108 BELL SYSTEM TECHNICAL JOURNAL 



peak-reading volume indicator is used, only about 5 per cent of the 

 listeners will now on the average hear distortion on the loudest program 

 passages, while if the r-m-s instrument is used, about 10 per cent will 

 hear distortion. To reduce the latter figure to 5 per cent would require 

 lowering the maximum volume level another decibel. Thus with this 

 criterion, the peak instrument has a slight advantage, as it would 

 permit the transmission of a 1 decibel higher average volume level for 

 the same likelihood of distortion being heard. 



The above statements assume that the observers and programs used 

 in the tests just described were representative of the listening public 

 and the programs they hear. Actually, the observers were trained by 

 experience in making many tests and were no doubt much more critical 

 than the average listener. Moreover, the conditions under which the 

 tests were performed, with the availability of frequent comparison 

 with the undistorted reference condition, were more conducive to 

 critical detection of overload than are average listening conditions. 

 These facts, together with the inevitable inability of the control 

 operator in practice to make his adjustments perfectly in anticipation 

 of the coming changes in the programs, tend to make the real practical 

 advantage of one instrument over the other considerably less than 

 shown by the tests. A further factor reducing the importance of the 

 small differences shown by the tests is the growing use of volume 

 limiting amplifiers at critical points in a broadcast system, such as at 

 the radio broadcast stations, which automatically prevent the trans- 

 mission of excessive levels. 



Another cumulative distribution curve is shown in Fig. 6, repre- 

 senting similar tests on a Western Electric 14B Program Amplifier. 

 This is a simple push-pull triode amplifier without negative feedback 

 and therefore having a more gradual cutoff than the 94B. (The gain 

 versus output power level curves at 1000 cycles per second are shown 

 in Fig. 7 for the two amplifiers.) It will be seen from Fig, 6 that the 

 data for the two volume indicators show no significant difference and 

 that the single curve equally well represents either set of data in the 

 region of interest. Somewhat fewer data are represented by this curve 

 and the agreement with the normal law is not quite so close as in the 

 previous case. 



The peak-reading instrument with the adjustment used in these 

 tests, although having characteristics similar to those usually proposed 

 for this type of device, is still far too slow in response to indicate the 

 true instantaneous peaks of the program wave. The question naturally 

 arises, therefore, whether any greater difference would be indicated if 

 the peak-reading instrument were made sufficiently fast in response 



