444 BELL SYSTEM TECHNICAL JOURNAL 



characteristic with the zero gain axis the calculation is the same as 

 that made previously in connection with Fig. 10. From the zero gain 

 intercept to the junction with the asymptote, where we can say that 

 design control is finally relaxed, there is, however, an additional interval 

 of nearly two octaves. Although Fig. 16 is fairly typical, the exact 

 breadth of the additional interval will depend somewhat on circum- 

 stances. It is increased by an increase in the asymptotic slope and 

 reduced by decreasing the gain margin. 



Relative Importance of Tubes and Circuit in 

 Limiting Feedback ^ 



The discussion just finished leads to the general conclusion that the 

 feedback which can be obtained in any given amplifier depends ulti- 

 mately upon the high-frequency asymptote of the feedback loop. It 

 is a matter of some importance, then, to determine what fixes the 

 asymptote and how it can be improved. Evidently, the asymptote is 

 finally restricted by the gains of the tubes alone. We can scarcely 

 improve upon the result secured by connecting the output plate directly 

 to the input grid. Within this limit, however, the actual asymptotic 

 characteristic will depend upon the configuration and type of feedback 

 employed, since a given distribution of parasitic elements may evi- 

 dently affect one arrangement more than another. The salient circuit 

 problem is therefore that of choosing a general configuration for the 

 feedback circuit which will allow the maximum efficiency of transmis- 

 sion at high frequencies. 



The relative importance of tube limitations and circuit limitations 

 is most easily studied if we replace (7) by 



^„. = 40logxo:^'-^S (10) 



where ft is the frequency at which the tubes themselves working into 

 their own parasitic capacitances have zero gain ^ and A t is the asymp- 

 totic loss of the complete feedback loop in db at/ = /^ The first term 



* The material of this section was largely inspired by comments due to Messrs. 

 G. H. Stevenson and J. M. West. 



' I.e., ft = TT^ , where Gm and Care respectively the transconductance and capaci- 



tance of a representative tube. The ratio -p? is the so-called " figure of merit " of the 



tube. The analysis assumes that the interstage network is a simple shunt impedance, 

 so that the parasitic capacitance does correctly represent its asymptotic behavior. 

 More complicated four-terminal interstage networks, such as transformer coupling 

 circuits and the like, are generally inadmissible in a feedback amplifier because of the 

 high asymptotic losses and consequent high-phase shifts which they introduce. 



