275 
of Metal Rods with rounded ends. 
To determine T we need a further reading, Q" , with a greater 
value of C 2 R 2 . We then get Q" = Q — Spark + y . 2 , so that 
Q'-Q"=yi -Vi or Vi = VC 2 + Q" - Q' (5). 
We then have, from (3), 
which gives the absolute value of the time free from all corrections. 
It would not be convenient to repeat this process for every 
experiment ; but it was carried through with the shortest steel 
rods, and, by comparing the result with that got by the normal 
method, the necessary corrections for the latter were obtained. 
To determine the best value of C 2 R 2 to use in this method, we 
may proceed as follows : — 
T=-G 2 R 2 \og e (l-y/VG 2 ), 
— — — C R x ^ x _ ^ 
.. dy - m y JVG t ) VGa > 
7} _ T 
or dT=*pe c ^dy, 
so that, for a given error in y, the error in T is least when 
T 
R e c 2 r 2 a minimum. 
Differentiating, with regard to R», we get the condition 
T 
c, it, 
-f Ro x — 
T 
(XR.r 
e GiRi = 0 or 
T 
C.R, 
- 1. 
For the case considered, T was about 1‘5 x 10~ 4 secs. R. 2 was 
therefore taken as 8000 ohms, and C 2 as '05 x 10 -fi farads. In 
determining the spark effect, R n was reduced to 300 ohms, so that 
T 
e C * R * = e~ 10 about, and is, consequently, quite negligible com- 
pared with unity. The figures obtained were as follows : — 
1(a) Normal experiment 1'002 ... Q 
Impact tests j (b) With condenser, R 2 = 3000 ohms 2'062 ... Q" 
((c) With condenser, R 2 = 300 ohms... 2'674 ... (/ 
(d) Charging condenser direct l'G76 ... VC» 
Each of the above figures represents the ratio of the galvanometer 
fling to the voltage, and is the mean of several observations, which 
are given in full in the tables. In the first place it is seen that 
(a) -I- (d) = 2'678, while (c) = 2'67I. From this we conclude that 
there is no sparking effect whatever, such a small difference being 
