Effect of Electrical Convection. 37 



As before, a Voss machine and battery of leyden-jars were 

 used to charge the disks. Instead, however, o£ connecting 

 each pole of the Voss machine to the inside coats of three 

 jars, one pole of the Voss was earthed and the other con- 

 nected to the inside coating of all six jars, the outer coats of 

 which were earthed. The employment of this method 

 rendered possible the developing of a higher potential, but 

 the convection current produced by the moving charged disks 

 was no longer reversed, but simply made and broken by the 

 reverser. The reverser was run at such a speed that this 

 make and break occurred about ten times a second. 



The potential to which the disks were charged was measured 

 by a Thomson electrostatic voltmeter having a range of 

 - 1200 volts. This instrument was carefully calibrated by 

 comparison with the standard guard-ring electrometer at the 

 Physical Laboratory of the Johns Hopkins University, 



The following diagram (fig. 1, p. 38) will make clear the 

 general arrangement of the apparatus. 



The letters indicate the following: — D, the disks and 

 induced coil. M l5 motor driving the disks. X, countershaft. 

 0, commutator and reverser. M 3 , motor driving the same. 

 L, leyden-jars. V, Voss machine. M 2 , motor driving the 

 same. E, electrometer. T, resistance-boxes and switch in 

 test circuit. S, high potential switch. K, key in galvano- 

 meter circuit. G, galvanometer. S', scale for same. 



Connexions are omitted for the sake of clearness. They 

 were essentially the same as previously used except that a 

 separate circuit had to be employed to carry the test-current 

 through the new reverser. 



A number of sets of readings were now taken to determine 

 the relation between the deflexion d of the galvanometer pro- 

 duced by rapidly reversing a known current in the test-coil, 

 and commutating the galvanometer terminals connected with 

 the induced coil ; and the deflexion D, produced by charging 

 and discharging the disks at the same rate, the galvanometer 

 terminals being commutated as before. The deflexions 

 actually measured were 2d and 2D (see preceding paper). 

 In each set of readings five determinations of both 2d and 

 2D were made with the disks running in each direction, 

 2d and 2D were determined alternately. In the first five 

 sets of readings the potential was kept practically constant, 

 and the speed of the disks varied from 9'9 to 92*4 revolutions 

 a second, and in the next six sets the speed was kept prac- 

 tically constant, and the potential of the disks varied from 

 905 to 5900 volts. To show how closely the various quan- 

 tities could be determined the following table, giving all the 



