FL UCTUA TIONS IN MICROPHONES A ND OTHER RESIST A NCES 209 



Figure 9 summarizes the results obtained in a typical set of measure- 

 ments on the noise generated in the transmitter cell before mentioned. 

 The resistance was varied by changing the height of the carbon layer 

 above the carbon which was in the conducting path. Each of the 

 plotted points is the average of nine observations, all of which occur in 

 a range of 1 db. The relationship plotted in Fig. 9 can also be ex- 

 pressed by Eq. (2) and we again find jS = 1.25. We shall see later — 

 Eq. (8) — that when the resistance of an aggregate is altered by chang- 

 ing the number of conducting contacts between the electrodes quite 

 another relationship between noise and resistance is obtained. Hence 

 our assumption regarding the nature of the resistance change in this 

 cell is consistent with the data of Fig. 9, and we believe that in this 

 experiment we have measured the average value of ^3 for all the contacts 

 in the conducting path and have found it to be in agreement with the 

 average value deduced from our single-contact measurements. 



Noise as a Function of Frequency 



For measuring the frequency distribution of the noise the filter 

 shown in Fig. 1 was replaced by a frequency analyzer ^ having a 

 constant band width of 20 cycles, the midpoint of which could be set at 

 any point between 50 and 10,000 cycles per second. The calibration of 

 the apparatus was checked by measuring the frequency distribution of 

 thermal noise which was constant over this entire range, in accordance 

 with theory. 



The results of the measurements on a standard carbon transmitter, 

 maintained at constant resistance and applied voltage, are shown in 

 Fig. 10 where ordinates represent the mean square noise voltage over 

 the 20-cycle band and abscissae represent the mid-frequency of the 

 band. It is seen that the experimental points fall on a straight line 

 having a negative slope of about 1.0. This relationship may be 

 represented by the equation 



AV} = Const. IFjF, (3) 



where A F? is the mean square noise voltage for the frequency band AF. 

 Integrating Eq. (3) between fixed limits we obtain: 



V} = Const, log (F2//^i), (4) 



which gives the total noise over the frequency range Fi to Fo. 



Figure 11 gives the results of similar measurements on a high- 

 9 T. G. Castner, Bell Laboratories Record 13, 267 (1935). 



