LOAD RATING THEORY 633 



channel distribution directly. Mr. M. E. Campbell effected this 

 addition by the use of phonograph records, the w-channel distributions 

 being determined by means of the instantaneous voltage sampling 

 apparatus previously mentioned. 



As material for this process, 16 high-quality phonograph records 

 were made of the outputs of commercial subsets through representative 

 subscriber loops. Both male and female voices were used. The 

 speech was furnished by reading magazine stories containing consider- 

 able conversational material, due precautions being taken that the 

 volume on each record was substantially constant throughout. A 

 calibrating tone was cut on each record to enable it to be played at any 

 desired volume and most of the volumes recorded were well below the 

 point at which the transmitter began to act as a voltage limiter. 



These individual records were then combined in groups of four, 

 with all records adjusted to the same volume by means of the calibrat- 

 ing tones, and re-recorded. Several such 4-voice records were made; 

 by combining them again in the same way, 16-voice records and finally 

 64-voice records were obtained. The instantaneous voltage distribu- 

 tions were measured before and after each re-recording to insure that 

 the recording process introduced no errors. A few minor discrepancies 

 were found, but all were small enough to be disregarded. Each single- 

 voice record appeared several times in a 64-voice record, but since the 

 phases of its dififerent appearances were random, this had no appreci- 

 able effect on the resultant voltage distribution. This was verified 

 by comparing the voltage distributions of the various possible 16- 

 voice combinations. By this process w-channel voltage distributions 

 were obtained for n = 1,4, 16 and 64. 



These distributions, together with a normal curve, are shown in 

 Fig. 2 in cumulative form. To show the curves conveniently to the 

 same scale, it has been necessary to plot for each case not the distribu- 

 tion of E, the rectified instantaneous voltage, but that of E/U, where 

 U is the rms voltage. The rms voltage, it will be remembered, is 

 directly related to the equivalent volume by equation (2). The 

 figure shows clearly the gradual transition from the single-channel 

 distribution to the normal one for large n, and also indicates that for 64 

 active channels the curve is normal within the precision of the measur- 

 ing apparatus. Hence, the normal distribution may justifiably be 

 used for any value of w > 64. 



Further significance is accorded the above data by plotting the 

 ratio of the voltage exceeded a fraction e of the time to the single- 

 channel rms voltage, as a function of the number n of active channels, 

 for several fixed values of e. From the data given, points on such 



