r 



NEGATIVE FEEDBACK 429 



data ^ published by Crosby indicates that in the case of fluctuation 

 noise the ratio of the maximum peak ampHtude to the root-mean- 

 square value is about 13 decibels. The corresponding figure in the 

 case of a sine wave is 3 decibels. Hence equality of carrier peak ampli- 

 tude and the highest peaks of the disturbance obtains when the ratio 

 of their root-mean-square values is 10 decibels. With feedback this 

 condition appears to define a fairly critical disturbance level above 

 which the output signal-to-noise ratio is very rapidly diminished. 

 Crosby has shown ^ that with systems employing amplitude limiters a 

 similar condition marks the point beyond which the noise improvement 

 realized at the lower disturbance levels is soon lost. This point he 

 has termed the "threshold of noise improvement." 



A less sharply defined break also occurs in the curve expressing noise 

 conditions in the absence of feedback. This is the result of a progres- 

 sive destruction, at the higher disturbance levels, of the balancing out 

 of amplitude efifects in the push-pull detector which is realized when 

 the noise is low. 



While direct comparison of the feedback system with an actual 

 amplitude modulation system was not possible with the equipment 

 used, it is thought that a comparison based upon theoretical considera- 

 tions may be of interest. The procedure is as follows: The noise ratios 

 shown in Fig. 15 for the system without feedback are, for disturbances 

 below the threshold value, 9.6 decibels in excess of those which would 

 be realized in a fully modulated amplitude system. A dotted line, 

 displaced from the linear portion of the measured curve by this amount, 

 is shown. The abscissae of the dotted curve do not represent the true 

 carrier-disturbance ratio which would obtain in the amplitude system 

 for the reason that, ideally, the intermediate-frequency amplifying 

 system would have a band width of but 8 kilocycles. In such a system 

 the signal-to-noise ratio would be equal to the carrier-disturbance ratio 

 except at the very high noise levels. Hence the intercept of the dotted 

 line with the axis of abscissas marks the point of equal carrier and 

 disturbance levels in this system. The difference of 11 decibels be- 

 tween this point and the zero point on the scale as drawn measures the 

 amount by which the disturbance at the rectifiers in the experimental 

 system exceeds that which would be found in the ideal amplitude 

 system.^ Consequently, if it is desired to relate the data of Fig. 15 

 to the disturbance ratio which would exist at the input to the detector 

 in the amplitude system, and hence to the signal-to-noise ratio in that 



^ Comparison of the areas under idealized curves representing the square of the 

 transmission through the intermediate-frequency systems in the two cases indicates 

 a difference of 10.1 decibels. 



