716 Mr. A. Campbell on Direct 



same resistance (at 0° C. say) as the thermometer, whilst the 

 arm AD is made adjustable within a small range so that it 

 can be made equal to AB at the zero-reading. Then if s, m y 

 and xq be the values found from a and /3 as shown above, 



T S 



AZ is set to be ttt-? and the shunt is made tt— . When 

 10 m Wm 



these adjustments are right to start with, then each t T q ohm 

 between Z and B represents 1° C. 



It will be remembered that in the above investigation we 

 neglected the powers of t above t 2 [equation (1)]. The result 

 of this is that the values of s, # , and m calculated as above, 

 give readings slightly in error, particularly at the higher 

 temperatures. It is not difficult, however, by one or two 

 trials to alter the value of s (and hence x and m) so as to 

 give readings whose accuracy is within two or three degrees, 

 from 0° to 1000° C. 



I have described this method mainly for two reasons : 

 firstly, that it led up to the discovery of the more exact way 

 (B), and secondly, because it is of some theoretical interest 

 to observe that the resistance of a platinum wire at varying 

 temperature can be very nearly represented by two resistances 

 in parallel, one of them having a constant temperature 

 coefficient and the other being invariable with temperature. 

 Perhaps the first part represents pure metal, while the second 

 (which is about ^ of the first) may be an alloy produced 

 by the impurities present. 



If the resistance-thermometer be of iron (at moderate 

 temperatures) or any other metal for which 



the temperature-reading curve can usually be brought very 

 nearly to a straight line by shunting the thermometer-arm 

 itself by a suitable invariable resistance. Neglecting the terms 

 above t 2 as in the preceding investigation, the requisite value 



2 N 



of the shunt would be R [-~— 1J. This value is not 



necessarily always positive, and thus it may happen that the 

 result is not attainable ( — it is possible for iron). As in the 

 case of metals like platinum, a better result can be got bv 



2 



trying various values of the shunt near to T\ t ( -jr — 1 j in this 

 case. ^ ' 



By introducing a little more complication, as in fig. 2, a 

 nearer approximation to linear variation with temperature 

 can be secured. 



