HILL: PHANTOM TELEPHONE CIRCUITS. 675

PHANTOM TELEPHONE CIRCUITS, AND COMBINED TELEGRAPH AND TELEPHONE CIRCUITS, WORKED AT AUDIO FREQUENCIES.

By J. G. HILL, Associate Member.

(Paper first received l&ih October, 1921, and in final form 28th April, 1922; read at THE INSTITUTION lGth March, 1922.)

SUMMARY. (2) The Equipotential method of providing simul- The superposing of additional circuits on telegraph and taneous channels of telegraphic and telephonic telephone conductors, so as to obtain two or more channels communication over the same wires, and the of independent communication from the same physical application of this method to the balancing circuit, now occupies a very important place in the design of telephonic relayed circuits. of such circuits. The loading of telephone circuits, combined with the (1) IMPEDANCE METHOD OF SIMULTANEOUS TELEGRAPH use of thermionic in them, now renders it AND TELEPHONE WORKING APPLIED TO SINGLE- possible to provide efficient long-distance telephonic com- WIRE WORKING. munication—including the provision of phantom circuits— on small-gauge conductors in underground cables carrying This method was first introduced by F. Van Ryssel- a large number of circuits. The provision of these cables berghe,* a Belgian telegraph engineer, in 1882. Com- involves the gradual replacement of overhead open telephone bined working is rendered possible by the different circuits in a large measure, and constitutes a revolution in impedance of coils and condensers respec- modern circuit provision. The object of this paper is to tively to high- and low-frequency currents. The review the present position of the art as applied to super- action depends upon the fundamental difference posed circuits worked at audio frequencies. of telegraph Morse signals and telephone speech The theory of simultaneous telegraph and telephone currents. working on the same circuit is first dealt with, and the Definition of telephone and telegraph signals, limitations of this method of working are stated. (a) Telephone signals.—The and currents The comparative efficiency of side and phantom circuits produced in a telephone circuit by the voice are assumed is then dealt with, and the principal problems arising from to vary in a simple harmonic manner. The voice the loading and relaying by amplifiers of phantom circuits covers a wide range of frequencies varying, say, between are outlined. The solution of the problems is indicated. 100 and 2 500 periods per second ; the mean frequency Finally an appendix is given, dealing with the derivation for purposes of calculation is taken at 800 periods per of formulae for the calculation of electric wave filters, for second, f application to the circuits dealt with. (b) Telegraph signals.—In the Morse code these con- sist of " dots " and " dashes." In hand signalling at the normal rate of 30 words per minute a dot has a INTRODUCTION. duration of 1/12 second and this will be used as a basis The problem of. utilizing telegraph and telephone of comparison. circuits to transmit two or more simultaneous electrical Explanation of the electrical action in an impeded communications over one circuit without interference combined telegraph and telephone circuit.—If an electric between them, has from an early date occupied the condenser of sufficient capacity (say 5/xF) is placed attention of those interested in telegraphy and telephony. in series with a telephone receiver in an ordinary During the past few years a considerable development telephone circuit, the attenuation of speech currents of some of the methods employed has taken place due to the condenser is so small as to be only just with far-reaching results ; and the object of this paper perceptible to the ear. The impedance of the condenser is to review the present position of the subject so far to very low-frequency currents is, however, very high. as the simultaneous transmission of telegraph signals On the other hand if a suitable inductive impedance and speech, at normal frequencies, and of phantom be placed in the line lead of a telegraph set the speed telephone working at such frequencies are concerned. of telegraph signalling is slowed down, but it is still In view of the very wide range of the problems possible to signal at hand speed on the telegraph circuit. involved it will be possible to deal only very briefly In addition to these direct actions in the telegraph with them. The methods most generally employed to and telephone circuit, and other direct actions which provide extra channels of telegraph or telephone com- will be shortly described, the action of each circuit on munication may be classified under the following the other requires consideration. The reduction in the heads : rate of rise of the telegraph signals reduces the normal * British Patents Nos. 1 303 and 2 466 of 1882, and 3 621 and (1) The Impedance or Retardation method of com- 5 503 of 1883. bined telegraphy and telephony over the same t " Conference Internationale des Techniciens des Administra- tions des Tel6graphes et Telephones de l'Europe." Paris, 1910. wire or wires. Comptes Rendus, p. 99. 676 HILL: PHANTOM TELEPHONE CIRCUITS, AND COMBINED TELEGRAPH tendency of those signals to cause disturbance in the matical point of view in the relatively simple circuit telephone apparatus via the condenser, in view of the shown in Fig. 2. simple fact that the disturbing signals vary in intensity in proportion to the rapidity of the rate of change Let E = a constant applied at A. of voltage at the terminals of the condenser. The v = the voltage at the terminals of the condenser retarding device in the telegraph circuit, however, has at the time t. a very high impedance to telephone speaking currents R = the resistance of the retardation coil. at the mean frequency of speech, and this high impe- L = the inductance in henrys of the retardation dance acts beneficially as regards the transmission of coil. speech, by reducing the transmission loss due to the Ri = the unvarying resistance of a long line telegraph apparatus acting as a shunt to speech currents. supposed to be non-inductive. The disturbing effect of speech currents on ordinary C = the capacity of the condenser in farads. Morse telegraphic apparatus is absolutely negligible. t = the time in seconds after the application of the voltage E to the circuit.

VAN RYSSELBERGHE SYSTEM. Then the equation connecting the voltage v with the time during the transient period when the E.M.F. is The working of this system depends fundamentally rising from zero to its steady value E, is on the actions described. A high inductive impedance RL is placed in the path of the telegraph apparatus dv R E and a condenser C in the path of the telephone apparatus, the two sets of apparatus being joined in parallel and connected to the line as indicated in Fig. 1 in which both ends of the circuit are shown. A single-current The corresponding equation for the value of v at any circuit is shown for simplicity, but double-current instant t after the application of E, is working is generally used. ERX _ l] (2) Line where 2 2/

- 2\L^CR]

E RL -oww° f 'III—IHH nnnnr<

Telephone Telegraph, Telegraph Telephone set set set set FIG. 1. FIG. 2.

The foregoing description shows in a general way If R\L = 1/CRi and R = the equation for v may that combined telegraph and telephone working on the be written same wire is possible, but it is of interest to study more closely the effect of the inductive impedance and the v = \E\\ — x/2e-tlcRsin {t/CR + {TT)} . . (4) condenser in order to find to what extent (if any) the retarded telegraph signals cause disturbance in the In any case if jS is imaginary the equation assumes a telephone set, and, further, how far the retardation form similar to (4). affects the telegraph signals themselves as regards Examination of these equations shows that the rate both shortening them and slowing down the rate of of rise of the voltage is logarithmic, and that by signalling. These matters will now be investigated. arranging suitable values of JB, L, C and Rx the rate In commercial combined telegraph and telephone of rise of the volts may be either advanced or retarded. circuits the apparatus arrangements are complex, as If RL represents a relay at the receiving end of a line may be seen by inspection of the diagrams here given, and C a condenser shunted by Rx at the same point and in addition the electrical action is materially we have the well-known arrangement of the shunted complicated by the fact that the permeability of the condenser as may be seen from Fig. 2. Fig. 3 gives iron cores and the effective resistances of the telegraph an oscillogram of the normal rise of current in an apparatus in circuit change materially with different unshunted Wheatstone telegraph receiver, and Fig. 4 current strengths. For these reasons experiment is shows the more rapid rate of rise obtained by adding usually preferred to calculation in determining the the shunted condenser to the same apparatus, the best values of the inductance and capacity required. adjustment being made to produce a signal as nearly The fundamental theoretical effect of the devices intro- square (flat-topped) as possible. duced to control the rate of rise of the current may, On the other hand, in order to produce the retardation however, be studied with advantage from the mathe- which is desired in a combined telegraph and telephone AND TELEPHONE CIRCUITS. 677 circuit, the following values which have been actually retardation which has just been examined has slowed used may be taken as an example : down the signal considerably. Fig. 3 shows a record of a normal signal passed through a Wheatstone receiver R = 50 ohms at a speed of 18 words per minute. In this case, in the L = 20 henrys first 1/100 second the current has risen to about 0-75 (7 = 5 microfarads of its full value, which is a much more rapid change than the case previously dealt with. If the value of the line resistance Rx be 500 ohms, the rate of rise is now greatly retarded as compared In order to see the effect of this changing current with Fig. 3. The variation of the voltage at the con- on the telephone apparatus, it should be noted that denser terminals with time is shown in Fig. 5. It is the telephone receiver has a marked sensitivity in the evident from these examples that the values of R, L neighbourhood of 800 to 1000 periods per second. and C must be carefully chosen. It is further evident If the maximum current passing through the Wheat- stone receiver is 10 mA there is a rate of change Ax of over 2 mA in the first 1/800 second (see Fig. 3) 1-0 and it is known that a current of this magnitude 0-8 0-6 100 0-4 0-2 O 001 002 0-03 004 005 0-06 007 Seconds FIG. 3. from the general equation (2) that in order to produce the least voltage at any instant t, E must be as small as the satisfactory working of the telegraph circuit will permit. It should also be noticed that although the value of R should be small in order to give a small time-constant yet the amount of v becomes less as R becomes greater, but of course the greater R may be unfavourable to the telegraph circuit. Although the more complex apparatus arrangement shown in Fig. 1 causes the effects just described to be modified, experiment clearly demonstrates the fact that relatively large values of L and C give good results.

20 -to 6O 80 1OO 120 500 500 5OO 50O 500 500 Time in seconds FIG. 5. 0-07 Seconds will produce a pronounced click in a telephone receiver. FIG. 4. The current attains 9/10ths of its maximum value after one-third of the full duration of the signal has It will now, however, be shown that it is not practicable elapsed. In the absence of combined telegraph and by an arrangement such as that shown in Fig. 2 to telephone working it is frequently necessary to increase eliminate entirely the disturbing effect of ordinary the rate of rise of signals instead of retarding them, Morse signals in the telephone circuit, due to the rate and the means of doing this have been dealt with. of change of voltage at the terminals of the condenser. It may be remarked that the retarding device RL An examination of Fig. 5 shows that in the first in Fig. 1 does not protect the telephone superimposed 1/100 second after the application of the E.M.F. the apparatus in the same circuit from the disturbing effects volts have risen to 18 per cent of their final value, of telegraph induction from contiguous uncomposited notwithstanding the heavy retardation, and such a rapid telegraph circuits. Inductive disturbances from such change of voltage would, with the voltages used in the circuits pass partly through the telephone apparatus ordinary Morse system, give rise to a click in the owing to the fact that the retarding inductance is in telephone receiver. the path of the telegraph apparatus only. On this If this is compared with the rapid rise of an ordinary account heavy induction from uncomposited contiguous Morse signal, however, it is at once seen that the telegraph circuits may affect the composited circuit. VOL. 60. 46 678 HILL: PHANTOM TELEPHONE CIRCUITS, AND COMBINED TELEGRAPH

The remedy is to place retarding devices in the circuits in practice that if the windings of the retardation coil which give rise to the disturbance. are joined up so as to oppose each other, balancing is much easier, whilst the telephone circuit may still be EFFECT OF THE RETARDING DEVICES ON THE RATE worked. The best position for the retardation coil in OF TELEGRAPH SIGNALLING IN AERIAL CIRCUITS. duplex circuits is in the lead common to the line and compensation circuit. Electric waves travel on unloaded aerial copper In addition to the rate of rise of the current, telegraph wires of small resistance with a speed approaching that signals have a periodicity, the wave-form being generally of light, whilst in heavily loaded underground circuits of a complicated nature. A Morse signal has a low the speed may be reduced to 8 000 miles per second. periodicity, and advantage may be taken of this fact The rate of telegraph working is largely influenced, so to tune anti-inductive apparatus as to eliminate or however, by the electrical sensitivity of the apparatus reduce inductive disturbance within a given band of employed, as well as its mechanical inertia, so that in frequencies, when the frequency of the disturbance is practice the rate of working depends on the combined sufficiently high. apparatus and line. From another point of view it The foregoing facts show the limitations of the also depends on the technical nature of the service system of combined telegraphy and telephony just utilized. For example, in simplex working a higher dealt with. all-round speed per channel is possible than in duplex working, and, again, a higher speed is obtainable per METHOD OF TELEPHONE RINGING. channel in duplex working than in quadruplex working. Further, it does not follow that the actual speed of The ordinary magneto generator, which is of low transmitting signals will give a proportionately eco- frequency—about 17 periods per second—is liable to nomical result, for the question of the operator's time interfere with the telegraph apparatus, and for this in preparing the message for transmission at the sending reason telephone signalling at a higher frequency is end and in preparing it for delivery at the receiving desirable. The best frequency is one which is high end enters into the matter, some systems involving enough to avoid interference with the telegraph appara- more time for this than others. tus, and at the same time is efficient for actuating the sensitive type of relay used. The frequency which For the purpose of the present problem, however, is found to be high enough for these purposes should the rate of rise of each signal is of immediate importance. not be exceeded, owing to the fact that high-frequency As an example, consider an aerial-line Morse Wheatstone currents cause inductive disturbance to other circuits ; circuit. It would be possible under the most favour- actually, a frequency near 150 periods per second is able conditions to obtain a speed of 600 words per found to be suitable. As a general rule some form of minute from this system, but it is preferred to work interrupter or buzzer is used to generate the calling at a speed which can generally be maintained with current on all types of composited circuits. Sometimes ease under unfavourable conditions, and 200 words per an interrupted direct current, actuated on the principle minute is frequently found to be convenient; this of the electric call-bell, is passed through the speaking is equivalent to 80 dots per second, so that a signal induction coil in order to convert the signals into an would last only 1/80 second. Referring to Fig. 5 it alternating current. will be seen that the voltage rises to only about one- third of its maximum value in that time, and it would The receiving device may be a loud form of telephone be impracticable to add a heavy retardation in such receiver with a horn, or a vibrating relay tuned to the a circuit. In practice, with the amount of retardation sending frequency, which brings a calling signal into necessary to reduce the noise in the telephone sufficiently operation. for satisfactory working, and bearing in mind the An arrangement developed by the Western Electric necessity for keeping the voltage of the telegraph Co. is as follows : The generator is of the vibrating-bar circuit as low as possible for the same reason, experi- type, the natural frequency of which is 140 periods per ment shows that 60 words per minute on an overhead second. The relay designed to receive the signals line working duplex can be obtained only under the consists of a bar tuned to the same frequency, and most favourable conditions. It should be mentioned actuated by the received currents, which pass through in this connection that if too much inductance is used bobbins wound on pole-pieces on each side of the reed. in the retarding device the telegraph signals may stick A light contact spring rests normally on the reed and and be sluggish. In order to provide a working margin makes contact with it. When the reed vibrates the of safety the Van Rysselberghe system is in fact generally contact is broken intermittently and this can be made worked only at hand speed. If this is assumed, elec- to release a relay connected in a local circuit. trical niters (see Appendix) may be used to cut off the The receiving apparatus must be so designed that currents of higher disturbing frequencies (see the it will not respond to the telegraph signals. example on page 690). It may be pointed out that the introduction of the GRADE OF TELEPHONE CIRCUIT OBTAINED. retarding device is further likely to cause difficulty A relatively poor circuit, unsuited for working on the in balancing telegraph duplex circuits, owing to the long-distance system, is usually obtained. This is due necessity of balancing in the compensation circuit of to several causes. A telegraph line wire may be iron the duplex, the effects of the retardation added to the and will, moreover, always be subject to a certain line. In order to minimize this difficulty it is found amount of disturbance. Generally the conductor will AND TELEPHONE CIRCUITS. 679

not be of a high-grade telephone type. If such a circuit is extended, complications arise, and , COMPOSITED CIRCUITS. involving further transmission loss, become necessary. Fig. 7 shows an arrangement used in England and This system works most successfully on relatively elsewhere for utilizing the two wires of a double-wire short circuits which do not require to be extended. circuit to form two duplex circuits. This arrangement If, however, the device is installed on a high-grade is known as composited working. In this country the copper conductor, the speaking range may be consider- arrangement is sometimes used to superpose telephone ably increased. call wires worked by telegraph, upon junction circuits . Since Van Rysselberghe's system numerous inventors worked on the common battery system (see Fig. 7), have exercised their ingenuity in producing the same or, alternatively, one wire is used as a telegraph call result by different means. As well-known examples wire, and the other for automatic signalling on the in this country may be cited the Phonopore circuit, junction circuit. In such a case each set of apparatus patented by Mr. C. L. Langdon-Davies (British Patent bridged on the telephone circuit involves a transmission No. 10 990 of 1884), and the Turchi-Brune system loss, which is usually not less than one mile of standard (British Patents Nos. 7 585 and 22 388 of 1903). cable, and may possibly exceed this figure. The loss Impedance device.—To give an idea of one of the is increased when similar circuits are extended to one many variations on the Van Rysselberghe method, another; it should therefore be ascertained, before

Line or Line or telegraph apparatus telegraph apparatus Coil Coil bridging bridging Transmitter Transmitter

I Receiver Receiver |

FIG. 6.

Fig. 6 has been prepared. This shows an arrangement doing this, whether the circuits have the necessary in which a bridging coil is placed in series with the transmission margin. receiver. When used in local circuits with a powerful It may be mentioned that* simplexed and composited transmitter it is capable of giving excellent results in sets have been used in this country for many years. circuits of suitable length. This apparatus suppresses (A simplexed circuit is a telegraph circuit worked in noise by sacrificing transmission efficiency in the receiving circuit, but if the receiver be made insensi- tive to the extent that it cuts down the volume of speech to the same degree as 20 miles of standard cable—which usually suffices to reduce loud noise to tolerable limits—there is still a possibility of utilizing the arrangement on many circuits, seeing that the commercial limit of speech is not reached in a quiet room until a total of 46 miles of standard cable is in circuit. It is of course desirable to keep far within that limit, and the margin available often permits this to be done on local circuits which do not require to be extended, when used in conjunction with a powerful transmitter. It should be noticed that " side tone " which pain- Repeating fully affects the ear when loud-speaking telephones are coil used in standard sets is practically reduced to negligible FIG. 7. proportions, owing to the circuit arrangement in this and similar cases. parallel over the two wires of a telephone circuit.) As an The Fullerphone.—This well-known device occupies a example I find that as far back as 1908, 200 circuits on prominent place in the literature of the subject, but the simplexed system were in use in this country. At the as it has already been fully described reference should same time a large number of circuits were in use for local be made to Major Fuller's paper.* telephone circuit working on a method devised by Mr. W. J. Medlyn which has largely been used in this country • " The Fullerphone and its Application to Civil and Military Telegraphy," Journal I.E.E., 1919, Supplement to vol. 57, p. 29. for this purpose and is described in Post Office Technical 680 HILL: PHANTOM TELEPHONE CIRCUITS, AND COMBINED TELEGRAPH

Circular E No. 10. The theoretical diagram is shown RELATIVE EFFICIENCY OF THE SIDE AND PHANTOM in Fig. 8 and the explanation is as follows : If on a short CIRCUITS IN AERIAL LINES. section of a pole route carrying a long main-line tele- graph circuit, we have also a local single-wire circuit The relative transmission efficiency of side and which it is desired to convert to telephone working, phantom circuits in aerial Lines depends essentially on such a circuit may be converted to double-wire working the arrangement of the wires on the poles. Every with little cost in the main-line section by utilizing in circuit has, of course, four electrical constants, namely, that length a long-distance telegraph circuit to form resistance, leakance, capacity, and inductance. Effec- the second wire of the local pair. The local single- tive, uniformly distributed values at alternating-current wire circuit and the telegraph circuit are crossed at voice frequencies are here assumed. intervals to give immunity from inductive disturbance The relative transmission efficiency of both side and due to contiguous telegraph circuits, and transformers phantom circuits is inversely proportional to the real are connected as shown. The section between the part f$ of the complex attenuation constant jS + ja transformers is the double-wire telephone circuit. and may be found from the following well-known Telegraph signals pass in parallel over the double- formula : wire telephone circuit thus formed, entering and leaving + j = {G + jcoC)] (5) the double-wire section at the centre point of the a transformers. By this method the continuity of the where jS = real part of attenuation constant, telegraph circuit is not interrupted. The working of a = wave-length constant, such a telephone loop is not disturbed by a superposed R = effective resistance per unit length, telegraph circuit if the A and B wires are of the same L = effective inductance per unit length, material and gauge. In the case shown, the telephone O = effective leakance per unit length, and circuit after leaving the main line is run on a single C = effective capacity per unit length.

a Wires on main telegraph line b on •o o p—2—£ :—CLI—

^Telephone AtoC

FIG. 8.

Side circuit Phantom circuit coil coil Phantom, circuit coil Side circuit coil jSide circuital B Ai Side circuit fl?2

Side circuit coil Theoretical arrangement of loading coils in circuit. Arrangement of loading-coil windings. FIG. 9.

wire which is connected to the transformers as shown. The transmission efficiency is also affected by the The possibility of using the single wire depends on impedance of the circuit. In the case of an electrically local conditions. long line the quantity ZQ is usually known as the characteristic impedance : and (2) TELEPHONE CIRCUIT PHANTOM WORKING. R + aiL The well-known equipotential method of superposing _ llff J \ now to be dealt with is of great importance and utility V V j (6) and of very wide application. Fig. 10 furnishes an example of the system. The earliest patent specifica- The minimum attenuation constant in unloaded tion in this country covering this form of superposing circuits is theoretically attained when we have the is that of Mr. F. Jacob (British Patent No. 231 of relation 1882). LO=CR (7) AND TELEPHONE CIRCUITS. 681

In phantom circuits where two wires are joined in length—say half a mile—and then cross them at parallel the resistance is generally half that of the regular and equal intervals of half a mile. This method side circuit, and the leakance double that of the side is being also adopted in this country. The transmission circuit. If, therefore, G is also doubled and L halved efficiency of the two anti-inductive methods described, as compared with the side circuit, equation (7) is satisfied both being in square formation, is nearly equal. and we have the conditions for minimum attenuation Experiments in this country showed a difference of for both side and phantom circuit. 2 per cent in the attenuation constant, being less to Now R and G are generally fixed for any combination that extent in the case of the wires run straight. of side and phantom circuits, and have the relative values given. Whether the minimum attenuation FOUR WIRES ARRANGED AT EQUAL HORIZONTAL constant is in question or not we have equal trans- DISTANCES ON THE SAME POLE-ARM. EFFECT ON mission efficiency, if, when the side-circuit constants TRANSMISSION. are C, O, L, R, the phantom-circuit constants are 2G, 20, \L, \R, and this furnishes a criterion for the The two side circuits being formed of adjacent wires comparison of transmission efficiency. As a general rule from left to right in each case and supposed to be crossed the capacity and inductance have not the proportion at equal intervals, it can be shown that the inductance shown. As is well known, the circuit inductance and L of the phantom circuit is greater than one-half that capacity vary according to a logarithmic function of of the side circuit, and that the capacity C of the the distance apart of the wires. The nearer together phantom circuit is less than twice the capacity of the the wires of the side circuit are, and in the case of the side circuit. As a consequence the phantom circuit phantom circuits the nearer together the side circuits has a lower attenuation than that of the side circuit, —which constitute the limbs of the phantom circuit— and this is verified in practice. the greater the capacity and the less the inductance, and vice versa. As a general rule the transmission TABLE 1. efficiency of a circuit is improved by separating the CAPACITY FORMULA FOR AERIAL LINES.* wires more widely, because the normal capacity of a circuit as compared with its inductance is in excess Wires arranged in square fonnation. from the point of view of the capacity and inductance Side circuit. Phantom circuit. required to produce the minimum attenuation constant. 1 1 The separation of the wires tends to bring both the C capacity and inductance nearer to the proportion that 4 log {d/r) c" = 2 log (d/2r) satisfies the minimum attenuation constant. There is, Inductance formulae. however, an accompanying disadvantage here, because L = 4 log {d/r) + fj, L = 2 log (d/2r) + fx/2 unless a circuit is perfectly balanced the separation of the wires increases inductive disturbance from sources r is the radius of a conductor, d is the distance between of disturbance outside the superposed four-wire system. the wires of the side circuit, and fjt, is the permeability of the conductor, assumed to be constant. WIRES ARRANGED ON THE POLE-ARMS AT THE FOUR Wires in a straight line on the same arm at equal CORNERS OF A SQUARE. EFFECT ON TRANS- distances, the wires composing the side circuit in each MISSION. case being crossed with each other at regular intervals. The relative transmission efficiency of two common arrangements will now be examined. If the four con- Side circuit. Phantom circuit, ductors of a phantom-circuit combination are arranged 1 log (V3d/2r) C = G = on the pole-arms so as to form a square as seen end-on, 4 log {d/r) ~ 2 log {d/r) log {3d/r) two wires being arranged on one arm and the other Inductance formulae. two immediately below on another arm, the side circuits being the diagonals of the square, it can be shown L = 4 log {d/r) + /x JD = 2 log (3-46d/r) + /x/2 theoretically that the relative values of capacity and inductance of side and phantom circuits are such RELATIVE EFFICIENCY OF SIDE AND PHANTOM CIRCUITS that the capacity of the phantom circuit is greater IN UNLOADED UNDERGROUND CIRCUITS OF THE than twice the capacity of the side circuit, and that MULTIPLE TWIN TYPE IN AIR-SPACE CABLES. the inductance of the phantom circuit is somewhat The first patent for this type of cable was that less than half that of the side circuit. It follows that of Dieselhorst and A. W. Martin (British Patent, the transmission efficiency of the phantom circuit is 12 526 of 1903). a little less than that of the side circuit. In these cables the relative transmission efficiency The arrangement described is that which has hitherto of side circuits and phantom circuits, respectively, been adopted in this country, the anti-inductive follows the general rule, and depends on the relative arrangement consisting in transposing the position of values of their inductance and capacity. Owing, each wire on the pole-arms so that a complete revolution however, to the fact that the wires making up the takes place between every four poles. side and phantom circuits in cables are closely packed An alternative anti-inductive method to that described, still retaining the square formation on the * See " Telephonic Transmission of Superimposed Circuits," by Kazukiyo Ogawa, Researches of the Electrotechnical Labora- four pole-arms, is to run the wires straight for a given tory, Department of Communications, Tokyo, Japan, Tuly, 1919. 682 HILL: PHANTOM TELEPHONE CIRCUITS, AND COMBINED TELEGRAPH together, the capacity is relatively much larger, and which is more efficient than the circuit on which it is the inductance much smaller than in the case of aerial superposed. wires, for the reason already given, and consequently the transmission efficiency of such unloaded circuits EXTENSION OF THE PHANTOM PRINCIPLE. is much less than those of the same conductor weight It should be noted that it is possible to carry the in aerial lines [see Formula (5)]. As a rule the capacity principle of phantoming further, and to obtain a second of telephone circuits in cables containing a large number phantom metallic circuit by combining two four-wire of wires is predicted from experiment, since it is largely cores and twisting them together in the same way affected by a factor depending on the form and dimen- as the two side circuits are twisted in order to form sions of the cable. Telephone air-space paper cables an ordinary phantom circuit. Eight wires would for phantom working are usually made up on the evidently be necessary for such a circuit, and the " multiple twin " formation, i.e. two telephone pairs principle may theoretically be continued, using 16 are twisted together to form a phantom pair, and the wires, 32 wires, etc., for successive phantom circuits relative capacity of side and phantom circuits may until the number of combinations is exhausted. In vary largely. The lay of the conductors is varied in practice, however, phantom circuits are limited to four-wire combinations at present. Loaded aide and phantom circuits. Explanatory diagram. SUBMARINE CABLES : CAPACITY AND TRANSMISSION RELATIONS. As regards submarine cables, which frequently contain only four wires, there is no difficulty in calcu- lating the capacity, which in the case of gutta-percha insulated cables is obtained from the formula

C = . . . . (8) logio {Did) Loaded phantom circuit Diagram showing electrical action. where D is the outer diameter of the gutta-percha, d the diameter of the copper conductor, and A is a factor proportional to the dielectric constant. The dielectric constant varies between 3 and 4. A figure often used is 3-6, and in that case A becomes 0*16 and the formula then gives the capacity of a single wire per nautical mile. Owing to their immersion in the sea the four wires XF of such cables are screened, and this simplifies the capacity relations ; for example, the capacity of a double- Expldnation of symbols Telephone wire circuit is one-half that of a single-wire circuit, Direction of current and the capacity of the metallic four-wire phantom Current passing through coil or transformer windings circuit is twice that of the side circuit. The capacity m inductive direction Current passing through coil or transformer windings of the four-wire earthed phantom circuit is four times in non-inductive direction that of a single earthed wire. As a consequence of Dotted lines indicate no current in that part of the circuit Solid lines show that a current is flowing these capacity relations a metallic loaded phantom XF Transformer ^ circuit has approximately the same efficiency as the S illustrates side circuit, dnd P phantom circuit working side circuit. The earthed phantom circuit, however, FIG. 10. is materially less efficient than the other two circuits, owing to the increased effective resistance, which is the usual way. Experiment on recent cables shows, observed in such circuits—e.g. an increase of about however, that the capacity of a phantom circuit has 2 • 5 ohms per naut per wire at 800 periods per second— usually a mean value approximately 50 per cent in owing to the interaction between the field of the excess of that of the side circuit. The natural induc- conductor and the metallic sheath of the cable, and the tance in telephone cables containing small-gauge con- effect of the return path consisting of sea water. (For ductors is so small as to be nearly negligible and will fuller details see Sir W. Noble's paper referred to on be left out of consideration. Artificial inductance is, page 684.) It will be remembered that the impedance of course, generally added to underground circuits of the side and phantom circuits varies according to which form part of the trunk or long-distance system, formula (6). The four wires in such a cable are usually and the problems of circuit balance presented by these arranged in quad formation, the diagonal wires forming loaded circuits will be dealt with further on. On the two side circuits and each side circuit forming account of the relatively small capacity of the phantom one limb of the phantom circuit. As seen end-on the circuit in such cables as compared with the capacity four wires are at the four corners of a square. In the of the side circuits, the phantom circuit is of a higher case of a four-wire loaded or unloaded submarine cable transmission efficiency than the side circuit, so that it is theoretically possible to obtain four telephone by superposing we actually obtain an additional circuit circuits without mutual interference. In the case of AND TELEPHONE CIRCUITS. 683 coil-loaded circuits two separate coils are used for every loading point (see Figs. 9 and 10). It will be the phantom circuit at each loading point, otherwise seen that the phantom-circuit coil is so arranged that the four-wire earthed phantom circuit would not be a current sent round each side-circuit coil in an loaded. The phantom coil arrangement referred to is inductive direction passes round the two halves of the the usual loading method of Messrs. Siemens Brothers phantom-circuit coil in opposite directions, so that no & Co. in such cases.* inductance is produced in the coil. On the contrary, currents flowing in the phantom-circuit coils in an COMBINED PHANTOM CIRCUIT AND COMPOSITED CIRCUIT inductive direction pass through the windings of the WORKING. side-circuit coils in a non-inductive direction. The It is possible under favourable circuit conditions theoretical electrical action in the loaded side and to combine the loaded phantom circuit, as shown phantom circuits is shown in Fig. 10 and it is thought in Fig. 10, with the composited arrangement shown that the diagrams will be found self-explanatory. in Fig. 7. For this purpose the telegraph sets arranged as in Fig. 7 may be added to the ends of the two RELATIVE TRANSMISSION EFFICIENCY AND CHEAPEST side circuits in Fig. 10. In order to secure a balanced COMBINATION OF LOADED SIDE AND PHANTOM arrangement it would be necessary for all the telegraph CIRCUITS IN UNDERGROUND CABLES. sets to have the same impedance at the same end of the The formula for the attenuation of a side and line. Such a circuit would be subject to the limitations phantom circuit when a>L is great in proportion to R pointed out in connection with the phantom circuit and coC great in proportion to G, which is the general and the composited circuits. It is believed that an equivalent arrangement is worked in America, but it has case, is not been used in Europe so far as the author is aware. (9) TRANSFORMERS IN PHANTOM CIRCUITS. This apparatus absorbs a certain amount of energy where RJLi is the effective resistance of the coil, and the effect on telephonic transmission requires divided by its inductance. Within practical working consideration. Moreover, the usual signalling fre- limits this ratio can be made constant for a side- quency is about 1 000 periods per minute, or approxi- circuit coil and for a phantom-circuit coil respectively, mately 17 per second, whereas the mean frequency of the two series of coils (side and phantom), however, speech is taken at 800 periods per second. The design being independent and separate; the ratio of the of a transformer of small dimensions to meet both circuit leakance G to the capacity G at a fixed fre- these requirements is difficult, and for that reason quency for different values of C is also constant. The there is an advantage in utilizing a frequency within other symbols have been already denned. Assuming the range of frequencies for which the transformer has RilLx and G/C to be the same for the side and phantom been designed. A transformer designed and sold by the circuits, the same value of j8 is obtained for both cir- Western Electric Co. and known as " type 4006A " is, cuits, if the side circuit has double the inductance of however, available for this purpose. At a frequency of the phantom circuit, and if the side circuit has one- 133 periods per second it is found to be very efficient half the capacity of the phantom circuit. As shown for signalling, and it is specially efficient at telephonic in the preceding paragraphs, however, the capacity of voice frequencies. the phantom circuit in underground cables is only The Post Office Specification which requires an out- about 50 per cent greater than that of the side cir- put of 80 per cent of the input energy at a frequency cuit, so that the arrangement just described does not / such that 2rrf = 5 000, and of 55 per cent at a fre- hold good. quency / such that 2TT/ = 100, is met by this appa- In practice, the relative efficiency of transmission of ratus ; it also fulfils the stringent condition imposed the side and phantom circuits is largely determined by for differential action when currents are sent through commercial conditions as well as electrical possibilities, the windings in opposite directions. the main considerations being as follows :— The transmission loss (core loss) observed in com- It is economical to space loading cofls as widely parative speech tests does not exceed the loss which apart as possible in order to obtain a given average takes place when speaking through one mile of the inductance per mile, but it can be shown that the standard cable. wider the spacing the more the higher frequencies It may be mentioned that the method of winding which enter into the composition of speech are cut to secure the differential action is one devised by the off, and beyond a certain cut-off point telephonic Post Office. speech becomes indistinct and impossible. Owing to Other manufacturers also make efficient transformers. the indeterminate nature of the frequencies entering into speech it is necessary to make experiments to LOADED PHANTOM CIRCUITS : ELECTRICAL ACTION. determine the widest permissible spacing. The rule It is assumed that the principles of loading are hitherto obtaining as a result of experiment in Great understood. When the phantom circuit is loaded a Britain is as follows : separate and extra coil is usually required for it at GDL = 25 (10) * J. G. HILL : " Two Loaded Phantom Circuits in a 4-Wire Submarine Cable," Journal of the Institution of Post Office Electricalwhere C is the capacity of the cable circuit in micro- Engineers, 1914, vol. 7, p. 197. farads per mile, D is the distance in miles separating G84 HILL: PHANTOM TELEPHONE CIRCUITS, AND COMBINED TELEGRAPH the coils and L is the inductance of the circuit is complex. In order to balance such a cir- in millihenrys. This rule applies to circuits of length, cuit perfectly the potential in the A and B wires of say, not exceeding 350 miles. Beyond that point the the side circuits (which are joined in parallel, see upper frequencies are too much attenuated and other Fig. 10) must be equal for all the frequencies of the phenomena become evident, making it necessary to voice and, unless the impedances of the A and B lines reduce the inductance without altering C or D in (in magnitude and phase) are equal at all points formula (10). Again, in practice it is economical to equidistant from the sending end, the want of electrical have only one set of manholes to accommodate both balance will result in overhearing if the permissible side and phantom loading coils. The most econo- limits of unbalance are exceeded. Now the loading of mical arrangement for the two side circuits generally the circuits augments the difficulties which may have leads to the greatest total economy of the four-wire existed in the unloaded pairs, principally because the combination, and the position of the manholes is circuits have thereby become more efficient carriers of therefore made to suit the side circuits ; this means energy, and also due to the shortened wave-length that there is no choice of spacing for the phantom (and this applies both to the circuit in which the over- coils. The only tiling that can be done for the phantom hearing arises and contiguous circuits which transmit circuit is to increase its inductance above the hypo- the augmented disturbance more efficiently than thetical value of one-half that of the side circuit up before). Alternatively we may say that if the circuits to the limit permissible by the spacing rule, and this are long enough, the mean voltage in a loaded circuit procedure gives a phantom circuit having a lower with the same applied E.M.F. is greater than in an attenuation constant, and therefore a higher trans- unloaded one, and any want of equality in the con- mission efficiency, than the side circuit, usually to stants of the A and B lines will result in a greater the extent of 15 to 25 per cent (owing to the relatively difference of potential between them, resulting in smaller capacity of the phantom circuit and its higher greater overhearing between the side and phantom degree of loading). circuits, and in cross-talk between the side circuits. For a given transmission equivalent—say S miles of In order to secure a perfect balance, the resistance, standard cable—between two given fixed points, such capacity, inductance and leakance of the four wires of a for example as a transmission equivalent of 15 miles phantom circuit must be theoretically equal. It is pos- of standard cable between London and Leicester, it is sible to a large extent to control the equality of the possible to find the cheapest possible combination of conductor resistance of the cable in the factory. The loading coil and cable (including side circuits and inductance and effective resistance of the loading phantom circuits) which will produce the required coils in a coil-loaded cable can also be very closely transmission efficiency. This is obtained from the balanced. In the loading coils used by the Post formula Office and manufactured by the Western Electric Co. a variation of not more than 0-25 per cent between the inductance of the two halves of the same loading where coil and not more than 0 • 1 ohm between the direct- P — the combined cost of one mile of cable pair and current resistance of the two halves of the same coil its loading, is guaranteed. A variation of not more than 2 per cent in the inductance of different coils is required A — RPX where R is the resistance of one mile of a known and selected conductor pair and Px its and is possible. The leakance of the cable can also cost, be efficiently controlled in manufacture, supplemented by care in laying the cable, and it does not cause B ~ y/{x/R)P2 where x = Lm/L. Lm is the maxi- mum loading of the conductor of resistance R appreciable difficulty. The electrostatic capacity of and L is the loading which when combined with each wire in a four-wire core of the cable cannot, R gives the required attenuation constant /?. however, be sufficiently equalized in an economical P2 is the cost per mile of providing L and manner in the factory to produce phantom circuits includes coils and manholes. which do not give rise to serious overhearing between side and phantom-loaded circuits in the same four- /) 2(2++/2). Herex2=Lm2/L2, wire core (or four wires making up the phantom cir- which is the maximum loading L??i2 for R2 divided by cuit). Special steps are therefore taken to equalize an inductance L2, which inductance when associated with R» in a cable pair gives the required jS. R the capacity of the wires in each four-wire group 2 during the process of laying. This is a recognized and L2 are the unknown values which may be derived from x * necessity in all countries where loaded phantom cable 2 circuits are used.* The steps taken result in reducing the cross-talk OVERHEARING AND CROSS-TALK DUE TO THE WANT between the side circuits, as well as making it possible OF UNIFORMITY OF THE ELECTRICAL CONSTANTS OF to use the phantom circuits. It may be noted that THE FOUR WIRES IN A PHANTOM CIRCUIT, WITH * S. A. POLLOCK : Journal of the Institution of Post Office Electrical NOTES ON THE METHOD OF ELIMINATING THEM. Engineers (April 1914 and January 1915). Under his personal direction the present British system of cable-capacity balancing It will be evident from the preceding description has been developed. that the circuit arrangement of a loaded phantom E. S. RITTER and A. MORRIS : " The Capacity and Insulation of Cables," ibid., vol. 12, part 4. • For proof see Chapter XII of "Telephonic Transmission," Also Sir W. NOBLE : " The Lohg-Distance Telephone System by J. G. HILL. (Longmans, Green & Co.) of the United Kingdom," Journal I.E.E., 1921, vol. 69, p. 389. AND TELEPHONE CIRCUITS. 685 the necessity for balancing arose primarily owing to not usually phantomed at present, but cross-talk in overhearing between loaded side and phantom cir- adjacent side circuits may nevertheless arise, par- cuits. It was formerly common practice to work ticularly at the sending end of the cable and at relay unbalanced and loaded cable pairs in cables which stations, notwithstanding the best possible balancing did not contain phantom circuits. of capacity. As a consequence it becomes necessary As a matter of interest it may be placed on record I to separate the sending circuits from the receiving that the first loaded phantom-circuit experiments in I circuits. It should be noted that the received cur- this country were carried out, and the possibility of rents although they may be of great amplitude are this type of loading demonstrated, between Liverpool I often materially less than the sent currents. and Manchester in 1908. As specially designed loading coils were not available for the phantom circuit coils, NOTES ON THE BALANCING OF CAPACITY IN telephone transformer coils were used as a substitute, the equal windings of a separate transformer being UNDERGROUND CABLES. used for each of the two side circuits which made up In view of the importance of this problem a con- the phantom circuit, i.e. an arrangement equivalent to i siderable amount of attention has been devoted to it two separate phantom coils for the phantom circuit ' in this country. The necessary conditions for pro- at the same geographical point, instead of one in the 1 ducing the required capacity balance have been fully well-known Western Electric Co.'s design. In 1912 it worked out, theoretically and practically. A printed was further demonstrated in this country that phantom j instruction (T.I.XIX Post Office Engineering Depart- circuits in the type of cable then available when using i ment) has been prepared, covering both branches of the best available loading coils, were subject to marked : the subject, and supplied to British manufacturers overhearing between the side and phantom circuits ; | who undertake the work of supplying the cables and but the difficulties were quickly overcome. balancing them on the road. Notwithstanding the complexity of the subject our EFFECT OF THE INTRODUCTION OF TELEPHONIC RELAYS manufacturers have met all past requirements, and it ON OVERHEARING AND CROSS-TALK. must be recognized that without their effective and Owing to the introduction of two-wire telephone willing co-operation the progress made would not relay working, a degree of accuracy of capacity balance have been possible. is now required which was not considered necessary In view of the detailed instruction named it is not for loaded phantom circuits without relays. The necessary to enter into details of the method in main reason is that the telephone relay augments question ; but the following general remarks may be the speech current in the circuit, and where there are of interest. several relays the attenuated current arriving at any As already mentioned, the form of cable utilized given relay station may be augmented to approxi- for loaded, phantomed and relayed underground cir- mately the same amplitude as that originally sent cuits is of the M.T. type and it is usually made up of from the office of origin, whereas in a long loaded conductors weighing 20 lb. or 40 lb. per mile per wire. circuit without relays the current is attenuated pro- The number of wires varies with the service require- gressively and continuously in accordance with the ments up to 308 pairs per cable. Each four-wire core compound interest law. As a consequence the mean in a cable is balanced in short lengths, because if the disturbing voltage in the circuit is much higher in the lengths taken were too long the electrical balance of the constants would not be the same for different fre- case of the relayed than the unrelayed two-wire cir- quencies. The complete unit taken for independent cuit. Contiguous relay circuits have also a greater balancing is a loading-coil section. The maximum transmission efficiency, and transmit disturbances with length of a section hitherto adopted in this country greater intensity than in the case of unrelayed circuits. is 2*6 miles, but present practice favours 1 • 6 miles and 1-125 miles. Each loading-coil section is sub- EFFECT OF FOUR-WIRE TELEPHONE RELAY WORKING divided into lengths of one-tenth of a mile, and these ON CROSS-TALK. are usually tested separately, and combined with other It is well, known that in some cases there is at similar lengths to produce the best result in the loading- present an advantage in utilizing four wires for one coil section. It can be shown that the impedance telephone underground loaded and relayed circuit, of a sufficiently short length of cable measured from two wires being used for the sending circuit and two one end of the cable with the distant end open depends for the receiving circuit. This is largely because such essentially on the capacity and leakance of the length an arrangement makes it possible to utilize a much under test, whilst with the circuit closed only the higher degree of relay amplification and to use smaller resistance and inductance enter into the measurement. gauge conductors than in the two-wire relay circuit, These facts form the basis of the procedure. The or, alternatively, to obtain a louder speech volume leakance, taking into account the high insulation now with the same type of conductor than is possible with obtainable, and the inductance of short unloaded the two-wire system (see British Patent No. 29165 lengths of cable are relatively unimportant as factors of 1913 by A. S. J. Van Kesteren). It is not difficult in causing overhearing in the type of cable considered. to foresee that the greater amplification of speech (An insulation of 10 000 megohms per mile per wire energy results in greater inductive disturbance in con- in air-space paper cable is now obtainable when tiguous circuits. In such cases four-wire circuits are measured by direct current, but the effective leakance 686 HILL: PHANTOM TELEPHONE CIRCUITS, AND COMBINED TELEGRAPH at 800 periods per second corresponds to only about artificial cable, the readings being expressed in terms 1 megohm per mile, wire to wire. The inductance of millionths of the current entering the cross-talk per mile per pair is 1 mH.) The capacity system of meter, instead of the more usual arrangement of read- a four-wire core may be represented by 10 capacities ing in miles of standard cable. It is, however, easy to between different wires and between wires and earth, convert from the one set of readings to the other.* and it can be shown that from six capacity measure- The instrument is arranged to have a constant send- ments the necessary combinations to avoid over- ing-end impedance of approximately 666 ohms, no hearing and cross-talk can be deduced (see Fig. 11). matter what equivalent attenuation length is in cir- These measurements are made by means of a double cuit ; this impedance is generally not the same as bridge specially designed for use on the road. Appa- the section of cable under test, and this complicates ratus is also designed for the rapid measurement of to some extent the interpretation of the results. The the degree of equality of the resistance of the con- comparative overhearing test is made by means of ductors. Special steps, however, are required in the telephones of standard type, by speaking on the cir- factory to reduce the out-of-balance of resistance as cuit giving rise to the interruption (say a side circuit) far as possible, and this is considered much prefer- which is closed by a resistance equal to the charac- able to compensation on the road by crossing. The teristic impedance, and listening on the disturbed circuit (say a phantom circuit). The degree of intensity of the sound is observed in the receiver and the tele- phones are then switched over to the cross-talk meter, which is altered until a sound of the same intensity c y=T= Z=T= is heard in the receiver as before. The number of millionths recorded on the dial of the cross-talk meter then indicates the magnitude of the overhearing. The relation between the capacity measurements fa) Capacity relations in a phantom circuit and the observed cross-talk is such that if the follow- ing values are not exceeded the result is usually satisfactory (see Fig. 11) :

2(p + Q) + w = 110 2(r + s) + v = 110 {p-q) = 80 u = 200 v = 200 This, however, assumes a uniform distribution of capacity unbalance. In abnormal cases these figures may not correspond closely enough to the mean over- hearing and cross-talk observed in normal cases. The overhearing and cross-talk must, however, at present always be taken as the ultimate criteria, inasmuch as (b) Symmetrical arrangement of the above they indicate the necessary commercial requirements. Conditions in 4-wire quad for no overhearing and no cross- As regards resistance unbalance there is no fixed talk. rule, but experience shows that the difference in resist- p = w — x = o s = x — y = 0 q-,z -2/=0 it = a - b = 0 ance of the a and b wires expressed as a percentage r - w - z — 0 v = c — d. = 0 of the loop resistance does not usually exceed a mean FIG. 11. value of 0-05. The question of capacity balancing has received considerable attention in America and reduction of road balancing of resistance to a minimum Germany. As regards the former see British Patents is therefore aimed at. Incidentally it may be mentioned Nos. 2 009 and 2 508 of 1913, granted to the Western that it is highly important that the cable constants, Electric Co. including capacity, should be made as uniform as Messrs. Siemens and Halske, Berlin, have patented possible in the factor}', in view of the exacting require- a process for balancing capacity in cables by the addi- ments of modern developments and the difficulty in tion of small condensers to the cable conductors at replacing non-uniform cables when faulty. The suitable distances (British Patent No. 147 013). The amount of out-of-balance of capacity and resistance condensers employed are made of rolled paper, inserted which can be permitted depend essentially on the and sealed in glass tubes, and joined in the cables at relation between these quantities and the corresponding loading points. They may be enclosed in separate amount of cross-talk and overhearing resulting from iron boxes or in loading-coil pots. It is claimed that them; tests of these quantities are therefore neces- the device is very successful.} sary. The cross-talk meter is used for this purpose. Professor F. Breisig has treated mathematically the This instrument, designed by the Western Electric Co. and manufactured in this country by the Cam- * See J. G. HILL : " Telephonic Transmission," p. 257. t A. EBELING : " Fernkabel und Verstarkung," Electrotech- bridge and Paul Instrument Co., is a distortionless nische Zeitschrift, 1921, vol. 42, p. 873. AND TELEPHONE CIRCUITS. 687 problem of overhearing and cross-talk in phantom In order that the voice currents through the differ- circuits.* He arrives at the conclusion that the over- ential windings of the transformer may be equal a hearing in short lengths of uniform cable such as load- complex problem requires solution. It will be remem- ing-coil sections with the receiving end insulated, is bered that the voice includes a large range of tones essentially due to capacity unbalance, and that speech of different frequencies, varying from, say, 100 to tests may accurately determine its magnitude if a more than 2 000 periods per second. All these tones suitably designed artificial cable is used. Details of are simultaneously impressed on the circuit, and the such an artificial cable are given. The main feature artificial networks which balance the up and the down of the artificial cable is that it is constructed of con- lines must be constructed to produce differential densers so as to vary in impedance with frequency in action through the transformer windings with all the same way as the actual cable under test, and that these frequencies simultaneously operating. Now, the its impedance is different from its two ends so as to characteristic impedance of the line is generally be the same as that of the circuits under test. One different for every frequency. The characteristic impe- end, for example, may be made the same as the phantom dance of a uniform unloaded line is given by formula circuit, and the other the same as the side circuit, f (6). In the case of a uniformly series-loaded line with the terminal loading coils spaced at a distance CONTINUOUSLY LOADED CABLES. DIFFICULTIES IN of one-half a loading-coil section from the ends of the BALANCING. line, the characteristic impedance ZQI is given by the following formula : When continuously loaded cables require balancing the process is much more complex than in the case /2Z sinh 6 + Za (cosh 9 + 1) 0 (12) of coil-loaded cables, owing to the fact that the induc- 2Z sinh $"+ Za (cosh 6 — 1) ' tance added by the iron wire also requires balancing. Q This subject is at present receiving special study in where 6 is the line angle between two loading coils, the British Post Office. ZQ is the characteristic impedance of the unloaded line, and Za the impedance of the loading coil. In EFFECTS OF THE NON-UNIFORMITY OF THE LINE both cases calculation from these formulae shows that CONSTANTS OF A LOADED TELEPHONE CIRCUIT the characteristic impedance changes with frequency in ON TELEPHONE RELAY WORKING IN THE SAME a smooth, regular way. Notwithstanding the com- CIRCUIT. plexity of the physical problem involved in balancing the multiple frequencies of the voice in unloaded or The difficulty which we are about to consider arises loaded lines, a comparatively simple balancing net- from the circuit arrangement which it has been found work can be designed which involves capacity, advantageous to adopt for telephonic relay working in inductance, and resistance, and which effectively both side and phantom circuits. It should be ex- balances the differential transformer and circuit pre- plained that when relays were first used in telephone viously referred to if the conditions of formulae (6) circuits at intermediate points the relayed voice and (12) are fulfilled. The network is constructed to current was applied to the circuit in bridge, through balance the circuit for a range of frequencies between the differential windings of a transformer, one half 300 and 2 000 periods per second. Filters are neces- of the windings being connected to the " up " line and sary to cut out frequencies outside the range for which the other half to the " down" line. Under these the circuit is balanced. The result of the application conditions if the " up " and " down " lines are not of these devices is that the repeaters used in double- of equal impedance the electrical action in the relay wire working are stable and give good articulation of windings is not differential and the is actuated, speech. Detailed examples of the design of balancing and sends a further current to line. As a result of this networks have been given.* continued action and reaction the amplifier may give out a loud note. This difficulty may to some extent In the case of unloaded, uniform cable lines the be minimized by reducing the sensitiveness of the line impedance which requires balancing may be taken relay, but the amplification of speech is not then so to be that calculated from formula (6). This involves great as it would otherwise be, or, alternatively, if constructing a network for the given circuit through the sensitiveness is not reduced the arrangement is the range of frequencies first specified. The calculated unstable. Further, if the out-of-balance currents sent impedance in this case usually approximates with into the relay are not sufficient to cause howling they sufficient accuracy to the measured impedance of that may interfere with the clearness of the amplified circuit, through the same range of frequencies dealt speech. If more than one relay is introduced into with. If, however, the circuit is series-loaded the a circuit the difficulties are increased. In order to values calculated may differ appreciably from the overcome these difficulties double-relay working was measured impedances. Fig. 12 shows a calculated introduced, i.e. the " up" and " down" lines were impedance curve of a loaded circuit in dotted lines, divided and a separate relay allotted to each, the line and the result of the measured impedance over the in each case being balanced by an artificial network. same frequency range is given in solid lines. The difference is due to the want of uniformity of the * " ijber das Nebensprechen In Fernsprechkreisen," Electro- • C. ROBINSON and R. M. CHAMNEV : " Gas Discharge Tele- technische Zeitschrift, 1921, vol. 42, p. 933. phone Relays and their Application to Commercial Circuits," t L. LICHTENSTEIN : " Uber das Nebensprechen in Kombinierten Professional Papers of the Institution of Post Office Electrical Fernsprechamtern," E.T.Z., 1920, 4 and 11 March. Engineers, No. 76. G88 HILL: PHANTOM TELEPHONE CIRCUITS, AND COMBINED TELEGRAPH

circuit constants. Formula (12) is based on the sup- inductance per section, and that the electrostatic posed uniform distribution of the capacity, leakance capacity per section should not vary more than \\ per and resistance throughout the circuit, and also sup- cent from the mean capacity per section. The same poses equal inductance and equal spacing of the loading general problem has also received careful attention in coils. Where these conditions are departed from, America. notably as regards uniformity of spacing and variation When the limitations of capacity, inductance and in inductance and capacity, reflection takes place at spacing here outlined are observed, the variations points of non-uniformity, and the effect varies with from the calculated curves—such for example as frequency. The result of these variations is shown in shown by Fig. 12—are materially reduced, the amplitude Fig. 12. Previous to the introduction of relay work- of the sinuous curve being reduced to negligible ing this phenomenon had comparatively little im- proportions. portance, and it may be mentioned that the circuit Before concluding it should be stated that it has which is here taken as an example and illustrated in not been overlooked that a system of high-frequency Fig. 12 was designed for pre-relay conditions, but such telegraphy and telephony has recently been developed a circuit is not sufficiently uniform to give the results in America, largely due to the labours of Major-General required in highly efficient relay working. It now G. O. Squier, and also the Western Electric Co. in becomes necessary to specify limits to the permissible co-operation with the American Telephone and Tele- variation of (a) the spacing distance of the loading graph Co. It would have been impossible, however, coils, (6) the inductance of any coil from the mean in any case to deal properly with the subject in the value of inductance per coil, (c) the capacity of any limits of this paper. High-frequency circuits are at present regarded, moreover, as supplementing, but 1300 not supplanting, the methods here described. The Impedance of loolb.loaded underground. Loading 133 millihenrys at z-566m spacing. subject is dealt with in Major-General Squier's pub- 1200 Curve A, Calculated characteristic impedance lications and has also been treated by E. H. Colpitts taking spacing; into account. and O. B. Blackwell.* 1100 Curve C, Measured impedance. r\ The study of superposed circuits, used in the widest sense, is one of increasing importance. It is not one for the application of cut-and-dried rules. The problems dealt with are so complex as to require special study of each case in all its aspects. The progress which has been made in recent years has resulted in the replacing, to some extent, of heavy-gauge overhead wires of limited electrical stability and carrying capa- city, by small loaded and relayed underground cir- 320 •lOO 480 560 640 720 800 880 960 1040 1120 1200 1280 cuits of uniform stability and efficiency. It must not be hastily assumed, however, that the reduction Frequency, in periods per second in cost is at all proportionate to the decrease in the FIG. 12. weight of copper per mile. The cost of loading such cables is considerable, and to this must be added the loading-coil section from the mean value of capacity cost of providing and maintaining relay stations, with per section. Experiment shows that leakance and the necessary equipment and staff. It may be stated resistance do not cause appreciable difficulty as regards as a broad proposition that a loaded circuit of a given present requirements of network impedance balances. transmission efficiency may be provided by a rela- The requirements as regards inductance in this country- tively large quantity of copper and a relatively small are that the inductance of any coil must not vary quantity of loading coils, or, conversely, within known from the mean inductance of each coil by more than limits, by a less quantity of copper and more loading 2 per cent, and this requirement is fully met in the coils. The engineering problem consists in providing highly efficient loading coils furnished by the Western the required efficiency at the cheapest cost. Similarly Electric Company. Further, the loading coils must with relay stations ; the more of these we provide be spaced at intervals which do not vary more than 2 per cent from the mean spacing distance. the less the amount of copper and coils which it is necessary to put into the loaded circuits for a given As regards capacity, a restriction of the amount of overall transmission efficiency, and vice versa; but non-uniformity permissible has not yet been fixed in there is a point beyond which it is cheaper to provide this country, but experiments made in Great Britain loaded conductors than relay stations. The ultimate point to the necessity of some limitation, of the same object is to find that combination which for a given order of accuracy as in the case of the inductance. number of years gives the required overall efficiency Recent extensive experiments made in Germany on the influence of inequalities in the construction of at the most economical cost. This is merely one coil-loaded telephone cables on their characteristic example of many problems which require solution. impedance by K. W. Wagner and K. Kiipfmuller * The design and calculation of such a system involves point to the fact that the inductance per section engineering problems of great complexity, as may be * " Carrier Current Telephony and Telegraphy," Journal of the should not vary more than 2 per cent from the mean American I.E.E., 1921, vol. 40, p. 310. See also "Multiplex Telegraphy and Telephony," by K. W. WAGNER, Eleklrotechnische • Archiv fur Elektrotcchnik, 1921, vol. 9, p. 461. Zeitschrift, 1919, vol. 40. AND TELEPHONE CIRCUITS. 089 seen from the details given in the paper. The ulti- will always be 12/I1 = e~n9, where e is the base of the mate result, however, should be a system of greater natural logarithms and 6 is the equivalent propagation efficiency and stability than any provided in the past. constant of the artificial section. Further, 6 is a The author is obliged to Mr. A. Morris for reading vector quantity such that the proofs and for suggestions. d=ja-\-p (13) where a represents the retardation of the wave per section, and jS the real part of the attenviation con- APPENDIX. stant. If, instead of closing the circuit through- an ELECTRICAL FILTERS. Zz =Z0tanh fi/2 Zz=Zntaiih 0/z In some cases, more particularly when relatively high speech-magnification is utilized, the higher un- balanced frequencies which are not dealt with in <*t) the impedance balancing network are liable to set up oscillations in telephone repeaters, and electric L filters may be employed to cut off all frequencies above yoo'o'O'o the range which experiment shows may be dispensed with without appreciable distortion of speech. (b) The method of calculation used for filters may also uC Sinh be utilized to calculate impedances for use at the end of composited circuits. Filters are also of considerable importance in con- nection with high-frequency guided telephony and telegraphy. It may therefore be of use to outline a relatively simple method of arriving at the funda- «•) mental data of these devices, more especially as the method is applicable to artificial circuits generally. The properties of such circuits are well known, and the object of the following notes is to show a method of obtaining from known data the principal details for the calculation of filters by inspection of tabulated formulae. The principal formulae for the calculation of the transmission characteristics of electrical filters may be written down by inspection of the known impedance of the three branches which make up an artificial T circuit, on the assumption that a source of sine- wave E.M.F. in the steady state is transmitted by the filter circuits. It will simplify the matter to recall the principal properties of such circuits, and they are as follows :— ((') If an artificial circuit is made up of any three inde- pendent impedances Zx and Z2 arranged as in Fig. 13 (a) where the two equal branches Z2 form the arms of the (S) arrangement, with the middle point earthed through an impedance Zx we have one section of the proposed = Mutual inductance artificial circuit. Suppose this section to attenuate between two equal the voltage and current transmitted through it. . If now halves .3 and j^ of the number of equal sections be indefinitely increased the same coil and joined in series, the impedance of the combination j£=Indiictance of coil. as measured from the sending end A will become con- stant, whether the distant end is opened or closed. FIG. 13. This impedance will be called the characteristic impe- dance and will be represented by the symbol ZQ. If impedance, one section be taken and earthed at the now a single impedance be constructed having the receiving end we have the relation value ZQ, and if this impedance be used to close one section or to close any number of sections, the impe- IxfJ2 = cosh d (14) dance as measured from A will always be ZQ in Now to determine the transmission properties of the same conditions of test. So long as the preceding filters we require to know the quantities just defined, conditions are fulfilled the ratio between the current namely, ZQ, 0, and cosh 6, as will be shown later by Ii at the beginning of any section and the current an example. These quantities may be written down Jo leaving the same section or any number n of sections directly in terms of Zx and Z2 if these independent 690 HILL: PHANTOM TELEPHONE CIRCUITS, AND COMBINED TELEGRAPH

impedances are arranged in an artificial circuit as in but the retardation per section during this range Fig. 13 (a) and have the following values assigned to remains constant with the value IT. them : Taking the following values for illustration, if C — 2-2 JHF and L = 10 henrys 1 sinh 0 Then when co = 300, j8 = 0, a = TT % = ZQ tanh |0 (16) (D =600, j§= 2-634, a = ir All the preceding formulae are proved in text-books ZQ decreases with frequency, becoming zero when on telephonic transmission. uPCL = 2. For values of co^CL greater than 2, ZQ is imaginary. To find ZQ in terms of Z± and Z2, by inspection of equations (15) and (16) we have Such a filter passes all frequencies below 48 per second without attenuation. Z% -f- 2Z2 x Zx The same method may be followed in more com- 2 — ZQ tanh |0 + 2Z0 tanh |0 x Zolsmh 0 plex cases.

- in ,_2 ELECTRICAL FILTERS WITH MUTUAL INDUCTANCE. [_ sinh2 0 <> sinh* 0 ~J ~ ° Suppose as a simple case of such a filter that we whence Zo = -\/{Z2 + 2Z±Z2) . . . . (17) have a continuous resistanceless inductive coil con- e Now e = sinh 0 + cosh 0 nected as shown in Fig. 13(/), the middle of the coil By inspection of (15) being connected to earth through a condenser; there will then be mutual inductance between the arms of ZQ/ZI = sinh 0 (18) the T. If Z2 is used as a general symbol for the im- To find cosh 0, we have pedance of each of the horizontal branches of a T cir- = Zo tanh \6 X sinh 0/Zo cuit, and Zi is the impedance of the vertical branch cosh 0 - 1 of the same circuit, excluding in each case the portion X sinh 0 = cosh 0 — 1 ; sinh 0 due to the mutual inductance M, it can be shown * that 1 whence -}- l = cosh0 (19) i = cosh 0 = (22)

.-. ee = sinh 0 + cosh 0 =.-. f° + (f? + l) . (20) . (23) It can further be shown that the results given in (22) and (23) are identical with those obtained from a simple The real part jS of the complex attenuation constant T circuit in which the independent value Z2—as in 0 is Fig. 13 (a)—is replaced by Z2 + jcoM = Z2M (say) in -log. mod . . (21) each horizontal arm of the T circuit, whilst at the same time Zx is replaced by the quantity Zx — jcoM = Zuj. The impedances Z and Z may be complex vectors. i.e. the logarithm of the modulus of the quantity in x 2 The mutual inductance may be positive or negative the brackets in (21). and have any desired coupling. The -j- sign is used As an example of the application of these formula? for convenience. Now the total impedance of the consider the simple case depicted in Fig. 13(6), which supposed resistanceless coil, which has a self-induction a filter with negligible resistance. is L and a mutual inductance 2M, is JOJL + 2ju)M, and We have by inspection of (15) to (20) and Fig. 13(6) Fig. 13(/) gives the fundamental data for the calcu- lation of the transmission constants of such a coil Cosh 0 = ^+1= _^(tC) + 1=1- ufiCL when used as a filter. Applying the preceding data 2Z Z = - o>2£2 + 2L/C we have Z2~ j(x>\L, Zi= —j/(a)C). In order to X 2 deduce the value of cosh 0 by means of a simple T circuit we proceed as follows :

e By analogy with Fig. 13 (a) and formula (19) we e = cosh 0 + sinh 0=1- + ZQfZx have

= 1 - of-CL + — j/lxiC

jS —- loge of the modulus of this quantity. a is the angle in radians per filter section such that _ uPCL-2 ee — ~-2{a)2CM+i) Examination of these equations shows that for all Proceeding in this way the following formulae and values of co2CL up to 2 such a filter transmits without any other similar formulae may be developed :— attenuation. The retardation a per section in the same Z\ = Z\M + 2Z1MZ2M [see formula (17)] range increases from 0 to IT. Above ufiCL = 2 there * G. W. PIERCE : " Electric Oscillations," chap. xvi. (McGraw is attenuation which increases rapidly with frequency, Hill Book Co.) AND TELEPHONE CIRCUITS: DISCUSSION. 691 that is ZQ = Z2 +juiM -f 2(Z2 + jtoM){Z1 - jcoM) The angle of retardation a associated with any fre- quency in the range of frequency obtained when B = 0 may be found from (27). Equation (28) has + 1 and [see formula (23)] — 1 as limits, and the cut-off frequencies may be deter- mined from these limits. sinh 6 = „ ° = - - jcoM (24) Examples. [sec formula (18)] Case of Fig. 13 (b). Cut-off limits If jcoM = 0 formulas (22) to (24) reduce to formula 2 (17) to (19). (a) f + 1 = 1 - a)*CL = 1, Special case. then = 0 and ct> = 0 If Zx — Zt or Z1M = Z>2M ZJZ2 = 1 and, by (19), cosh 6 is constant for all frequencies, since Z^\Z^ = 1. From the detailed study of the formulas given, the (b) 1 - ufiCL = - 1, or u£CL = 2 and a> = electrical constants necessary for various forms of niters may be deduced. The effect of apparatus joined to the ends of such i.e. there is no attenuation for frequencies between the niters follows the usual known transmission laws cut-off limits having the values a> in (a) and (b). applicable to equivalent T circuits. Case of Fig. 13 (c). Cut-off limits To DETERMINE THE RANGE OF FREQUENCY IN WHICH Z2 1 2 = 1, A FILTER HAS NO ATTENUATION, AND THE CORRE- (a) - + 1 = 1 - oflCL SPONDING ANGLE OF RETARDATION PER FILTER SECTION. or — Since 6 = jo. + j3 by (13) we may write e0"a + P) = cosh (ja + £) + sinh (/a + /?) = (cosh j8 cos a + j sinh j8 sin a) + (sinh jS cos a + j cosh j3 sin a) . (25) when cosh/?=l, sinh /? = 0, and j8 = 0, (25) then Case of Fig. 13 (/). becomes _afiOL _2 _ ei°- = cos a + j sin a . . . . (26) -2(co2CM + 1) For the case of no attenuation we have therefore 2 by (20) or a> C7(L + 2M) = 0 = cos a + j sin a (27) This is indeterminate ; there is no lower frequency (i*)® • limit, and a> = 0. when 0, " is real and we have a>*CL - 2 ^2 -f 1 = cos a (28) V[C{L _ In the case of (a) in each instance a = 0. (29) Z, = j sin a In the case of (b) in each instance a = TT.

DISCUSSION AT THE INSTITUTION, 16 MARCH, 1922. Sir Andrew Ogilvie: This paper is of a very tech- of trunk or long-distance working in those days, as nical character, and perhaps I ought to apologize for well as the special limitations of the Van Rysselberghe offering a few remarks on the historical and commercial system, gave it only a very limited range of usefulness aspects of the systems, while leaving the technical for long-distance work. I suppose that its chief use aspects to be dealt with by those who are more fitted has been for railway purposes, but in a certain number to speak on them. I remember the enthusiasm in of cases at the beginning of the present century, when telegraph circles which followed the Van Rysselberghe the Post Office first undertook the extension of the invention. At that time the telegraph industry was telephone system in rural districts, the system was used beginning to apprehend the competition which would for converting small village telegraph offices to telephone arise from telephones, and this seemed a method of working, thereby enabling them to be linked up to the salvation by means of which those who owned telegraph telephone system of the country. During the war the systems might get a cheap share in the new business. Fullerphone development of the method was very use- Those hopes were not fulfilled. Manifestly the Van ful for military purposes, not as a specially good means Rysselberghe system is not one which lends itself to of communication, but because it was secret and could ordinary exchange working, and the general conditions not be overheard on enemy listening sets. The simplex