SS6 Dr. W. H. Eccles on Coherers. 



detector. It will be found that 



Lan = Jjn 2 -+- mp + rn 

 and U=n?r — mnp + 2m 2 p , . . . (15) 



when, as implied by (3), we remember that 



mp =np 1 . 

 Now if (3) be differentiated (15) will be seen to become 



t i d € 

 hb = ri- 

 de ' 



Hence the coefficient of Wi in equation (14) 



qc(w— m)( ~ — + -)-=?. • • • ( 1G ) 



\rn-\-mpQ nj de 



It is convenient to think of c lt the impressed current 

 through the detector, as the independent variable. It is, of 

 course, readily connected with e ? the impressed voltage by 

 equation (3) or by fig. 4. 



Then as c x is varied the factor n—m has as graph a para- 

 bola with its vertex at the origin and with the tangent there 

 horizontal. This parabola rises only slightly within those 

 values of the current ever passed through detectors. The 

 second factor has as graph, when values of c 1 are again taken 

 as abscissae, a slowly falling curve. As a rule the graph of 

 the product of these two factors appears to be a slowly rising- 

 curve. Now the third factor dc/de is the gradient of the 

 e, c, curves shown in figs. 7, 8. Hence, for a self-restoring 

 coherer we expect that if we draw curves connecting w and 

 € — that is the watts in the telephone, and the voltage applied 

 steadily to the detector — we expect to find w rising to a 

 maximum at a value of e rather beyond the steepest part of 

 the steady-current e, c, curve. This deduction is amply 

 borne out by the experimental curves of figs 2 and 6. I am 

 not able at the present moment to enter into an exact 

 numerical examination of the whole matter ; but such rough 

 values as have as yet been obtained leave little room for 

 doubt that the quantitative adequacy of the above reasoning- 

 is as perfect as its qualitative sufficiency. It may be re- 

 marked, in closing, that the term subtracted from W 1 in the 

 bracket of equation (14) has its maximum at about the same 

 value of c x as gives the maximum gradient in the e, c, curve 

 of the detector. This deduction also has to some extent been 

 experimentally confirmed, but the lability of coherers is so 



