CARRIER TELEGRAPH METHODS 493 



reasonable that there could likewise be a number of speeds where maxima 

 and minima occur in the distortion. 



It is practically impossible to sift out and measure all causes of irregulari- 

 ties in a reasonable time. Therefore the irregularities have been shown 

 exactly as measured in all the following curves wherein noise and interchan- 

 nel interference were absent. 



Transmission fluctuations due to the causes just mentioned were small 

 compared to those encountered when strong interference or noise was pres- 

 ent, because the latter varied greatly with time. In resistance noise, for 

 example, peaks of great amplitude occur occasionally, although most of the 

 time the fluctuations are relatively minor. It was necessary to observe for 

 several minutes the distortion measured in the presence of resistance noise 

 before one could be sure of finding anything approaching the maximum dis- 

 tortion; and the longer the period of observation, the greater was the peak 

 distortion. In order to complete the testing in a reasonable time, watching 

 periods were restricted to five minutes per observer and his maximum distor- 

 tion reading was recorded. The results of two such obser\'ation periods for 

 the same noise condition were averaged to determine a point for an experi- 

 mental curve of distortion vs. noise-to-carrier ratio. Such points when 

 plotted failed to lie in a perfectly smooth curve, but a smooth curve disre- 

 garding irregularities was drawn through the available points in what was 

 estimated to be the correct location. The curves are described under the 

 heading "Xoise Tests" and may be used for comparison purposes, but are 

 not an exact measure of the worst distortion to be expected over a long 

 period of time. A similar procedure was followed for distortion measure- 

 ments with interchannel interference. 



Distortion vs. Speed Tests 



The distortion mentioned throughout this paper is the absolute value of 

 the maximum total distortion measured with the test sentence, and for 

 brevity is merely called distortion. Except where otherwise specified, the 

 arrangements were previously adjusted to have zero bias on reversals at the 

 same speed. In these tests line noise and interchannel interference were 

 absent. 



Frequency-Slnft A rrangemeuts 



Limiler 



Some preliminary measurements on a frequency-shift arrangement having 

 the loss characteristic of curve A of Fig. 1, with carrier varied abruptly from 

 1920 to 1990 cycles, showed that the limiter has little effect on the speed of 

 the channel whether or not channel filters are used. The fact that low dis- 

 tortion was measured without channel filters indicated that the channel and 



