1024 THE BELL SYSTEM TECHNICAL JOURNAL, SEPTEMBER, 1953 



Tolerance to Echo 



People vary greatly in their reaction to echo of given magnitude and 

 delay and it is therefore necessary to treat them statistically, that is, 

 what we need are the two statistical terms which are usually used to 

 describe a mass of data, i.e., "average" and "standard deviation." 



"Average" is simply the familiar algebraic average of the data — no 

 talker is average but it is possible to obtain the average of a lot of talkers. 



"Standard deviation" is a number which tells in general terms how the 

 individuals spread out on both sides of the average. Usually: 30 to 35 per 

 cent will be between the average and one standard deviation below the 

 average; 30 to 35 per cent will be between the average and 1 standard 

 deviation above it. At least 45 per cent will be between the average and 

 two standard deviations below it and at least 45 per cent will be between 



Table I 



the average and two standard deviations above it. Very few, if any, will 

 be outside of plus or minus three standard deviations. 



Judgment tests under controlled conditions and with a number of 

 observers (talkers), under conditions simulating connections to sub- 

 scribers near the toll office, have given the basic data on these effects 

 in Table I. (These data are slightly different from some published 

 earlier because of recent ree valuations. Further studies are now under 

 way and may indicate some further changes.) 



An analysis of all the test data indicated that the observer judgments 

 conformed fairly well with a normal law curve having a standard devia- 

 tion (Do) of 2.5. This means, for example, taking the 40 millisecond delay 

 condition, that while on the average a 17.7 db loss was required in the 

 echo path to make the echo just tolerable, some 30 to 35 per cent of 

 observers could tolerate 2.5 db less loss. Another 30 to 35 per cent were 

 sensitive enough to need 2.5 more loss for satisfactory echo condition. 

 Practically no observer was so sensitive as to require 3 X 2.5 = 7.5 db 

 more than 17.7 db, and practically none was so tolerant that he could 

 permit 7.5 db less. 



