206 BELL SYSTEM TECHNICAL JOURNAL 



charge accumulated during a mark would be dissipated during the 

 following space, and since both marks and spaces continually change in 

 relative lengths, this would give rise to characteristic distortion. This 

 difficulty could be taken care of by having a relay operating in unison 

 with the receiving relay and serving to disconnect the leakage resistance 

 Re during spaces in the manner shown in Fig. 6. Actually, however, 

 the presence of the channel filters causes the received signals to resemble 

 more nearly Fig. 275. This leads to a further difficulty due to the 

 fact that the charging intervals ah, cd, etc., are shorter than the periods 

 a'b', c'd', etc., during which the receiving relay is closed. To remedy 

 this, the compensator relay is biased towards spacing by means of the 

 resistance AB (Fig. 6) and associated battery; the operating impulses 

 being first rounded off by the resistances, condenser, and inductance 

 in the wave shaping circuit to make them susceptible of such time bias. 

 A necessary condition to be fulfilled by this circuit is that it should 

 supply enough energy to the compensator relay to operate it even 

 under conditions of extreme bias. 



Drift 



Returning to Fig. 275, we may consider the bias of the compensator 

 relay as equal to a'h' — ah for the shorter signal or c'd' — cd for the 

 longer one. These biases are, of course, equal time intervals which 

 we will denote by 5. From this it can readily be shown that in a given 

 period — one second for instance — the grid will be conducting during a 

 longer time for a group of long signals than for a group of short ones. 

 Thus, let it be assumed for example, that c'd' = 2a'b' and let n be 

 the number of marking conditions of length a'h' in a given interval; 

 we have for the cumulative charging time in the two cases: 



Ta = n{a'h' - 8) 



whence 



Tb-^"^ (c'd' -8) = n( a'h' - | . . 



J- B = 1 A -r -^ • 



It follows that the charge on condenser C is greater for the longer 

 signals. If, therefore, the receiving circuit is adjusted to give unbiased 

 signals when dot signals at the rate of 11 d.p.s. are received, similar 

 signals at 23 d.p.s. will be biased positively. Experience shows that an 

 adjustment which gives zero bias with 11 d.p.s. dots will give sub- 

 stantially unbiased signals with the standard test-sentence; hence 



