ENGINEERING RESEARCH INSTITUTE TEE UNIVERSITY OF MICHIGAN ANN ARBOR Phase I Report COMPUTER COMPONEJTS DEVELOPMENT.~~ ~~^. ('.'..'..'.. ":.' L.........ari sen SIGNAL CORPS iPROCUREMENT OFFICE Project 2452 ~. i-WASHIINGTON 25~, D. C. September 1956

G -

The University of Michigan * Engineering Research Institute TABLE OF CONTENTS Page LIST OF ILLUSTRATIONS iii ABSTRACT iv OBJECTIVE iv DESIGN PHILOSOPHY 1 INVESTIGATIONS OF VACUUM-TUBE DRIVER CONFIGURATIONS THEORETICAL COMPARISON OF VACUUM TUBES THE CASCODE CIRCUIT 5 EX)PERIMENTAL PERFORMANCE OF THE 436A TETRODE COMPARED WITH CASCODE 437A TRIODES 9 TYPICAL OPERATION OF CASCODE 437A'S 10 DIRECTION OF FUTURE DEVELOPMENT 10 THEORETICAL STUDIES OF DIODE GATING STRUCTURES 11 INTRODUCTION 11 REVERSE TRANSIENET 11 REVERSE TRANSIENT MEASUREMENTS i8 PRELIMINARY INVESTIGATION OF LOGICAL CONFIGURATIONS 23 DIRECTION OF FUTURE ANALYSIS 28 ii

The University of Michigan * Engineering Research Institute LIST OF ILLUSTRATIONS Table Page I COMPARISON OF FIGURE OF MERIT OF VACUUM TUBES 6 II FIGURE OF MERIT OF CASCODE TRIODES WITH AN OVERALL GAIN OF 20 9 III COMPARISON OF THE 436A TETRODE AND CASCODE 437A TRIODES 10 IV DISTRIBUTION OF DIODE TYPES BASED ON RECOVERY TME MEASUREMENTS 20 V REVERSE TO FORWARD RESISTANCE RATIO 22 VI LOGIC STAGE VOLTAGE EFFICIENCY (A) 26 Figure Page 1 The cascode circuit. 7 2 Cascode equivalent circuit. 7 3 Reverse transient measurement. 13 4 Theoretical junction diode. 14 5 Reverse transient characteristics. 19 6 Reverse transient'characteristics of a 1Nl91. 24 7 Logic configuration. 25 iii

The University of Michigan * Engineering Research Institute ABSTRACT This report represents the work performed under Phase I of the contract. The object of Phase I was to determine a reasonable design philosophy and to conduct preliminary investigations to verify the soundness of the design approach. Under Phase I, studies were limited to techniques and components which could be suitably incorporated into dynamic switching circuits utilizing semiconductor diode logic and vacuum-tube power devices. Conventional transformer designs are being considered to obtain the necessary impedance transformations. The remainder of this report contains a discussion of the design philosophy which is being followed in the development of high-speed switching components. This is followed by descriptions and results of component tests which have been made. Also included in this report is a description of a promising circuit configuration utilizing diode logic and a cascode power amplifier. The report is concluded with a brief outline of the direction of future effort. OBJECTIVE The object of this program is the engineering development of a basic set of switching elements which may be used to fabricate certain specific logical configurations. It is desired that the circuitry developed be capable of operating at switching rates which are an order of magnitude greater than rates presently employed in conventional digital switching circuits. This implies that the circuitry must operate effectively at a switch rate of 10 mc/sec. Another objective of the program is the determination and clarification of basic limitations of the component parts of the circuitry. iv

The University of Michigan * Engineering Research Institute.DESIGN. PEILOSOPHY It is ami ost:impossible -to idesign highspeed switehing circuitry.on a hit:ormiss 8basis, Thi s is due prima;rily to the inadequcis or.limitations of.the basic ee.is No neof the basic-:eeentdioes tues, transformers., etc.'are pefect.:Th s uccessful circuit design.must unction orrectly in spite -of thee basgi difficulties. Te ask of- the circuit desiger is the optimization of the circuit;:eign:so theat the bst pOible performance is:obtained. In other'words: th de-sii ger.must be acutely.awre of the limitations -of t.he b'asic cmpnet, and the design ust accot for theselimitatio Co.mpIonen limitations are th primary factors goerning the specifica to..n of the basic design phlosophy Let us then begin the discussion of de. gni philosopy with. a:short exlanation of the component limitations Perhapls the worst limitation arises from th capacitan-e associated withtht;e basic cQmpnents', All cmponents such as tubes and diodes have shut capacitc etween elem.There also exists additional capacitance in the form of stray apacitance due to wiring. In order to produce a voltage pulse at a given poirnt,. it is necessary first to charge the capacitance at that point An em inration Of a typical dynamic circuit configuration would reveal that capacitanei is charged from three emponents, dioe gates, tubes, or trasfoimrs, The gates and the tubes will approximate constatnt. t generators while the output fram the traisformers approximates a coastat voltage generator* Witha constant current g.eratr th effct of capacitance isobtained fraom th: following equations:.=e'o =.i / dt if i a= 0, t < i''I, t2O then | -.e = I/C t The rat:io of e/At i determined by the ratio of I/C. ~...1

The University of Michigan ~ u Engineering Research Institute As a base' consider the standard -onemegacycle dynamic -circuiitry u seed in the SEAC, MIDAC, and vios other ompuers.The goal is to ine the oeration sped.of this basic circuit by at least a factor of te. Can th e capacitane.at the variius critical points e decrieased.y a factor of ten? The answer is'obviously s..It is appaent. that the eapacitance whi:ch pres etly.xists cannot. be stalxially redu small redtiu m of capaci tance is obtaiiable if -ore direct wiring is:.pleoyed. Iowever, this reuction.ay be nullified if the t Abe-: lectle for the power.amplifer has greate irn teelectro de' ciapacitae han -the 6AN5 Next let us consider the effets obtained if the circuit action. is. speded p byincr ing th availabl charing current I If perfect diodes werie available, this wuld be a feasible solution.except from the viewptint of poer disipation. ower presently available diodes suffer unsirable transiet effects uring th switching interval. In. particulart the traient:tdy. The..iode does not recover instanl y, due to the existence of the minor'ity.crriers in the dio: at.he time of switchinag. The diod is not iswitcheduntil the:e carriers have diffused out of the active diodderegion. The number of carrier found in the:diode at; the:time of switching iS pr.opor tioal to the magnitde of the forar.d current Thu, for any given diode th fat:st reovery time is obtaind with the minimum forward curent The re cvry tim will roughly increase in proportion to a increas in for-ard current. Preirminary studies have indicated that back recovery effets are not " as serious as might first h expected.E..en so, it is still important t. obXt!ain fast recovery times This mans that onl small increases in charging cnurrt can:-be obtained The increase in charging urrent will neer approach an ord er of m.'agnitu de One consideration.rie.ns. If the voltge Ae is reduced.a faster operation -ca be achid Thivs. means tha a gate designed to give a l v pulse at I mc/sec should funtion adeqtely t 10 mc if the voltag.swig is.red.ced to T.. This is. the.deired low-curret ati..teo reduction of.the gate.vo:ltage.swig must be accpanied by a prootional i se in the g of the: pwer.amplifier tube. In.the MIAC:1 m/sec circtuits the 6A5 tube has an effective gm of about 7000 p.hos. A IDAC gate structure operating at 10 me/see requires a pentode-ty tube with. a gm of 70,t000.nhos It is posible withexi.stiig tubes to achieve a gm. of 50o000.mhos. Lt us now exame in -ewhat more detail the effects of gm and gate.voltage changes on the performace of a dynamic.switc hing circuit. A transformer of optimum design will delivr a secondary current of inmagnitude ~ ^~1- at2'~ ~ -t -e t4ldeTh, uc -g Ths sth srdlv*:- rrt gter o.2e

The University of Michigan ~ Engineering Research Institute where - is t primary current-, Cp is the capacitane seen at the primary, Eo is the magnitude of the pulse from the secondary,, T iS the rise time desired., g is the current reqired for one gate drive, and N is the maximum permissible.:mber of gate drives., -Now Ip = gmE g, wher Eg is sthe grid swing.. Th loss occurring in the gating can be accounted for by setting Eg = A Eo A<1 Eg may also be- expres ed in:terms of c:rrent a^nd caacitance: B IT'Eg cg where Cg i-s the capacitance seen by the output of the gating circuit, and. B is the efficiency -of the gating configuration Combining the previous formulae yields i..^ ^AB N 2 T2 AB 4CpCg or M2 T2 E where. M.=, a figure of merity and a g pg C, E = AB, the gate efficiency dfined as the product of the gate:..structue.. curre and oltagetransfer ratis 35

The University of Michigan * Engineering Research Institute For the puripose of illistration assusme' T = 10o8 e, E.= 1/.5 and M 5= x 109 rad /sec, For the. above alues N = 25, It is noted that rat iefficient gate structure was assumeda Alsoy the formla does not accun for transfo loases which certaainl are not negligibe at 10 ic,/se In spit f varius losses. which.have been negl cted-t abe e calculation.. is imprtan It in dicate that O10nme dy c circuitry- whih will provide an aquae n er of gate dr:~ives.c.an be ahieved., Little" has beien said about the transforme'r de ign, The:ultimat'e -effiienicy of a traformer at high:frqueies seems to be primarily depende n the cre mat l. To e s.e- differnt edof cons uctim:produce better transformer s. but at high freq cies he core material s to remain the dominant factor, It appa that the chice of a core material ust be a comromise. -pe ab ilityand core losses. Existing ferr-ites which have high perm ilit also have high losses; those with ow losses have lower permeabilities~. It is impoible at this time to redict the exact behavior of high fequency ore mateial for the reduction of permeability is peap Counterhal:aied by the ruced ses, which shold tend to increase the time rate of chage of flux deL ity B etn thougi AB has dereased. E xperimentation.an s tudy are reuire d to determimne the ptimum tr anforier In brief,. the isign philosopy.adpted seeks to minimize th gating curres and the gatevoltage swi The power-amplifier design effort is dir'ead toward a circuit havin a vry high g ad an ex e fi gur: of erit,. -nod a nd. cas oded triodes m bst suited for this application. The typel ofgate structure tuoBe used.r the11 nuomberof drive reuo.ired is. at present an.open qestio. We.shall determin optim- gatestr ur for hi;hf uency operation.and- the maximm number -of gate drives available - ThFIG<ATIONS O:F -VAC11l....TUBE DRIV R CONFIGURATI NS:E'II:CAL C:OARISON OF VACI TE As stated earlier in:.:his port, the.-r of gate-'drives," N,. for serial dynic c.ircuitry ma eexpressed as _~~~~~..:..,.......

The University of Michigan ~ Engineering Research Institute where ~h~ ri gm:~M = -~- -, rad/sec, T p= puls rise timea s:ec,: ad E.= gate eff iciency, T he qantity.M. may be dfined as.a vac tube figure of merit for this.type of circuitry; in order to obatin the largest rmer of gate drives at a given repftiton rate, M s hould.be as large as possible, Csequently, a.rvey of available tubes has been conducted:with a view toward a maximum M, The results of this:.ury are presente in Table I-* From TablX I it is clear that the best of the conventional tubes is| the 43A tetode:. owever, some of the triodes available have extremely large traecolld.ctances and would have a better figre of merit than the.46A if it were. not for the Miller effect capaeity. The cacode circuit describd in the next section. reduces the Miller capacitance ad therey icrases the figure of merit. TECASCODE CIRCIT The cascode circuit of Fig.. 1 repesents n approach which rduces the inpit capacitance due. to:the Milir effect A high fie of merit is thereby obtained fromi the verryhighgm triodes The Miller effect in this circuit depeds on the gain to the plate of tube oe K; K1 is small because the plate lo.a for tube one is the impdan.e looking into the cathode of.tube two, which is a small impendance airately eqal to l/g. Thus e. w mae write Cin = Cgk + (1 ) Cpg (1) In ordr to determine te input capacitance. one m.stfirst determine the im p ane Re looking ito' te catho.of.tb two* Re maybe lf from a.onsideration thequivalIetircuit.of Fig, 2. The following equations are writteen by inspection of the equivalet circuit:..-.~: (Eg El!) Ep = -.Eg- Ip r tp Eg ( +1) (4) E,, ^ - (^^^^^ ^^~~I

TABI I C01PARISOF OF.KF.M F' OF VAC'UM TUBES 6AI5.6AK5 6AIQ6 404A 435A 418A 436A 5857 417A 437A 416A 0.075 0.01 0.03 0,.05 0.0o 0.05 o.07 0,004 1.8 35. 1. 2 cin(Fvf) 9.0 3.9.10 0 7*0 7.9 16 15 9.5 90 11*5 6- 5 cU (f) 4.8 2X85- 2.0 2. 5 2,.9 2;,7 535 2.2 -48.9 0.7 0 "o~ g (ma/volt) 8 5 9 12.5 155 25 30 20 24 45 50 (1) M = (ra/sec) 1210 1490 2010 0 240 13790 4130 4410 (2) (2) (2) "~go,, _,,i^._.,,1 -_-_,_,_,,_.. ~......'..: 0...' (1) Secondary emission- type (2) See section. on the cascode circeui t. -I 3~ ew

The University of Michigan * Engineering Research Institute ~two I _tube.._. t. be -f^~\VA one _________i __ ): Figo. 1..' The- cascodo e l circ uit., C + ~p7, Pi ~ L..~ E Fig. 2:' Cas.ode equivalent:.circui-t. Subst~itut ing (4) into (]) ol-has - ^ ( ^ e) ^' ~' ~~~~~~~~~7 ~

The University of Michigan ~ Engineering Research Institute The impedance we desire is evidntly the ratio of voltage to curent at that point: Re = Ei (6)'P Substituting (4) and (5) into (6), one has finally rp + R Re = + 1 (7) The gain K1 is that of a triode with plate load Re* rp + Re. (rp + RL) (n + 2) rp+ RL Therefore, from (1) the input capacitance is C. - c^^ (^+ ) ^ ^^R L +' P -. Ci Cgk + + 2) rp R ] g p If RL ~ (p4 + 2) rp, this may be simplified to Ci = Cgk + (2.+ RL/rp) Cgp (10) The figure of merit of the cascode circuit is also dependent on t.he output capacitance, which is evidently the capacitance plate-to-grid plus some fraction of the capacitance plate-to~ocathode since this latter capacitance does not have the full output voltage impressed. The effect of Cpk will be that of a capacitance whose size is Cpk multiplied by the ratio of the voltage impressed on Cpk to the total output voltage. Thus Cout Pg +2- pk (11) where K2 = overall gain. The gain K2 is K2 = Eo/Ei = (RLIp)/Ei (12) Substituting Ip from (8), one has K2 = ~p.( p + 2) R ~ (1) If RL ~.(. + 2) rp, this reduces to 8

The University of Michigan * Engineering Research Institute K2 - (r) RL =. gm RL. (i4) Sustittuting the alues for K and. K2 into the formula for output capacitance c laasia:( I+ 1) RL - (r + RL) o.."(':'. +:l )'"RL C pk opg+ (C1g +*/(R RL) c pk() The figre os-of mereit for partie-ular tubes in a.ascode configratio with an ovea ll gain -of 20 are presented. in Table Ie Of these tub- the 416A is evidently the est A test eircuit usng the 437A triode has be stdied ex erimenally in- ord to t t feaibility of:th ca e ir c uditd.AB1E II FIGJURE OF M. IT -OF CASCOD]E TRIODES IH AN OVIBALL GAIN OF'20 _-.-~:.............417A 437A 416-A Cin (IA) 13f5 20.1 8 Cout' (fc) 5P3.28 4.4 2 gm (ma/volt) 24 45 50 M (rad/3Sec ) 360o 0 4780 12,50 P TE PERFM:ANCE OF THE 436A T ODE COMPARESD ITH CCAS E 437A TR The theoretical perfon o th: cascd 437A-is is slihtly better (aproxim ly 14) than that of.a single 4, Exal peormane has been lsered with a 31 tranformer.matched to a 10-ob.se.ondary load as t h plat-e load.The tr'nsfrmer..was:wod on a Ferroxcube 3-C ferrite cup coreo this is.t the- est p.ossib:l. trasfo-e r but: it e to:omw vaum.t.s. Thegrid input -co sised.of puls at ap poxixma tely a 10.6.c/sece rpttion- rat with a pul e d:a.ion ofabout 0.e03 T, bias:, plate' supply,. a int: ignal voltage wr. adj td fo each te so the te maxim.um signal ouput. was.bae nd within th. rating otf thee tube. -'Te-results' gvel in.Table III show t. the outp ent pe r volt of drive is approximately. 16 moe for the 4A'.. Thus thei 437A"Ts maintai-ntheir theor'tical superiority-in experiment. ~..-9, ~ 1.. ~~~~~~~~

The University of Michigan Engineering Research Institute |-: 1 X11 TABL III MPARISON OF TE 436A T RODE AD CA D:437A TRIODES B+ Bias Grid Sirralet Eg, volts I E g 437A 250:26 5.2 3o 69.2 436A 2a0.4 4.6 26.0 56':.' /.:.:... - ". I TICAL OPRATION OF CASCODE 4437A'S -:e large grid s.ing U:s.ed in the foregoing section for comparing the 43-7A.and 436 tubs would be impractical in a c omputer iruit beause 20ma gates would e reuired to produce t- 5.,,grid swing at 10 me A more reali-stic 4ma gate will deliver 1 v to the gid; with a.O. 5-v bias, the cascode circuit will then produce about 100 ma^ n ethe I'oh load thru the: tr fomer at 10.6 mc. Again it is emphasized that the transformer core,material use in these l0ic iexeriments was int'ened for- -mc operation; it is anticipated that the useof ferrites.deigned for higfrequency application will result in a btert transformerl. DIRECTION OF EVURE DEVELOPMENT Pr Ampifie Ath h th'e differnce in performance of the 436A.and 37A tbes is aps not great enoh to j if:y the us of two tubes in the casco:de circit, the.416A will teoretically perform thee times as well a the 43i6A. Sincoe the-.cascoe configurationi hwas w well with. onte tbe tpe, it is reasale to expct that te teo tial daae of the 416A will be.realizLed f ture xper tal pulse. aplifieer will consist of cascade 1416A triodes.. to be dieted toward studies- of gating config. ation using nowideal diodes TanSf r Dsi er ss which:o e just begii.ing will csit o: a sf reially.aailable core materials ad samp trsormers: In paicu e seek.an optimum transformer desig.n compatile:with tlhe cascLde power ampifier. ~ esi Xa Q ~

The University of Michigan ~ Engineering Research Institute TBRE':ICAL ~S-TIDES OF DIODE GATi STUCTES i.NTRODUCTION The bsic MDAC package has, for rpose of analysis, bee broken ow into three partS. itnput logic cofiguration, tube,- and tranforer This section deals particularly with the probles.asociate with the input -logic ircuit operation a highl clock repetition rates; To date the major portion of the work d on gate desig ha assumed ideal diodes However, it was felt:at the beginni that.coiderabe at-te tin should be directed towa-d coasideing the diodes as nolinear circuit- ele mets Th, a diode test programas begu to evaluate currently available diii.es unde citio m ore tpical of the anticipated gate t ucture tha those conditions specified by the manufacturer.| The poiatcontct:diod e bit a sharp revere current spike wen switched frm a forward to a.rv bias codition,. is this reverse tra._ sient and the associatd recove time which are of major intret in the.de.e~ Sign of high-sped swithing cirits. The approach taken to this problem has been first to study the factors tffetieng recovry d seo to cnside.r- th analytically and experimentally the impjrtaance of the.ieal iode behior.in the orall ircuit oprati.n. Althugh there is a.::t diode: the conclusion to date' is that if.fowart.c.uri'-rents. through t di-e li-ted to- 2.ma. an good potcontact diode will work s.atisfactorily-at plse widths of.04 bsea. RVR SE TNS.T If ome assumptio are made regarding the iode, it is possible to obtain. a solution for this reverse current as a function of time, 1, If the crosssectionl dilmesis are much greater than the diode thickness then the problem becomes. one dimensiona 2. A point-contact diode can be approximaed b.a jnction dio.de ~;.11. T-est.:~sults show this t:o' b:rfasoie asasumptio, If'p th conce..trFt,-on;of P N' th it ac" be assu that r".'.ent.flo'w at the j'UnCion:.-e h p ole cm:a', en.. -CZ~ - Ll

The University of Michigan ~ Engineering Research Institute Since the hole current at -the junction is equal to the total current, and if an expression is known for the current -at the junction, the reverse current at any cross section is completely described, Current flow is composed of conduction and diffusion components' I =q jp Ep - q D(16) x where q = charge of an electron or hole, Ip = hole mobility in the N region, Ep = magnitude of the electric field in the N region, (P Dp = diffusion constant for holes in the N region, and p= hole density gradient in the N region. ax At the junction, current flow is by diffusion only: Tr = -q D (17) r P ^ Ix=O X Thus, to solve for the reverse current it is then necessary to solve the continuous-flow or one-dimensional diffusion equation (18) for the proper set of boundary conditions associated with the reverse current transient: Ip P -P = no (18)'- = ~ + D_ _ ) at T, P 2x where P = hole density in N region, Pn = hole concentration in N region under conditions of no externally applied bias, and p = lifetime of holes in N region. The reverse transient is composed of durations T1 and T2. (See Fig. 3.) After a semiconductor diode has been passing current in the forward direction for a period of time sufficient to allow the hole density P to reach its steady-state value, the concentration gradient is as shown in Fig. 4Ia. Whe: the bias is reversed, the conc:entration gradient assumes successive conditions at times tl, t2, t13 and t4. 12

The Universityof Michigan ~ Engineering Research Institute test diode f(av) --.:25-k suare using Te5ktronix type 105 Outxut Trt = reverse, recovery t:jme ha: time in pusec for 1r = 33'IR Trr^ = + T2 for junctin:diod-S r 1 r Irrt' = T2 for point^contact dlioa b where T1 < T2 -. r:1J_ ~ m&t~e Fig 35. Reverse transient iri measurement. -:~ 13~

The University of Michigan Engineering Research Institute - - — x L Ste..t ateoe Concentrat-ion (a) I X o-le Concertration. f R eerse Biaas -Fig- Du,tual dbe t2 (b) -__ _ __ __ Ho~leo Conce ntratn for Reverse Bias Codition p P = Oatx =x p= Oat x Lo:Fig!.. 4, Theoretical junction diode.

The University of Michigan * Engineering Research Institute During Ti the hole diffusion out of the N region is limited by the resistance of the external circuit. Thus T1 i a function of the external circuit resistance and the previous forward current.* Reverse recovery time measurements have shown that for-point-contact and small junction diodes T1i T~ 2.One possible reason for this would be the relatively small volume of P and N material used in the construction of the above type diode as compared to the larger volumes used in such junction diodes as the G*E. 1N91, where TI does become an important factor. Since T1 ~T2 here, we can limit our interest to the time T2' Reverse recovery time (Trrt) is equal to T2 for the diodes considered. (See Fig. 3.) Thus we are intestrsted n solving the diffusion equation for the following boundary conditions (see Fig 4): R = 0, x = 0 and P = Oy x = L at t = O, to obtain P = f(xt) t ~ 0. The procedure is to solve Equation. 18 for the bound conditions given and thus -'evaluate two of the three resulting constants. However, in order to satisfy the third constant it is necessaryto require Equation 19 to equal Equation 20 at t = 0 P = f (x,t) fort > O (19) P = f (x,t) for t< 0. (20) Equations 19 and 20 both result from the solution of Equation 18, but Equation 20 is the steady-state solution for hole density before switching bias conditions. The mathematical steps have been omitted as routine methods of solving partial differential equations are used? sn h (Lo/L) P -(19) 2 sinX [.L( - ) n max = m t L -= xxp jLt1. f (LT )j. (20) *Someties reference is made to holes being:swept out of the N region. It is true that those holes adJacent to the junction are "swept out," but the majority of the N region holes'di-e"t as a result of recoabination -~.15

The University of Michigan ~ Engineering Research Institute Then from Equation 17 at x 0 -and Lo/L < 1-, I -= 2 q P x [(n ex) (21) where. 2 The diffusion constant is related to hol mobilty by Einstein'ts relati:on,(22.) Dp= p: ):' 2 (L~o,._k_ where Pn m~aax mj=maximum. hole conceentration: on. the- N side of the junction.-:der steady-*state tforwsd bias cinditions, Lo == length of N rgion; L = p T = absolute temperature in degrees, and k = Boltzmann'- s constanta Certainly as small 3 as possibl is desired, owTever, c are should be taken in draing conclusi epeially in regad teerate ffects Th model taken is quite simple, and sueh considerations as srface rcombina tion have been. neglected.. oweve:, thr a hr e three conclu.taca be drawn::1 Slnce hole moility. () ts les than electron mbility (te), there would be a gain n recovery time if the current was pr3imarily elect tron ci i.t rater thah hole urrent a as saed here. 2: Te hole moility of silicon (xp) is less than the hole mobility for gerxani L. T- ere fore, diode made from basic geranim may be s.uperr:o silicon diodes of idfl.di l ^onstruction. 3.Most certainly small dimensions are very important as Lo appears as a squalq d term..16

The University of Michigan * Engineering Research Institute There are a number of factors that affect reverse recovery time (Trrt ) 1l Trrt is a function of the diode physics by Equation 21, 2. Trrt is a function of steady-state forward current (If). Notice in Equation 21 that the term Pn max appears. By Boltzmann's equation governing diffusion across the junction, q Eex Pn max = Pno RT (2 E5x is the external forward bias, and is a function of forward current (If) 3. Trrt is a function of pulse rate (prf)o If the pulse rate is such that the forward-current steady-state condition is not reached, then the reverse recovery time is certainly less. 4o Reverse recovery time has been assumed equal to T2, but if T1 is not negligible, circuit resistance then becomes much more important. 5. The back'bias Eb also affects back recovery time, but to a much lesser extent than would be expected. Considering all the above, Trrt = fIl(f f,,RC Cj ). (24) The junction capacitance (C.) has been included. The junction between the N and P regions is also referred to in the literature as the depletion layer. This physical capacitance is to be distinguished from an P'effective hole-storage capacitance" resulting from the hole storage in the N region. If a correlation were found between measured capacitance and recovery time, it would indicate the hole-storage effect would be unimportant compared to charge storage of Cj. Recovery time and capacitances have been measured; and it has been found that there is no correlation, as would be expected. It can be shown that junction capacitance is a function of back bias (E). The junction between the two types of crystalline impurities is characterized by a lack of holes and. electrons in what is known as the depletion layer. Adjacent to this region there are fixed ions that are not neutralized by the mobile carriers in the respective regions. This situation is then exactly equivalent to the parallel-plate capacitor where 17

The University of Michigan ~ Engineering Research Institute C = A.c/s E = dielectric:contant, A = crosssetion al a:rea and:s = distance between plates. As Eis 8 in.read the depletion er widens and thp capaeitan C:decreases.. RI'ER SE TRANSIIT. MASUR TS Eqlation 24 imariz'e: the fators afeSting revere recovery time: Fig. 5 illustrates the type of ata tan on a numbe r of types of point.-ctact dioe: The H2182 dioe ha en ed for illtration becaue of its relative conSisftecy. Notice the following, 1. Trt is ilndependent of circuit resistance...2, Trt increses monotonically with increased forar current 3, The jtion capaeitace Cj and the revere recov.ry time T as fctions of b.are.appaently related However- data from oter diodes haeshown Cj vs E to be essentially onst for all poitcontact dio while Trr vs cres vary betwee the dashed limits:hwn t only for various diode types. but within a particlar type. Thu i Trrt f(Cj). 4, Trrt ine'sed with -ed ced E, indi that.a large:-ampit de input pulse to the tand gae.would be. desire if Trt were critical in the "'and" gate struct.re'. Although:exeptions t-o each' the above have ben ob sr parti larlly amo.ng the N191 and l7 the rem s can be ma rearding p.inl-Otaot diee-in genralt The fU. llwing types bee t'est 1N1171191,- -Fach 1 19, H1.1^. D2:109*, HD... 2B18*, and a nw si ic:. -ct ion di.de by ughes with a,r go'ad recovery tim. If:onsistency o. haracteristic treads can be.dfined as wll ehaved, the stared -(*) diodes h^ve this characteristie of ldependablity Table IV is incd3t to:support the following prosition. That is. good recvery times may be exeted among all types fl pointcontact dioes, Thrfore,:othr propertie will probably be the determining factors in picking a good diode:^ 18

The University of Michigan * Engineering Research Institute Diode under test Tektronix (',-j~ ~ Key type 105 R To input of Diode Symbol square ^Eb^r Tektronix-514AD CRO HD2182-I o o o waive I, I Cin 20Zf HD2182-2 000 generator H D 2182-3 X X X PRF-25KC If(average) HD2182-4 ~-~ e Eb =back bias If = peak for current Cj =junction capacitance 14 12 I0 8.12 J6'S W #2 6n.10 4 t.08 2.06o_ I I, I.0 lt Ii I I 0 2 4 6 8 10 12 14 IK 1.5K 2.0K 2.5K Eb RC (OHMS).20.18 \4mEb\ I =4ma (i f \ Eb=9volts n.1 14.12..12 L.i3.08 o -.06.06 #2.04.04 _.02 _ _ 0 2 4 6 8 101214 0 I 2 3 4 5 6 7 8 Eb If (PEAK) IN ma Fig. 5. Reverse transient characteristics. ~19

The University of Michigan ~ Engineering Research Institute DISTRIBUTION OF DIODE TYPES BASED ON RECOVERY TIME'[MEAS!ERMEDTS Trrt If =2 ma I = nma L 117111 111N191 150 2109 1i17 1191 P11131 150109 282.00.01 02 05.:04 x x.05 xxxx xxx xx x.06 x x x xx xxxx x.07 x x x.08 x x x x x x xxxx.09 xx x x.10 x x xx x..1i x.12 x x x.153 x x. X14.15,16.17..18 o19.20.Eb 9 volts in all cases 20

The University of Michigan ~ Engineering Research Institute From Fig. 3, a/b= IRO/If. Let RgoB initia back resistance and Rf = forward steady-state resistance. Then from Fig. 3, if a/b =, RB = Rf, and if a/b < 1, RBO > Rf. Thus, if the diode in switching from forward to back bias can be assumed to be a nonlinear resistance element of the form RB0 eat the importance of a/b = 1 is apparent. Two diodes, H.D, 2109 and Hughes Si-Junction, have very desirable a/b ratios (Table V). To illustrate the importance of RB0 on ca simple two-input and gate (truth = 1 = a pulse), the following experiment was tried: For a good diode, RB0 103s Q is conservative. Therefore let RD =RBo et/Trrt = RBg during the entire pulse. Let Rf = 10 Q + 22.5 v + 22.5 v ~ -1.5 v -22v5 v +22 v I_' -1.5 v! -5 v -L.5 v Vertical: 1 volt/cm Horizontal:.1 psec/cm 10 resistor pulled down to 1~ -21 ~I I ~10 1 1 1O K 1'":' llll' I IOK.. "'"r...... P1..................-22.. v. ~~~~<J -~~~~~^ O -104 resistor pulled down ~ ~~ ~ simult T ~ 21 ~' -__

The University of Michigan * Engineering Research Institute TABLE V REVERSE TO FORWARD RESISTANCE RATIO Di oa..... a'l' to __.......... IfZ, l' = 2 a 4 ma IN!7 -6, 6,.4 7.8:467 8 1,17,.867 9.75.433 IN191 - 4 - 3 183 5.3.175 6 d35 *.267 F191 - 1.566 36 2 -. 566 36 3.67.44 4. 7.5 5.67.4o 150 -1 60.375 2.95.65 *HDf 210(9 1,267.14 2.67 4HD 2182 1 I 113 4867 2 1413 490 3,07 183 4 1.11 883 b _= l9 volts - *l, "1 4,78 1 2.972.513 4.930.628 -*These. de have proven to be of Bsuperior quality,. ~t1S ii c on unction...^.~22

The University of Michigan * Engineering Research Institute point in regard. to easurent of rev trsets i f artie ular m rtae tht is the smag itud.'f the scope input capaci ce (Fig 35) Figre 6a.dat a t tak-n on a 11911 at high valuof forward currt Udr- the t. es conditions indicate Fie 6b illus.trates the effect of inceas scope c:aacitance Figire 6cs the Hughes published a a for the 31191,. indicating a make-d disa""eent;betwen thes mea d vas and publihed data as the differ by a factor of 3* An increas in' scpe int capacitan e. coul well.account for these differen.aces PRElt0M Y IVESTIGATION OF LOGICAL CONFiGURATIONS Consider first tle gatig structure of Fig* 7a. For optimum d-esign diodes at various points in -the circuit require different charactitics. For diode A it is desired to hav-h t-e sama voltage:rop across it for various vtalueS.of forard.current as the ptpose of this dioe is to eiamp point (2) as othrgate's are pulsed- That i, the slopeof the static characteritic crve is large or AE/AI > 0 Tf Fb At = 20 ma the followinE dta we oai Dio~de T/p S:l'ope ED 2182 2 French 1N191 10 RD 2109 6.66 IN117 6o00 9l 3*:5 Hughes Si.t-ion 2 8.1 is noticed in Table VT that the frwa resistan:eas Rf, at lower| values of do not differ by -an..orld of magnitude" as indicated here. Tie change-in slope' at low value of forward current accounts for this difference., N0otice 1 that diode D of Fig. 7a is also a grid clamp, but hre A0E/41 is| no longer of major concern as this diode, being on the "'Qr' output, is elther

The University of Michigan * Engineering Research Institute Tektronix type 105 To input of El20v square wave Eb Rc Tektronix 514AD CRO R =2.5KQ generator C in= 12,f c PRF 25 Kc/sec C If average.24 o.23- o 1.8. 0~~~~~~~~o.22 1.6 o 73.21 -o 1.4 Cd 0) 0.20 1.2 f= 20ma 4.. I,.,.1 9. 1.0 0.18 0.8.1 7.6 0.16 _ _ 4 __ I I I 4I I I 0 5 10 15 20 25 0 0 10 20 30 40 50 60 70 80 If (PEAK) IN ma Cin (u./f) ADDED (A) (B) 30 MA ~1i'~ ~ ~3.5. sec ~ 0.5, sec O.I/ sec.3 400K 50K ~^ | /~Trrt (C) Fig, 6. BReverse transient characteristics of a 11191"

The University of Michigan * Engineering Research Institute -1.5v +22.5v Diode A A 4.7K 14.77K1I 4X 4.7KQ -22.5 v _ _ _ _ _ _ IC g 1.5v 1. E -22.5v -15v 1 T IOK1 1 (b)^ 0-1..5v ti 97~., Logic — -:cofigurati:on ~ ~-~25.:~~~~~~R.. IO,

The University of Michigan ~ Engineering Research Institute TABLE VI LOGIC STAGE VOLTAGE EFFICIENCY (A) DiodW ^ T iD.lC Volt Lelul Re iiJa CKT' Diode* 2 1''L IT A CRonito IN117 2.25.4 5 -105. -.35 156:4 3,4 1.6 1.90 47.5 No Tube HD 2109 2.25;-1.5 -1i00.,.30 134 4 35 2.1 2040 60% No Tube KD 2182 2.25.1.15.-1o00 -.30 154 4 3;4 1.6 1.9 47.5% No Tube Si.J, 2.,25.l- I5 1.4- 00.'-55 244 4 5,'4 2:40 2,.05 511. 3 No Tube 1N117 2.25 -.-15:1.00..-40 178 4 5.4 1,3 1.50 375% With Tube EHP 2109 2.25 -1,5 -1.00 -.30 134 4 353 1.5 1,80 45% With Tube Si-J 2.25 -1.5.-1.00-,,55 244 4 52. 1,6 1,.65 415.5 WithTube N1117 4..-1.5:.90.50 111 4 5.4 1N5 15 375%.NO Tube H 2109 4,5.-.5.:,90 -...40 8 4 389 4.' 1.9 2.0 50% NO Tube D'2182 45.1.:5 -.90..355 77.8 4 353 1,3 145 36,5 7No Tube iJ 45..5 -.. 90.80 178 4 35 2.3 2. *0.50% No Tbe Nli7 4, 5 With Tube HP 2109 4,-5 With Tube Si-J 4.5 With Tube 1N117 6 75 -16 -90,50 7 4 3. 1.3 1.3.37 5 NoT'PD 2109 6.75 -.lo6. 90 -9Q 50 74 4 35.5 1-,8 1. 8 45 Tu HD 2182 6,75 -1.6 - -90 -.,*40 59,3 4 3.2 092 1,0.25 No Tube Si-J.6o75 -1.6 -,90 +,40 59.3 4 3,5 1.1 2..0 350 No Tbe 11117' 675 1.6 -,.90...350 74 4 33 1,1,.1 27,5% With lbe HD 2109 6. 75 -1.06:-.90, 50 74 4 3.4 1, 4 1,4 355 With Tube r- -. SitJ 6.75 1,.6.: ".,60 88.9 4 3.4 1.6 1.5 37,5% With Tube If s steady-state for ward current thrh the grid cl.ap diode..All of the;above represel t t.best of te parcular d.iode t.yp Le 9 ~:759 50 743.~ 1 8 0. N2Tb

The University of Michigan ~ Engineering Research Institute 0cnducting or it is not In the work to date, it has been found important when designing for noise -lipping to consider the d c voltage drop across diode D as IE(2)dC 1IE(s)d-c I I E(2)dC (4)d-c The transient effect of diode B is not critical for two reasons, First, during the transient time RBO e becoms large after a vry short period, Second when all gates are pulsed, the reverse current telds to charge the grid capacit~ance. In both circuits of Fig.7 it is the output clamp diode that is.of major concern, as it is this diode that-.steals a portion of the charging cur rent I. In neral, the principa pproblem is to.g t the grid up If currnt I is needd to charge Cg, then R and the pullldown upply E1 are chosen such.a to discharg Cg in the desiri tim Therefoie as u ing C can be dis charged, the frward transiints of diodes, A, B,- aand and th reverse trasife -of diode C:fter the' pulse can be neglected. Using eassenially the cireuit of Fig, 7a as a on.e0 inp ut ain gate to detrmine possible transient ffects n th diode B position, it was found that at If = 205 ma all point-contact dioes prformed'qally well When additional input was added to simulate the logical function AB, the noie- through the- gte for.a v input puIlse of,02.sec duration wa fod to be.approxmately 3 v, If capacitae were put in parallel with the pulsed iand: diode, the following noise otputs were observed at point (2) C.9.Noise Voltage 0 3 10 1 20! e5 The circuit of Fig. 7b is a simplificeation f the circuit of Fig. 7a. Sinle the-:- resitanee of the transformer widing is 0 2, the clamp diode may be -eliminated by pulling down the econ dary to - 1.5 v. Formerlyj the pulse transformer, had to supply 21 + e- to raise the potential o.f point (1)* Hoever, in the present.ircuit, ttransfo rmer'ned pu out only the sfficient current to the parallel combination of the:dampi'ng resistance an'd wring capacitance- to cut off the "land" diode. The:current effic-iency B a-s defined in tIe introduction now becomes eqa to 1: wheas.fo rly it as:at best e. ual to I/2. 27

The University of Michigan * Engineering Research Institute Sinee te grdlaclsmp diodle is of maj:r concern- the:data o Table. VI'have: be'i en.Usi ngthe ci eircuit in Fig 7b: to d.teri e the voltag:effi cie.nyof:a single tt r i or ro ypesof utput.clamp diodes:.A = voltageieffiiency = p!: amplitude.at t pi (3): pulse alude at poin (1).h ouput colu () has ben tcorrected for thi diffre in nois:e clipping due: to voltag.drop across the clamp dio For the tests; 4 v. of noise clipping has been taken as. tandrd. VToltage eiff iey dreass with ieed fr ard. c n trogh t he lam.diide as xh ree t ry ti:eas'tae rlghly compar fo'r all. poi ontact di it.is ths s ttiat that it is the ab rati -or la:rge..that is t i: critical:dioe prop With th tub inued the voltage ffiiency of the cirit reases: about 10 de to the grid drawing rrt inga potion of the pulae to:otain the high values of g. In lcocluion, th h.D2109 ofs the bst psibilities fto the foillowing.rea.ons:~:1. Low steady.state forward resistance:ompared to the silicon: junction,,: 2, Small a/ rtiio. *. Good transient properties. Mo-re of t'hese dios as r on order as we origially had.only a few samples.' DIRECION.0F' ANALYSI Ftur e-ffort.is t to b plaed on the realizatim of efficient.highL-...fre;quc:y gating structes th exprimental andical.studies ae c t.plated An aaytical studr of.an.or-an-or gating satru.ctue: realize.d with n'on-id.eal dio'dehas has be t no sultare prestly aail The effects of eapacitance ar include in the m.ode1.tey lr s.e:sults of th. urisistudy shld yild valab d-.1:sign data: tainiigg to high-freqecy ga ting structue usiing exis sting non/ideal diodes o * 28 --....~ QQ ~~~~~~~~~~~~~~~~~~o