408 



SCIENCE. 



N. S. Vol. XI. No. 272. 



These only affect the deflection corres- 

 ponding to the unbalanced differences in the 

 ratios of the coils intercompared so that 

 even if the errors in interpolation should 

 be relatively large, the error in the ratio 

 will be exceedingly small. With the ratios 

 adjusted by means of shunts to within 

 TTo-ufiro of equality an error of 1%, 5 to 10 

 times that actually existing, would produce 

 an error of one part in 1,000,000 in the cal- 

 culated ratio of the coils compared. 



The Wheatstone-Kelvin bridge entirely 

 eliminates the resistance of the parts con- 

 necting the lower mercury cups of these coils 

 by means of shunting a second ratio coil 

 across the resistance to be eliminated, the 

 battery contact being transferred to its 

 middle point. The inequality of the two 

 halves of the ratio coil is eliminated by its 

 reversal. The remaining sources of error are 

 possible variable contact resistances in the 

 mercury cups and possible differences in the 

 insulation resistances between them. The 

 bottoms of the cups are accurately surfaced 

 and these sources of error are shown to have 

 no importance by interchanging the posi- 

 tions of the two coils compared. Thermo- 

 electromotive forces are eliminated by bat- 

 tery reversal. The heating effect of the test 

 current was shown to be quite negligible, 

 less than yo^^nnroinr ^O'' ^' current of .03 am- 

 pere for one ohm mangauin coils, since a 

 current of 1 ampere through such a coil pro- 

 duces a heating of the coil above the tem- 

 perature of the petroleum bath in which the 

 comparisons are made, of approximately 1 ° C. 



A specimen of the results which may be 

 obtained under rather unfavorable condi- 

 tions is given below. Four one ohm coils 

 were intercompared in the 6 possible com- 

 binations, 6 additional measurements being 

 made with the left and right coils inter, 

 changed with the following results : 



Observed Cal. Eesiduals 



L E DiSerences. Diff. Obs.-Cal, 



1402— 1403 = — 3.8 X10-« ohm —3.8 0.0 



EL —0.1 



1402- 



L 



1402- 



R 



1402- 



L 



1402- 



E 



1402- 



L 



1403- 



E 



1403- 



L 



1403- 



E 



1403- 



L 



1404- 



E 



1404- 



-1403= — 3.9 



E 

 -1404=— 4.6 



L 

 -1404 = — 2.8 



E 

 -1405 = — 0.8 



L 

 -1405 = — 1.0 



R 

 -1404 = + 0.2 



L 

 -1404 =—0.6 



E 

 -1405 =-1-3.4 



L 

 -1405 =-1-2.5 



E 

 -1405 = -1-3.4 



L 

 -1405 =-f 3.0 



— 3.7 —0.9 



-f 0.9 



— 0.8 0.0 



— 0.2 



— 0.2 ;-fo.4 



— 0.4 

 -f-3.0 +0.4 



— 0.5 

 + 3.2 +0.2 



— 0.2 



Since resistance comparisons can be made 

 to such a verj' high order of accuracy, to 

 within 5^0^% at least, even in the case of 

 properly designed .0001 ohm standards, all 

 interested may assure themselves that 

 every effort will be made by the Office to 

 provide itself with primary standards of 

 reference meeting the highest scientific re- 

 quirements. 



STANDAEDS OF BLECTRO-MOTIVE FORCE. 



The unit of electro-motive force, the volt, 

 is legally defined as WW of the electro- 

 motive force of a Clark Standard cell at 

 15° C. 



The official standards of electro-motive 

 force for a government should obviously be, 

 as far as possible, independent of any error 

 due to impurities of the chemicals used. A 

 large number of cells were set up with the 

 purest commercially obtainable materials, 

 from a number of independent sources. 

 The work of purifying these materials by 

 special methods was begun. Cells have also 

 been set up with some of the purifi.ed mate- 

 rials, though much still remains to be done. 



The intercomparisons so far made indi- 

 cate a most satisfactory agreement of all the 



