182 ABSTRACTS OF PAPERS

THE THEORY AND PRACTICE OF REGENERATIVE BRAKING OF D.C. LOCOMOTIVES,WITH PARTICULAR REFERENCE TO MULTIPLE-UNIT OPERATION

By O. H. HAHN, Ph.D., M.Sc.(Eng.), Graduate.

(ABSTRACT of a paper which was published in February, 1948, in Part II of the Journal.)

It can be shown that, if the speed of a locomotive whendescending a gradient is to remain constant, its regeneratingcharacteristic must be

where Eg — Voltage generated, volts.Ia -- Armature current, amp.7]0 — Overall efficiency.k3 — A constant.

This curve cannot be obtained completely in practice, becausethe maximum values of flux and current are limited. On theright-hand side of Fig. 1, the hypothetical voltage/current

If, amp

Hypothetical

7a,amp

Fig. 1 .—Development of stabilizing line.

characteristic is shown for a particular speed. The actual charac-teristic can be developed as shown by the broken line. Con-sidering the limitations already referred to, a working part ofthis curve will practically coincide with the hypothetical curveover which the braking power will remain constant. Variationsin the load lines are reflected by changes in the field currents.Stability is determined by the ratio of the field current to thearmature current, the points giving that ratio which yields con-stant speed forming what may be called the stabilizing line.

The fundamental circuits employed for obtaining the droopingcharacteristic are shown in Figs. 2 and 3. If the differentially-compounded exciter is worked with an unsaturated field, thecharacteristics of the two systems are similar, as previouslyshown by the author.*

In multiple-unit regeneration, good load-sharing is essential,as otherwise a driver, gauging the regeneration in the underloadedunit he may be driving, might notch up his controller further andoverload the following unit, causing a trip-out of the over-current releases, whereby the braking effect would be suddenlyremoved altogether and the train allowed to gather speed.

To obtain good load-sharing in multiple-unit regeneration, notonly must the magnetic characteristics of the motors be similar,

Dr. Hahn is with Siemens (South Africa) Ltd.* "Excitation Circuits for Regeneration in D.C. Locomotives," Railway Gazette,

l"43 3 p 2o

Fig. 2.—Separately-excited system with stabilizing resistor.

Fig. 3.—Separately excited system with differential-compound winding.

Traction motor flux A Exciters

A B

Settingrequired

Fig. 4.—Matching for load equality of two units, with motor andexciter characteristics differing slightly.

ABSTRACTS OF PAPERS 183

but their stabilizing lines must have the same slope. Slightdifferences in the characteristic can even be compensated bysuitably matching the stabilizing lines, which must, however,remain parallel.

The matching of two units, as well as the interdependence of

the various factors for equality of load, is shown in Fig. 4.Unit A is regarded as the standard and has a regulating circuitwith a stabilizing resistor as in Fig. 2, while Unit B has thedifferentially-compounded exciter of Fig. 3, and in addition hasa slightly different magnetic characteristic as shown.

IMPULSIVE INTERFERENCE IN AMPLITUDE-MODULATION RECEIVERS

By D. WEIGHTON, M.A.

(ABSTRACT of a Radio Section paper which was published in March, 1948, in Part III of the Journal.)

The paper discusses the theoretical limitations of impulse-noise suppression applied to amplitude-modulation receiversand is particularly concerned with the maximum signal/noiseratio at which suppression may be made effective.

It is shown that, with certain simplifying assumptions, theeffect of impulsive noise at the demodulator stage of a receivermay be written in the form of a carrier frequency equal to thatat the middle of the receiver band modulated with an impulsedefined by the Fourier transform of the low-pass equivalent ofthe receiver band-pass filters. From this expression it followsthat the average value of the noise is independent of bandwidth,that the peak value is proportional to bandwidth, and that inthe presence of a signal the noise is of random amplitude andsense. It is shown that, if the bandwidth of the r.f. and i.f.stages of the receiver is great compared with that of the I.f.section, then the noise output is independent of r.f. bandwidthif the receiver is linear throughout. With a non-linear elementhaving a rising characteristic (e.g. square-law detector) the noiseoutput is increased by increasing the r.f. bandwidth, and with anon-linear element having a falling characteristic (e.g. noiselimiter) the noise output is reduced by increasing the r.f. band-width. The latter feature forms the basis of most circuits forthe suppression of impulsive noise.

Since any noise limiter must have a linear response over therange of amplitude of the signal in order to avoid distortion,there exists a threshold signal/noise ratio, above which thelimiter is not effective. The behaviour of the receiver at noiselevels greater than threshold depends upon the particular circuitemployed for limiting, and the paper is not concerned withbehaviour in this region. The threshold conditions may, how-ever, represent the worst case that can occur, and for this reasonthe signal/noise ratio at the output terminals of the receiver

Mr. Weighton is with Pyc, Ltd.

under threshold conditions is of considerable interest. It repre-sents a theoretical limitation to the performance of each classof noise suppressor and is independent of circuit details.

Noise suppressor circuits may be classified generally asamplitude limiters, in which the impulse is distinguished fromthe signal simply by its greater amplitude at a particular stageof the receiver; differential limiters, in which the rate of rise ofthe leading edge of the impulse is the distinguishing feature; anddelay limiters, where the circuit operates on the shorter totalduration of the pulse.

The threshold signal/noise ratio is calculated for each typewith the following results:

Circuit class ificat on

Amplitude limiterDifferential limiterDelay limiter

Threshold signal/noiseratio

BXIB20-9(Bi/B2)2015(Bi/B2)2

where Bx is the bandwidth of the amplifier preceding the limiter,and B2 the bandwidth of the following amplifier, reckoned ashalf bandwidth in the case of r.f. and i.f. stages. For differentiallimiters of higher order the threshold signal/noise ratio is pro-portional to the appropriate power of the bandwidth ratio.

Some experiments were carried out with the object of checkingthe formulae for amplitude and differential limiters, and theresults are found to be in fair agreement.

A comparison is made between suppressed amplitude modu-lation, frequency modulation and pulse-width modulation, inrespect of impulsive noise. The results are presented in theform of curves showing the effect of noise level and band-width ratio on the signal/noise ratio obtainable in each case.