JULY 1937 ·209

THE USE OF AMPLIFIERS (REPEATERS) IN TELEPHONY. . . .' ,

by W. SIX and H. MpLDERS.

Summary. Following a discussion of ,the principle of two-wire and four-wire repeaters foramplifying speech currents in telephony, three different types of repeater are described,viz., a four-wire repeater equipped. With triodes, a four-wire repeater equipped withpentodes and inverse feed back, and a two-wire repeater also with pentodes and inversefeed back. . ,;

In the two-wire circuit, two repeaters ll:re con-nected to the line at the same time, the input of onerepeater being in parallel with the output of thesecond repeater. The repeaters in .the. variousrepeater stations are thus linked through a twin-coreconductor.In the four-wire circuit, telëphone traffic in the

one direction is entirely distinct from that in theopposite direction. Hence two twin-conductors arerequired and at the terminals the input of onerepeater is connected in parallel to the output of the'othe~.. Paralleling the. input and output, particularly

.with two-wire circuits, is not possible without'further precautions. For the amplified signals wouldpass from one repeater to the input .of the otherrepeater,. where they could again be amplified, sothat in fact the repeaters would commence tooscillate. This !s prevented by means of so-calledbalancing network shown in fig. 2, which fundament-ally is a bridge circuit. The four arms of the bridgeare formed by the line impedancê 1, the artificialline N2 and the two impedances bet~een the points

1) W. Six: The use of loading coils in telephony, Philips .T hu R 1 353 1936 a - band b - c " these impedances aré determinedec .' ev., , .

In a previous article 1) published in this Reviewit has already been pointed out that before the in-troducti~n of loading coils the range 'of trarisrpi~si.on

. '. !

in telephony was fairly small. While the adoptionof coil-loaded cables increased the' range over whichcalls ~ould be made, it did not entirely remove all'restrictions to telephonic intercommunication. Anoutstanding advance in developing an internationaltelephone service was in fact not mad~ until 1920when the general principles of amplification teelmol-ogy were worked out, -. . '!

By inserting loading coils in telephone cables the, attem{àtion can, certainly \ be considerablyreduced, but ovet very great distances the weak-ening of speech currents is still excessive. Fortransmission over long distances, it is imperativetherefore to supply additional energy to the trans-mission line by introducing a type of relay action,In the first application of this principle thellne wasinterrupted at a partienlar point, the 'incomingline connected to a telephone receiver and the out-going line to a microphone, the diaphragms of the .receiver and the microphone being linked mechan-ically (B row n relay). The signals reaching thereceiver then cause the microphone diaphragmto vibrate, so that an amplified signal is transmittcdthrough the outgoing line. The energy' transmittedthrough the first part of the line is therefore usedto. operate the relay, while the energy passed through

. - I .the second part is furnished by the' microphone

_'battery.Nevertheless the general application of amplifiers

of repeaters to telephone technique only becamepossible when the production of the triode ampli-fying valve provided an inertia-free and efficientrelay, which also permitted a high amplificationratio to be obtained.

Two-Wire and .Four- Wire Circuits

As stated above the incoming signal is utilisedto operate the relay, which in the case of the triode(three-electrode valve) means that the incomingline is connected to the grid of the valve through

a transformer. 'I'he outg.oing fue is connected tothe anode circuit through another transformer. Thiscircuit will naturally operatei:n: one direction only;to make two-way traffic possible, two methods canbe employed, either a' two-wire or a four-wirecircuit may be adopted (seefig. 1).

- -- - a

,

::,

:- ,-- - b

.22442or, ..

Fig. 1. a) General scheme of a, two-wire circuit.b) General scheme of a four-wire circuit.

210 . PHILlPS TECHNICAL REVIEW .vol. 2, No. 7

by the input impedance of the repeater. The twotransformer windings a - band b - c are similar,so that the impedances between a - band b - c also

- .22441

Fig. 2. Balancing network: The impedances of the artificiallines NI and N2 are equal to the impedances of the lines 1 and 2.In this way the output signalof V. produces no voltage atthe input of VI and vice versa. •

are equal. Furthermore, the impedance of the ar-tifièial line NI is made equal as far as possible tothe line impedance for all frequencies. If a potentialdifference obtains between the points band d(output voltage of V2), the potential differencebetween a and c must be zerp, since the ,bridgeis balanced. No voltage is therefore applied to theinput of VI' On the other hand, a signal incomingover the line 1 must be passed to the input of VIand after amplification transmitted to the line 2.But ' it is difficult to make the impedance of theartificial line exactly equal to the line impedanceat all frequencies, since particularly with coil-lo'aded cables this impedance is non-uniform ina certain frequency range (see jig. 4), owing to theirregular intervals between the loading coils and thedifference in their self-inductance values.

The propagation of electric waves through cables hasalready been' discussed theoretically in this Review 2)., From equation (2) of the first article, we can readily deduce

that for a wave passing through a cable the ratio V: I, whichis termed the impedance, is represented by:

Z = V' 1 = "IR + j w L .... ,. (1). , G+jwC'

where R is the resistance, L the inductance, G the dielectricconductivity (insulation loss) and C the capacity per unitlength of the cable. This expression also applies to coil-loadedcables provided that the frequency is low, i.e, the wavelengthis long, compared to the intervals between the loading coils.

But if this is not the case, the fact that the self-inductancesare not uniformly distributed over the cable must be taken intoconsideratiorr, The cable is then more satisfactorily representedby the equivalent circuit' shown in jig. 3, which is identical to

2) W. Six: The use of loading coils in telephony, Philipstechno Rev. 1, 353, 1936, which is the first article; andBal t h. van der Pol and Th. J. W eye r s, ElectricalFilters V, Philips techno Rev. 1, 367, 1936, which is thesecond article.

the low-pass filter discussed in the seco'Îld article. By analogyto the discussion in the seriesof articles .dealingwith electricalfilters, it may be expected that the attenuation in loaded

I T I2

.• 22483

Fig. 3. Equivalent circuit' for a coil-loaded cable. If theuniformly-distributed capacity is replaced by' condensers, asystem equivalent to a low-pass filter is obtained.

cables will remain small up to nearly a specific cut-off fre-quencyof:

1)1---'-1 - :n;11LC

and will then increase rapidly. \Vhere no losses occur, theimpedance, which in the notation used in the second articlequoted must be represented by Z:r', should also becomeinfinite at the cut-off frequency and be expressed as a functionof the frequency as follows:

Z == Z':r

These equations for the low-pass filter cannot be applied.to cables without some modification, since adequate con-sideration is not given to the fact that the capacity is dis-tributed uniformly along the line. 'I'he.qualitative deductionsmade therefrom are, however, correct, as may be seen from,jig. 4, where thc attenuation and the, ilTlpeda~ceare plotted

db0,36 .2.2436

2800

._L

f ~I-"'"

1/ _l1

11 •

1Le: ,..,.,....- ,

0 1000 2000 3IXXJ 4000 sao 6000

II/_

L_LJ_

2A1If v

Ir'o_/_

1 ...- ./V

<, I--

'I

0.32

0,28

0,24

0.20

0,16

0,12

JL3200

Hz

2400

2000

1600

1200

800o ox 2000 3000 4(X)() 5000 . 6(}{X) Hz

Fig. 4. Attenuation of an impedance in (1) a lightly-loadedcable, and (2) a heavily-loaded cable.

1· 44 millihenries at intervals of 1830 m.2 - 177 millihenries at intervals of 1830 m.

JULY 1937 AMPLIFIERS IN TELEPHONY

In addition to the input: impedance, the im-pedance of the artificial line must also be equal to Z,as already stated above. Fig. 5 shows diagram-

as functions of thc frequency for a low-loaded cable and ahigh-loaded cable.

Close to the cut-off frequency, the impedance varies con-siderably even with very ~mall changes in the self-inducta~ceand in the intervals between the loading coils; it gives a wavycurve when plotted as a function of the frequency if the coilsare not equally spaced, and for this reason it is difficult 'tokeep the balancing I?-etworkin balance in this frequency range.

If the bridge circuit is not fully balanced at aspecific frequency, then although this circuit will.still produce a certain attenuation, part of the signaloriginating from 'one repeater will be .passed backto the input of the other repeater. As a result thegain is limited, for in the complete ciréuit which ismade up of the two repeaters and the two balancingnetworks a certain attenuation must be presentand must be greater than the total gain in thecircuit; if this were not the case the amplifiers wouldstart .oscillating.' .If the amplification ratio of Vl and. .V2 is, for matically how these two requirements can be satis-

instance, 20 decibels, then the attenuation in the fied. In this diagram Z is the impedance of the line,balancing network must he greater than. 20 decibels Zl the input impedance of the' rep,.eater1 Z2 theat all frequencies. But, as already indicated above, output impedance of the other repeater and N. this is difficult to achieve close to the cut-~ff the artificial line; el' is the voltage in the line andfrequency of the cable. A filter is, 'thereforë, con- e2' the voltage generated in the amplifying valve.nected in front of the amplifiers Vl and. T1, and If the impedance of N is equal to Z, the voltage e2serves to produce a marked attenuation of the fre- will not cause current to flow ill the branch z,quencies lying above the band required for: satis- since the bridge is balanced: If Zl' .1/2Zand Z2=factory acoustic intelligibility. The amplification 2 Z, it may be found by caléulatiori that the voltageratio at frequencies above this band will, in con- e2 produces no current at all through the branchsequence, be much reduced, so that the attenuation N"and that the terminal impedance of the line isof the balancing network can also be made much equal to Z:smalle,r. ' With four-wire circuits, somewhat different con-"Another important point in the case of two-wire ditions obtain. Naturally a certain. attenuation must

repeaters is with regard to the input jmpedance, be present also in this circuit: which' is made upi.e. the impedance between the points a and d, of repeaters, lines and balancing networks. But inwhich. must be as nearly equal to the line impedance this case, the attenuation of the .telephone linesas possible. The greater the difference between these also is included in the circuit. If the' repeater gainimpedance values, the greater will he the fraction is made so great that the voltage is slightly lowerof the incoming signal which is reflected back to at the end of the line than at the beginning, therethe preceding repeater, thus further reducing its can be no question of oscillation, even when astability. Rednetion in the stability by reflection balancing network is not used. The only disadvan-can also be described in another way, viz,., by _tage without such a network is that' the signal maysaying that the line 'impedance seen from the' return to the input of the line and thus give risepoints a - d is determined by the termination of to echo phenomena. For this reason balancingthe circuit at the preceding repeater situated at networks are used in these cases also, but thethe opposite end of the cable and that therefore artificial lines need not be given the same accuracythe balancing network will only be equivalent when in simulation as is required with two-wire circuits,the line is terminated at that point by its impedance. The amplification ratio at the i~djvidual repeater'If, therefore, the impedance of the line is Zand

3). the input impedance of the repeater is W, thecoefficient of reflection will be:

.22488

Fig. 5. With this choice of impedances in: the balancing net-work, the bridge is not only balanced but at the same time theline is terminated by its impedance.

. Z-W .F = 3).

Z+W

A cable with impedance can he substituted electricallyby a' voltage source in series with •.an impëdance Z. If àvoltage source ewith an internal impedance Z is shorted byan impedance W, the voltage applied to W will beeW/(Z+ W). If Z = W, the voltage -will be 1/2 e. Thismayalso he regarded as a reflected voltage equal to1/2 e [1- 2W/(Z + W») = 1/2 e (Z- W)/(Z + W.

211

212 PHILIPS TECHNICAL REVIEW Vol. 2, No. 7

stations is .in this case limited by cross-talk phe-nomena between the telephone channels as well as bythe interference level. In practice a gain of 12 to 15decibels can he obtained with two-wire repeatersand of 30 to 40 decibels with four-wire repeaters.

Frequency Characteristics of Repeaters

In general, the gain at all frequencies should bemade equal to the attenuation produced by' theline, i.e. within a frequency band from 300 to 2500cycles, which as a rule is adequate for telephonechannels. With coil-loaded conductors, the, attenu-ation increases considerably at higher frequenciesparticularly in the neighbourhood of the cut-offfrequency. In a heavily-loaded line, i.e. one equippedwith coils with high, self-inductance values andlocated close together, the attenuation in the speechband is on the average smaller, but on .the otherhand the cut-off freqeuncy is lower, w\ille withlightly-loaded lines the attenuation' is gr~ater butthe cut-off frequency is 'higher. '

It is, therefore, necessary in the case of heavily-loaded lines to insert a wave filter or .networkbetween the lines and the repeaters or ui the reo'peaters themselves, whose main purpose ,is tocounteract the increase in attenuation close to thecut-off frequency and at the same time provide aslight balance for the drop in the attenuation atlow' frequencies. This measure is not necessarywith lightly-loaded lfu.es.

Choice of Repeater System

There is still a wide divergence of opinion indifferent countries with regard 'to the relativesuitability of the various systems for producinga gain in transmission viz., two-wire or four-wirerepeaters or light or heavy loading. In -Hollandthe method which has been adopted generallyduring recent years consists in the use of four-wirerepeaters and light loading 5). Although' doublethe number of conductors are required, with thefour-wire system, it yet. allows a much higher gainto be obtained than the two-wire system, so thata much greater attenuation is permissible in thecables with the same number of, repeaters, andin consequence a smaller copper cross-section can

,4) For every sectionof line joining two repeaters, the averagepower at the output of one repeater is much greater thanthat at the input of the other repeater. This difference inlevel is equal to the gain of the repeater. For lines withopposing directions of speech, the level is high in one lineand low in the other line. Cross-talk will therefore be thegreater, the greater the' difference in level, i.e. the greaterthe gain.

, 5), A new telephone system in Holland, by A. H. de Voo g t,Post Office.Electr. Engin. Journ., 25, 195, 1932-33.

be used. Moreove!, as already stated above, thefour-wire repeater is much simpler to design,'since in it the balancing networks with artificiallines are required only at the terminals of theline and need not be made with the sameaccuracy as in two-wire repeaters; in addition,rejector filters to cut out high frequencies, whichare necessary with two-wire repeaters, can be dis-pensed with in the case of the four-wire type.

The method of Ioading adopted in Holland inwhich 65-millihenry coils are located at intervalsof approximately 3.68 km, has a cut-off frequencyof roughly 3400 cycles, which' is sufficiently abovethe speech frequencies, so that the attenuationin:the cable requires compensa~ion only in the rangefrom 300 to 800 cycles. The wide interval betweenthe loading coils has resulted, moreover, In amarked saving in the cost of the coils.

Inverse Feed back

Inverse feed back or negative reaction has beenemployed in amplifier technique for many years,.and consists in feeding back in phase oppositionpart of the output voltage to the' input of theamplifier. In these circuits the gain is naturallyreduced, but their advantage p.es in the fact thatat the same time the distortion is lower and the gainis made more stable, in other :wo~ds'it is made lessdependent on the characteristics of the amplifyingvalve and the feed voltages. For some years inversefeed back has been used in all amplifiers employedin carrier-wave telephony in which an extremelylow distortion is essential in view of cross-modu-'lation between the different transmission channels.'The Dutch Post Office authorities first employedthis circuit in intermediate . repeaters for low-frequency telephony, its chief advantage beingnot so much that the distortion is reduced but thatthe gain is constant. Reduced gain which is in-separable from inverse feed back is not really adrawback, since by using a valve with a higheffective amplification (pentode) roughly the samegain can be obtained as was realised in the pastWith a triode.

Description of Three Types of Repeater

Three types of repeater have been developedin the Philips Laboratory, all three being designedfor indirectly-heated amplifying valves, The use

6) C. J. van L 00 n: Improvements on radio receivers,Philips techn, Rev. 1, 264, 1936.

7) A new feed-back repeater, by G. H. Bas tand E. H.Stieltjes, Post, Office Eleetr. Engin. Journ. 28, 225,1935-36.

JULY 1937 AMPLIFIERS IN TELEPHONY 213

of these valves considerably simplifies the circuitcompared with triodes having directly-heatedfilaments as hitherto in use and in which thefilaments of four valves were always connected inseries. For these latter a separate battery for thenegative grid bias was necessary, while at the sametime complicated circuits had to be adopted todecouple the repeaters from each other so as toavoid cross-talk. The simplified circuit enables avery compact repeater unit to be obtained. Theserepeaters are all-mains fed from an alternating-current supply.I) The first repeater developed was a four-wirerepeater with triodes without feed back. Arepeater unit, containing four repeaters of thistype, is shown in jig. 6.

Fig. 6. Repeater unit with 4 four-wire repeaters equipped withtriodes.

II) The fundamental circuit of the four-wirerepeater with feed back is shown in jig. 7. Inthis repeater a pentode is used as an amplifier,and the feed back voltage is tapped from therheostat R. This voltage is determined by the

.22440

600Il j ~ 600fl

Fig. 7. Fundamental circuit of the Philips four-wire repeaterwith pentodes and inverse feed back.

resistance, so that the gain is controllable. Between300 and 4000 .cycles the characteristic of therepeater is a straight line within 2 decibels (seejig. 8).

db.2248050

fdb I

4

~

NORM

20 ..........MIN

fa

050 fOO 200 300 500 fOOO 2000 400 0 10000 Hz

Fig. 8. Gain in the four-wire repeater with inverse feed backplotted against the frequency.

To obtain satisfactory matching to the line, theinput and output transformers are terminated withresistances equal to the impedance of the telephoneline, viz, 600 ohms. Strictly speaking, the terminalresistance together with the input or outputimpedance of the amplifier in parallel to it shouldbe made equal to the impedance. The input andoutput impedances are, in this case, much greaterthan 600 ohms and hence practically without effecton the combined value. The non-linear distortionfor an output of 50 milliwatts is shown in jig .. 9

%5,----,-----,-----,-----,---- 224:1f

Fig.9. Non-linear distortion in the inverse feedback repeaterplotted against the gain at a power output of 50 milliwatts.

as a function of the amplification. Throughout thewhole range this distortion is below the limit statedin the Standards of the Comité Consultatif Inter-national des lignes téléphoniques à grande distance(e.C.I). The principal advantage offered by thisrepeater, as compared with the preceding repeaterwhich i~ equipped with triodes, is in its markedindependenee of the feed voltages, as may be seenfrom Jig. 10. These curves in fact show that thepentode amplifier with negative reaction stillsatisfies !peC.C.1. standards for lines with more than12 repeaters when the anode and filament voltagesdrop to 60 per cent, while in the case of the triodeamplifiers the voltage must not drop below 82 per

214· PHILIPS TECHNICAL REVIEW

cent if they are still to .conformwith these standards.Ill) The fundamental layout of a two-wire

40 50 . 60 70 80 100%90

aQ2

0.aaaaaodb

,I -~ -y/"" ./

1--- -:::-..f..1'--- 1---- V-- ~lcl_3

/ /,4 -- =t t- V -- (CL-=.~ I.6 J

I I,8 I I,9 I I

A

8

.22432

Fig. 10. Gain in decibels below the normal value plottedagainst the filament and anode voltages in percentages of thenormal value. T - triode amplifier without negative reaction; ,P - pentode amplifier with negative reaction. A - C.C.I. limitsfor lines containing more than 12 repeaters. B. - C.C.I. limitsfor lines with 12 repeaters or less.

1

Vol. 2, No. 7

tional to the output' voltage. The feeding backof the current raises the internal resistance of therepeater, while it is reduced 'by the voltage back-feeding 8). By. a suitable combination of thesetwo' reactions the required internal resistancecan he obtained in the repeaters and at thesamè time the requisite degree of negative reactionrealised. In this repeater the internal resistancehas been made equal to the line impedance, sothat the output transformer need not be terminatedby .a 600. ohm resistance as is necessarywith the four-wire repeater. In the latter, halfof the energy output was lost in the terminalresistance, while in the two-wire repeater the wholeof the en~rgy output is available for useful work.In, fig . .1~l the impedance (magnitude Izl and

phase angle tp) of this repeater is plotted against

+ +

Fig. lL General circuit diagram of Philips two-wire repeaters with pentodes and inversefeed back. NI and N2 artificial lines. E - Balancing networks. F - Filters.

repeater with 'inverse feed back is shown in fig. 11.In this repeater; a combined current and voltagefeed baék has been used, in other words thenegative-reaction voltage is made up of twocomponents: one component is tapped fromthe resistance R and is hence proportional to theoutput current, and a second component takenfrom the potentiometer P and is hence prop or-

8) Negative reaction or inverse feed back consists in feedingback to the input the voltage variations occurring in oneor other of the circuit components at the output side of anamplifying valve, so that these variations become reduced.If the component is so chosen that the voltage applied toit is proportional to the output voltage, the alteration inthe output voltage, caused for example by altering theexternal resistance load connected to it, will be less, i.e.the effective internal resistance is reduced. If the feed-back voltage is. made proportional to the outputcurrent, this current becomes less dependent on the valueof the resistance load, and the internal resistance is in-creased,

2

U444

the frequency. The broken line represents the im-'J/.

2500

"--e

2000 . r-, r-- -- -,1000

zlSOO

1 0200 400 6oo

~t _tvc:::.+-

I20 . IlP 300

I I

800 fOoo 1200, 14oo 1600 1800 2000Hz

I II II

-

Fig. 12. Fulllines: Impedance of the two-wire repeater (mag-nitude IZI and phase angle cp). Broken line: Impedance of acable with L3 mm cores and loaded with 177-millihenry coilsat intervals of 1830 m. , .

JULY 1937 AMPLIFIERS IN TELEPHONY 215

.2244$

0,2--- -- -- -- -- ------ -- -- C£L

\1 "',5 <,r---

0,1

0,

0,0

oo 200 400 600 800 1000 1200 1400 1600 1800 2000 Hz

Fig. 13. Coefficient of reflection of the two-wire repeaterplotted against the frequency, when using a cable for whichthis repeater has been designed.

pedanee of the cable for which the repeater wasdesigned (diameter of cores 1.3 mm; loading coilsof 177 millihenry at intervals of 1830 m).The reflection coefficient F = Z - WIZ + Wis

db25

.22435

20 ft~ VK

0 ~ P\V \

5

0200 500 1500 2000 2500 Hz1000

Fig. 14. Gain in two-wire repeater containing pentodes, plottedagainst the frequency. J - Repeater with rejector filter for highhigh frequencies. 2 - Attenuation of the associated cable. 3 -Repeater corrected by means of a balancing network.

plotted against the frequency infig.13. Throughoutthe whole frequency range this coefficient is con-siderably below the limiting value specified bythe C.C.1.

Fig. 14 gives the frequency characteristic of thisrepeater. Curve 1 is the characteristic of the repeaterwith rejector filter for cutting out high frequencies.Curve 2 is the attenuation of the cable, whilecurve 3 is the characteristic of the repeater whenusing a balancing network which compensatesfor the increase in attenuation in the neighbourhoodof the cut-off frequency of the cable.Finally, a photograph of a repeater of this type

is reproduced in fig. IS.

Fig. IS. Photograph of a Philips two-wire repeater with pen-todes and inverse feed back.