100 ABSTRACTS OF PAPERS

speech sounds like chanting. These results are as good as couldbe expected on the basis of the theory previously presented, butthey prove that the principle must be modified before the systemcan be applied to telephony to give reproductions which arenatural as well as intelligible.

A method of achieving the necessary improvements is described.It is shown that the two annoying factors, roughness andmonotony, which are mutually exclusive, arise from the samefundamental cause. The frequency convertor in its present formhas a set of "preferred frequencies" which are multiples of afundamental, fixed by the data of the device. These are theonly frequencies which are correctly reproduced. A modifiedsystem of frequency compression is proposed, in which the dataof the device are continuously controlled by the pitch of thespeaking voice, so that the pitch and its overtones are always"preferred frequencies" and are reproduced without distortion.It is expected that this modification will assure perfect transmis-sion of the vowels and of the individual characteristics of the

speaker. Devices are suggested, rather different from thepresent experimental frequency convertor, in which the adjust-ment to the pitch of the speaker's voice is carried out rapidly,continuously and automatically. It is shown that for malevoices this perfected method reaches its limit at about eightfoldcompression, and a frequency band of about 400 c/s is probablythe narrowest which allows the transmission of intelligible andnatural speech. Female voices may allow twelvefold compres-sion without appreciable loss in quality.

This system, which is based on the tolerances of hearing andon the regularities in human voice production, promises eco-nomies up to about 85 % in bandwidth. At least in principle itappears possible to cut the waveband down to one-half again,by making use of the insensitivity of the ear to the phase ofacoustical oscillations. But it is shown by theoretical considera-tions that "phase utilization," though possible in principle, isnot practicable as it leads to apparatus of extraordinary com-plexity and delicacy.

AN IMPROVEMENT IN END-FIRE ARRAYSBy F. K. GOWARD, B.A.*

(ABSTRACT of a Radio Section paper which was published in November, 1947, in Part III of the Journal.)

A conventional end-fire aerial, n\ long, has a uniform distribu-tion of amplitude along it, and a total phase change of 2im, thephase changing linearly with distance. Thus the radiation fromall points along the array adds in phase in the end-on direction.Hansen and Woodyard have shown that a conventional arraymay be improved by having a total linear phase change of(2« + 1)TT, still keeping the amplitude distribution constant; themain beam is then narrowed, but the side-lobe amplitude in-creases. Still further improvement may be effected by super-

• Telecommunications Research Establishment.

posing on a Hansen and Woodyard array a further arraywith a smaller uniform-amplitude distribution and a linearphase change of {In + 3)7r; the side lobes of this latter arraylargely cancel those of the first array. The net result is anarray giving a main beam narrower than that of a Hansenand Woodyard array and yet having side lobes smaller thanthose of a conventional array. This improved array has aphase change of (In + 3)TT along it, but the change is notlinear with distance, the amplitude distribution being taperedtowards the ends.

NEUTRAL EARTHING OF THREE-PHASE SYSTEMS, WITH PARTICULARREFERENCE TO LARGE POWER STATIONS

By J. R. MORTLOCK, B.Sc(Eng.)> Member,* and C. M. DOBSON.*

(ABSTRACT of a paper which was published in December, 1947, in Part II of the Journal.)

Controversy still exists as to the best method of earthing theneutrals of power systems, and before a decision can be madein any particular case it is desirable that the advantages and dis-advantages of the various possible methods should be appreciated.Economically the complete adoption of a specific arrangementmay not be possible; nevertheless the risk entailed in the de-parture should be appreciated and assessed. The leadingfeatures of the various methods are summarized in Table 1 andconsidered individually below.

For a simple system such as that shown in Fig. 2, an apprecia-tion of the phenomena accompanying a particular method ofearthing the neutral is most easily obtained from the sequence

* British Thomson-Houston Co., Ltd.

networks, Fig. 1. The switch S, Fig. 1, is open under normalconditions and closed for the particular fault; it is evident thatonly single-phase to earth and 2-phase to earth faults are ofinterest as only in these cases is the zero-sequence network in-volved, and, in particular, attention can be confined to the pointsGo, representing the earth, and No, representing the neutral;moreover, in the case of 2-phase to earth faults the negative-sequence network parallels the zero-sequence network and con-sequently acts as a discharge path for any charges trapped on Co.

SYSTEM WITH NEUTRAL INSULATEDThe major disadvantage of leaving the neutral insulated is that

arcing grounds can occur with a considerable build in voltage;