Mr. R. Appleyard on Dielectrics. 153 



sively as R T and ~R e in (1), we find as a mean value 



log« = I'96344 (2) 



The temperature 6 may be given some standard value, 

 say 20° C; and from a curve drawn from the fifteen observed 

 values of resistance and temperature the corresponding re- 

 sistance R 20 maybe obtained. The resistance at 20° C, found 

 in this way, was, for paraffin-paper, 



R 20 =3670 megohms; 



equation (1) in this case becomes 



log E T = 3*56467 + (r-20) (1-96344), . . (3) 



from which may be deduced the resistance corresponding to 

 any temperature t. It is interesting to see how nearly the 

 results calculated from (3) agree with the observed values, as 

 can be done by drawing the two curves which represent the 

 observed and calculated resistances respectively. 



The agreement between these two curves is so close, consi- 

 dering the conventions which always have to be adopted in 

 tests upon dielectrics, that it has been suggested to me that a 

 simple logarithmic curve may perhaps represent the true state 

 of affairs, the observed results being assumed more or less at 

 fault. This seems improbable for the following reason: — In 

 equation (1) a is really the ratio of two resistances at an 

 interval of 1° C, and it is presumed* that this ratio, i. e. 



Br 



-ti T +l 



is constant for all values of t. This might be true for some 

 ideal dielectric whose structure was perfectly definite at all 

 temperatures, and which made invariable contact with its 

 electrodes under all conditions ; but there is no reason why a, 

 should be constant for a substance having the instability of wax. 



Temperature Tables. 

 One use of these results is to enable the resistances of 

 condensers to be reduced to the resistance at some standard 

 temperature. A similar correction has always to be applied 

 for gutta-percha and india-rubber cores. It is usual to 

 draw up a table, assuming a constant over a given range, and 

 to supply a correcting logarithm for each degree of tempe- 

 rature. This may be sufficiently accurate for small ranges of 

 temperature ; but I would suggest that a better way is to work 

 from the observed results directly, making no assumptions 

 * See Appendix II. 



