of Duplex Telegraphy. 533 



other, at all events as far as this investigation is concerned. 



Consequently the safest plan will be to make each influence as 



small as the circumstances will allow it. 



The disturbance S x for any constant eqkv \/b and any given 

 SK will obviously become smaller the larger RK is selected. 

 Supposing R + K constant, whatever that value finally may be, 

 RK has a maximum for R=K ; and the very same condition 

 will obviously make the disturbance S 2 smallest. 



S 3 offers no best condition ; this expression only shows that 

 it has an absolute maximum with respect to b, namely (as 

 R = « + d-f 6) for b = a + d. 



Thus we are informed that whatever relation between b and 

 a + d may be finally chosen, b — u-{-cl should not be selected, as 

 otherwise any given variation of \ would have the greatest 

 possible disturbing effect on the balance. But b = ot + d being 

 the condition for the maximum magnetic effect in the compen- 

 sation circuit, it is hereby established that for the sake of regu- 



arity of signals, which under all circumstances is to be consi- 

 dered of paramount importance in duplex telegraphy, the mag- 

 netic effect in the compensation branch must not be achieved 

 in the most economical manner, but quite the reverse. This, 

 as the compensation circuit has actually to produce wholly or 

 partly the duplex signals, is a testimonium paupertatis for the 

 compensation method, and proves it in this respect inferior to 

 both the double balance and the differential method. 



R=K 



is the regularity condition for the compensation method ; 1. e. 



In order to make the disturbance of balance by a variation of 

 the resistance in both the circuits absolutely as small as possible, 

 the total resistance of the compensation circuit should be equal to 

 the total resistance of the line-circuit*. 



If we now substitute in Si for K the value R, and in S 2 for 

 R the value K, we get 



* This result is against the adopted view, for Dr. Gintl, as well as others 

 after him, has always treated the compensation circuit as a kind of local 

 circuit, i. e. giving to it as low a resistance as practice allows. But this is 

 Nearly wrong : for if E. is made very small as compared with K, the ba- 

 ance becomes unstable. This fact explains, to a certain degree, the failure 

 tvhich has attended the application of the compensation method for duplex 

 working, because the method was tried under the most unfavourable quan- 

 itative arrangements. 



