468 Mr. W. E. Sumpner on the Variation 



amperes before each observation, and then increased to the 

 particular value of A at which the value of L was observed. 



The annexed table indicates the results obtained by the 

 swing method : — 



A. 



*v 



L c . 







•243 



•0290 



2 



•230 



•0261 



4 



•083 



•0103 



5 





•0177 



6 



•016 



•0155 



8 



•013 



•0120 



10 



•012 



•0097 



The values obtained with diminishing currents (*. e. for the 

 curve P'A') were slightly lower. They were arrived at by 

 increasing the current to + 10 amperes and then diminish- 

 ing it to the value of A, at which L p was required. 



The values of L c were much easier to obtain than those of 

 1j p , for the slightest variation in the main magnetizing 

 current A was sufficient to render it impossible to obtain 

 L p until the previous magnetizing processes were gone 

 through again. Accumulators were used to provide a con- 

 stant E.M.F. ; but it was nevertheless found difficult to keep 

 A sufficiently constant, so that the test-swings for L p were 

 taken as quickly as possible to prevent values intermediate 

 between L p and L c being obtained. These considerations 

 partially account for the unsatisfactory nature of the results 

 obtained with the Ferranti field-magnets ; for, at the time 

 those experiments were made, the necessity of keeping the 

 current quite steady until the swing measuring L was ob- 

 served was not realized. The numbers given do not show the 

 slight initial increase in the value of L p for small increasing 

 currents, but more recent experiments have done so. 



The values of L e were also determined by means of the 

 secohmmeter of Professors Ayrton and Perry. This instru- 

 ment necessarily only measures coefficients for cycles of 

 magnetism. The value of A was varied from zero to 10 am- 

 peres positive, thence to 10 amperes negative, and thence to 

 zero again. The value of each coefficient was determined for 

 three different speeds, and the mean taken. The value of the 

 coefficient appeared to diminish slightly as the speed increased. 

 (This effect has been observed in several cases when the mag- 

 netic circuit is wholly or partially composed of iron, and is 

 probably due to magnetic lag.) The coefficients L c obtained 

 for a complete cycle of current values are given in the four 



I 



