FORCE OF THE WESTON NORMAL CELL IN SEMI-ABSOLUTE VOLTS. 171 
Hence substituting these values we get 
-4 = +1215122 
-4 G 5 ^ 5 P 5 =+ 1215 
— 4 G 7 gr 7 P' 7 = - 57 
and the total = +1216280 
YI. The Bifilar Suspensions and the Determination of their 
Directive Force. 
(a) The Suspensions and the Theory of the Method. 
The bifilar suspensions were made of hard drawn copper wire, diameter 0'0450 cm., 
length 80 cm., and were rigidly clamped 3T cm. apart. They had been hanging for 
over ten years under the tension of the suspended coils arranged as in actual use, and 
it was thought that the wires would be exceptionally satisfactory in their behaviour. 
An account of the influence of elastic changes in the part of the controlling couple, 
due to torsion, is presented in the section on the deflection measurements. It should 
be noted here that the examination and treatment of this problem is of fundamental 
importance to the whole investigation. A full discussion of the matter is given by 
the writer in a paper on “ Increased Accuracy in the Use of Bifilar Suspensions ” in 
the ‘ Philosophical Magazine ’ for September, 1911. 
The suspended system was placed with its centre coincident with that of the fixed 
coils, and the wires were adjusted symmetrically for use according to the “ method of 
tilting” described by Gray.* The position of the coils could be easily checked to 
within the desired limits of accuracy by simple optical methods. Each wire was 
finally clamped at both the top and the bottom. 
If the controlling couple due to the bifilar suspensions is taken as B sin 6 for a 
deflection 6 then it can be shown that 
tj _ Wct 2 2pd i 6 (q) 
4(1—x—z) l sin 6 
for symmetrical suspensions, where W is the weight of suspended system, “ a ” is the 
distance between the bifilars, “ l” is their length, “ x” is the constant correction for 
the effect of rigidity at the ends, “z” is the distance through which the suspended 
system is raised for a deflection 6, “ d” is the diameter of the wire, and p is Coulomb’s 
torsional coefficient. 
Under the conditions of our investigation it was found that an error of less than 
0 
5 parts in 1,000,000 in B would be introduced by taking z = 0 and ——- =1. If we 
J 6 sm 9 
* Gray, ‘Absolute Measurements in Electricity and Magnetism,’ vol. 1, p. 242. 
z 2 
