cami 174A pa ad Taek One Statistical Cireuit Design: Confirmation of Design Using Compnter- Controlled Test Sets By GD. HAYNE ant S. YANG (sip avd comer 3, 10) ‘Phe ability fo evataate the pesforaunon of Bingar network elements seal iconv tense wile 9 eects uplimizaton ond verfcn on ofthe eyo erin earring nyutene contrat ty digital computers provide a nen copabilty Jr Sinking the steve of neevork dee, ieadboard devel- tpmone and chacurterizetion sod ja Te such 2 toy dist he Fotcarie wodaced wi more nenely Mest the reguivement of the stone ivcobich they ate gal Tce ihage ie effet ty surorporat= She tie ncnsuriny nyse the elowrtbiae relating tho measured Ggovotitios ty the systom. performance pamamater werd during the Sesion stage Guapat of a poser id for debug ve spateun performs parameters a oat fame pois Tevetapovent wodede and a move valid bawie for oveptng or veeetany yreduee am fortory text. Sintstoat analy ue an sid Sahin Cort hats hy svaluairg te veto finnahipe betncen component toisrunaes, memevoment errors aud the falcata eyaten perioemone® parameter Communisgtion systems, beth analog sud digit, often aes inex ncuron to which system erlornaer i highly eenstive. In cases the telatiorshyy llien network wharactristies ond system Derforusinee are comple, compere nee bein ieneacingly se for Ebon snd eyetets desige, Exp wf such clesigs are given in fther papers in thie hs. The design races Hs pot complete nti he design inteat i ves, at by tatarorenents of the hreadboxed fodcls and finally by mescarewente 6” the manufastuned networks, Won phgeical networks to be evaliatelslireet measurement of ie een ater perform res having w ystems available 1198 ru aun svetaan muutnucan Jocadad, ABR 107 5 test insted, To Ie assist tert, one would also require thas the physien) evstem repent all nevemary worst eaten. Thie fpprageh fs bron tre, hit st Tne ehorent ile sash ap te dlouly in obtaining a nominal or wore-cace eset, th difeulty fn mintaining the system, and the luck of infocmetion un a network thet a failed a ts, "These difieies ave lrgely overcame with the test. method de- sevibecl in this paper. Using ths method, the acoverk is mewsared on 4 general purjowe test eb cout by a compater, Fem these smonsurements 6 the network, parameters ape ealeulated which prediet the porfarmance of the weiwerk in system anal which provide diet Snformation about the nebwork, Thy sections which follow sive exe temples of euch tete and the steps necessary co implement the tate fad estab thie wally ‘Wits the generel purpose computer qpersted test ses, we evaluate the Hinear netanee wsametars ‘om imertion Foe and hase lara proluced by Ue uctork wlan rewurel it & stable vonnenion Tn many’ oes, the Titan pacar sl for netic syle a Iie direct seared. Tr oer evar, Iualornntin of the imertion Towa pa dt ruined ‘Ae suggesed ia tke intocuction, the eebtigaship Welw lass an phase speasuremente and eystam per‘armunee wish the network Sage-ted can be quite complex. in face, cxumples exist where setting Timss on inserion lose ead phaee (buses on component tolerance taher than spin peeformange raat “good” netmorke toe fe feoved a "had novo {a he secepltl. Opa oF the sssters per Tormaner in real time previdee a more valid basis for seoepsing oF rejecting ucbworks in faelory tein ard provider poor ald for tebuaziog breed nslely acing evelopment, ‘What ie needed to inlement the apevem performance test then js 1 eomputetional Tink beliaen sgsiem performnnee aril netrork Tost find poe. Fartnnlely, this Tisk has meaily be eealihed tee eases where ncbwere synshess,optimizutien, or tlerance analy fs were dove om a compuler uli fystam performance 8 the design foal, What remcins isto exact the requied computation section of the dcsgn progeam, yt ito the te et eamputer, and eouple i to the maseurement progesm through a ramon se: of parameter “After th sytem pesfosmance el iv iplenented, we tl nals corms OF pase 1109 its validity ack o! exact pszespundnneeottzen He online mens furemenls and actu systems avforeance wit tue nore fected fs eau by (3) unwerainties fm the loar measnrementy on {ii} error snultiphegeion in Uhe astern peremrtarealetation Uncertnvn in the bar mesures ave m fnesion of euch quae tice as tet ffeaueney, ination Iss, and measarerent sveraging time, ‘Theao uneertaities ean be relisbly wedieted. Bere in the system peeameter euleullion coalt in theory be datormined hy experiment Touensutemens| but this wuld be very me ronsuming anc worsttase Scale would be difleal by obtain, ‘The Trwnework used an the ‘Tolerance Analysis Program: is eomvesicat tool fur providing» statistical model of whe elaionship between muxsunl system per formance an aetnal eystam pesformanoe. This Se decueedl in mare ‘etal n Section V The next eesiow wit) desi, in sorte detail, the implementation fot test for expt (D2) channel bao her ‘The D2 Chonwwl Rass is used fue Line alvsion auliesing wt Atunltplesing ia u pues coe “ocilasien transmision system. The oilers born testes ane ve noes ring. ties with & Sevodicaly opera smiteh, Ts the design ste, filter compoxent Nhlnes wece optininel to tee requitemensa ixposed on the tandem Combination of ilor and swine ue, “amteked crunefer tanetion.” Historically, sceeweanee of the Miler weuld 99 bused on meusoe Inenis of insertion Toes under oont?minne excasion Hower, eon ptr stadice by HN Ble pot. tthe possibly of passing had Drvluet tie, feed STF lest end rejrng poo! product ti pasted ES test en wg iertom nes tests Piguet Lebo ch reste ie 1 Compson: tetas seed tar Sac wad ern es si 1200, amr rms svnomae nnenrtekte sormst, APRIL oA of Batons wadiow for Se demi ler whew al ecto olemente sce deviated from nosainal with form dissutions of =2 percent tnd “1 percent. Note cht in fhe 2 peroent eaec, B6 porsent of the food prosuce foil the insertion Ines ts, dn the & pereent cage, 1 percent of the bud product pases the incrtion lose tet. Bven though the correspondence. between the tmp teen impnnves dor th tighter tolerances cewally use in the aystam, the cauchsion is resehed that {proper filter agepan tet ret Ie Inst dst on the saitehe toate taetion (TP) 3 Gowerat Approach Figure 2 ehiaws ibe eoliguenion of Use soe lene used in ue 2 system, Two diferent approaches are available for ealelating the SST ofthe fille. "The nate variate uel ef FR, Mastevmnaraeo snd MLL, Liou' "ean ave us eames elution bus eeuire eae kom ede the network topology and clement vases inching el ares le- mers, Tro other apsroneh, vast by W. H Bamaatt! CA. Deaoes: Tel, Grow" MR. ctor) and 2. E, Fleleches, exsressre the SSE ae e tunction of + paramctee of tho netvor, 2 parasnsters ran be entity drive fromm sss aud pis negaarsieme atthe te=ninale 8 the newwork. The velie of n cupasver in the nswvork is also reid, bt fais 00 ie rudy obtain, Fig, Sntched ls Sor oul cad nainean “main eto: emt aeoel (2° Huaipse neon o cnseusustios oF DRSKOS 1201 ‘The STF of the desulipley flee, Helis), end the 81 analtile Mer, (ja ane exorene by Mlesscher as solid shotn: = period of awitehing froqueney, = input expan, ane B ~ souros impedance of matiplewr Aw approximation to fhe easton wove br yiven by Fleece as Ha) HG 7 Re sali Outs? (YT ol zai @ Since the STF formes wed or the Alter test Toquation (211 ‘Sree only cppcosimite Tule, 5: xe important ¢o coat Ue Secursey of the eppinninatiny, Reeds (rant eatin 11} amd 12 ‘pere compared ith realle ni raleslations wing the etace enable Trecho! af ML. Liou, This compasiber botwect he canoe methods hosted a mgximum disepaney of O09 aR jn oe passband end (0429 inthe zeoet bem 122 Mecensrmeat Maths 1 perform the ealelation indicated by equation (2), thee values Tend ove value of suze sequin for tah Trequeney at whieh 0 ie fr the case Where Hence 12 som we svsomE HECHSIEAL JOURSAL, APKOL LOL so, ond “The value of capacitance Cy Wig. 24) me ako bo Keown, This ie decermined by measuring 2, us just deseribed abu feauency where the eerie incuotanee [2 ¥e 1 reenunce with ite ditribused capacity fo that the romaine of Dae iva vitally dismneted, At the resonant frequene, = to Fea) BR TR ‘Pee messurements of Cy, Zz, and fn are implemented by auto- rmscieally evitching « high impedes probe tw muonaure Fy/Be mh a/¥ athe appropriate frequen, a ‘43 Vadidation of SOP Feet ‘The ST tet i validated by reviewing the factors eontibucing w tzzor in the SEF moceurement ana performing teste to guinsusuranee ‘thatthe net error eel, aia Determination of Capacitance ‘The value of aupaetaner obtained by: measuring se(0) ix shown fon Fig, 4 lgtier with the valies obiained fram mensirementa on fr admittance bridge, The diteropaney in the tito vale Was lee than 12 perosnt and perturb the SH hy Lee than OCO8 a8 in the par bund. The reproducibility of the meusuremest of Cy is beter thon 05 percent = Fig. 8 or enero, coxpmesrios oP DESKS 1268 ese sana Probe Betas ‘Aer optinsnig levels fr dhe Wek esdeof betavea prt ee and licearity, tie anearty resin che probs were seasoned (0 be 8 than 0.0} aB nd core ‘eam probe noiae ere os than OL AB in the STF asjeer band, Lawtirg rors fom the five wnbe input impede were cansgemed onl weve estimated to introduce lee than ‘O01 UB escoeiathe ST passband a Accuracy of fd 2 Meanurements ‘Admittance metstremente were milton denaliplex filter using an adnittanee brige eapable of 01 perv aorurary. From these Ieuanements sy and sy were wsinate! und compare with valuee tbtained from Toe aud phase meueutements ene agreement is within (Od percent. When bridge eseurements ace msde directly ox the indiidal componente of Ue iter and STF is eseulted, the dis fropeney with tie STE obtained fron low gad phase mrascrements irae lores as 00 AR inthe pashan. Much of this cfferenee is due to error inthe eee raodel sc in ralelating 2 paremcers. 44 Sensitivity of STP to Lage and Phase Measurement Herre "The Tolerwise Analysis Propmin: (TAPH was weed to determine the “amplilieation” of meseutement error inert isthe computation fof STF from low acl pluse daca, Figure § gives the results for & 1204 me muy sretuse raeusIOMt JOURNAL, sem Jon lemulipier Mar in which crore ix exe monsurement (F/Eos WeiVs in Tog. sh wore sssumed to be 0.08 B aod 0.2" with the sign of th error randomly chon, ‘This rom inelied 17,500 eases end vir- tually 100 pereen of the SLY loce vals Te within 0.1 dB af the forest value, We alsa not thot w Ol porwent ear in [oss and hase (005 4B, 02*) bes been amplified co x 1 pervnt enor in ST. Th a econ rn, Int and phase ervors 1/8 as great produeed errors in STF 3 ay qroet, Siilar sult wone obtained nt other passhand fre- ‘quero whoee TAP nuns were made. Slightly grosterepreuds were observed in the Mller eee, mon, but the diferences were not sige nifeeat, 25 Summary of D2 Test ‘The SEF tot an the sampler operated tase sat provides a more valid noscptanee lst of the D2 nud ane ceraneple Sere than, (he ingerton loss tae commonly etd for vers ofthis vype, The STF EFS ee AES te cusinaTins oF ens 1206 4 rnousined i 0 shard ane of es tas 009 0 an error of leet swe 02D 'n the evi bans. ing computer coz ny few switching the smneuting abe, Uke RTE zesceurcment 1,2, 8.4 fxd SHE is obrained in Hares 2 pena, even EyLauuaan ie ‘The ces evolution of the T2 wastoer ie covered elector jn hic inexes* This digi. aysbn urloaeh provides a etant exemp.c of ‘sign ites Reving a very earaplex-rbvivmliy lo lee ane sbcse hnevourrmenle, Iw she (2 oqalizer tet used during dev elopnect, insertion dose gad pase meszuremerta eure truueforrnl y= eepoure, eye ypuning, mv! stor ve fon svat aceon Lavin ‘ravping able logis ext vury ng ele leuceratee. "Th program to ot! le senwfonetion wae excentilly fe some program need for fv epalians design. Lz the [2 Saray lox, ve png Ras chosen ne te ce: pavamete:, ead the cried program wis swe sn ‘eh lo alone Ie pregran. size and rupsine time, ‘Tas rogate af TAP pnnlgis on the eye opening racustement ad some vertcacin by scalylval lv “aves show that, for a zocaa] fete, enrnra of Ik] sezeons in ler nad tae eae a rection 1 bal yerce ia the eye pte. Un he st pep geN Used, easare ‘hon preision ie eontaliet lv LT orteenk cr better and tho eye loon ia ereiaated foe 3 eus of ols Il al loaperature bt 5 ine Tu te (a0 TAP enalyses previously tore. the efsete of meus scriens cross le caleated BTE sed ry epening Ware rooedered, foe the somincl vefwnes ony a tae sel fury romero leo hows the staleseal varieton 2 Ree neserkes tUnsTeos ening, frovs componsit witiaus “Ty ees deh the hme af aceepanoe Un tiny Lt € no-stage TAP anclysis $s use Tw the fest sin, e#tsurk components exe vated covolicg to ex pectal sation W!sibene avd che aaa wre rian of the olan pereormance pastas te ecru ‘Ths pve trom Qhis Shop le Is and pha veluee weasel wit Ie womnal meter Und ith the nevorke saeessing tbe wpper ace Tower Eats of Ube system purenmnarce patarseter Th tae seeone. ete, 3740" ra re miads wag The apron & sovstoal yavanna im magus oe wal gen (tara depend or rob 120) non nn meen reomns Jou, Atm LL swe is}. The ale hore eve tania 0° eytem pater fee suruinden. Tigire 6 given he wet'te for the D2 demaliplee ve, This wusysis provides medal of che die buon af meaeaee amonts be eral inthe Suen ith che oneal ee tl Shows fie reletone ip Webwewr ces inte ed le Unita, Note ut a requreraent of 2UD peveent sted or the werk th tear het thee Taais ant ie within Ge ermew rages, “There 9 one tzaGeol Coal hou be moutiened. Lone ie ooceoened wich e minim. eo clwerk, he vol of tasting with higher or loner eearany act be some ap 9) ereaving or inereesing the compoaaat fucerames. Toy posible easton are indicatod ia Fig. 7 Tu Ure tat requiem ou asitacke waed in eyatome have ot victory been olan nin the etaedanine af ayer pat =| ig, ¢ Daan of log eee fr Be wh wet