REGENERATION THEORY AND EXPERIMENT 



693 



stricted frequency range as compared to the null method, but it 

 covered the region of particular interest in the experiments conducted 

 for the purpose of testing the stability criterion. Through its use, 

 measurements over its frequency range could be made in a few minutes 

 time, whereas corresponding measurements by the more precise null 

 method required three to six hours. Of course the time intervals cited 

 do not include time occupied in setting up and adjusting the apparatus. 



Test Amplifier and Experimental Results 

 Test Amplifier 



The stability criterion indicates three distinct conditions of in- 

 terest, one of which is unstable, the other two being stable. The un- 

 stable condition (1) is that in which the transfer factor curve encloses 

 the point (1, 0). Two stable conditions are those in which (1, 0) is not 

 enclosed by the curve, but in which (2) the curve crosses the zero phase 

 shift axis at points greater than unity, and (3) the curve does not cross 

 the zero phase shift axis at points greater than unity. Condition (2) 

 is of particular interest because while it is judged stable on the basis 

 of Nyquist's criterion it would appear to be unstable on the basis of the 

 older transfer criterion discussed in the first and second sections. 



For test purposes an amplifier was designed which, upon variation 

 of an attenuator in the feed-back path, would satisfy each of the three 

 above conditions in turn. The amplifier schematic is shown in Fig. 9. 



Fig. 9 — Circuit diagram of the feed-back amplifier used in testing the stability 

 criterion. The dashed line indicates the point at which the loop was broken for 

 measurement of the transfer factor. At the left of this line is shown the resistance 

 attenuator provided to vary the gain around the feed-back loop. 



