42 BELL SYSTEM TECHNICAL JOURNAL 



single-frequency output of the system (taken to be 9 decibels above a milli- 

 watt at zero level). This output corresponds to an input just reaching the 

 peak limiters. Thus in a single link the idle circuit noise is down about 68 

 decibels from the full-load sine wave. For five links in tandem, the readings 

 are at least 8 decibels below the accepted limit (29 decibels on the left-hand 

 scale) for noise at the end of a 4000-mile circuit. Quantizing noise is found 

 to increase approximately three decibels for each doubling of the number of 

 links, as is generally the case with other forms of random noise. 



For a single channel, or a small number of channels in tandem, the idle 

 circuit noise varies considerably with the vertical centering of the coding 

 tube. This may be understood by noting that if the zero-signal operating 

 point effectively rests near the center of a "tread" on the quantizing stair- 

 case, (Fig. 4) a small amount of power hum or other disturbance may simply 

 move it back and forth on the same tread, in which case quantizing noise is 

 entirely absent. On the other hand, if the operating point is near a ''riser" 

 the small disturbances may cause it to joggle from one tread to another, 

 producing noise. The measurements given in Fig. 27 were obtained with a 

 very quiet input circuit and with the centering adjusted for maximum noise. 

 The "joggling" was produced largely by residual power hum. 



A-B tests to compare PCM transmission over a single link with direct 

 transmission over a noise-free circuit of the same audio band were carried out 

 using a wide range of talker volumes. In such tests only a few experienced 

 observers were able to pick the PCM path consistently. When the PCM 

 circuit included five tandem links most observers could tell the difference, 

 but all judged the quality to be more than satisfactory for toll service. 



Crosstalk. It has been pointed out earlier in this paper that interchannel 

 crosstalk in a PCM system can occur only in the terminal equipment. 

 Considerable care was exercised, particularly in the design of the time- 

 division parts of the system, to hold individual sources of crosstalk to 70 

 decibels or better. As a result, measurements using a single-frequency test 

 tone and a current analyzer have shown the overall crosstalk from any one 

 channel to any other to be down 66 decibels in the worst cases. 



Very severe tests have also been made in which a loud talker was connected 

 to ten channels of a group simultaneously and crosstalk into either of the 

 two remaining channels (one odd, one even) was listened to, and measured 

 with the 2-B Noise Set. In such tests unintelligible crosstalk could be de- 

 tected, which seemed to consist of changes in the quality of the quantizing 

 noise occurring at a syllabic rate. The 2-B readings averaged about a 

 decibel above the quantizing noise of a single idle channel with occasional 

 peaks reaching the 17-decibel point on the reference scale. 



In tests involving 24 talkers in 12 simultaneous conversations, crosstalk 

 was practically undetectable. 



