CONTINUOUS ELECTRIC CALORIMETRY. 
83 
deflection of 50 centims. on either side of the zero. This method of reading 
eliminates a number of small uncertainties, and makes it possible to approach an 
order of accuracy of 1 part in 100,000 in the deflection measurements. 
(12.) The Bifilar Suspension. 
In order to minimise the effect of imperfect elasticity, the wires of the suspension 
were made of hard drawn copper of high conductivity, as fine as was consistent with 
constancy of length (safety factor 5), and were rigidly. clamped at a distance of 
3 "2 centims. apart, after the tensions had been adjusted to equality. The length 
of the suspension was 80 centims., and the effect of torsion of the wires was 
comparatively small. The directive force of the bifilar depended almost entirely on 
gravity, and remained constant to 1 in 20,000 for several months. The effect of 
current heating of the suspension was tested by observing the time of oscillation 
with the working current of '5 ampere flowing in opposite directions in the two 
windings of the small coils, and was found to be less than 1 part in 20,000. 
To determine the moment of inertia, the small coils were fitted with a co-axial 
brass tube about 2*5 centims. in diameter, and 50 centims. long, containing a pair of 
cylindrical inertia weights, 150 grammes each, which could be clamped at the centre 
of the tube, or at the ends. The distance between the weights in their extreme 
position is the most important measurement. This was determined by a very 
accurate pair of steel callipers by Brown and Sharpe, reading to ‘01 millim., with 
a range of 50 centims. This gave sufficient accuracy for the preliminary measure¬ 
ments, but it was intended to employ a Whitworth measuring machine for the final 
series. In observing the oscillations, the times of passage were recorded on an electric 
chronograph with a standard clock rated from the observatory. The periods o 
oscillation, with the ends of the weights flush with the ends of the tube, and with 
the weights in contact at the centre of the tube, were 11'5385 and 6'7857 seconds 
respectively, and the observations on different days, when corrected for temperature, 
did not vary by more than two or three in the last figure. 
(13.) Mean Radius of the Large Coils. 
The mean radius of the pair of large coils was determined from the length of the 
copper tape with which they were wound. This method is not satisfactory with soft 
annealed copper wire owing to stretching, but the hard rolled copper tape could be 
wound without any tension, and did not undergo any change of length. This was 
verified by graduating the tape itself on a 50-foot comparator, the errors of which 
were known ; then winding the coil for trial, and unwinding and measuring the tape 
again, which was found not to have changed in length by more than a tenth of 
a millimetre in each 50 feet. The tape was supported horizontally on the polished 
surface of the comparator, and measured under a tension of 6 kilogs. Young’s 
m 2 
