468 Special Reports 



between the terminals of the standard resistance, which was in series with the instru- 

 ment, was determined by the potentiometer method. The current through the sine 

 galvanometer was obtained from a lead storage battery; the maximum currents used 

 were from 0.14 to 0.15 ampere, the single deflection-angles being from 65 to 74. The 

 values of horizontal intensity determined by the sine galvanometer are in absolute 

 units, the degree of precision depending upon the magnitudes of any errors in the elec- 

 trical measurements and in the determinations of the constants for the sine galvanometer. 

 The formula (cf. pp. 391-392) for the measurement of horizontal intensity, H, by 

 the sine galvanometer is 



jr _ G,EC 

 R t > sin u 



where (7, = the constant of the sine galvanometer at the temperature t, E = voltage 



across the standard resistance, C=\ 1+ -(LR) [cotu ; 1 I R t , = resistance of 



L 2 \ / sin uj J 



the standard resistance at the tempprature t', u = single deflection-angle of the magnet, L 

 = scale reading when magnet is deflected clockwise, 72 = scale reading when magnet is 

 deflected counter-clockwise, and h = angle through which magnet is turned when tor- 

 sion head is turned through an angle/. According to calculations made by Dr. Barnett, 

 G is known to 1 part in 30,000. The temperature coefficient of G per degree centigrade 

 is 9xl0~. An error of 1 in temperature of the coils will, therefore, make an error 

 in G of less than 1 part in 100,000. The electrical instruments used in the preliminary 

 work were calibrated by the United States Bureau of Standards with a precision some- 

 what better than 0.01 per cent each for the potentiometer and for the standard cell 

 and of about 0.005 per cent for the standard resistance. The several temperature 

 coefficients are well determined; because of the excellent insulation of the Standardizing 

 Magnetic Observatory variations in temperature during the comparisons were very 

 small and gradual. Thus the combined maximum error for a single determination as 

 regards the electrical measurements might be about one part in 4,000, but as it is unlikely 

 that the errors are all in the same direction the actual mean error for a complete obser- 

 vation is doubtless less than one part in 10,000. Thus the error for a determination 

 of H is probably not more than one part in 7,000. 



Throughout the comparisons the author had the counsel of his colleague, Dr. S. J. 

 Barnett, who also took part on June 2 in the preliminary work and in the set-up of the 

 accessory apparatus. The observations with the sine galvanometer were made con- 

 tinuously during the intervals during which Mr. Fisk observed with the C. I. W. standard 

 magnetometer No. 3; Messrs. Fleming and Peters observed with the sine galvanometer 

 on June 3 and 4, and Messrs. Fleming and Ives on August 2, 4, 5, and 8. Since a complete 

 determination with C. I. W. sine galvanometer No. 1 rarely took over 2 minutes of time 

 the results constitute practically diurnal-variation series for the periods of comparison, 

 and simultaneous mean values applying for the intervals of oscillation and deflection 

 observations (at 3 distances, 25, 30, and 40 cm.) with magnetometer No. 3 were readily 

 obtained. To eliminate any question as to possible disturbing effects of any slightly 

 magnetic parts of the instruments 3 foot-screw orientations were used. The results 

 indicate that there is no such measurable effect. Stations and instruments were also 

 interchanged as indicated in Table 20F, which gives the data obtained from the com- 

 parisons. 



A few observations were made also on June 2 but these are not reported in Table 

 20F for the reason that the reversing switch used was not in good condition and great 

 trouble was experienced because of large and rapid variations in the electromotive 

 force of the storage battery on that day because the battery was drawn upon unexpect- 



