TRANSACTIONS OF SECTION G. . 759 



but also on the inertia of the fly-wheel and the load on the engine. A Watt 

 governor acting directly on the throttle or expansion valve does not hunt if designed 

 for stability, but does not give the same speed at heavy as at light loads. If made 

 isochronous it has either a dash-pot or a relay. In either case the action of the 

 governor is not instantaneous, and a time-lag is introduced. This causes hunting, 

 especially if the load is light and the fly-wheel heavy or fast-running. If the iso- 

 chronous governor works a slow-acting relay the hunting may be so serious that 

 the steam supply alternates between complete cut-off and full supply. The remedy 

 is to use either an isochronous governor with no relay, so that the engine gets either 

 full steam or none, or to use a stable governor acting direct with a supplementary 

 relay gear to secure the same speed at full and light loads. The first method is 

 not employed in large slow-running engines, as the fly-wheel would have to be 

 enormous. The second is utilised in the Knowles supplementary governor, but a 

 second governor is unnecessary, as one can fill both functions. 



4. On Engineering Laboratory Instruments and their Calibration. By 

 David S. Capper, M.A., Professor of Mechanical Engineering, King's 

 College, London. 



The reliance to be placed upon observations made with measuring instruments 

 evidently depends primarily upon the accuracy with which those instruments 

 record. Neglect of this fundamental truth often leads to inaccurate and erroneous 

 deductions from experiments which are themselves of the highest scientific value ; 

 not infrequently the whole value of observations may be destroyed by insufficient 

 care in the calibration of the instruments used. The subject is therefore one of 

 some importance. The author describes the chief sources of error in some of the 

 most common engineering investigations, and their probable value, and points out 

 some of the possible methods of correction where such exist. For example, in 

 engine trials there are many possible sources of error. Most of these may be reduced 

 in percentage value by continuing the trial for a sufficient period. But this is not the 

 case with errors which may occur in the indicators, gauges, or spring balances used 

 in the determination of power. In these, unless properly calibrated before trial, 

 very serious errors may be introduced, amounting in some cases to 5 and 6 per 

 cent, of the total power indicated. It is therefore absurd, even if proper precau- 

 tions have been taken, to rely upon horse-power measurements to two places of 

 decimals. 



Similarly with regard to tension and compression experiments with standard 

 10-inch bars. Here calibration of the testing machine is extremely difficult, and 

 can in general only be carried out over a small portion of the range of the experi- 

 ments. Deductions have therefore to be made from the less to the greater, with 

 the result that small errors in the calibration will tend to be magnified. Vertical 

 testing machines have fewer sources of error, and can be calibrated with more 

 certainty, than horizontal machines. Extensometers are, however, much more 

 easily applied to a horizontal bar than a vertical, and variable jockey weights, 

 which are requisite if the same accuracy is to be maintained at low loads as at 

 high, are also more readily adapted to horizontal machines. 



Extensometers can be made and calibrated well up to the accuracy of the 

 testing machine. With standard bars and a measuring instrument true to the ten- 

 thousandth of an inch, the modulus can be relied upon to the second significant 

 figure. It is doubtful if more can be obtained without very special construction 

 and calibration of the testing machine. 



The difficulty in bending experiments, again, lies in the accurate application of 

 load. Unless the beams are very short or of unmanageable cross-sections, the 

 load measurement must be very delicate if readings approaching the accuracy of 

 those in tension are to be obtained. It is possible that some of the discrepancies 

 in published beam experiments may be due to this cause. 



The paper deals shortly with other cases where calibration is specially needed, 

 to -which the limits of this abstract do not permit a more extensive allusion. 



