APPLICATION OF PRINTING TELEGRAPH 



613 



from the transmitting station would be more than those given in the 

 table for half of the time in each month. 



TABLE II 



From these figures it is apparent that, under the conditions given, 

 satisfactory all-year-round transmission could probably not be ob- 

 tained over a radius of more than a hundred miles. To obtain the same 

 grade of copy at a distance of 400 miles, as this assumed set-up could 

 give at 100 miles, would require an increase in radiated power of about 

 25 db, making about 316 kilowatts radiated. 



Development of systems and tests of the kind involved in obtaining 



information such as the authors have reported above have required the 



cooperative effort of a considerable number of engineers of the British 



General Post Office and of various parts of the Bell System. In solving 



many of the problems of telegraph signal transmission Mr. J. Herman 



was particularly active. 



Appendix A 



In deriving Curve A on Fig. 6 between "cumulative per cent of ob- 

 servations" and " Rochester signal-to-noise ratio at the voice-frequency 

 filter output" from the Houlton noise data, the following facts were as- 

 sembled and coordinated. In the first place, it was determined from 

 the analysis of a large number of observations of loop noise at Houlton, 

 that the magnitude of the noise is random and that its distribution 

 obeys the Normal Law of Probability frequently used in engineering 

 studies, provided the values of noise are in each case expressed as the 

 number of decibels the "warbler" noise is above one microvolt per 

 meter. Since each observation requires about the same time to com- 

 plete and the observations are made at the same fixed times each day, 

 the process really becomes one of sampling and the "per cent of obser- 

 vations" is equivalent to the "per cent of time" for the period covered 

 by the tests. Then if, as in the Rochester tests, the radio signal 

 strength is substantially constant, the signal-to-noise ratio (expressed 



