A NEW CURRENT \VKIGHER, ETC. 525 



An idea of the effect of the convection currents of air rising from the fixed and 

 suspended coils was obtained from observations on the balance pointer when the 

 forces acting on the suspended systems were in opposition. In such a case small 

 variations in current strength have no measurable effect on the total force. With the 

 Ixilance case covered with blankets and practically uniform radiation in all directions 

 (the observations were made at midnight), the mean doubled rest- point of the balance 

 pointer was deduced from 108 readings as 20(5 '7. These readings were taken in three 

 sets. The first set of 36 readings gave 206 '2 as the rest-point ; the second set were 

 taken immediately after the first and gave 205*9 ; there was an interval of half an hour 

 between the second and third sets, the mean of the latter being 208 '0. The average 

 ditfrieiice l>etween the first 36 readings and 206'2 is 0'8, so that the mean of a few 

 iv.ulings is associated with a large probable error. In addition there was difficulty in 

 maintaining a very steady current through the heated coils ; the rest-point of the 

 balance was subject to drift ; and the difference of temperature between the coils and 

 marble and between fixed and suspended coils introduces serious difficulties in the 

 calculation of the mutual induction. 



The rest-point of the balance is very constant when no current is flowing through 

 the coils and lias not passed for some hours previously ; it is also very constant for 

 the first 20 minutes after the circuit has been completed. The resistance of the coils 

 increases considerably in this period, but observations proved that a current constant 

 to 2 iii 1,000,000 and often to I in 1,000,000 could be maintained for four minutes, in 

 which interval the resistance of the balance coils increases about 0'12 ohm, and the 

 sliding contact of the mercury trough passes from the most to the least favourable 

 position for adjustment. In this interval three readings of the balance pointer could 

 always be taken, and experience has shown such readings to be remarkably accurate. 

 This method was adopted. 



(D) Electrostatic Effect*. Electrostatic effects of sufficient magnitude to produce 

 a readable deflection of the balance }>ointer were not antici]>ated. The mean electro- 

 static potentials of the various pairs of coils are, of course, different, hut the maximum 

 variation between any part of one suspended and any part of one fixed system is less 

 than 36 volts when a current of 1 ampere is flowing. A test was made by connecting 

 the upper coils of one fixed system and the lower coils of the other fixed system to 

 the + pole of a lottery of 110 volts; the other coils of the balance were connected 

 to the pole and to earth. No difference was observed in the rest-point of the 

 Ki lance, and hence there could be no disturbing effect due to electrostatic attraction 

 ltween the fixed and suspended coils. When, however, the balance coils were placed 

 in series and a current of 1 amj)ere passed through them, a difference in the rest-point 

 of 07 scale-division was always observed on reversing the current ; this was found to 

 be a measure of the difference of the electrostatic forces between the suspended coils and 

 the metal guard-discs d, fig. 6 (Plate 8), underneath them. The difference of mean 

 potential of the coils on the suspended systems is 62 volts ; the metal rings are about 



