EXPERIMENTAL STT'DY OF HEAT LEAKAGE. 



761 



diflferential thermometer involved, two were made necessary by 

 reason of accidents necessitating repairs to the thermometers which 

 involved changes in their constants. Each of the values of R^i 

 (a single one excepted) by interchange 

 is the mean of from three to five deter- 

 minations with the same set-up , of appa- 

 ratus, but with varying flow rates. The 

 greatest average deviation involved in 

 any of these five sets of interchange values 

 of /?67 was 0.00018 ohm, (0°.009 C); the 

 least, 0.00007 ohm (0°.0035 C); the mean, 

 0.00010 ohm (0°.005 C). Thus the inter- 

 change method gives results which are 

 more consistent among themselves than 

 they are with the results obtained by 

 calibration. It cannot be supposed, how- 

 ever, that this signifies a real difference 

 of the indicated magnitude in the values 

 of the normal resistance by the two 

 methods, for the reason that uncontrol- 

 lable fluctuations in this resistance be- 

 tween calibrations are at least as great 

 as the discrepancy between its values 

 as obtained by interchange and by 

 calibration. The size of these fluctua- 

 tions is shown in Fig. 10, which exhibits 

 the behavior of Rqi at 0° C. over a period 

 of eighteen months. It is partly because 

 it is not always convenient to calibrate 

 the thermometer immediately before or 

 after (or both before and after) a set of 

 runs that the interchange method has 

 been used to eliminate the normal resist- 

 ance ; partly also because, by this method, 

 the thing eliminated is the normal re- 

 sistance under actual conditions of use. 

 This last reason was of more weight 



before experience had proven that the use of the differential ther- 

 mometer under pressure did not affect its normal resistance, as 

 conceivably might have been the case: in calibration, of course, the 

 normal resistance is determined at atmospheric pressure. 







O 



o3 

 O 



d 



c3 





a 



P 

 o 





