Empire Telegraph Communications

EMPIRE TELEGRAPH COMMUNICATIONSBy K. L. WOOD, Member.*

(Paper first received lQth June, and in revised form 10th November, 1938; read before THE INSTITUTION \5th December, 1938, andbefore the NORTH MIDLAND CENTRE 28th February, 1939.)

SUMMARYThe paper deals with the engineering aspect of Empire

telegraph communications. In order that the size and scopeof these communications may be realized, the author beginsby setting forth the components of the telegraph operatingcompany, showing the systems contributed by each section.

The engineering organization within the operating companyis then dealt with, the training and necessary qualificationsof telegraph engineers in the present service being outlined.

Returning to the general aspect of Empire communications,the main traffic routes are discussed, with details of a numberof main-line circuits. Some of the engineering requirementsof the flow of traffic are touched on.

Having dealt briefly with the system, the paper proceedsto enumerate the methods by which the collection and thedelivery of messages are effected. The methods of signallingand the suitability of the different codes under varyingcircumstances are explained.

The author then presents a short history of progress incable telegraphy. Modern methods of signal shaping,balancing, and synchronous-regenerator working on cablecircuits are reviewed. Where well-known methods are inuse the descriptions are brief, full technical explanationsbeing given only where information has not previously beenpublished except in patent specifications. The items describedin detail include the capacitance magnifier, the fork relay, theautomatic bias corrector, and the automatic scrutinizer.

There follow brief notes on recent progress and cableproblems which are still being investigated.

A brief outline of the methods of operating the wirelesstelegraph circuits is given, with details of the applicationof regenerator working to these circuits. A description ofa method of double-frequency keying of wireless circuitsfollows. Finally, reference is made to the effects of varyingcosmic conditions on wireless signals.

(1) INTRODUCTION: COMPOSITION AND SER-VICES OF OPERATING COMPANY

On the 29th September, 1929, the telegraph communi-cations of the Empire, as detailed below, were, by Act ofParliament, placed in the hands of a single operatingcompany. At the same time an Imperial Communi-cations Advisory Committee was set up, the membersof which were nominated by the British and DominionGovernments. This operating company took over thesystems belonging to the following companies anddepartments:—

The Eastern Telegraph Co., Ltd., and four associatedcompanies.

The Eastern Extension, Australasia and ChinaTelegraph Co., Ltd.

The Western Telegraph Co., Ltd., and four associatedcompanies.

• Cable and Wireless, Ltd.

The telegraph operating section of Marconf's WirelessTelegraph Co., Ltd., with several subsidiary services.

The Imperial Cables, and the English end qf the beamwireless services formerly operated by t:he BritishPost Office.

The Pacific Cable Board's cable system and the WestIndian wireless services associated therewith.

In addition to the above, 8 other telegraph companiescame under the control of the operating company. Thecomponent parts of the operating company are furtherdetailed in Appendix 1.

In Appendix 2 will be found maps showing: (a) Themain cable connections of the Eastern and associatedcompanies before the merger (Fig. 37). (b) Other cableand landline systems incorporated in the merger (Fig. 38).(c) Wireless services as established by the Governmentand by the Marconi Co. before the merger (Fig. 39).(d) Wireless services opened since the merger (Fig. 40).The first two maps show how effectually the State-owned cables filled in the gaps in the network establishedby the companies, thus providing a comprehensiveEmpire cable system. The fourth map reveals that con-siderable progress has been made in extending bothEmpire and Empire-foreign services.

In addition to the main systems indicated on the maps,there are various small wireless circuits and ship andaircraft services. The cable systems in many parts ofthe world have been reorganized. The terminals ofcertain wireless circuits in South Africa, India, Canada,Egypt, and Australia are operated by local companies.Except in the cases of Australia and Canada, the cablesystems in these countries are also operated by the localcompanies, by arrangement with the operating company.Telegraph connections also exist between Great Britainand foreign countries on certain wireless circuits operatedby the Post Office. The Post Office also maintains abroadcast telegraph service on long and short wave-lengths. Radio telephony in the United Kingdom iscontrolled and operated by the G.P.O., but the overseasterminals in the Dominions are controlled either by theabove-mentioned local companies or by the operatingcompany. The latter also operates some colonial-foreignwireless telephone circuits, such as Bermuda-New Yorkand Kingston-Miami-New York, these services linkingup with the normal United States network and withother countries.

The operating company also handles abroad the shoreend of various ship services, some of which are fittedwith direction-finding apparatus. At Nairobi, Bermuda,and Bahrein, Adcock direction-finding stations for airservices are in operation.

638]

WOOD: EMPIRE TELEGRAPH COMMUNICATIONS 639

Other activities included are the London end. ofradio-phototelegraph services with New York, BuenosAires, Melbourne, and Tokio; also land telephony inCyprus and Peru, and broadcasting in Kenya.

The present telegraph system comprises 155 090nautical miles of submarine cable, 125 wireless circuits,and 161 overseas branches.

With regard to cable circuits, only those of mostrecent date are of the high-speed loaded type in which thecopper conductor is surrounded by high-permeabilitynickel-iron alloy. The duplicate cables across thePacific, and the duplicate Cocos-Perth cable, are of thistype. Since two or more cables are already laid on all

became possible to economize efficiently at variousstations where both wireless and cable circuits wereoperated.

The merger brought together the two head-officeengineering departments, and, for a period, duplication ofengineers-in-chief and of their deputies was unavoidable.As knowledge of the capabilities and problems of the twosystems accumulated, economies were effected by com-bining the duties of various officers. The eventualorganization of the head-office engineering departmentis shown in the accompanying chart (Fig. 1). The headsof the department consist of an engineer-in-chief, adeputy, and two assistants. The general administration

CONTACT VIITHMfl«/flC£f»S OF

OTHER DEPARTMENTS

GEHERBl ADMINISTRATION

IN THE FIELD

ENGINEER-IN-CHIEF

Deputy

AssistantE-m-C

mmm

AssistantE-in-C

\flccounts \

jfontneta tSloreo |

N{ Traffic TrafficRouting '

DEPARTMENT

1Plant

Maintenance

L 1ExperimentalDevelopment

Labs.

IConditions

Tata

L " 1'.irrespondena

Records

\ iTrafficLiaison

Engineer

LaboratoryWorKShopa

OutsideProgress &rWfents

1CableshipMovementst Cable'Repairs

Vectrtctomafloat

25

1Coble

branches( • m U K150 Abroad

t |tnoiiec**S0U*liah

1WirelessBranches0 in UK,32 Abroad

1OverseasCompaniesIfldvtioriWorK)'

EngineersISO

Fig. 1.—Engineering organization.

important routes, it would not be economical to replacethem by one high-speed cable in which an interruptionwould dislocate the service.

(2) ENGINEERING ORGANIZATION WITHINTHE OPERATING COMPANY

When the cable and wireless companies were united,the problem at once arose of combining two distinctengineering staffs, one trained mainly in cable working,and the other mainly in wireless working, the obviousaim being to avoid overlapping and to make the best useof the experience and skill available in both sections underthe new administrative conditions. The problem wasmet by retaining specialists on each class of work and byselecting a certain number of the younger men, who hadshown themselves to be sound both practically andtheoretically, and training them in both wireless andcable work. After a reasonable period of time it thus

section contains staff from both services, the majority ofwhom have received training and experience in bothcable and wireless work, though some degree of speciali-zation is necessary in many cases. Specialization, alliedto a certain amount of general knowledge, is requiredin most of the other sections of the department.

A number of the posts are temporary appointments;selected foreign-service engineers are appointed to thegeneral administration and development sections of thehead-office department for periods up to 3 years, inorder that the maximum number shall have experienceof head-office methods and that the qualifications of anumber of engineers shall come under review, enablingfurther selection to be made when vacancies occur in thepermanent positions in the department.

From Fig. 1 it will be seen that a close liaison is main-tained between the traffic and engineering departments.Cable interruptions and bad wireless conditions arealways setting up combinations to which the best-fitting

640 WOOD: EMPIRE TELEGRAPH COMMUNICATIONS

key has to be found, by prompt co-operation between thetwo departments, for carrying the traffic expeditiously.

The operations of the cable maintenance departmentare very important; it is that department's duty toadvise the engineer-in-chief of sections requiring renewal;the amount and type of cable likely to be required; andthe state of stocks on ships, on shore, and under manu-facture. Every repair means an alteration in historydiagram, splice list, and chart. These together form acomprehensive record of each particular cable. Thehistory diagram is a record of every renewal and repairto the cable since it was originally laid, the splice listcontains the principal electrical and mechanical data ofthe existing cable, and the chart shows its position on theocean bed.

The training of engineers for the service is of primaryimportance. The method employed up to the presentis to select healthy young men who have reached orapproached matriculation standard and have shownpromise in scientific subjects as well as a facility forforeign languages. Their first training is in telegraphoperating; during this course they also attend lectureson the fundamentals of electrical theory and receivesome mechanical training. Those who do not respondto the engineering training may continue as operators,with the possibility of promotion on the administrationside. Those who show engineering ability, especiallyon the practical side, will be given a course lasting 2years, during which they are instructed in wireless andcable theory and are thoroughly trained in the use ofall the apparatus of which they are likely to have charge.This training includes the attainment of sufficientskill to manufacture and repair any component part ofthe mechanical apparatus. Proficiency in cable faultand break localization is also attained.

From Fig. 1 it will be seen that contact is maintainedbetween the engineer-in-chief and the staff manager.The staff manager is responsible for the selection, generaltraining, and subsequent movements of staff, includingengineers. In view of the great distances involved, theimportance of interchangeability of staff in making themovements of personnel as economical as possible will berealized.

(3) ROUTESAlthough it is primarily the work of the traffic depart-

ment to regulate the routes and the flow of traffic, the•engineers are required to provide the necessary facilities,particularly as regards the instantaneous provision ofalternative routes to prevent delay when cable interrup-tions or unexpected wireless fades occur. Details of someof the more important traffic routes will therefore begiven, together with maps (see Appendix 2), on which forclarity the Empire and Empire-foreign routes are shownseparately. Fig. 41 (Appendix 2) shows the main inter-Empire cable and wireless routes.

It will be observed that almost all the Empire routesfrom the United Kingdom are served both by cable andby wireless. London stands out clearly as the focalpoint of the whole system. During certain periods of the•day much of the traffic between distant parts of theEmpire passes through London; for example, SouthAfrica-India traffic may travel via the Cape Town-

London beam and the London-Bombay beam or via thecable chains over the same route.

The following are typical alternative routes to Aus-tralia :—

(i) By beam, either long or short route to Sydney andMelbourne, or relaying by Canada on Montreal beam.

(ii) Via Cape beam or cable and the cable chainDurban, Rodriguez, Cocos, Perth, Adelaide, etc.

(iii) Via the cable chain Halifax, Montreal, Vancouver,Fanning, Suva, Auckland, Sydney (or Suva, Norfolk,Southport, Sydney).

(iv) Via the cable chain Gibraltar, Malta, Alexandria,Suez, Port Sudan, Aden, Seychelles, Colombo, Penang,Singapore, Batavia, Cocos, Perth, Adelaide.

The choice as between wireless and table is primarilydependent on the sender's instructions, but the particu-lar route is regulated in part by the outpayments thathave to be made on the various routes, and depends verymuch upon the time of traffic flow, which again is mainlydependent upon the business times in various parts ofthe world. During the night, for instance, on a routewhich is served both by wireless and by cable the cablecircuit may carry the bulk of the traffic, since the majorityof the cable chain stations may have to be kept fullymanned to serve other routes.

Early homeward Australian traffic normally arrivesvia the Adelaide-Durban-Cape route, and the trans-Canadian route, so freeing the Singapore route for home-ward Far-East traffic. The beam takes its share later.Thus the two methods of transmitting messages areutilized to the best advantage.

Fig. 42 (Appendix 2) shows the main Empire-foreignroutes. A large proportion of the Empire-foreign trafficflows over portions of the routes shown on the previousmap, and some of the routes are necessarily duplicated.London again stands out as the focal point.

The London-Rio de Janeiro and London-Buenos Airescable circuits are particularly interesting from an engi-neering point of view. They are carried on a single cablechain as far as Ascension Island, the signal elements inthis chain being collected from two automatic transmit-ters in London in the ratio of one signal element fromthe instrument transmitting to Buenos Aires to two fromthat transmitting to Rio. At Ascension the beats forBuenos Aires are picked out from the aggregate, auto-matically prolonged in length, and retransmitted; the gapsin the remaining elements are closed in and the signalssent on regularly to Rio. In the reverse direction thesignals from the two cables are combined into the singleAscension-London chain and sorted out into their respec-tive receivers in London, but the ratio in this case is 2 to3, owing to the superior reception conditions at Ascension.

(4) THE COLLECTION, TRANSMISSION, ANDDELIVERY OF MESSAGES

As London is the focal point of Empire communica-tions, collection and delivery problems require moreattention in London than at any other branch.

Traffic for transmission via London is collected by thefollowing channels: (i) Branch offices in the main pro-vincial cities or towns using private teleprinter lines orprivate telegraph line (see Fig. 43). (ii) Offices of theG.P.O. inland telegraph system and of the Irish tele-


graphs, (iii) London branch offices, by teleprinter, orby pneumatic tubes from those city offices which aresituated close to the main offices, (iv) Direct fromcustomers by private teleprinter line, telex service,private telephone line, public telephone, counters, ormessengers.

Urgent and competitive traffic is most expeditiouslyhandled, both in acceptance and in delivery, by meansof private telephone lines.

Fig. 2 shows the increases in the numbers of messagesaccepted and delivered over the telephone at the Londonmain branch during the years 1933 to 1937. The in-creases are shown as percentages of the 1933 figures.The average number of messages delivered over the tele-phone is about 9 per cent more than the average numberreceived by this method. Messages received in thetelephone room are conveyed by chutes, tubes, and beltsdirect to their respective forwarding circuits, where theyare numbered, converted into the form of perforated tape,and transmitted.

All messages from London to points carrying anappreciable amount of traffic are given a prefix and num-bered in series. For instance, a London-Alexandria

= 60•oo

£40oin<uui

s120•00

§ .

AcceptancesDelivery

1933 1934 1935 1936 1937

Fig. 2.—Increase in the number of messages accepted anddelivered over the telephone.

message will bear the prefix AX followed by a number.The consecutive arrival of these numbers is checked byAlexandra, and any query is addressed direct to London.Thus, should a message be diverted from the normalLondon-Alexandria circuit for any reason, an inter-mediate station, such as Gibraltar, receiving this messagein the form of perforated tape merely passes over thetape to a Gibraltar-Alexandria circuit without checkingthe number, the message arriving at Alexandria almost asquickly as on a direct circuit.

Transit traffic in London is dealt with by automatic per-forators. The perforators of various circuits are grouped,and the message tapes are distributed by hand into f-in.diameter tubes leading to the appropriate transmittingposition. These tubes take the perforated tape withoutfolding. They were first used by Mr. Christensen, of theGreat Northern Telegraph Co., for the transfer of tapebetween the Extension and the Northern companies'offices in Shanghai.

The automatic transmitters are connected by landlineor by underground cable to Porthcurno, Bodmin, Dor-chester, Ongar, and Grimsby, the choice of route being

decided by the traffic supervisor and the appropriatewireless transmitter by a control-room engineer in theLondon building.

The delivery of messages is carried out by substantiallythe same means as the collection of messages.

(5) SIGNALLING CODES USED INTRANSMISSION

The three-positional " cable code " using the standardmorse alphabet is in use on nearly all the cables of theservice. A dot is represented by a negative impulse toline, a dash by a positive impulse to line, and a space byno current, all signals being of equal length. The cablecode has two great advantages over other codes in generaluse: (i) The majority of instrumental errors are readilydetectable, (ii) Intelligence can be transmitted over along submarine cable more rapidly and under more stableconditions by cable code than by any other code. Thedevices by which signalling speed with the Morse Codeor the 5-unit code can be brought to approach the maxi-mum cable-code speed of a cable produce instability andlimitations with regard to duplex balance and change ofspeed.

With regard to the first point, errors in transmissioncan practically be confined to (a) one or more elementslost, (b) one or more elements inserted. In cable codethe loss of a beat at the beginning or end of a letter willresult in a double space, which is easily detected. Theloss of a beat other than the first or last in a letter willresult in the breaking-up of one letter into two, producingin a 6-letter code word, a 6-letter word, detectable as amiscount. The insertion of a beat in the space betweenletters will run two together, either making an unprint-able combination or producing a 4-letter code word.

In 5-unit code the insertion or dropping of a beatgenerally results in the substitution of one letter foranother, a fault usually quite undetectable in a codeword.

With regard to the second point, with morse signallingon a long submarine cable it is necessary, in order toconvey the same amount of intelligence in a given time,to increase the number of bauds transmitted by 152 percent, i.e. by 3-6 to 9-0 bauds per letter. This can beattained only by attenuating the dots until they are notappreciated by the primary receiving instrument, andby re-inserting them locally. A three-positional relayis required, since the morse signals in effect arrive as athree-positional code, i.e. space-zero-mark.

The zero element, being the result of a balance betweenmark and space at the far end, is susceptible to bias, andon a reduction of speed the indication of transmitteddots is no longer zero but an attenuated reversal. Thus,with receiving conditions at their optimum value for thebest speed on a long cable, a reduction of speed results inloss of stability until a point is reached wnere the attenu-ated reversal is too large to allow the relay to remain inan inoperative position but is yet too attenuated tooperate the relay correctly. Fig. 3(6) is a graph of themovements of the relay pointer under this critical con-dition. The dotted horizontal lines show the points atwhich the relay will operate its local circuit; it will benoted that any displacement of zero or any balance


disturbance will cause interference with the robustcontrol of the local circuits of the relay.

Inability to reduce speed without change of receivingconditions is a grave disadvantage, since it prevents theeasy interpolation of a fast cable in a comparatively slow-working chain. Again, a reduction of speed which maybe desirable for balance or other purposes producesinstability instead of improving a circuit. Cable codedoes not suffer from this defect, since the zero unit remainszero at all speeds.

The use of morse double-current signalling at thesending end introduces difficulties in the duplex balanceowing to the quick reversal, and greater perfection ofbalance is necessary owing to the greater number offrequencies introduced.

Five-unit code suffers from the same disabilities.Although the increase of bauds necessary to equal theperformance of cable code is not so great as with morse,the fact that the local vibrating relay must be held upby a mark or space of 2 bauds' length instead of 3, as

«JU Lnji_nruLetter 'A' Letter'D"

voice-frequency channel or a limited wireless circuitby over 30 % as compared with the Morse Code.

A further elaboration of this code (see Section 7) hasbeen made for use on those beam wireless circuits whichare worked on a channelled regenerator system.

(6) PROGRESS AND PRESENT METHODS INCABLE TELEGRAPHY

GeneralNo important long cables have been laid in recent

years, as the service has been able to cope easily withexisting traffic.

It is sometimes asserted that improvements in cablesystems were not made until wireless competition wasexperienced. Substantial progress was, however, made

Selection points

I I I J I I , , I i

LetterV

( b ) - 1 1 1 1Dot selection points

I I < 1 \ \

Fig. 3.—Double-current morse on cables.(a) Transmitted morse signals.(b) Received signals (dotted lines show " no-man's-land'' of receiving relay)(c) Output of receiving relay which controls vibratory system.

in morse, prevents full advantage being taken of thesmaller number of bauds required per letter.

A two-positional code having some of the advantagesof the cable code, and known as the " double-currentcable code," has been brought into use on landlinecircuits and wireless circuits where regenerator systemsare worked. In its simplest form, as used on landlinecircuits and on some wireless circuits, the dot con-sists of a marking impulse and the dash of a spacingimpulse, but the zero interval for a space signal isreplaced by a 50 % mark signal, followed by a 50 %space signal. This code* has been described briefly ina previous paper,•}• but a further explanation is addedhere to clarify later developments.

Fig. 4 compares the double-current cable code with thenormal cable code. It will be observed that when thedouble-current cable code is in use it is necessary toseparate the dot and dash selection points of the regener-ator apparatus, a dot being selected three-quarters ofthe way along the unit signal period and a dash one-quarter of the way along. The double-current cablecode has the advantage that a dash occupies only 2bauds, thus increasing the traffic-carrying capacity of a

• H. V. HIGCITT: British Patents Nos. 267180 and 294715.t H. KINGSBURY and R. A GOODMAN: "Methods and Equipment in Cable

Telegraphy," Journal I.E.E., 1932, vol. 70, p. 477.

mt t t t ! t t t t I t i

Dash selection points

Fig. 4(a) Normal cable code.(6) Double-current cable code: same signals as in (a).

before as well as after the advent of wireless. As early as1900, relay reception with automatic re-perforation wasaccomplished, while in 1912 mis-selection due to distor-tion was minimized by mechanical delay of the selectingpoint to the centre of each beat—a principle which isincorporated in present-day " regenerator" systems.By 1914 a number of cables were equipped with regener-ator apparatus and a beginning had been made in linkingsections together, thus laying the foundations of thepresent world-wide chain systems.

The abnormal demand upon communications person-nel during the War precluded any experimental workfrom 1914 to 1918. At the same time the expansion oftraffic and the availability of morse re-perforators ledto a change-over from cable code to morse. Defects inthis system were experienced and, when research againbecame possible, work was continued with cable code on asystem employing continuously synchronous apparatusinstead of the start-stop synchronism of the 1912 appa-ratus. By 1923 the " regenerator " had been perfected,and during the next 4 years the appropriate instrumentswere manufactured and supplied to 120 stations, effect-ing an immediate improvement in reliability and aspeed-up of the service.

The principles of synchronous regenerator working are


well known and have been fully described in the Journal.*Special problems, however, arose in the Empire service.The number of cables which had to be linked up to formsingle regenerator chains was so much larger than on anyother cable system that it was necessary to develop moreprecise methods of speed control and synchronization.

R5 and R6 with values of some thousands of ohms. Thecross circuit is composed of the two windings, Wl andW2, of the receiving instrument (described later), withthe shunted capacitor C3 of the receiving-circuit networkinserted between the two windings in order to achievesymmetry in relation to earth. Across this combination

Short Longsea earth sea earth

Fig. 5.—Complete duplex network on submarine cable.

Since that date various devices have been incorporatedin regenerator equipment which have had the effect ofincreasing working speeds and further improving thereliability of the system.

Figs. 5 and 6 show respectively the complete duplexnetwork and the component parts of the regeneratorsystem as at present operated. Means of measuringsignal distortion are incorporated in the regeneratorapparatus. While the regenerator itself is capable oftransmitting practically perfect signals from badly dis-torted received signals, it was soon realized that theundue distortion of certain signals limited the speed.This drawback led to a close investigation into the net-works used, with the object of finding the best arrange-

is the inductor L2, with series resistor R7, in parallelwith a second system of bridge arms, R8 and R9, havingan adjustable point connected to earth. These armsform what is known as an " extra apex."

Between the transmitting key or transmitter K andthe normal ap. v: of the bridge is an inductor Ll. Thefunction of this ->ductor is to raise the effective voltageat the head of the c ble during the signalling of the higherfrequencies and also to change the wave-form of thepotential applied to tL - cable and artificial line from asharp rise to one appro ;mating to a sine wave, thussubstantially reducing quic. and violent currents causedin the cross circuit by small e. -ors in the duplex balance.Fig. 7 (see Plate 1, facing page 644) shows the wave-

Sendingterminal3

ci.c lock -controlled

reed

Tltransmitter

Duplex andSendingShapingnetwork

Cable Duplex andreceivinf-

Mcapacitymagnifier

Intermediatestation

Keyboardperforator __Signal impulses

Speed control impulses""""Bias correcting impulses

Fig. 6.—Component parts of regenerator system.

ment and the optimum values of the sending and receiving" shaping " conditions.

From Fig. 5 it will be seen that the system is based onthe original Muirhead capacitance bridge. The ratioarms Cl and C2 of the bridge are shunted by resistors

• H. H. HARRISON: "Developments of Machine Telegraph Systems andMethods of Operation," Journal I.E.E., 1980, vol. 68, p. 1869.

forms of the voltage at the head of the Porthcurno-Gibraltar No. 3 cable, for signalling (i) with and (ii)without the inductor Ll. The peak at the beginningof each signal in Fig. 7(6) which introduces balancingdifficulties, is absent in Fig. 7(a), and the potentialover the main portion of the signal is increased.Formulae are available for calculating the best value of


inductance at the apex and the resultant increase inpotential at the head of the cable, but so many factors,such as the lack of homogeneity of the cable, have to betaken into account that these formulae give startingvalues only; the best values, as found by practical trial,usually differ widely from the calculated values.

The effect of Ll on the received signals on the samecable is shown in Fig. 8 (Plate 1). The increase in thesize of the signals of high frequency is very marked,amounting in this case to over 20 % of the signal.

Unshunted capacitance bridge arms practically preventthe passage of earth currents into the cross circuit. Onthe other hand, when a succession of beats of the samepolarity is sent by the distant end, the received currentfalls away, and the receiving relay is not maintained in amarking position. S. G. Brown devised a method ofcorrection in which a local current flowing in an auxiliarywinding of the receiving relay increased at the same rateas the fall-away; this arrangement necessitated a delicatebalance of forces, and proved troublesome in practice.The shunting of the capacitors by resistors of high valueprovides the necessary hold-up current from the distantend, but it permits the entry of earth currents. Effectivemeans have, however, been devised in the regeneratorsystem for dealing with the zero-wander resulting fromthese currents.

The addition of the shunts had the further effect ofgiving rise to a slow wave during signalling, owing to thecharging-up of the terminal and artificial-line capacitanceat the receiving station by the received current; thiswave was reduced to a negligible amount by the additionof the " extra apex " resistance arms R8 and R9 at bothends of the cable, as shown in Fig. 5.

Fig. 9 (Plate 1) illustrates the effect on the receivedsignals in the Porthcurno-Gibraltar No. 3 cable of re-moving the extra apex at each end. The normal signals areas in Figs.9(a) and 9(g); Fig. 9(6) shows the signals receivedafter the extra apices had been removed, and Fig. 9(c) thesignals after compensation had been made elsewhere in thenetworks to regain the best shape of signal. The slowwave referred to shows up to the greatest extent in theselast signals when a series of beats of the same sign arebeing sent. Sections of the signals from Fig. 9(c) havebeen enlarged to show this point more clearly, and astraight line has been added on the tapes for the samereason. Figs. 9(d) and 9(e), Plate 2, show the beginningand end of the series of signals of the same sign. Thequick dip at the beginning of the series, which shows upagain in Fig. 9(/), can be eliminated without the introduc-tion of extra apices, but its elimination results in evengreater wandering of the remainder of the steady signals,and a consequent general wandering of the signal zero.If extra apices are not used it is therefore necessary toeffect a compromise.

The shunted capacitor in series and the inductor inparallel with the relay windings are adjusted so that theincrease in impedance in series with the ceil and thedecrease in parallel are such that, although the currentfrom the cable due to a succession of beats of the samepolarity is increasing in value as the cable becomescharged, the current through the windings remainsconstant.

Referring to Fig. 6, the clock-controlled reed Cl

controls the speed of the automatic transmitter Tl.This transmitter, for the purpose of easy interchange oftraffic between wireless and cable circuits, is madecapable of sending cable-code signals from either Wheat-stone or cable-code slip at will. The transmitter sendsinto the cable through the duplex terminal apparatus,as shown in Fig. 5, and after passing through similarequipment at the distant end the shaped received signalsare amplified by the magnifier M and are passed on to thecontact-making secondary relay Rl . The contacts ofthis relay operate more robust relays R2. These relaysassist in the operation of the bias corrector B, attachedto the secondary relay Rl, and also operate the selectormagnets either of a re-transmitter T2, or of a perforatorP, or of both, according to the position of a switchingdevice D. The re-transmitter T2, kept in synchronismby the synchronizer S, re-forms the distorted signalsoperating its selector magnets, and, by means of cam-operated contacts, sends on signals with practicallyperfect formation into the next cable.

The perforator P, also running synchronously, startsperforating as soon as its selector magnets are broughtunder the control of the main relays by the insertion, atthe transmitting end, of a special combination of signalswhich operates the switching device.

The relays R2 also operate a relay in the synchronizerS; this relay, when operated, discharges a capacitor inone or other of two paths according to the phase positionof a contact device driven by a motor under the controlof the clock-controlled reed C2. The relative speeds ofthe two clock-controlled reeds Cl and C2 are thuscompared, an excess of impulses in one path indicating adifference in speed. Means are provided in the synchro-nizer for utilizing these impulses to make appropriatealterations in the speed of the motors of T2 and P sothat their operative parts act substantially in the centreof each received signal.

While a message is being automatically received on theperforator at an intermediate station of a chain of re-generative repeaters, or during pauses in the terminaltransmission, it is possible for the intermediate station toforward a message without interfering with the syn-chronism of the chain. A switch that can be turned byhand connects a small automatic transmitter T3, incor-porated in the re-transmitter, to the selector magnets ofthe re-transmitter, thus initiating signals in the nextcable. Should the terminal resume transmission beforethe local message is finished, the switch is automaticallyreturned to the through working position and the feed ofthe perforated tape of the local message is stopped.

At the terminal of a chain of cables the re-transmitteris omitted and the whole of the traffic is received on thereceiving perforator P (Fig. 6); the perforated tape fromthis is fed into a Creed printer, so producing romancharacters. This method is now giving way to a newprinter which translates the cable-code signals directlyinto roman characters without interposing the perforatedslip.

An automatic scrutinizing device is fitted to bothperforator and re-transmitter. This device gives warningwhen the signals have deteriorated to such an extentthat danger of failure is approached, the normal causesof deterioration being changes in the duplex balance,

WOOD: EMPIRE TELEGRAPH COMMUNICATIONS

Plate 1

(a)

Fig. 7.—Wave-form of transmitted signals, (a) with apex inductor LI, (b) without apex inductor LI.

(a) v v •••/

Fig. 8.—Wave-form of received signals, (a) with apex inductor LI, (b) without apex inductor LI.

S >••'••. : ' ^ V

••••J - J •-;

Fig. 9.—Received signals with and without extra apices in sending and receiving networks.

I.E.E. JOURNAL, VOL. 84.

(a) Normal signals.(6) Extra apices removed at both ends.(c) Compensation made elsewhere in networks to regain best shape of signals.

(Facing page 644.)


Plate 2

Fig. 9(d).—Enlargement of beginning of steady signal in Fig. 9(c).

Fig. 9(e).—Enlargement of end of steady signal in Fig. 9(c).

% **

f 0>«i»nii»,

Fig. 9(/).—Enlargement of signals from Fig. 9(c). Fig. 9(#).—Enlargement of signals from Fig. 9(a).

Fig. 11.—Filtering effect of magnifier coil.(a) Input of magnifier amplified to normal signal size by valve amplifier.(b) Output of magnifier from same input.


Plate 3

LrVV\/v'VVJV~v

* \Fig. 17.—Secondary-relay signals, showing rapid removal of heavy bias

\^%r\"hJ^

Fig. 19.—Action of automatic scrutinizer(a) Received signals(6) Retransmitted signals.

Fig. 20.—Comparisons ot the frequencies of two reeds when the voltage applied to one is varied.

(a) With normal adjustments(6) With new method of adjustment.

twuwwLnawu\^^

ky

Fig. 22.—Records of double-current cable code signals on Bombay-London beam.

Plate 4


r\y~\.Di

1 2 3 4 5

Fig. 35.—Multiple echoes.

This is a continuous record, the tape being cut for ease of reproduction.

Fig. 36(a).—Forward and backward echoes.

1 2 3 4 5

1 2

1 2

3

3I

i

i

4

4

4

1

5

5

As transmitted

Backward echo

Forward echo

As received

I I I I I I

Fig. 36(b).—Forward and backward echoes.


excessive bias, speed sway, and poor contact. Thewarning usually takes the form of the lighting of a lampand the augmentation of the signals on a local record atthe point where the device functioned. In particularcases, however, a single-stroke bell may be rung, and oncircuits where no permanent record is running the devicemay be used to start up a local record. The attention ofthe watch engineer is thus drawn to some deficiency inhis apparatus before errors actually occur, and in this waystoppages are avoided which, in some cases, would neces-sitate the attention of as many as 16 other engineersdistributed at stations on a chain of cables. In addition,therefore, to the substantial reduction of stoppages andthe prevention of many errors, this device has enabledstaff economies to be effected.

Although the modern apparatus may appear to be

4 0 0 -

Details of ApparatusThe clock-controlled reed* and automatic transmitter!

are fully described elsewhere.

The capacitance magnifier.As the speed of transmission of signals through a cable

is increased, a point is reached where attenuation becomesso great that reliable operation of contact-making relaysis no longer possible. Some means of amplifying thesignals before passing them to the relay is thereforenecessary if an increase in speed is to be achieved.

It might be thought that valve amplification wouldprovide the best solution of this problem, but the meritsof the moving coil as a simple and efficient filter of theunwanted higher frequencies present in the duplexbalance outweigh all other considerations.

1936 1937 1938

200

1936 1937 1938

Fig. 10.—Graphs of stoppages at main-line station having 20 duplex circuits.

delicate and complicated, in' actual practice it is normalto run as many as 17 sets—2 terminal and 15 intermediatestations all working in synchronism—on a chain of cablesreliably and without undue attention.

Fig. 10 shows a typical stoppage graph at a branch with20 duplex circuits equipped as described, the graph cover-ing the period from January, 1936, to September, 1938,inclusive. The abscissae represent in minutes per monththe total stoppages on all circuits due to this particularbranch. Stoppages caused by cable interruptions arenot included. At this branch, submarine cables areconnected to landlines, so that the difference between theupper and lower graphs represents stoppages due todifficulties in maintaining the duplex balance on thecable sections. The average stoppages work out at0 • 69 minute per circuit per day of 24 hours in onedirection and 0-23 in the other.

VOL. 84.

Fig. 11 (see Plate 2) shows a comparison between a badduplex balance outness as applied to the input of amoving-coil magnifier and the outness as applied to thesecondary relay after the filtering effect of the magnifiercoil. The large reversals represent the size of normalsignals. In each case the records were obtained througha valve amplifier, the gain being adjusted to give approxi-mately equal sizes of signal.

The capacitance magnifierj (Fig. 12), however, com-bines the use of valves with the moving-coil principle.A light aluminium pallet P, carried on an aluminiumarm AA, is attached to, or in some cases harnessed to,the main coil of the magnifier. The vanes of the palletare adjacent to two fixed sets of plates E, termed thefixed electrodes, a movement of the magnifier coil MC

• H. H. HARRISON: Journal I.E.E., 1930, vol. 68, p. 1377.t British Patent No. 2627S—1913. \ British Patent No. 277740.

41


increasing the capacitance between the pallet and oneset and decreasing the capacitance between the palletand the other set. The two sets of fixed electrodes areconnected to the grids of two rectifying valves VI andV2 by short lengths of wire. Oscillations from a self-oscillatory valve circuit V3 are tapped off on a capacitorpotentiometer and applied to the pallet. The rectifyingvalves, together with two resistors of approximately1 000 ohms each, form a bridge circuit across which isconnected the coil of the secondary instrument.

A movement of the pallet increases the mean potentialapplied to one grid and decreases that applied to theother grid, causing variations of the plate currents in thevalves and a consequent flow of current in the coils ofthe secondary instrument FC.

The rectified current in the cross circuit is substantiallyproportional to the movement of the pallet, up to a limitof 1 mA of rectified current with the degree of magnifi-cation that is normally used. This means that a con-siderable wandering of the magnifier pallet due to earthcurrents may take place without appreciably distortingthe output signals—apart, of course, from zero-shift.

The capacitance magnifier will operate on a current of

Fig. 12.—Capacitance magnifier.

2 microamperes, corresponding to an input voltage of1 millivolt. Normal received currents on a cable afterpassing through the shaping networks are greater thanthis, being of the order of 10 microamperes for a longcable and more for shorter ones, giving, with a magnifi-cation factor of 7 or 8 to 1, ample current to work thesecondary relay. In special cases an amplification of90 to 1 has been obtained from this instrument.

Secondary relay.This relay, known as the " fork relay," is of the

moving-coil type but it has the unique feature that thecontact elements make contact when at rest. The move-ment of the relay coil in one direction beyond a predeter-mined amount breaks contact on one side and maintainscontact on the other, thus upsetting the balance of abridge circuit and operating-one local relay. Movementin the other direction causes an opposite upset, so oper-ating the other local relay.

The tongue of the fork relay, attached to the wire-suspended coil, consists of two short gold-wire antennaearranged like a 2-pronged horizontal fork, the antennaebeing adjustably compressed between two silver pillarswhich are kept in a state of vertical vibration. Theamount of compression determines the angle through

which the coil has to move before one wire breaks con-tact, and is therefore comparable to the transit of thetongue of an ordinary relay. Reliable contact-breakingcan be carried out with a current of 60 microamperes

RA

Fig. 13A

through the 500-turn 500-ohm coil. The movement ofthe coil can be read for balance and signal-shape purposesby light reflected from a mirror harnessed to the coil.

Bias corrector.*As already mentioned, the shunts on the capacitance

bridge permit earth currents to reach the receiving net-work, and the necessary bias correction is most con-veniently carried out on the secondary relay.

A plan view of the bias-adjustment mechanismattached to the fork relay is shown in Fig. 13A, and anelevation of the magnetic-clutch mechanism in Fig. 13B.The silver contact pillars CP are carried on a platformattached to a shaft which is V-pivoted at a point Pbelow and on the same axis as the moving coil; thisshaft can be turned by a bias screw BS through a leversystem L, in which there is no backlash, so as to bringthe pillars into such a position that equal movements ofthe coil can be made on each side before a contact isbroken. The bias screw can be turned by means of asmall magnetic clutch MC (Fig. 13A) which operates in onedirection or the other according to which magnet isimpulsed.

0RA

Fig. 13B.—Bias corrector attached to secondary relay.

The schematic connections of the bias-corrector systemare shown in Fig. 14. The magnets RA and RB, whichoperate the clutch, are differentially wound, the windingsbeing connected to the dot-and-dash relays operated bythe secondary relay, and to a pair of relays RE and RI

• British Patent No. 240906.


which are operated by the advance and retard of thesynchronizer.

Fig. 15(a) is a graphical representation of the move-ments of the relay antennae, the movement being plottedvertically against regular time-intervals shown hori-zontally, corresponding to the times of arrival of perfectsignals. A dash bias resulting from a steady earth

Madefor dot

Increasedotbias

for dash

Increasedashbias

In parallelwith

synchronizemagnets

Fig. 14.—Diagram of connections of bias corrector.

current is assumed to exist. The horizontal dotted linesdenote the position at which the relay operates its localcircuit, the resultant contacts of the main dot and dashrelays being shown in Fig. 15(6). When the local circuitis operated a capacitor discharge energizes the syn-chronizing magnets of the synchronizer momentarilyto alter the speed of T2 and P (Fig. 6); at the same timethe discharge operates the magnets RE or RL (Fig. 14)of the bias corrector. Under dash bias conditions, asshown, it is arranged that the current from the capacitorC (Fig. 14), resulting from the operation of RL and thedot relay, flows in such a direction in the windings ofRA and RB that the two windings of RA assist eachother, and those of RB oppose each other. This causesthe magnetic clutch to operate the bias screw in such adirection as to remove the dash bias. An early dashsignal has a similar effect. If a dot bias exists, so causingearly dots and late dashes, RB will be operated, movingthe bias screw in the opposite direction. If, when thebias is correct, both dots and dashes are early or late

owing to speed-difference, the false step taken in onedirection by, say, an early dot, is in practice countered bya step in the other direction resulting from an early dash.

The amount of correction applied per impulse can bevaried in the lever system and by adjustment of thecapacitor C; it is adjusted so that in all but the worstmagnetic storms the movements of the relay coil can be

Fig. 15(a) Received signals with bias resulting from earth current.(b) Resultant main dot and dash relay contacts.

followed without producing sufficient distortion to causetrouble.

Fig. 16 shows typical variations of earth currentsduring a period of 3 days on the Harbour Grace-Porth-curno cable under normal conditions.- The earth-currentamplitude is compared with the signal amplitude. Fig. 17(Plate 3) shows signals obtained from the fork relay whenan additional potential of 18 volts was suddenly intro-duced into a long cable working with a 60-volt sendingbattery. With normal adjustments of the bias correctorthe bias was removed in 12 seconds. It will be realizedthat the normal earth current changes reasonablyslowly, and that an instantaneous change of as muchas 18 volts corresponds to severe magnetic-storm con-ditions. In practice, the fork relay with bias correctorhas functioned satisfactorily with earth currents causingbias outnesses up to 3 or 4 times the signal amplitude atthe receiving end.

Calling and switching device.For complete flexibility on a long regenerator chain

it is necessary that the terminal transmitting stationshould have complete control of the chain and should beable to call the attention of any individual intermediatestation at any time. With these objects, a selectivecalling and switching device is provided at intermediatestations. Two magnets, operated directly from the dotand dash relays, control two camshafts by means ofpawls and ratchet wheels.

70 voltsSignal amplitude (dot)

27fci\A1200

28fc-hJune

0600 1200

29fci\June2400 0600

I

30b-hJune

1200

70 voltsSignal amplitude (dash)

Fig. 16.—Earth-current variation on Harbour Grace-Porthcurno cable during a period of 72 hours:hourly observations at Porthcurno.


A dot signal feeds the dot camshaft forward one toothof the ratchet wheel, in which position it is held by adetent; at the same time the detent of the dash ratchetwheel is lifted, allowing the dash camshaft to return tozero. In a similar manner a dash signal feeds the dashcamshaft forward one tooth and resets the dot camshaftto zero. Thus normal traffic continually resets the cam-shafts, and does not allow a prolonged feed on one side.

In order that the device may be made selective, camson the dot and dash camshafts periodically operatecontacts which short-circuit the dash and the dot feed-magnets respectively. By fitting different short-cir-cuiting cams to each call device on a chain, each stationmay be made to respond to a different call. By buildingup the appropriate number of impulses either camshaftcan be used for any desired switching function, such ascalling attention on through circuits which are notstaffed, the temporary switching-in of re-perforators, orthe directing of traffic over selected routes.

100 VDash t Dot

,uu 1 uuT

I00V +

UUR2

Fig. 18.—Diagram of connections ot automatic scrutinizer.

The delay due to switching signals is very slight, theline time occupied by call being of the order of only 4seconds on a circuit working as slowly as 1 000 signalelements per minute.

Automatic scrutinizer.A schematic diagram of the connections of this device

is shown in Fig. 18. A relay Rl is operated by both dotand dash signals. The attraction of the armature atthe beginning of a signal causes the capacitor C to becharged, and its release at the end of the signal causesthe capacitor to be discharged. These charge anddischarge kicks, taking place at the beginning and endof a signal, may pass through a short-circuiting contactSC or the windings of an alarm relay R2 according towhether the short-circuit is made or broken at thatinstant. The short-circuiting contact is controlled by acam mounted on the main shaft of the re-perforator orthe re-transmitter, breaking the contact once per signalelement, and the contact mechanism is rotatable aboutthe cam in order that its phase may be set relative to theselection point of the instrument. When the phase iscorrectly set, the contact is broken for a short periodwhich embraces the selection point, and, should eitherend of the signal approach the selection point, the

charge or discharge kick passes through the windings ofthe alarm relay R2 instead of across the short-circuitingcontact. Fig. 19 (Plate 3) shows a practical case of theaction of this device. Records are shown of the incomingand outgoing signals at a branch on a main-line circuit.The incoming signals, shown at (a), have been distortedby a sudden dash bias at the point indicated by anarrow, though the distortion is not sufficient to causefailure of the re-transmitted signals. The automaticscrutinizer has indicated the danger point by anincrease in the size of the record of the outgoing signals.

The alarm relay R2 may be used to operate an alarmlamp or bell, to start up records, or to indicate on arecord the faulty signal which caused it to be energized.

Channelling.In recent years the number of non-loaded cables on

which 2-channel systems have been installed, operatingeither at even or at uneven speeds, has grown consider-ably. The conditions which justify the use of channellingare: (i) when the normal speed of the chain, worked singlechannel, is too high for the satisfactory use of a printer;(ii) when cables capable of high speed are included in alower-speed chain; (iii) when it is necessary to combine thesignals of two lower-speed cables into one faster cablein an emergency due to cable interruptions.

Since all signals in cable code—dot, dash, and space—are of the same length, even speed channelling consistssimply in allotting alternate signal periods to twoseparate transmissions. Two transmitters running athalf aggregate speed are phased correctly and theirmarking periods are reduced to less than 50 % to providethe alternate transmissions. At an intermediate stationthe re-transmitter runs at aggregate speed and passes onthe combined signals. At the terminal the signals arereceived on two re-perforators running at half aggregatespeed, the selection points being phased into alternatesignal periods. A re-transmitter may be substituted forone or both of these instruments by automatic switching,or permanently where channels are split into two separatecables. At the present time 16 cables are permanentlychannelled by this method, and apparatus is availablefor channelling during emergency on many more.

Channels bearing a simple speed relation to each othermay be provided by adaptations of the even-channelmethods described.

As previously mentioned, particular use has been madeof uneven channels on the London-Rio de Janeiro andLondon-Buenos Aires chain, where the aggregate speedof the system as far as Ascension Island is capable ofthe combined speeds of the Ascension-Rio de Janeiroand the longer Ascension-Buenos Aires sections. Recep-tion at Buenos Aires is limited by interference, a longlength of the Ascension cable being laid in shallow water.Consequently it has been found economical to channel atspeeds bearing the ratio 2 : 1 in the London-Rio deJaneiro-Buenos Aires direction and 3 :2 in the Rio deJaneiro-Buenos Aires-London direction. The fast chan-nel is known as the " A " and the slow as the " B "; thuswith 2 : 1 ratio the cycle of unit signal periods allottedto the two transmissions is AAB, while with 3 :2 ratioit is AABAB. Modifications of standard regeneratorapparatus are necessary only at the terminal and splitting


stations, as relaying stations on the chain pass on theaggregate signals only, picking out the signals of eachchannel for record purposes by means of cam-operatedcontacts. The divergence of channels, as in the London-Rio de Janeiro-Buenos Aires case, presents no difficultiesas far as synchronization is concerned; on the other handthe combination of channels (as at Ascension in the Riode Janeiro-Buenos Aires-London direction) can beachieved only by some " storage " system on one channelor by the speed control of the terminal transmitters fromthe outward signals, owing to the impracticability ofmaintaining identical speeds. The latter method is notvery successful over long chains, as sway and consequentphase trouble are introduced.

" Storage " by re-perforation was first used, but thehigh .operating costs led to the introduction of " capacitorstorage."* With this method a margin for speed varia-tion between the terminal transmitters is ensured byproviding a loose link in one channel at the combiningstation through impulses stored in capacitors. Eachincoming impulse from one cable is led into a capacitorby means of a distributor under the control of thesignals on that cable, is stored, and is later picked upby a distributor under the control of the signals on theother cable, whence it is combined in correct sequencewith the signals from the second cable. In the standardunit 60 capacitors are used, 30 for dots and 30 for dashes.Thus, if the storage is set in the centre of the marginavailable—i.e. a storage delay of 15 impulses—oneterminal controlling clock may gain or lose a time corre-sponding to 15-unit signal periods with respect to the otherbefore re-adjustment of the delay becomes necessary.In practice it is found that observation of the delay isrequired at the most once in about every 8 hours.

Variable-ratio channelling! has been evolved to meetthe needs of a particular case where circuits of variousspeeds bearing no simple relation to each other are com-bined over a section capable of a speed equal to twice thatof the fastest chan'nel.

Improvement in speed control.The method of bringing the speed of the reed under

control of the clock has been described in a previouspaper. $ Briefly, any slight deviation of the reed speedis corrected by the angular displacement of a permanentmagnet in relation to an armature carried at the ex-tremity of the reed. The advent of increased speedsand of channelling showed that, although the averagespeed was maintained, sudden jumps would occur whichhad the effect of crossing channels during pauses insignalling. It was known that the addition of thepermanent control increased the sensitivity of the reedto variations in the voltage of the supply, and thistrouble was traced to momentary changes in voltagewhen a new bank of cells was switched on, and similarcauses.

Investigation has shown that a relationship can beobtained between the natural frequency of the reed andthe control exercised by the permanent magnet, wherebythe change in reed frequency can be reduced to a neg-ligible amount even when a variation as great as 5 %

occurs in the voltage of the supply. The action may bedescribed as follows: Considering the effect of a reduc-tion in supply voltage on the reed alone, the amplitudeof the reed vibration will become less and the frequencywill increase. The presence of the permanent magnetproduces the same effect, but in a degree depending onits angular position and its axial distance from theiron armature on the reed blade. Thus it is possibleso to adjust the permanent magnet that, as voltagedrop takes place, the resultant reduction of amplitudecarries the reed out of the field of influence of thepermanent magnet by precisely the amount necessaryto counteract the increase of speed due to voltage drop.

A record of the behaviour of a reed when the voltageis varied is obtained by combining in the coil of a siphonrecorder the output of the reed and that of a second onerun from a steady supply during the trial. By thismeans a change in phase of the two reeds will be shownas an alteration in the amplitude of the siphon move-ment across the tape.

The specimens in Fig. 20 (Plate 3) obtained from trialscarried out at a station show clearly in the first case thechange of phase due to a 5 % voltage variation with thereed adjusted in the original manner, and in the secondthe steady running achieved by the new method ofadjustment.

The direct printer.*The direct printer already mentioned is now coming

into use at terminal stations. The selection of theappropriate type bar and the printing of the character,immediately upon the reception of all the elements ofthe complete letter, eliminates the delay of l-l£ minutesinvolved in the use of a perforated-tape-operatedprinter. The reception of a special signal combinationwill, however, bring a perforator into action whenrequired for a message which has to be re-forwarded onanother circuit.

The printer is designed to operate in conjunctionwith a synchronous system. Driving power is derivedfrom a direct-current motor, but speed control is obtainedfrom a phonic motor constantly phased to the incomingsignals. When used with the regenerator system, thephonic motor is driven from the synchronizer unit inthe same manner as a re-perforator.

Signals from the main circuit relays are fed to theprinter without regeneration, the instrument itselfperforming this function. The formation of the charactercombinations is determined by three electromagnets onthe selector head; one of these magnets is energized bydots, the second by dashes, and the third, called thespacing magnet, is energized by every impulse received,its function being to differentiate between signallingand spacing conditions.

Selection takes place once per unit signal period, thephasing of the printer being set so that selection occursin the centre of each period. The presence of an impulsewill cause the displacement of a dot or dash comb and,at the end of the character, the letter space causes theactuation of a spacing comb or gate permitting theexploration for the permutation set-up to take pi ace.

• British Patent No. 378090.t British Patent No. 45427G. t H. H. HARRISON: loc. e%t.

* This instrument is fully described in British Patent Specifications Nos.463975, 463976, and 463977.'


In the next unit period one exploring blade descendsinto its aligned slot, causing that particular combinationto be stored in a slide. In this or the next unit period,depending on whether the printing shaft is geared forlow- or high-speed operation, this slide is engaged byone of the type hammers and the character is printed.

A combination that is not printable is indicated on theprinted copy by a plus sign, audible warning being alsogiven when this takes place. The printer has attaineda speed of well over 100 words per minute and may beconstructed to print from either cable, morse, or equal-length codes.

Localization of Breaks and FaultsThe older methods of testing for breaks—such as the

Kennelly two-current test to false zero and Schaefer'sscale-zero test—are now used as confirmatory testsupon the Lloyd and Lloyd-Black break tests.* Faultsare still localized by the loop test where a second cableis available, and also by overlaps. Work has beencarried out recently upon corrections to the overlaptest with a view to minimizing errors due to low insula-tion. More accurate localization of high-resistance faultsis thus obtained. Higgitt's loop methodf of localizinghigh-resistance faults has proved of great use, particularlyon underground landlines.

Records of the accuracy of the localization of faults,and breaks, taken by a large shore station having 12cables of different lengths and characteristics, show thatover a period of 3£ years the average error, expressed as

Error in ohms, is onlythe ratio

0-26%.Resistance in ohms to break or fault'

Protection of CablesWith regard to the causes of breaks and faults,

trawlers and corrosion are about equally responsible.It is hoped that a fair number at least of the trawlerfaults may be prevented by the use of the plough methoddeveloped by the Western Union Telegraph Co. Muchwill depend on the possibility of finding suitable pathsfor ploughing in various parts of the world, and alsoupon finding a successful method of dealing with cross-ings. To gain the greatest economy it should be possibleto plough-in a light type of cable.

In dealing with corrosion the tendency is to concen-trate upon protection of the armouring from contactwith sea water. Corrosion is most active in warmwaters; it is caused both by chemical constituents onthe surface of the sea bed and by the products of marinegrowths which attach themselves to the cable. Thegalvanizing of the iron or steel wires is of protectivevalue in proportion to the thickness of the coat. Thecoating specified by the British Standards Institution isconsidered too thin for the purpose, and as far as possiblea more thickly coated wire is used. Under deep cold-water conditions galvanizing appears to be effective inprotecting the armouring, whereas under the corrosiveconditions mentioned above the zinc does not form aself-protective coat but appears to be dissolved at therate of about 0-001 in. of thickness per annum. For

• Electrictan, 1919, vol. 83, p. 498. f Ibid., 1921, vol. 86, p. 96.

this reason, as well as to protect the galvanizing fromdamage when being handled and to provide a highercoefficient of friction on the paying-out and picking-updrums, some form of strong adherent waterproof cover-ing is required. Where causes of chafe are not present,smaller-diameter armour wires with a more efficientcovering than was previously used are therefore beingemployed.

The galvanized wires, which must be weathered orproduced with a matt surface, are tar-dipped and eachwire is further protected by an impregnated cotton tapewound round it. After the formation of these coveredwires into the armouring of a cable, the cable is coveredwith a serving of hessian tape having on its inner surfacea covering of an inert waterproof substance; in theprocess of wrapping and compounding, the inert sub-stance is forced into the interstices of the armouring.Applied outside the tape is a compounded serving oftarred jute. With this protection it is very difficult forthe water to reach the galvanized iron or steel wire insuch a way that circulation can take place to bring freshcorrosive products to bear upon the susceptible portionof the armouring.

Investigations into the cause of the reduction of thezinc covering of armour wires when cable is stored intanks ashore and on ships, tend to show that the presenceof sulphate-reducing bacteria is a likely source of thetrouble. The irregularity of the incidence of thisphenomenon would appear to support this theory.Active investigation of the subject is proceeding.

Duplex BalancingAlthough from theoretical considerations there would

appear to be little difficulty in obtaining a duplexbalance upon a submarine cable, the practical conditionsare such that the exact measurement of balance error,the calculations therefrom of the position and amountof corrections to the network, and their application andre-measurement, take so much time that the processbecomes uncommercial. A skilled man can produce agood workable balance in a quarter of the time by roughobservations of the balance error at various frequencies,and by using his experience in determining the alterationsnecessary to the artificial line.

Refinement of balances is progressing steadily asopportunities present themselves. Many balances havebeen obtained in the past by the use of a number ofadjustments such as small resistors inserted betweensections of the line, shunts over sections, etc. It hasbeen found, however, that as a general rule the cleareran artificial line is of additional adjustments (known asfakes), i.e. the more nearly it is allowed to remain inits designed similarity to the cable as regards ratio ofresistances to capacitance, the more stable the balance.Further, the art of obtaining a steady balance lies inthe accurate application of adjustments which will causethe inductance of the cable (which decreases withfrequency owing to the return current through theearth closing-in on the cable) to be imitated in thecombination of resistance and capacitance in the artificialline. These adjustments consist of the insertion ofresistors between certain groups of capacitors near thebeginning of the line and the earth.


One of the main difficulties in the determination ofthe values of the resistors and their exact location lies inthe want of homogeneity of the cable as regards theratio of resistance to capacitance. As the cable systemwas extended the copper conductors became larger andlarger in order to obtain greater speed. As it is quiteimpossible to stock all these types for repair work theportions near the end of a cable, where repairs are mostfrequent, often consist of a patchwork of different types.This enhances very much the difficulties in the applicationof calculated values, since compromises have to beeffected.

Many attempts have been made to put balancing ona scientific basis. A few years ago D. C. Gall made aninvestigation of the problem and evolved a methodbased on a.c. measurements.* The principle of thismethod proved to be sound, but for the practical reasonsgiven above it has not been adopted.

Instances of recent increases of speed, due, in the main,to refinement of balance, are given in Table 1.

Table 1

Cable

Aden-ColomboColombo-Penang No. 1Porthcurno-Gibraltar

No. 4Porthcurno-Madeira .

Signal elementsper minute

fromfrom

fromfrom

1

11

846 to250 to

412 to114 to

11

11

019413

693500

Increase

2013

2034

0/

/o/o

0/

/o/o

(7) WIRELESS TELEGRAPHYGeneral

The five transmitting stations and four receivingstations controlled by the operating company in Englandare shown in Fig. 43 (Appendix 2). The telegraphand facsimile circuits handled by these stations are setout in Table 2.

Destination

BangkokBelgradeBerneBeyrouthBogataBombay .Buenos Aires ICairoCape TownFalkland IslandLas PalmasLisbonMadridMaracayMelbourneMelbourne

Table 2f

Transmittingstation

DorchesterOngarOngarDorchesterDorchesterTetney DDorchester DDorchester DBodmin DDorchesterOngarOngarOngarDorchesterTetney DOngar D

Receivingstation

BrentwoodBrentwoodBrentwoodBrentwoodBrentwoodSkegnessSomertonSomertonBridgwaterBrentwoodBrentwoodBrentwoodBrentwoodBrentwoodSkegnessSomerton

Destination

Melbourne FMontrealMoscowNairobiNew YorkNew York FNew YorkParisRio de JaneiroSalisburySantiagoShanghaiShanghaiSofiaSofiaStamboulTeheranTeheranTokioTokio FVienna

Transmittingstation

Ongar DBodminOngarOngar DCarnarvonDorchesterOngar DOngarDorchesterOngar DDorchesterDorchesterOngar DOngarCarnarvonOngar DOngarDorchesterDorchesterOngar DOngar

D

D

D

D

Receivingstation

SomertonBridgwaterBrentwoodBrentwoodSomertonSomertonBrentwoodBrentwoodSomertonBrentwoodBrentwoodSomertonSomertonBrentwoodBrentwoodBrentwoodBrentwoodBrentwoodSomertonSomertonBrentwood

• British Patent Specifications Nos. 262991, 301340, and 302474; alsoElectnctan, 1920, vol. 85, p. 736.

t A directional aerial is indicated by D and a facsimile circuit by F.

All these circuits terminate in London. The majorityof the links between the nine stations and Londonconsist of underground land cables worked on a tone-channel system; and only a few physical circuits areretained. The London ends of the lines are terminatedin a control room where the circuits are monitored anddistributed to the various sending and receiving positionsin the main instrument room.

The majority of the wireless circuits are worked byautomatic transmission and manual reception, operatorsbeing employed to transcribe the undulator signals intotypewritten script. On certain circuits, however, auto-matic reception using regenerator methods is possible.A description of these is given in a later section of thispaper.

Wireless messages at an originating station such asLondon are normally prepared in the form of a' perforatedtape and are fed through a Wheatstone transmitter, theimpulses from which operate a relay in the controlroom. This relay controls one of the outward tonechannels to the selected transmitting station, where thesignals are filtered and made to key the appropriatetransmitter by means of either a valve or a relay circuit.At the control switchboard of the transmitting stationare lamp devices which give an immediate indication,both audible and visible, when the transmitter fails toemit a signal from the line.

At a receiving station the homeward tone channels arenormally controlled b y the incoming signals withoutthe use of relays, and thus a possible source of localnoise is eliminated.

Automatic reception is employed on several importantcircuits. The system used has, to a large extent, thedetectability of error enjoyed by cable code. Thecircuits are usually worked with two channels on a time-basis system, the elements of the channels being inter-leaved. The code used is similar to the double-currentcable code described previously, but after a period ofquite successful working with the normal double-current


cable code it was found that the code could be modifiedwith advantage for use on wireless circuits. Thesemodifications are described later.

On the circuits which are worked with regeneratormethods poor conditions are covered by making use ofthe Verdan principle. Both channels carry the samesignals, but one set is delayed at the transmitting endby capacitor storage and the other set an equal amountat the receiving end. The delay is usually set at 2seconds, since this period has been found to cover themajority of short fades and atmospherics.

The increase in the number of wireless servicesoperated during recent years is indicated on the mapsin Appendix 2.

The accumulation of information on working condi-tions has led to steady progress in the methods ofoperating the various services; and prolonged tests onthe majority of circuits have enabled the best types ofaerials to be selected to suit particular conditions. Inthis connection progress has been mainly in the modifi-cation of standard designs and the use of more efficientdirectional aerials.

In co-operation with manufacturers, marked improve-ments have been effected in the design of transmitters,particularly in the direction of controlled-frequencydrive, giving increased stability of transmitter frequencyand permitting the use of more selective receivers.Unbalanced absorber keying, in which only a portionof the power used on " mark " is absorbed by the keyingcircuit during " space," is extensively adopted. Sufficientpower only is absorbed on the space signal to cut offradiation and at the same time to keep keying reactionson the supply system down to a reasonable figure. Theresultant saving of power is considerable.

In special cases methods of frequency modulation areused for overcoming fading troubles.

Verdan working, while catering for the eliminationof either fades or extras, cannot deal with conditionswhere both these factors are present at the same time.Recently experiments have been carried out on " doublefrequency keying,"* by means of which the intelligenceis in effect conveyed twice by one transmission. Thismethod includes means by which all types of inter-ference may be converted into a single form of mutilation(e.g. fades), as is required for automatic repetition ofsignals by the Verdan system. Considerable work onsimilar methods has been carried out elsewhere, f butthe practical results achieved so far have been limited.

A further advantage of the double-frequency system isthat the normal cable code can be used in place of thedouble-current cable code. A full description of themethods adopted is given later, together with circuitdrawings. Means for automatically adjusting the receiv-ing oscillator frequency, employing the magnetic-clutchmechanism fitted on cable relays, are also described. Itwill be appreciated that the problems of frequency controlare considerably greater on a telegraph circuit than on atelephone circuit, since a pause in signalling may meana complete cessation of signal, during which control isinoperative. In addition, owing to the closeness of the

* British Patent No. 480289.t A. BARKER and H. C. A. VAN DUUREK: British Patents Nos. 388625

and 419526.

frequencies used, double-frequency keying has called forgreater stability of heterodyne in receivers.

Receiver design as a whole is as up to date as possible,automatic gain control, modified limiter circuits, etc.,being used generally. For ships' services such detailsas automatic searching over a band of, say, 595 to 605metres are of interest.

^Normal diversity reception is adopted where justifiedby considerations of space and cost.

Dual reception of two wavelengths (i.e. frequencydiversity as opposed to space diversity) is also used, butis a more expensive method, as two distinct transmittersare required. Where spare apparatus is available themain use of dual transmission and reception is to main-tain a continuous service during times at which a changeof wavelength is made.

The main circuits, both transmitting and receiving,retain the Franklin vertical beam aerial in a modernform; no other form of aerial has proved to be equallyeffective, having regard not only to signal fields but alsoto signal/noise ratios and to the reduction of echo.This is of extreme importance during maximum sunspotyears. Considerations of cost, however, have led tothe adoption of the Franklin series phase aerial on manyrecent circuits. Rhombic aerials and other types areused in certain cases.

A general policy of separate transmitting and receivingsites has been followed, owing to the multiple servicesrequired, such as ships' services, direction-finding forships and aircraft, fixed telegraph services, and radio-telephony. Reasons of economy in buildings and staff,however, have led to the adoption of common siteworking of wireless and cable circuits in certain cases.This arrangement has raised many problems, includingthe suppression of " noise " created by the cable appar-atus, the provision of remote control of transmittersfor starting up and shutting down, and the use ofautomatic wave-change switches.

Centralization of a number of services at one receivingor transmitting point has given increased flexibility withthe aerial systems available. Thus during maximumsunspot years the London-Buenos Aires circuit maybecome unworkable for many hours over the directroute with the normal wavelengths available. However,advantage can be taken of the Tokio aerial systems atLondon and Buenos Aires to work over the long route,where such wavelengths may be usefully employed.The necessity of installing additional aerial systems isthus avoided. Similar use is made of other aerialsystems, such as London-Bangkok and London-Maracay.

Cosmic data are carefully correlated with transmissionand reception conditions, enabling the most efficient useof wavelengths to be made to suit the changes in the solarcycle. A few typical data charts are shown later inthe paper. A continuous check is also kept at Brentwoodon transmitter frequencies, upon both the company'sand other transmitters.

Double-current Cable Code, and Method ofOperation on Wireless Circuits

The double-current cable code, already described, wasfound to have some disadvantages when used on a wire-


less circuit, as comparatively long periods of spacingsignal were possible when no signal was transmitted.Thus on a circuit working with two even-speed channelsa figure 0 on each channel might result in a quiescentperiod of 10 unit time-intervals. In certain circum-stances this introduced difficulties with regard to thegain-control apparatus at the receiving wireless station.There was also a tendency for atmospherics and quickfades to turn dashes into dots, and vice versa.

As the result of suggestions made by R. Bruford thecode was therefore modified in such a manner that thelongest possible unbroken space periods and the longestpossible unbroken mark periods have been reduced to2 unit time-intervals. Referring to the two channelsas Channel A and Channel B, on Channel A a dot

possibility of a dot being changed into a dash, or viceversa, by atmospherics or fades is considerably reduced.The suppression of half a dot or dash signal will resultin the signal being dropped. The spaces may, of course,be mutilated and be changed into a dot or dash, butthis change should be readily detectable.

The code also adds to the secrecy of wireless messages.Fig. 22 (Plate 3) illustrates this point. A record of signalsreceived on the Bombay circuit is shown, togetherwith the separated channel signals. For clarity, 50 %channel signals are shown.

The regenerated channel signals operate relays whichnormally actuate a perforator feeding into a tapeprinter. In special cases, calling and switching devicesare operated and the signals fed into re-transmitters.

A dash ! d o t

selection points iAdot I*

selection poinfcsi i

(a) Combined ! f "channel signals!

B dashselection points

Bdotselection points

B ! A I Bspace'dash!dash|Space! dot

! ! •dot

B ; Adot i dash

i

\\

B I A > Bdot 'spacejspace

,,>. A channel r\o) regenerated |

cable codesignals

B channel(c) regenerated,

cable codesignals

Fig. 21.—Double-current cable code signals on wireless.

consists of 50 % space followed by 50 % mark, adash consists of 50 % mark followed by 50 % space,and a space signal of 100 % space. On Channel Ba dot is the same as the Channel A dash, a dash thesame as the Channel A dot, and a space signal consistsof 100 % mark. Combined channel signals are illustratedin Fig. 21 (a), the A channel transmitter signalling theletters AA and the B channel, transmitter the letter B.Thus the first signal period shown consists of a dot fromthe A transmitter (space/mark), the second signal periodis a space from the B transmitter (100 % mark), thethird a dash from the A transmitter (mark/space), thefourth a dash from the B transmitter (space/mark), andso on.

The signals are received on synchronous regeneratorapparatus similar in principle to that used on the cablecircuits, but owing to the necessity of uneven selectionpoints a cam-and-contact method of selection is substi-tuted for the mechanical selection of the cable apparatus.These selection points are shown in Fig. 21(a), and theresultant regenerated channel signals in Figs. 21(6)and 21(c). In each case the dashes are delayed, afterselection, by capacitor storage to bring them into phasewith the dots. The A channel signals are selectedduring the marking periods and the B channel signalsduring the spacing periods, with the result that the

The direct printer, already described, will shortly bein use on these circuits.

Transmitting-Station MonitorThe wireless transmitting-station monitor is included

in this paper as an example of a number of warningdevices incorporated in both wireless and cable circuits.The device is designed to provide a warning in the eventof improper operation of the transmitter under conditionswhere the usual protective apparatus would fail tofunction.

A differentially wound relay is arranged with onewinding energized from the transmitter keying circuitand the other from the radio check circuit. The inputsfrom the two circuits are adjusted to balance, but if theradiation fails to follow the telegraph signals, or variesin value, the balance is destroyed and the relay operatesto ring an alarm bell. The device will give warningin the event of an aerial or feeder fault, failure of anyone valve in the magnifier stages, and failure or partialfailure of the auxiliary rectifiers for the magnifier stages.

Various methods of connecting the relays are possible,and the choice must depend on the system of lineworking and of keying the transmitter. The mostgenerally applicable method of energizing the winding(A) which is operated by the incoming line signals is to


key a valve whose grid circuit is joined in parallel withthat of the sub-absorber, the line winding of the relaybeing in the anode circuit (Fig. 23). The monitor isthen independent of the method of line working or

Fig. 23.—Transmitter monitoring relay: line winding circuit.

transmitter keying and does not appreciably increasethe current across the keying-relay contacts, while theworking current in the monitor relay winding is not inany way controlled by current values required in othercircuits.

The second winding (B) of the differential relay—thewinding energized by the radio signals—is normally

240 V

Transmitterfeeder

Fig. 24.—Transmitter monitoring relay: circuit of radiopick-up winding.

joined in series with the resistors of the radio pick-upcircuit. The pick-up is adjusted to give a currentthrough this relay winding which balances that flowingin the line winding in the mark-valve anode circuit ofthe valve keying unit. Fig. 24 shows the connections,and the type of radio pick-up used. The relays arefitted with change-over contacts. These are joined upto show a green light under normal conditions and to

ring an alarm bell and show a red light when the relayoperates.

Keying a Tone Channel on a Landline Circuitfrom a Wireless Receiver

Many methods are in use for keying a tone channelon a landline circuit from the received wireless signals.Keying circuits have been improved steadily with aview to " cleaning up " the wireless received signals andreducing distortion due to multiple-ray and similareffects. Fig. 25 shows an improved type of tone-sendercircuit. After the signals have been rectified in thereceiver they are passed through two limiting stages,the first of which has provision for bias correction(i.e. a 0-01- to 0-1-JUF variable capacitor C), and thesecond of which is necessary to reduce - the slopingfront of the signal resulting from the bias correction.The anodes of the oscillator and the second limiterstage are fed through a galvanometer to give a biasindication, balance being obtained by means of a variableresistor in the second limiter anode. The actual tone-sender portion of the unit is similar to the band-passtone-sender normally used, except that battery valvesare employed.

Such tone senders can be grouped at a convenientpoint away from the receivers, for it is found that theextension leads produce no appreciable interference andthat the capacitance of the maximum lengths of leadslikely to be required has a negligible effect on the signalbias. The oscillator keying wave-form is square, andany malformation of the wave-form due to the actionof the bias-correction capacitor is not reflected in theoscillator anode. As the output wave-form of theoscillator is square, the listening point before the filtermay be used as a checking point for the existing audio-frequency check bridges.

Double-frequency Keying of WirelessTransmitters

The following is a description of a method, at' presentbeing developed, of operating short-wave wirelesscommunication channels with a view to providing meansby which all types of interference may be convertedinto one form of mutilation (e.g. fades) as required for

Test(space)

+8V

O-24V

Fig. 25.—Valve-keyed tone sender.

W O O D : E M P I R E T E L E G R A P H C O M M U N I C A T I O N S 655

automatic repetition of signals by the Verdan system.These developments have the ultimate object of enablingthe direct printing of messages to be adopted on all mainwireless circuits.

Transmitter .The transmitter (Fig. 26), which, following the conven-

tional design, includes a master oscillator (M.O.) fordriving the main amplifier at the required frequency, isarranged to emit either a spacing or a marking frequency(differing by about 600 cycles per sec.) in accordancewith the position of the tongue of the control relay.For example, when the position of the relay tongue issuch as to make the first valve conductive, the secondvalve—a diode—is arranged to be non-conductive, thefrequency of the master oscillator being then substan-tially unaffected by the small capacitor plate C. Inthe other position of the relay tongue the first valveis non-conductive and the diode is then conductive,

The high-frequency amplifier, intermediate-frequencyamplifier, audio-frequency amplifier, and second hetero-dyne are of conventional type and call for no specialcomment. The remaining portion of the receiver ispeculiar to the special type of signalling, and will there-fore be dealt with in greater detail.

The output of the audio-frequency amplifier containsthe mark and space signals, and operates (a) into adouble-wave rectifier for securing automatic gaincontrol of the high-frequency and intermediate-frequencyamplifiers; and (b) into high- and low-pass filters toseparate the mark and space signals. These in turnoperate into the combiner unit and automatic frequency-control unit, the latter being preceded by suitableband-pass filters.

Combiner unit.As will be seen from Fig. 28, the combiner unit consists

of a rectifier (VI, V4), a limiter (V2, V5), and a coupling

Fig. 26.—Double-frequency keying system transmitter.

with the result that the plate C reacts on the masteroscillator in such a manner as to produce the desiredfrequency-change. Although this frequency-differenceis capable of accurate adjustment, the receiving apparatusas described permits a variation of ± 15 % to ± 20 %of the mean setting.

Receiver.The receiver, which may be of either the superhetero-

dyne or the " straight " type, includes the followingfeatures in addition to an audio-frequency amplifierand high- and low-pass filters for amplifying andseparating the mark and space signals:—

(a) Automatic gain control of the earlier stages of thereceiver, comparable in degree with that used forreception of telephony.

(b) A combiner unit for automatically operating theoutput relay by the mark and/or space signal, (i) underconditions of reception where one or other frequency iscompletely absent owing to selective frequency fading,or (ii) where the degree of fading (or interference) isexcessive, in such a manner as to convert all types ofmutilation into one form (e.g. fades)'.

(c) Automatic frequency control of the first heterodynecircuit so as to maintain sensibly constant the beatfrequencies in the high- and low-pass filters.

Fig. 27 gives a schematic diagram of the receiver.

stage (V3, V6) for each signal (mark and space), thecoupling stages being interconnected by a " dead-beat "multivibrator (V7, V8), the anode circuits of which carrythe coils of the output relay.

The multivibrator connected in this manner can becontrolled (i) by the mark signal alone, (ii) by the spacesignal alone, the form of connection being such that thesignals are automatically reversed by the multivibratorso as correctly to control the output relay, (iii) by boththe mark and space signals simultaneously. Moreover,atmospherics or other interference joining togethercharacters of one signal will, in the presence of the othersignal, leave the output relay undisturbed. Such signalsare shown in Fig. 29.

I t will be seen that, since the gain of the high-frequencyand intermediate-frequency amplifiers is controlled byeither the mark or the space signal, whichever is thestronger under conditions of " selective frequencyfading," the receiver will operate on " peak " signals.Normally, therefore, a degree of protection from inter-ference and cross-modulation and from overloading ofthe receiver is secured. Certain types of interference,e.g. a partial fade of any particular signal element asdistinct from complete absence of signal, may producefaulty operation of the output apparatus. Therefore,in practice, a certain degree of frequency modulationon both the space and mark signals is desirable, which

€56 WOOD: EMPIRE TELEGRAPH COMMUNICATIONS

Automatic gam control

Input

1H F amplifier I F amplifier

i h e t(controlled)

A F amplif ier

HPfilter(mark)

LP filter(space)

2ndhet

Combiner unit(markai%space

Output

BP f i l ter

8 Pf liter

Automaticfrequency

control unit(marksn%r space)

Automatic frequency control

Fig. 27.—Double-frequency keying system- receiver

HT +

Mark

SpaceJ

H.T.+ ,o 11-

Fig. 28.—Double-frequency keying system combiner unit.

JUUlfi

{Space

ru in j LJiJinJ i__:c»*.n«.: ; J ;output

Fig. 29.—Combined signals.


in turn necessitates a somewhat larger difference of meanfrequency of these two signals.

When the field intensity of the peak signals becomestoo low to permit satisfactory printing, the connectionof the combiner unit may be modified by switch operationso that a mark is recorded only when the mark signal,as received, is on and the space signal off. Under allother conditions (e.g. fading or mutilation of eitherfrequency channel, or a correctly received space signal,i.e. space on and mark off) a space is recorded. By thismeans automatic repetition of signals on the Verdanprinciple will cater simultaneously for atmosphericsand fades.

A convenient method of modifying the combinerunit for this purpose is to disconnect the anode of

Q)T»

"aE<oa>>3TO<U

Filter unit nF.I. I I

1

11 ,

/-A/

Filter unib1 1 F.2

1 1

1J I

1

V/+A/

f

Fig. 30.—Response curves.

valve V6 (Fig. 28) from its resistor-capacitor combinationand connect it to the anode of valve V2, the multivibratorthen being actuated only by valve V3.

Under these conditions valve V3 will be conductive anda mark recorded only when valves V2 and V6 are non-conductive, i.e. the mark signal on and the space signaloff. Under all other conditions a space will be recorded.Alternatively, the same effect may be obtained by theintroduction of a multi-electrode valve, in which Grid 1is controlled by the mark and Grid 2 by the space signal.

Automatic frequency control of receivers.On the receivers a form of automatic frequency control

is used which is applicable equally to telephony, todouble-frequency keying, and to the normal type oftelegraph transmission in which there are quiescentperiods during signalling.

The receiver oscillator is adjusted in the first instanceso as to produce, either directly, or indirectly in con-junction with one or more auxiliary oscillators, a beatfrequency / (Fig. 30). Energy at this frequency ispassed through two discriminating filters Fl , F2 (Fig. 31),

the outputs of which are connected to two a.c./d.c.convertor units. The output circuits of the latterconsist of valves (V3, V6) and anode coils (Cl, C2), theH.T. supply being either alternating current or inter-rupted direct current. The coils Cl and C2 form the

Filter unitF l

AC/DCconverbor

Filter umtF2

AC/DCconvertor

240V 50

Fig. 31.—Automatic frequency control.Ci, C2 are bias-corrector electromagnets.

windings of the electromagnets of a magnetic-clutchmechanism as described in connection with the bias-correcting unit shown in Fig. 13(6). In this case thethreaded main spindle attached to the clutch disc ofthe corrector controls the position of a capacitor vaneEl. A sectional drawing of this mechanism is shownin Fig. 32.

When a magnet is energized as a result of one of theoutput valves being made conductive, the resultantfield passes through the adjacent soft-iron portions ofthe armature spider and the portion of the clutch discbetween them. The armature is pulled against thedisc and towards the magnet poles simultaneously,causing the disc to move with the armature. When thearmature current ceases and the field collapses, thearmature is returned to its central position by controlsprings, but as it is no longer attracted to the disc the

To main oscillator,grid or anode

Fig. 32.—Automatic frequency-control mechanism

latter and the capacitor vane are left in their newposition. The above cycle of operations continues, asa result of the interrupted nature of the current in theoutput valve, until such time as the valve is renderednon-conductive.

With the receiving oscillator adjusted so as to producea beat frequenc)r / through the filters, the output valves


of the a.c./d.c. units are adjusted so as just not toconduct. Any tendency for the beat frequency tochange is accompanied by an increase of current throughone unit, causing its ouptut valve to conduct andresulting in the operation of the magnetic clutch andtherefore the capacitor vane El, the direction of move-ment depending upon which of the output valves isactuated. The resultant change of the capacitance ofEl E2 is arranged to react on the constants of theoscillator (Fig. 27) so as to effect the necessary changeof frequency for the desired constant beat frequency / .

Satisfactory operation of the corrector can also beobtained by initially setting the output valves to conductand by balancing the resultant flux set up in each coil.The arrangement described above is, however, preferable.The range of correction can be varied between widelimits in accordance with the size of the variablecapacitor and the sensitivity of the a.c./d.c. convertorunits.

As an alternative to the method indicated in Fig. 31for obtaining the interrupted nature of current in thecoils Cl and C2, the H.T. supply may be direct current andthe coils Cl and C2 may each form part of an oscillatorycircuit associated respectively with the valves V3 andV6, the oscillation amplitudes of which may be varieddifferentially in accordance with the tendency offrequency / either to increase or to decrease.

For the reception of double-frequency signals thecorrection is applied to one only of the two frequencies(mark or space).

If, at the transmitter, an elementary oscillator be usedin conjunction with the master oscillator—the one formark and the other for space—and if the differencefrequency be maintained constant in the mannerdescribed above, the resultant beat frequencies in theoutput stages of the receiver will be constant and cantherefore be used in association with narrow-bandfilters, with consequent improvement of signal/noiseratio.

Effect of Cosmic Conditions on WirelessTelegraph Circuits

A vast amount of data is collected regarding theconditions of working on all the main wireless circuitsin order that the best methods may be evolved to suiteach particular circuit, and also that some attempt maybe made to predict the conditions likely to be met inthe future. The following two examples are illustrativeof this work.

Effects of solar cycle.The optimum wavelengths for any given route are

dependent upon the season and the time of day; anadditional and important factor is the degree of solaractivity, which follows approximately an 11^-year cycle.The last sunspot maximum period was in 1928-29, andthe minimum period in 1933—34. The years followingthe latter period have been accompanied by a markedincrease in activity, which on a half-yearly basisexhibited a well-defined peak for January to June, 1937.Records for the past 100 years reveal that, whilst theaverage length of cycle was 11-4 years, the changefrom mimimum to maximum activity occurred over an

average period of only 4-5 years; a study of wirelessdata for the year 1938 indicates that it is not improbablethat the year 1937 may prove to have been the maximumpeak of the present cycle.

As an example of the effects of the cycle on short-wavecommunication, the curve of Fig. 33(6) shows hoursduring which a wavelength of 16 metres proved usableon the London-Bombay route for the period 1930-38for the months centring on the equinoxes fi.e. Februaryto April, August to October, inclusive), whilst Fig. 33(a)

120

•§80

2 4 0

~» (Average value11 Jan/June 1938)

1930 4(a)

7 1938

2400+

52000-13

1600

1200

££0800

U| J 0400

Season equinox

A

1930 I 2 3 4 5(b)

1938

Fig. 33.—Effect of sunspot activity on utility of 16 metreswavelength on London-Bombay circuit.

indicates the average sunspot activity for the sameyears. The correlation is less pronounced at the openingof the route (AA') than at its closing (BB'), owing tothe fact that the rate of ionization associated withsunrise is considerably greater than the rate of recombina-tion around sunset. The change in conditions resultingfrom a change in activity is still more pronounced forroutes in higher latitudes, and Fig. 34 shows the per-centage time of each day for which the 16-5- and 69-metrewavelengths were employed on the London-Montrealcircuit in the years 1934 and 1937. The intermediatewavelengths are omitted for the sake of clarity. It willbe noted that increased activity, with its accompanyingincrease of ionization, enhances the utility of the shorterwavelengths.


Echo.At certain times and seasons, attenuation of wireless

signals may be so low over both the major and minorarcs of the great circle embracing the two terminals thatsignals may arrive both by the forward and by thebackward route, and even encircle the globe some two orthree times. As an example of these conditions, Fig. 35(Plate 4) shows the reception at Bridgwater of two con-secutive dots transmitted from Cape Town (Aj), togetherwith a sequence of three echoes—Bj, C ,̂ and Dx—ofthe main transmission, as a result of the rays encirclingthe globe three times. During the interval AXA2 thetransmitter was quiescent, after which the transmissionof two more dots produced a similar sequence of echoes.The distortion of the pair of dots AjA2 received over

50

2§20al0

16 5m

Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec

50

ho-u

0-10

(b)l937I65m

Jan Feb Mar Apr May June July Aug Sept Oct Nov Dec

Fig. 34.—Extreme wavelengths used on Canadian beam.

the relatively short direct path is attributed to thereception of multiple signals, to which the route isknown to be susceptible at the time and season inquestion. The echo pairs tend to become progressively" thinner " as the steeper-angle multiple rays becomemore and more attenuated with increasing distance,reception of the later echoes being confined substantiallyto shallow-angle rays.

A further example of such phenomena is indicated inFig. 36(a), Plate 4, where a similar transmission (i.e. twoconsecutive dots 1, 2, repeated at regular intervals,originating in this instance at Bombay) resulted in thereception at Skegness of additional signal elements 3,4, and 5. In this case the mutilation was caused by acombination of forward and backward echo, the speedof signalling being such that five equally-spaced dotswere produced at the receiver, the individual pairs ofdots being as shown in Fig. 36(&). The delay periodof the forward echo is of the order of 0-14 sec, and thedifference between the lengths of the forward andbackward routes is such as to produce a delay of about0-09 sec. on the backward echo. The ratio of thesetwo delays is seen to be substantially equal to the ratioY/X, i.e. 1-5.

It should be emphasized that the above records wereobtained under test conditions, i.e. with the receiveradjusted to maximum sensitivity and- the automaticgain controls removed, the purpose of the tests being todisclose the presence of any abnormal route conditions.In practice, the direct signals are generally so muchstronger than the auxiliary ones that automatic gaincontrol and similar methods can be satisfactorily usedto reject partially, if not completely, the unwantedsignal components.

ACKNOWLEDGMENTSIn conclusion, the author desires to acknowledge with

thanks the courtesy of Cable and Wireless, Ltd., ingiving permission for the production of this paper, andalso the assistance of his staff, with particular referenceto Mr. W. L. Langford, Associate Member, and Mr. A. M.Humby.

APPENDIX 1

Detailed List of Companies and Systems Incor-porated in the Communications Operating Com-pany, and Companies Associated with the GroupEastern Telegraph Co., Ltd.: Group including Eastern

and South African Telegraph Co., Ltd., Europe andAzores Telegraph Co., Ltd., West African TelegraphCo., Ltd., African Direct Telegraph Co., Ltd.

Western Telegraph Co., Ltd.: Group including WestCoast of America Telegraph Co., Ltd., Pacific andEuropean Telegraph Co., Ltd., London Platino-Brazilian Telegraph Co., Ltd. (liquidated), River PlateTelegraph Co., Ltd.

Eastern Extension Australasia and China TelegraphCo., Ltd.

Marconi's Wireless Telegraph Co., Ltd.Direct Spanish Telegraph Co., Ltd.Indo-European Telegraph Co., Ltd. (liquidated).British East African Broadcasting Co., Ltd. (liquidated).Indo-European Telegraph Department (now non-

existent).Societe Anonyme Beige des Cables Telegraphiques.Direct West India Cable Co., Ltd.Halifax and Bermudas Cable Co., Ltd.West India and Panama Telegraph Co., Ltd. [now Cable

and Wireless (West Indies), Ltd.].Cuba Submarine Telegraph Co., Ltd.Pacific Cable Board's Cables.The West Indian Cable and Wireless System worked by

the Pacific Cable Board.The Imperial Atlantic Cables.Post Office Beam Services., The Company has controlling interests in:—Indian Radio and Cable Communications Co., Ltd.Marconi Radio Telegraph Co. of Egypt.Overseas Communications of South Africa, Ltd.Companhia Portuguesa Radio Marconi.Peruvian Telephone Co.Guayaquil Telephone Co.

as well as interests inAmalgamated Wireless (Australasia), Ltd.,

and a number of foreign wireless companies.

660 WOOD: EMPIRE TELEGRAPH C O M M U N I C A T I O N S

APPENDIX 2

ROUTES OF CABLE AND WIRELESS SERVICES FORMING THE TELEGRAPH COMMUNICATIONSYSTEM OF THE" BRITISH EMPIRE

Fig. 37.—Eastern and associated Companies' systems prior to the merger.

Fig. 38.—Other cable and landline systems incorporated in the merger.

Fig. 39.—Wireless services prior to the mergerG.P.O. beam servicesServices of Marconi Co


p

Fig. 40.—Wireless services opened since the merger.

Fig. 41.—Routes of main inter-Empire traffic.Cable and landline.Wireless.

VOL. 84.Fig. 42.—Routes of main Empire-foreign traffic.

42


)Dundee

gh

Nev/castle

Liverpod/Wfendiesler <^fetncy

Sheffield PHSkeffness

e^errorthcurno

OTelegraph offices ©Wireless stations • Cable houses

Fig. 43.—Landline system in the British Isles.

DISCUSSION BEFORE THE INSTITUTION, 15TH DECEMBER, 1938

Sir George Lee: The apparatus used in submarine-cable telegraphy has an extraordinarily high degree ofprecision as regards finish and manufacture and in theaccuracy of. the balances, great advances having takenplace in the last decade.

I have been very interested in the author's descriptionof the division of traffic between the cable and the wirelesssendees. It is quite clear that the combination makes foreconomic utilization of the two services, and they are nowcomplementary to each other instead of being com-petitive. I should be interested to know what happenswhen one terminal or the cable associated with it goes outof action owing to a fault. When the terminal goes backinto action, how is the chain picked up again ?

It is mentioned in the paper that the capacitancemagnifier has the property of acting as a filter; low-passfilters would carry out that part of the operation muchmore simply and certainly, but the capacitance magnifierseems to have the advantage of isolating the cable fromthe amplifiers.

The author's description of the way synchronization ishandled is very interesting; I should like to ask whetherthe use of crystal oscillators has been considered for thispurpose. The degree of constancy of crystal oscillatorsis most remarkable; they keep constant to a fraction ofa second over long periods of time.

Perhaps the author would say whether trailing elec-trodes, as we call them, have been used for tracing faults

in submarine cables on his system. They are applicablemainly to comparatively shallow waters; we find themvery useful in the North Sea, where cables become sanded-over and the pick-up gear cannot get hold of them. Themethod is to pass a low-frequency alternating currentalong the cable, and to trail the electrodes some distanceaway from the ship by a paravane arrangement. Thealternating current can be detected by means of amplifiersand telephones on the ship until it has passed the fault.

I notice that the word " baud " is used in the paperas a time-interval, and I should like to argue in favour ofobserving international agreements on this subject, whichdefine " baud " as a speed.

We have had the same sort of experience as the authorin regard to armoured wires. I received a report a fewweeks ago on the difference in the armoured wires of twocables laid side by side off the Shetlands. One was laidin 1885 and the other in 1902, but the condition of thearmouring on the 1885 cable is at present far better thanof that on the 1902. Similar effects are noticed on asingle cable. I was present at the picking-up of a60-year-old cable in the Irish Sea. There were hundredsof yards of beautiful galvanized armouring, and thensuddenly a patch from which the wires had nearlydisappeared. It is suggested in the paper that the effectis due to bacteriological action, which is an interestingpossibility.

I am interested in the author's description of the time-

Empire Telegraph Communications

Documents

Transcript of Empire Telegraph Communications