172 PHILIPS TECHNICAL REVIEW Vol. 5, No. 6

A NEW PUSH-PULL AMPLIFIER VALVE,FOR DECIMETRE WAVES

by M. J. O. STRUTT andA, van der ZIEL. 537.543 :621.385

,For the reception of signals with wave lengths shorter than 1 metre electro~ic valves areneeded with very low input and output damping, steep slope and Iow noise resistance.The socalIed acorn pentode satisfies the first requirement but fails to satisfy the lasttwo. 'In order to satisfy these last requirements electrode systems with larger -dimensionsmust he used. At frequencies above lOBcycles/sce., however, such systems possess toogreat input and output damping. It is shown in this article that the imput and outputdamping of such systems can be considerably lowered by connecting two systems in apush-pull circuit in a suitable way. After an explanation of the construction andproperties of the push-pull amplifier valve EFF 50 developed on this principle, the pos-sibilities of its employment are dealt with. In particular a discussion is given of the resultswhich can be obtained upon application of this valve in amplifying and mixing stages.

Methods of measurement have been developedin this laboratory in ~ecent years by which curr~nt~,voltages and impedances at very high frequencies(up to about 1.5 . 109 cjs. corresponding to a wavelength of 20 cm) can' be accurately' determined.These measurements have in the first place beenapplied to electronic valves in order to investigate,the behaviour of these valves in detail I). On thebasis of the insight obtained, new valves have been'developed which give better results at very highfrequencies than those used until now. After anexplanation of the limitations in the applicationof the existing electron valves, the constructionand properties of one of these new types of valves,namely the push-pull amplifier valve. EFF 50, IS

discussed in detail. '

The functi~n of an electron valve

The electron valve serves in general to amplifythe incoming signal and to transmit it in some formor other to the following stage. In the case of ahigh-frequency amplifier, for instance, the require-ment is that the character of the voltage transmittedshall be a faithful enlarged image of the incomingvoltage within a' certain frequency region. In the. case of a mixing stage, on- the other hand, it isdesired that the modulation of the transmittedvoltage shall have the same character as the mod-ulation, of the input voltage, while the characterof the voltage itself 'is quite different; the latter has,namely, a much lower frequency' than the inputsignal.

In addition to the quantitative requirement of asatisfactory enlargement of th~ amplitudes, thereis the qualitative requirement' that the desiredrelation between output voltage and input voltage. shall be realized as fully as possible. According to

1) M.,J. O. Strutt: Moderne Kurzwellen-Empfangstech-nik, Verlag Springer 1939;ModerneMehrgitter-Elektronen-röhren, 2nd edition, Verlag Springer 1940.

•whether the amplitude of the input voltage is verylarge or very small, two kinds of deviations may,here occur. At very great amplitudes due to the

. curvature of the valve characteristics, the relationbetween input and output voltage, ~r between inputand output modulation, deviates from true pro-portionality, or, in other words, distortion occurs.The relation is thus indeed unambiguous, but doesnot have the desired form. At very small amplitudesthe reserve is true: the above mentioned distortiondoes not occur, but the relation between inputand output voltage is no longer absolutely jsharp:with á fixed input voltage the output voltage maystill exhibit certain fluctuations, which are caused,by incidental phenomena in the valve. If thesefluctuations are of the same order of magnitude asthe-signal voltage obtained at the output, there canno longe)' be the least question of a relation betweeninput and output voltage. These fluctuations, whichare. usually called noise, thus set a lower limit .tothe input voltage at which the valve can be used 2).Since this limit is found to be of essential signifi-cance for the usefulness of a receiving installation,we arrive at the conclusion that, as far as thefunction of an' electronic valve is concerned, thevalve must amplify. as ml~cli as possible and give aslittle noise as possible.In the following we shall see how far valves

of ordinary construction are capable of fulfillingthis function.'

Specification of the requirements made of electronicvalves

The amplification of an electronic valve can'be expressed quantitatively by four quantitieswhich 'connect the output voltage and current withthe input voltage and current. This is done. by

2) The factors which determine the noise of a receiving setwere dealt with in detail in a previous article (M. ZiegIer,Philips techno Rev; 3, 189, 1938).

JUNE 194·0 PUSH-PULL AMPLIFIER FOR DECIMETRE WAVES 173

means of the equations:

~a - A Vg + B Va ~

Lg = CVg + D Va ~

which have been given and discussed' previouslyin this periodical "]. A, B, C and D are in generalcomplex quantities depending upon. the frequency:they are called the characteristic admittances of anamplifier valve. The admittance A is nothing otherthan the slope ofthe valve, B is the output'admit-·tance, C the input admittance and D the reactionadmitta~ce. If the admittances are known, the am-plifying properties of the valve can be calculatedin every connection. The noise properties are how-ever not given thereby, and must be consideredseparately.The most efficient way' of characterizing the noise .

of an electronic valve is by the introduetion of theconcept of noise resistance. The definition of noiseresistance is based upon the fact that a "noisevoltage" ,appears in every resistance, i.e. that at itsextremities certain voltage fluctuations occur due.to the thermal motion of the electrons in the in-terior of the resistance. In effective magnitude thes~fluctuations are' proportional to the square root of· the resistance value and the tcmperature. By thenoise resistance R of an electronic valve is under-stood that resistance whose volta-ge fluctuations (atroom temperature) are of the same magnitude as

· those voltage fluctuations which would have to beintroduced between grid and cathode .of an elec-tronie valve in order to cause the anode currentto fluctuate as much as actually occurs due to thenoise of. the valve.If we now return to the function of the electronic

valve, it may be seen immediately' from the abovethat it is desirable tö make the slope A as steep aspossible and the noise resistance R as small as pos-sible. In order to understand the influence of theoutput admittance B and the input admittance C,we must consider the connection in which the valveis used. From' equation (1), however, it follo~simmediately that the real parts of these admittancesgive rise to energy losses at the output and inputends, which is undesired in principle, so that it is.important to make these admittances as small aspossibl~.

Special cases .

We shall confine ourselves here to the specialcase of high-frequency amplification, making adistinction between

· 3) See in thls' connection the article by M. J. O. St rut t,and A. van der Ziel, Philips techno Rev. 3, 103, 1938.

(1)

1) the high-frequency amplifier stage of a cascadeamplifier (two identical' stages following oneanother), and .

2) the high-frequency amplifier stage of a super-heterodyne receiver (which. is followed by amixing stage). 'We may begin by dealing with both together

with reference to the scheme represented in fig. 1.The input and output circuits are formed by tuned .oscillation circuits with the, impedances Zl and Z2'In parallel with Zl is the input impedance I/CI. Inparallel with Z2 is the output impedance. liB and,via the transformer T, the input impedance l/C2 ofthe following stage. Although I/ZI' I/Z2' Cl' BandC2 are in general complex, we shall in the followingalways confine ourselves to the real parts, since bycorrect tuning of the circuits the imaginary partscan, always be made to compensate each other.

'Fig. 1. Diagrammatic representation of a high-frequency,amplifier 'stage connected to a second stage via a trans-former T, Zl and Z2 are jhe impedances of the input and out-put circuits, respectively, of the first stage; Cl and C2 arethe input admittances of the first and secondamplifier valves,respectively, B is the output admittance of the first amplifiervalve. '

If the amplification, z.e, the relation betweenthe output voltage Eu and the input voltage Eiof the' given circuit, is calculated for, a transfor-mation ratio n; one finds:

An

This expression as a function of n has a maximumfor n2 = (B + I/Z2)/C2, and the maximum ampli-fication amounts to:

(2)

It is often possible to make Z2 ~ liB. On thiscondition:

Eu AEi 2 YBC2

If 2 VBC>A, the ratio Eu/Ei becomes less 'thanunity, and there is therefore no longer any ampli-

174 PHILlPS TECHNICAL REVIEW Vol. 5, No. 6

fication possible. If a given amplification is desired,for instance by a factor v, the condition" '

}'BC2 < A/2 v . . .... (3);

holds for the input and output admit'tances, and itmay therefore be seen that ,the two admittancesmay not exceed a certain magnitude.

As to the noise, it is more difficult to reach anumerical formulation of the requirements made.These requirements depend of course entirely upon

. the intensity of the available input signal. We may,however, state that a decrease in the noise of elec-~ronic valves is only of importance as long as thisdisturbance is not weak compared with the. noisedue to other sources. Besides the electronic valvesit may be seen in the figure that there are othersources of noise, namely the impedances Zl' I/Cl'u», Z2 and 1/C2•·

Let us first consider the noise from the first am-plifier valve and the first oscillation circuit. In fig. 1we see that between grid and cathode of the firstvalve the impedances Zl and I/Cl are in p~rallelwith each other, thus at the same spot as theimaginary noise resistance R. From this it followsthat the noise from this noise resistance R must besmall compared with the noise of the two impedancesZl and I/Cl connected in parallel. If this conditionis satisfied, the contribution of the first valve to thenoise is small compared to that of the other source~of noise mentioned. This requiremerrt thus indi-'cates an upper limit for the noise resistance R.If we now consider first the case where I/Cl islarge compared with Zl our requirement amounts to

When, however, Zl is large compared with 'l/CIone .would expect as condition

noise properties so obtained shall suffer as littledisadvantage as possible f~om the following stages,it is found, that- a ·lower limit must be .set to theamplification of the first stage. We must now makea distinction between the cases (1) and (2) men-tioned above. . 'ln case 1) ,(cascade amplification) the' input·

resistance and noise resistance of the second valveare about as large as those of the first valve, thesame is thus true of the fluctuations which occur inthe second stage .When we represent the total fluctuations of valve

plus circuit of each stage by a noise resistance Rt, thenoise resistance which, connected to the input ofthe first stage, represents the noise of both stagestogether, amounts to:

where v is the amplification. If v -has the value 3,for example, the total noise resistance at the inputof the first stage becomes only about 10'per: centgreater due to the presence of the second stage, andthe equivalent voltage fluctuations become only 5 .per cent greater, since these fluctuations are pro-portional to the square root of the noise resistance.In cascade amplification, therefore, an amplifi-cation of 3 must be considered as sufficient in con-nection with the noise.In the case of a mixing valve in the second stage,

the total noise resistance of this stage is generallyconsiderably greater than with a high-frequencyamplifier valve in the second stage, for .inst.ance, afactor 10 greater. If the requirement is again madethat the influence of the second stage onrhe noiseshall be appreciably less than that of the first stage,the .amplification must be .made considerablyhigher, 5 to io :times instead of 3 times. . .

Limitations of electronic valves at high frequencies'. .. ~From a more detailed consideraticn it is found that , In the foregoing we have stated the requirementsin the last condition Cl must be replaced by Cc, i:e. which must be met by the admittances of an elec-by that part of the input admittance which is due tronie valve if it is to fulfil its function. Passingto the transit time of the .electrons.. The calculation, .. on to higher frequencies. it becomes more andwhich we shall not reproduce here, shows that the more difficult to fulfil this requirement.product RCc is a measure ..of the maximum value As has been explained in detail in .the article'of the ratio between signal voltage and noise voltage quoted about the behaviour of electronic valves,which can be obtained by adapting the first stage while the slope A remains .unehanged, at leastto a given aerial. The larger this product, the less in absolute value, up to veryhigh frequencies, bothuseful the valve. the input and the output admittances (we refer

Until now we have given the requirements f?r always to the real parts) increase steadily withmaking the noise due to the first valve as small as increasing frequency, usually proportional to thepossible compared with the noise from the input square of the frequency. The result is that thecircuit. If we now make the ..requirement that the left-hand side of equation (3) becomes steadily

JUNE 194,0 ' PUSH-PULL AMPLIFIERS FOR DECIMETRE WAVE 175

larger with increasing frequency, so that the'equation no longer holds for valves of ordinaryconstruction for frequencies above about 108cycles/sec.The cause of the undesired increase in input

and output admittance lies, as explained in thearticle referred to, in two phenomena: '1) the finite transit time of the electrons hetween

the electrodes in the valve;2) the self-inductions and mutual inductions 'of the

connecting wires of the various electrodes.If it is desired to keep the input and output

admittances small at high frequencies" therefore,it is necessary to combat these two phenomena.

The acorn pentode

, A very 'ohvious method of decreasing the induc-'tive effects and the transit, times' is to diminish

'the dimensions of the valve. The result of thetechnical development which took place 'in thisdirection several years ago is the so-called acornpentode, whose admittances are given in .fig. 2 asfunctions of the frèquency.

The acorn pentode may he used for amplification 'purposes up to about 2 X 108 c/s. (1.5 m). At'this frequency the input admittance e is 1/4 000 n,the output admittance ,B = 1/10000 Q and the'slope A = 1.4 mA/V. With the help of equation (3)it may be calculated that at this frequency an am-plification is possible by a factor

, v = A/2 JIBê = 4:6.If, in agreement with experience, it is assumed that

~

5

3

V11 rf

~-.

III I31/

.'7

1 I/ II'/

2

5

2

5

3

2

, Fig. 2. Input-resistance (curve 1)and output resistance (curve II)of the acorn pentode 4 672 as function of the frequency in cjs.

the admittances increase with the square of thefrequency; then for the case of cascade amplification(v = 3) an upper limit of about '2.5 . 108 c/s.(1.2 m) can be calculated. ,The noise resistance of the acorp.pentode amounts

to about 8000 ohms 4). The product Ree at 2 ' 108c/s is thus about 2, which may be consideredtolerable. With increasing frequency" however,this product rapidly increases, so that as far asnoise is concerned the application of the acornpentode is subject to about the same limitationsas set by the amplification.A f';uther reduction of the dimen~ions of the valve,

which theoretically would mea~ an improvement, isat present not yet possible technically. Moreoverit is 'very much a question whether, the expectedimprovement would be realized. It is essentialthat the decrease in size' of the valve should not. ,

result in a decrease in steepness of slope. Theoret-ically this requirement could be fulfilled, since theslope is entirely determined by the relations be-

, tween the dimensions. Actually, however, there area itumber of effects which, with decreasing dimen'-sions, have an unfavorable influence upon the slope

of a valve, so that the, ratio Adne will increaselittle or perhaps not at all llpon further reductionof the dimensions. In the present position of tech-nology the acorn pentode must be considered asthe most, satisfactory oompromise- as far as thedimensions of the valve are concerned, so thatit will not be possible to develop .a valve in thisdirection which is suitable for the amplification ofhigher frequencies than 2 . 108 el«.

On the principles discussed above- we shallnow point out a new line of development which willcarry us farther than the line employed until now'of decreasing all the dimensions.

A new type of high-frequency amplifier valve

'As we have seen, for frequencies higher than2 X 108 c/s; 1.5 m) it is desirable to have a valve,with a steeper slope and a lower noise resistance thanthe .acorn pentode. Valves with steep slope canbe constructed by applying various devices. One ofthese is the enlargement of the surface of the cathodewhile keeping the distance between the electrodesin the valve constant. This method is analogousto the connection in parallel of a' number of elec-

4) It must be noted here th~t the noise of a valve is actuallysomewhat stronger than is 'indicated by the above-men-tionednoiscresistance. With the high frequencies in questionthe intensity of the noise of the valve is a function of theimpedance between grid and cathode. The above-mentionedvalue of B000 .Q holds for a very small value of this im-pedance. .

176 PHILIPS TECHNICAL REVIEW Vol. 5, No. 6

trode systems. It therefore also leads to a decreasein input and output :resistance so that no advan-tage is gained. When this method is eliminatedthere remains as second possibility the method.used in the acorn pentode, namely by decreasingthe distance between: control grid and cathode,using at the same time a thinner grid wire and alower pitch. In this case the input and output resist-ance need not become smaller. Starting with elec-tronie valves of ordinary' construction we haveapplied this second device. The distance betweengrid and cathode was reduced to 90 fL, and the.slope thereby obtained: amounts in the normaladjustment of the valve to 11 mAjV at an anodecurrent of 10 mA.An increase on the slope. of a valve involves

automatically a lowering of the noise resistance'.The greater the slope the' smaller the amplitude ofthe 'grid A.C. voltage necessary to cause a givenfluctuation of the current. In the construction ofthe new high-frequency amplifier valve, however,an attempt was made to combat noise even morefully by decreasing the noise A.C. itself. This couldbe done be keeping the screen grid current low.

The fluctuations ofthe anode current in a pentode are causedpartly by the fact that the electron current. emitted by thecathode itself fluctuates, and for an,important part also theyare caused by the fluctuations of the current distributionbetween screen grid and anode 6). The latter fluctuationsmay be kept small by providing that the screen grid currentitself amount to only a small percentage of the total cathodecurrent. In this respect an improvement was obtained in theconstruction in question by taking thinner screen grid wiresand making the pitch of the screen grid gr~ater than is usualin ordinary high-frequency amplifier valves. This has indeedas result that the screen grid fulfils its shielding function lesseffectively, but it is just the amplification at very high fre-quencies which is found to suffer no serious disadvantage fromthis. . ,The screen grid has the function of shielding the anode from

the control, in order that a change in the anode voltage mayhave no influence on'the anode current (high output resistance}and on the control grid voltage (small reaction).In a valve of ordinary construction the output resistance

is of the order of üia.o at low frequencies.No advantage couldbe taken of such a high output 'resistance in thè case of the'amplification of very high frequencies, since the resistance ofthe tuned circuits connected in parallel amounts only to'

6) Sec in this connection the article by M. ZiegIer; Philipstechno Rev. 2, 329, 1937. In that article the phenomenawhich give riseto the noise in receivingvalves are discussedin detail. From the discussion given the following formulacan he derived for the noise resistance of a pentode:

R = 4·000 [1 +- 5 '182 ( Ia )].0,Sa Sa 182 + Ia

where the value of the slope Sa must be .suhstituted in, mAfV and the anode current la and the screen grid cur-rent 182 in m.á.:

several thousand .0. As to the static output resistance, there-fore, there is no objection to making the shielding hetweencontrol grid and anode less effective than is usual in valvesintended for lower frequencies.The same result is reached as far as reaction is concerned.

The reaction can he described by a capacity Gug betweencontrol grid and anode, and is kept as small as possibleby thescreen grid. At high frequencies, however, the reaction isfound to be caused to. a large extent, not by the capacitybetween control grid and anode, but by the mutual capacities,self-inductions and mutual inductions of the various connec-tions of the electredes. The effect of this additional source hasin many cases a greater absolute value and a sign oppositeto that of the effect of the capacity Gag. In these cases an.increase of Gng is immediately permissible and it even leads'to a decrease in the total reaction.

By the decrease of the screen grid current, to-gether with the increase in the slope, the. noise re-.sistance was reduced to 600 n, thus to about 1/.13of the value of the noise resistance of the acornpentode. For the smallest input voltage to be am-plified, which is proportional to ,the squ~re rootof the noise resistance, this means a gain by a factor3.5.As to the amplification, the improvement is

even more important compared with the acornpentode. The slope has increased from 1.4 to 11,.thus by' a factor 8, and when unput and outputimpedance are kept constant the same is true ofthe amplification.

The performance at high frequencies with theabove-described construction is, however, not to beconsidered satisfaètory. While the influence of thetransit times will indeed be only relatively slightthanks to the short distance between control gridand anode, the inductive 'and capacitive effects ofthe connection wires are no smaller than in ordinaryelectronic valves, and these may have the well-known damping effect on the input and outputcircuit.

It has been found by measurements and calcu-lations that this damping. in the case of ,valveswith s~eep slopes is mainly due to i?ductive effectsof the cathode connections (inside and outside the,valve) and is proportional to the slope. Since we arehere concerned with a valve of very steep slope it istherefore necessary to combat the influence of ·theinductive effects of the cathode connections.

The push-pull amplifier valve EFF 50

The elimination of the inductive' effects of thecathode connections is found to be possible by aspecial application of the push-pull principle. Twoidentical electrode systems fused together into asingle bulb are provided as shown in fig. 3a with acommon cathode co~nectio~. The voltage biases of

JUNE 1940 PUSH-PULL AMPLIFIER FOR DECIMETRE WAVES

A'H'D' f B f D"H/~II356.18

Fig. Cl a) Diagram of the electrode arrangement of the push-pull amplifier valve EFF 50. Two .identical pentodesystems with a common cathode connection Bh, andindirect heating (.ff) are mounted in an all-glassbulb.

00 k 0g' gilI 1

0gl 9 gP2 2

Oa' a"0f f0 0

b) Arrangement of the electrode pins in the base ofthe push-pull amplifier valve EFF 50.

corresponding electrodes are chosen equal, while theA.C. voltage of the control grid is supplied in oppo-site phase to the two halves of the system. Theanode A.C.'s of the two halves of the system will thenexactly cancel each other in part Bk of the cathodeconnection, so that inductive effects can only occurin the section k'kk" ofthe cathode connection, whichcan be kept very short.The input voltage acts over the terminals D' Dil;

the resistance between these terminals may there-fore be considered the input resistance. The inputresistance is obviously twice as high as that of asingle system. Moreover, due to the employment ofthe push-pull principle by which the connectionwire Bk is rendered ineffective, the input resistanceof each separate system is increased by a factor ofabout 2.5, so that a total increase of the inputresistance by a factor 5 is obtained. The remainingpart of the input damping must be ascribed chieflyto the transit time of the electrons between cathodeand control grid. The output resistance betweenA' and A", which is affected to a smaller degreeby the transit times of the electrons, amounts toeven more than five times that of a single valve;it amounts, for instance, to ten times.Another advantage of the push-pull principle

is that the input and output capacities of the wholevalve amount to only half of the capacities of eachsystem separately, which makes these capacitiesnot appreciably greater than in the acorn pentode.

Construction of the push-pull valve

A form has been chosen for the constructionof the push-pull valve which is also used for modernreceiving valves for ordinary broadcasting wavelengths, namely the so-called all-glass constructionwhich has previously been described in this period-ical "}. This may be seen in jig. 4 together withseveral developmental models of different construe-tion.

G) See the article: A new principle of construction Ior radiovalves, Philips techno Rev. 4, 162, 1939.

Fig. 4-,Acorn pentode 4672 (left) and a number of developmental models of the push-pullamplifier valve EFF 50. In the case of the first four of these models (second, fourth, fifthand sixth from the left) the systems are assembled end to end; ill the final model (thirdfrom left) the systems are side hy side. The two systems are surrounded not only by a glassbulb, but also by a common gauze cage which has been removed in order to show theassembly to better advantage. To the right a centimetre scale.

177

178 PHILlPS TECHNICAL REVIEW Vol. 5, No. 6

The connecting. leads of the electrodes in theall-glass construction are' all led .out through a flatbase plate. As may be seen ÎIl: fig. 3a the, push-pullvalve has nine connections, the arrangement of thenine pins in the base is ,shown in fig., 3b.

The connecting pins which are. fused into. the glass baseare often made of chrome-iron for reasons connected with. . . .glass technology. Due to the high magnetic permeabilityof this material considerable skin effect occurs at frequenciesof the order of lOB els and higher, so that the resistance atthese frequencies becomes several hundred times the resistancefor direct current. A decrease of the resistance can be obtainedby covering the surface of' the pins with a non-magneticmaterial, such as copper or silver.It is important to keep the resistance of the pins low: be-

cause this resistance contributes to the input and outputdamping. If, for example, we consider the input end, theadmittance between the connecting pins is given by the sub-stitution scheme sketched in fig. 5, 'where Ts represents theresistances and L, the self-inductions of the grid connections,while Ckg stands for the capacity between control grid andcathode in each system. The re~part of the admittance ofthis circuit is: ,7'

1 Tsoo2 CklRp 2 (1:- 002 LsCkg)2

and thus gives rise to a damping which is proportional to T.

and to the square of the frequency (rs also increases with thefrequency and is approximately proportional to' lÇ).

Measurements carried out on valves with pins of copper-covered wire indicated that Rp was stilllarger than 10 000 nat a frequency of 3X lOB eis, '

k'

0"JS6S!1

D'

Fig. 5. Substitution diagram for the admittance between theconnection terminals of the two control grids.

-Applieatlon of the push-pul~ amplifier valve forhigh-frequency amplification

, In the applica tion of the push-pull valve forthe amplification of high-frequency signals, various, cases can he distinguishedvsuch as' thc amplificationof 'a carrier wave 'with a very broad modulation "band (tclevision) and tlie amplification of a carrierwave modulated with, sound. Although the newpush-pull valve also offers certain advantagesfor the first mentioned application, we shall confine, ourselves to the' second, which is at the momentthe most important in the wave region below 2 in.

Infig: 6 a diagram is given showing the principle~f the amplifier. The input signal is induced in thecoil Sp of a tuned oscillation circuit, whose extrem-

rties are connected to the grid connection pinsD' and Dil of the push-pull valve 7). The condenserCl we shall for the time being leave out of consider-ation, as well as several other components whosefunctions will be discussed later.

Fig. 6. Connections of a high-frequency amplifier with push-.pull valve. The input voltage is applied to the two controlgrids in opposite phase; the output voltage is taken off fromthe two anodes in opposite phase. The self-inductions L serveto diminish the input damping, while by a suitable choice of the 'condenser Cl and Ca the amplification Eu/Ei ca~ he affected.Sp self-induction, with C, C tuned to the frequency of the A.C.voltage to be amplified; R" cathode resistance, bridged by ablock condenser Cb in order to obtain a negative control gridvoltage; R grid leakage resistance; R; supply resistances forthe scr~en grids; Rn supply resistances for the anodes.

, The D.e. voltages for the screen grids and anodesare 'taken from a positive voltage source via theresistances Rs and Ra, respectively. The anode A.C.passes, via the anode terminals A' and A", to thetuned output circuit giving the output voltage Euon the latter.If S is the slope of the valve, the following form~a

is found for, the amplification of the stage:

(4),

where Zu is the impedance between the anodes. Thevalue of Zu depends upo~ the output impedance ofthe push-pull amplifier valve, upon the impedanceof the oscillation circuit: Sp C C and 'on the inputimpedance of the nex~ stage in th~ circuit. Hereagain,we must make a distinction hetween cascadeamplification and superheterodyne amplification.. If in the, case of cascade amplification we goas far as frequencies higher than 3 . 108 els. the

7) The input circuit SpCC may he considered as a symbolfor some kind of aerial circuit, 'for instance for a dipolecombined with a transmission line' of suitable length.In any case it is advisable to employ a connection which issymmetrical with respect to earth. By this means it ispossible with the help of a dipole to obtain an input voltagetwice ,as great as in the ordinary input connection which isearthed at one .end, and where often therefore only halfof a dipole can be employed.

JUNE 1940 PUSH-PULL AMPLIFIER FOR DECIMETRE WAYES 179

imput impedance of the following stage becomesso low, even with a valve of the type EFF 50, thatit is a serious obstacle to obtaining satisfactory am-plification. It is found, however, that this dampingcan be eliminated to a large extent by including aself-induction in the screen connection. The A.C.voltage of the screen grid thereby generated pro-.duces, via the capacity between screen grid andcontrol grid, a current in the control grid circuitwhich is opposite in phase to the control grid volt-age, so that the real part of the input admittanceis reduced 8).When the second stage is a mixing stage, care

must also be taken that its input admittance is suf-ficiently hIgh. Two diodes in push-pull connection

Fig. 7. Installat:ion for the measurement of input and outputadmittance as well as amplification of push-pull valves atfrequencies up to 6· 108 cis (50 cm). The meters at: thetop indicate the control grid voltages and the anode and screengrid currents. Underneath are two micro-ammeters, belongingto the diode voltmeters for the input A.C. voltages of the twocontrol grids with respect to the cathode; below these, twosimilar micro-ammeters for the output end. 1 oscillator, 2tunable input circuit consisting of two Lech er conductorswith earthed covering; 3 valve to be measured; 4- tunableoutput circuit, 5 wave meter.

Fig. 8. Simple arrangement for the measurement of the ampli-fication of push-pull valves in a circuit according to fig. 6at a frequency of 3 . 108 cis. 1 diodes for the measurementof the input A.C. voltage. 2 diodes for the measurement of theoutput A.C, voltage. The other symbols in the figure correspondto those of fig. 6.

may for instance be used as mixing ·stage, or oneEFF 50 in a connection which will shortly be dis-cussed. In both cases at a frequency of 3 X 108 els.a value of Zu of about 3 000 Q can be obtained.Since at this frequency the slope corresponds ap-proximately to its static value of 11 mA/V, the am-plification v = 1/2 X 11 X 10-3 X 3000 = 16.5.

In testing this result experimentally a compli-cation occurred which we shall discuss in detail,because it involves a point which mayalso be im-portant in the practical application of the valve.If we consider fig. 6 we see that the input voltageas well as the output voltage is measured at theextremities of a coil. In the above calculation wehave assumed that these voltages correspond tothe voltages between the control grids gl" s," andbetween the anodes a'a", respectively. For fre-quencies like those in question this is, however, nottrue, since the connecting wires of the coil to thegrids (or anodes) possess considerable self-induction.This has already been pointed out in the discussionof the influence which the resistance of the gridconnection pins has on the input damping, and wemay use the substitute circuit, fig. 5, there derivedin order to calculate the relation between the voltageEi across the coil and the true input voltage Ei'between the control grids. If the series resistance ïs

8) In principle the application of this method is not limited.to the push-pull amplifier valve, hut offers a possibilityof eliminating the input damping in the ease of any ampli.fier valve. In practice, however, this method bas the ob-jection for most valves that the self-induction needed fora satisfactory lowering .of the input damping is 20 to 30times as great as with the push-pull valve EFF 50. Self-inductions of this order of magnitude, at the frequencies ofdecimetre waves, exhibit resonance phenomena which giverise to undesired effects.

180 PIfILIPS TECHNICAL REVIEW Vol. 5, No. 6

Fig. 9. Complete receiver for signals with a frequency of about 3 . 108 cl«. 1 dipoleaerial; 2 box containing the circuit of fig. 6; 3 mixing stage; 4 intermediate and low-Ire-quency part. To give some idea of the dimensions a slide rule about 25 cm long may beseen to the right.

lS hereby neglected, we find for sufficiently largevalues of Cl:

E-t _ 2 L C--l-w s kg

Ei'

and an analogous relation holds for the output end.It may be seen that the voltage at the extremitiesof the coil is in general smaller than the voltagebetween the electrodes.

As to the measurement of the amplification, thecondensers Cl and Ca, were so chosen that in theinput as well in the output circuit, at the frequencyfor which the amplification was being investigated,series resonance occurred between these capacitiesand the self-inductions of the connecting wires(about 4.10-8 henries), so that the total impedanceof the connection between the coil and the elec-trodes is zero. In that case Ei and Eu correspondsto the actual input and output voltages, respec-tively, so that equation (4) can be tested. The meas-urements were carried out with the help of appara-tus especially developed for that purpose (see.figs. 7, 8 and 9 with the text under the figures),and gave the theoretically expected results at fre-quencies of 3 . 108 cis. It may be concluded from

this that, within the limits of the experimentalerror, which may be taken as about 5%, the slopeof the valve does not change up to frequencies of3 . 108 cis. As stated, at a frequency of 3 . 108 cis,an amplification by a factor 16.5 is obtained whenan output circuit w it.h an impedance of 3 000 0 isused. At a frequency of 5 . 108 cis (60 cm wavelength) the amplification might still amount to 8under certain circumstances. These figures, whichby no means represent an upper limit since imped-ances higher than 3 000 0 can also be realized,are appreciably larger than what can be attainedwith the help of the acorn pentode.

As for noise also, the push-pull amplifier valve isappreciably better than the acorn pentode. At afrequency of 3 . 108 cis the electron part of theinput admittance for the push-pull amplifier valveis Ce = 1/600 0-1, while 'the noise resistanceR = 1 200 D, both values calculated between gl'and gin (see figs. 3a and 6). The acorn pentode, onthe other hand, has at the same frequency an inputadmittance of 1/2 000 D-1, almost entirely due tothe transit times of the electrons, and a noiseresistance of 8 000 O. The product Ree, which,as already mentioned, determines the noise prop-

JUNE 1940 PUSH-PULL AMPLIFIER FOR DECIMETRE WAVES

Fig. 10. Cir~uit of a·ri:tixing~tage with the help of the push-pull amplifier valve EFF 50.The input voltage Ei is applied to the two control grids in opposite phase. The same is donevia the small capacities Cbl with the oscillator voltage generated by the triodes T' andTil. The necessary oscillator voltage is of the order of 1 V, while the necessary negativebias is 2 V higher than the oscillator voltage. Tbe intermediate-frequency'output voltageEm is taken from the two anodes in the same phasc with the help of the intermediatefrequency circuit LmCm . RI; grid leakage resistances of the triodes; Cg grid condensers.of the triodes; CoCoSpo, oscillator circuit. The othe; symbols cor~espond to those of fig. 6.

'èrties of a valve 9), therefore, in the case of theacorn pentode, in spite of its lower input admittance,is twice as unfavourable as iJ?-the case of the push-pull amplifier valve. '

For the practical application 'of the push-pull amplifiervalve it is very important to know whether or not any slightdifferences between the t\VOhalves of the valve which mayoccur during manufacture, are detrimental to the performanceof the whol~ valve .. This was investigated experimentallywith the help of the, measuring arrangement shown above.It was found that large differences jn tlie anode currents (forinstance 5 mA in one half and 15 mA in the other) cause theamplification, the input resistance and the output resistanceto change by several per cent at the most.This good performance mayalso be explained theoretically.

Since at the high frequencies of decimetre waves the capacitiveimpedances between cathode, control grid and anode are smallcompared with the internal resistance of the valve, the anodeA.C. in the two halves of the system will be exactly in oppositephase, if care' is only taken that the external impedance ofboth systems with respect to earth is the same, and that theanode A.C. voltage varies exactly in opposite phase in the twosystems. Both conditions can be satisfied by providing thatthe two halves of the coil Sp (fig. 6) are sufficiently stronglycoupled with each other.Although slight asymmetries have little' unfavourable effect

on the amplification of the valve, it remains neverthelessundesirable that one half of the valve should give much moredirect current than the other half. This can be prevented by asuitable choice of the supply resistances of each of the screengrids (see fig. 6).

D) As we' havc seen the input resistance can be made verylarge by the introduetion of a self-induction in the screengrid connection, and at first glance it might be supposed 'that a favourable influence might be exerted hereby on theratio between signal and noise. This is found, however, notto be the case. Closer investigation shows that the ratiobetween signal voltage and noise voltage does not becomebetter by the application of this device.

Application of the push-pull amplifier valve inmixing stages

In receiving' sets it is also possible to use the valveEFF 50 as a .mixing valve. Several different con-nections "have been tried our, for this purpose. Oneof these is shown_infig. 10. ~~v~ral of the desirableproperties of the valve, namély the slight noiseresistance, the steep slope and the possibility ofobtaining a high input resistance are also usefulin this application. With the scheme of fig. 10, withsuitably chosen voltages for each half of the valve,a so-called conversion slope of 2.8 mA/V can beobtained. This conversion slope indicates the ratiobetween the amplitude of the intermediate-fre-quency signal in the anode current and the ampli-tude of the high-frequency signalon the controlgrid. The amplification (intermediate-frequencyoutput voltage divided by high-frequency inputvoltage) is, in the scheme shown, equal to the prod- ,uct of the conversion slope of one half of the valveand the impedance' of the output circuit for theintermediate frequency used.If the mixing stage is connected behind a high-

frequency amplifier stage, it is desirable to makethe input resistance of the mixing stage highby the use of self-inductions in the screen gridcircuit, as was already explained.The noise resistance of the push-pull amplifier

valve when used in a mixing stage amounts toabout 5 000 Q ,between the input terminals, whi~hmay be considered unusually low. If the inputresistance, of the valve is increased, to improve ~heamplification of the first stage, the noise of coursealso becomes stronger.

f

181