492 BELL SYSTEM TECHNICAL JOURNAL 



commercial type, they are believed to be representative. This distortion 

 was erratic, depending upon the speed and wear of brushes and commutators. 

 The distortion measured at the receiving relay could not be corrected by sub- 

 tracting the distortion of the sent signals because during miscellaneous 

 signals maximum distortion in the received signals might occur on a different 

 transition from that in the sent signals. The small errors which existed in 

 the sent signals are therefore believed to have been neither serious in their 

 effect on the measured distortion, nor the sole cause of irregularity in the 

 data. 



The accuracy of the distortion measuring set itself was in the order of 1 per 

 cent, as determined by measuring known amounts of bias in signals sent from 

 a special distributor. Usually two observers took independent readings 

 which were required to check closely or the observations were repeated. The 

 average of the two observations was taken as the final measurement. 



Although the line loss was constant in these tests, amplifiers, oscillators, 

 power packs, and telegraph batteries were subject to slight voltage varia- 

 tions. Precautions were taken to reduce all variables as far as was practica- 

 ble, yet it seems likely that the telegraph transmission measurements may 

 be slightly in error due to such variations. 



The individual sources of error mentioned in the last three paragraphs 

 seem reasonable and sufficient to account for most of the irregularities in the 

 following curves of telegraph transmission vs. speed. Yet it did not seem 

 fair to draw smooth curves and neglect the irregularities, because these can 

 also be due to the telegraph system itself, as was found by careful and re- 

 peated measurements. For example, it is known that relay performance is 

 erratic and depends upon the speed. Chattering of relay contacts and 

 periodic vibration of the armature have appreciable effect upon the distor- 

 tion and can cause irregularities in distortion vs. speed characteristics, 

 particularly at the higher speeds. Furthermore, such irregularities may not 

 be wholly reproducible in repeated measurements due to changes in the relay 

 temperature or contact surfaces and due to the occasional readjustments of 

 relays. Another cause of irregularities in a distortion vs. speed characteristic 

 may be the loss and phase characteristics of the channel filters. For example, 

 consider an ideal transducer^ which is distortionless at a speed s near the 

 cut-oflf. It is also distortionless at speeds such as s/2, s/3, s/4, s/5, etc. At 

 intermediate speeds distortion may exist, so that a curve of distortion vs. 

 speed would show irregularities with minima at these optimum speeds. 

 (Some computations by one of the writers for a frequency-shift arrangement 

 using idealized filters show irregularities in the distortion vs. duration curve 

 for a single dot.) No attempt was made to shape the channel filter charac- 

 teristics to be perfectly distortionless at a particular speed; and it seems 



