ON STANDARDS OF ELECTRICAL RESISTANCE. 485 



least ; but successive discharges were occasionally found to differ by as much 

 as two divisions, though this amount of discrepancy was rare. It was due 

 partly to the residual effect of former charges in the condenser, though gieat 

 care was taken to avoid this, partly, it is believed, to slight changes in the 

 electromotive force of the battery (which was not in very good order, the 

 discharges being generally less toward the end of a set of experiments), and 

 partly to sHght motion of the needle at the moment of taking the discharge. 

 This last source of error made it impossible to make the observations in Lou- 

 don ; even in the country the needle was seldom, if ever, absolutely stUl, 

 though the oscillations were generally less than one division. The variation 

 of the electromotive force and resistance of the battery when taking a per- 

 manent deflection was another source of error. Owing to the great inertia 

 of the swinging parts, no observation could be taken until the cm-rent had 

 been flowing for at least a minute, and often more ; and, especially when small 

 resistances were used, the deflections visibly diminished with time. Owing to 

 all these causes, I do not depend on the results obtained as certainlj- accurate 

 within less than one per cent. This is the less to be regretted, as the capacity 

 of a mica condenser is very ill defined within wide limits, owing to absorption. 

 The condenser used consisted of 38 plates of mica, about 0-003 in. thick, and 

 having a circular piece of tinfoil 3 in. in diameter cemented to each side of the 

 mica, -ndth a piece of each tinfoil projecting beyond the mica so as to join all 

 the upper tinfoils and all the lower tinfoils together, and form the inner and 

 outer armature of the condensers. This plan has for some time been practised 

 by Mr. Latimer Clark, and makes a very constant and well-insulated condenser, 

 extremely easy to adjust roughly by altering the number of the mica plates, and 

 for small corrections by cutting away portions of the tinfoil from the top plate. 

 Mica, like all other solid dielectrics with which I am acquainted, apparently 

 absorbs electricity to a A"ery large extent, and continues to do so for a long- 

 time, discharging it at first rapidly, but at the last very slowly indeed, so that 

 a complete discharge is not effected for hours. The total capacity of the con- 

 denser varies therefore as the time varies during which it is charged, and the 

 apparent discharge varies with the time during which we measure it; for 

 instance, if we merely observe the discharge due to a momentary contact, 

 "we shall obtain a different result from that given when we maintain the 

 contact all the time the needle is swinging ; the result will also vary in the 

 latter case with the time of oscillation of the galvanometer needle. If the 

 needle oscillates slowly, it -will be acted upon by a greater qiiantity of elec- 

 tricity than if oscillating rapidly. Thus, in one experiment, the deflection, 

 when the discharging contact was permanently maintained, was 1C6 divisions, 

 when a momentary contact Avas made by a blow it was only 156°. AVhen 

 the contact was made for about 1-7 second the deflection was 161, and when 

 the contact was maintained for 3*4 seconds the deflection was 164 ; the 

 maximum deflection of 166 was reached after 5 seconds : these experiments 

 show that when the needle had travelled two-thirds of its maximum 

 distance, the current being- discharged exercised a very sensible influence on 

 the deflection. The ballistic formula is therefore not strictly applicable to a 

 case of this kind, and a different result would be obtained with a galvano- 

 meter oscillating either more or less quickly than the one I used. It seemed 

 therefore unnecessary to take great precautions or to aim at any high degree 

 of acciu-acy ; and my object has simply been to provide a unit for cable-testing 

 which shall be approximately equal to the ideal standard chosen by the Com- 

 mittee, and wliich can be used with at least as great accuracy as those copies 

 of knots of Atlantic or Persian Gulf cables hitherto used. 



