154 BEIDGMAN. 



range of frequencies, with particular care at the three highest fre- 

 quencies, as error here will affect the extrapolation. The check is not 

 as good as that above because of the greater error of the A.C. readings, 

 but the requirements were met within the errors of reading, which 

 were 0.5 mm. The A.C. intensity was varied over a range of 10 or 

 20 fold, and its intensity determined by means of a thermo-couple. 

 At the greater intensities of this range, the difference. between A.C. arid 

 D.C. settings might be of the order of 1 cm. 



For the actual measurements the A.C. was made as large as possible 

 in conformity with these requirements, in order that the sensitiveness 

 might be as high as possible. The requisite intensity varied somewhat 

 at the different frequencies, (there was resonance at about 1000 cycles 

 in the telephone) but as a rough average was about 0.02 amp. This 

 is of the order of l/30 of the D.C. current near its maximum, so that 

 the COS" term, which must be negligible, is of the order of 5 X 10~^ 

 of the steady heating term. The D.C. shift of zero near the maximum 

 was of the order of 15 cm. of bridge wire. 5 X 10~* X 15 is 0.075 mm., 

 which is thus beyond the possibility of detection or error. 



Equality of A.C. and D.C. Zeroes. If the bridge is properly set up 

 and free from capacity and inductive effects, the D.C. zero for small 

 direct currents and the A.C. zero (that is, the A.C. setting in the 

 absence of D.C.) should be the same. This condition was satisfied 

 within the limits of reading at the three highest frequencies. The 

 settings could be made and read to 0.1 mm. for both A.C. and D.C. 

 It was only when the A.C. settings were made in the presence of D.C. 

 that there was room for as much uncertainty as 0.5 mm. At the lower 

 frequencies the two zeroes might sometimes differ by as much as sev- 

 eral tenths of a mm., but this does not affect the extrapolation or the 

 final results. 



Indepcndejice of Rate of Cooling. If the extrapolated difference 

 between A.C. and D.C. readings is due to a real departure from Ohm's 

 law, and is not in some way connected with the microphone action, 

 which depends on the heating, then for a fixed D.C. density the extra- 

 polated difference of readings must be independent of the steady rise 

 of temperature, which may be made to vary by changing the speed 

 of the cooling water. That this condition is satisfied is shown in 

 Figure 5, in which the difference between D.C. and A.C. settings is 

 plotted against the reciprocal of the frequency. The D.C. heating 

 current was the same for the two curves, but the rate of flow of the 

 cooling water was different, the steady rise of temperature of the 

 upper curve being 1.4 times as great as that of the lower. 



