344 



BELL SYSTEM TECHNICAL JOURNAL 



The right-hand portion of the curve is determined by the audibility 

 of the beat-note, and its position will of course depend upon the masking 

 effect of the noise background. Theory has indicated that the flutter 

 frequency portion of the curve should drop with the noise level, but it 

 is evident that, with such a small difference in carrier amplitudes as 

 is indicated in the figure, the results would not be appreciably differ- 

 ent were the noise level to be reduced. On the other hand a noise 

 level which is down only 10 db from a 30 per cent modulated signal is 

 equivalent to a modulation of nearly 10 per cent. This is an extremely 

 objectionable noise level, so objectionable, in fact, that under the condi- 

 tions of the tests it was very unpleasant to listen to. Consequently, 

 it did not seem worth while to run curves similar to that of Fig. 5 for a 

 number of different noise levels. Instead, a set of observations was 

 made with a fixed carrier frequency difference of 2 cycles and a variable 

 noise level. The results are indicated by the lower curve in Fig. 6. 



f ■ — ■ — 



<; (» 



> I 1 1 1 1 1 1 1 1 1 1 



^' 



5 10 15 20 26 30 



NOISE LEVEL IN DECIBELS BELOW 30 "/o EQUIVALENT MODULATION 



Fig. 6 — Carrier ratio for perceptible flutter as a function of noise level. The upper 

 curve is for the square law and the lower curve for the linear detector. 



With a noise level equivalent to a 30 per cent modulation, a carrier ratio 

 of only 2 : 1 is necessary to reduce the flutter to a barely detectable 

 amount. At low noise levels, down 20 db or more from 30 per cent, the 

 flutter could hardly be detected but there was noticeable a "bumping" 

 sound which was due to the rather violent motion of the cone of the 

 loud speaker at a frequency of 2 cycles. This was partially eliminated 



