552 PROCEEDINGS OF THE AMERICAN ACADEMY. 



introduced, with the above arrangement, one-half the percentage error 

 is introduced as regards the knowledge of the difference between S 

 and 2\ Let us assume that we are comparing two twenty-ohm resist- 

 ances differing by one two-hundredth of their value (or .1 ohm) with 

 leads of half an ohm each to one of the coils. If we should disregard 

 the lead resistance entirely, an error of 1 ohm or 5 per cent would result 

 with the ordinary method ; here the error would be approximately 



/41 40\ 1 



•^ U ~ 40 J = Too "^"^'' "' '^^^ P'^' '^'^'- 



The error has been reduced to one four-hundredth of the other value. 



The knowledge of the lead resistance is thus of less imjDortance, in so 

 far as the difference is less than one side of the entire bridge. 



The actual value of the lead resistance is determined most satisfactorily 

 by getting the balancing points on the bridge wire with the galvanometer 

 connected first to the end of the platinum wire by means of leads C and 

 E, and second to the ends of D and F, where they connect to the heavy 

 bridge terminals. 



The necessary accuracy in the knowledge of the lead resistance being 

 greatly reduced, we are able to have the leads of much higher resistance, 

 and thus decrease greatly the possibility of heat being conducted to or 

 from the platinum by the comparatively heavy low resistance leads. 



We have now the problem of arranging a resistance *S', whose value in 

 terms of some particular coil can be varied from a fractional part to four 

 or five times the resistance of the platinum at 0°C., and in steps small 

 enough to permit the use of a low resistance bridge wire. A box of 

 series resistances introduces calibration problems that are sure to give 

 trouble, especially when some of the resistances bear a ratio to the unit 

 of 100 : 1 or 1000 : 1. 



A simple and more direct method, but yet more tedious as regards 

 computation, is to arrange a set of five, ten, and twenty-ohm coils (whose 

 values in terms of a ten-ohm unit can be determined with great conven- 

 ience and accuracy) in series and parallel combinations the resistances 

 of which increase in steps of .1 or .05 ohms from two or three, to thirty 

 or forty ohms. The resistance of any complicated combination can be 

 computed with a little care, and the errors are especially small, as they 

 depend upon the comparison of coils in the ratio of 2 : 1 instead of 

 100 : 1. 



As finally built the bridge consists (Figure 3) of copper bars and 

 forgings 19 X 14 mm. in cross section, supported on a board of first- 



