MEASUREMENT OF CAPACITANCE 431 



were all so selected as to have negligible temperature coefficients over 

 this range no temperature correction in the capacitance values nor in 

 the value of K was necessary. 



A well aged 3 dial condenser box which had just been calibrated at 

 the Bureau of Standards was checked on the bridge at the time of the 

 second series of tests. In this way a comparison of the true capaci- 

 tance of the bridge condensers as determined by the Bureau of Stand- 

 ards method and as determined by the present method is afforded. 

 This comparison is shown in Table V. 



TABLE V 



Comparison of Accuracy of Primary Standards by Determination of K and 

 BY Bureau of Standards Calibration 



K for Bridge at 1000 K for Bridge at 1000 

 cycles by method of cycles by comparison 

 this paper: — ■ 2/27/26 with Bureau of Stand- 

 to 3/19/26 ards: 3/1/26 

 Average (27 determina- 

 tions) 1.00034 (18 values) . . 1.00038 



<r 000047 .00005 



3<7 00014 .00015 



^ 000027 .000035 



K for Bridge at 2000 cycles 



by method of this paper: — 



3/19/26 



Average (14 determinations 1.00032 



a 000035 



3(r 00010 



^ 000027 



Note: K by method of this paper was determined by the following bridge con- 

 densers .1, .2, .3, .4, .04, .06, .07, .08, .09. K by comparison with the Bureau of Stand- 

 ards values on condenser box No. 26962 was determined by all the settings of the .01 

 and .1 dials of the bridge. 



Discussion of Results 



In the discussion which follows an attempt will be made to point out 

 the various sources of error which may creep into such a determination 

 and to state briefly what precautions were taken to guard against them. 



1. Frequency Errors 



The primary error in the values of frequency used was to be found 

 in a variation of the standard fork from its nominal value. The 

 average frequency of this fork integrated over 24 hours against the 

 Arlington time signals can be held constant to ±.001 per cent. While 

 this tells us little about the fluctuations from the mean value over 

 short periods it is at least reasonable to assume that they will be of the 

 same general order as the variation in the average. As a check on this 



