182 BELL SYSTEM TECHNICAL JOURNAL 



the noise-to-signal ratio at the receiver output will accomplish a 

 certain increase in usefulness of the circuit. 



It is difficult to set down noise-to-signal ratios that may be em- 

 ployed to distinguish between satisfactory and unsatisfactory service. 

 .Such requirements would be different in the cases of telephone, tele- 

 graph and broadcast circuits. They would also depend to a consider- 

 able extent upon the facilities that it is technically and economically 

 reasonable to provide. During times of magnetic disturbance radio 

 telephone circuits are continued in service when noise conditions are 

 very appreciably worse than would be tolerated on wire telephone 

 circuits. In this emergency situation it is, of course, necessary to 

 maintain service on the radio links as long as communication can be 

 carried on with a reasonable degree of satisfaction. 



Without a quantitative definition of the boundary between satis- 

 factory and unsatisfactory service in terms of noise-to-signal ratio it 

 is possible to determine from an analysis of past operating experience 

 what percentage of the time a certain circuit was unsatisfactory. With 

 such information available it would be useful to know how much this 

 percentage could be reduced by the application of transmission im- 

 provements. There is developed below a form of "reliability" 

 curve that makes it possible to estimate approximately the efTect of 

 such transmission improvements in terms of decibels upon the per- 

 centage of unsatisfactory or lost circuit time. 



As a background for the following discussion it will be helpful to 

 review briefly the conditions experienced on a typical short-wave 

 circuit and the way in which these are related to the present analysis. 

 For example, the instability of short-wave transmission over the North 

 Atlantic path is well known. There are days when these trans- 

 atlantic short-wave signals are remarkably good and others during 

 times of magnetic disturbance when they are exceptionally poor. 

 Between these two extemes is a wide range of circuit conditions. The 

 situation is illustrated in idealized fashion by Fig. 1. The ordinates 

 here represent average noise-to-signal ratios as measured on successive 

 days at the receiver output and curve A of Fig. 1 (a) shows how this 

 average might vary over an interval of many days. A certain noise-to- 

 signal ratio such as is indicated by the horizontal line B might be speci- 

 fied as the highest value tolerable for a useful circuit according to some 

 predetermined standard. Then the width of the cross-hatched inter- 

 vals C represents the lost circuit time. 



Fig. 1 (b) is the same as 1 (a) except that here a transmission im- 

 provement of X db has been applied so that the noise-to-signal ratio 

 at the receiver output is on all days reduced x db and the curve A is 



