972 THE BELL SYSTEM TECHNICAL JOURNAL, SEPTEMBER 1952 



RATIO OF LIFE PATH TO SEGMENT LENGTH 



Fig. 16 — Noise correlation. The dotted curve is calculated for lumped noise 

 sources. 



In view of these considerations there seems to be very satisfactory agree- 

 ment between the experimental results and the model. 



Another type of experiment involves noise measurements at frequen- 

 cies high enough so that the transit time of a hole across a segment is 

 an appreciable fraction of a cycle. In this case the correlation between 

 noise voltages from adjacent segments can be improved by putting a 

 time delay in one channel of the measuring circuit. Measurements were 

 made by taking the noise voltages from the two segments through sepa- 

 rate amplifying channels having identical pass bands extending from 17 

 to 24 kilocyles. The outputs of the two channels were put on the vertical 

 and horizontal plates of a cathode ray oscilloscope, forming a sort of 

 Lissajous pattern. The patterns differ from those obtained with sinus- 

 oidal voltages in that the elliptical figures are filled in solid, due to the 

 continual variation in amplitude of the noise. A phase shifting device is 

 included in one channel, and as the phase is shifted to give optimum 

 correlation, the elliptical pattern narrows down and approaches a line 

 inclined at 45°. For a quadrature phase shift, the pattern becomes cir- 

 cular, and in practice this setting can be determined more precisely 

 than the in-phase setting, largely because the backgroinid noise in the 

 circuit is less troublesome. With the phase shift for optimum correlation 

 determined, the delay at the center of the pass band is easily calculated, 



