AN EXPERIMENTAL. TRANSMITTER FOR ULTRA·SHORT·WAVE RADIO ... Bound... · TRANSMITTER FOR RADIO-TELEPHONY among other reasons, the influence of the self-induction of the cathode feeding

APRIL 1946 121

AN EXPERIMENTAL TRANSMITTER FOR. .ULTRA·SHORT·WAVE RADIO·TELEPHONY WITH FREQUENCY MODULATION

by A. van.WEEL. ~2I.396.5 : 621.396.615.14

,.A transmitter working with frequency modulation has been developed. for an experimentalultra-short-wave radio-telephonic' link between the Philîps factories at Eindhoven andthose at Tilburg, Holland, on a wave length of 90.5' cm for one direction.and 99 cm fortheother. Modulation with frequencies from 12 to 204 kc/sec, for 48 telephone calls at once,takes place on a carrier wave with a frequency equal to 1/9 of the desired transmitterfrequency, with a maximum frequency swing of 67 kc/sec. By frequency multiplication

. a signal is obtained with the desired transmitter' frequency and with a frequency swing', .9 times as large. This article points out the advantage of this method and describes howthe frequencies have been chosen and the transmitting stages arranged. A brief descriptionis also given of the construction of the transmitter, attention being drawn in particular to

" } ,the distribution of the circuiting ov;e:.:two separate panels and the simplification of the

- wiring, both of which have been ,made possible by the application of a new method ofcoupling between successive stages, further explained in this article.

An experimental radio-telephony link hetweenthe Philips factories in Eindhoven and those atTilburg 1) has existed for several years. This hasbeen working latelyon a wave length of 90.5 cmfor one direction and 99cm for the other. The linkis so arranged that the radio installation canfunc-·tion as an entirely automatically acting link in thetelephone network. Neither the person phoningnor the operators of the telephone exchangethroughwhich the connection passes need to' be aW;!lr~of-the fact that the calls are transmitted by wirelessinstead of by .cable.Since the installation was first set up there have

been important developments both in the field of'telephony and in short-wave transmltting tech-nique. In"telephony-there is an increasing tendencyto use a singlepair of conductors in a cable for the.transmission of a large number of calls at thesame time by means of a carrier telephone system 2).The. frequency band of about" 3000 c/sec widthnecessary for transmitting a call is modulated on aspecific"carrier", the carriers for the different calls("channels") lying 4000 c/sec apart (only one sideband is used), for instance at ... ; 32 kcjs"ec,36 kc/sec, 40 kc/sec, etc. It was' found desirableto have the radio-telephone link referred to aboveequipped for such a carrier telephone system,namely for 48 chamiels, in connection with theavailable carrier telephone apparatus. The 48 chan-nels, which cover.the frequency'region from 12 to204 ké/sec, have to be modulated as a'wholeon theradio wave, which in this case serves as a "pair

~) C. G. von Lindern and G. de Vries: An ultra shortwave telephone link between Eindhoven and Tilburg,Philips techno Rev. 2,171,1937. .

~) See for example the articles published in this' periodicalabout carrier telephony: Philips techno Rev. 4, 20, 1939;ti, 325, 1941; 7, 83, 104, 184, 1942.

of conductors" in the link. The transmitter andreceiver had therefore to be made suitable for thisvery wide range of modulation frequencies (up to200 kc/sec). . .At that time there was an important development

~ the field of ultra short waves « 10m), vizthe gradual supersedingof the oldmethod of ampli-t~de modulation by frequency modulation.rr:h~advantages of this method ledus also to rebuild'ourult~a-short-wavelink for the newmethod.All these facts led to a completereconstruction

'ofthe transmitter and receiver of this experimentalcommunication, not only as regards electricalconnections but also in the actual construction.Of the old installation only the Ya gi directional"aerilJls for transmitter and receiver1) could beretained unaltered.In this article the transmitter in its present form

will be described, while the receiver will be dealt'with in' a subsequent article.

The method of modulation

The advantages of frequency modulation overamplitude modulation are twofold:1) In a~plitude.modulation . of ultra short. waves it is almostin~vitable'that also an undesired frequency modu-lation occurs; the difficulties created by such amixed modulation are avoided by employing purefrequency modulation. 2) Frequency modulation,compared with amplitude modulation, gives anappreciable improvement in the ratio between theintensity of the signal and that of the. fluctuationnoise3).

3) Fo~ readers who wish to know more about frequency modu-lation we refer to the articles by Th. J. We y e r 8 inthe preceding numbers of this periodical: Philips technoRev. 8, 42 and 89, 1946.

122 PHILIPS TECHNICAL REVIEW VOL. 8, No. '4

In order to make full use of the second advantage,i.e. in order to reduce noise 'to the lowest 'possibleIevel., it is necessary that the maximum frequencyswing, i.e. the largest deviation occurring from the'average frequency, should he ahout ten times aslarge as the highest modulation frequency to betransmitted. In our case this was not less than 0.2megacycles/sec, . so that the ~aximum frequencyswing had to amount to 2 ~c/sec. Since for theundistorted transmission of a frequency-modulatedoscillation, a side band of at least 11/2 times the'frequency swing on b~th sides of the, carrierfrequency has tó be transmitted, the transmitter. and receiver would have to be adapted for a fre-quency band with a total width of 6 Mc/sec. Becauseof the difficulties that would have been involvedin getting such a great hand width, we confinedourselves to a frequency swing of the emitted signalequal to about three times the highest modulationfrequency, In this way one arrives at a band widthof ~ Mc/sec, which is well possible in practice, whilethe ratio hetween intensity of signal and, that ofnoise ~s still 14 db greater thari it would he withamplitude modulation ..

With frequency-módulated transmitters it iscustomary to apply the actual modulation processto oscillator connections which' oscillate ai a lowfrequency, By frequency multiplication, where theaverage oscillator frequency and the frequencyswing are ihcreas~d in the same proportions, thedesired transmitting frequency. is then ohtained.This' method, whicli we also' employed, has the

, advantage that 'it is easier to 'obtain the required,proportionality between frequency swing and inten-

, sity of the modulated signal. This can be explainedas follows.

Modulation is effected.by detuning the oscillatorcircuit with the help' of a so-called reactancevalv.e 3). This is a normal multigrid valve, between.the cathode and anode of which the circuit to bedetuned is connected, while part of the circuitvoltage is applied to one of the grids with 90° phase.displacement. The anode AC is then' shifted 900

in. phase with respect to the anode AC voltage, inother words the valve acts as a reactance (capacita-tive or inductive, according as the grid AC voltageis shifted +900 or :"":"'900in phase). To take a specificcase, let us assume that the tube acts as a capacity,Cl' The reactance I/mCl is equal to the ratio of theamplitudes of anode AC voltage and anode AC..while the frequency W at which the circuit connectedoscillates is 'determined by Cl together with theremaining circuit capacity Co' Cl is now varied hyapplying the modulating signal voltage to a second

, .

.grid of the valve, whereby' the slope,· and withit the amplitude of the anode AC, is-changed in the'rhythm. of that voltage. The' detuning, .dw, of thecircuit obtained with respect to' the frequency Wo'with Cl = 0 is given by

.dw = WOCI•. 2Co

We will now consider the influence of the oscil-lator frequency to he chosen, Wo' In connection with'the linearity of the modulation, only a certainchange in slope of the reactance valve is permissible,namely such that there is no deviation frOJII.thatsection of the valve characteristic where the slopeis by sufficient approximation proportional to thegrid voltage. Corresponding to the permissiblemaximum slope, which depends only on the pro-perties of the valve' and n 0 t on the oscillatorfrequency Wo' is a certain maximum anode AC, thus- because the 'anode AC voltage' is fixed - a certainmaximum value of the reactance I/woel, Iikewiseindependent of Wo' The maxhnum detuning to beobtained .dm, which according to the formula isproportional to the greatest value of Wo Cl' is there-fore independent of the frequency Wo' be-cause Comayalso he considered as a constant. Inorder to make the detuning by the reactance valveas great as possible, Cowill be kept as small as pos-sible, and thus will he limited tö the unavoidablecapacities of valve 'and wiring.' The choice of Wois then realized with the self-induction of the circuit.

Since, therefore, the same absolute frequencysweep can be obtained 'with a low oscillator fre-quency as with a high one, if in thè manner describedo~e begins with a low oscillator frequency and acorrespondingly small frequency sweep, it. willactually be easier to ensure the necessary linearityof the modulation. We shall revert later to the exactchoice of oscillator frequency:

The connections

The connections are in push-pull arrangementfor various reasons: variationa in the feedingvoltage are much less manifest as undesired modu-lation, since they act on. the two halves of themodulator in the same' phase instead of in counter-phase; furthermore one avoids .the strong high-frequency currents. which otlierwise flow in theearth connectionsand therefore (since all the pointsto he earthed cannot be.connected to the same pointof the chassis) also through the chassis, causing allkinds of undesired couplings, etc; finally, by usingpush-pull amplifier valves several diffieulties occur- 'ring at very high voltages are diminished, becaûse,

TRANSMITTER FOR RADIO-TELEPHONY

among other reasons, the influence of the self-induction of the cathode feeding connections ismuch smallers], Infig. 1 a hlock diagram is givenofthe transmitter connections. We shall first consideronly that part drawn with heavy lines. The low-'frequency signal from the telephone exchange (the 'term "low-frequency" is here used in a relativesense, the frequencies of this signal heing as highas ahout 200 kc/sec.) is led to two reactance valveswhich influence the. frequency of the push-pulloscillator. The signal ohtained from the oscillator,ite. the frequency-modulated oscillator voltage, is

explained above that the oscillator frequency musthe chosen' as low as possible. From the. frequencyvalues indicated in fig, 1 in the different stages itmay he seen that the choice fell upon 36.9 Mc/sec.This still seems'relatively high, but it must he takeninto account that the receiver for the same, con-nection is situated close to the transmitter. Thereceiver works on the superheterodyne principlewith an intermediate frequency of IS Mc/sec; Inorder to prevent interferences in the reception it isnecessary that neither the oscillator frequency ofthe transmitter nor any harmonics of it shall fall

I '

Fig. 1.Block diagram of the transmitter. The modulating low-frequency voltage is appliedat M, R reactance valves, 0 oscillator, Tl> T2 frequency-triplicator stages, Vi- Va amplifierstages. At A the frequency-modulated output voltage is fed to 'the aerial. The numbers.in the blocks indicate the frequencies in Mc/sec.at which the stages work; above each blockthe type of valve used for that stage is given. The part of the diagram drawn with thinlines serves to keep the average oscillator frequency constant. This is 'explained later (secfig. 3). '

led to a push-pull valve, connected as a frequencytriplicator. Then the signal is amplified and co'n-ducted to anothe~ frequency triplicator. The signalhere reaches its final frequency, after which it isfurther amplified in two end stages and sent tothe aerial.

Tm: choice of oscillator frequency

The reason for the multiplication of the frequencyin, stages hy a factor of three lies in the nature ofthe connections. In push-pull connections the out-put voltage can contain, in principle, only the oddharmonics, of the frequency of the grid' AC voltageapplied, so that hy :filtering out the harmonics inquestion a frequency multiplication hy a factor3, 5 or 7, etc. can be ohtained. For the sake of effi-cient output the lowest factor has heen chosen.Since the transmitting frequency, if we confineourselves for the moment to one single call direction,was fixed at 3.32.1 Mc/sec (90.5 cm wave length)the 'possible choices of oscillator frequency were332.1~ 110.7 - 36.9 - 12.3 - 4.1 Mc/sec, etc. We have

') M. J. O. Strutt and A. van der Ziel: A new push-pull amplifier valve for decimetre waves, Philips technoRev. 5. 172, 1940. ., .

in 'the intermediate-frequency hand of the receiver,which hand, according to the ahove figures, must'extend from 17 to 19 Mc/sec.Àn oscillator frequencyof 12.3 or 4.1 Mc/sec. would, it is true, also answer

. this condition, hut the transmitting apparatus in, question had to he so designed-rthat later on,"ifnecessary. the connections huilt up with the samestages could a.lso he. used, for telephone links onother wavelengths hetween about 1.50 and 0.90 m,simply hy changing slightly the oscillator frequency,In ord~r to avoid once for all the danger of inter-ferences in the receiver it was therefore decidedto place the oscillator frequency ahove the inter-mediate-frequency 'hand' of the receiver. At thesame time of course the possibility had to he con-sidered of choosing that intermediate frequencyitself lower; in the discuesionof the receiver it willhe shown, why this was not done.

It is perhaps advisable to stress the point that the require-ment mentioned above, that for a favourable ratio betweensignal and noise the frequency sweep must be several times as'large as the liighest modulation frequency occurring, is onlyapplicable for the signal emitted by the aerial. In our case,as a consequence of the frequency multiplication, the maximumfrequency swing' in the oscillator stage amounts to only 1/9

123

124 PHILIPS TECHNICAL REVIEW VOL. 8, No. 4

of the final value, i.e. 1/9 X 3 X 200= 67 kc/sec. If a modu-lating signal with the frequency oi 200 kc/sec and with thelargest permissible amplitude is applied to the oscillator, theoscillator delivers an oscillation with a frequency fluctuating200000 times per second between 36.833 and 36.967 Mc/sec.In the case of the frequency multiplication here employedthe rhythm of the fluctuation, i.e. the modulation frequency,naturally remains unaltered: the following stage delivers anoscilla:tion with a frequency fluctuating 200000 times persecond between 110.5 and 110.9 Mc/sec, etc.

The valves used

The valves used are indicated over the blocks inthe diagram of fig. 1. Except for those in the threelast stages they are all normal receiving valveswith only a low energy dissipation so that a compactassembly is possible. The last stages must of coursehave transmitting valves in order to producethe power required for transffiission.

Nevertheless, the energy amplification in thelast two stages is only slight, as may be seen fromthe wattage figures given in fig. 1.This is due to thehigh frequency at which these amplifier stages haveto function. It would seem obvious to ask why theonce amplified signalof 110.7 Mc/sec is not firstamplified further to the desired finallevel and thengiven the necessary frequency amplification in thelast stage. This, however, is impossible, because ifthat were done very high AC voltages (of the orderof 500 Volts) would have to be applied to the gridof the last valve; in order to function as frequencymultiplier the valve must work in class C, thus withvery high negative grid bias. I~ has been found inpractice that with the very small distance betweengriil and cathode in these short-wave tubes thehigh voltages mentioned lead to disturbances:breakdown may occur or the insulation between gridand cathode may be damaged (especially at veryhigh frequencies).

Fig. 2 is a reproduction of two X-ray photographsof the type of short-wave transmitting valve used,the QQE 06/40. It is a double tetrode in which thetwo balanced systems have a common indirectlyheated cathode and a common screen grid. Thepower on a wave length of 3 m is 40 W; at 1 mabout 30 W.

Maintaining the constancy of the oscillator frequency

The part of the diagram in fig. 1which is drawnwith thin lines serves to keep the average oscillatorfrequency constant; it is given again separatelyin fig 3.

The usual method of synchronizing the oscillatorvibration directly with the characteristic oscillationof a quartz crystal or a harmonic of the same

could not be employed in our case since the oscil-lator vibration is already frequency-modulated uponits formation, and consequently when keeping the

Fig. 2. X-ray photographs taken in twomutually perpendiculardirections of the ultra-short-wave push-pull transmitter valve,type QQE 06/40, used in the three last stages of the trans-mitter. The cathode connection is made very short thanks tothe two electrode systems having a common indirectly-heatedcathode.

oscillator frequency constant a certain margin hasto be left for the frequency swing. For that reasonthe following method was chosen.

A small part of the output voltage of the oscillatorstage, of which the average frequency fo mustamount nominally to 36.9 Mc/sec, is tapped off and

Fig. 3. Block diagram of the connections for keeping theaverage oscillator frequency constant. K quartz crystal whichkeeps the frequency 8.725 Mc/sec of the crystal oscillator KOconstant, M mixing stage in which the fourth harmonic ofthis frequency (34.9 Mc/sec) is mixed with the frequencyfo R:j 36.9 Mc/sec of the transmitter oscillator 0, V amplifier,D discriminator, P pentode, F coil with ferromagnetic corecoupled magnetically with the circuit self-induction of theoscillator O.The numbers in the blocks indicate thefrequenciesin Me/sec.

•TRANSMITTER FOR :t;tADIO-TELEPHONY 125

iD. a mixing valve mixed with an AC voltage of the.very constant frequency of 34.9 •Mc/sec from avibrating crystal. The output voltage of this mixingstage has an average frequency f1 = fo - 34.9 Mc/sec.If fo is exactly'. 36.9 Mc/sec f1 = 2 Mc/sec; if fodiffers slightly from 36.9 Mc/sec f1 exhibits the sameabsol~te difference from 2 .Mc/sec. Furthermore, the"intermediate frequency" f1 is of course frequency-modulated with the telephone frequencies. in thesame way .as .the high frequency fo applied. Thisoutput voltage of the mixing stage is now amplifiedto a discriminator connection. This produces a DCvoltage' proportional to the' difference between theav~rage frequency f1 of the applied voltage and thefixed frequency fd = 2 Mc/sec at .which .the dis-criminator 'il;iset. An AC voltage is superposed' on

,l ".

temperature, etc. The fact that this forms no ob-stacle to the regulating 'action. of the whole -is dueto . the discriminator reacting to the absolutechanges of the oscillator frequency. Even if thediscriminator frequency fd should drift. propor-tionally just as much as the oscillator frequency fodoes with no regulation, the absolute variationsof t~e latter are still 36.9/2 R:j 18 times as large asthe drifts of fd.

The coupling between the successive valves

For the coupling together of the various tripli-:cator and amplifier stages a new method has beenemployed . which': offers important advantages:In order to explain the particulars ofthis coupling,let us consider figs. 4a :and b in which a coupling

Fig. 4. a) Diagram of the usual method of coupling two amplifier valves BI" and B2; L-Ctuned parallel circuit. b) New method of coupling. Ca. Cg. and La. Lg are the .internalcapacities and self-inductions of the valves involved in the coupling. Tlie elements neces-sary for determiping the DC voltage situation are not drawn. -. . .!

the DC voltage,' namely. the 0 iginal modulation,.voltage, which, however, is suppressed by a low-pass filter. The DC voltage is now applied to the'grid of a pentode. The anode current of this valve,which is thus proportional to the absolute deviationof the average oscillator frequency from the pres-cribed value of 36.9 Mc/sec, flows through a coilwound around a core of ferrornagnetic material 5)magnetically coupled with the selfinduction coilof the oscillator circuit. Owing to the fact that theanode current changes "the premagnetization andthus the permeability of the coil core, it affects theself-induction 'of the oscillator coil in the sense. that'the change' in the average oscillator frequency,to which the anode curren:t is proportional, isopposed. In this way it proved possible to reduce

. the drift 'of this frequency by a. factor 30.The "fixed" frequency fd at which the discriminator

is set is, of 'course, nót actually fixed either, butsubject to some variation due to the drift of the'elements of· the discriminator connections with

6) .·It is necess~ry t~ u~e afer~omagnetic materi~l the.losses ~fwhich aresufficiently small, even at high frequencies.

according to the usual method and, ~ne accorwPgto the new methoq are .shown side by side, For thesake, of simplicity ordinary (not push-pull] stages are .assumed and all elements only: of Importance fOFthe DC voltage situation are omitted, In theordi-nary coupling the anode. of the' first and' the gridof the second valves ar~ connected directly witheach other, wlÛle betweeit this .point and earth atuned parallel circuit is connected (the internalself-inductioria La and 'Lg of the valves, which areindicated wi!;]i·dotted lines in figs. 4a. and- b; mayfor the present be disregarded). In the new. coupling;on the other hand, the tuned circuit is formed bythe self-induction Land the connection in seriesof the internal valve capaeities Ca and Cg. Infigs. Sa ana 'b the situation fn the two cases is shownstill more concisely. Let us first consider fig. 5b •The high-frequency voltage between -the points'A and B is mainly determined by the oscillationof fhe circuit, hut the potentials of these two pointsfluctuate in' opposite phase With respect to earthdue to the o})trusion of earth .potential. at poin): 0.in the middle of the circuit capacity. From this it-

_-------r-----------. ~ - ------- ---~- -----------,--------

126 •PHILIPS TECHNICAL REVIEW VOL. 8, No. 4

follows that also in the middle of the circuit self-înduction there must bè-a point (P) which remainsat earth potentlal during oscillation' and thus has no

'--- " ---~-;;i~,~..L· ~ -- \.' -'.,,_ \.-'- . _-- --....._ . ---"'!f"

Fig. -5 -. a) More detailed diagram of the usual coupling circuit.The anode current ia of the first; valve and the grid voltage VIIfor the second valve are respectively 'applied to, and takenfrom the same pair of terminals. .

b). The same for the .new method of .coupling. Here ia isapplied to-the "terminals" of ·Cà.while v8 i,Staken from theterminals of Cs' Since the point 0 lying in the middle of thecircuit capacity remains at earth pótential, there is also apoint P on the self-induction with the same property.

high-frequency voltage 6). The position of P on theself-induction L is determined by the ratio of the

_capacities Ca and Cg. On the other hand, turning. to fig. Sa, it is evident that in the old coupling'connections no such non-voltage point can be found.Even if the self-inductions La and Lg are taken intoaccount this fact is not altered.. The occurrence of a non-voltage point in the new

method of coupling offers the possibility of dividingthe transmitter constructionally into two' parts at anon-voltage point between two valves and assem-bling the two parts, for example, in separate panels -.The connection between the 'two panels then carries .no high-frequency voltage (high-frequency currentdoes, however, flow through it) and therefore noundêsired couplings or radiation can occur, Ifsuch a division were made at a point which was notvoltage-free, the connection would have to be shiel-,ded against the effects mentioned, and a large extracircuit capacity would thereby be introduced, ingeneral resulting in a, loss of. amplification. Forthis reason i~ is very difficult to divide the connec-·tions when the old method of coupling is used. ,

Another advantage of the new method of couplingis of particular importance at very high frequencies ..Since in this case the capacity and/or self-induction .of the coupling .eircuit m~st become very small, thecontributing, unavoidable capaeities and self-inductions of the valves begin to play an importantpart. Fig. 4a cannot then be reduced to' the simplesituation of fig. Sa, because of the presence ofLa and Lg, which can no Ionger be ignored. The ordi-

G) Strictly speaking this is not true. If the losses of the cir-cuit elements are taken into account it is found that thereis a certain residual voltage. This is very small, however,compared with the circuit voltage.

nary connections are now in principle more compli-cated than a simple L-C circuit, .and in practiceit proves difficult to obtain sufficient amplificatiorrwith them. This is understandable' when it is bornein mind that at very high frequencies Lg mayalready be approximately in resonance with Cg.The tuned' parallel circuit L-C, whose task it is tofurnish a high impedance for the circuit frequency, isthen as it were short-circuited by the low impedanceof the series circuit Lg-Cg. Measures for over-coming these difficulties are known but they onlyresult in making the circuit still more complicated,more extensive and more difficult to manipulate.W~~n we' compare this method with the new coup-ling according to fig. 4b, we see that the presenceof La and Lg does' not alter anything in the principleof the connections and only makes. it necessaryto choose the external self-induction L somewhatsm~Uei- than thë total circuit self-induction desired.Even if Lg or perhaps only a part of it is of itselftuned to Cg, in the main everything remainsunchanged, It means really nothing else than thatthe above-mentioned .non-voltage point then liesexactly at the valve terminal or somewhere in theinternal grid feed, connection, as the case may he.It is then' no longer possible to divide the connee-tionà at that point, but the action of the .amplifierstage is pot at' all affected. It would only becometroublesome if a negative value should- be requiredfor the external self-induction' L. But since Ca is-usually much smaller than Cg, the self-inductionnecessary to tune Ca, i.e. the self-induction be-tween the anode and the non-voltage point, is gen-:erally large enough to cover, in addition to theself-induction La of the internal anode feed con-nection, a positive external self-induction as well.

Fig. 6. Example of a push-pull amplifier stage according tothe new method of coupling. Between the anodes of the push-pull valve Bl and the control grids of the valve B2 two self-Inductions L are connected. The capacities Co only have thefunction of separating the DC voltage positions of anodes andgrids. The circuit -is tuned with the variable capacity Cl'The anode DC is supplied through the resistances Ra;Ra may not therefore be chosen very large; in spite of this inorder not to obtain any undesired damping of the circuitthese resistances are connected at the non-voltage points..There. is no objection, to choosing large grid leakagerestst~nces RB' .

TRANSMITTER FOR RADIO-TELEPHONY 127

Fig.6 shows the complete connections of apush-pull amplifier stage with the hew couplingmethod. Several details are explained in the textof the figure.

Construction

If a transmitter is to serve as part of a earnertelephone system it is desirable that its external. form should be adapted to that of the carrier tele-phone apparatus. In the case in question the trans-mitter had to be housed in a rack of certain dimen-sions with the components mounted on panels of acertain length and width which could be slidinto the rack from front and rear. Similarity inexternal appearance, however, can never be com-plete. In the case of the carrier telephone apparatus

- Fig. 7. The transmitter and receiver assembled in a rack. Fromtop to bottom: a panel (of double depth) containing the high

. tension supply unit, two panels with the transmitter, twopanels with the receiver. All the panels are constructed assliding drawers. On the front plate of the upper transmitterpanel may be seen the valves of the three last stages, within,between the coupling connections, shielded by caps. On eachpanel is a meter with which the cathode currents of all thevalves can be checked by means of a switch. In the open spacesat the back of the rack identical drawers with a transmitterand receiver can be inserted to serve as reserves.

Fig. 8. Part of the two transmitter panels seen from the rear.The two small coils in each panel on the extreme right (indi-cated by arrows) are the parts of the two coupling coilsdividedinto two. In the lowest drawer, pulled halfway out. may beseen the plug pins and sockets.

the available space in every panel is almost com-pletely filled with the parts of the different filters,modulators, repeaters, etc. Because of the muchhigher frequencies, the components of the trans-mitter, however, must in general be spread out overa certain area in order to provide sufficient mutualintervals to prevent undesired couplings. One mayspeak here of surface assembly in contrast tothe volume assembly in the apparatus forcarrier telephony. In this way one arrives at theconstruction shown in jig. 7. The components aremounted side by side in each panel in a singleplane, a "front plate". This lies fairly deep in therack, so that the different valves, coil cans,Lecher systems, etc. situated on the front of theplate are well protected against undesired contactor shocks without it being necessary to place acover plate in front of the rack, which would hinderthe dissipation of the heat given off by the valvesto the air.The-uppermost panel ofthe installation is actually

double, occupying the entire depth of the rack inorde; to offer space for the high tension supplyunit mountedtherein. This is placed at the top inorder that the rather large amount of heat developedin it shall not cause any difficulties in the trans-mitter part proper. The transmitter also containstoo many components to be housed on one normalpanel. As a consequence of the above-mentionedmethod of coupling between successive stages thetransmitter could without difficulty be dividedand housed in two separate panels one above the

128 PHILlPS TECHNICAL REVIEW VOL. 8, No. 4

other. This can he seen in fig. 7 underneath the hightension supply unit. Fig. 8 shows the rear of thetransmitter panels with the parts of the two coup-ling .coils (push-pull connections, see fig. 6). ineach panel. Below the transmitter panels is the

\corresponding receiver, also divided between twopanels. On the other side of the rack two exactly'identical panels with a complete trans:q1itter andreceiver 'can he inserted to serve as' reserves,and in case of a breakdown these' can beswitched on by a single touch of the hand orentirely automatically.

Upon inserting a panel all connections of thevarious signal and feeding voltages are automati-

cally made by means of plug pins in the rack' andsockets in thè panels. The lead wires to the. plug'pins are led out at the side, where the wiring harnessfor the mutual connection of the panels is mounted.The details can be seen in figs. 7 and 8. On eachpanel a measuring instrument is mounted on whichthe cathode current of each valve can be checkedby means of a switch. By this' means it is possibleto localize the cause of any interruptions quickly,while the ageing of the valves can also easily beascertained in good time. This is necessary becausethe same requirements as to reliability are made ofth~ transmitting apparatus as are made of thetelephoiie apparatus itself.