92 MU. F. E. SMITH ON THK CONSTRUCTION OP 



4 metres. The sensitiveness was. even greater than with the double bridge, but owing 

 to the nature of the method, the error was probably greater. 



In practice, P was not initially introduced into the circuit. This was to avoid any 

 warming of the mercury by the current. For it, D could be substituted by the 

 switch O. The two currents were maintained until in a very steady state ; P was 

 then thrown into the circuit and a measurement made in the following manner : 



Balance between P and Z was first obtained by an adjustment of R. The resistance 

 of X was greater than that of any of the mercury standards employed, and by an 

 adjustment of the shunt W, a balance could now always be secured between X and Z. 

 In some cases this was possible with P'. Afterwards the currents were reversed and 

 balance again restored in the two cases, usually by some slight adjustment of W. 



The value of P was thus known in terms of X (or, possibly, P'). In a similar 

 manner, X was evaluated in terms of P', the measurements being thus rendered 

 complete. As before, a correction for the platinum-foil junctions has to be made. 

 The thermal effects were usually equivalent to a difference in the shunted values of 

 0-006 per cent. 



Each tube was thus measured by the Kelvin bridge and potentiometer at least 

 three times. For each observation the tube was separated from its connectors, all 

 the parts re-cleaned, dried, and fitted together again, as already described. Apart 

 from the probable error of the observations, the consistency of the results is therefore 

 entirely dependent on the state of the tube, the purity of the mercury employed, and 

 the accurate adjustment of the connectors. Temperature variation is here ignored, 

 since the repetition of this can certainly be secured. 



The cleaning and drying of the parts, if always similarly conducted, should leave 

 the tube always in the same state, that is, the thickness of air and liquid films should 

 remain constant and be identical with those present when L and W were determined. 

 The mode of ensuring that the axis of the tube is in line with the axes of the 

 connectors has already been described. This method, however, was an afterthought. 

 Originally, this adjustment was attempted when the tubes were empty. Approxi- 

 mate consistency in the resistance measurements was obtained, but variations of 

 0*003 per cent, were not uncommon, and occasionally differences as great as O'Ol per 

 cent, resulted. The reason for these variations was not at first apparent. An accident 

 to one of the connectors, however, led to some interesting measurements. One of 

 these was slightly rotated by mistake, but the measurement was made without any 

 attempt to rectify the displacement. The value of the resistance was 0'03 per cent. 

 greater than usual, being equivalent to adding 0'2 millim. to the length of the tube. 

 While in the ice box, the same connector was rotated through various angles, and 

 measurements of the resistance made. The variation was as great as 0'05 per cent. 

 The explanation, of course, lies in the deflection of the stream lines, the value of the 

 resistance varying both \\itli the displacement of the axes and with the position of the 

 potential lead with respect to the section. 



