March, lUlS] A Lecher System — Experimental 



161 



Remembering that the thing that is experimentally deter- 

 mined is Xs the percents of error in column 25 are remarkabl}^ 

 small. x\veraging for all the sets we may say approximately 

 that one per cent, variation in Xs with its consequent variation in 

 /s according to the formula 



= />— I 15 



L = p 



O + Cf— a) 



where .v is the frequency number 1, 3, 5, 7, etc., and a the 

 corresponding natural number 1, 2, 3, 4, etc., changes the 

 value of Ko/ K for values of s^\, 3, 5, 7, 9, 11, respectively 

 8, 11, 15, 22, 32, 44 per cent. This serves again to emphasize 

 the importance of an accurate measurement of Xg. 



Figure 9. 



We have gone into the method of these calculations with 

 some detail because of the dependence placed upon the figures 

 in the theoretical paper by one of us that follows this paper. 



Two things stand out from this discussion of Table I. 

 The value of Ko/k is very constant for small values of y\ and a 

 small error in the wave length determination makes a large 

 error in the constancy of Ko//c, the errors being larger propor- 

 tionately for the smaller wave-lengths. The first point was 

 made by Blake and Sheard and we thus confirm it. For the 

 first three sets of the table the average error in X^ is only 0.13 per 

 cent. 



