Exploitation of Initial Conditions to Achieve FluxGain and Nondestructive Readout in

Balanced Magnetic CircuitsE. E. NEWHALL, MEMBER, IEEE, AND J. R. PERUCCA

Summary-A simple experiment on a magnetically-balanced Fig. 1 [3]- [5]. It does not represent the observed sensi-circuit, consisting of a flux source and two equal-length parallel tivity to controlling current or the flux gain and non-paths is outlined. It is observed that if a controlling signal causes one destructive read-out phenomena described herein. Atof the parallel paths to be preferred and this controlling signal is re-moved before switching is completed, the parallel path originally first glance, it appears as if the reasons for this failurepreferred continues to be preferred. The core equivalent of a bal- are twofold. In the first place, since the circuits areanced circuit is then examined to determine if this phenomenon is balanced, path length differences are minimized, andpresent there as well. It is found that the basic behavior of a bal- phenomena which normally would be obscured by pathanced circuit and its core equivalent can be made the same, provided length differences tend to dominate the behavior.care is taken in setting up the same initial conditions in both circuits.It is shown that initial flux levels are important, as well as the rate Secondly, the elements are most often operated in aat which these initial levels are established. An attempt is then made partially set state, which historically has been difficultto relate, in a qualitative way, the circuit properties observed to to represent simply and adequately.known properties of a single core. It is observed that the properties First consider the experiment shown in Fig. 1. Herewhich will be exploited here are present in most standard ma- two balanced circuits are shown. The circuit shown interials; however, a balanced circuit tends to exaggerate these Fig. 1(a) consists of a source of flux in the center, con-properties.

In Section II experiments on two balanced circuits interconnected nected to two equal length parallel paths. In the secondwith a 1:1 turns ratio coupling winding are described. This configura- circuit there is a flux source at both ends, and the equaltion has been found to have two stable states, the final state being length parallel paths are placed between the flux sourcesdependent on the initial condition in one of the elements. Signal [3]. In both circuits it is clear, with the area ratiosamplification has been achieved over at least a 4:1 range of drivecurrents. These results are readily explainable by reference to the shown, that energization of the reset winding wll satu-element characteristics described in the first section. rate all flux-carrying legs. Some reflection shows that

Section III of this paper describes nondestructive readout in bal- energization of the drive winding alone will cause eachanced circuits. As before, balanced circuits are used to exaggerate of the parallel paths to switch equally. With the areasproperties present in most standard materials. The properties ex- shown the parallel paths will go from saturation to neu-ploited here are identical to those used in constructing the bistable tralcircuit described above. Initial tests on a circuit element suitable foruse in an electrically-alterable nondestructive-readout memory are Consider the way in which the splitting of flux at thedescribed. These tests include a so-called high-speed flux source, drive time is influenced by the initial conditions in thewhich has permitted sets as low as one line, or one maxwell, to be circuit. A preset winding is used for this purpose. Dur-used at a 2 Mc rate, without heating, and without requiring time or ing the preset, one of the parallel paths is partiallyamplitude limited drive currents.

switched while the other is shuttled. The cycle thenI. THE INFLUENCE OF INITIAL CONDITIONS ON THE consists of reset, preset and drive. The volt-second area

BEHAVIOR OF A SIMPLE BALANCED CIRCUIT appearing on the output winding was recorded at theAND ITS CORE EQUIVALENT time of preset and at the time of reset. The area was ex-

FqOR CIRCUIT PURPOSES, the dynamic be- pressed as a percentage of the volt-second area whichhavior of a core has been represented successfully would appear in this winding if all the flux switchedby a "switching resistance" in parallel with a cur- was diverted to one of the parallel paths. The results are

rent sink [1], [2]. This model may be thought of as shown in Fig. 1(c) for the circuit in Fig. 1(a). For exam-arising from direct examination of a 1/r vs H char- ple, suppose the difference flux seen on the output wind-acteristi. The curen sk p e.finite d- ing at the time of preset is 40 per cent of a full tip. If

namic~~ .thehod whl.obnn h tagtln a at the time of drive the flux switched split equally be-

ture of this characteristic with an assumption of con- tween the parallel paths then at the time of reset westat vltae otpu duingswichig yeld a onsant would again see 40 per cent of a full tip on the output

vle for the "switching resistance." This conventional winding. However from Fig. 1(c), for a 40 per cetpemode fais trepeset th balnce ciruit shon i set, we see 51 per cent of a full tip at the time of reset.

This indicates that the one leg which was preset is pre-Manuscript received November 8, 1963. ferred at the time of drive, and this preference exists forThe authors are with the Bell Telephone Laboratories, Inc.,alpetsesthn6prcntoafuli.

Murray Hill, N. J. alpeesls hn6 e eto ultp278

Newhall and Perucca: Gain and Readout in Balanced Magnetic Circuits 279

2 AREA UNITS RESET DRIVE

I AREA UNIT

EQUAL 2 SATURATIONLENGTH2NSAT /

RAT N

PATHSIPRESET 2 4 OUTPUTI PRESET * /POIVE

-DRIVE / ~~OUTPUT K/ SAUAINRESET `DIELENGTH-c FLU SE

PATHSFLXST-(a)

DRIVE RESET

(a) (b)POSITIVE REMANENCE-_ ,SLOW DRIVE BEHAVIOR

1001

w-_80 HIL H

70 -w

IL ~~~~~~~~~~~~~~~~~~~~~~~~~~~~-NEGATIVEREMANENCEwa.w SO

(b)40-0 Fig. 2-Core behavior when driven from a voltage source.

Z30-

e 20 - connection with Fig. 1. Since the initial wall area in thepreset side is largest and assuming that the total flux

z 10 reversed is proportional to the product of wall area andw

o Y wallve i andpreal that bot parallel and0 10 20 30 40 50 80 70 80 90 100 velocity, realizing both pathsPERCENTAGE SET IN OUTSlDE PERIPHERY PRIOR have the same mmf applied, and hence the same wall

TO DRIVE velocity, it is not unreasonable to expect a greater flux(c) ~~~~~~~reversal in the preset side.

Fig. 1-Effect of initial conditions on the behavior of a balanced re n thepresetnside.circuit. (a) One form of a balanced magnetic circuit. (b) An We next consider the core equivalent, [11], of aalternate form of balanced circuit. (c) Output flux change ob- balanced circuit shown in Fig. 3 and inquire whether orserved in balanced circuit (d) during reset, plotted vs flux changeintroduced during preset and the pulse sequence; reset-preset not the same phenomenon exists. The core equivalentdrive. shown is not an exact equivalent as we now have parallel

paths of 10 area units and a flux source of 1 area unit;There are several mechanisms which contribute to however, the circuit still demonstrates the same de-

this phenomenon. One of these mechanisms is displayed pendence on initial conditions. In this experiment, allin the experiment shown in Fig. 2. In this experiment a cores were initially reset. Next, with the relay contactcore was driven from a low impedance source so that the open, a preset was applied. This preset, as before, par-current which flowed was determined primarily by the tially sets one of the equal length cores. The percentageback voltage generated by the core. The relationship be- set of the core is plotted as an abscissa in Fig. 3. Nexttween flux and drive then gave the re-entrant loop the relay contact was closed and the 1-unit area coreshown. Now it is known that a plot of wall area vs driven from negative saturation to positive saturation.set would appear as shown in Fig. 2(a) [6]- [8]. With only The volt-second area on the output winding was re-this in mind, the re-entrant loop shown is readily ex- corded at this time and expressed as a percentage of theplained. Initially, since the wall area is small, the re- volt-second area which would exist if all the fluxquired back voltage can only be generated by a high switched went into one of the equal length cores. This iswall velocity. We assume that the flux reversal in unit plotted as an ordinate in Fig. 3. It is clear that for 5time is proportional to the product of wall area and wall isec time limited prepulses, most of the flux switchedvelocity. The high wall velocity can only be sustained at the time of drive prefers the preset core, as long asby a large overdrive. As flux is switched the wall area the preset is less than 74 per cent of a full set. Thebuilds up and the necessary back voltage can be gener- preference is particularly strong for presets in the rangeated by a lower wall velocity, which in turn can be sus- from 10 to 60 per cent. The preference is much less iftamned by a lower overdrive, hence the drive field de- we use an 80,usec amplitude limited preset.creases. As the core approaches positive saturation, the We hypothesize that two distinct phenomena influ-wall area begins to decrease, and we must have an in- ence the behavior of this circuit. In the 80 ,u~sec presetcreasing wall velocity to generate the necessary back case we hypothesize that we merely increase the initialvoltage. This in turn can only be sustained by a larger wall area during the preset interval. In this case, thedrive field. The same sort of explanation can be used in explanation given previously in connection with Fig. 1

280 IEEE TRANSACTIONS ON ELECTRONIC COMPUTERS June

-J RESET PRESET B

xB REA OUTPUT

RELEAYRE(a

LOPOSERVEDI UNIT AREA 10 UNIT AREA UNDER PREPULSE

CONDITIONS

FLUX SOURCE/

(a)80 JASEC CURRENT

LIMITED PREPULSE

FULL. SET UNIT/AREA CORE (a)

5/.ISEC WIDE PREPULSE

B

0

-J 80/ISECWIDEPREPULSE ~~~~~~~~~~~~~~~~~~~LOOPOBSERVEDU. 80/.ISECWIDEPREPULSE ~~~~~~~~~~~~~~~~~~~~UNDERPREPULSE

tu ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~CONDITIONSU Hz

ILJIL

zLUU

cl: ~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~~IjuSECTIMELIMITEDuj-50 ~~~~~~~~~~~~~~~~~~~~~~~~~~~~PREPUL-SE

(b)Fig. 4-Behavior of a single core under slow and

fast prepulse conditions.

0 10 20 30 40 50 60 70 80 90 10PERCENTAGE SET OF 10 UNIT AREA CORE PRIOR TO DRIVE applying a slow ramp of current and measuring the(b) associated flux change. The experiment was repeated for

Fig. 3-Effect of initial conditions on the behavior of various amounts of preset flux holding the switchingthe core equivalent of a balanced circuit. onstant at IresetThe obseve reductionitime constant at 1 jxsec. The observed reduction inthreshold field is well known, particularly in a two core

is sufficient. If we apply a short, fast, time limited pre- per bit memory scheme where this reduction in thresh-set, however, it is argued that we move the walls we old reduces considerably the allowable disturb cur-created before and in addition nucleate walls which were rents.' It is also clear from these experiments that thisnot previously present. Here we assume that the num- same phenomenon plays a significant role in the be-ber of walls nucleated is a function of the overdrive. havior of the core equivalent in Fig. 3.Furthermore, if the preset is removed while some of the We assume that the same phenomenon influences thewalls are still moving, upon subsequent application of circuit of Fig. l(a). Again both paths have the samedrive these walls move easily at a field significantly less mmf applied to them at the time of drive; however, onethan the normal coercive field. Thus we hypothesize path contains many low coercive force walls, while thethat a fast preset leaves many low coercieve force walls second path does not. As in the core equivalent, thein the preset side. When the drive is applied, very little preference depends strongly on the rate at which pre-coupling-loop current needs flow to move these walls. setting is done. This strong dependence of balanced cir-The back voltage generated by subsequent switching cuits on the rate at which things were done in the pasttends to keep the coupling-loop current low, and this makes modeling difficult. This dependence is removed ifcurrent causes only a small amount of switching in the the paths start in a saturated state.core not prepulsed The relationship between sheet circuits and their coreTo support this hypothesis, additional experiments equivalent has been treated in a qualitative manner in

have been performed on a single core. The behavior of a an attempt to identify the significant phenomena whichsingle core under fast and slow preset conditions is must be considered in mapping balanced sheet circuitsshown in Fig. 4. In this experiment, the core was first into core circuits and vice versa. It is very clear in thereset. Then in the case of Fig. 4(b), a 1-g1sec time-limitedpulse was applied so as to partially set the core. The sub- 'Tetrsodithdrcinopsteotedivisvnsequent minor loop followed was then measured by softer than the threshold shown here.

1964 Newhall and Perucca: Gain and Readout in Balanced Magnetic Circuits 281

cases considered that the initial conditions must be DIMENSIONS IN MILS 120 40 40 80mapped carefully if this correspondence is to be main- 0c 0tained. If this correspondence is to be established in a U 0 QUquantitative way, then for every flux and mmf in the 20''.core circuit, an mmf and flux must exist in the sheet (a)circuit with the same time behavior. Measuring flux cor- 4T

respondences is straightforward; however, measuring)mmf correspondences is more difficult. Initial attempts |2)in this direction have been partially successful using the EACH ENDmmf probe shown in Fig. 5. Whenever this probe is PHASE EC ENSEplaced across a portion of a magnetic circuit it can be Lshown that the instantaneous value of the integratedvoltage wave shape is proportional to the instantaneous EAC3TENDdynamic mmf in the circuit between the points of con- (b)tact, provided the probe does not disturb the circuit _ _ _ _ _ _significantly. Wave shapes which have been observed '-FLLLTIPusing such a probe correspond in shape to the coupling- x

loop currents in the core equivalent. Absolute measure- 5jments have been difficult to take due to variations in,contact pressure. 0 _F~~~~~~~~~~~~~~~i I.&SEC

SQUARE LOOP mrrlf PROBE FERRITE W DFERRITE OF PERMEABILITY /U1

FULL TIP

Hi ~~ ~ ~ ~ ~ ~ cjH,dll.1, 1 IHI0 10 20 30 40 50 60CYCLES

(c)Fig. 6-Behavior of a coupled pair of balanced circuits.

Fig. 5-Magnetomotive force probe.

The dimensions shown in Fig. 6 are sufficient to preventII. BEHAVIOR OF A COUPLED PAIR OF bit interaction. Consider interconnecting a pair of bal-

BALANCED CIRCUITS anced circuits as shown in Fig. 6(b). It is clear from theThe way in which balanced circuits can be arranged discussion of Fig. 1 that phase 1 will saturate the top

to achieve directivity and gain using a 2:1 turns ratio in balanced circuit and drive the bottom circuit to neutral.the coupling loop has been described before [4]. Experi- Phase 2 will reverse the procedure. Now it has beenments here show that gain may also be achieved using found that the coupled pair of balanced circuits hasa 1: 1 turns ratio. The amount of gain obtainable is not two stable states dependent on the initial conditions.at present as great as that which can be achieved using The buildup from a small set to a large set for a "1"a 2:1 turns ratio; however, it is now clear that the and a "O" is shown in Fig. 6(c). This shows that forphenomenon described here assists in the operating of perturbations on the order of 10 per cent of a full tip,circuits coupled with a 2:1 turns ratio and may have the buildup will be to a "1" or to a "0" dependent on theto be taken into account if accurate models are to be sense of the initial tip. These buildup curves were takenestablished for these circuits. It is to be noted that gain by first saturating both balanced circuits using a satu-has been achieved previously in square-loop ferrite cir- rating winding not shown. During this interval, the cou-cuits by others using a 1:1 turns ratio [9], [10]. In the pling loop current is limited only by the unavoidablecircuit to be described, gain can be achieved with a coupling loop resistance. Then phase 1 was applied indrive-current range of at least 4:1. coincidence with a small tip signal input to the lower

Consider the series connected balanced circuits shown balanced circuit. The tip signal was then removed andin Fig. 6 [3], [4]. The basic idea is the same as that the buildup over subsequent cycles allowed to takediscussed in connection with Fig. 1; however, now 14 place. In the absence of any tipping, signal buildup stillbalanced circuits are energized from two flux sources. occurred, but in a sense determined by the natural un-The balanced circuits have been found to operate inde- balance of the elements. This in effect served as a noisependently as long as the width of the region between input.holes is significantly smaller than the hole diameter.2 The explanation here is straightforward, keeping in

mind the element characteristics described in Section I.When the initial perturbationl is driven from the lower2Experimental results associated with hole interaction are dis-eeetbc oteupreeeth inlcue

cussed in [12]. eeetbc oteupreeet h inlcue

282 IEEE TRANSACTIONS ON ELECTRONIC COMPUTERS June

one side of the receiving element to be preferred. This but of opposite sense were applied after the write phase.preference continues after the element being driven to The flux at the time of reset was recorded. It is clearsaturation has completed switching. Note that with the that approximately 35 ma turns of disturb is sufficientdrive turns ratio shown, the element receiving informa- by itself to switch around the periphery of the balancedtion will continue to switch after the element transmit- circuit.ting information has finished switching. As long as the Next, two additional windings were placed on theflux gain achieved in this fashion is greater than the re- structure as shown in Fig. 8(a). These are labeled inter-sistance loss, buildup will occur. rogate drive and interrogate reset. A normal writingSome reflection shows that the preference for the side cycle was first performed. This left the region between

originally preferred should cause the coupling loop cur- balanced circuits neutral. A constant dc bias was thenrent to reverse toward the termination of the transfer applied to the interrogate reset winding. It is clear thatcycle. A representative coupling loop current wave this tends to hold the structure in the state which existedshape is shown in Fig. 6(c), and it is clear that this is at the termination of the write cycle. Next, an inter-indeed the case. The amount of the allowed reversal will rogate drive phase was applied which tended to saturatedepend upon the coupling-loop resistance and induc- the structure. However, this phase was time limited sotance. It is to be noted that once the transmitting ele- that the information originally set into the structurement shuts down, it presents a short-circuited turn to the was not completely erased. Upon removal of the inter-receiving element which would tend to cause equal-flux rogate drive, the interrogate reset sent the structuresplitting between the parallel paths. As long as some back to neutral. The structure was then ready for an-coupling loop circuit resistance and inductance is pres- other interrogate drive pulse. The resultant waveent, the natural tendency of one side to continue to shapes for a "1" and "0" are shown in Fig. 8(b). Hereswitch is not inconsistent with the circuit constraints. nondestructive interrogation is taking place at a 1.8The number of cycles required for buildup has been Mc rate. The output pulses are approximately 120 nsecfound to decrease with an increase in coupling-loop long. Taking into account the wave shapes observed, itresistance, as is to be expected from the above argument. is clear that the leg preferred at the time of write is theThe final stable operating level has been found to be preferred leg at the time of interrogate drive. In ex-surprisingly high, 81 per cent in the case of a "0," and plaining these results, we suppose that by limiting the67 per cent in the case of a "1." Transfer time has been extent of switching at the time of interrogation, themade as fast as 500 nsec without any basic change in the low coercive force walls generated in the preferred pathabove results. Faster transfer times were not possible at the time of write, are not swept out, and for this rea-with the pulsers available. The material used here was son this path is preferred at the time of interrogate driveMnMgCd ferrite. The same result has been obtained and the time of interrogate reset. It should be notedusing MnMgZn ferrite. that although time limiting has been used here to limit

the flux delivered at the time of interrogate drive, fluxIII. NONDESTRUCTIVE READOUT OF limiting by controlled geometry has also been used suc-

BALANCED CIRCUITS cessfully.The phenomena outlined in Section I are hypoth- It has been found that increasing the amplitude of the

esized to be directly responsible for the nondestructive- interrogate drive and interrogate reset may change thereadout capacity to be described here. The exact man- level of the nondestructive readout signal. However, thener in which these phenomena act to permit nondestruc- sense remains correct over a wide range of drive ampli-tive readout is not as clear as the manner in which they tudes and flux levels. In fact, the information can onlyenter into the experiment described in Section II. In be completely erased if the structure is saturated. Afact, other phenomena not yet well understood could multiplicity of stable nondestructive readout levelsconceivably be dominant here. exists dependent on the drive conditions. An indication

Consider the circuit of Fig. 6(a) to be wound as shown of the way in which the stable level is reached is shownin Fig. 7(a). The erase winding initially saturates the in Fig. 9(a) and (b). The flux level of both sides of thebalanced circuit. Energization of the write winding in balanced circuit is shown just after writing and aftercoincidence with a signal input to the tip winding will successive interrogate cycles. The dotted lines show thecause the balanced circuit to store information as a history of each of the parallel paths. Fig. 9(a) was takenclockwise or counterclockwise perturbation fron neu- using a tip mmf of 13 ma turns while Fig. 9(b) was takentrality. Curves which show the control milliampere using a tip mmf of 26 ma turns. The interrogate driveturns vs difference flux measured at the time of reset are in both cases lasted for 1 jxsec. The final stable levelshown in Fig. 7(b). Curves for two different values of appears to be the same in both cases.write current are shown and are labeled "no disturb The phenomenon described above might be incor-current." To obtain the other pair of curves, the same porated into an electrically alterable high-speed non-pulse sequence was applied with one addition. A multi- destructive-readout element as shown in Fig. 10. First,plicity of disturbs of amplitude equal to the tip current consider the so-called "high-speed flux source" shown in

1964 Newhall and Perucca: Gain and Readout in Balanced Magnetic Circuits 283

"POS. SATURATION TIP = 13 MILLIAMPERE TURNSINTERROGATE DRIVE =1/LSEC

ERSE50RA

T 5T WRITE 0trx z32Tal 73T

TIP" UDISTURB |-OUTPUTj

(a)

w~~~~~~~~~~~~~~~~100 _ 5r: 0

/ \

wU) N DISTURB / ffi 800 MA ,ERASE Z 0NO DISTURBCURRENT -I >J1IZ ( 75 __ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _ _

3D // // \ \ < TIP= 26 MILLIAMPERE TURNSoa j INTERROGATE DRIVE= IACSEC_j ~~~~~~~~~~~~~DISTURBZU | >_ . 200 MA WRITE \U_ 5Q / CURRENT >

< 506 2 2 4 8 5 4

v~~~~~~~~O~~0 /

WRITE-- freatbl one c MA WRITErmwNERGCURRENTA

< au 25TIP AND DISTURB CURRENT _

INADJUSTED TO BE THE SAME_x

0 8 16 24 32 40 48 56 64 72 /' \

TIP-MILLIAMPERE TURNS 5 ~ \(b) r

Fig. 7-Tip disturb characteristic of balanced circuit. NEG. SATURATION

CYICLES OF INTERROGATION

Fig. 9-Curves showing the walking which takes placebefore a stable nondestructive level is reached.

WRITE--INTERROGATE

RESET -ERASE- 4 'V=AE NON-DESTRUCTIVE

_Xn gEC/DI. WRIT ST READ

INTERROGATE IT = REAIDRIVE

3TURNSA SEACH END TIP OUTPUT

(a) .*-- -~~~~~~~~~~~~~~~~ - ~~~DESTRUCTIVE-OUTPUT VOLTAGE. 0--o- A ~~~~~~~~~~READ~~OUTPUTVOLTAGE. ~~WRITE A WRITE~_

--A0.5 VOLTS/DIV.ERSEAE______ __ __ ___CURRENT = i EAE11 ERS200MA/DIV. C--~~~~~~~T= 0.20

/ASEC/ DIV.OUTPUT OUTPUT

INTERROGATE RESET =100 MA D.C. \INTERROGATE ( hINTERROGATE DRIVE SHOWN ABOVE DRIVE(a b

VOLTA G E ~~~~~~~~NON-DESTRUCTIVEREAD

0.5 VOLTS/DIV.CURRENT=200 MA/DIV. 0.---NON-DESTRUCTIVE

T = 0.20 ~~~~~~WRITE 0-READUSEC-/DIV. ~~~~~~~~~~~ERASE

284 IEEE TRANSACTIONS ON ELECTRONIC COMPUTERS

Fig. 10(a). The source is capable of delivering a fixed this type of flux source is shown in Fig. 10(c). Here a fluxamount of flux to an output leg or output winding at source of this kind has been placed at both ends, andhigh rates of speed without significant heating and four series connected balanced circuit interconnected be-without time limiting. Consider an element with the tween.initial flux pattern shown in which areas A, B and C are In constructing a store it is proposed that an inhibitequal. Then energization of the write winding will send logic type of addressing, [3 ], [5 ], be placed between theB from neutral to saturation, and send C from satura- flux source and the words. In this case, an electrical con-tion to neutral. Leg A will merely be shuttled. Subse- nection will be made between one high-speed flux sourcequent energization of the erase winding will send C and a multiplicity of words connected to a common loop.from neutral to saturation, sending B back to the con- The address input will inhibit all sheets except one, anddition shown in Fig. 10(a). Thus, drive of B sets up C that sheet will accept flux from the flux source in anand vice versa. Under these circumstances, the output amount dependent on whether erasing ornondestruc-leg is oscillated back and forth a fixed amount, depend- tive interrogation is to be accomplished. Initial experi-ent on the structure geometry, and yet the drive current ments suggest that this can be accomplished satisfac-need not be time limited. More important, no leg goes torily.from negative saturation to positive saturation, thusreducing heating. Some reflection shows that many ACKNOWLEDGMENTdifferent area combinations are allowed. For example, The authors are indebted to J. A. Young, J. E.with A = 48, B =20, C = 40, leg C will be oscillated 15 Schwenker and T. H. Crowley of Bell Laboratories,per cent of the distance from negative to positive satura- Murray Hill, N. J., for their encouragement and sug-tion. Such a flux source has been constructed and leg C gestions. The authors are also much indebted to F. J.lengthened so that series connected balanced circuits Schnettler, F. R. Monforte and W. W. Rhodes of thecan be inserted. Destructive reading and writing using Metallurgical Research Department for their coopera-irreversible switching has been performed in such a tion and to J. N. Brown and J. C. Stuart of Bell Labora-structure at a 2 Mc rate without significant heating. tories, Burlington, N. C., for a substantial developmentThe material was MInMgZn ferrite with a coercive effort in connection with the series-connected balancedforce of about 0.20 oersted. The drive was approxi- circuits shown.mately three times the threshold as measured from a1/7- vs H characteristic. Flux reversal was achieved in REFERENCES50 nsec. These fast switching times with moderate over- [11 M. Karnaugh, "Pulse switching circuit using magnetic cores,"drives were possible because flux reversal is accom- PROC. IRE, vol. 43, pp. 570-584; May, 1955.

[2] U. F. Gianola, "Integrated magnetic circuits for synchronousplished by many walls moving at modest velocities sequential logic machines," Bell. Sys. Tech. J., vol. xxxix, pp.rather than a few walls moving at high velocities. The 295-332; March, 1960.

[31 R. M. Averill, P. S. Kopel and E. E. Newhall, "A word organizedlow velocities do not require excessive overdrives. The memory which uses a guided flux for reading and writing,"flux level oscillated back and forth was of the order of presented at Solid State Circuit Conference, Philadelphia, Pa.;

1961.2 maxwells. [4] E. E. Newhall, and J. R. Perucca, "The Use of Balanced Magnetic

This high speed source might be combined with a non- Circuits to Achieve Energy Gain and Directivity," Presentedat Solid State Circuit Conference; 1963.

destructive-read source as shown in Fig. 10(b). Here [5] E. E. Newhall, "The Use of Balanced Magnetic Circuits tothe identity of legs A and C is maintained, as is the Construct Digital Controllers," Presented at International

Magnetics Conference; 1963.identity of the erase and write windings. However, leg [6] N. Menyuk and J. B. Goodenough, "Magnetic materials forD has been placed in parallel with leg B. The pattern digital computer components," J. Appl. Phys., vol. 26, pp. 8-18;

1955.after energization of the erase winding would then be as [7] D. Nitzan, "Flux Switching in Multipath Cores," Stanfordshown in Fig. 10(b). Writing could then take place as Res. Inst., Menlo Park, Calif., Rept. No. 1; November, 1961.

[8] D. Nitzan and V. W. Hesterman, "Flux Switching in Multipathusual, sending leg C to neutral, and B to saturation. If Cores," Stanford Res. Inst., Menlo Park, Calif., Rept. No. 2;leg D is energized some reflection shows that as long as November, 1962.

9I H. D. Crane, "A high-speed logic system using magnetic ele-leg A is held saturated, then the flux switching in D will ments and connecting wire only," PROC. IRE, vol. 47, pp. 63-73;switch in C, sending C halfway from neutral to negative January, 1959.

[10] R. H. Tancrell, "Impulse selection for core logic," J. Appl.saturation, as is desired if nondestructive readout is to Phys., suppl.; March, 1961.be accomplished. Subsequent energization of the write [11] J. A. Baldwin, "Circuits employing toroidal magnetic cores asanalogs of multipath cores," IRE TRANS. ON ELECTRONICwinding will re-establish a neutral condition in leg D COMPUTERS, vol. EC-11, pp. 218-223; April, 1962.and a neutral condition in leg C. Another nondestruc- [12] JT N. Brown, Jr. and E. E. Newhall, "The Storage and Gatingof Information Using Balanced Magnetic Circuits," Interna-tive-read cycle could now be initiated. An element using tional Magnetics Conference, Washington, D. C.; April, 1964.