DESIGN OF SYSTEM — NORTH ATLANTIC LINK 39 



Computed Noise 



When the repeater design was estabhshed, the remaining theoretical 

 work on signal-to-noise consisted in refining the determination of the 

 repeater output and input levels; computations of system noise, and com- 

 parison of this with the objectives to establish the margin available for 

 variations; and determination of the necessary measures in manufactur- 

 ing and cable laying so that these margins would not be exceeded by the 

 deviations from ideal conditions which would occur. These deviations 

 assumed great importance, because they tended to accumulate over 

 the entire length of the system, and because many of them were unknown 

 in magnitude before the system was actually laid. 



The repeater levels and the resultant system noise were computed as 

 follows : 



It was recognized that the output levels of different repeaters would 

 differ somewhat at any given frequency. The maximum allowable output 

 level of the highest-level repeater was computed by the criterion that the 

 instantaneous voltage at its output grid would be expected to reach the 

 load-limit voltage very infrequently. This is the system load criterion 

 estabhshed by Holbrook and Dixon. ^ It premises that in the busy hour, 

 the load-limit voltage (instantaneous peak value) should be reached 

 0.001 per cent of the time, or less. It is probable that the level could be 

 raised 2 db higher than the one computed in this way without notice- 

 able effect on intermodulation noise. 

 An important factor in the Holbrook-Dixon method is the talker 

 volume distribution. Because of the special nature of this long circuit, 

 a careful study was made of the expected United States talker volumes. 



First, recent measurements of volumes on long-distance circuits were 

 examined for the relation between talker volume and circuit length. 

 They showed a small increase for the longer-distance circuits. This rela- 

 tion was extrapolated by a small amount to reach the 4000-mile value 

 appropriate to the New York-London distance. 



Second, an estimate was made of the probable trends in the Bell Sys- 

 tem plant in the next several years, which might affect the United States 

 volumes on transatlantic cable calls. 



The result of this was a "most probable U. S. volume distribution". 

 This distribution, which had an average value of —12.5 vu at the zero 

 level point, with a standard deviation of 5 db, agreed very well with one 

 furnished by the Post Office and based on calls between London and the 

 European continent. A further small allowance was then made for the 

 contribution of signaling tones and system pilots which brought the 

 resultant distribution to an average value of — 12 vu at zero level of the 



