638 Mr. H. E. Schmitz on the 



The values of W, r, and x could all be determined with a 

 high degree of accuracy; the accuracy of the value of E 

 depends therefore mainly on the measurement of s. If z is 

 the least recognizable current through the galvanomer and 

 g the resistance of galvanometer and thermo-circuit, then we 

 find 



E — sxi = + (g + s) z, 

 whence proportionate error in E = y — - . — . . In the present 



s XI 



measurements it was found convenient to use a moving-coil 



galvanometer, in spite of its high resistance (600 ohms), 



The value of s was from about 50 ohms to about 400 ohms. 



Taking z as the current corresponding with a deflexion 



of a millimetre on the galvanometer scale, its value was 



3 13 



jqio ampere. Approximate values for x (jqq) and i (ffQ^ 



7 z 



ampere) give — — h for —.. Hence, taking the maximum value 



q-\-s Xl 



(13) for ' — — , we find as the proportionate error in E (or 



in s) 1 in 1000. 



This result assumes absence of current through the galva- 

 nometer when the galvanometer circuit is completed by 

 closing the key K after removing all sources of electromotive 

 force, a condition not easy to realize in practice. In the 

 earlier measurements the resistance AB (fig. 1) was provided 

 by a new compensation apparatus of the type described in 

 the Pliysikal. Zeitschr. vol. i. p. 167. The fractions of an ohm 

 in the instrument used were provided by exposed wires of 

 " Kruppin," and the various external and internal contacts 

 almost always produced an appreciable thermo-voltage ap- 

 parently not avoidable by any precautions. In the later 

 measurements the resistance AB was provided by two new 

 "precision" resistance-boxes (Siemens & JBalske, No. 19530 

 and No. 19531) reading from 0'1 ohm upwards. This arrange- 

 ment gave much more satisfactory results. The chief error 

 was now due to the key K, which in this case consisted of 

 two mercury cups connected by a copper wire (unfortunately 

 not well amalgamated) attached to a sealing-wax handle. 

 It was found that a current, large enough, to cause a con- 

 siderable error in the measurements, might be produced at 

 the contact of copper and mercury if a fresh surface of 

 mercury were not constantly exposed. 



