324 Mr Campbell, Discontinuities in Light Emission. 



The total current through either cell {i) is given by 



i = eQyN (12). 



The magnitudes of the various measured instrumental constants 

 in the experiments which have been made are approximately as 

 follows : 



(These values are not the most favourable that can be obtained 

 experimentally, so that our estimate is once more a minimum 

 estimate.) 



a = 6 = 0-2, R = 0% (7=500, s = 10«, ^ = 100. 



r^ 



Assuming that the value of | f^ {t) dt can be calculated from 



Jo 

 (11'), it is found to be about 2 x 10^ el Hence for the fluctuations 

 of the balanced cells illuminated by independent sources 



0^ = 2 {(o' + ^) N .2 .lO'e^ 



= 4 . 10^ . €1 



a) 



_ = 16, 



or Je^^ = 4 (about). 



This value is just on the limit of the range within which 

 detection is possible — it means that the electrometer spot will be 

 on the average about 4 divisions distant from its mean position. 

 But this is a minimum estimate : as a matter of fact such pre- 

 liminary observations as have been made with two independent 

 sources (two Osram lamps) indicate that, with these values for 



the instrumental constants, Jd^'^ is about 18 — which is of the 

 right order of magnitude. 



§ 17. A brief explanation must be given of the manner in 

 which the observations were taken and the value of 6t^ deduced 

 from them. 



This quantity is the arithmetic mean of the squares of the 

 deviation of the indicator from its mean position for all times 

 after a period T from the moment that the electrode is insulated. 

 As was pointed out in the previous paper, the most satisfactory 

 way of estimating djf'^ would be by recording the motion of the 

 spot photographically and integrating the resulting curve (with 

 the ordinates squared) graphically. But since, at least at 

 present, no great accuracy has been required, the observations 

 have been taken by eye at equidistant intervals of time (4") 

 indicated by a metronome. 200 such observations constitute one 

 series, of which the probable error is, as has been shown, 1/V200. 



