[shaw] THE KELVIN-VARLEY SLIDE POTENTIOMETER 



99 



One set of values obtained were d= +168, d' = + 351 and d" = -246 



and hence -~ = .99887. Repetitions of this measurement, taken at 



different times during three months, showed the following constancy : 



. 99885 

 .99887 

 .99888 

 .99887 

 . 99890 

 .99885 

 .99887 



Mean .99887 



The small variations in this case are almost entirely due to the 

 contact resistance between the stud and the revolving contact maker. 

 It is very necessary to note that the spring is making proper contact 

 all round before using the instrument for accurate work. This is 

 easily ascertained by applying slight pressure on the contact springs 

 when the instrument is connected in the circuit for taking the main 

 observations. If there is any appreciable change it will at once be 

 shown by the galvanometer, but this error can be eliminated by 

 carefully polishing the tips of the studs with chamois and by slightly 

 tightening the spring. A variation as large as two parts in 10,000 has 

 been found, which was entirely due to this cause. 



The constant k must be determined from time to time during the 

 calibration in order to note any slight unequal temperature changes 

 in the circuit. This is rapidly done by obtaining a balance and then 

 changing P by one ohm, k being given by the reciprocal of the differ- 

 ence of the two deflections. 



The successive observations of the hundred deflections for equa- 

 tions (8) are taken as indicated above, but in the course of observation 

 a slight further error is introduced. A small current must necessarily 

 be flowing through the vernier coils and the auxiliary arms continu- 

 ously while the successive coils of the main slide only receive it tem- 

 porarily. With the application of two volts, it was found that there 

 was sufficient heating effect to alter appreciably the effective value 



P 



of V -y during the calibration. A simple method of correcting for 



this error can be obtained in such a manner as the following. The 

 first ten readings are taken and then the first one is repeated. Any 

 slight change can readilv be applied as a correction to reduce all 



P 



subsequent readings to those that would have been obtained with F -pl- 

 under the initial conditions. This would again be checked at 20, 30, 

 40, etc., and the extra corrections would accumulate as the calibration 

 proceeded. The choice of the gap between the repetition observations 

 can be greater or less than ten according to the magnitude of the 



