MEASUREMENT OF CAPACITANCE 433 



be large enough to cause serious error in some of the values, one coil in 

 particular having an extremely large temperature variation for re- 

 sistances of this type. By determining the temperature coefficient of 

 the coils used and making appropriate temperature corrections it was 

 possible to reduce the uncertainty due to this cause to less than .001 

 per cent. The second source of error in resistance values resulted from 

 the effect of humidity on the coils. They were impregnated with shel- 

 lac, which absorbed moisture and swelled sufficiently to cause changes 

 in resistance with humidity large enough to efifect the results seriously. 

 By measuring the coils both before and after each series of tests it was 

 possible to keep the error due to this cause below .001 per cent. The 

 use of potted coils would undoubtedly eliminate this difficulty com- 

 pletely. The third resistance error present arose from uncertainties 

 as to the series resistance of the condensers Ci. These values were de- 

 termined in terms of the bridge air condensers by the step-up method, 

 on the assumption that the air condensers in the bridge have no con- 

 ductance, it being eliminated by the method of measurement and by 

 virtue of the special construction of the bridge. 



In the fourth place the accuracy with which our primary standards 

 of resistance are calibrated by the Bureau of Standards must be con- 

 sidered. These calibrations have been found consistent from year to 

 year to about ±.001 per cent., and accordingly can probably be relied 

 on to that value in the future. The results furnished by the Bureau 

 are based on their primary standards, which agree with the primary 

 standards of leading European nations to better than ±.002 per cent, 

 hence there is little likelihood of their changing their values by the 

 latter amount. 



4. Capacitance Errors 



The accuracy with which the ratio of any two capacitances may be 

 determined in such an investigation is dependent upon the consistency 

 of the bridge calibration by the step-up method, A detailed discussion 

 of the consistency of this calibration is beyond the scope of this paper. 

 As an indication of the order of the precision with which it is possible 

 to obtain a comparison between two condensers by this method, the 

 results of measurements on a standard condenser box on three different 

 bridges all calibrated by means of the step-up are shown in Table VI, 

 The value of ±Za, which is taken as the measure of the accuracy with 

 which measurements can be reproduced, is ±.004 per cent as deter- 

 mined from 54 individual measurements. On this basis the error in 

 the ratio of two condensers compared by this method will be less than 

 ±.0056 per cent (V2 X .004), Actually the error in the comparison 

 of the values of two condensers by measurement on such a calibrated 



