Conductivity cf some Dielectrics. 493 



The voltage scale was calibrated as follows : — 



( 1) The ballistic galvanometer G 3 was calibrated through- 

 out its range by means o£ a set of mica condensers charged 

 to pressures up to about 10 volts. The condensers, which 

 varied from 0'05 to 0'8 microfarad, were standardized against 

 each other, and the voltmeter used was standardized by 

 means of a potentiometer. G 3 was thus standardized for 

 quantity both with and without a 2000 ohm shunt, and the 

 ratio of the two sensitivities found. In both cases the scale 

 was almost uniform. 



(2) The steady deflexion of the dead-beat galvanometer 

 G 2 was compared with the discharge throw of G 3 for various 

 voltages .of the high-pressure system. The scale of G 2 was 

 thns found to be almost uniform for pressures above 

 700 volts. At lower pressures the deflexion per volt was 

 slightly smaller. The readings of the two galvanometers 

 were most "Consistent. As G 2 measures the average value of 

 the voltage, and G 3 the instantaneous value at the instant 

 of discharge, this shows that the fall in pressure between two 

 maxima of the A.C. supply is very small when no specimen 

 is connected. With a conducting specimen of course a 

 certain fall must occur. By far the most extreme case 

 occurred in the case of celluloid just previous to its failure. 

 Making an allowance for the current passing to earth 

 through the guard-ring, the total leakage current through 

 the specimen must have been about 70 to 80 microamperes, 

 which would reduce the potential difference of a condenser, 

 0*017 microfarad capacit} r , by from 80 to 90 volts in the 

 interval, 0'02 second, between maxima. As the average 

 pressure on this occasion was 1250 volts, the maximum 

 variation of pressure was about 7 per cent. In most cases 

 the variation must have been negligible. 



(3) The actual scale was found by replacing the induction 

 coil by the 200 volt A.C. mains, which provided an almost 

 steady " peak '' voltage of about 280, as read on a Kelvin 

 electrostatic voltmeter connected to the system. The cor- 

 responding deflexion of G 2 (about 16 scale-divisions) was 

 carefully noted, and hence, allowing for the small reduction 

 in sensitivity at low voltages, the voltage scale was found. 



It was thus found that for pressures above 700 volts the 

 sensitivity was 16*2 volts per scale-division, or with a 

 200 ohm shunt 40"0- volts per scale-division. 



It was found necessary to place an earthed metal screen 

 between the high-pressure system and the galvanometers to 

 reduce the effects of draughts of ionized air. A strip of 

 earthed tinfoil was also summed to both surfaces of the 



