Unit of Candle-power in White Light, 



79 



the lamp was likely to have taken place. When, however, 

 the comparisons were spread over a long period of time, the 

 N.P.L. value assigned to any lamp was taken as the mean 

 of the initial and final measurements on it. 



The following table shows the value of the N.P.L. unit in 

 terms of that of the Reichsanstalt, Bureau of Standards, and 

 Laboratoire Central, deduced from the mean of all the lamps 

 tested : — 



Table VI. 





Katio. 



Difference from 

 accepted value. 



Bureau of Standards/N.P.L 



Eeichsanstalt/N.P.L 



Laboratoire/N.P.L 



0-90 oa 



1-00 2 



+0"3 3 p.c. 



+0- a p.c. 



A higher accuracy than 0*25 per cent, cannot be claimed 

 for these intercomparisons, and the results, therefore, show- 

 that within the limits of accuracy of the experiment, the 

 units of candle-power at the four laboratories realized in 

 lamps operating at 1*5 watts per candle are in virtual agree- 

 ment. The chief interest of the result lies in the fact that 

 different methods were used at the several laboratories for 

 carrying out the comparison. At the Laboratoire Central 

 the comparison was made by measuring the lamps directly 

 against standards of about 3*5 or 4*0 watts per candle — thus 

 the large colour difference was bridged in one step. The 

 Laboratoire Central values given are based on tests by an 

 observer with normal colour vision. A second observer with 

 abnormal colour vision obtained readings differing by 2 per 

 cent, from these. By the method adopted at the Reichs- 

 anstalt a faintly coloured blue glass is placed between the 

 photometer and the lamp of lower efficiency of the two 

 under comparison so as to equalize the colours. This blue 

 glass is calibrated once and for all, and the constants so 

 determined are used in deducing the candle-power of the 

 high-efficiency lamps on test. 



The values assessed by the Bureau of Standards are the 

 mean of comparisons taken (a) directly and (b) by means of 

 a blue glass calibrated photometrically. In this calibration 

 the mean of a number of observers was taken in fixing the 

 transmission constants of the slass. 



