIRCUITS THEORY and ELECTRONIC CIRCUITS, Szczyrk, Octobr 0-3,199, Warsaw, Poland, 199 pp. 448-453. 16. Paradowski L., Piniężny A., Fornalik J.: Tim-frquncy fin analysis using th comprssiv rcivr concpt, Procdings of Intrnational Symposium on Surfac Wavs in Solid and Layrd Structurs and National Confrnc on Acoustolctronics, Moscow-St.Ptrsburg, Russia, 17-3 May 1994, IEEE Ultrasonics, Frrolctrics, and frquncy Control Socity, St Ptrsburg 1995 s. 06-14. 17. Paradowski L., Piniężny A., Instantanous spctrum analysis using tim intrval
18 Kawalc A., Piniężny A. masurmnts, Procdings of th XIII IMEKO XIII WORLD CONGRESS "From masurmnt to innovation", Torino, Italy, Sptmbr 5-9, 1994 Vol.1, s.456-461. 18. Paradowski L., Piniężny A., Radar signal procssing in tim microscal using frquncyto-tim and tim-to-frquncy procdurs and transducrs, Th Procdings of Th 7th Intrnational Confrnc on Signal Procssing Applications & Tchnology, Boston, Massachusstt, USA, Octobr 7-10, 1996, Millr Frman Inc., 1996, vol., pp.1500-1504. 19. Sobol N.V., Dolbnja, E.V., Kočmasov, V.N., Analogovy fur'-procssory i ich primnni; Zarubžnaja radiolktronika, nr 4, (1983), pp. 47-69. 0. Tsui J.B.Y., Microwav rcivrs with lctronic warfar applications; John Wily and Sons, Nw York, 1986, pp. 78-38. 1. Tsui J.B.Y., Rsarch in lctronic warfar (EW) rcivrs; Procdings of th Fourth Confrnc on MODERN RADAR PROBLEMS, Zakopan, Octobr 19-, 1993, Warsaw, Vol.II, pp.358-367.