FEEDBACK AMPLIFIER DESIGN 451 



obtained through the shielded input and output transformers T\ and 

 T2. The three stages in the ix circuit are represented in Fig. 23 as single 

 tubes. Physically, however, each stage employes two tubes in parallel, 

 the transconductances of the individual tubes being about 2000 micro- 

 mhos. The principal feedback is obtained through the impedance Z/j. 

 There is in addition a subsidiary local feedback on the power stage 

 through the impedance Zr. This is advantageous in producing a 

 further reduction in the effects of modulation in this stage but it does 

 not materially affect the feedback available around the principal loop. 



The elements shown explicitly include resistance-capacitance filters 

 in the power supply leads to the plates and screens, cathode resistances 

 and by-pass condensers to provide grid bias potentials, and blocking 

 condenser-grid-leak combinations for the several tubes. In addition to 

 serving these functions, the various resistance-capacitance combina- 

 tions are also used to provide the cutoff characteristic below the useful 

 band. The low-frequency asymptote is established by the grid leak 

 resistances and the associated coupling condensers and the approach 

 of the feedback characteristic to the asymptote is controlled mainly by 

 the cathode impedances and the resistance-capacitance filters in the 

 power supply leads to the plates. The principal parts of the circuit 

 entering into the ij.^ characteristic at high frequencies are the interstage 

 impedances Z\ and Z2, the feedback impedance Z/j,^^ the cathode im- 

 pedance Zk, and the two transformers. The four network designs are 

 shown in detail in Figs. 24, 25, 26, and 27. 



The joint transconductance, 4000 micromhos, of two tubes in parallel 

 operating into an average interstage capacity of 14 mmf, as indicated 

 by Figs. 24 and 25, gives an/t of about 50 mc. The parasitic capacities 

 (chiefly transformer high side and ground capacities) in the other parts 

 of the feedback loop provide a net loss. At, of about 18 db at this fre- 

 quency. Since the asymptotic slope is 18 db per octave the intercept 

 of the complete asymptote with the zero gain axis occurs about one 

 octave lower, at slightly less than 25 mc. This is a relatively high 

 intercept and may be attributed in part to the high gain of the vacuum 

 tubes. The care used in minimizing parasitic capacities in the con- 

 struction of the amplifier and the general circuit arrangement, including 

 in particular the use of single shunt impedances for the coupling and 

 feedback networks, are also helpful. 



" The relative complexity of this network is explained by the fact that it actually 

 serves as a regulator to compensate for the effects of changes in the line temperature. 

 (See H. W. Bode, "Variable Equalizers," Bell System Technical Journal, April, 1938.) 

 The present discussion assumes that the controlling element is at its normal setting. 

 For this setting the network is approximately equal to a resistance in series with a 

 small inductance. The fact that the amplifier must remain stable over a regulation 

 range may serve to explain why the design includes such large stability margins. 



