UDK 621.3:(53+54+621 +66)(05)(497.1 )=00 ISSN 0352-9045 Strokovno društvo za mikroelektroniko elektronske sestavne dele in materiale 22009 Strokovna revija za mikroelektroniko, elektronske sestavne dele in materiale Journal of Microelectronics, Electronic Components and Materials INFORMACIJE MIDEM, LETNIK 39, ŠT. 2(130), LJUBLJANA, junij 2009 UDK 621.3:(53+54+621 +66)(05)(497.1 )=00 ISSN 0352-9045 INFORMACIJE MIDEM 2 o 2009 INFORMACIJE MIDEM LETNIK 39, ŠT. 2(130), LJUBLJANA, JUNIJ 2009 INFORMACIJE MIDEM VOLUME 39, NO. 2(130), LJUBLJANA, JUNE 2009 Revija izhaja trimesečno (marec, junij, september, december). Izdaja strokovno društvo za mikroelektroniko, elektronske sestavne dele in materiale - MIDEM. Published quarterly (march, june, september, december) by Society for Microelectronics, Electronic Components and Materials -"föiOEM. Glavni in odgovorni urednik Editor in Chief Dr. Iztok Šorli, univ. dipl.inž.fiz., MIKROIKS, d.o.o., Ljubljana Tehnični urednik Executive Editor Dr. Iztok Šorli, univ. dipl.inž.fiz., MIKROIKS, d.o.o., Ljubljana Uredniški odbor Editorial Board Dr. Barbara Malič, univ. dipl.inž. kem., Institut "Jožef Stefan", Ljubljana Prof. dr. Slavko Amon, univ. dipl.inž. el., Fakulteta za elektrotehniko, Ljubljana Prof. dr. Marko Topič, univ. dipl.inž. el., Fakulteta za elektrotehniko, Ljubljana Prof. dr. Rudi Babič, univ. dipl.inž. el., Fakulteta za elektrotehniko, računalništvo in informatiko Maribor Dr. Marko Hrovat, univ. dipl.inž. kern., Institut "Jožef Stefan", Ljubljana Dr. Wolfgang Pribyl, Austria Mikro Systeme Intl. AG, Unterpremstaetten Časopisni svet International Advisory Board Uredništvo Informacije MIDEM MIDEM pri MIKROIKS Stegne 11, 1521 Ljubljana, Slovenija tel.: +386(0)1 51 33 768 faks: + 386 (0)1 51 33 771 e-pošta: Iztok.Sorli@guest.arnes.si http: //www. m id e m-d rustvo. si / Prof. dr. Janez Trontelj, univ. dipl.inž. el., Fakulteta za elektrotehniko, Ljubljana, PREDSEDNIK- PRESIDENT Prof. dr. Cor Claeys, IMEC, Leuven Dr. Jean-Marie Haussonne, EIC-LUSAC, Octeville Darko Belavič, univ. dipl.inž. el., Institut "Jožef Stefan", Ljubljana Prof. dr. Zvonko Fazarinc, univ. dipl.inž., CIS, Stanford University, Stanford Prof. dr. Giorgio Pignatel, University of Padova Prof. dr. Stane Pejovnik, univ. dipl.inž., Fakulteta za kemijo in kemijsko tehnologijo, Ljubljana Dr. Giovanni Soncini, University of Trento, Trento ^ Prof. dr. Anton Zalar, univ. dipl.inž.met., Institut Jožef Stefan, Ljubljana Dr. PeterWeissglas, Swedish Institute of Microelectronics, Stockholm Prof. dr. Leszek J. Golonka, Technical University Wroclaw .Naslov uredništva Headquarters Letna naročnina je 100 EUR, cena posamezne številke pa 25 EUR. Člani in sponzorji MIDEM prejemajo Informacije MIDEM brezplačno. Annual subscription rate is EUR 100, separate issue is EUR 25. MIDEM members and Society sponsors receive Informacije MIDEM for free. Znanstveni svet za tehnične vede je podal pozitivno mnenje o reviji kot znanstveno-strokovni reviji za mikroelektroniko, elektronske sestavne dele in materiale. Izdajo revije sofinancirajo ARRS in sponzorji društva. Scientific Council for Technical Sciences of Slovene Research Agency has recognized Informacije MIDEM as scientific Journal for microelectronics, electronic components and materials. Publishing of the Journal is financed by Slovene Research Agency and by Society sponsors. Znanstveno-strokovne prispevke objavljene v Informacijah MIDEM zajemamo v podatkovne baze C0BISS In INSPEC. Prispevke iz revije zajema ISI® v naslednje svoje produkte: Sci Search®, Research Alert® in Materials Science Citation Index™ Scientific and professional papers published in Informacije MIDEM are assessed into C0BISS and INSPEC databases. The Journal is indexed by ISI® for Sci Search®, Research Alert® and Material Science Citation Index™ Po mnenju Ministrstva za informiranje št.23/300-92 šteje glasilo Informacije MIDEM med proizvode informativnega značaja. Grafična priprava in tisk BIRO M, Ljubljana Printed by Naklada 1000 izvodov Circulation 1000 issues Poštnina plačana pri pošti 1102 Ljubljana Slovenia Taxe Perçue UDK621.3:(53+54+621 +66), ISSN0352-9045 Informacije MIDEM 39(2009)2, Ljubljana ZNANSTVENO STROKOVNI PRISPEVKI PROFESSIONAL SCIENTIFIC PAPERS M.Bizjak: Vpliv sten ob kontaktih z oblokom na izklopno zmogljivost nizkonapetostnega odklopnika 65 M.Bizjak: The Influence of Lateral Walls Near Arcing Contacts on Breaking Capacity of Low-voltage Circuit-breaker M.Fras, J.Mohorko: Simulacija komunikacijskih sistemov v realnem času z realno komunikacijsko opremo v simulacijski zanki 71 M.Fras, J.Mohorko: Real Time Communication Systems' Simulation With Real Communication Devices in Simulation Loop S.Klampfer, J.Mohorko, Ž.Čučej: Mehanizmi paketnega uvrščanja kot osnovni pogoj za zagotavljanje kvalitete storitev v omrežju 78 S.Klampfer, J.Mohorko, Ž.Čučej: IP Packet Queuing Disciplines as Basic Part of QoS Assurance within the Network S.Penič, U.Aljančič, D.Resnik, D.Vrtačnik, M.Možek, S.Amon: Metoda za določanje koeficienta c/31 tankih piezoelektričnih filmov 85 S.Penič, U.Aljančič, D.Resnik, D.Vrtačnik, M.Možek, S.Amon: Cantilever Method for Determination of D Coefficient in Thin Piezoelectric Films M.Šalamon, T.Dogša: Model detektorja kaotičnosti 93 M.Šalamon, T.Dogša: The Model of Chaoticness Detector H.Abdul-Majid, Y.Yusoff, R.Musa, T.Kong, M. Sulaiman: Nizkocenovno vezje na čipu za odčitavanje pH vrednosti 100 H.Abdul-Majid, Y.Yusoff, R.Musa, T.Kong, M. Sulaiman: A Low-Cost Single-Chip Readout Circuit for pH Sensing J.Tušek, A.Šarlah, A.Poredoš, D.Fefer: Optimiranje magnetnega polja v magnetnem hladilniku 105 J.Tušek, A.Šarlah, A.Poredoš, D.Fefer: Optimization of the Magnetic Field in a Magnetic Refrigerator Z.Živkovič ,M.Hribšekand D.Tošič: Modeliranje SAW kemičnih senzorjev hlapov 111 Z.Živkovič .M.Hribšekand D.Tošič: Modeling of Surface Acoustic Wave Chemical Vapor Sensors Andrej Kosi, Mitja Šolar: Delitev in izbiranje DVB-ASI signala v redundantnlh DVB-T/H oddajnih sistemih 118 Andrej Kosi, Mitja Solar: DVB-ASI Distribution and Selection in DVB-T/H Redundancy Systems MIDEM prijavnica 123 MIDEM Registration Form Slika na naslovnici je skupek fotografij in slik iz posameznih prispevkov v tej številki. Front page is constructed of photos and pictures taken from contributions published in this issue. VSEBINA CONTENT Obnovitev članstva v strokovnem društvu MIDEM in iz tega izhajajoče ugodnosti in obveznosti Spoštovani, V svojem več desetletij dolgem obstoju in delovanju smo si prizadevali narediti društvo privlačno in koristno vsem članom,Z delovanjem društva ste se srečali tudi vi in se odločili, da se v društvo včlanite. Življenske poti, zaposlitev in strokovno zanimanje pa se z leti spreminjajo, najrazličnejši dogodki, izzivi in odločitve so vas morda usmerili v povsem druga področja in vaš interes za delovanje ali članstvo v društvu se je z leti močno spremenil, morda izginil. Morda pa vas aktivnosti društva kljub temu še vedno zanimajo, če ne drugače, kot spomin na prijetne čase, ki smo jih skupaj preživeli. Spremenili so se tudi naslovi in način komuniciranja. Ker je seznam članstva postal dolg, očitno pa je, da mnogi nekdanji člani nimajo več interesa za sodelovanje v društvu, seje Izvršilni odbor društva odločil, da stanje članstva uredi in vas zato prosi, da izpolnite in nam pošljete obrazec priložen na koncu revije. Naj vas ponovno spomnimo na ugodnosti, ki izhajajo iz vašega članstva. Kot član strokovnega društva prejemate revijo »Informacije MIDEM«, povabljeni ste na strokovne konference, kjer lahko predstavite svoje raziskovalne in razvojne dosežke ali srečate stare znance in nove, povabljene predavatelje s področja, ki vas zanima. O svojih dosežkih in problemih lahko poročate v strokovni reviji, ki ima ugleden IMPACT faktor.S svojimi predlogi lahko usmerjate delovanje društva. Vaša obveza je plačilo članarine 25 EUR na leto. Članarino lahko plačate na transakcijski račun društva pri A-banki : 051008010631192. Pri nakazilu ne pozabite navesti svojega imena! Upamo, da vas delovanje društva še vedno zanima in da boste članstvo obnovili. Žal pa bomo morali dosedanje člane, ki članstva ne boste obnovili do konca leta 2009, brisati iz seznama članstva. Prijavnice pošljite na naslov: MIDEM pri MIKROIKS Stegne 11 1521 Ljubljana Ljubljana, junij 2009 Izvršilni odbor društva UDK621,3:(53+54+621 +66), ISSN0352-9045 Informacije MIDEM 39(2009)2, Ljubljana VPLIV STEN OB KONTAKTIH Z OBLOKOM NA IZKLOPNO ZMOGLJIVOST NIZKONAPETOSTNEGA ODKLOPNIKA Martin Bizjak Iskra MIS, Kranj, Slovenija Kjučne besede: nizkonapetostni odklopnik, napetost obloka, tokovna omejitev, kratki oblok, delitev obloka, deion komora, prepustni tok, joulski integral izklopa (integral l2t), vpliv plastične mase na izklopno zmogljivost. Izvleček: Izklopna zmogljivost nizkonapetostnega odklopnika je odvisna od razvoja električnega obloka in njegovih značilnosti. Odklopnik naj pri izklopu kratkostičnega toka zagotovi čim manjšo amplitudo prepuščenega toka in čim manjši joulski integral (integral l2t). To je omogoča učinek tokovne omejitve kratkostičnega toka. Tokovno omejitev določa razvoj obločne napetosti, ki naj dovolj hitro naraste preko neke kritične vrednosti. Hitrost naraščanja obločne napetosti in njena maksimalna vrednost sta odvisni od hitrosti odpiranja kontaktov, ki prekinjajo varovani tokokrog, formiranja zanke obloka med kontakti in delitve obloka na ustrezno število delnih oblokov. Računska simulacija poteka toka in napetosti obloka med kontakti z ad hoc računskim modelom razkrije, katere značilne količine obloka odločujoče vplivajo na tokovno omejitev pri izklopu. S simulacijo ponazorimo časovni potek obločne napetosti in prepuščenega toka, kot ga registriramo eksperimentalno na oscilogramu pri preizkusu izklopne zmogljivosti. Nekatere vrednosti, ki jih potrebujemo za računsko simulacijo najdemo v literaturi, druge pa vzamemo kot variabilne parametre. Iz primerjave simulacijskih in eksperimentalnih rezultatov sklepamo na vplive, ki so posledica hitrosti odpiranja kontaktov in značilnosti v odklopniku vgrajenega materiala. Tako je bil kvalitativno ugotovljen vpliv plastične mase iz obloge kontaktno-obločnega prostora (s primerjavo dveh vrst materiala) na izklopno zmogljivost obravnavanega tipa odklopnika. The influence of lateral walls near arcing contacts on breaking capacity of low-voltage circuit-breaker Key words: low-voltage circuit breaker, arc voltage, current limiting, short arc, arc splitting, arc chute, let-through current, joule integral (l2t integral), influence of plastic material on breaking capacity. Abstract: The breaking capacity of low-voltage circuit breaker is closely related to the development of arc voltage and its magnitude during breaking of short-circuits current. The magnitude of let-through current shall be reduced and the joule integral minimized as well in orderto achieve significant current-limiting effect of circuit breaker. Current limiting requires the magnitude of arc voltage to increase in the sufficiently short time interval to the relatively stationary value above the certain critical voltage. Both of them are dependent on the movement of contact members during opening and on the elongation speed of arc column in the initial phase as well as on the subdividing of arc in the arc chute. Numerical simulation of the variation of arc voltage and let-through current during the breaking time reveals the influencing quantities of arc crucially affecting the current limiting action. The numerical procedure simulates the current and voltage waveforms normally recorded in oscillograms of breaking test shots. Some values necessary for the calculation were found in literature, others are used for simulation as variable parameters. The mutual combination of experimental and simulation results led us to the conclusion which physical effects, either movement of contacts or characteristics of materials (metallic and non-metallic) are dominant at break. By this way a significant influence of plastic material forming lateral walls around arcing contacts on the breaking capacity of circuit breaker under discussion was indicated qualitatively by comparing effects of two types of plastics. 1. Uvod Pri nastanku nenormalnih stanj v nizkonapetostnem energetskem tokokrogu, kot je pojav prenapetosti ali tokovne preobremenitve v kratkem stiku, je treba tako stanje čim hitreje odpraviti ali tokokrog zaščititi pred njim. Prevelik tok skušamo hitro prekiniti z odklopnikom, kar pa v nizkonapetostnem energetskem tokokrogu traja vsaj toliko časa, da trenutna vrednost toka doseže prvo ničelno vrednost. To preide izmenični tok na »naravni« način v vsaki polperiodi izmeničnega toka in odklopnik bi lahko uspešno opravil izklop toka, če bi po prehodu toka skozi »naravno ničlo« obdržal tok na ničelni vrednosti še naprej. V tokokrogih z veliko razpoložljivo močjo (in z velikim razpoložljivim tokom /p v kratkem stiku) pa bi izklop trajal že predolgo in bi bil vpliv preobremenitve z joulskim integralom (integral l2t) in z konično vrednostjo Id prepuščenega toka že kritičen. Zato odklopnik pri operaciji izklopa omeji tok po amplitudi (ki je bistveno manjša od amplitude izmeničnega toka) in po času do prve tokovne ničle (bistveno krajši od trajanja polperi-ode izmeničnega toka), po kateri mora biti tok prekinjen, to je: ostati na vrednosti nič. Izklop je opravljen z učinkom »tokovne omejitve« /1/. Časovni potek prepustnega toka od trenutka, ko se kontakti odklopnika začnejo odpirati, je odvisen predvsem od naraščanja napetosti med kontakti. Režo med razmikajoči-ma se kontaktoma kontaktnega para takoj premosti električni oblok. Kot aktivno komponento tokokroga ga izkoristimo za omejitev prepustnega toka izklopa /2/. Delovanje odklopnika v nenormalnih stanjih v tokokrogu, npr. pri kratkem stiku, ne moremo poljubno preskušati »in situ«, ampak v simuliranih razmerah, kar lahko opravljamo v specializiranih preskuševališčih. Tak laboratorij razpolaga z viri izmeničnega enofaznega ali trifaznega toka velike kratkostične zmogljivosti z možnostjo nastavitve razpoložljivega toka na nekaj 10 kA ali celo več kot 100 kA in z nastavitvijo preskusne napetosti do 1000 V. Preskusni 65 Informacije MIDEM 39(2009)2, str. 65-70 M. Bizjak: Vpliv sten ob kontaktih z oblokom na izklopno zmogljivost nizkonapetostnega odklopnika tokokrog, v katerem preskušamo delovanje odklopnika, vsebuje ohmske in induktivne komponente, s katerimi simuliramo razmere v realnem tokokrogu v kratkem stiku. Pogoje preskusa glede na namen odklopnika podajajo ustrezni elektrotehniški standardi /3/, /4/. S posebnim stikalom lahko vklopimo preskusni tok pri vnaprej izbranem faznem kotu sinusne napetosti vira, s čimer simuliramo nastanek kratkega stika pri poljubnem tokovnem faznem kotu. 2. Potek izklopa s tokovno omejitvijo Sposobnost odklopnika, koliko velik kratkostični tok še uspešno prekine in ostane v stanju delovanja, je podana z njegovo kratkostično izklopno zmogljivostjo. Zmogljivi od-klopniki opravijo izklop v prvi polperiodi izmeničnega toka. Tokokrogi, ki dajo v kratkem stiku večji razpoložljivi tok (prospective current) /p, imajo zaradi večje induktivne komponente ¡mpedance manjši faktor moči, toda amplitudatoka v prvi polperiodi je lahko večja od amplitud v naslednjih periodah. Pri tokih nad 10 000 A pri nekaterih faznih kotih začetka kratkega stika je povečanje amplitude znatno. Na grafu Slike 1 je računsko simuliran vklop toka v pretežno induktivnem tokokrogu pri faktorju moči 0,3 in pri različnih vklopnih faznih kotih. Prva amplituda pri neugodnem vk-lopnem faznem kotu naraste glede na ostale za faktor 1,3. fazni kot [rdi Slika 1: Vklop toka v pretežno induktivnem tokokrogu (faktor moči 0,3 J pri različnih vklopnih faznih kotih, računska simulacija Pri nenadnem povečanju toka, ki ga povzroči nenormalno stanje v tokokrogu, odklopnik reagira samodejno: pri nastavljenem faktorju nadtoka, ki je običajno med 10- in 15-kratnikom toka normalne obremenitve, sproži mehanizem za odpiranje kontaktov. Poleg tega ne glede na nastavljeno prožilno vrednost se na kontaktnem paru v stiku poveča odrivna sila zaradi toka, ki teče skozi majhno stično ploskev med kontaktnima površinama in ta prevlada nad silo kontaktnega stiska, tako da razmakne (odpre) kontaktaktni par (contact blow-off) /5/. V odklopniku za tok delovanja v normalnem stanju pri 25 A se odskok kontaktov zgodi pri trenutni vrednosti toka kakih 1000 A. Vgrajeni magnetni sprožnik deluje z nekaj več zakasnitve od trenutka, ko trenutna vrednost toka preseže nastavljeno vrednost: udarna igla kotve nadtokovnega magnetnega sprožnika, ki jo požene magnetno polje toka v njegovi tuljavi, pospeši odpiranje gibljivega kontakta do hitrosti nekaj m/s. Hkrati sprosti še napeto vzmet mehanizma za izklop, ki požene vklopno-izklopno ročičje, da dokončno razmakne kontakte in jih obdrži odprte na ustrezni medkontaktni razdalji Izklopljenega stanja. Ta zadnji proces je zaradi vztrajnosti relativno velikih mas mehanizma najpočasnejši, vendar pote-kavčasovnem merilu milisekunde. Skupni učinek zaporednega delovanja vseh navedenih izklopnih dejavnosti je velika hitrost pri odpiranju gibljivega kontakta, ki doseže ob ločitvi kontaktnega para od 2 m/s do 5 m/s, v izklopljenem stanju pa je med njima razmik od 9 mm do 10 mm. V trenutku ločitve kontaktov napetost na kontaktnem paru takoj naraste na 10 V do 20 V in se na oscilogramu poteka napetosti na kontaktnem paru jasno opazi kot skokovita sprememba vrednosti. Velikost »stopnice« je odvisna le od vrste kontaktnega materiala, ker v reži velikosti 0,1 mm oblok gori v kovinskih parah s površine kontaktov. Med kontakti z materialom na osnovi srebra z dodatki, npr. Ag/Ni, Ag/C, in na bakru (Cu) znaša (16 ± 1) V /6/. Pri nekoliko večji reži velikosti kak milimeter napetost naraste na 30 V do 35 V; ta vrednost je odvisna tudi od vrste plina, ki obdaja kontakte, ker pri tej medkontaktni reži oblok gori pretežno v okoliškem mediju /7/. Dolžina obtočnega stolpca reda velikosti 1 mm pri napetostnem gradientu približno 2 V/ mm vzdolž reže /8/ prispeva k celotni napetosti zanemarljivo malo. Dokler dolžina obtočnega stolpca ne preseže kritične vrednosti, je obtok zasidran med kontakte in se ga ne da speljati drugam. Ko med odpiranjem kontaktov reža med kontaktnim parom naraste preko kritične dolžine (pribl. 2 mm), lastno magnetno polje obtoka že lahko obtočni stolpec ukrivi v tok ali del zanke in ta se pod vplivom magnetnega polja giblje v smeri Biot-Savartove sile po podaljških kontaktnih delov, kontaktnih »rogov« in obtočnih letev. Potencialno razliko na kontaktnem paru z obtokom med njima podajamo kot »napetost obtoka«. V začetku odpiranja kontaktov v stanju negibljivega obtoka (1 ms do 2 ms od trenutka ločitve kontaktnega para) je približno sorazmerna naraščanju med-kontaktne reže (po oceni 2 V/mm x 5 m/s), v stanju gibljivega obtoka, pa začne naraščati veliko hitreje. Značilni časovni potek napetosti obtoka v stanju odpiranja kontaktov je razpoznaven na oscilogramih izklopa kot »stopnica« napetosti na kontaktnem paru pri prehodu napetostne razlike na sklenjenem kontaktnem paru zaradi toka skozi prehodno upornost na kontaktnem mestu, oz. pri zaprtih kontaktih, na napetost »kratkega obtoka« pri ločitvi kontaktnega para. Pogosto namesto enega kontaktnega para v odklopniku uporabljamo dvojni kontaktni par, ki prekine tokokrog na dveh zaporednih mestih. Gibljivi kontaktni del, lahek in za mehansko izvedbo enostavnejši »kontaktni mostiček«, premosti dva mirujoča kontaktna dela, tako da nastaneta pri izklopu dva zaporedna obtoka. Na oscilogramu obtočne napetosti se pri ločitvi kontaktnega mostička od mirujočih kontaktov opazi dve zaporedni napetostni »stopnici«, visoki po kakih 15 V . Med naraščanjem trenutne vrednosti toka se povečuje tudi presek električno prevodnega obtočnega stolpca. Prosto goreči obtok ima pri toku reda 1 kA do 10 kA premer med 10 mm in 20 mm /9/. Kontaktne dele nizkonapetostnega 66 M. Bizjak: Vpliv sten ob kontaktih z oblokom na izklopno zmogljivost nizkonapetostnega odklopnika Informacije MIDEM 39(2009)2, str. 65-70 odklopnika obdajata bočno dve stranski steni, ki sta razmaknjeni za kakih 5 mm. Slika 2 prikazuje razporeditev sistema para mirujočih kontaktov (na sliki je zaradi preglednosti prikazan le eden od para) in kontaktnega »mostič-ka«, vloženih med stranski steni, ki se na obeh koncih podaljšujeta v deion komori s paketom kovinskih lamel, med katere se oblok razdeli na delne obloke. Dvojni kontaktni par poskrbi, da pri razmaknitvi kontaktnega mostička nastaneta dva izklopna obloka, med vsakim kontaktnim parom eden. Vsak od njiju nekaj časa gori med razmikajočim se kontaktnim mostičkom in mirujočim kontaktnim delom enega in drugega para, potem pa se začneta bolj ali manj sočasno premikati proti koncema kontaktnega mostička in proti obema deion komorama. Pred vstopom v komoro oba obloka gorita v prostoru med stranskima stenama. Skozi vsakega od obeh obločnih stolpcev tedaj teče tok velikosti nekaj kA in glede na toku ustrezni premer prosto gorečega obloka, povzet iz /9/, lahko ugotovimo, da ga stranski steni omejujeta s precej manjšim razmakom in stiskata njegov presek. Oblok se ob stiku s steno učinkovito hladi, pri tem pa se material stene na površini segreje na nekaj 100°C. Taka temperatura presega mejo termične obstojnosti večine plastičnih mas. Površina stranskih sten začne emitirati snovne komponente z izparevanjem in ab-lacijo. Napetostni gradient obločnega stolpca se zaradi vpliva sten lahko precej poveča in s tem tudi celotna napetost obloka/10/. Slika 2: Gibljivi kontaktni mostiček med stenama kontaktno-obločnega prostora z parom deion komor Dokončno pa obločno napetost povečamo na želeno vrednost v deion komori, kjer oblok razdelimo na več zaporednih delnih obtokov. Napetost obtoka l/0biok lahko uravnamo s številom lamel deion komore, ki določa število delnih obtokov ndo- Za obtok med jeklenimi lamelami z razmakom 2,5 mm je bila izmerjena napetost med 26,5 V in 27,5 V, pri ocenjenem gradientu 2,25 V/mm dolžine obloka /8/. Za delni obtok v deion komori nizkonapetostnega odklopnika, ki ima lamele običajno razmaknjene za 1 mm, lahko za oceno napetosti celotnega obtoka L/0biok vzamemo napetost delnega obloka 28 V, U0biok pa je v približku (1) sorazmerna številu delnih obtokov nd0: LUiok = 28 ■ nd0 [V] (1) Velikost napetosti obloka v deion komori in čas, v katerem to vrednost doseže, pogojujeta učinek tokovne omejitve pri izklopu. Z njim želimo doseči omejitev konične vrednosti prepuščenega toka Id in čim manjši joulski integral l2t, ki je določen z (2): 'o i2t = JVck (2) o kjer je / = i(t) tok med skozi odklopnik od trenutka nastanka v času i = 0 do trenutka to, ko i(t) doseže ničelno vrednost, i(to) = 0, in jo obdrži poljubno dolgo, i(t > to) = 0. Če je čas trajanja izklopa to krajši, je tudi vrednost l2t običajno manjša. 3. Ocena vplivov obločne napetosti na tokovno omejitev Vpliv velikosti napetosti obtoka U0biok in časa naraščanja L/obiok (t) do Uobiok na potek prepuščenega toka i(t) pokaže računska simulacija izklopa, kot je prikazana na Slikah 3a, 3b, 3c, 3d, 3e in 3f. S simulacijo so raziskane razmere v pretežno induktivnem tokokrogu, ki ga napaja vir izmenične napetosti 50 Hz z vrednostjo Us = 230 V in pričakovano variacijo + 10%, pri razpoložljivem toku /p = 10 kA, kjer je faktor moči 0,45 z pričakovanim odstopanjem - 0,05. V neugodnem primeru pri Us = 253 V, faktor moči 0,40 kratek stik nastopi pri napetostnem faznem kotu 60°. Stikalna pot skozi odklopnik ima neko impedanco, ki je pretežno ohm-ska in je določena z zgornjo mejo dopustne disipacije P < 2,0 W pri nazivnem toku odklopnika /n. Ko trenutna vrednost i(t) doseže 1000 A (izmerjeno v primeru obravnavanega tipa odklopnika), se zaradi elektrodinamičnega (blow-off) učinka kontaktni par loči, napetost med kontakti u0biok (t) takoj naraste od 0 v f = 0 na 15 V in narašča po t najprej linearno v skladu s predpostavko, da je v začetku odpiranje kontaktnega para linearno po času t, potem pa hitreje proti asimptotični vrednosti. Za presojo učinka hitrosti naraščanja Uobiok (t) proti l/obiok , je opravljena simulacija izklopa za asimptoto u0biok (f) pri 1 ms in pri 2 ms. Simulacija izklopa za obe predpostavljeni vrednosti je prikazana na Slikah 3a, b in c za asimptoto u0blok (f) pri 2 ms, na Slikah 3d, e in f pa pri 1 ms od nastanka kratkega stika. Po tem času se obtok nahaja v deion komori razdeljen med lamelami komore na delne obtoke, zato je napetost obtoka po času konstantna, iv0biok(f) = ^obiok. Vpliv vrednosti Cobiok je obdelana s simulacijo pri vrednostih 250 V, 350 V in 500 V. Te vrednosti so izbrane glede na izbrano napetost vira Us = 253 V: prva je približno enaka Us, druga je blizu temenski vrednosti Us (358 V), zadnja pa je približno enaka 2 Us- Rezultati simulacije za U0b\ok = 250 V so prikazani na grafih Slike 3a in Slike 3d, za C/obiok _ 350 V na grafih Slike 3b in Slike 3e, za Uoblok = 500 V pa na Sliki 3c in Sliki 3f. 67 Informacije MIDEM 39(2009)2, str. 65-70 M. Bizjak: Vpliv sten ob kontaktih z oblokom na izklopno zmogljivost nizkonapetostnega odklopnika Napetost vira Napet.obJoka . Prepušč.iok o Napetost vira Nap-etobloSca a. Prepušč.iok Slika 3a: Simulacija izklopa lp = 10 kA pri U0biok = 250 V, čas dviga u obtok 2 ms. Slika 3d: Simulacija izklopa \p= 10 kA pri U obtok 250 V, čas dviga u obtok 1 ms. ¡rsfotf-Soc Uj-5 3 £OV o Napetost vka 7 Napet, obfcka ¿i Prepušč.tok Slika 3b: Simulacija izklopa lp = 10 kA pri U obtok : 350 V, čas dviga u obtok 2 ms. Slika 3e: Simulacija izklopa \p = 10 kA pri Uobiok ' 350 V, čas dviga u0btok 1 ms. 5 ODG 450Q 4000 3500 3C00 r- 2000 O 1500 h 1CG0 * 500 _) S' o -KO «„ -I0C0 -15C0 -2000 -2500 -3000 -35C0 MT V • Napetost vira ' Napet obSoka . Prspu.šč tok 0 1234 56 7 8 3 10 II i [ms! & <:< A A r* A ft fi A & A & A A A i! A A A A $ Napetost vira v Napet, ob loka i Prepušč.tok Slika 3c: Simulacija izklopa lp = 10 kA pri U0biok = 500 V, čas dviga u0biok 2 ms. Slika 3f: Simulacija izklopa \p= 10 kA pri U obtok = 500 V, čas dviga u0biok 1 ms. Pri presoji vplivov hitrosti razvoja obloka iz »kratkega« ob-loka do popolnoma na delne obloke v deion komori »razdeljnega« obloka iz rezultatov simulacije razberemo, da hitrost prehoda iz začetne faze obloka v končno vpliva predvsem na konično vrednost prepuščenega toka /d, če 68 M. Bizjak: Vpliv sten ob kontaktih z oblokom na izklopno zmogljivost nizkonapetostnega odklopnika Informacije MIDEM 39(2009)2, str. 65-70 je l/obiok > l/s ■ V2. Če torej l/0t>iok ne dosega vsaj temenske vrednosti vira Us, je učinek tokovne omejitve pri izklopu majhen in izklop bo mogoč šele v trenutku blizu prve naravne ničle sinusnega toka. Pri večanju U0biok 2 Us je učinek tokovne omejitve že zelo velik, saj tok /(f) upada od /d —^ 0 že s hitrostjo začetnega naraščanja proti /d. Večji učinek tokovne omejitve v praksi ni sorazmeren vložku, s katerim bi ga dosegli. Da omejimo joulski integral izklopa l2t, je treba predvsem zadostiti kriteriju, da je l/0biok večja ali vsaj približno enaka temenski vrednosti napetosti vira Us, velikost Id je v šibki korelaciji z vrednostjo l2t. Zato pri izklopu Izmeničnega toka 50 Hz (s polperiodo 10 ms) čas naraščanja obločne napetosti pod 1 ms nima odločilnega vpliva na vrednost l2t, če vrednost L/0biok ne dosega kritične vrednosti, dokler ga omejujemo na vrednost, ki je za red velikosti manjša od trajanja ene polperiode izmeničnega toka. Pri ekstremno dolgih časih naraščanja u0biok (t) L/obiok v trajanju 1/2 polperiode ali dalj, seveda učinek tokovne omejitve na velikost l2t tudi pri preseganju kriterija za vrednost U0biok nima uporabnega vpliva. 4. Rezultati preskusov izklopa velikih tokov Nizkonapetostni odklopniki za splošne namene zaščite inštalacij ali motorjev se izdelujejo za različne vrednosti na-zivnih tokov. Za določeno območje nazivnih tokov, npr. po stopnjah za In od 2 A, 2,5 A, 4 A, 6,3 A, 10 A ... do zgornje vrednosti 25 A so istega tipa, to je iste velikosti in konstrukcije, za /n od 63 A navzgor so drugega tipa, ki je večji in drugačne zgradbe. Zaradi funkcionalnih zahtev so včasih med različicami istega tipa, ki pripadajo različnim vrednostim /n, nekatere razlike v materialu in izvedbi posameznih vgrajenih delov. Pri večjih vrednostih nazivnih tokov so potrebni večji preseki vodnikov ali termično bolj odporen material in podobno. Zaradi neizbežnega kompromisa pri spremembi materiala se včasih na preskušanju prototipov ali izdelkov v redni proizvodnji pojavijo neželeni stranski učinki. Na osnovi rezultatov simulacije izklopa, kot so prikazani na Slikah 3(a ... f) so bile določene smernice za konstrukcijsko zasnovo kontaktno-obločnega sistema odklopnika za zaporedje nazivnih tokov /n do 32 A. Odklopniki za nekaj največjih vrednosti /n so bili preskušeni na kratkostično izklopno zmogljivost v laboratorijskih pogojih pri nastavljeni napetosti vira 253 V z razpoložljivim tokom /p = 25 kA. Preskusi izklopa so bili opravljeni pri vklopnih faznih kotih laboratorijskega »kratkostičnega« toka a, a + 30° in a + 60°. Kot a je bil nastavljen na sinusoidi napetosti vira, vk-lopni fazni kot preskusnega toka pa je odvisen od fazne diference med napetostjo in tokom v preskusnem tokokrogu. Pri vsakem preskusu izklopa (operacija »O«) je bil posnet oscilogram prepuščenega toka /(i). Potek u0biok (f) na oscilogramu ni bil registriran zaradi nevarnosti za merilni sistem, ker bi bilo treba za to meriti napetostno razliko med deloma kontaktnega para v prisotnosti obloka. Iz registriranih oscilogramov je bila določena vrednost Id, iz- klopnl čas to, ki je časovni interval med vklopom preskusnega toka in dokončno prekinitvijo toka, ko tok doseže ničelno vednost in ostane na tej poljubno dolgo, s pomočjo vgrajenega integratorja izračunan joulski integral l2t. Preskus je bil opravljen na vsaj treh odklopnikih z istim na-zivnim tokom. Za preskušance z /n 14 A, 18 A, 23 A, 27 A in 32 A so merski rezultati urejeni kot zveza med izklopnimi količinami in nazivnim tokom /n: {fo, b, l2t} = f(/n). Grafično jih podaja Slika 4: za odklopnike nazivnih tokov /n z vrednostjo 14 A, 18 A, 23 A, in 27 A korelacijo med izmerjenimi količinami to, Id, l2t in /n lahko prikažemo kot funkcijsko zvezo. Za /n = 32 A pa rezultati odstopajo od te za več, kot je interval deviacije izmerjenih vrednosti. o Anipiprep.toka j 7 Izkiopnf čas j s Integral R Slika 4: Vrednosti značilnih izklopnih količin za odklopnike zln= 14 A, 18A, 23A, 27A in 32A Razlog za opisano razlika je v zgradbi tipske različice odklopnika za /n = 32 A glede na različice za /n < 32 A, kar se odraža tudi v zveznem spreminjanju izklopnih količin to, Id in l2tz ln do 32 A in z nezveznostjo pri različici za 32 A. V tej je bilo treba zagotoviti večji kontaktni stisk in večjo odpornost kontaktov na zavaritev pri velikih tokih, zaradi velike termične obremenitve odklopnika pri nazivnih pogojih obremenitve pa je bilo treba izbrati za stene kontaktno-ob-ločnega prostora in nosilec gibljivega kontakta termično odpornejši plastični material. Ta običajno v stiku z vročim obločnim medijem manj degradira in zato manj emitira v plazmo obloka. Eksperimentalni rezultati (Slika 4) kažejo, da so se pri odklopniku za 32 A vse vrednosti prikazanih izklopnih količin prekomerno povečale glede na pričakovane, ki bi jih dobili z ekstrapolacije rezultatov za 14 A, 18 A, 23 A in 27 A. S pomočjo rezultatov simulacije izklopa, prikazanih na Slikah 3 (a - f), presojamo, daje pri odklopniku za 32 A povečanje vrednosti l2t verjetno posledica manjše Uobrn, kar hkrati podaljša tudi izklopni čas to, povečanje Id pa posledica počasnejšega naraščanja u0biok (f) do končne l/0biok, ki ga pogojuje počasnejše odpiranje kontaktov Hipoteza, da ima v odklopniku za 32 A vgrajen plastični material za stene kontaktnega prostora manjši učinek gašenja obloka, je bila preverjena eksperimentalno na odklopniku za /n = 18 A. Ta v standardni izvedbi izkazuje dobro 69 Informacije MIDEM 39(2009)2, str. 65-70 M. Bizjak: Vpliv sten ob kontaktih z oblokom na izklopno zmogljivost nizkonapetostnega odklopnika izklopno zmogljivost, ki dobro korelira z rezultati preskusov na odklopnikih ostalih nazivnih vrednosti. Zato sta bili za preskus vpliva plastičnega materiala na Izklopno zmogljivost odklopnika izdelani dve različica odklopnika za 18 A: prva, v kateri so bile stene kontaktno-obločnega prostora iz po-liamida (PA) in druga, v kateri je bil uporabljen polifenilen-sulfid (PPS). V obeh različicah odklopnika za 18 A so bili izdelani po trije preskušanci ene in druge verzije in preskušeni pri enakih pogojih. Izmerjene vrednosti značilnih izklopnih količin to, Id in l2t so prikazane na grafu Slike 5. Pri obeh verzijah je konična vrednost prepuščenega toka Id v mejah merske nenatančnosti enaka. Vrednost l2t pa je za preskušance z vgrajenim PPS materialom skoraj dvakrat večja, kot pri različici z vgrajenim PA materialom. Pri tem se je tudi izklopni čas to povečal iz povprečno 4 ms na povprečno 7 ms. Rezultati preskusa očitno kažejo kvalitativni vpliv materiala na stene kontaktno-obločnega prostora na izklopno zmogljivost odklopnika. ŠJ 100 plastična masa PA plastična masa PPS A Ó £ ' . * * » ♦ * > Ampí.ptep.íoka 7 izklopiti Cos č. Integra! !2t merilni inštrumentarij. Zato smo za razlago pojavov, ki so registrirani na oscilogramih, opravili še simulacijo izklopa z ad hoc računskim modelom. S hkratno obravnavo eksperimentalnih in simulacijskih rezultatov smo kvalitativno razpoznali vpliv plastičnega materiala za stene kontaktno-obločnega prostora na oblok pri izklopu kratkostičnega toka in izbrali za praktično uporabo ustreznejši material. 6. Literatura /1/ M. Lindmayer: Schaltgeräte, Springer, 1987 /2/ G. Burkhard: Schaltgeräte der Enenergletechnik, I.Auflage, VEB Verlag Technik, Berlin, 1985 /3/ Standard SIST EN 60947-2: 2006, Nizkonapetostni odklopniki /4/ Standard SIST EN 60898-1: 2004, Inštalacijski odklopniki /5/ E. Vinaricky: Elektrische Kontakte, Werkstoffe und Anwendungen, 2. Auflage, Springer, 2002 /6/ A. Erk, H. Finke: Über das Verhalten unterschiedlicher Kontaktwerkstoffe beim Einschalten prellender Starkstrom-Schaltglieder, ETZ-A 86 (1965), pp.297-302 /7/ E. Vinaricky: Das Abbrand-und Schweissverhalten verschiedener Silber-Grafit-Kontaktwerkstoffe In unterschiedlichen Atmosphären, disertacija, TH Wien, 1994 /8/ H. Klepp: Über den Einfluss der Löschkammerkonstruktion auf die Lichtbogelöschung In Schützen grosser Nennstromstärken, disertacija, TU Carolo-Wilhemlna, Braunschwelg, 1982 /9/ P. G. Slade: Electrical Contacts, Marcel Dekker, 1999 /10/ M. Bizjak: Model ablacijskostabiliziranega oblokavodklopniku, disertacija, Univerza v Ljubljani, Fakulteta za elektrotehniko in računalništvo, 1992 3 4 5 oznaka preskušanca Slika 5: Odvisnost značilnih izklopnih količin to, Id in l2t od vrste plastičnega materiala ob kontaktih, izklop lp = 25 kA 5. Sklep Značilnosti izklopnega pojav pri preskusu kratkostične izk-lopne zmogljivosti se odražajo v oscilogramih časovnega poteka napetosti med kontakti in prepuščenega toka. Merjenje medkontaktne napetosti in napetosti obloka zahteva poseg v merjeni vzorec, ki ni zaželen, nevarno pa je tudi za dr. Martin Bizjak, univ. dipl. ing. fizike R&R, Iskra MIS, d. d. Ljubljanska cesta 24a 4000 Kranj e-mail: martin. bizjak@iskra-mis. si Prispelo (Arrived): 02.09.2008 Sprejeto (Accepted): 09.06.2009 70 UDK621,3:(53+54+621 +66), ISSN0352-9045 Informacije MIDEM 39(2009)2, Ljubljana SIMULACIJA KOMUNIKACIJSKIH SISTEMOV V REALNEM ČASU Z REALNO KOMUNIKACIJSKO OPREMO V SIMULACIJSKI ZANKI Matjaž Fras, Jože Mohorko Univerza v Mariboru, Maribor, Slovenija Kjučne besede: simulacija, taktična omrežja, realni čas, radijska vidljivost Izvleček: V članku smo predstavili nov pristop k simulacijam komunikacijskih omrežij, ki omogoča simulacije v realnem času z realno komunikacijsko opremo v simulacijski zanki. Pri tem pristopu je mogoče v simulacijsko okolje priključiti realne komunikacijske naprave, ki lahko komunicirajo s simuliranimi napravami ali pa z drugimi realnimi napravami preko simuliranih povezav ali omrežij. Tovrstne simulacije omogoča simulacijsko okolje OPNET Modeler z dodatnim novim modulom SITL (System-in-the-Loop), ki omogoča povezavo med realnimi in simuliranimi napravami. Uporabnost tovrstnih simulacij smo predstavili na primeru »real-sim-real« tipa simulacije, kjer realne komunikacijske naprave komunicirajo preko simuliranega omrežja v realnem času. V tem primeru smo, preko simuliranega brezžičnega omrežja na virtualnem terenu, prenašali realen promet digitalne telefonije (VolP). Real time communication systems' simulation with real communication devices in simulation loop Keywords: simulation, tactical network, real time, radio visibility Abstract: This paper presents the new concept of the communication networks' simulations, which enable real time simulations, with the real communication equipment in the simulation loop. In the context of project »Modeling of Command and Control information systems« we develop the simulation system, which can be used in the tactical network's planning and evaluation process. This simulation system is based on the OPNET Modeler simulation tool, which is combined by our utilities (TPGen, ModCom), that enables traffic modeling of Command and Control information system (C2IS) used in Slovenian army /1, 2/. We developed also the expert system for automatic simulation results analysis to evaluate tactical networks' performances /3, 4/ . In the context of this project we also research the possibilities for real time tactical network simulations, where real tactical units' computers with installed C2IS communication are connected to the simulated wireless communication infrastructure on virtual terrain. The first such simulator is Battlespace communication network planner and simulator, which are described in /5,6/. This simulator is based on special hardware communication equipment controlled by OPNET simulation tool, where real traffic is influenced by simulation results in real time. In our research we used, Instead to the special hardware, the new OPNET Modeler's software module System-in-the-Loop (SITL) /7, 8/. This module allows interconnections and packets exchange between real and simulated communication devices or networks. This simulation concept is one of the most important novelties, in the last years and it was also presented in the OPNET conference OPNETWORK 2007 in Washington /10/. Figure 1 shows all possible types of the SITL simulations. Figure 2 shows the example of the real-sim-real SITL simulation type, where real computers communicate through the simulated network. The mechanisms of real time SITL simulations with required software and hardware equipment are shown on the block scheme from Figure 3. In our SITL simulation test case of transmitting voice over IP (VoIP), we connect two external laptop computers across the simulated network as shown in Figure 4. The simulated network, which was modeled in OPNET simulation tool, Is more detailed shown in Figure 5. The connectivity between external laptops was tested using ping utility as shown in Figure 6. As VoIP application /15/ we used freeware D-Voicer /12/ installed on real laptop computers. Figure 7 shows the user interface of active D-Voicer application on both, client and server, side. The results of SITL simulation, where real digital voice stream is transmitted through simulated network are shown in Figures 8 and 9. In the third part we present the proposed concept of real time SITL tactical wireless networks'suitable for Slovenian army. The simulation concept is similar as in VoIP example case, described in previous section. The main difference is in the used communication applications, which are forthis case Command and Control Information System (C2IS). In Slovenian case this C2IS consists from Iris Replication Mechanism (IRM) in conjunction with Sitaware graphical interface/1,2, 16/as shown in Figure 10. Tactical radios of military units we modeled, in OPNET Modeler simulation tool, using standard wireless routers connected over SITL gateway to the real tactical computers. Communication parameters of wireless routers are defined according to communication parameters of real tactical radios and modems. In OPNET Modeler simulation tools are considered influences modeled virtual terrain on wireless links using different radio wave propagation models/13, 19-22/. Simulated tactical network can be very realistically visualized using 3DNV visualization tool as shown in Figure 11. 1. Uvod Visoka kompleksnost sodobnih komunikacijskih sistemov in kratek čas, ki je na voljo za iskanje kakovostnih rešitev, narekujeta potrebo po simulacijskih orodjih, ki so namenjeni simulacijam telekomunikacijskih omrežij, naprav, protokolov. Takšno potrebo je začutila tudi slovenska vojska ko je razpisala projekt1 »Modeliranje taktičnih informaci- jskih sistemov poveljevanja in kontrole (TISPINK)« v okvirju katerega smo razvili simulacijski sistem, ki omogoča kakovostno načrtovanje, ovrednotenje in optimizacijo taktičnih komunikacijskih omrežij. Simulacijski sistem /1, 2/ temelji na simulatorju OPNET Modeler, ki smo ga nadgradili z aplikacijami TPGen, za avtomatsko modeliranje prometa med vozlišči omrežja v TISPINK, ModCom za avtomatsko modeliranje taktičnih radijskih naprav za prenos podatkov, Ek- 1 Rezultati, predstavljeni v tem članku so rezultat raziskav narejenih v okviru CRP projekta "Modeliranje informacijskih sistemov poveljevanja in kontrole", financiranega s strani Slovenskega ministrstva za obrambo. 71 Informacije MIDEM 39(2009)2, str. 71-77 M. Fras, J. Mohorko: Simulacija komunikacijskih sistemov v realnem času z realno komunikacijsko opremo v simulacijski zanki spertnim sistemom /3, 4/ za avtomatsko analizo in vrednotenje simulacijskih rezultatov, ter Taktičnim predvajaln-ikom /23/ za nazorno vizualizacijo rezultatov ekspertnega sistema. V takšnem simulacljskem sistemu je najtežavnejši problem natančno modeliranje kompleksnih aplikacij, kakršne so uporabljene tudi v TISPNK. V opisanem simu-lacijskem sistemu smo ta problem poenostavili s statističnim modeliranje podatkovnih izvorov. V tem članku bomo pokazali drugačen pristop, kjer se izognemo modeliranju na aplikacijskem nivoju tako, da v simulacijsko zanko, v realnem času, povežemo realno komunikacijsko opremo. Eden prvih takšnih simulatorjev je predstavljen v člankih /5, 6/, kjer so avtorji predstavili simulator taktičnih omrežij v realnem času z imenom »Bat-tlespace communication netvvork plannerand simulator«. Izdelan simulator temelji na simulacijskem okolju OPNETv povezavi s specialno v ta namen razvito komunikacijsko opremo, ki omogoča emulacijo vplivov simuliranih radijskih naprav na podatkovne prenose med realnimi komunikacijskimi napravami In aplikacijami. V naši raziskavi smo za tovrstne simulacije uporabili namesto specialne emulacijske strojne opreme programsko rešitev, ki jo ponuja najnovejši OPNET modul System-in-the-Loop (SITL) /11 /, katerega smo že na kratko opisali v/9/. Modul SITL omogoča izmenjavo paketov med realnimi in simuliranimi napravami ter predstavlja zadnjo večjo novost na področju simulacij komunikacijskih omrežij, ki je bila predstavljena na konferenci OPNETVVORK2007 /10/. V članku bomo predstavili nekaj rezultatov, ki smo jih pridobili v fazi testiranja tega koncepta, pri načrtovanju simulatorja TISPINK. V fazi testiranj smo izvedli »real-sim-real« tip simulacij, kjer realne komunikacijske naprave komunicirajo preko simuliranega omrežja. Preko simulatorja (računalnik z nameščenim simulacijskim okoljem OPNET Modeler in dodatnim SITL modulom) smo povezali dva realna prenosna računalnika. Komunikacijo med priključenima prenosnikoma smo najprej testirali s pomočjo ICMP ping aplikacije, nato pa smo izvedli še prenos digitalnega govora (VolP). Kot VolP aplikacijo smo uporabili D-Voicer /12/, ki omogoča prenos digitalnega govora z uporabo širokopasovne internetne povezave. V drugem delu tega članka smo predstavili zasnovo sistema za simulacijo taktičnih omrežij slovenske vojske. V tem primeru so preko simulatorja povezani reali taktični računalniki, z nameščeno specialno programsko opremo za sisteme poveljevanja in kontrole, kot je npr. replikacijski mehanizem IRM /1, 2/. V simulatorju je modelirano brezžično radijsko omrežje na virtualnem terenu, kjer je upoštevan vpliv terena na širjenje radijskih valov/13/. Takšen simulacijski sistem omogoča zelo realistično urjenje poveljniškega kadra. Članek je sestavljen iz naslednjih poglavij. Drugo poglavje opisuje modul SITL. Tretje poglavje opisuje »real-sim-real« tip simulacije, kjer realne komunikacijske naprave komunicirajo preko simuliranega omrežja. Uporabnost tega kon- cepta smo prikazali na primeru prenosa digitalnega govora (VolP) preko preprostega simuliranega omrežja. V četrtem poglavju je predstavljena zasnova trenažnega sistema poveljevanja in kontrole Slovenske vojske. V zaključku so podani predlogi za možnosti uporabe tovrstnih simulacij. 2. Modul SITL Simulacijo v realnem času z realnimi napravami v simulacijski zanki smo izvedli s pomočjo OPNET modula System-in-the-Loop (SITL) /9, 11/. Modul SITL omogoča povezavo in izmenjavo paketov med realnimi omrežnimi napravami in simuliranimi omrežnimi napravami v realnem času. Obstajajo tri tipične konfiguracije povezovanja zunanjimi omrežnimi napravami, ki jih prikazuje slika 1. • Realno omrežje z realnim omrežjem (preko simuliranega omrežja) m realno omrežje B<4 ifl G f 1 p» simulirano omrežje realno omrežje Simulirano omrežje s simuliranim omrežjem (preko realnega omrežja) i. \ V' simulirano omrežje rodlno omrežje simulirano omrežje Simulirano omrežje z realnim omrežjem y <-.,> >• M m simulirano omrežje realno omrežje Slika 1: Možnosti povezav realnih in simuliranih naprav ali omrežij s pomočjo modula System-in-the-Loop /10/. Najzanimivejši primer povezav predstavlja prvi tip simulacije (»real-sim-real«), kjer povežemo realne komunikacijske naprave preko simuliranega omrežja. Simulirano omrežje vpliva na realne pakete v obliki zakasnitev, izgub, podvojitev, itd. SITL omogoča, s pomočjo prevajalnih funkcij, povezavo med simulacijskim okoljem z OPNET Modelere-jem ter Ethernet fizičnimi vmesniki gostujočega računalnika. S pomočjo prevajalnih funkcij se realni paketi pretvorijo v simulacijske pakete (in obratno). V primerih uporabe protokolov, ki s SITL niso podprti, pa lahko tudi sami napišemo svojo prevajalno funkcijo/11/. V verziji OPNET Modeler 14.5 so, v kombinaciji s SITL, podprti naslednji protokoli: Ethernet, IPv4 in IPv6, ICMP, ICMPv6, OSP-Fv2, RIFV1, RIFV2, TCP, UDP, okrnjen FTP. V primeru tipa simulacij, kjer realna omrežja komunicirajo preko simuliranega omrežja, SITL praktično podpira vse potrebne protokole, saj simulirano omrežje uporabimo le kot transportno omrežje. Primer takega simulacijskega scenarija je prikazan na sliki 2, kjer realni klient dostopa do realnega strežnika preko simuliranega omrežja, predstavljenim z internetnim oblakom. Na SITL vmesniku, kjer je priključen na simulacijo realni prenosni računalnik, se paketi preoblikujejo v simulacijske, ti paketi potujejo preko 72 M. Fras, J. Mohorko: Simulacija komunikacijskih sistemov v realnem času z realno komunikacijsko opremo v simulacijski zanki Informacije MIDEM 39(2009)2, str. 71-77 simuliranega interneta, nato se pri drugem SITL vmesniki, ponovno pretvorijo v realne pakete, ki potujejo do realnega strežnika. Slika 2: Primer uporabe »real-sim-real« tip simulacij, kjer realni gostitelj in strežnik komunicirata preko simuliranega omrežja v simulacijskem okolju OPNET. Slika 3 prikazuje obravnavanje paketov na simulacijskem računalniku v primeru uporabe v SITL modula. Realne pakete, ki prispejo na Ethernet vmesnik (mrežne kartica NIC) in so namenjeni za simulacijo, blokira na poti do lokalnega operacijskega sistema požarni zid. Hkrati se ti paketi, s pomočjo gonilnikov VVinPCap /14/ zajamejo, filtrirajo in preusmerijo preko SITL prehoda v simulirano omrežje. zunanja realna komunikacijska 11 oprema NIC Simulator Slika 3: Blokovna shema simulacijskega sistema z nameščeno programsko opremo OPNET in SITL, kjer se izvaja simulacija v realnem času. Modul SITL je primeren za uporabo v naslednjih primerih /10/: opazovanje vplivov simuliranega omrežja na realno omrežje in obratno. simuliranje vpliva obremenjevanja realnega omrežja s pomočjo simuliranega generatorja prometa, razvoj in preizkušanje prototipov novih protokolov in naprav z povezavo realnih naprav ali protokolov z simuliranimi. 3. zunanje krmiljenje simuliranih naprav v realnem času testiranje zmogljivosti novih komunikacijskih protokolov. testiranje skalabilnosti s povečanjem števila realnih naprav v omrežju s pomočjo dodatnih simuliranih naprav itd. Prenos realnega digitalnega govora preko simuliranega omrežja Metodologijo simulacij tipa »real-sim-real« bomo pokazali najprej na primeru prenosa digitalnega govora preko simulirane omrežne infrastrukture. Na simulacijski računalnik smo priključili dva prenosnika (Prenosnik 1 in Prenosnik 2) preko dveh mrežnih vmesnikov (NIC 1 in NIC2). Princip priključitve zunanjih prenosnikov na simulator s potrebnimi nastavitvami prikazuje slika 4. Ed£ Lšcerise Prenosnik 1 IP 192.168.1.6 Prenosnik 2 IP 192,168.2.6 Slika 4: Povezava zunanjih realnih prenosnikov (Prenosnik 1 in Prenosnik 2) na simulator, kjer se izvaja simulacija komunikacijske infrastrukture. Konfiguracija z dvema mrežnima karticama na simulacijskem računalniku je nujna v primeru, ko so internetni naslovi obeh računalnikov iz istega podomrežja. V nasprotnem primeru bi zadostoval, na simulacijskem računalniku, en sam vmesnik, na katerega bi priključili oba prenosnika preko zvezdišča. S temi ukrepi zagotovimo, da med zunanjimi napravami ni direktne komunikacije. Potem, ko smo povezali zunanje računalnike na simulator in nastavili komunikacijske parametre (IP naslovi, nastavitev požarnih zidov, itd.), smo se lotili še modeliranja simu- 73 Informacije MIDEM 39(2009)2, str. 71-77 M. Fras, J. Mohorko: Simulacija komunikacijskih sistemov v realnem času z realno komunikacijsko opremo v simulacijski zanki liranega omrežja, preko katerega bo potekala komunikacija z realnimi zunanjimi računalniki. Vsimulacijskem okolju OPNET imamo na voljo posebno knjižnico s SITL gradniki kot sta SITL prehod (SITL_Gateway) in SITL povezavo (SITLJink). SITL prehod omogoča prehod paketov med simuliranimi in realnimi komunikacijskimi napravami. Na SITL prehodu lahko izbiramo prevajalne funkcijo za pretvorbo paketov ter definiramo filter za zajem paketov na mrežni kartici. S tem lahko določimo katere pakete bomo usmerili v simulacijo in katere blokirali. Pomembna je tudi definicija izvora IP paketov, kjer določimo, s katere mrežne kartice bodo zajeti realni paketi. SITL povezava se uporablja za povezavo med SITL prehodom in prvo simulacljsko napravo. Slika 5 prikazuje simulirano omrežje, ki smo ga uporabili v našem primeru za prenos digitalnega govora med realnima prenosnima računalnikoma (Laptop 1 In Laptop 2). Simulirano omrežje je sestavljeno iz dveh stikal ter štirih usmerjevalnikov. nik 1 in Prenosnik 2) smo namestili Voice over IP (VolP) aplikacijo D-Voicer/12, 15/. D-Voicer je preprosta aplikacija, ki omogoča prenos govora z uporabo širokopasovne internetne povezave. Uporabniški vmesnik te aplikacije je prikazan na sliki 7. VolP storitve pretvorijo analogni govor iz mikrofona v digitalni signal, katerega nato prenašamo po omrežju. Na drugi strani omrežja se digitalni signal ponovno pretvori v govor in reproducirá na slušalkah. ■DnHHBHEUi i Computer rwne oí IP address Port ,¡182.168.2.6 i 5007 '":•:■:■ - : Disconnect n. J* ? Cormected. codec: speex 16 KHz D D-Bross MBMBMHMMSBg,. Computer name or IP address Pot t f _ 'l ^ni 17 Ejt m f ? ■ Cormected. codec: spees 1 b KHs D-Bross Slika 5: Simulirano omrežje uporabljeno v simulacijskem primeru za prenos digitalnega govora. Povezljivost med prenosnikoma preko simuliranega omrežja smo najprej testirali s pomočjo ICMP aplikacije ping, ki smo jo izvedli na enem izmed obeh prenosnikov. Komunikacija med realnima računalnika je možna le v primeru, daje simulacija aktivna in so vsi komunikacijski parametri pravilno nastavljeni. Na sliki 6 je prikazan primer uspešnega testa povezljivosti. Slika 6: Preverjanje povezljivosti med realnima računalnikoma (Prenosnik 1 in prenosnik 2) povezanima preko simuliranega omrežja. Ko smo vzpostavili povezavo med obema računalnikoma, smo izvedli še preizkus prenosa digitalnega govora preko simuliranega omrežja. Na prenosna računalnika (Prenos- Slika 7: Uporabniški vmesnik aplikacije D-Voicer za primer oddajanje digitalnega govora na prenosniku (Prenosniki) in sprejemanje tega signala na drugem prenosniku (Prenosnik 2) preko simuliranega omrežja. Po izvedeni simulaciji lahko na podlagi zajetih statistik, v OPNET Modelerju, opravimo podrobno analizo rezultatov. Slika 8 prikazuje realni promet digitalnega govora izmerjenega na SITL prehodih, ki predstavljajo točko vstopa podatkov digitalnega govora v simulirano omrežje (zgornji graf) ter točko izstopa (spodnji graf). Iz rezultatov simulacij vidimo, da je promet, ki ga oddaja prenosnika 1 enak tistemu, ki ga sprejema prenosnik 2, kar pomeni, da ni bilo izgubljenih podatkov. Slika 9 prikazuje izmerjene zakasnitev (zgornji graf), ki nastajajo v simuliranem omrežju pri prenosu digitalnega govora, ter promet izražen v paketih/s, ki smo ga izmerili na simuliranem stikalu 2. 74 M. Fras, J. Mohorko: Simulacija komunikacijskih sistemov v realnem času z realno komunikacijsko opremo v simulacijski zanki Informacije MIDEM 39(2009)2, str. 71-77 30000 25000 M AjuUULA ÎUUJJUJ Â1AJJ. /L 1 - ■ - 20000 ! i aj 15000 i f3 ; . sprejet promet 1?. iaptop 1 10000 5000 f 0 0 20 40 60 80 100 120 čas (s) 30000 25000 \r.Ji /l/-, jv; h a ,..>l A /f, /s, A , K À Š, -I«» ! 20000 : J ................................■ - 3 : F ..........; a 15000 i QJ Z V 10000 .........J................promet îaptopu 2 2 f. .......... ' 5000 j I 0 ■ ' -5000 20 40 • ......60 ' 80....... 100 • 120 čas{s) Slika 8: Promet digitalnega govora v bit/s izmerjen na SITL prehodih pri vhodu in izhodu iz simuliranega omrežja. om , /tv^l/m i lf <» ! » 0.015 i • r ' Q : uk! to enrUafOMllte/ 0 20 40 60 80 100 120 čas (s) 70 60 pjïhrt pfonu t iki t kc i.. i 50 ! i . 1 "z 40 ■ ra ¿t 30 O. 20 10 0 ! 0 20 40 60 80 100 cas(s) 120 Slika 9: Prvi graf prikazuje zakasnitve digitalnega govora pri prenosu preko simuliranega omrežja. Drugi graf prikazuje poslan promet v paketih/s preko drugega stikala v simuliranem omrežju. Iz rezultatov na sliki 9 vidimo, da so zakasnitve v simuliranem omrežju pri prenosu digitalnega govora minimalne. To je tudi za pričakovati, saj so bile uporabljene simulirane povezave s kapaciteto 100Mb/s minimalno obremenjene. 4. Načrtovanje lastnosti sistema za simulacije taktičnih omrežij v realnem času V zadnjih letih se je vse bolj pojavila potreba po simulacijah v realnem času z realno opremo v simulacijski zanki. Ta trend je še posebej izražen na vojaškem področju. Eden pravih eksperimentov na tem področju je »Battlespace communication network planner and simulator« /5, 6/, kjer je bila razvita namenska programska in strojna komunikacijska oprema, ki je omogočila povezavo med realnim in simuliranim komunikacijskimi napravami. Koncept, ki ga bomo predstavili tukaj, temelji na programski rešitvi, kjer smo namesto posebej razvite strojne opreme uporabili v prejšnjih poglavjih predstavljen OPNET modul SITL. Na osnovi tega koncepta želimo zasnovati trenažer taktičnih informacijskih sistemov poveljevanja In kontrole (TISPINK) (C2IS - Command and Control Information Systems), ki bo omogočal zelo realistično urjenje poveljniškega kadra. V takšnem sistemu bi povezali realne taktične računalnike, na katerih je nameščena TISPINK programska oprema (IRM in Sitaware), s simuliranim taktičnim radijskim omrežjem na virtualnem terenu. Na sliki 10 je predstavljena zasnova takšnega simulacijskega sistema. Vsaka taktična enota ima vsaj en taktični računalnik, ki se povezuje z ostalimi enotami preko TISPINK sistema. V trenažnem sistemu smo vse taktične računalnike povezali v Ethernet lokalno omrežje preko zvezdlšča. Osnovo TISPINK programske opreme slovenske vojske predstavljata grafični vmesnik Sitaware in replikacijskl mehanizem IRM /1, 2/, ki skrbi za izmenjavo podatkov med podatkovnimi bazami enot. Obe programski opremi sta produkta Danskega proizvajalca Systematic /16/. V isto lokalno omrežje povežemo tudi računalnik, na katerem je nameščena simulacijska programska oprema OPNET Modeler z dodatnim modulom SITL. V simulacijskem okolju OPNET na virtualen teren postavimo komunikacijske modele taktičnih enot (za vsak taktični računalnik po en model). Taktične radijske postaje so modelirane z brezžičnimi usmerjevalniki, katerim lahko nastavljamo parametre, kot so oddajna moč, tip antene, modulacija, frekvenca kanala, kapaciteta kanal, itd. Trenažni sistem temelji na že predstavljenem tipu »real-sim-real« SITL simulacij. V takšnem trenažnem sistemu se bodo realni TISPINK podatki replici-rali preko simuliranega brezžičnega omrežja v simulatorju. Simulator bo zelo realistično vplival na realne pakete v obliki zakasnitev, izgube podatkov zaradi radijskih motilnikov (jamerjev), izgube radijske vidljivosti, itd. TISPINK operaterji bodo komunikacijske probleme zaznali na enak način, kot da bi bili na realnem terenu z realno komunikacijsko opremo. Takšen sistem omogoča, zraven realističnega treninga poveljnikov in TISPINK operaterjev, tudi kakovostno 75 Informacije MIDEM 39(2009)2, str. 71-77 M. Fras, J. Mohorko: Simulacija komunikacijskih sistemov v realnem času z realno komunikacijsko opremo v simulacijski zanki načrtovanje radijskih omrežij in razvoj taktičnih strategij, saj je mogoče že v laboratoriju predvideti velik del komunikacijskih problemov, ki bi se drugače pokazali šele na terenu. •^-'—i ENOTA 1 0 ô "-"/p, „m o o pà jm' ■J À o hub jft^l simulator OPNLl-SITL i*. S//7 - 3 ■> i F * c j,) < 0 —" , . if " " "I - i", J" ■> ft) Fig. 5. Used different queuing disciplines upon VoIP and HTTP traffic WO. ■ CO. CBWRi. MWRR ■ DYVRIi ^ Rctay in the network (seconds) 025- tm- (LIS-- aro-- QJ05--- _ 14» 15t 16« 1ft 18« 19. 30« Fig. 6. Time avarage global delay in the network 82 S. Klampfer, J. Mohorko, Z. Cucej: IP Packet Queuing Disciplines as Basic Part of QOS Assurance Within the Network Informacije MIDEM 39(2009)2, str. 78-84 Figure 7 shows the amount of VoIP dropped packets, when using different queuing schemes. As we mentioned above, best results in that case we obtain with CBWFQ method which have fixed guaranteed amount of bandwidth. Then follow WFQ, DWRR, MWRR and CQ queuing scheme. The opposite situation is when we take in to the consideration delays. There selection of CBWFQ introduces biggest delay, because majority low level traffic must wait. Fig. 7. Tested wired network arhitecture which is a copy of real network On the Figure 8 we will show, how combined queuing method PQCBWFQ improves the delays comparison with before presented delays, shown on figure 6. As we can see on Figure 8, PQ-CBWFQ delay is smallest than WFQ delay, but on figure 6 has ordinary CBWFQ method bigger delay than WFQ, observed in whole ether-net segment. Such combinations can perceivable improves network performances. Similar effect that is shown as on Figure 8 can be seen also in Figure 9 for VoIP delay. ■ PQ;CBW['Tj_ ■ WK> VoIP Dcky (seconds) 1.8-- time (sec) Fig. 9. VoIP delay (seconds) for combined PQCBWFQ method comparison with WFQ Using combined queuing method the delay is also reduced in comparison with ordinary WFQ queuing for VoIP traffic. Delay in VoIP application plays an important role on quality of perception. The smaller it is, the better voice quality can be offered. VflRQ BDWRR œwfla a oa MWRR Dsopcnl IFtrifOc (pjcketv'w) ■ PtfcCBWFq a wro Efhemcf Del«' (sceoiuls! Qm 10s Om 1 Si Dm 20 s Fig. 8. Ethernet delay (seconds) for combined PQCBWFQ method comparison with WFQ 7. Conclusion During many simulation runs and graph analysis we can say that queuing policy discipline influences on quality of service for the network applications. In many cases CQ queuing discipline was the best choice, in the case, when we have only two traffic flows WFQ has been the best choice, but looking from other perspective, when we must handle multiple traffic flows CBWFQ was the best solution. Such method (CBWFQ) has also disadvantages; in our case we defined only one class with bandwidth amount 9Mbit/s reserved for VoIP, rest of bandwidth belongs to majority low priority HTTP traffic. In that point, majority traffic hasn't enough bandwidth and must wait, what influences on common delay. This is the main reason why CBWFQ has highest average delay in network. No matter to that delay, VoIP delay is constant during simulation because of ensured bandwidth by defined class. Looking from other side, if we want fairness queuing discipline which serves fair all applications then we use WFQ or CQ mechanism, but if we want only that highest priority traffic flows pass through the network, we should use priority queuing PQ. 83 Informacije MIDEM 39(2009)2, str. 78-84 S. Klampfer, J. Mohorko, Z. Cucej: IP Packet Queuing Disciplines as Basic Part of QOS Assurance Within the Network Delays in CBWFQ case can be reduced using PQ-CBW-FQ queuing scheme. This is a combination of a priority PQ and a CBWFQ queuing mechanisms. PQ-CBWFQ is a queuing method from family that offers smallest delays. Such queuing allows that specific flow class defined with IP priorities is served as strict priority queue. Highest priority class is served before all other priority classes. Such combination and their improvement from view of Ethernet delays are shown on Figure 9. Our simulations show that we must look for solutions also in using combined queuing methods. All other available combinations represent a challenge for further researches on that area. /7/ Kian Meng Yap, Alan Marshall, Wai Yu, "Providing QoS for Multimodal System Traffic Flows in Distributed Haptic Virtual Environments," Queen's University Belfast /8/ Internetworking Technology Handbook - Quality of Service (QoS), Cisco Systems /9/ OPNET Modeler Techical Documentât /10/ Samuel P. Morgan, "Queuing Disciplines and Passive Congestion Control in Byte-Stream Networks," IEEE Transactions On Communications /11/ http://www.cisco.com/en/US/docs/ios/12_0/qos/configura- tion/guide/qccq.html /12/ Maria Johanna Gerarda van Uitert, "Generalized Processor Sharing Queues," Masterdam 2003 References /1/ S. Klampfer, J. Mohorko, and C. Žarko, "Vpliv različnih načinov uvrščanja na karakteristiko prepustnosti omrežja," ERK 2007 /2/ M. Fras, S. Klampfer, Ž. Čučej, "Impact of P2P traffic to the IP communication network's performance," IWSSIP 2008 /3/ T. Subash and S. IndiraGandhi, "Performance Analysis of Scheduling Disciplines in Optical Networks," MADRAS Institute od Technology, Anna University /4/ Jernej Kozak: Podatkovne strukture in Algoritmi. Društvo M FA SRS, Ljubljana. /5/ Lary L. Peterson and Bruce S. Davie, Computer Networks, Edition 3, San Francisco 2003 /6/ Stefan Bucheli, "Compensation Modeling for QoS Support on a Wireless Network," Master degree thesis Saša Klampfer, Jože Mohorko, Žarko Čučej University of Maribor, Faculty of Electrical Engineering and Computer Science, Maribor, Slovenia sasa. klampfer@uni-mb. si Prispelo (Arrived): 28.10.2008 Sprejeto (Accepted): 09.06.2009 84 UDK621,3:(53+54+621 +66), ISSN0352-9045 Informacije MIDEM 39(2009)2, Ljubljana CANTILEVER METHOD FOR DETERMINATION OF D31 COEFFICIENT IN THIN PIEZOELECTRIC FILMS Samo Penič, Uroš Aljančič, Drago Resnik, Danilo Vrtačnik, Matej Možek, Slavko Amon Laboratory of Microsensor Structures and Electronics (LMSE), Faculty of Electrical Engineering, University of Ljubljana, Ljubljana, Slovenia Keywords: piezoelectric, characterization, cfej coefficient, PZT, simulation, FEM, ANSYS Abstract: A cantilever method for characetrization of thin piezoelectric films is proposed. Using the proposed cantilever method, piezoelectric coefficient d31 of thin film piezoelectric material on various samples was determined. Cantilever based characterization method provides a fast comparison of different piezoelectric material samples, since multiple samples can be mounted simultaneously on the testing structure. It is shown how, when combined with numerical simulation, piezoelectric coefficient d3^ can be determined from fitting measured voltage response with simulated response. Exact knowledge of geometry and material properties of cantilever and samples proved to be important in order to determine piezoelectric coefficients with sufficient accuracy. Stainless steel cantilever was adequately characterized by measuring its Young's modulus. Silicon properties are adequately determined by published data. Mechanical properties of PZT layers are on the other hand more difficult to acquire, since they are rather dependent on the actual PZT preparation procedure and composition. Nevertheless, we expect that error here introduced is small due to very thin PZT layer compared to stainless steel cantilever and silicon substrate. To Improve the proposed method, based on numerical simulation results, guard chips were mounted at the side of the cantilever to reduce stress variation over samples. Determined values of piezoelectric coefficients c/31 for PZT layers under test were in reasonable agreement with results available in the literature. Metoda za določanje koeficienta c/31 tankih piezoelektričnih filmov Kjučne besede: piezoelektrik, karakterizacija, d3i koeficient, PZT, simulacija, silicij, FEM, ANSYS Izvleček: V članku je predstavljena metoda za karakterizacijo tankih piezoelektričnih plasti. Z uporabo ročice smo določili piezoelektrični koeficient cf3i tankih piezoelektričnih filmov. Metoda omogoča hitro primerjavo lastnosti različnih materialov, ter s pomočjo numerične simulacije hkratno karakterizacijo parametra c/31 večih vzorcev. Poznavanje geometrije in materialnih lastnosti ročice in vzorjev je ključno za natančno določitev piezoelektričnih koeficientov. Mehanske lastnosti jeklene ročice smo določili z meritvijo Youngovega modula, za mehanske lastnosti silicijevega substrata pa smo uporabili podatke v literaturi. Mehanske lastnosti tankih PZT plasti so težje dostopne, saj se razlikujejo zaradi same zgradbe PZT keramike ter njene priprave. Zaradi tanke plasti PZT materiala, ocenjujemo, da je napaka pri uporabi vrednosti za debele materiale zanemarljiva. Na osnovi simulacij smo predstavljeno metodo izboljšali z dodatnimi stranskimi čipi, ki izboljšajo homogenost stresa na vzorcih. Vrednosti za piezoelektrični koeficient c/31, ki smo jih dolovili s predlagano metodo, se ujemajo s podatki iz literature. 1. Introduction When designing a new product or device, proper material selection is of basic importance. Material properties are also used in numerical analysis, when predicting device behavior. In case of piezoelectric microstructures, the properties of thin film piezoelectrics are influenced by chemical composition and other parameters of piezoelectric manufacturing process. It is thus important to have means for analyzing specific samples of piezoelectric thin films. Due to unique properties of piezoelectric effect, piezoelectrics are important materials in micro-electromechanical system (MEMS) technology, used for actuation or sensing, energy harvesting etc. Characteristics of piezoelectrics, especially piezoelectric coefficients d, play important role in device design, simulation and behavior prediction. In general, thin film materials used in microengineer-ing behave differently than bulk, thus requiring an adequate characterization of their core properties. Several methods are in use /1 / and new ones are being developed. Selec- tion of the appropriate material for a certain application requires comparison of different materials using datasheet specifying core information about these materials. Properties of piezoelectric materials vary with chemical composition, preparation technique e.g. sintering temperature and other effects. These influences present difficulties for comparison of different materials prepared by different methods, of different thicknesses and possibly from different producers, usually taking plain datasheet information from catalogue as a starting point. To overcome this obstacle a comparative method for characterization of different thin film piezoelectric samples bonded to a stainless steel cantilever is proposed. The relative response of different piezoelectric samples to the same mechanical stress gives immediate comparison of their basic properties such as sensitivity and linearity. Furthermore, coupling the measured results with numerical simulation based on finite element method (FEM) enables determination of absolute value for piezoelectric coefficient c/31 ■ 85 Informacije MIDEM 39(2009)2, str. 85-92 S. Penič, U. Aljančič, D. Resnik, D. Vrtačnik, M. Možek, S. Amon: Cantilever Method for Determination of D31 Coefficient in Thin ... The paper presents in detail the proposed technique for thin film piezoelectrics characterization and introduces a comparative method for simultaneous evaluation of multiple piezoelectric samples based on numerical simulation in combination with measured results. The result of this characterization is the absolute value of c/31 coefficient for multiple samples and comparison of piezoelectric response to mechanical stimulus. The method is practically tested on different thin film Lead Zirconate Titanate (PZT) chip samples prepared on silicon substrates. Measured results are matched with numerical simulation and piezoelectric coefficients are determined using ANSYS finite element analysis software. 2. Basic properties of piezoelectrics Piezoelectrics are materials that respond to the applied mechanical stress with electric voltage on the electrodes. This is called the direct piezoelectric effect, which serves as a basis for sensors and generators. The effect can be reversed and it is then called converse or inverse piezoelectric effect. Here mechanical strain is induced when voltage is applied. The response is dependent on the polarity of applied voltage and can therefore vary between elongation and contraction. Equations that describe electromechanical relations in a piezoelectric material are given in Voight notation with relations /2/ {7} = [c]{S}-[e]{£} {D}=[e]T {5}-[»]{£} 0) where {7} is stress tensor, {S} strain tensor, {£} electric field vector and {D} electric displacement vector. Material properties are described with stiffness matrix [c] which includes information about Young's modulus Y and Poisson ratio a of the material, with piezoelectric stress matrix [e] (superscript T denotes matrix transpose) related to piezoelectric strain matrix [c/] and with permittivity matrix [n]. Piezoelectric strain coefficients c/,/ and piezoelectric stress coefficients e,yare related with stiffness coefficients c,y by matrix equation [e]= [c] [d]. Piezoelectrics can be used for sensing or actuation, depending on whether the applied input load is mechanical or electrical, respectively. The two modes of operation can also be used interchangeably which makes piezoelectrics extremely versatile electromechanical materials since the same structure can act as a sensor or an actuator. Though the effect is reversible, certain considerations must be taken into account during the design of the structure /3/. 3. Piezoelectrics characterization 3.1 Bulk piezoelectrics characterization A complete characterization process of bulk piezoelectric material includes determination of stiffness coefficients cu (including Young's modulus Yand Poisson ratio a), permittivity (h//) and piezoelectric coefficients (c/,/). Most widely used method adopted as IEEE standard for piezoelectric characterization is the resonance method /4/. For such characterization, piezoelectric material is prepared as a flat rectangular plate between two electrodes, forming a capacitor. The capacitor impedance Z is measured at different frequencies. From Z(f) diagram, the resonant (fr) and anti-resonant (fa) frequencies are found. Then, the elastic compliance (inverse stiffness matrix) and piezoelectric coefficients for practical purposes usually 0(31 and c/33 can be derived /1 / . Direct methods for determining piezoelectric coefficients da include deformation measurements when voltage is applied to the electrodes. These methods are used to quantify the direct and converse piezoelectric effect. Direct methods are also used to investigate the behavior of the piezoelectric material in terms of hysteresis and nonlinear-ity, thermal behavior and aging. Mechanical deformation measurement of piezoelectric sample vs. applied voltage is used to determine piezoelectric coefficients d,y, calculated from relation in Voight notation Sy= dy Ei/1/. A different method for measuring piezoelectric coefficients da is based on direct piezoelectric effect. Here, sample is mechanically loaded, therefore the bounded electric charge becomes free, ready to flow out from the electrodes /5/. Electrodes are short circuited and electric displacement D is measured. Piezoelectric coefficient c/,y is here calculated from equation in Voight notation D/ = dy T, /1 /. In order to determine the relative permittivity nr, capacitance measurements are carried out at low frequency, usually 1 kHz and for low AC voltage excitation levels, ranging few mV /1 / . The relative dielectric constant is then calculated as Ct 121 where f is thickness of piezoelectric layer, A electrode area, C measured capacitance and ho permittivity of free space. 3.2 Thin film piezoelectrics characterization In general, the properties of thin film materials can differ significantly from its bulk counterparts. Therefore, adequate characterization of piezoelectric thin film properties is essential. Thin film characterization methods are usually based on similar principles as for bulk. The prevailing methods use converse piezoelectric effect where electrically excited thin piezoelectric film results in mechanical displacement, which is typically in the order of a few angstroms /6/. Sometimes the direct piezoelectric effect is used. Thin film piezoelectric together with electrodes are deposited on a substrate wafer and fixed in a rigid frame above pneumatic pressure cavity /7/. Pressure in the cavity is varied thus applying different mechanical stress to the piezoelec- 86 S. Penič, U. Aljančič, D. Resnik, D. Vrtačnik, M. Možek, S. Amon: Cantilever Method for Determination of D31 Coefficient in Thin ... Informacije MIDEM 39(2009)2, str. 85-92 trie layer. The charge integrator is used to measure the induced charge which is used in combination with excitation pressure to determine piezoelectric coefficients c/,y. For determining Young's modulus of thin piezoelectric films, several approaches exist. One of the possibilities to characterize mechanical thin film properties is presented in /8/. The experiment consists of loading a membrane with a line load applied to the middle of the span using nanoin-denter. A Mireau microscope interferometer is used to observe fringes that are formed on the loaded sample. Using a CCD camera these fringes are recorded and strains determined. From known stresses and strains in the material, Young's modulus can be determined. 3.3 Cantilever method for characterization of thin piezoelectric films In this case, characterization method is focused on piezoelectric coefficient c/37 using direct piezoelectric effect. In the proposed characterization method we introduce a cantilever with mounted piezoelectric samples on silicon substrate, with exact control of deflection. Mounting several samples simultaneously to the same cantilever provides us a comparison of piezoelectric responses of various piezoelectric materials to the same stimulus. This provides fast and accurate comparison of different piezoelectric materials appropriate for R&D work. When comparing responses of different materials, relative comparative method is usually sufficient and sometimes preferred to comparing absolute values due to its simplicity. However, determination of absolute values of piezoelectric coefficients is also possible, upgrading the proposed method with analysis of mechanical setup using appropriate numerical simulation as shown later. For this purpose, finite element analysis (FEA) software ANSYS was used. Mechanical properties of piezoelectric and silicon were taken from literature /9,10/. Permittivity was determined from capacitance measurements. 4. Experimental setup Experimental setup consisting of rectangular cross-sec-tion cantilever with mounted samples is shown in Fig. 1. Due to the simplicity of cantilever with rectangular cross-section, also analytical expressions for stress distribution exist, enabling comparison with numerical results. Proposed characterization method uses samples with thin film piezoelectric capacitor structure on silicon substrate, mounted on stainless steel cantilever. The selection of optimal samples placement is essential, usually selected for high sensitivity as the region of maximum stress distribution in the beam still having sufficient uniformity. Stress decreases in cantilever longitudinal direction towards the cantilever free end where it reaches zero. Therefore, the samples are mounted in the region of maximum stress being at the root of the cantilever. Following our simulation results, care must be taken not to induce an excessive error in the placement of samples. PZT samples" Guard chips iHHHHHH Fig. 1: Top view of the cantilever with mounted samples and side guards: (a) schematic, (b) photograph For adequate characterization of piezoelectric thin film samples, high repeatability of sample loading is essential. The testing cantilever setup, together with bonded samples represents such a test structure. Stainless steel was selected as the material for cantilever, providing possibility of high repeatable deflections. Furthermore, stainless steel cantilever is mechanically resistant and can be reused after replacing samples. During characterization, samples are often exposed to higher mechanical stresses as during the normal sensor or actuator operation. To achieve such a wide measurement range, cold rolled austenitic stainless steel (1.4310) was selected for the cantilever realization. This material has an extended elastic range due to a special treatment during the fabrication. In this case, the cantilever returns to its initial position even after extremely large deflections. To achieve large measured range of stresses for samples under test, the mechanical part of testing system has to provide adaptability. Therefore, 10 cm long and 18 mm wide stainless steel strips (cantilevers) of thickness 0.5 mm were cut by milling and then pressed between two rigid stainless steel plates acting as a fixed support. In this approach, the cantilever length is adjustable, resulting in increased measured range with high repeatability and accuracy. To illustrate the characterization of piezoelectric samples with described experimental setup, various thin PZT layers were deposited by sol-gel method on silicon chips covered by Pt/Ti as reported elsewhere /11/. Gold electrodes were placed on top of PZT layer by sputtering and shaped by shadow mask method. Thin Ti and Pt layers with thicknesses of 10 and 100 nm respectively are not significant for the overall mechanical properties of the relatively thick samples and were thus neglected in numerical simulations. 87 Informacije MIDEM 39(2009)2, str. 85-92 S. Penič, U. Aljančič, D. Resnik, D. Vrtačnik, M. Možek, S. Amon: Cantilever Method for Determination of D31 Coefficient in Thin ... As an example of the proposed characterization procedure, three samples with two different thicknesses of PZT layer were introduced, marked as samples PZT1a, PZT1b and PZT2. Due to our numerical simulations, two dummy guard chips were added at cantilever sides to achieve better stress uniformity over the samples (Fig. 1). To assure a reliable transfer of induced mechanical stress from the cantilever to the PZT samples, a strong and stable bond between the cantilever and the samples has to be achieved. Therefore, an epoxy adhesive (UHU endfest 3000) with high bonding strength of 3000 N/cm2 was used for PZT samples bonding. The extended elastic range of the selected stainless steel, in the combination with the mentioned adhesive enable highly reliable loads on testing samples, up to the silicon tensile strength. In addition, samples fixed with the selected adhesive can be easily removed at relativly low temperatures what makes the testing cantilever reusable /11/. for cantilever test structure with samples is built using AN-SYS proprietary scripting language APDL. Meshing is done using built-in automatic mesh generator. The resulting hex-ahedral mesh of simulated test structure is shown in Fig. 3. Local improvement of the mesh was done manually to refine mesh in structure critical regions such as thin PZT layer and to avoid badly shaped elements. The test structure basically consists of several different layers - stainless steel (SS) cantilever, silicon (Si) substrate chip, metal and PZT layer. Electrodes and interface layers were neglected at mechanical simulation due to their small thicknesses. For modeling SS and Si materials, three-di-mensional SOLID95 elements were used. PZT layer was modeled with SOLID226 elements with capability to couple mechanical and electrical quantities using piezoelectric effect. mm Fig. 2: Experimental setup: Taylor-Hobson traversing table and micromanipulator are used to achieve high deflection repeatability. To achieve highly repeatable stresses, testing cantilever with bonded samples is mounted on the fixed part of modified Taylor-Hobson 150mm Traversing Table, as shown in Fig. 2. The computer controlled worktable is motor driven in both directions, but can also be moved manually. Straight-ness accuracy of the worktable is within ±1 pm over the full 150mm range. In order to assure deflection repeatibil-ity, a micromanipulator with 8 mm tall pointed pin is mounted at the top of the worktable, as described in detail elsewhere /12/. Voltage response of PZT samples is measured by Semiconductor Parameter Analyzer HP4155A, including SMU and PMU Generator Expander HP41501A. For determination of piezoelectrics permittivity, capacitance on test capacitors is measured with HP4284A Precision LCR Meter at various frequencies, at excitation amplitude 1 V and DC bias 0 V. 5. Numerical modeling For the purpose of simulation, commercial FEM modeling and simulation software ANSYS was used. Simulator input Fig. 3: Generated mesh of cantilever with 3 bonded samples and two side guards. When we take into account material symmetry, general form of stiffness matrix [c] for ceramics, permittivity matrix [n] and piezoelectric coefficients matrix [d] can be simplified /1/. [c] = C11 ^12 cl3 0 0 0 c12 Cu C13 0 0 0 CI3 c, 3 C33 0 0 0 0 0 0 C44 0 0 0 0 0 0 C44 0 0 0 0 0 0 (cn ~ci \ 00 [n] = 0 ht 0 0 0 n3 (3) (4) [d] = 0 0 0 0 dl5 0 0 0 0 d15 0 0 i 31 d3i ¿33 0 0 0 (5) Due to the lack of exact information in the literature, mechanical properties of thin PZT layer were approximated by bulk values. Therefore, values cu = 13.91010 Pa, c33 = 11.51010 Pa, cM = 2.56-1010 Pa, c13 =7.43-1010 Pa, 88 S. Penič, U. Aljančič, D. Resnik, D. Vrtačnik, M. Možek, S. Amon: Cantilever Method for Determination of D31 Coefficient in Thin ... Informacije MIDEM 39(2009)2, str. 85-92 c12 = 7.78'1010 Pa were taken from literature /10/. Due to the small thickness of PZT compared to the cantilever and Si substrate, the error introduced is negligible. SS material is usually considered isotropic. The Young's modulus of SS material was measured using nanoindenta-tion method /13/. The measured value of SS Young's modulus is Y= 167.56 GPa. Silicon is very well known material. Due to Si crystal symmetry, it is described by 3 stiffness coefficients Cn, C12 and C44. In our case Si was modeled using anisotropic symmetric matrix with coefficients cn = 0.1657-106 Pa, c12 = 0.06 39-106 Pa, c+4 = 0.0796-106 Pa /14/. Due to the longitudinal stress dominating in our case as confirmed by our numerical simulation, only piezoelectric coefficient c/31 was taken into account. Boundary conditions for cantilever at FEM simulation were fixed support on the cantilever left side (deflection and its derivative equal to 0) and free deflection on the right side. To allow simple load variation, the deflection was described in the program as a parameter. Electrical ground boundary condition was set on the bottom electrode. Standard sparse direct linear solver was used for solving the model having 85000 elements with 4 basic variables (degrees of freedom) of the problem: electric potential and displacements in x, y, z direction. Sparse direct solver is a robust and fast solver for linear and nonlinear analysis, appropriate when poorly shaped elements are present in the model, such as the high aspect ratio (thickness vs. width) elements in the model of PZT layer. The sparse direct solver is based on a direct solution of equations by elimination, as opposed to iterative solvers where the solution is obtained through an iterative process that successively refines an initial guess to the final solution that is within a prescribed tolerance of the final solution. Direct elimination requires the factorization of an initial very sparse linear system of equations into a lower triangular matrix followed by forward and backward substitution. Drawback of this solver is that it requires a significant amount of memory, thus it is not suitable for larger scale models with more than a half million variables. Because sparse direct solver is based on direct elimination, poorly conditioned matrices do not pose difficulty in producing the solution /15/. Direct solver was chosen for our simulated approach since it does not exceed the recommended number of equations and there was enough computer memory available to perform computation. Simulations were performed on Intel Core Duo 6600 64-bit processor architecture with 4GB RAM memory, running at 2.4GHz. A single simulation run with chosen solver required typically 6 minutes. 6. Procedure for determination of piezoelectric coefficient d3i The described experimental setup was used to deflect cantilever. Corresponding voltage response of the mechanically loaded PZT samples was measured with parametric analyzer as described previously. Numerical simulator was configured as discussed in the previous section, to translate the test structure into numerical model. The characterization of piezoelectric effect and related c/37 parameter was performed by fitting the measured voltage response with simulated response: c/37 parameter value was varied in the simulator until a good match between measured and calculated voltage response was found. The value of c/37 that provided best fit throughout all deflections between calculated and measured voltage response was selected as the final result for the piezoelectric coefficient c/37. 7. Results and Discussion The PZT samples capacitance was measured using LCR meter at frequencies ranging from 20 Hz to 10 kHz, at excitation voltage of 10 mV. A relatively small dependence of capacitance vs. frequency was detected (Fig. 4). Measured capacitance value at 1 kHz was taken, as stated in /1 / . Top electrode area was measured under the microscope. PZT layer thickness was measured after the fabrication of the layer. From data given in Table 1 the relative permittivity nr for samples was calculated. Samples PZT1 a and PZT1 b are built on the same PZT layer differing only in their electrode position, regarding to the cantilever support (Fig. 1). Electrode of PZT1 a was located 2.7 mm from the support, while the electrode of PZT1 b was located 5.4 mm from the cantilever support. Sample PZT2 was prepared with modified processing for double thickness of PZT1, resulting in changed value of PZT permittivity. The electrode location for PZT2 was the same as for PZT1 a, 2.7 mm from the support. a. 4 05 O PZT 1a and PZT1b PZT2 Frequency [Hz] Fig. 4: Measured samples capacitance vs. frequency. 89 Informacije MIDEM 39(2009)2, str. 85-92 S. Penič, U. Aljančič, D. Resnik, D. Vrtačnik, M. Možek, S. Amon: Cantilever Method for Determination of D31 Coefficient in Thin ... Table 1: Measured sample parameters and calculated relative permittivity of PZT layers Sample PZT Thickness [nm] Electrodes [mm] Capacitance [nF] Rel. permittivity PZT la 740 0.87 7.67 737 PZT lb 740 0.87 7.67 737 PZT2 1554 0.87 2.40 484 Simulated stress profile in PZT layer is shown in Fig. 5 (simulation path is shown In the inset). From Fig. 5 can be concluded that electrode exact position is important when performing characterization of multiple samples. Following our numerical simulations results, to minimize the difference of stress profile in neighbor samples, two longer guard chips are added at the sides, as shown in Fig. 1. Calculated stress distribution in the cantilever and samples is shown in Fig. 6a. Position on Si chip [mm] Fig. 5: Simulated longitudinal stress profile in PZT layer vs. position on Si chip. The effect of guard chips is quantified in Fig. 6 and Table 2. As shown, the absolute stress in samples is decreased when guard chips are present. However, stress uniformity over the samples improves significantly. The relative difference in stress in both cases, without and with guard chips, was calculated between central and side samples. Guard chips thus provide more homogenous stress conditions on all samples. According to the piezoelectric effect, voltage response is proportional to the stress, what is described by piezoelectric coefficients. Calculated voltage response of PZT samples due to calculated stress is given in Fig. 6b. Measured time dependent voltage response of PZT samples during testing is shown in Fig. 7. Here, the cantilever ABC Fig. 6: Simulated longitudinal stress distribution in stainless steel cantilever and silicon chips (a) and corresponding voltage on top of PZT layer due to accumulated charge (b). Positions on the chips A, B and C show where stresses were compared. was deflected to predefined values using the micromanipulator as previously described. At start, the cantilever was first manually deflected over the desired deflection value, and then after this it was released to rest in final position determined by micromanipulator. Similar procedure was applied also during the end of testing. Consequently, voltage spikes always occurred at the start and at the end of loading. As also seen in Fig. 7, the response for constantly deflected cantilever slowly decreases with time, probably due to piezoelectric internal effects such as leakage and recombinations, and due to external effects such as input im- Table 2: Improvement of the stress uniformity over samples when guard chips are used. Stress in samples without guard chips ("MPa] Stress in sam pies with guard chips [MPa] Position Positions A, C Rel. difference Position Positions A, C Rel. difference 13.19 14.74 10.5 % 10.04 10.08 3.7 % 90 S. Penič, U. Aljančič, D. Resnik, D. Vrtačnik, M. Možek, S. Amon: Cantilever Method for Determination of D31 Coefficient in Thin ... Informacije MIDEM 39(2009)2, str. 85-92 pedance of HP4155A connected to the sample. Therefore, measurement of the response was done after the spike settled down, typically after 10 seconds. 0.25 0.20 0.15 0 05 0.00 -0.05 Fig. 7: Measured voltage response vs. time during testing Measured voltage response results are graphically displayed in Fig. 8. The response amplitude is dependent on electrode distance from the cantilever support and is in correlation with simulated stress profile in PZT layer shown in Fig. 5. The voltage on PZT1 a is thus considerably higher than voltage on PZT1 b. PZT2 that differs in thickness and material properties produces response that is slightly higher than with PZT1 a. 0.20 0.18 0.16 0.14 JT 0.12 4) o> 0.10 ro "o > 0.08 0.00 0.04 0.02 0.00 2 4 6 8 10 12 14 16 18 Deflection [mm] Fig. 8: Measured voltage response of PZT samples vs. deflection. Determination of piezoelectric coefficient c/37 was done by using numerical simulation as described previously. Successive simulations were performed for various values of coeffi- Table 4: Measured properties of PZT layer. Sample Relative permittivity - ñ Piezoelectric coef. -du TpC/Nl PZT la Til 66.1 PZT lb 737 -66.1 PZT2 484 -20.7 cient d3i until close agreement between simulated and measured voltage response was obtained. Measured and simulated responses at various deflections for best values of piezoelectric coefficient d3i are given in Table 3. The summary of measured values for relative permittivity rir and piezoelectric coefficient c/37 for PZT materials under test is given in Table 4. Results obtained are in reasonable agreement with available values from the literature /7/. Graphical representation of measured and simulated voltage responses vs. deflection for all three PZT samples are shown in Fig. 9. In the range of measured deflections, the simulated response displays linearity while it is slightly distorted for measured values, probably due to measurement error. 8. Conclusion Using the proposed cantilever method, piezoelectric coefficients c/37 for various thin film piezoelectrics were determined. Cantilever based characterization method provides a fast comparison of different piezoelectric material samples, since multiple samples can be mounted simultaneously on the testing structure. Furthermore, when combining experimental data with numerical simulation, piezoelectric coefficient c/37 can be determined by matching simulated results with voltage response measurements. Exact knowledge of geometry and material properties of cantilever and samples proved to be important in order to measure piezoelectric coefficients with sufficient accuracy. Stainless steel cantilever was adequately characterized by measuring its Young's modulus. Silicon properties are adequately determined by published data in the literature. Mechanical properties of PZT layers are on the other hand more difficult to acquire, since they are rather dependent on the actual PZT preparation procedure and composition. Nevertheless, we expect that error here introduced is small due to very thin PZT layer compared to stainless steel cantilever and silicon substrate. To improve the presented method, based on numerical simulation results guard chips were mounted at the side of the cantilever to PZT la PZT lb PZT2 Deflection Meas[mV] Sim [mV] Meas[mV] Sim [mV] Meas[mV] Sim [mV] 3.175 mm 35.5 35.24 24.9 27.80 41 37.38 5.715 mm 60.0 63.40 50.8 50.10 63.7 67.30 8.255 mm 95.4 91.62 76.9 72.43 92.5 97.19 10.795 mm 121 119.8 93.9 94.70 125 127.1 13.335 mm 144 148.0 114 117.0 158 157.1 15.875 mm 173 176.2 141 139.1 187 186.9 Voltage response of measured samples cantilever deflected cantilever undeflected PZT13 PZT2 10 20 30 40 Time [s] 50 60 Table 3: Measured voltage response and simulated values for different deflections of cantilever with best fit value for d3i parameter. 91 Informacije MIDEM 39(2009)2, str. 85-92 S. Penič, U. Aljančič, D. Resnik, D. Vrtačnik, M. Možek, S. Amon: Cantilever Method for Determination of D31 Coefficient in Thin ... 2 4 6 8 10 12 14 16 18 Deflection [mm] 2 4 6 8 10 12 14 18 18 Deflection [mm] Fig. 9: Graphical representation of measured and simulated voltage response of PZT samples vs. deflection. reduce stress variation over the samples. Determined values of piezoelectric coefficients d3i for PZT layers under test were in reasonable agreement with results available in the literature. Acknowledgment Authors would like to acknowledge Electronic Ceramics Department - K5, Jožef Stefan Institute, Slovenia for PZT samples preparation. This work was supported by Ministry of Higher Education, Science and Technology and Slovenian Research Agency. References /1/ T.L. Jordan, Z. Ounaies, "Piezoelectric Ceramics Characterization". NASA/ CR-2001-211225 ICASE report to NASA Langley Research Center. Report No. 2001-28, September 2001. /2/ Ansys Inc. "Ansys Inc. Theory reference". Ansys Inc., 2005. /3/ S. Penič, U. Aljančič, D. Vrtačnik, D. Resnik, M. Možek and S. Amon "Numerical modeling of PZT/SiC>2 microcantilever with in-terdigitated electrodes", Proc. 43rd International Conference on Microelectronics, Devices and Materials and the Workshop on Electronic Testing, Bled, Slovenia, September 2007, pp. 63-68. /4/ "IEEE Standard on Piezoelectricity, (IEEE Standard 176-1987), Institute of Electrical and Electronic Engineers, 345 East 47th St, New York, NY 10017. /5/ K. C. Kao, "Dielectric phenomena in solids", Elsevier Academic Press, San Diego, California, 2004. /6/ J.T. Dawley, G. Teowee, B.J.J, Zelinski and D.R. Uhlmann "Piezoelectric Characterization of Bulk and Thin Film Ferroelectric Materials using Fiber Optics". MTI Instruments application note, http://www.mtiinstruments.com/ /7/ J.F. Shepard Jr., P.J. Moses and S. Trolier-McKinstry "The wafer flexure technique for the determination of the transverse piezoelectric coefficient (d3i) of PZT thin films". Sensors and Actuators A, vol. 71, 1998, pp. 133-138. /8/ H.D. Espinosa , B.C. Prorok, M. Fischer "A methodology for determining mechanical properties of freestanding thin films and MEMS materials". Journal of the Mechanics and Physics of Solids, vol. 51, 2003, pp. 47-67. /9/ Efunda, http://www.efunda.com/ /10/ X.J. Zheng, Y. C. Zhou and J.Y. Li "Nano-indentation fracture test of Pb(Zro.52Tio.48)03 ferroelectric thin films". Acta materialia, vol. 51, 2003, pp. 3985-3997. /11/ U. Aljancic, B. Malic, M. Mandeljc, M. Vukadinovic, D. Vrtacnik, D. Resnik, M. Mozek, M.Kosec, S. Amon "Cantilever as Testing Structure for Characterization of PZT Thin Films on Pt/Si Substrates". Proc. 42nd International Conference on Microelectronics, Devices and Materials and the Workshop on MEMS and NEMS, Strunjan, Slovenia, September 2006, pp. 271-276. /12/ U. Aljancic, M. Vukadinovic, D. Resnik, D. Vrtacnik, M. Mozek, S. Peniv, S. Amon "Cantilever Characterization Method for Static Behavior of PZT Thin Films". Proc. 43rd International Conference on Microelectronics, Devices and Materials and the Workshop on Electronic Testing, Bled, Slovenia, September 2007, pp. 115-120. /13/ S. Penic, U. Aljancic, D. Vrtacnik, D. Resnik, M. Mozek, M. Makovec, R. Bosnjak and S. Amon "FEM modeling of piezore-sistive force sensor for medical retractor and design verification". Proc. 6th EUROSIM Congress on Modelling and Simulation, Ljubljana, Slovenia, September 2007, p. 158. /14/ A. M. Fitzgerald "Practical Issues in Finite Element Analysis of MEMS". Ansys Workshop, March 2006. /15/ Ansys Inc. "Ansys Inc. Basic Analysis Guide". Ansys Inc., 2005. Samo Penič, univ. dipl. inž. el. mag. Uroš Aljančič doc. dr. Drago Resnik doc. dr. Danilo Vrtačnik mag. Matej Možek prof.dr. Slavko Amon University of Ljubljana, Faculty of Electrical Engineering, Laboratory of Microsensor Structures and Electronics Trzaska 25, Ljubljana 1000, SLOVENIA e-mail: matej.mozek@fe.uni-lj.si Telefon: 01 4768 303, Telefax: 01 4264 630 Prispelo (Arrived): 12.08.2008 Sprejeto (Accepted): 09.06.2009 92 UDK621,3:(53+54+621 +66), ISSN0352-9045 Informacije MIDEM 39(2009)2, Ljubljana MODEL DETEKTORJA KAOTIČNOSTI Matej Šalamon, Tomaž Dogša Univerza v Mariboru, Fakulteta za elektrotehniko računalništvo in informatiko, Maribor, Slovenija Kjučne besede: kaotična vezja, bifurkacijski diagram, modeliranje, simulatorji električnih vezij. Izvleček: Kaotično vedenje elektronskih vezij je mogoče analizirati tudi s pomočjo simulatorjev analognih vezij. Začetek kaotičnega režima delovanja vezja lahko ocenimo s pomočjo univerzalne Feigenbaumove konstante in bifurkacijskih točk, ki jih odčitamo iz bifurkacijskega diagrama. V prispevku je predstavljen model detektorja kaotičnosti, ki omogoča direktno tvorjenje bifurkacijskega diagrama s samim simulatorjem SPICE /6/. Brez tega bi sicer potrebovali dodatni program, ki bi omogočal avtomatsko izvajanje simulacij vezja pri različnih vrednostih bifurkacijskega parametra ter shranjevanje ekstremnih vrednosti signalov. The Model of Chaoticness Detector Keywords: chaotic circuits, bifurcation diagram, modelling, circuit simulators. Abstract: Almost all circuits under specific circumstances (choice of parameters, initial conditions, input signals etc.) can become chaotic /1 /. Chaotic circuits can be used in the cryptography or as a test that shows simulators non-repeatability /8/. In general, chaotic behaviour of the electronic circuits is not tolerated. Simulation experiments allow us to determine chaotic responses for various sets of parameters, calculation of Lyapunov exponents, entropy, construction of Poincare planes, construction of bifurcation diagrams etc. The determination of the chaotic regions requires two steps. First one is the manual construction of a bifurcation diagram which is a tedious work. While carefully changing the bifurcation parameter and running the simulation we must find peak values in the selected output signal. The bifurcation points are values of the bifurcation parameter where a qualitative change in circuit dynamics is noticed. In the bifurcation diagram we can observe bifurcation sequences or period-doublings that construct a tree (Fig. 1). If a period-doubling bifurcation from period-1 to period-2 occurs at a bifurcation parameter value a r and from period-2 to period-4 at a parameter value 32, then we can make a prediction of the parameter value a» where chaos should appear (eq. (1)). However, we can not expect that this prediction is totally accurate. Firstly, because it is based on numbers a j and a2 determined by the simulation and secondly, we have assumed that all measured bifurcation ratios are described by the Feigenbaum's constant. Nevertheless, after the bifurcation diagram is constructed it is easy to calculate the parameter value a« which is a reasonable estimation of the region where chaos begins. This simple calculation is the second step toward the determination of the chaotic region. Our goal was to simplify the construction of the bifurcation diagram by using the circuit simulator. The process of generating the bifurcation diagram was modelled by the special circuit that was called the chaoticness detector. With this bifurcation points based model we can easy detect the presence of circuit's chaotic behaviour. This paper is structured as follows. In the second chapter the description of the bifurcation diagram and its use is presented. In the third chapter we describe in the detail the model of the chaoticness detector which comprises analogue memory where the peak values (Fig. 3) are stored and the model of time varying bifurcation parameter. The basic components of the analogue memory model are two capacitors and switches. The switches ensure that within each signal's period the capacitors store the maximum and minimum values of the signal. In the fourth chapter we present the chaoticness detection of the Chua's circuit /9/. While the simulation runs, bifurcation parameter alters in time from the selected start to stop values. Signals peak values are kept in the peak values memory and with appropriate time step they are also stored in the output file. Based on these stored data we can easy plot a bifurcation diagram (Fig. 6) that enables a parameter a„ calculation. The chaoticness detector can be used as an effective tool for analysis of the chaotic behaviour on the level of circuit simulators. The concept of the detector can also be used as a starting point for the practical implementation. V prispevku smo se osredotočili na detekcijo kaotičnosti elektronskih vezij s pomočjo bifurkacijskega diagrama. Ker obnašanje analognih vezij najpogosteje analiziramo s pomočjo simulatorjev analognih vezij /6/, smo izdelali model detektorja kaotičnosti, ki je primeren za uporabo v tovrstnih simulatorjih. Namenjen je izračunu ocene mejne vrednosti bifurkacijskega parametra, pri kateri se običajno obnašanje vezja spremeni v kaotično. Delovanje modela temelji na načinu detektiranja kaotičnosti s pomočjo bifurkacijskih točk. Ta način je opisan v drugem poglavju. Bifurkacijske točke odčitamo iz bifurkacijskega diagrama, ki ga dobimo na osnovi analize, vvezju izbranega signala. Iz njega izločimo vse lokalne ekstremne vrednosti (minimume ali maksimume) in jih, v odvisnosti od spremenljivega bifurkacijskega parametra, grafično prikažemo v bi-furkacijskem diagramu. Ker simulatorji ne omogočajo iskan- 1 Uvod Odkritje determinističnega kaosa v sredini šestdesetih let je pritegnilo pozornost številnih raziskav različnih znanstvenih področij: biologije, meteorologije, fizike, astrologije, kemije, medicine, elektronike, kriptologije. Analize različnih nelinearnih sistemov potrjujejo obstoj fenomena kaosa z značilnimi univerzalnimi zakonitostmi, lastnostmi in pojavi /2/, /3/. Eden takšnih pojavov je pojav bifurkac-ij, katerih število postane v primeru kaotičnega obnašanja neskončno veliko. Kaotičnost je mogoče zaznati /1/, /2/, /4/, /5/, /6/, / 7/ npr. z opazovanjem trajektorij v prostoru stanj, z različnimi analizami časovnih potekov signalov v vezju, s Poincar-ejevimi prerezi atraktorjev, z izračuni Lyapunovih eksponentov, z entropijo, z analizo bifurkacijskih diagramov itd. 93 Informacije MIDEM 39(2009)2, str. 93-99 M. Šalamon, T. Dogša: Model detektorja kaotičnosti ja lokalnih ekstremnih vrednosti, je za to potrebno uporabiti še dodatna matematična orodja /8/, /7/, ki omogočijo naknadno analizo signalov, dobljenih z več predhodno izvedenimi simulacijami vezja, pri različnih vrednostih bi-furkacijskega parametra. Dodatnim programom in naknadnim analizam se lahko izognemo z uporabo modela detektorja kaotičnosti, s katerim je mogoče bifurkacijski diagram tvoriti s samim simulatorjem. Model detektorja kaotičnosti, ki je opisan v tretjem poglavju, sestavljata model časovno spremenljivega bifurkacijskega parametra in model analognega pomnilnika ekstremnih vrednosti. V četrtem poglavju je predstavljen konkretni primer detek-cije kaotičnosti Chujevega oscilatorja /9/. 2 Detektiranje kaotičnosti Kvalitativne spremembe v obnašanju nekega vezja, ki nastopijo zaradi spremembe njegovih parametrov, imenujemo bifurkacije. Parameter, s katerim je na te spremembe mogoče vplivati, imenujemo bifurkacijski parameter. Vrednosti bifurkacijskega parametra, pri katerih te spremembe nastopijo, imenujemo bifurkacijske točke. Bifurkacij je več vrst /1 /. Pri vezjih višjega reda so bifurkacije lahko zelo kompleksne. Različne vrste bifurkacij lahko vodijo v kaotično obnašanje vezja, če se le-te zaradi spremembe bifurkacijskega parametra pričnejo ponavljati. Prehod bifurkacij v kaotični režim delovanja je mogoče zelo nazorno prikazati s pomočjo bifurkacijskega drevesa oziroma bifurkacijskega diagrama. Zgled takšnega diagrama prikazuje slika 1. Vrednosti lokalnih ekslremov spremenljivke stanja «; «!\v\ bifurkacijski "t ryirnmi'ler Slika 1: Bifurkacijski diagram in bifurkacijske točke ai, a2— a-. Fig. 1: The bifurcation diagram and the bifurcation points ai, a2... a». Bifurkacijski diagram predstavlja odvisnost lokalnih ekstremnih vrednosti izbrane spremenljivke stanj od vrednosti bifurkacijskega parametra. Vrednosti bifurkacijskega parametra, pri katerih nastopijo kvalitativne spremembe obnašanja vezja, so označene kot bifurkacijske točke: a?, a2...a». Pri vrednostih bifurkacijskega parametra manjših od a?, se vezje obnaša kot običajen oscilator, kar kaže zvezno naraščajoč potek ekstremnih vrednosti. Spremenljivka stanja zavzame v tem primeru le eno ekstremno vrednost, kar pomeni, da vezje oscilira s konstantno amplitudo. Pri bifurkacijskem parametru, večjem od a?, nastopi prva cepitev - bifurkacija. Vezje se obnaša kot oscilator, katerega izhodni signal je sicer periodičen, znotraj periode pa sta prisotni dve različni amplitudi oziroma lokalna ekstrema. Naslednje cepitve, ki se pojavijo v točkah a2, 33, a4 itd., so vse pogostejše, kar v praksi otežuje njihovo natančno določitev. Neperiodično, kaotično obnašanje vezja nastopi šele v točki a«, ko postane število različnih lokalnih ekstremov neskončno. Ker je to točko z meritvijo ali simulacijo nemogoče natančno določiti, jo lahko na osnovi univerzalne konstante a ter poznavanja točk a? in a2, le ocenimo /10/: ¿H (1) 8 je Feigenbaumova konstanta oziroma limita vrste, ki jo tvorijo razmerja bifurkacijskih točk: :(«2-«i> 5 slima* a*"' =4,6692016... (2) Gre za univerzalno in eksperimentalno določeno konstanto, ki se pojavlja v vseh kaotičnih sistemih. Ker so bifurkacijske točke zmeraj povezane s to konstanto, lahko na osnovi poznavanja vsaj dveh ocenimo vrednost bifurkacijskega parametra ain tako ocenimo začetek kaotičnega vedenja vezja. Detekcija kaotičnega režima delovanja vezja je torej povezana z iskanjem in določitvijo lokalnih ekstremnih vrednosti izbranega signala vezja, določitvijo bifurkacijskega diagrama ter z oceno bifurkacijske točke a». 3 Model detektorja kaotičnosti analognih elektronskih vezij Kaotično obnašanje vezij lahko analiziramo s pomočjo simulatorjev analognih vezij, v povezavi z dodatnimi matematičnimi orodji, ki omogočajo dodatno, naknadno obdelavo rezultatov simulacij. Tej se lahko izognemo z uporabo dodatnih modelov za simulator, ki omogočajo neposredno obdelavo vmesnih rezultatov simulacije. Tovrstni modeli predstavljajo vezja, ki se lahko priključijo ali dodajo k vezju, katerega obnašanje želimo analizirati. Običajno vsebujejo poleg osnovnih gradnikov (uporov, kondenzatorjev, tuljav, napetostnih in tokovnih virov...) še elemente, krmiljene na osnovi različnih matematičnih izrazov ali ukazov posebnega skriptnega jezika1 /11/. 1 Npr. ICL - (Interactive Command Language) skriptni jezik simulatorja lsSpice4. 94 M. Šalamon, T. Dogša: Model detektorja kaotičnosti Informacije MIDEM 39(2009)2, str, 93-99 Prispevek predstavlja model detektorja kaotičnosti, ki omogoča direktno tvorjenje bifurkacijskega diagrama s samim simulatorjem. Brez tega bi sicer potrebovali dodatni program, ki bi omogočal avtomatsko izvajanje simulacij vezja pri različnih vrednostih bifurkacijskega parametra ter detekcijo in določitev ekstremnih vrednosti v simuliranih časovnih potekih. Model detektorja sestavljata: model časovno spremenljivega bifurkacijskega parametra in model analognega pomnilnika lokalnih ekstremnih vrednosti. Bifurkacijski parameter je lahko v splošnem poljuben parameter vezja, npr.: upornost, kapacitivnost, ojačenje, temperatura, napetost. Njegovo spreminjanje v izbranem intervalu lahko zagotovimo s pomočjo spremenljive napetosti ali toka. Izvedba modela časovno spremenljivega bifurkacijskega parametra je odvisna predvsem od vrste samega parametra in možnosti deklaracije njegovih vrednosti. Zaradi tega je potrebno model spremenljivega bifurkacijskega parametra prilagoditi obravnavanemu kaotičnem vezju in ga ni mogoče posplošiti. V četrtem poglavju je opisan primer modela časovno spremenljive upornosti, uporabljene kot bifurkacijski parameter Chujevega oscilatorja. Ker simulatorji analognih vezij ne vsebujejo modelov, ki bi omogočali iskanje in shranjevanje posameznih ekstremnih vrednosti signalov, smo za model detektorja kaotičnosti potrebovali model analognega pomnilnika lokalnih ekstremnih vrednosti. 3.1 Model analognega pomnilnika ekstremnih vrednosti signala Analogni pomnilnik ekstremnih vrednosti vhodnega signala uVh(t) omogoča začasno shranjevanje njegovih ekstremnih vrednosti. Ekstremna vrednost signala se najprej de-tektira in nato zadrži na izhodu vezja Uizh(t) vse do naslednje, nove, ekstremne vrednosti. Način pomnjenja ekstremnih vrednosti temelji na polnjenju in praznjenju kondenzatorja C, z napetostjo vhodnega signala uVh(t), preko stikala S, kar prikazuje slika 2. Stikalo S, ki je krmiljeno s krmilno napetostjo Ukrmiina(t), se mora vklopiti tik pred nastopom ekstremne vrednosti, izklopiti pa v trenutku, ko ekstremna vrednost nastopi. Trenutek vklopa stikala in časovna konstanta polnjenja kondenzatorja, določena z vklopno upornostjo stikala Ron in ka-pacitivnostjo kondenzatorja C, morata zagotoviti, da se kondenzator v času vklopa napolni na napetost, čim bližjo ekstremni. S "'"A "T"/" \ \ \ I I-/-" "'*('> 4= C v v v \/ / < t j v o-1- Slika 2: Način delovanja pomnilnika ekstremnih vrednosti. uvh(t) in uIZh(t) - časovna poteka vhodnega in izhodnega signala. Fig. 2: A signal peak values storage method. uVh(t) and Uizh(t) are an input and an output signals. V trenutku prisotnosti ekstremne vrednosti se stikalo izklopi. S tem se povezava med vhodom in kondenzatorjem prekine, kar povzroči praznjenje kondenzatorja preko izk-lopne upornosti stikala Roff. če je ta dovolj velika, se vred-nost napetosti na kondenzatorju, do naslednjega ponovnega vklopa stikala, bistveno ne spremeni. V takšnem primeru predpostavljamo, da se v času izklopljenega stikala na kondenzatorju ohranja konstantna, ekstremna vrednost vhodnega signala. Kadar želimo shranjevati tako minimalne kot maksimalne ekstremne vrednosti, moramo uporabiti dva kondenzatorja in dve stikali, ki ju je potrebno ločeno krmiliti. Model analognega pomnilnika lokalnih ekstremnih vrednosti vhodnega signala, primeren za simulator2 SPICE, prikazuje slika 3. Na vhod (in) modela je priključen vhodni signal, iz katerega se izločijo njegove lokalne ekstremne vrednosti. Minimalne se ohranijo na izhodu npeak, maksimalne pa na izhodu peak. Slika 3: Model analognega pomnilnika ekstremnih vrednosti signala, primeren za simulator lsSpice4. Fig. 3: The lsSpice4 model of the peak values analog memory. Kondenzator C1 služi začasnemu shranjevanju minimalnih vrednosti, kondenzator C2 pa maksimalnih. Z ustreznim krmiljenjem stikal X1 in X2 zagotovimo, da se oba kondenzatorja pravočasno napolnita in zadržita ekstremni vrednosti. Krmiljenje stikal se izvaja na osnovi izračunanih trenutnih vrednosti prvega in drugega odvoda vhodnega signala. Oglejmo si podrobnosti. Ekstremne vrednosti vhodne napetosti nastopijo zmeraj v trenutkih, ko je vrednost prvega odvoda vhodnega signala 2 lsSpice4 verzija 8.11. 95 Informacije MIDEM 39(2009)2, str, 93-99 M. Šalamon, T. Dogša: Model detektorja kaotičnosti enak nič, detekcija lokalnih minimumov oziroma maksimumov pa zahteva še poznavanje vrednosti drugega odvoda vhodnega signala. Obstoj lokalnega minimuma je pogojen s pozitivno vrednostjo drugega odvoda, lokalnega maksimuma pa z negativno vrednostjo. Izračun trenutnih vrednosti prvega in drugega odvoda vhodnega signala omogočata diferenciatorja A2 (prvi odvod) in A3 (drugi odvod). Zaradi lažjega razumevanja krmiljenja stikal v nadaljevanju predpostavimo, daje vhodni signal uVh(t) kosinusne oblike, s časovno spremenljivo amplltudo/Vt) in frekvenco f(t): uvh(t) = A(t}cos(2-K-f(t}t) Njegov prvi odvod opisuje izraz: (3) -2-7l-A(t) duvh(t) _ dAit) dt dt df(t) cos(2 • K ■ f(t)-t)~ dt i+ /(j0)-sin(2-n ■/(*}*) (4) Če predpostavimo, da so časovne spremembe amplitud in frekvence dovolj majhne, se izraz (4) lahko poenostavi: ^~-2-k-A(t)-f(t)-sm(2-k-mt) (5) dt Vidimo, da je amplituda časovnega poteka odvoda v takšnem primeru kar sorazmerna produktu trenutnih vrednosti frekvence in amplitude vhodnega signala. Če želimo stikali X1 in X2 krmiliti z vrednostmi odvodov, je potrebno te predhodno normirati. Normiramo jih s predvideno maksimalno vrednostjo frekvence fmax in amplitude Anax vhodnega signala: (2-n ■ A^ ■ /max). Na podoben način izvedemo še normiranje drugega odvoda. Normiranje prvega in drugega odvoda izvedeta napetostno krmiljena vira B4 oziroma B5. Stikali X1 in X2 sta neposredno krmiljeni s krmilnima napetostma v vozliščih v(vpisimax) in v(vpisimin), ki jih dajeta krmiljena vira B1 in B2. Vklopljeni sta, kadar sta krmilni napetosti pozitivni, izklopljeni kadar sta enaki nič. Upornost stikal ob vklopu Ron je obratno sorazmerna vrednosti krmilne napetosti. Če izberemo, daje vrednost krmilne napetosti ob vklopu 1/Ron in, da je Ron=1£2, je časovna konstanta polnjenja kondenzatorja: ■C (6) določena samo z vrednostjo kondenzatorja C. t on —r-on Napetost krmiljenega vira B1 skrbi za vklop in izklop stikala X1 oziroma določa interval polnjenja kondenzatorja C1 ter interval zadrževanja minimalne vrednosti na izhodu npeak. Podobno funkcijo ima krmiljen vir B2, s katerim je krmiljen vklop in izklop stikala X2 oziroma interval polnjenja kondenzatorja C2 ter interval zadrževanja maksimalne vrednosti na izhodu peak. Stikali se vklopita, če so izpolnjeni določeni pogoji. X1 je vklopljeno kadar je vrednost drfigega odvoda pozitivna in hkrati vrednost prvega odvoda negativna in večja od negativne referenčne napetosti Von. Referenčno napetost zagotavlja napetostni vir B3 in je prisotna v vozlišču v(VON). Podobno velja za krmiljenje stikala X2 preko katerega se polni kondenzator C2. Vklopljeno je samo, če je vrednost drugega odvoda negativna in hkrati vrednost prvega odvoda pozitivna ter manjša od referenčne napetosti Von- Referenčna napetost predstavlja tisto vrednost prvega odvoda, pri kateri se mora stikalo vklopiti, da se bo kondenzator v času "f napolnil na ekstremno vrednost napetosti vhodnega signala z dopustnim odstopanjem ±Au. Če zahtevamo, da se pri maksimalni frekvenci vhodnega signala in pri spremembi ekstremne vrednosti iz OV na 1V zadržana napetost na kondenzatorju ne sme razlikovati od prave ekstremne vrednosti za več kot ±0,001%, lahko izračunamo potreben čas polnjenja kondenzatorja C: At ■ a cos (l-Am) _ a cos (0,999) _ 2,56 360 -f ~ 360 -f ~ 360- f J max J max J m.i (7) Sedaj lahko izračunamo še referenčno napetost Von pri kateri se mora stikalo vklopiti, da bo čas polnjenja kondenzatorja enak A t. Ker se kondenzator prične polniti zmeraj za čas Af pred nastopom ekstremne vrednosti, ko je vrednost odvoda enaka nič, lahko ob upoštevanju Izraza (7) in normiranega prvega odvoda, izračunamo vrednost referenčne napetosti: Vm = sin(2-re • /max -t0N)= srn 1 2-fr. --A t \ J max J J ■ srn ^ 2 - flcosil -Am)Y) rr :■ 1--i-1 = 54,02m F l 360 JJ (8) Vidimo, da je ob zgornjih zahtevah referenčna napetost konstantna oziroma neodvisna od frekvence vhodnega signala. Zahtevajmo še, da se kondenzator napolni na 99,9% končne vrednosti v času 6,9-ton- če upoštevamo, dajeta čas enak času Ai in, dajef?cw=1£2, lahko izračunamo vrednost kondenzatorja C=C1=C2 pri kateri bo, pri maksimalni frekvenci vhodnega signala, odstopanje zadržanih ek-stremnih vrednosti od dejanskih, manjša od Au=0,999V: C = - A t acos (1-Au) 1 (9) 6,9-Ron 6,9-360 -fmm-R0N 969,3 Pri frekvencah nižjih od fmax bo omenjeno odstopanje kvečjemu manjše od ±0,001%. Kakor hitro pogoja za vklop stikal nista izpolnjena, stikali izklopita in s tem prekineta povezavo med vhodom in kondenzatorjema. V času izklopljenih stikal se na kondenzatorjih ohranita trenutni ekstremni vrednosti vhodnega signala. Ti se bistveno ne spremenita, če izberemo dovolj veliko izklopno upornost (Roff) stikal oziroma dovolj veliko izklopno časovno konstanto: 96 M. Šalamon, T. Dogša: Model detektorja kaotičnostl Informacije MIDEM 39(2009)2, str. 93-99 ^off — roff 'c (10) Natančnost opisanega modela pomnilnika je odvisna od dopustnega odstopanja zadržanih ekstremnih vrednosti, izbranih vrednosti časovnih konstant ob vklopu in izklopu stikal, dopustnega maksimalnega časa integriranja3 in izbranega časa trajanja simulacije. Večja natančnost je pogojena z dlje trajajočo simulacijo. 4 Rezultati detekcije kaotičnosti Chujevega oscilatorja V številnih znanstvenih prispevkih zasledimo različna vezja, ki se lahko obnašajo kaotično. Gre za preprosta RLC-vezja, raznovrstne oscilatorje, kapacitivno" preklopna vezja, digitalne filtre, flip-flope, adaptivna sita, napajalnike in pretvornike, močnostna vezja/1/. Najpogosteje obravnavano kaotično vezje je prav gotovo Chujev oscilator /9/ (slika 4), ki smo ga v prispevku obravnavali kot vzorčno vezje v katerem smo, s pomočjo modela detektorja kaotičnosti, detektirali mejo med periodičnim in kaotičnim osciliranjem. Detekcijo kaotičnosti smo izvedli na osnovi analize časovnega poteka napetosti4 na kondenzatorju C1. Pri tem smo uporabili simulator lsSpice4, verzijo 8.11 /11/. Za bi-furkacijski parameter smo izbrali5 upornost R2, ki smo jo spreminjali v intervalu med 1800£KR2<1860iž, s čimer R2 VI V2 Slika 4: Chujev oscilator /9/. Fig. 4: The Chua's oscillator/9/. 3 Parameter tmax pri časovni analizi. 4 Enake rezultate bi dobili z analizo napetosti na kondenzatorju C2 ali toka skozi tuljavo L1. 5 Bifurkacijski parameter je lahko tudi kapacitivnost kondenzatorja C2. 97 Model detektorja kaotičnosti Slika 5: Model detektorja kaotičnosti v povezavi s Chujevim oscilatorjem. Fig. 5; The Chua's oscillator wired to the model of chaoticness detector. smo dosegli vse pomembne kvalitativne spremembe v obnašanju vezja: običajno harmonično osciliranje vezja se je preko bifurkacij spremenilo v kaotično. Mejno vrednost bifurkacijskega parametra, pri kateri se nekaotični režim delovanja vezja spremeni v kaotičnega, smo detektirali glede na enačbo (1), na osnovi prvih dveh bifurkacijskih točk v bifurkacijskem diagramu. Slednjega smo določili s simulacijo vezja, prikazanega na sliki 5. Ta Informacije MIDEM 39(2009)2, str, 93-99 M. Šalamon, T. Dogša: Model detektorja kaotičnosti slika prikazuje povezavo Chujevega oscilatorja z modelom detektorja kaotičnosti, ki ga sestavljata časovno spremenljiva upornost R2 in model analognega pomnilnika lokalnih ekstremnih vrednosti. Časovno spremenljivo upornost smo modelirali s pomočjo časovno odvisnega napetostnega vira V1 in upornosti R2. Napetost vira V1 oziroma napetost v vozlišču v(Rsp) smo spreminjali linearno od vrednosti 1860 do 1800. Padajočo napetost smo povezali z upornostjo R2 tako, da smo vrednost upora R2 opisali z izrazom: r=v(Rsp)+1 p. Glede na osnovno frekvenco oscilatorja (f=3751Hz), določeno z vrednostjo tuljave L1 in kondenzatorja C1 ter pričakovano maksimalno amplitudo signala v opazovanem vozlišču (Amax=1V), smo izračunali potrebne parametre modela pomnilnika ekstremnih vrednosti. Pri izbrani maksimalni frekvenci opazovanega signala /max=3,8kHz, smo po enačbi (9) izračunali še vrednosti kondenzatorjev C1 in C2: C1 =C2=C=271.5nF. Za vrednosti vklopne in izklopne upornosti stikal smo izbrali: R0n=1O, R0ff=1TiX Izvedli smo časovno analizo (TRAN), z naslednjimi vrednostmi parametrov: isfep=1/3751, tstart=50-tstep, tmax=6,Q RoN-C/Q, tstop=4s. Z izbrano vrednostjo parametra tstep smo dosegli, da so se v izhodno datoteko shranjevale ekstremne vrednosti le enkrat na periodo opazovanega signala. S tem smo onemogočili nepotrebno shranjevanje velikega števila enakih ekstremnih vrednosti znotraj periode. S parametrom tstart smo določili čas pričetka shranjevanja ekstremnih vrednosti v izhodno datoteko in s tem izločili nepotrebno informacijo o začetnem prehodnem pojavu v vezju. Parameter tmax, ki določa maksimalni dopustni korak integriranja, smo izbrali glede na najmanjšo časovno konstanto v vezju. V našem primeru je to časovna konstanta polnjenja kondenzatorjev v modelu pomnilnika ekstremnih vrednosti, ki znaša 1,87|js. Končni čas simulacije tstop smo izbrali tako, da se je v času simulacije, pri spremenljivi vrednosti upora R2, v izhodno datoteko shranilo dovolj ekstremnih vrednosti, potrebnih za določitev bifurkacijskega diagrama. Z izbranim časom tstop=4s smo tako dosegli, da se je upornost R2 vsako periodo zmanjšala za 4mQ, kar predstavlja tudi izbran korak spreminjanja bifurkacijskega parametra. Rezultat izvedene simulacije sta bifurkacijska diagrama, prikazana na sliki 6. Pri bifurkacijskem diagramu, označenim s številko 1, predstavljajo vrednosti na ordinatni osi vrednosti minimalnih napetosti, pri diagramu označenem s številko 2 pa vrednosti maksimalnih napetosti na kondenzatorju C1. Vrednosti na abscisni osi predstavljajo vrednosti bifurkacijskega parametra oziroma upornost upora R2. 1.30 1 10 -840111 S v; 900,-n * -1.21 a $ < 700m -1.64 5C0m -2.04 Slika 6: Bifurkacijska diagrama Chujevega oscilatorja, dobljena s simulatorjem isSpice4. Fig. 6: The bifurcation diagrams of the Chua's oscillator, simulated with the lsSpice4 simulator. Bifurkacijske točke, ki jih lahko odčitamo iz enega ali drugega bifurkacijskega diagrama, so zapisane v tabeli 1. Table 1: The bifurcation points of the Chua's oscillator. Bifurkacijska točka Vrednost upora R2 Vrsta trajektorije ai 1851 25 O Pojav limitnega cikla s periodo 2 a2 1835,79 Q. Pojav limitnega ci kla s periodo 4 as 1832 50 Q, Pojav limitnega cikla s periodo 8 Natančnost položaja posameznih točk bifurkacijskega diagrama je odvisna predvsem od izbrane natančnosti merjenja lokalnih ekstremnih vrednosti, izbranega koraka bifurkacijskega parametra in maksimalnega časa integriranja. Z ozirom na prej izbrane zahteve pričakujemo, da se bifurkacijska točka pojavi kakor hitro se sosednji ekstrem-ni vrednosti amplitud razlikujeta za več kot ±0,001%. S pomočjo dobljenih bifurkacijskih točk in enačbe (1) smo ocenili, da bo osciliranje Chujevega oscilatorja kaotično, kakor hitro bo vrednost upornosti R2 manjša od 1831,57 5 Sklep Pri približevanju bifurkacijskega parametra k vrednosti, ki ločuje običajno in deterministično naključno - kaotično vedenje, postane vezje izredno občutljivo. Zaradi hiper-občutljivosti te mejne vrednosti natančno ni mogoče predvideti. S pomočjo bifurkacijskega diagrama jo lahko le ocenimo. V prispevku smo predstavili model detektorja kaotičnosti, ki omogoča direktno tvorjenje bifurkacijskega diagrama s O VNPEAK 0 VPEAK 98 M. Šalamon, T. Dogša: Model detektorja kaotičnosti Informacije MIDEM 39(2009)2, str, 93-99 pomočjo samega simulatorja SPICE. Tvorjenje bifurkaci-jskega diagrama bi sicer zahtevalo uporabo dodatnega orodja, ki bi omogočalo avtomatsko izvajanje simulacij vezja pri različnih vrednostih bifurkacljskega parametra ter shranjevanje ekstremnih vrednosti signalov. Pokazali smo, da lahko z ustreznimi parametri modela detektorja in časovne analize dosežemo 0,001 % natančnost izračuna bifurkacijskih točk oziroma mejne vrednosti bi-furkacijskega parametra, ki ločuje običajni režim delovanja vezja od kaotičnega. Uporabnost modela detektorja kaotičnosti smo predstavili s primerom detekcije kaotičnosti Chujevega oscilatorja. Sicer je model detektorja mogoče uporabiti tudi pri ocenjevanju neponovljivosti rezultatov simulacij /8/ in kot izhodišče za praktično implementacijo. 6 Literatura /1 / M. J. Ogorzatek: Chaos and complexity in nonlinear electronic circuits, World Scientific, Series A, letnik 22, 1997. /2/ P.Faure, H. Korn: Is there chaos in the brain? I. Concepts of nonlinear dynamics and methods of investigation, letnik 324, št. 9, september 2001 , str. 773-793. /3/ T. S. Parker, L.O. Chua: Chaos: A tutorial for engineers, Proceedings of the IEEE, letnik 75, št. 8, avgust 1987, str. 982-1008. /4/ C.W. Wu, N. F. Rul'kov: Studying chaos via 1-D maps-A Tutorial, IEEE Transactions on Circuits and Systems I: Fundamental Theory and Applications, letnik 40, št. 10, oktober 1993, str. 707 -721. /5/ M. Suneel: Electronic Circuit Realization of the Logistic Map, Sadhana - Academy Proceedings in Engineering Sciences, Indian Academy of Sciences, letnik 31, št. 1, februar 2006, str. 69-78. /6/ D. C. Hamill: Learning about chaotic circuits with SPICE, IEEE Transactions on Education, letnik 36, št. 1, februar 1993, str. 28 - 35. /7/ M. Stork, J. Hrusak, D. Mayer: Chaos in Simple Nonlinear Systems and Chaotic Systems Simulation and Implementation, Applied Electronics, 2006, International Conference on 6-7 Sept. 2006 str. 193-196. /8/ M. Šalamon, T. Dogša: Problem neponovljivosti simulacij električnih vezij, Informacije MIDEM, 2004, letnik 34, št. 1, str. 11-17. /9/ M. P. Kennedy: Three steps to chaos. II. A Chua's circuit primer, IEEE Transactions on Circuits and Systems I: Fundamental Theory and Applications, letnik 40, št. 10, oktober 1993, str. 657-674. /10/ R. C. Hilborn: Chaos and Nonlinear Dynamics, an Introduction for Scientists and Engineers, Oxford University Press, 2000, Second Edition. /11/ Intusoft: ICAP/4 lsSpice4 user's guide, Intusoft 1988-1996. Doc. dr. Matej Šalamon Izr. prof. dr. Tomaž Dogša oba UNIVERZA V MARIBORU FAKULTETA ZA ELEKTROTEHNIKO, RAČUNALNIŠTVO IN INFORMATIKO 2000 Maribor, Smetanova 17, Slovenija E-mail: matej.saiamon@uni-mb.si, tdogsa@uni-mb.si Prispelo (Arrived): 01.09.2008 Sprejeto (Accepted): 09.06.2009 99 UDK621,3:(53+54+621 +66), ISSN0352-9045 Informacije MIDEM 39(2009)2, Ljubljana A LOW-COST SINGLE-CHIP READOUT CIRCUIT FOR PH SENSING Hasmayadi Abdul-Majid, Yuzman Yusoff, Rohana Musa, Tan Kong Yew, Mohd-Shahiman Sulaiman MIMOS Berhad, Technology Park Malaysia, Kuala Lumpur, Malaysia Key words: ROIC, ISFET sensor, sensor interfacing circuit, single-chip sensor solution Abstract: A readout interface circuit for ISFET-based pH sensor with temperature compensation is described. The ROIC, together with our in-house developed enhancement-type ISFET could lead to single-chip pH sensor solution, leading to low cost sensor for precision agrl-and aqua-culture. Temperature coefficient (TCF) cancellation technique has resulted in ISFET temperature dependency of being a negligible 0.2mV/°C over a temperature range of 0°C to 65°C - close to ideal case of 0 mV/°C. Preliminary results of sub-blocks that were fabricated in earlier runs indicate close resemblance of test results to the simulated results hence the conclusion that the silicon is functioning properly. Nizkocenovno vezje na čipu za odčitavanje pH vrednosti Kjučne besede: ROIC, ISFET senzor, senzorsko vmesno vezje, senzorna enem čipu Izvleček: V prispevku opišemo vmesno čitalno vezje (ROIC) za pH senzor na osnovi tranzistorja ISFET. Vezje ROIC skupaj z ISFET tranzistorjem lahko pripelje do cenene izvedbe pH senzorja na enem čipu, primernega za široko uporabo. Temperaturna odvisnost tranzistorja je zanemarljivih 0,2 mV/°C v temperaturnem območju 0°C do 65°C, zahvaljujoč tehniki ničenja temperaturnega koeficienta, TCF. Preliminarne meritve na izdelanih blokih kažejo na dobro ujemanje s simuliranimi rezultati. 1. Introduction Advancement in the microelectronics and semiconductor technology has led to various research and development activities in the field of sensor and electronics interfacing. One such area is FET-based pH sensor development, or also known as Ion-Sensitive Field Effect Transistor (ISFET) development. Slight modification of a field effect transistor by replacing certain layers with pH sensitive materials such as silicon nitride, aluminum and tantalum has resulted in a miniaturized FET-based pH sensors (ion-sensitive field-ef-fect transistor, ISFET)/1 / . This also promises the possibility of lab-on-a-chip solution, i.e. integration of sensor and the electronics on the same silicon. The end product is targeted to be low cost and suitable for field deployment. An ISFET sensor is cheap and has good sensitivity characteristics. These features, coupled with the fact that it is silicon-based, have made these sensors as one of the main candidate for single-chip implementation. On top of the positive sides of ISFET pH sensor, among the problems faced in ISFET pH sensor development is reproducibility, stability, drift and temperature dependency /2/. Improvements made in each of the problems will, thus, make the use of ISFET sensor more appealing. Analog readout interface circuit (ROIC) plays an important role in acquiring and processing the signal produced by ISFET sensor and passing it to a data converter for further signal processing. Many research works in the area of ISFET ROIC design have been reported in open literature /3/-/5/. Some of the approaches used as part of the sensing mechanisms are bridge type technique /3/, complementary ISFET-MOSFET pair technique /4/ and Constant-Voltage-Constant-Current (CVCC) technique /5/. CVCC technique has the advantage in terms of robustness and suitability for precision agriculture application. In view of precision agriculture application where the sensors are placed in a field for a long period, with periodic data collection time at a center, temperature effects and drifting are some of the issues that need to be handled by the ROIC circuit properly. This means, a need for temperature sensing and cancelling the effects either at the ROIC stage or at the DSP stage for a true pH reading. The latter could be expensive in terms of processing power, leading to the solution not being suitable for field deployment where power is a concern. To do correction due to temperature effects at the ROIC stage, work in /5/, although using CVCC technique but it uses resistor temperature sensor. Resistor temperature sensor is resistive-based, hence producing thermal noise. Due to very low signal being read from the sensor, the thermal noise generated might hinder a single-chip solution hence could potentially increase the cost of the sensor solution. Our work solves this problem through the use of parasitics bipolar transistor to sense This work was supported by the Malaysian Government through the Ministry of Science, Technology and Innovation (MOSTI) and Ministry of Finance (MoF) under the 9th Malaysia Plan (9MP). All of the authors are with MIMOS Berhad, a national R&D institute located at Technology Park Malaysia, Kuala Lumpur, Malaysia 100 H. A. Majid, Y. Yusoff, R. Musa, T. K. Yew, M. S. Sulaiman: A Low-cost Single-chip Readout Circuit for pH Sensing Informacije MIDEM 39(2009)2, str. 100-104 the temperature. /4/ cancels out the temperature effect by using complementary ISFET-MOSFET pair technique. In this work, CVCC technique is implemented for sensing a signal from sensor due to its robustness. Parasitic bipolar is used as a temperature sensing element to correct the actual pH reading (a temperature-independent pH measurement). This also solves the problem faced by /5/, hence making it suitable for single-chip implementation. This paper is divided into four sections. Section 2 briefly describes the ISFET pH sensor. The proposed readout interface circuit with temperature coefficient cancellation method is presented in section 3. The SPICE simulation results of readout interface circuit with and without temperature compensation are presented in section 4. Finally, the conclusions are summarized in section 5. 3. pH Sensor interfacing circuit The proposed ROIC for ISFET pH sensor with temperature compensation consists of three componets: a parasitic bipolar vertical pnp temperature sensor, constant-volt-age-constant-current (CVCC) circuit and a summing circuit (Fig. 2). The principle of this readout circuit is to cancel the positive temperature coefficient (+ve TCF) of ISFET with the negative temperature coefficient (-veTCF) of parasitic bipolar vertical pnp device. Temperature coefficient of both devices can be combined through a summation circuit. As a result, a temperature-dependent characteristic of an ISFET is improved compared to the solution proposed by /5/. A non-temperature independent solution such as the one by /5/ is not suitable for deployment in an uncontrolled environment such as agriculture. 2. Isfet Device A basic cross-section of an ISFET structure is shown in Fig. 1. It is similar to the conventional MOSFET device, except for the standard metal gate which is replaced by a reference electrode and an electrolyte /1 /. In base or acidity (pH) sensor, a sensing membrane sensitive to hydrogen ion concentration such as SiaNU, AI2O3 and Ta20s is used as gate insulator /3,4,5/. A change of pH or hydrogen ion concentration in an electrolyte induces a change in the electric field in the insulator-semiconductor interface and hence changing ISFET threshold voltage (Vt). Thus, different levels of pH can be represented by the drain currents (Ids) of an ISFET device. 11+ P^Si n+ Fig. 1. Basic cross-section of Ion Sensitive Field Effect Transistor (ISFET) To investigate the characteristic of ISFET, an established behavioral macro-model from previous publication is used in the design simulation. The model was firstly introduced by Sergio Martinoiaand Guiseppe Massobrio in 1999/6/ . It is based on a modified SPICE MOS transistor model in which the threshold voltage is influenced by the gate surface potential induced by hydrogen ions present in electrolyte flowing over the gate structure. It is suitable for the simulations of the device operating in a relatively wide range of temperature and hydrogen ion concentration. 1, -A/VV op 2 . >f vdd op 3 Fig. 2. Readout circuitry with temperature compensation 3.1 Constant Voltage Constant Current (CVCC) Circuit In order to ensure the ISFET device operates in linear region at fixed VDS, a CVCC circuit is implemented. The schematic of this circuit is shown in Fig. 3. v2 O 1k wv--- Fig. 3. Constant voltage constant current (CVCC) circuit It consists of three operational amplifiers OP1, OP2 and OP3. OP1 is connected to the source of ISFET while OP2 101 Informacije MIDEM 39(2009)2, str. 100-104 H. A. Majid, Y. Yusoff, R. Musa, T. K. Yew, M. S. Sulaiman: A Low-cost Single-chip Readout Circuit for pH Sensing is connected to the drain and provides a feedback to the ISFET gate. The gate voltage serves as an output signal in response to electrolyte solutions. In addition the feedback gate voltage, which is connected to a reference electrode will establish a stable electrolyte potential with respect to the surface potential of ISFET sensing gate. The varying of the gate voltage will restore the drain current as electrolyte parameters are changed. Another amplifier, OP3, serves as an output buffer. 3.2 Vertical PNP Temperature Sensor In order to develop temperature sensor using CMOS compatible process with no required additional mask, a parasitic bipolar pnp device is employed. The circuit configuration of a bipolar pnp temperature sensor is shown in Fig. 4(a). Fig 4(b) shows the layout structure of bipolar vertical pnp. Area used for the emitter is 100|jm2. Generally, the base-emitter voltage of parasitic bipolar pnp provides a negative temperature coefficient. By adding and controlling the value of the two resistors (R1 and R2) that are connected to the parasitic bipolar pnp, the temperature coefficient of the temperature sensor can be tuned closely to the negative of the ISFET temperature coefficient. 3.3 Summation Circuit A summation circuit consists of a simple amplifier with several resistors. Fig. 5 illustrates the circuit configuration of the summation circuit employed in this work. This summation circuit is used to combine the output of ISFET and bipolar pnp temperature sensor. The positive temperature coefficient (+TCF) of ISFET is cancelled by adding to negative temperature coefficient (-TCF) of pnp temperature sensor. 10 k 10 k j_ -A/VV- V2 O-......- WV 10 k V1 Fig. 5. Circuit schematic of summation circuit. 4. Results Fig. 4(a) Schematic of bipolar temperature sensor. Fig. 4(b) Layout of bipolar vertical pnp device. © 2008 MIMOS Berhad In this work, the ROIC is simulated and fabricated using MIMOS 0.35|jm process technology. Fig. 6 shows measured data for the ROIC. The average sensitivity is 50.16 mV/pH, showing adherence to the specification which is from 50 to 55 mV/pH. Fig. 7 shows measured output voltage for 3 different pH values, confirming the functionality oftheROICs. Fig. 8 shows test setup for the ROIC. Values for R1 and R2 are set to bias the ISFET'sVds to 0.5V. Fig. 9 and Fig. 10 show the simulation result of the output voltage signal from the ISFET and temperature sensor, respectively. This result is obtained for the temperature range of 0°C to 65°C. The simulated temperature coefficient (TCF) for ISFET is +1,420mV/°C. Meanwhile the designed temperature sensor gives temperature coefficient (TCF) of -1.418mV/°C. The readout circuitry is completed by feeding both the temperature sensor and ISFET output voltages with different gains into a summing circuit that mutually offset their temperature coefficient and produce a temperature independent output signal. To validate the improvements of the proposed compensation method, the output voltage signal with and without compensation is shown in Fig. 11 and Fig. 12 respectively. For pH 4, 7 and 10, the corresponding temperature coefficient (TCF) before compensation in 0°C to 65°C ranges is 1.42mV/°C. On the contrary, temperature coefficient (TCF) with compensation has a near zero slope (temperature independent) curves. Thus, the temperature sensitivity compensation method presents a more accurate pH measurement. 102 H. A. Majid, Y. Yusoff, R. Musa, T. K. Yew, M. S. Sulaiman: A Low-cost Single-chip Readout Circuit for pH Sensing Informacije MIDEM 39(2009)2, str. 100-104 ROIC Sensitivity Vs. 35 Chips X a > B0 50 40 m 30 a) 20 O 10 O a 0 10 15 20 25 30 35 Chip ID Fig. 6. Measured ROIC's sensitivity. © 2008 MIMOS Berhad Output ROIC (pH 10.7&4) Vs. 35 Chips — 2.5 I 2 Š 1.5 1 1 -pH10 - pH7 pH4 20 Chip ID Fig. 7. Measured of ROIC's output voltage for three pH values. © 2008 MIMOS Berhad îç-rrçi. si'fvsor > <> < * ¡"i I I I » 1 JL r> I i --------2ffrer^ïirï.....— VL_*s * V2_Éta Rid^i* D RtfJElKlhUlt VûUJ r S I'D GNDA VDDA VDOCs Fig. 8. Test setup for the ROIC. © 2008 MIMOS Berhad (V) T empereur 6(0«)) 0.0 5 0 ; 0.0 15.0 20.0 25 0 30 0 35 0 -SO 0 45 0 50.0 55.0 GO.O 65.0 Ternpe>o C OLD Ik 3. The form of permanent magnets for the generation of a magnetic field in a magnetic refrigerator So far there have been 17 prototype magnetic refrigerators built and tested with the magnetic field being generated with permanent magnets. Of these 13 magnetic refrigerators had the magnetic field sources designed on the basis of a classical magnetic circuit and 4 had magnetic refrigerators designed on the basis of a Halbach array. /2/ 3.1. Magnetic circuit In general, the magnetic circuit is designed with a magnetic field source, which in our case was a high-quality permanent magnet, a magnetic flux conductor, usually a soft ferromagnetic material that conducts and directs the magnetic flux, and an air gap, which provides access to the generated magnetic field. Atypical example of a magnetic circuit is shown in Fig. 2. The basic purpose of the magnetic circuit is to generate as large a magnetic field density in the air gap as possible. This is achieved by directing the magnetic flux in such a way that it forms a closed path, with the losses being as small as possible. In order to generate the largest possible magnetic field density in the air gap, this gap should be as small as possible to reduce the losses of the magnetic flux to the surroundings. h MAGNET v m Fig. 2: Example of a simple magnetic circuit Fig. 1: Schematic illustration of the magnetocaloric effect 3.2. Halbach array Conventional structures of permanent magnets, like the magnetic circuit in Fig. 2, can generate a magnetic field density of up to 1 Tin the air gap. Until the discovery of the so-called Halbach array in the 1970s, to generate larger 106 J. Tusek, A. Sarlah, A. Poredos, D. Fefer: Optimization of the Magnetic Field in a Magnetic Refrigerator Informacije MIDEM 39(2009)2, str. 105-110 magnetic field densities (of about 2 T and more) required the use of electromagnetic devices. However, permanent magnets arranged in a Halbach array can achieve magnetic field densities that were previously only possible with an electromagnet. The basic concept of a Halbach array is in the structure of several permanent magnets that are magnetized in such a way as to cooperate in generating a stronger, more coherent magnetic field density, which can be even larger than the remanence of the permanent magnets being used. A circular Halbach array, which is the basis of some structures for the generation of a magnetic field in a magnetic refrigerator, is shown in Fig. 3. /3/ With the discovery of the Halbach effect a lot of researchers from all over the world started to investigate the effect and its applications. Since then there have been many modified Halbach arrays, with which even larger magnetic field density than are possible with the classical circular Halbach array have been reported. In the early 1990s Leupold et al. /4/ presented the possibility of augmenting the magnetic field density in the air gap by using a soft ferromagnetic material in the air gap of a circular Halbach array. The development of magnetic refrigerators has indicated that modified Halbach arrays are very convenient for the needs of magnetic refrigeration. Among the modified Halbach arrays that are the most promising for magnetic refrigeration, two in particular stand out. Both were developed by Lee et al. /5, 6/ and could generate about 2.5 T of magnetic field density in a 10-mm air gap. Fig. 3: Magnetic field density generated with a circular Halbach array of 8 permanent magnets 4. Development and analysis of the magnets' structure for the generation of a magnetic field in the magnetic refrigerator in the LHT The generation of a magnetic field in the device for magnetic refrigeration in the Laboratory for Refrigeration (LHT) in the Faculty of Mechanical Engineering is designed on the basis of two parallel magnetic circuits (Fig. 5). This geometry was chosen based on price and the future trends in the development of magnetic refrigeration, which is focused on the development of rotating magnetic refrigerators with several regions of large magnetic fields /2/. In this way we can ensure a higher operating frequency of the magnetic refrigerator. First, we carried out a numerical simulation of the magnetic field that is generated by the magnet structure and then we measured the magnetic field, with the intention being to confirm the numerical results. In this way we obtain the exact values of the magnetic field density in the magnet structure. 4.1. Numerical simulation The finite-element method is the most useful way to numerically simulate a magnetic field because there are several commercial programs that can solve very complex problems without the need to develop specific algorithms. We decided to use the FEMM (Finite Element Method) program for the numerical simulation of the magnetic field. FEMM is able to solve two-dimensional, low-frequency electromagnetic problems /7/. In its solving phase the program solves the appropriate Maxwell equations for each node of the numerical mesh. The Maxwell equations are a system of six equations that describe the electromagnetic field. We only work with a static magnetic field, so only three of the Maxwell equations are required in our case. If we considerthe magnetic vector potential (Amag), we could write down only one equation, which is the basis for the numerical simulation of a static magnetic field with the FEMM program /7/: Where is the permeability of a vacuum, which is 4%-10~7 H/m, and j is the electric flux density. We simulated the magnetic field that is generated by the structure of permanent magnets and soft ferromagnetic material with the FEMM program. First, we had to optimize the geometry, because we want to use the minimum amount of material, and at the same time we want to have the strongest magnetic field possible in the air gaps, which means in the areas where the magnetocaloric material would be magnetized, and the smallest magnetic field possible in areas where the magnetocaloric material would be demagnetized. We also optimized the height of the air gaps, 107 Informacije MIDEM 39(2009)2, str. 105-110 J. Tusek, A. Sarlah, A. Poredos, D. Fefer: Optimization of the Magnetic Field in a Magnetic Refrigerator because from the magnetic point of view we want to have the smallest air gap, but from the refrigeration point of view we want to use the largest amount of magnetocaloric material possible, and so for this reason we need to have the largest air gap possible. The optimization was performed by plotting charts of average magnetic field density in the air gap as a function of the different dimensions of the basic parts of the structure (the width and the height of the magnet, the thickness of the external ring and the height of the air gap) /8/. The scheme of the structure with optimized dimensions is shown in Fig. 4 in two-dimensional form. The depth of the structure in axial direction is 170 mm (external ring) and 90 mm (magnets and inner yoke), while in area of air gaps it is further focused on 55 mm. „mid Steel 200 mm i 'j ! l! Steel LJ el IdFeB 40 MGC ¿KtfFeB 40 IViGOe -lO^ETSteel Fig. 4: The scheme of the structure for generating the magnetic field with marked optimized dimensions and used materials The magnet structure is designed on the basis of four neo-dymium-iron-boron permanent magnets (Nd-Fe-B with 40 MGOe maximum energy product). These magnets are currently some of the strongest permanent magnets available, based on their maximum energy product, which is the most important factor when selecting a permanent magnet. As a soft ferromagnetic material for conducting and focusing the magnetic flux we used low-carbon 1010 steel, which is magnetically not ideal, but we chose it because of its low price and good forming properties. The structure has four air gaps with a strong magnetic field and four areas of low magnetic field where the magnetocaloric material is circulating during the operation of the magnetic refrigerator. After defining the final geometry of the magnetic structure we simulated the magnetic field that is generated by it. When the program completed the calculations the results were outputted in graphical form with the distribution of the magnetic field density (Fig. 5) or with a graph of the magnetic field density (B) as a function of distance (Fig. 6); this represents the circle where the magnetocaloric material would be placed during the operation of the magnetic refrigerator and is shown in Fig. 5. 1 ' P Fig. 5: Distribution of the magnetic field density 0,9 n I \ |B|,Tesla 0.8 -0.7 - 0.5 " 0.4 0.3 0.2 0.1 y w w w v C 10 20 Length, or 30 40 50 60 Fig. 6: Magnetic field density as a function of distance, represented by the circle that is shown in Fig. 5 We can see from Fig. 6 that in the air gaps, which means in the areas where the magnetocaloric material should be magnetized during the operation of the magnetic refrigerator, the magnetic field density is 0.98 T and suitably homogeneous for efficient operation. At the same time the magnetic field density in the areas where the magnetocaloric material should be demagnetized is a little less than 0.05 T, which means that the magnets and the carbon steel are far enough away from the demagnetization areas so we have a suitably low magnetic field density. 4.2. Measurement of the magnetic field density When we established the final geometry of the magnetic structure for the generation of the magnetic field, we built it into the magnetic refrigerator. Out of a desire to know accurately the magnetic field we also measured the magnetic field density in the magnet structure. The measurements were made using a three-axis magnetometer with an integrated three-axis Hall probe (SENIS 108 J. Tusek, A. Sarlah, A. Poredos, D. Fefer: Optimization of the Magnetic Field in a Magnetic Refrigerator Informacije MIDEM 39(2009)2, str. 105-110 transducer x-H3x-xx_E3D-2.5kHz-0.1-2T/9/), which is the most appropriate for this kind of measurement because of its accuracy and small dimensions. The structure and the measurement points are shown in Fig. 7 (the front supporting plate covers the view to the basic elements of the structure). The magnetic field density was measured at 40 measuring points, which are marked in Fig. 7. The measurement points are arranged in such a way that three measurement points are in the middle of each air gap, two in the internal edge and two in the external edge of each air gap, while three measurement points are in each demagnetizing area. In this way we cover the whole of the circle In which the magnetocaloric material is situated. Fig. 1: Structure for generating the magnetic field in the magnetic refrigerator and the points where the measurements were made At each of the marked measurement points we measured the magnetic field density and the results are shown In Fig. 8. For comparison the results obtained with the FEMM program are also presented in Fig. 8. MEASUREMENT VALUES (and their comparison with numerical calculated values) measurement t-% i n ^rn I ir ll t ! t i jl ]l I f I li (' I t II It i 1 | > 1 1 V 1 f ,1 \, 3 1, h \\ Il \ il .1 V, / I1 I % // ^y* V ■ y 0 10 20 30 I [cm] 40 50 60 70 Fig. 8: Magnetic field density in the structure It is clear that the magnetic field density in the air gaps is 0.97 T and in the demagnetizing areas it is around 0.05 T. In addition, to estimate the homogeneity of the magnetic field density in the air gaps, which is very important for the efficient operation of the magnetic refrigerator, we also measured the magnetic field density at different heights (radial direction) and depths (axial direction) of the air gaps. We concluded that the magnetic field density varies a lot with the height in the air gap. In the middle of the air gap the magnetic field density was almost perfectly homogenous (Fig. 8), but close to the magnets and far from the magnets, which means on the upper and lower edges of the air gaps, the homogeneity of the magnetic field density is much worse and varies in the air gaps by as much as 0.2 T. At the same time the homogeneity of the magnetic field density is much better for different depths of the air gaps, because at the front and back edges of the air gaps the magnetic field density is 0.95 T. The uncertainty in the measurement results is a combination of two factors. First, is the uncertainty due to the accuracy of the magnetometer, which is 0.1 % of the linear measurement range (0-2 T). On this basis we can calculate the relative measurement uncertainty due to the accuracy of the magnetometer, which is between ±0.1 % and ±2.3 %. The accuracy is the poorest in the air gaps and the best in the demagnetizing areas, where the values of the magnetic field densities are the smallest. Second, Is the uncertainty that is caused by the positioning of the magnetometer's probe during the measurement. We were not able to use mechanical positioning because of the compactness of the structure, which is why the measurement was made manually. The inaccuracy due to the positioning of the probe is the main contribution to the uncertainty of the measured values in the intermediate areas, where the inhomogeneity of the magnetic field density is at its greatest, whereas in the air gaps and the demagnetizing areas, because of the good homogeneity, the error in the positioning was negligible. We were able to estimate the absolute accuracy due to the positioning of the probe as ±2.5 mm. On this basis and with the distribution of the magnetic field density in the intermediate areas we calculated the relative measurement uncertainty in the intermediate areas due to the positioning of the probe to be between +5 % (near the air gaps) and ±46 % (near the demagnetizing areas). This latter value is large and so In those areas the measured results are clearly not very accurate. 5. Conclusion If we compare the results obtained with the FEMM program and the measured values of the magnetic field density (Fig. 8) we can conclude, on the basis of the numerical results, that our structure provides a 0.93 T change in the magnetic field density. On the basis of the measurement results the structure provides a 0.92 T change. The small 109 Informacije MIDEM 39(2009)2, str. 105-110 J. Tusek, A. Sarlah, A. Poredos, D. Fefer: Optimization of the Magnetic Field in a Magnetic Refrigerator difference between the values means that the agreement is very good, and so we can confirm the suitability of the FEMM program for estimating the magnetic field that is generated by the symmetrical magnetic circuits. The difference between the measured and the calculated values is in the range of the measurement uncertainty of the magnetometer we used and the uncertainty due to the positioning of the Hall probe. Furthermore, the reason for the deviation of the results can be attributed to three sources of error. First, is that the FEMM program allows only two-dimensional simulations, which can cause some error. Second, is that some changes were made to the structure, i.e., the inhomogeneity of the structure (e.g., the screws and the holders for the magnets), but these were not considered in the simulation. Third, we did not know accurately the properties of the permanent magnets and the carbon steel that were used in the simulation. This is why we used these assumed materials in the simulation. With the development and analysis of the structure we were able to obtain sufficiently accurate values for the change in the magnetic field density that is possible with this magnet structure. This represents the basic information for further analyses and research on our magnetic refrigerator. References: /1/ A. M. Tishin, Y.I. Spichkln: The Magnetocaloric Effect and its Applications. Institute of Physics Publishing, London, 2003 /2/ K.A. Gschneidner Jr., V.K. Pecharsky: Thirty years of near room temperature magnetic cooling: Where we are today and future prospect. International Journal of Refrigeration, vol 31, 2008, str. 945-961 /3/ K. Halbach: Design of permanent multipole magnets with oriented rare earth cobalt material. Nuclear Instruments and Methods, vol. 169, 1980, str. 1 - 10 /4/ H.A. Leupold, A.S. Tilak, E. Potenzianl.il: Adjustable Multl-Tesla Permanent Field Sources. IEEE Trans. Magn., 1993, str. 2902 - 2904 /5/ S.-J. Lee, J.M. Kenkle, D. Jiles: Design of Permanent - Magnet Field Source for Rotary - Magnetic Refrigeration Systems. IEEE Transactions on Magnetics, vol. 38, 2002, str. 2991 - 2993 /6/ S.-J. Lee, D. Jiles, K.A. Gschneidner, Jr., V. Pecharsky: Permanent magnet structure for generation of magnetic fields. United States Patent, Patent No.: US 6,680,663 B1 ; 2004 /7/ D. Meeker: Finite Element Method Magnetics, version 4.O., User's manual, 2006 /8/ Jaka Tusek: Developing of a regenerator for the magnetic refrigerator. University of Ljubljana, Faculty of Mechanical Engineering, Ljubljana, 2007 /9/ D.R. Popovic, S. Dimitrijevic, M. Blagojevic, P. Kejik, E. Sohurig, R.S. Popovic: Three-AxisTeslameter With Integrated Hall Probe. IEEE Transactions on Instrumentation and Measurement, vol. 56, 2007, str. 1396 - 1402 Jaka Tušek, univ. dipl. inž. str. Dr. Aten Šarlah, univ. dipl. inž. str. Prof. Dr. Alojz Poredoš, univ. dipl. inž. str. University of Ljubljana Faculty of Mechanical Engineering Laboratory for Refrigeration Aškerčeva 6, SI-1000 Ljubljana, Slovenia E-mail: jaka. tusek@fs. uni-lj. si Prof. Dr. Dušan Fefer, univ. dipl. inž. el. University of Ljubljana Faculty of Electrical Engineering Laboratory for Magnetic Measurements Tržaška cesta 25, SI-1000 Ljubljana, Slovenia Prispelo (Arrived): 12.03.2009 Sprejeto (Accepted): 09.06.2009 110 UDK621,3:(53+54+621 +66), ISSN0352-9045 Informacije MIDEM 39(2009)2, Ljubljana MODELING OF SURFACE ACOUSTIC WAVE CHEMICAL VAPOR SENSORS Zdravko Zivkovic1, Marija Hribsek1 and Dejan Tosic2 institute Gosa, Belgrade, Serbia 2School of Electrical Engineering, University of Belgrade, Belgrade, Serbia Key words: SAW, surface acoustic wave, vapor sensor, polymer Abstract: New approach to modeling and analysis of transversal surface acoustic wave (SAW) chemical vapor sensors is presented. The sensor is modeled as a two-port device with parts represented by equivalent circuits. Change of output voltage, or frequency, as a function of vapor concentration is calculated. The model is general and includes propagation losses which are usually neglected in analysis methods. Closed form expressions for vapor concentration estimations are obtained. Simulation results are compared to experimental data. The approach enables better insight in the sensor operation and therefore the optimal design of vapor sensors. Modeliranje SAW senzorja kemičnih hlapov Kjučne besede: SAW, površinski akustični valovi, senzor hlapov, polimer Izvleček: V prispevku predstavimo nov pristop k modeliranju in analizi transverzalnega SAW senzorja kemičnih hlapov. Senzor z dvema priključkoma modeliramo z nadomestnimi vezji. Izračunavamo spremembo izhodne napetosti ali frekvence v odvisnosti od koncentracije hlapov. Model je splošen in upošteva izgube, kijih druge analize zanemarijo. Simulacije primerjamo z merjenimi rezultati. Naš pristop omogoča boljše razumevanje delovanja senzorja zatorej tudi možnost optimalnega načrtovanja kemičnih senzorjev hlapov. 1. Introduction In the last two decades surface acoustic wave (SAW) chemical vapor sensors have found numerous applications due to their compact structure, high sensitivity, small size, outstanding stability, low cost, fast real-time response, passivity, and above all their ability to be incorporated in complex data processing systems. They can be used for in situ monitoring and sensing systems. /1,2,3/ The basic principle of SAW sensors is the reversible sorption of chemical vapors by a solvent coating which is sensitive to the vapor to be detected. It is interesting that a SAW-based sensor system is used as a volatile organic contamination monitoring system for the satellite and space vehicle assembly rooms in NASA. SAW sensors have been able to distinguish organophosphates, chlorinated hydrocarbons, ketones, alcohol, aromatic hydrocarbons, saturated hydrocarbons, and water /1 /. SAW sensors are particularly useful for wireless monitoring in harsh environment /4/. Surface acoustic waves were discovered in 1885 by Lord Rayleigh and are often named after him as Rayleigh waves /5/. A surface acoustic wave is a type of mechanical wave motion which travels along the surface of a solid material, referred to as substrate. The amplitude of the wave decays exponentially with distance from the surface into the substrate, so that the most of the wave energy is confined to within one wavelength of the surface /6,7/. The velocity of acoustic waves is typically 3000 m/s , which is much lower than the velocity of the electromagnetic waves. A basic SAW device was originally developed in 1965 /8/. It consists of two interdigitai transducers (IDTs) on a piezo- electric substrate such as quartz. Each IDT is a reversible element made of interleaved metal electrodes, which are used to launch and receive the waves: an electrical signal is converted to an acoustic wave and then back to an electrical signal. An IDT is a bidirectional transducer that radiates energy equally on both sides of the electrodes. Consequently, theoretical insertion loss introduced by an IDT is at least 6 dB. Starting around 1970, different kinds of SAW devices were developed for applications in pulse compression radars, satellite communications and signal processing systems, mobile radio, and cellular telephones /9,10/. There are very broad ranges of commercial and military system applications that include components for radars, front-end and IF filters, CATV and VCR components, cellular radio and pagers, synthesizers and analyzers, navigation, computer clocks, tags, and many, many others /11,12/. SAW devices have found numerous applications outside their conventional fields, communications and signal processing. In the last two decades considerable work has been done in the development of SAW sensors of different types. High quality SAW filters are used as temperature, pressure and stress sensors /4,13/ as well as chemical and biosensors /14,15/. Generally, two types of chemical SAW sensors are used: transversal (or delay) and resonant. Analysis of SAW devices can be approached to in three ways: (1) exact analysis by solving the wave equation, (2) approximate analysis by means of equivalent electro-me-chanical circuits, and (3) approximate analysis via the delta function model /9,pp.55-122/. It is well known that the exact analysis of SAW devices using surface wave theory is very complex (even in the case of a free surface) 111 Informacije MIDEM 39(2009)2, str. 111-117 Z. Zlvkovic, M. Hribsek, D. Tosic: Modeling of Surface Acoustic Wave Chemical Vapor Sensors /6,7,10,14/. It starts from the second Newton's law and a set of partial differential equations. The equations are solved for the appropriate boundary conditions and relations between mechanical and electrical quantities of a piezoelectric substrate. In the most general case, e.g. in the presence of electrodes on the surface, the Maxwell's equations for the electromagnetic field should be taken into account, as well. Consequently, the exact analysis can be effectively applied only for IDTs with a small number of electrodes. The simplest approximate method of analysis is based on the delta function model. It gives the results relatively fast, but its use is limited to small loads and substrates with lower coupling constants. Better approximate methods use equivalent circuit models for IDTs, where the analysis tools known in electrical engineering can be applied. In these methods the accuracy depends on the complexity of the model. The closed-form solutions are derived for simple IDTs on quartz and lithium-niobate /9, pp.55-122,15/. Recent development of MEMS-based SAW chemical sensors, on new piezoelectric materials, also utilize equivalent circuits but only for modeling frequency characteristics of uniform IDTs /16,17/. In addition, closed-form solutions for more complex IDT structures have been developed by means of advanced electrical engineering analysis methods /18,19,20,21/. One of the main objectives in a chemical sensor analysis is derivation of formulas which connect the change of electrical signals (e.g., voltages and frequency shifts) and chemical quantities (e.g., vapor concentration). The existing analysis approaches are usually: (a) the exact analysis via the wave equation /2,3,7/ and (b) the analysis based on published formulas derived from the wave equation /22,23/. The most complete treatment of the exact analysis has been reported in /7/. Generally, chemical SAW sensors have been analyzed mainly from the chemical point of view with less attention given to the relations between the electrical and chemical quantities and matching conditions at the electrical ports. Typically, the analysis that is based on published formulas (which connect electrical signals and chemical compounds) neglects many properties of a real SAW delay line, such as propagation losses, technological constraints, and production tolerances. This is the reason why some researchers perform more experiments than needed, or have difficulties in explaining discrepancies between the expected and measured values /24/. In this paper, a new modeling algorithm for the analysis of transversal chemical SAW sensors, based on the electrical equivalent circuit method, is presented. The algorithm develops in a straight forward manner explicit general relations between electrical signals, voltages or frequencies, and vapor detection estimations taking into account properties of real SAW devices, which are usually neglected. The whole sensor is modeled as a two-port network consisting of three parts: (1) the input interdigital transducer, (2) the delay line that is the sensing part, and (3) the output interdigital transducer. The transducers are modeled as three-port networks and the delay line as a two-port network. This paper focuses on the essential problem of modeling the delay line with acoustically thin films and the influence of the gas concentration on its behavior and, consequently, on the sensor's output voltage or frequency. The proposed algorithm is used for the vapor concentration estimation and the results are compared with the experimental data. It has been shown that the concentration prediction is better if the properties of practical SAW devices are properly taken into account, especially at higher center frequencies. 2. Principles of SAW Sensor Operation A transversal SAW chemical sensor can be schematically presented as in Figure 1. It consists of two interdigital transducers on a piezoelectric substrate, such as quartz. Piezoelectric materials are anisotropic, which will yield different material properties versus the cut of the material and the direction of propagation. Commonly used substrates are ST-cut quartz and llthiumniobate. ST-cut quartz crystal is cut at a specified angle (0°, 132.75°, 0°) to the crystal-lographic axes so that it has a small or vanishing dependence of wave velocity upon temperature at room temperature /7/. The SAW propagation is in the x-directlon with velocity v = 3158 m/s. Usually, Y-cut lithiumniobate crystal is used. The propagation is in the z-direction with velocity v = 3488 m/s, but the temperature dependence is not negligible. input output transducer transducer Fig. 1. The basic configuration of a chemical SAW sensor. (Acoustic absorbers and matching networks are not shown.) A chemically sensitive thin layer is placed between the interdigital transducers on the top surface of the piezoelectric substrate. The surface wave is induced by an electrical signal applied to the input IDT. The output signal (voltage) is taken from the output IDT. The velocity and attenuation of the wave are sensitive to mass and viscosity of the thin layer. The purpose of the thin layer - a polymer film - is to absorb chemicals of inter- 112 Z. Zlvkovic, M. Hribsek, D. Tosic: Modeling of Surface Acoustic Wave Chemical Vapor Sensors Informacije MIDEM 39(2009)2, str. 111-117 est. When the chemical is absorbed, the mass of the polymer increases causing a change in velocity and phase of the acoustic signal, which causes a change in amplitude and frequency of the output voltage at the load impedance Zi_. Acoustic absorbers (not shown in Figure 1) should be appropriately placed on the substrate edges to damp unwanted SAW energy and eliminate spurious reflections that could cause signal distortions. The IDTs are identical with uniformly spaced electrodes of equal lengths and equal ratio of electrodes width and spacing. The number of electrodes defines the frequency bandwidth of a SAW device. The electrode's length and number, and matching networks at the electrical ports, should be chosen to match the IDT input resistance, at the center frequency fo, to the load resistance R|_ and the generator resistance Rg. In that case, the overall minimal loss due to IDTs is 12 dB. The wavelength corresponding to the center frequency equals the distance between the electrodes of the same polarity. The center frequency and the bandwidth are determined by the IDTs geometry and the substrate type. The middle part of a SAW sensor is a delay line, generally treated as lossless. However, its losses can be neglected only for lower frequencies and small delays (small distances between the transducers). The transfer function of the delay line is normally assumed unity, although this may not be true for high frequencies (f > 0.5 GHz) or if there are films in the propagation path /12,p.1.6-10/. In communications, in electrical filtering applications, the distance between the IDTs is small. Quite opposite, in chemical sensors this part is essential and must have a certain length, usually 100-200 wavelengths /7/, which should be taken into account. 3. New Model of SAW Chemical Sensors The configuration presented in Figure 1 can be modeled by a general equivalent electro-mechanical circuit given in Figure 2. The IDTs are three-port networks and the sensing part is a two-port network designated by DLin Figure 2. The characteristic SAW acoustic impedance of the unloaded substrate is designated by Zo and the acoustic impedance due to the mass loading of the thin film IS ¿-m- Z0 = Apsv (1) Zm=AmPmv <2> where A is the substrate cross-section area through which the waves propagate, rs is the mass density of the piezoelectric substrate, v is the SAW velocity in the piezoelectric substrate, Am is the cross-section area of the thin film, and rm is the mass density of the film. Zg = Rg and Z[_ = Rl are purely resistive electrical impedances of the generator and the electrical load, respectively. IDT 1 Fi DL F 2 IDT 2 V le va, Vg Fig. 2. The equivalent circuit of a SAW sensor. Since the analysis of IDTs using equivalent electrical circuit models is well-known, the focus of this paper is to model the sensing part properties. The key observations relevant to the chemical sensor analysis are the following: (a) the sensor operates near the center frequency and (b) the IDTs are uniform with equal length electrodes. In that case, the IDT driving-point admittance at the electrical port, Yim = GAf) + iBAf) + pKfC0, j = yFl, where C0 is the static capacitance, can be calculated using well-known formulas /12,p.1.6-6/. Ga(f0) = 8k2f0CsWaNBa (/0) = 0 (3) where k is the piezoelectric coupling coefficient, fo is the center frequency, Cs is the capacitance per unit electrode length, Wa is the electrode length (that is, the width of the wave front), Np is the number of electrode pairs. Equation (3) is used for designing proper matching of IDTs at the electrical ports. It should be noted that the characteristics of IDTs outside the narrow band around the center frequency are of no interest. The output voltage across the load Vout is proportional to the mass loading of the sensing part. First, the output voltage in the presence of sensing material (polymer without vapor) is calculated and it serves as a reference voltage Vb, also referred to as the baseline voltage. The difference of the output voltage in the presence of vapor and the reference voltage is proportional to the vapor concentration. In some applications the output voltage is directly measured, but usually the sensor is a part of a more sophisticated system. In that case two equal SAW sensors are used: one is vapor-free and serves as a reference, the other one is exposed to vapor and actually performs the sensing function. The two SAW sensors are embedded into electrical oscillator circuits and the frequency shift between the oscillators is proportional to the gas concentration. Using an electronic circuit called the mixer the voltage proportional to the vapor concentration is obtained from the frequency shift. In any case the voltage across the load has to be found, which implies that the electrical transfer function of the sensor has to be determined. According to Figure 2 the electrical transfer function can be expressed as follows: n/)= ^Ul v„ K out F2FX F, F Va (4) 113 Z. Zivkovic, M. Hribsek, D. Tosic: Informacije MIDEM 39(2009)2, str. 111-117 Modeling of Surface Acoustic Wave Chemical Vapor Sensors where F1 and F2 are mechanical forces. Since the transducers are identical, Vout/F2 is the conjugate complex value of F-\/Vg, and these terms are only functions of frequency. Therefore, \nf)\=\rl3(ft F, F (5) where T^if) = V0utIF2 represents the transfer function of the transducer. Fortunately, in this case, since the sensors work close to resonance in matched conditions at the input and output, the elaborate work of computing T\skf) can be avoided. Instead of Eq. (5) a much simpler expression can be used: |n/o)|=fe(/o)f (6) At resonance, for a delay line without the polymer film, and negligible losses, F2/F1 = 1 and | T13(f0) 12 =1/4 12dB. Therefore, the relative variation of the output voltage l/out due to the mass loading is equal to the relative variation of F2. The delay line of Figure 2 can be schematically represented as shown in Figure 3. V Z m V + F, Fo where pm and hm are the density and thickness of the thin layer, ps is the density of the piezoelectric substrate, and Kw is a coefficient that depends on the technological process and implementation of the sensor. Kw is defined as the ratio of the polymer film width INm and the width of the SAW front Wa: Kw= Wm/ Wa- The components of the wave decay exponentially inside the substrate and the penetration is of the order of one wavelength. Therefore, in Eq.(9), instead of the substrate thickness, one wavelength Xo is used. From the last equation Af can be determined as: Af = -^f02Kv Psv (10) The last equation shows that the higher sensitivity will be obtained if the center frequency is higher, thickness and density of the film larger, and the substrate density and velocity smaller. This means that quartz (ps = 2.62 g/cm3) is a better choice for the substrate than lithiumniobate (ps = 4.7 g/cm3). Furthermore, if ST-cut quartz is used temperature dependence can be neglected /7/. Using the last equation the frequency shift, or the output voltage change, due to the polymer sensing film (without vapor) can be determined: AVr out Vn Af = Pp^p /0 Psv /o^v (11) where pp and hp are the density and thickness of the polymer, respectively. The reference voltage is ^b^O-Kutb^O 1- AVr out I vn (12) Fig. 3. The equivalent circuit of a mass loaded delay line. By analogy between electrical and mechanical quantities, the relative variation of F2 and the relative variation of velocity, forZm much smaller than Zo, are determined by AF, A v_F2 -F 20 F*, 20 F- 20 Z0 + Zn zm «z0 (7) where F20 denotes F2 without mass loading, and v is the corresponding wave velocity. Using Eqs. (6) and (7) the variation of the output voltage due to the mass loading, can be calculated as follows: AV, out V V (8) Where Vo is the output voltage without the mass loading. Using Eqs. (1),(2),(7),(8) and the well-known relationship between wavelength, velocity and frequency, the frequency shift due to the mass loading can be calculated as: /0 Av — Z_ AV, out Zn vn PnA Ps^O (9) Since Al/out is very small, Vb is very close to Vo- Mass sensitivity Sm is an important characteristic of SAW sensors and is defined as Sm = Af/A{pphp) /22/. According to Eq.(11), assuming Kw = 1, it follows: A/ 1 A(ppAp) Psv ti (13) The mass sensitivity is determined only by the substrate (ps and v) and the geometry of IDTs (fo). When vapor is absorbed, an additional voltage change occurs. Using the same reasoning and the fact that hp is much smaller than Xo, the voltage change due to vapor in sorbent phase, can be calculated as A K vap Pvap^p Vu foKv (14) 'b Psv where pvap is the density of the vapor in sorbent phase. The sorbent phase of a volatile chemical compound is that part of the compound which is absorbed by polymer. Since the reference voltage shift without vapor AVout is known, Eq.(11), the last equation can be expressed as follows: AFvap pvap AV0l Vh Vn (15) 114 Z. Zlvkovic, M. Hribsek, D. Tosic: Modeling of Surface Acoustic Wave Chemical Vapor Sensors Informacije MIDEM 39(2009)2, str. 111-117 From Eq.(15) concentration of the chemical compound in vapor phase can be predicted using the known relation between the concentrations in sorbent and vapor phases /22/: K = CjCy (16) where K is the partition coefficient, Cs is the concentration of the chemical compound in sorbent phase (in the sorbent coating /14,p.291/), and Cv is the concentration of the chemical compound in vapor phase (concentration in the ambient /14,p.291 /). Vapor sensitivity depends on the choice of the sorbent coating material, polymer, and its strength of sorption, which Is given by the partition coefficient K. The voltage shift AVvap as a function of concentration Cv can be obtained as Sm ( Hz ng/c-i AK vap Vu KC„, 1 AVr out Pp V0 (17) The concentration in Eq.(17) is in g/cm3. It should be noted that different concentration units can be used in literature, but the units used here are as in /14,p.291/. Consequently, the value of K depends on the concentration units used. Equivalent relationship holds for Af and the frequency shift due to vapor Afvap\ A/vap =^CvA/7pp (18) Equations (14)-(17) are derived for a delay line with negligible propagation losses. These equations contain normalized voltages (dimensionless quantities) so that they remain valid even for lossy delay lines. However, the propagation losses affect voltage shifts and voltages (quantities in Volts) and should be taken into account. The propagation loss is a nonlinear function of frequency, substrate and delay. According to /6,10/ the propagation loss for quartz can be calculated using a^s=(2.15/G2Hz+0.45/GHz) (19) where a^s is the attenuation coefficient in dB/|is and feHz is the frequency in GHz. The propagation loss ade (in dB) is the product of the attenuation coefficient and the delay x (in |j.s): B =ai^st. Any voltage Vora voltage shift AV obtained from a lossless model should be divided by the factor a =10°. The results are presented in Figure 5, where dots represent the measured data /1/. According to /1, Figure 3/ the correction factor is Kw = 0.76. The prediction is in a good agreement with the experimental results. TCE (ppm) 14 000 12000.......................... 10 000 8000 6000 4000 2000 ¿00" 0.05 0.10 0.15 0.20 Fig. 5. TCE concentration versus the normalized voltage shift. The proposed method is verified using experimental data from /24/. The experiments were carried out without any predictions, for devices on quartz at the center frequencies (39.6, 99, 132, 198, 264) MHz, different polymers, and three gases which simulate warfare chemical agents. Polymers were polyisobutylene (PIB), polyepichlorohydrin (PECH), and polydimethylsiloxane (PDMS) deposited by the spin coating technique. Space between the transducers was 1500p,m with the aperture 1800|am. The characteristics were measured directly using E-5061A network analyzer. Predictions are made at fo = 99 MHz for PECH and dichloromethane (CH2CI2, DCM) with pp = 1.36 g/cm3, hp = 0.24mm, K= 1020743, MW= 85g/mol, Vm = 24.46 l/mol. The delay x is calculated from the distance between transducers diDT = 1500|j,m and the wave velocity v = 3158 m/ s: x = d\oj/ v = 0.475|is ~ 0.5jis. The correction factor 1 / a due to attenuation loss is calculated according to Eq.(19), /ghz = 0.099, (2.15fG^Hz+0.45fGHz), t = 0.5, adB = a 1|isx, a = 10adEs/20, 1 /a = 0.996, as expected since /6 / the frequency is lower than 500 MHz and the delay is only about 0.5|is. For 5 ppm concentration of DCM, the predicted frequency shift is 584 Hz and the measured value is 574 Hz/24/. Difficulties were encountered in measuring the concentration of the same gas on PIB /24/. That can be explained by the proposed model: the constant K for PIB is about three times smaller than for PECH and therefore the detected voltage, which represents the gas concentration, will be also three times smaller, which gives insufficient voltage for the analyzer to be detected. 5. Conclusions A new model for acoustically thin SAW transversal chemical vapor sensors has been developed. The model Is based on electrical equivalent circuits of SAW devices, which connect electrical signals and chemical compounds, and takes into account important properties of real SAW devices, such as propagation losses, technological constraints, and production tolerances. The unique feature of the model is a set of closed form analytic expressions for vapor concentration estimations. The expressions explicitly relate the vapor concentration, substrate parameters, and center frequency. They enable insight into the influence of the sensor design parameters on the sensor performance. The presented method predicts very efficiently and correctly the frequency and voltage shifts due to the vapor concentrations in chemical sensors. The simulation results, based on the proposed model, are in a good agreement with the experimental results. The results presented can be used in future for the design of optimal sensors for a given vapor. Acknowledgements The authors would like to thank the Ministry of Science and Technological Development of Serbia for financial support under the project number TR 11026. References /1/ Ho, C.K.; Lindgren, E.R.; Rawlinson, K.S.; McGrath, L.K.; Wright, J.L. Development of a Surface Acoustic Wave Sensor for In-Situ Monitoring of Volatile Organic Compounds. Sensors 2003, 3, 236-247. /2/ Wohltjen, H.; Dessy, R. Surface acoustic wave probe for chemical analysis. Analytical Chemistry 1979, 51, 1458-1464. /3/ Wohltjen, H. Mechanism of operation and design considerations for surface acoustic wave device vapour sensors. Sensors and Actuators 1984, 5, 307-325. /4 / Pohl A. A Review of Wireless SAW Sensors. IEEE Transactions on Ultrasonics, Ferroelectrics, and Frequency Control 2000, 47, 317-332. /5 / Rayleigh L. On waves propagated along the plane surface of an elastic solid. Proc. London Math. Soc. 1885, 17, 4-11. Farnell, G.W. Elastic Surface Waves. In Surface Wave Filters; Matthews, H., Ed.; John Wiley: New York, USA, 1977; pp. 1-55. /7 / Martin, S.J.; Frye, G.C.; Senturia, S.D. Dynamicsand Response of Polymer-Coated Surface Acoustic Wave Devices: Effect of Viscoelastic Properties and Film Resonance. Analytical Chemistry 1994, 66, 2201-2219. /8 / White, R.M.; Voltmer, F.W. Direct Piezoelectric Coupling to Surface Electric Waves. Appl. Phys. Lett. 1965, 7, 314-316. 116 Z. Zivkovic, M. Hribsek, D. Tosic: Modeling of Surface Acoustic Wave Chemical Vapor Sensors Informacije MIDEM 39(2009)2, str. 111-117 /9 / Matthews, H. Surface Wave Filters; John Wiley: New York, USA. 1977; pp. 443-476. /10 / Morgan, D.P. Surface Wave Devices for Signal Processing; Elsevier: London, UK. 1985; pp. 15-57. /11 / Campbell, C. Surface Acoustic Wave Devices and their Signal Processing Applications; Academic Press: San Diego, USA. 1989, pp. 238-315. /12 / Golio, M. The RF and Microwave Handbook, Second Edition; CRC Press LLC: Boca Raton, USA. 2008; pp. 1.6.1-1.6.15. /13 / Seifert, F.; Bulst, W.E.; Ruppel, C. Mechanical sensors based on surface acoustic waves. Sensors and Actuators 1994, A44, 231-239. /14 / Ballantine, D.S.; White, R.M.; Martin, S.J.; Ricco, A.J.; Zellers, E.T.; Frye, G.C.; Wohltjen, H. Acoustic Wave Sensors: Theory, Design, Physico-Chemlcal Applications; Academic Press: San Diego, USA. 1997, pp. 1-7. /15 / Smith, W.R., etal. Analysis of Interdigltal surface wave transducer by use of equivalent circuit model. IEEE Transaction on Microwave Theory and Techniques 1969, 17, 856-864. /16 / Rufer, L.; Torres, A.; Mir, S.; Alam, M. O.; Lalinsky, T.; Chan, Y. C. SAW chemical sensors based on AIGaN/GaN piezoelectric material system: acoustic design and packaging considerations, in Proceedings of the 7th International Conference on Electronics Materials and Packaging, EMAP 2005, 2005, Tokyo, Japan, pp. 204-208. /17 / Rufer, L; Lalinsky, T.; Grobelny, D.; Mir, S.; Vanko, G.; Oszi, Zs.; Mozolova, Ž.; Gregus, J. GaAs and GaN based SAW chemical sensors: acoustic part design and technology, in Proceedings of the 6th International Conference on Advanced Semiconductor Devices and Microsystems, ASDAM 2006, 2006, Smo-lenice, Slovakia, pp. 165-168. /18/ Debnath, N.; Ajmera, J.C.; Hribšek, M.F.; Newcomb, R.W. Scattering and Admittance Matrices of SAW Transducers. Circuits, Systems and Signal Processing 1983, 2, 161-178. /19 / Hribšek, M.; Tošič, D. An Improved Algorithm for Analysis of Uniform SAW Devices, in Proceedings of the 26th Midwest Symposium on Circuits and Systems, 1983, Puebla, Mexico, pp. 243-246. /20 / Hribšek, M.;Tošič, D. An Improved algorithm for Analysis of SAW Pulse Compression Filters, in Proceedings of the 8th Microcoll Conference, 1986, Budapest, Hungary, pp. 373-374. /21 / Hribsek, M. Transfer Function of a SAW Device with Apodized Transducers, in Proceedings of the International Symposium on Circuits and Systems ISCAS'82, 1982, Rome, Italy, pp. 636-638. /22 / Grate, J.W.; Klusty, M. Surface Acoustic Wave Vapor Sensor Based on Resonator Devices, NRL Memorandum report 6829, May 23, 1991, pp. 1-38. /23 / Grate, J.W.; Zellers, E.T. The Fractional Free Volume of the Sorbed Vapor In Modeling the Viscoelastic Contribution to Polymer-Coated Surface Acoustic Wave Vapor Sensor Responses. Analytical Chemistry 2000, 72, 2861-2868. /24 / Joo, B.-S.; Lee, J.-H.; Lee, E.-W.; Song, K.-D.; Lee, D.-D. Polymer Film SAW Sensors for Chemical Agent Detection, in Proceedings of the 1st International Conference on Sensing Technology, Nov. 21-23, 2005, Palmerston North, New Zealand, pp. 307-310. Zdravko Živkovič, Marija Hribšek * Institute Goša, Milana Rakiča 35, 11000 Belgrade, Serbia. E-Mail: zdravko.zivkovic@ymail.com Dejan Tošič School of Electrical Engineering, University of Belgrade, Bulevar kralja Aleksandra 73, PO Box 35-54, 11120 Belgrade, Serbia. E-Mail: tosic@etf.rs * Corresponding author: marija.hribsek@yahoo.com; tel.: +381-11-2413332; fax: +381-11-2410977 Prispelo (Arrived): 20.02.2009 Sprejeto (Accepted): 09.06.2009 117 UDK621,3:(53+54+621 +66), ISSN0352-9045 Informacije MIDEM 39(2009)2, Ljubljana D VB-AS I DISTRIBUTION AND SELECTION IN DVB-T/H REDUNDANCY SYSTEMS Andrej Kosi, Mitja Solar University of Maribor, Faculty of Electrical Engineering and Computer Science, Maribor, Slovenia Key words: Telecommunications networks and systems, DVB-T/H Digital Video Broadcasting - Terrestrial/Handheld, Redundancy system, DVB - Digital Video Broadcasting, ASI - Asynchronous Serial Interface, DVB-ASI distributor, DVB-ASI selector Abstract: DVB-ASI signal is used as a standard interconnection interface in digital broadcasting equipment. In article main focus is on DVB-T/H transmitter and redundancy systems, where specific demands regarding quality, size, price and flexibility are present. On more important transmission sites usually different types of redundancy systems are used for higher reliability. The most known types of redundancy systems are dual drive (DD), N+1 and N+M. When redundancy system is used the need for additional DVB-ASI signals and control over signal is demanded. For this purpose DVB-ASI distributor and DVB-ASI selector units have been developed. Basic DVB-ASI distributor module has one DVB-ASI input and four DVB-ASI outputs. For achieving distribution to more outputs chaining of modules is used. Basic DVB-ASI selector module allows selection from four inputs to one main output and one spare (monitoring) output. With combination of selector modules more complex selections are possible. Decision for four outputs for DVB-ASI distributor and four inputs for DVB-ASI selector is a tradeoff that was made on basis of known demands in dual drive and usual size of N+1 system. In this way costs are reduced and with modular approach extension of system is still possible. Delitev in izbiranje DVB-ASI signala v redundantnih DVB-T/H oddajnih sistemih Kjučne besede: telekomunikacijski omrežja in sistemi, digitalna video radiodifuzija (DVB) DVB-ASI signal, digitalna prizemeljska televizija (DVB-T), mobilna prizemeljska televizija (DVB-H), redundantni sistemi, DVB-ASI delilnik, DVB-ASI selektor Izvleček: V digitalni radiodifuzije je DVB-ASI signal standardiziran in najbolj razširjen način distribucije do oddajnih točk oziroma oddajnikov DVB-T/H signala. V članku je poudarek na DVB-T/H oddajnikih in redundantnih oddajniških sistemih, kjer so specifične zahteve za kvaliteto, velikost, ceno in prilagodljivost oddajniškega sistema. Na večjih in pomembnejših oddajnih točkah se pogosto uporabljajo redundantni sistemi oddajanja, ki zagotavljajo višjo zanesljivost oddajniškega sistema. Najbolj razširjeni sistemi so oddajniki z dvema gonilnikoma (tako imenovani Dual Drive), N+1 sistem (sistem kjer je na N oddajnikov na voljo en rezervni oddajnik) in N+M sistem (na N oddajnikov je M rezervnih oddajnikov). Kljub temu, da N+M sistem zagotavlja visoko stopnjo zanesljivosti se v praksi najpogosteje uporabljata sistem oddajnika z dvema gonilnikoma in N+1 sistem. Ne glede na uporabljen sistem je osnovna naloga zagotoviti nemoteno oddajanje signala v primeru napake na oddajnem sistemu. Na oddajni točki je običajno na voljo en izvor DVB-ASI signala na posamezen oddajnik. Signal je potrebno zaradi podvajanja sklopov razdeliti na več enakih signalov in imeti možnost izbiranja vhodnega signala. Na tržišču sicer obstaja nekaj produktov, ki pa so običajno namenjeni za uporabo v večjih distribucijskih mrežah in so zato neekonomični za uporabo v zgoraj omenjenih redundantnih sistemih. Zaradi večje prilagodljivosti pri izvedbi različnih tipov in različnih velikosti redundantnih sistemih smo rešitev zasnovali modularno. Osnovni modul DVB-ASI delilnika deli DVB-ASI signal na štiri izhodne DVB-ASI signale. V primeru, da potrebujemo več izhodnih signalov zaporedno povežemo posamezne module. Pri DVB-ASI selektorju osnovni modul omogoča izbiro DVB-ASI signala med štirimi DVB-ASI vhodi. S pomočjo kombiniranja več osnovnih modulov DVB-ASI selektorja je možna razširitev izbire števila vhodnih signalov. Pri sami izvedbi rešitve je veliko pozornosti potrebno posveti pravilni zasnovi in izbiri komponent, saj imamo opravka z visokofrekvenčnimi signali. Popačen-je, ki nastanejo pri prehodu visokofrekvenčnega pravokotnega signala skozi koaksialni kabel zmanjšamo s pomočjo kabelskega izravnalnika. V nadaljevanju zmanjšamo še odstopanje (trepetanje) urinega takta. Pri DVB-ASI selektorji je za razliko od DVB-ASI delilnika pred časovnim sklopom dana še možnost izbira vhodnega signala. Amplitudno in časovno izboljšan signal je potem razdeljen in amplitudno ojačen na več izhodnih DVB-ASI signalov. 1 Introduction For distribution of digital television and mobile television content from television studio to transmitter different distribution approaches (optic cables, wireless, satellite, coaxial cable and other) are used. Input signal is transformed (if needed) to standard input for DVB-T/H (Digital Video Broadcasting - Terrestrial/Handheld) transmitter regardless which distribution path is used. As standard input for transmitter DVB-ASI (Digital Video Broadcasting - Asynchronous Serial Interface) signal is used. Formal document describing professional interfaces for Digital Video Broadcasting is ETSI TR101 891 (European Telecommunications Standards In- stitute) /1/. Physical characteristic of the DVB-ASI are similar to digital video signal SDI (Serial Digital Interface) defined by organization SMPTE (Society of Motion Picture and Television Engineers) under designation 259M. DVB-ASI defines interface and way of transmission of digital data. Signal is transmitted serial over 75i2 coaxial cable with nominal amplitude of 800mVpp at source. DVB-ASI is using 8B10 encoding and 270 Mbps bit rate /2/. Equipment used for transmission of SDI signal is not always compatible with DVB-ASI signal and using such equipment in transmission chain can cause problems. One of possible causes could be wrong clock detection, polarity of signal (DVB-ASI is polarity sensitive and SDI is not) and others. 118 A. Kosi, M. Solar: DVB-ASI Distribution and Selection in DVB-T/H Redundancy Systems Informacije MIDEM 39(2009)2, str. 118-122 Depending on size and importance of transmitting site different redundancy systems are used for achieving higher reliability of transmitting system. In most cases one of the following solutions is used for redundancy systems: dual drive (DD), N+1 and N+M. In case of dual drive system dividing DVB-ASI signal for two PA (Power Amplifier) drivers is needed. Basic concept of dual drive redundancy system is shown in picture 1.1. In case of N+1 system dividing and selection unit is needed. For this purpose we need N dividing modules, one for each transmitter. One DVB-ASI output signal is used for transmitter and other one as input for selection unit. With selection unit appropriate DVB-ASI signal is selected for reserve transmitter. Basic concept of N+1 redundancy system is shown in picture 1.2. In case of failure of one transmitter, ASI selector will receive information from control unit about which input to select. DVB-ASI input DVB-ASI distributor PA driver 1 RF output 1 I * I , PA driver 2 RF output 2 2 Proposed solution On basis of presented demands for redundancy systems DVB-ASI distribution (DVB-ASI divider) and DVB-ASI selector units have been developed. To have same basic construction block for all redundancy systems optimum solution is DVB-ASI distributor with four outputs per module and DVB-ASI selector with four inputs per module. Besides main output DVB-ASI selector has one spare monitoring output. To extend distribution to more outputs, chaining of modules is used. To extend number of inputs for DVB-ASI selector combining selector modules is also possible. 2.1 DVB-ASI distributor Basic function of DVB-ASI distributor is dividing input signal on more equivalent outputs. Simply passive dividing of signal is not appropriate solution because in equipment that is following passive dividing system wrong detection could happen. For this reason active dividing is used. Example of active dividing system is shown on picture 2.1, where separated blocks with their functions are presented. Fig. 1.1: Example of dividing DVB-ASI signal for dual drive system. DVB-ASI input 1 DVB-ASI DVB-ASI distributor 1 input 2 DVB-ASI DVB-ASI distributor 2 input N DVB-ASI distributor N v ir TX1 RF output 1 TX2 RF output 2 TX N RF output N ASI selector TX R RF output R Fig. 1.2: Example of dividing and selection of DVB-ASI signal for N+1 system. N+M redundancy system is extension of N+1 system. On each dividing module more outputs are used and M selection units are needed. Although N+M system ensures higher reliability it is rarely used because of higher price and complex architecture. If on transmitting site additional reserve DVB-ASI signal is present and one of described redundancy system is used, then demands for dividing and selecting units are doubled for every system. DVB-ASI input Cable equalizer Re-clock Cable driver 1 Cable driver 2 DVB-ASI outputs Fig. 2.1: DVB-ASI distributor module. When high frequency digital signal is traveling through long coaxial cable signal will loose amplitude and shape because higher frequencies are more attenuated as lower ones /3/. On picture 2.2 /3/ example of dependence between frequency and attenuation for 100m coaxial cable is shown. Attenuation depends on frequency and cable length. Because of this property input signal is first recovered against cable length. For this purpose cable equalizer with inverse cable characteristic for compensating cable influence on signal is used. Insertion loss for 100-motcf Balden 1C34A -6.0Q£t01 \ 1.00E+00 J \ ! \ I.OOE-iOe 1.00E*0? 1.00E+08 Frequency (MHz) 1.00E+09 1.002+10 Fig. 2.2: Attenuation of signal against frequency for 100m long coaxial cable. Long cables and different equipment in transmission chain have also influence on embedded clock of DVB-ASI sig- 119 A. Kosi, M. Solar: Informacije MIDEM 39(2009)2, str. 118-122 DVB-ASI Distribution and Selection in DVB-T/H Redundancy Systems nal. Such influence is manifested as jitter in signal. With reclocking embedded clock is recovered again. Depending on usage of ASI distributor reclocking function can be disabled. Such improved signal is then divided to four output signals. At the end of circuit cable driver is amplifying signal to nominal amplitude of 800 mVpp. On picture 2.3 DVB-ASI signal is shown after passing long coaxial cable. If same signal is then used as input for DVB-ASI distributor the signal on output will be recovered to signal shown in picture 2.4. For such measurements special oscilloscopes with infinitive persistence and color display are used. One of such measuring instrument is Agilent Infiniium MS08104A that was used for measuring DVB-ASI signal before and after usage of developed DVB-ASI distributor and DVB-ASI selector module /7/. Fig. 2.3: Example of DVB-ASI signal after traveling through long coaxial cable. Fig. 2.5: DVB-ASI selector module. Table 3.1: Demands for DVB-ASI equipment. Transmitter output characteristics Units Output voltage (p -p) mV 800 ± 10% Deterministic Jitter (DJ) (p -p) % 10 Random Jitter (RJ) (p-p) % 8 Return loss dB under consideration Max. rise/fall time (20 -80%) ns 1,2 Receiver input characteristics Units Min. sensitivity (D21.5 idle pattern) mV 200 Max. input voltage (p -p) mV 880 Sn (range: 0,1 to 1,0 x bit rate) dB -17 Min. discrete connector return dB -15 loss (0,3 MHz - 1 GHz) Because DVB-ASI physical characteristic are similar to SDI signal descriptions and measurements proposed from EBU (European Broadcasting Union) document Tech 3283 (Measurements in digital component television studios) can be used /5/. On picture 3.1 basic DVB-ASI signal characteristic and measurement definitions are shown. Overshoot Fig. 2.4: Example of DVB-ASI signal on picture 2.3 after using DVB-ASI distributor. 2.2 DVB-ASI selector For DVB-ASI selector module similar components as for DVB-ASI distributor are used. On picture 2.5 basic block of DVB-ASI selector module are shown. Input DVB-ASI signals are first reconstructed against coaxial cable deformation with cable equalizer. Selection block is main part of DVB-ASI selector. In this part selection of Input signal is done. After selecting signal and reclocking cable driver is amplifying signal to nominal amplitude 800 mVpp. 3 Measurements and basic technical data In table 3.1 demands for professional equipment that is using DVB-ASI according to European standard EN 50083-9 are presented /4/, Rise-time Fig. 3.1: Basic measurements for DVB-ASI signal. Besides basic measurements for DVB-ASI signal important information is jitter. Jitter is short term variation of digital signal from their ideal position in Time /6, 8/. There are several ways how measuring of jitter is done. On basis of type of measure we know periodic jitter, cycle - cycle jitter, TIE - time interval error. Another way to visually represent jitter is eye diagram. Quick estimation of signal characteristics is possible from eye diagram. On picture 3.2 /reference 6 Figure 4.2.1 b, 4.2.3b, 4.2.4c and 4.2.5a/ typical eye diagrams with their distributions for different types of jitter are shown. In the upper left part of picture 120 A. Kosi, M. Solar: DVB-ASI Distribution and Selection in DVB-T/H Redundancy Systems Informacije MIDEM 39(2009)2, str. 118-122 random jitter is shown. In the upper right periodic jitter is shown. In the lower left part of picture data depended jitter and lower right duty-cycle depended jitter is shown. u LI Fig. 3.2: Typical eye diagrams for different types of jitter. On picture 3.3 and 3.4 results of measurements are shown. For measuring instrument Agilent Infiniium MS08104A was used /7/. Both pictures are presenting typical values measured on DVB-ASI distributor and DVB-ASI selector modules. On picture 3.3 amplitude charac- teristic are shown with their current, mean, minimum and maximum values. Measured values for peak to peak amplitude (Vp-p), overshot and averaged amplitude (DC offset) are visible. Fig. 3.3: Example of measure results. On picture 3.4 time and eye diagram characteristic are shown with their current, mean, minimum and maximum values. Measured values for jitter, eye width, eye height, fall time and rise time are visible. i r 1 )) ' 'i ' ¥ A « J- 1 . OO av H ' - • i ® Fig. 3.4: Example of measure results. In table 3.2 basic technical data for DVB-ASI distributor are shown. 4 Conclusions In the article a concept for distribution and selection of DVB-ASI signal in DVB-T/H transmitting redundancy systems was shown. Solution with costs effective DVB-ASI distributor and DVB-ASI selector was presented. DVB-ASI distributor module has one DVB-ASI input and four DVB-ASI outputs per module. DVB-ASI selector module has four DVB-ASI inputs, one main and one monitoring DVB-ASI output per module. Because of modular concept different 121 A, Kosi, M. Solar: Informacije MIDEM 39(2009)2, str. 118-122 DVB-ASI Distribution and Selection in DVB-T/H Redundancy Systems Table 3.2: Technical data for DVB-ASI distributor. Data Value Number of inputs 1 for each module Signal type DVB-ASI 270 Mb/s Connector BNC (IEC169-8) Impedance 75 ohms Return loss 18 dB to 540 MHz Number of outputs 4 for each module (up to 3 modules in one rack) Output amplitude 800 mV± 10% DC offset 0,0 V± 0,5 V Overshoot <10% of amplitude Rise time <1,2 ns Jitter < 0,2 UI peak-to-peak External dimensions 19" rack, 1U, 180 mm depth Mains AC voltage 85-264 VAC, 47-60Hz Power consumption <10 VA In table 3.3 basic technical data for DVB-ASI selector are shown. Table 3.3: Technical data for DVB-ASI selector. Data Value Number of inputs 4 for each module Signal type DVB-ASI 270 Mb/s Connector BNC (IEC169-8) Impedance 75 ohms Return loss 18 dB to 540 MHz Number of outputs 2 for each module (up to 2 modules in one rack) Output amplitude 800 mV± 10% DC offset 0,0 V± 0,5V Overshoot <10% of amplitude Rise time <1,2 ns Jitter < 0,2 UI peak-to-peak Communication RS232/RS422/RS485 Communication protocol ASCII External dimensions 19" rack, 1U, 180 mm depth Mains AC voltage 85-264 VAC, 47-60Hz Power consumption <10 VA types and sizes of redundancy systems are possible. Additional flexibility in planning or changing structure of redundancy systems is achieved. Usage of developed products is possible in any other system where distribution or selection of DVB-ASI signal is needed. Measuring results confirmed accordance to different standards for professional equipment and DVB-ASI signal. 5 References /1/ ETSITR 101 891 V1.1.1, Digital Video Broadcasting (DVB); Professional Interfaces: Guidelines for the implementation and usage of the DVB Asynchronous Serial Interface (ASI), 2001-02. /2/ Asynchronous Interfaces For Video Servers by Karl Paulsen, November 2003, http://www.tv-technology.com/pages/ s.0069A.1519.html /3/ Use equalization to drive digital video through long cables, Mark Sauerwald, National Semiconductor Corp, June 2006, http:// www.planetanalog.com/article/printableArticle.jhtml7articielD =188702457 /4/ European Standard EN 50083 Part 9: Interfaces for CATV/ SMATV headends and similar professional equipment for DVB/ MPEG-2 transport streams, CENELEC European Committee for Electrotechnlcal Standardization, March 1997. /5/ EBU document Tech 3283, Measurements in digital component television studios 625-line systems at the 4:2:2 and 4:4:4 levels using parallel and serial Interfaces (SDI), December 1996. /6/ Understanding and Characterizing Timing Jitter, Tektronix, 2003 http://www. tek.com/Measurement/scopes/jitter/55W_16146„ 1.pdf /7/ Infiniium 8000 Series Oscilloscopes Superior Signal Viewing and Analysis, Agilent Technologies, 2007, http://cp. literature.agilent. com/litweb/pdf/5989-4271 EN.pdf /8/ Eye Patterns in Scopes Peter J. Pupalaikis, Eric Yudln, LeCroy Corporation, 2005 http://www.lecroy.com/tm/Library/WhitePa-pers/PDF/Eye_PatternsJn_Scopes-designcon_2005.pdf Manuscript submitted on August 26, 2008. This work was supported by company ELTI d.o.o Slovenia. A. Kosi is with the ELTI d.o.o, Panonska 23, Gornja Radgona, Slovenia (e-mail: andrej.kosi@uni-mb.si). M. Solar is with the Faculty of Electrical Engineering and Computer Science in Maribor, Smetanova 17, 2000 Maribor, Slovenia (e-mail: mitja.solar@uni-mb.si). Prispelo (Arrived): 04.12.2008 Sprejeto (Accepted): 09.06.2009 122