E. J. SMITHMEMBER AlEE

A Transient Analyzer For Magnetic

Amplifiers

Fig . 4 it is seen that for Vl=O a certainoutput VM, small but not quite zero, wasmetered. In fact, with VSl = 0, no firingof the converter occurs in its gating half­cycles, but still the flow of magnetizingcurrent through Rc gives a certain con­verted signal v'c to the multiplier ; thelatter's area of least ordinates AVI is notbrought to zero, and the multiplier doesfire at some time near the end of its gatinghalf-cycle with an output substantiallyindependent of V2• A direct use of theoriginal circuit of Fig. 3 had caused aneven larger error in the VM reading be­cause of the drop of pre firing magnetizingcurrent through RM • To avoid this, RM

was displaced in the position shown in thesketch of Fig. 4; in this manner its dropof magnetizing current over the range oflow multiplier outputs nearly averages tozero in consecutive half-cycles. In par­ticular, for any VI, zero output is truly ob­tained when V2=0. In the experimen­tation the rectified signal voltages wereobtained from two variable transformerseach feeding through a half-wave recti­fier into a resistor Rs l (or Rs2) connectedbetween ground and signal rectifier 0"'1

(or O"S2) . (This is needed to provide apath for the presetting magnetizing cur­rents flowing against the signal voltages.)In the experimental circuit Rsl and R.2

were chosen as Rat= RS2 = 2,000 ohms. Asa consequence, at full output the powerlevel of either signal source was about 1/4

Synopsis: The transient analyzer describedis a simple and reliable electronic devicewhich operates by generating a number ofpulses exactly equal to the number ofcycles required for the output current orvoltage of a magnetic amplifier to changefrom its initial value to a predeterminedvalue. The response time in cycles is ob­tained directly by recording the output ofthe analyzer with an electronic counter orconventional recording device,

THE RESPONSE TIME of a mag­netic amplifier is defined as the time

required for the rms or rectified average

of the gross output power. Since it isgenerally desirable to reduce the over-allsignal power requirements, tests werealso performed with RSI =RS2= 10,000ohms, with results not very different fromthose shown in Fig. 4, except for lowvalues of V2• (At low values of V2 thewave form of V.2 was significantly dis­torted from the sinusoidal by the drop ofmagnetizing current.) Also with RSI =

R.2=2,000 ohms, voltage measurementsdirectly at the transformer terminalscorresponded substantially to the readingsof the meters VI and V2 across the resistors.This was no longer true for Rsl = Rs2 =10,000 ohms ; in this case the experi­mental results plotted in terms of trans­former voltages departed badly fromlinearity over the first one-third of thescale of the abscis sae .

A more proper design of the circuitcomponents would improve all this asthe magnetizing currents (about ± Imilliampere) could be reduced by the useof cores of lesser mean length (the availa­ble cores had more than twice the diam­eter needed to accommodate the wind­ings used .) The voltage drops in ques­tion could also be made less significantpercentagewise by the use of higher gateand signal voltage levels, allowed by thecore cross section but limited in the ex­perimentation by rectifier ratings. Andlarger power outputs could have been ob­tained from a reduction of R M • But the

No Discussion

value of output current or voltage tochange a prescribed percentage of thedifference between the correspondinginitial and final steady-state values.When the magnetic characteristic of thecore material approaches the rectangularB-H-Ioopshape, the wave form of the out­put current exhibits the typical rapid riseas one core saturates at some angle aMA,

and sinusoidal form after saturation,until the end of the half cycle . Underthese conditions, the peak value of loadcurrent is independent of the average orrms value when

circuit described, while performing theoperations assigned remarkably well, isnot primarily a power amplifier.

Nevertheless a variety of means isavailable to circumvent this question.The operation from low-level signalsources can be made possible by means ofinput cathode followers. Also, if docsignal voltages (and square-waved gatevoltages) are available, no need is feltfor the resistors Rs l and R.2 as the magne­tizing current then can flow directly intothe signal source.

Finally, one may aim more directly ata reduction of the magnetizing currentsin question . Since the magnetomotiveforce required by a core over intervals ofincreasing or decreasing fluxes is prac­tically constant, it may not be difficultto provide a large part of it by an auxiliarywinding fed from some suitable currentsource during presetting and prefiring or,more simply, during presetting only.Procedures of this kind are mentionedin the references.

References

1. ON TH E MECHANICS OF MAGNETIC AMPLIFIBROPERATION, Robert A. Ramey. AlEE Transactions.vol. 70, 1951 , pt . II , pp. 1214-23.

2. ON TH E CONTROL OF MAGNETIC AMPLIFIERS.R. A. Ramey. Ibid. , pp , 2124 -28.

3 . THE SINGL E· CORE MAGNETIC AMPLIFIER AS ACOMPUTER ELEMENT, R. A. Ramey. AlEE Trans­actions, vol. 71, pt. I , 1952 (January 1953 section).pp. 442-46.

7r

a'\{A~2

Unfortunately, conventional recordinginstruments are better suited to the meas­urement of cyclic peak values, and can beadapted to the measurement of the rmsor average value response time only atthe expense of considerable inconvenienceand difficulty. An electronic cyclic inte­grator which permits direct measurementof the rectified average value of the output

Paper 53-283, re commended by the AlEE MagneticAmplifiers Committee and approved by the AlEECommittee on Technical Operations for presenta­tion at the AlEE Summer General Meeting, At­lantic City, N . J.. June 15-19, 1953 . Manuscriptsubmitted Marcb 12, 1953; made available forprinting April 23, 1953.

E . J. SMITH is with the Polytecbnic Institute ofBrooklyn, Brooklyn, N . Y.

The study of Magnetic Amplifiers is being under­taken at the Microwave Research Institute of thePolytechnic Institute of Brooklyn, under thesponsorsbip of the Office of Naval Research contractN6ori-98 , task order IV .

The assistance of Mr. Victor Boros is gratefullyacknowledged.

SEPTEMBER 1953 Smith-Transient Analyzer for Magnetic Amplifiers 461

tel

: PHASE(}AOJUST

tDI(E)

(Although these remarks apply only toa single-ended magnetic amplifier, themethod itself can be applied to a push­pull amplifier by comparing the satura­tion angle of one reactor with the refer­ence angle.)

The function of the transient analyzer isto compare the angles aMA and an everycycle, and to indicate the result by gen­erating an output pulse every cycle ifaMA>an, but no pulse if aMA<O:li for abuild-up transient; and by generating apulse every cycle if aMA<aR, but nopulse if aMA>an for a decay transient.The transient analyzer becomes operativeonly after the switch initiating the tran­sient is thrown. Therefore, the numberof pulses generated by the analyzer afterthe switch is thrown is equal to the num­ber of cycles required for the magneticamplifier output to change from the initialvalue to the reference value. The re­sponse time of the magnetic amplifier isobtained directly by counting the numberof pulses generated by the analyzer duringthe transient. An electronic counter ofthe conventional type (multiple decadecounters employing Eccles-Jordan flip­flop circuits, as used in electronic digitalcomputers) was found to be a very con­venient means for counting the pulses ;however, a standard-type recording oscil­lograph could be used equally as well,but with perhaps less convenience.

The comparison of the angles aMA

and an is made during one half of the cycle

and when the output current is greaterthan the reference value

tfl

Fig. 1. Methodof determinationof transient re­sponse of a mag-

netic amplifier

A . Arrdngementof megnetic em­plifier end instru­ments for per­forming measure-

mentB. Tvpicel out­put weve form ofd mdgnetic ernpli­fier during d tren-

sient

(B I

Block diagram of the transient analyzer, illustrating wave forms at various points

(AI

(3)

IS)

tKI

Fig. 2.

aniscalled the reference angle. When thecore materials are not ideal the currentwave does not jump abruptly at satura­tion; however, if aMAis taken as the angleof maximum slope of the wave, the opera­tion is the same as though the materialswere rectangular.

The transient analyzer generates avoltage pulse of very short duration onceevery cycle . The phase of this referencepulse is next made to coincide with thereference angle by a manual adjustment.When the output current is less than thereference value

REFERENCEADJUSTMENT

(ORI

IIII

6

AC POWERSOURCE

INITIATETRANSIENT

INPUT

OUTPUTOF MA.

I REFERENCEII

°R 1

PRINCIPLE OF OPERATION

The operation of the transient analyzeris best described by making reference toFig. 1. The values of control-circuitvoltage E ChEc2, or current correspondingto the desired initial and final steady-statevalues of load current or voltage are firstestablished. The output of the magneticamplifier is then set at the reference value(that is, 63, 90, etc. per cent of the totalchange) by applying a suitable signal tothe control circuit. Corresponding tothe reference value of output, the magne­tic amplifier saturates at some angle

Description of Transient Analyzer

taken over each cycle has been previouslydescribed.F' With this method, theresponse of the amplifier in terms of cyclicaverage values is portrayed on the screenof a cathode-ray oscilloscope, is photo­graphed, and the response time deter­mined in the usual manner.

The device to be described in this paper(called the transient analyzer) is alsoelectronic, but is very simple and reliableas compared with the above device, andpossesses accuracy of a high order. Theoutput of the transient analyzer is a seriesof pulses equal in number to the number ofcycles required for the load current tochange from its initial value to any arbi­trarily specified value, called the refer­ence value. The reference may be speci­fied in terms of rectified average or rmsvalues by the use of a suitable meter, byperforming a manual adjustment priorto the initiation of the transient.

462 Smith-Transient Analyzer for Magnetic Amplifiers SEPTEMBER 1953

CI3 (Ie)

F=---gUTPUT

1R28

T5

~------------,

i~ : :

10 I I

IFROM IICONTROL TO II MA I

'I I

~9 ~O~ML------~THROW FOR:oECAY­BUILD UP-

[sSj

R7

T2

R2b9 R2\ 1 CI1. R2

R6

RI

(8)

RI7

C9

T3

R3C3

TIB+

(~) ~f1rof'RI C2 - _.!INPUT R 3

}R4

CR - +~I]-= -=-

B-o---------.----+--~+---~~_t_----------

A-C)II 400 60~""",_REF. = i<I (H

(FINE> =­REF ADJ.

B-O------------I-------4I~-------.._-------

Fig. 3. Circuit diagram of transient analyzer

(the negative half in Fig. 1). If the outputof the magnetic amplifier is alternatingcurrent, as shown in Fig. 1, the signalinto the transient analyzer can be ob­tainedfrom the load (that is, load voltage) .If the signal is full-wave direct current(rectified direct-current), then the signalto the transient analyzer must be takenas the voltage across one reactor winding(see waves in upper left corner of Fig . 2).In any case (that is, full-wave directcurrent, full-wave alternating current,half-wave, or push -pull alternating cur­rent or direct current), the signal canalways be taken from a winding of onereactor. The reason for the above isthat the analyzer operates by comparingthe reference pulse with some negativepulse derived from the magnetic amplifierwhich occurs only once during a cycle .This requirement is always met by takingthe signal from a winding on any corein the amplifier. The input impedanceof the transient analyzer is high enough(one megohm) to prevent detrimentalloading of the circuits .

The reference pulse phase angle lXR isreadily adjusted without the aid of anoscilloscope as follows:

1. The output of the magnetic amplifier isset at the reference value by applying anappropriate steady-state control signal, andobserving the output with a suitable a-c or

d-e meter.2. The transient analyzer is unclampedand the phase adjustment varied untilpulses are obtained from the output ter­minals. (The electronic counter countscontinuously as indicated by neon bulbs.)3. The phase adjustment is then varied(coarse and fine adjustments) until theoutput from the analyzer just ceases. Thereference angle lXR is now adjusted properly.4. The analyzer is again clamped, thecounter reset to zero, and the transientinitiated by throwing the switch. At thecompletion of the transient, the reading ofthe counter is exactly equal to the responsetime in cycles.

DESCRIPTION OF CIRCUITS

The internal operation of the transientanalyzer can best be explained by refer­ence to the block diagram of Fig . 2. Theinput signal A is clipped,* differentiated,and amplified, resulting in a sharp nega­tive pulse of short duration B, whichoccurs at phase angle lXMA' An a-c voltageC having the same frequency as themagnetic supply is passed through aphase shifter, which is equipped withcoarse and fine adjustments for varyingthe phase. The phase shifted a-c wave D

*The negative half cycle is removed if the negativedrop in the signal occurs on the positive half cycle(input from a reactor winding) and the positivehalf cycle is removed if the negative drop occursduring the negative half cycle. See (A) in Fig. 2 ;for circuit details see Fig. 3.

is clipped E, differentiated, and ampli­fied, resulting in a sharp negative pulseof short duration F; this is the referencepulse. The phase of the reference anglelXR is determined by the phase adjust­ments previously described. The twopulses Band F are fed into oppositeplates of a conventional Eccles-Jordanflip-flop circuit. The operation of theEccles-jordan! circuit is such that a nega­tive pulse received at one plate triggersthe stage, causing that plate voltage todrop to its low value and the voltage ofthe opposite plate to rise to its high value.The circuit then remains in this condi­tion until a negative pulse is received atthe plate of the second (opposite) tube, atwhich time the plate voltage values of thetwo tubes reverse.

The voltage from one plate of the flip­flop circuit G is used to control the gatestage. When this plate is high the gateis open and allows to pass through it asignal appearing at its input terminals 1.When this plate is low the gate is dosedand the incoming signal I is blocked .

The two input connections to the flip­flop circuit are interchanged for build-upand decay transients. For a build-uptransient, the pulse derived from themagnetic amplifier B is connected totrigger the flip-flop in such a direction asto open the gate, and the reference pulse

SEPTEMBER 1953 Smith-Transient Analyzer for Magnetic Amplifiers 463

spending to A2. GI and G2 correspondto signals A2 (or A3) and AI, respectivelyfor a decay transient in which the refer­ence angle F lies in between these twolimits. The magnetic amplifier sourceis 20 volts rms, and is shown in all oscil­lograms to the same scale.

Fig. 4. Oscillograms of voltage wave formsat the various points in the circuit and block diagrams

The primary aim in developing thetr ansient analyzer was to keep the cir­cuitry as simple as possible, and yet pro­vide a device which would be suitable fortesting a wide ran ge of magnetic ampli­fiers. Certain limitations exist ; thesewill be brought out in the following.

The sensitivity of the circuit was foundto be sufficient to allow satisfactory opera­tion on signals obtained from typical 10­volt 60-cycles-per-second (supply voltageand frequency) half-wave and full-wavemagnetic amplifiers, using Mumetal andHipernik-V core materials. For muchsmaller signals, an additional stage of pre­amplification would be necessary. Whilethere is virtually no upper limit to theamplitude of th e signal that can be ac­cepted, it may be necessary to restrictthe amplitude in certain cases, where dis­torti ons or spurious pulses exist in theoutput of th e magnetic amplifier wave ;thi s can always be done with a high­resistance voltage-divider netw ork .

The logic of the circuit operation re­quire s that the input from the magneticamplifier be sufficient to trigger the flip­flop stage for all values of aMA realizedduring the transient. This condition be­comes difficult to fulfill when either theinitial or final value of CiMA approaches '11',

even though the signal strength may beadequate over the major portion of thetransient. While the logic of the designcould be changed to take care of thissituation, the additional circuit com­plexity makes the change undesirableand contrary to the original aim of sim­plicity. Furthermore, the case for whichCiMA approaches 'II' is a very unusual one,and its exclusion does not impose a seriouslimitation on the usefullness of the device.At this point it may be asked whetherthe aforementioned limitation on a ,VA

does not exclude the application of thetransient analyzer to push-pull magneticamplifiers of the null-balance type, inwhich the initial or final steady-stateoutput is usually zero. The answer isnegative, because with such circuits thesignal is obtained not from the output,but from a winding on one core . Hence,even at null balance, the signal CiMA

is not likely to approach '11'.

The accuracy of the transient analyzer

Operating Experience and RemarksG2

and F pulses are interchanged by throwingthe reversing switch shown in Fig. 2.Th e operation of th e transient analyzerthen proceeds as prev iously described.

The complete circuit diagram of thetransient analyzer is shown in Fig. 3.Conventional pulse-circuit techniques areused ; therefore , no detailed descriptionis necessary . The particular circuit de­sign shown is intended primarily for 60and 400-cyc1es-per-second operati on, butcould be modi fied for higher frequencyuse with minor changes. Wave forms ofvoltages at various points in the circuitare shown in Fig. 4. (Oscillograms BandF are obtained with the tube T2 removedfrom its socket.) The letters A to Krefer to the circuit points shown in Figs .2 and 3. Al and A2 are the outputvoltages across the load of a Hipernik-Vfull-wave a-c amplifier for two values ofcontrol signal ; A3 is the voltage acrossthe load windin g of one reactor corre-

c

th e gate will be closed when the sensingpulse I arri ves at the input to the gate ;under these conditions, no pulse appearsat the output of the transient analyzer .For a decay transient, the roles of the B

the gate will be open when the sensingpulse I arrives at the input to th e gate ;therefore, a pulse appears at th e outputJ and K of the device. On the other hand,during a build-up transient, if F occursafter B, th at is

It follows immediately that if B occursafter F, that is

F is connected in such a direction as toclose the gate. The signal input (calledthe sensing pulse) to the gate I occursat angle a" where

464 Smith-Transient Analyzer for Magnetic Amplifiers SEPTEMBER 1953

AlEE COMMITTEE REPORT

Water Cooling Systems of Mercury-ArcRectifiers

depends upon how accurately the refer­ence angle etR can be set according to theprocedure previously given . In everycase tested, the reference adjustment(the point at which the "counting" justceases) could be set as accurately as themeter (indicating output current or volt­age) could be read .

Appendix. List of Compcnents-:Transient Analyzer

Tubes T1 to TS-6SN7Transformer X 1-120 volts primary,

80 volts center tappedsecondary, 60 and 400cycles per second

Potentiometers . .P1-850 kilohms, 4 wattsP2-20 kilohms, 4 watts

Crystal Diode . . . CR-IN34

Synopsis: This report is intended as aguide on the design, application, andoperation of water cooling systems formercury-arc rectifiers. Its contents arebased on the accumulated experienceacquired by rectifier manufacturers andusers over many years. The report dealswith the various engineering and operatingaspects of rectifier cooling systems . Severaltypes of water cooling systems are de­scribed, and their application is discussed.The various factors influencing corrosionare reviewed. The characteristics of cool­ing waters are discussed and recommenda­tions are given on the chemical treatmentof coolants, the application of antifreezesolutions, and the maintenance require­ments of cooling systems . The results of asurvey on the operation and maintenanceof cooling systems in representative rectifierinstallations are also included .

THER E are two general types of mer­cury-arc rectifiers in operation in the

United States and Canada: the multi­anode type, with 6, 12, or 18 anodes witha common mercury cathode in one tank;and the single-anode type, with one anodeand its cathode in an individual tank.Nearly all new rectifier installations madein the last 10 years or longer are of thesingle-anode ignitron or excitron type.

Switches Sl, S2, S4, SS, S6-dou­ble-pole-double-throw, noneutral positionS3-single-pole-double­throw, neutral position

Power Supply . .. B +, 150 volts, 40 milli­amperesB -, 100 volts, 10 milli­amperes

Resistors (1 watt, carbon, ±5per cent)

R13, RH, R28 10 kilohmsR3, R17, R2S 18 kilohmsR6, R7 24 kilohms

(matched with­in 2 per cent)

RS, R19, R27 100 kilohmsR23 150 kilohmsR8, R9 and RlO, Rll 240 kilohms

(matched with­in 2 per cent)

R2, R1S, R21, R22 .470 kilohrnsR1, R18, R20, R24, R26 . . 1 megohmR4 . . . . . . . . . . . . . . . . . . . •1. 3 megohms

No Discussion

For this reason, the report refers pri­marily to the single-anode type of recti­fier; however, its contents apply gener­ally to both types.

A standard rectifier consists of anassembly of 6 or 12 tubes with the aux­iliary equipment. The rectifier coolingsystem, which is the subject of this report,is one of the essential auxiliaries, and itsperformance is an important factor in theoperation of the rectifier.

The flow of current through a mercury­arc rectifier tube is accompanied by avoltage drop called the arc drop, whichmay have a value in the range of 15 to25 volts, depending on the tube designand rating. The product of the currentand arc-drop voltage is a power loss whichis converted into heat. Most of the heatis transferred to the cooled surfaces of thetube by radiation and by condensationof mercury vapor.

For satisfactory operation, the heatgenerated by the losses has to be removedand the tube temperature maintained inthe optimum operating range. Thetemperature of the cooled surfaces con­trols the mercury vapor density in thetube. Excessive temperatures are likely

R12, R16 2.2 megohmsCondensers (±1O per cent)

Cll 56 micromicrofaradsC4, CS 100 micromicrofaradsC9 230 micromicrofaradsC12 360 micromicrofaradsC3, ClO 0 .001 microfaradC2 0 .002 microfaradC8 0.003 microfaradC6 0 .007 microfaradC7 0 .05 microfaradC1, C13 0 . 1 microfarad

ReFerences

1. A CYCLIC INTEGRATOR, S. Roman. Thesis,Polytechnic Institute of Brooklyn, N . Y ., 1951.

2 . A STUDY OF MAGNETIC AMPLIFIERS. PAKTIV-TRANSIENT RESPONSE, E . ]. Smith. ReportR-23850. PIB-183. M icrowave Research Institute,Polytechnic Institute of Brooklyn, N. Y.

3. THEORY AND ApPLICATIONS OF ELECTRONTUB ES, H . ] . Reich. McGraw-Hili Book Com­pany, New York. N. Y ., 1944.

to cause arc backs. If the temperatureis too low, the arc-drop voltage may in­crease and voltage surges might be pro­duced due to arc starvation, when themercury vapor density is insufficient forconduction of the tube current.

Glass-bulb rectifier tubes and somemetal-tank tubes are cooled by air forcedover the tube surfaces, with a fan orblower. However, most of the rectifiersin the United States and Canada arecooled by circulation of water throughwater jackets or tubular coils. Some tubedesigns have copper tubing soldered onthe outside surface; others have internalsteel coils . Combinations of externaland internal coils or of internal coils andexternal jackets are also used. Thewater-cooled parts of sealed tubes areusually made of stainless steel.

Rectifier Cooling Systems

The following four types of water cool­ing systems are used for mercury-arcrectifiers : direct raw water cooling sys­tern; direct raw water cooling system withrecirculation; heat exchanger cooling

Paper 53-288, recommended by the AlEE Elec­tronic Power Converters Committee and approvedby the AlEE Committee on Technical Operationsfor presentation at the AlEE Summer GeneralMeeting, Atlantic City, N . ] ., June 15-19, 1958.Manuscript submitted March 17, 1953; madeavailable for printing April 27, 1953.

The personnel of the Working Group on WaterCooling Systems of Mercury Arc Rectifiers of theAlEE Committee on Electronic Power Convertersare : E.]. Remsheid, Chairman, Mrs. B. O. Buck­land, A. P . Colaiaco, A. G . Dickinson, G. N.Hughes . J. Klotz (alternate), J, B. Nayler, R . J.Nix, M. W. Rew, Mark Sareault, and HaroldWinograd.

SEPTEMBER 1953 Water Cooling Systems of Mercury-Arc Rectifiers 465