MEASUREMENT OF TELEGRAPH TRANSMISSION 147 



istic distortion may be obtained by determining the difference between 

 the systematic distortion measured with selected recurring signals and 

 the average distortion with miscellaneous signals. Fortuitous dis- 

 tortions may be taken as the difference between the total and sys- 

 tematic distortions obtained with recurring signals. In order to 

 distinguish between the variable bias and the true characteristic 

 distortion in the case of level-compensated circuits measurements 

 may be made with the compensator disabled and with it functioning. 



For tests in the field where it is desired to measure systematic 

 distortion effects, as for instance in connection with equalizing, several 

 simple signals corresponding to certain especially selected teletype- 

 writer characters are employed. In each of these signal combinations 

 there are only two transitions; therefore it is convenient to observe 

 the effect of the remnants of one transition upon the next transition. 

 As discussed in the Appendix, the characteristic distortion obtained 

 for miscellaneous signals is a function of the distortion obtained with 

 the simplest characters; if there were no distortion on the simple 

 characters no characteristic distortion effects would be expected 

 when miscellaneous signals were transmitted. The process of equal- 

 ization, therefore, consists in adjusting the transmission characteristics 

 of the line circuit to reduce the characteristic distortion measured 

 on six special characters to a minimum. The teletypewriter char- 

 acters which are used for this purpose are Blank, T, 0, M, V and 

 Letters; the corresponding signals are shown in Fig. 1. The distortions 

 of these signals are observed at the receiving end on a measuring set 

 operating on the start-stop principle or a portable systematic-distortion 

 measuring set having an integrating meter, as will be described more 

 fully below. 



In equalization testing, each of the six signals is sent repeatedly for 

 the time required to determine the total systematic distortion — 

 generally about 50 repetitions. In analyzing the results the bias 

 component is assumed to be substantially the same for all of these 

 signals and whatever difference is observed from one signal to another 

 is taken as being due to characteristic distortion. It is found generally 

 that the result for the O signals, which are practically unbiased six- 

 cycle reversals (see Fig. 1), is not radically different from the result 

 obtained for bias with reversals at 6 or 23 d.p.s. (shown in the lower 

 part of Fig. 1), the results being expressed, of course, in the same 

 terms, as for instance in per cent of a 23-cycle dot. The largest 

 distortion is usually found on either the Blank or Letters character, 

 this being reasonable because usually the remnants of transients 

 practically disappear within a few dot lengths. Sometimes it is 



