NEGATIVE VOLTAGE FEEDBACK AND THE STABILIZED GAIN AMPLIFIER 



subject of Graph 12. C is chosen so that at low frequencies the effect of C 

 and i?2 is negligible, but at 17,500 c/s the reactance of C is small and the two 

 resistances are effectively in parallel. The gain is much reduced. At some 

 intermediate frequency the complex load draws a leading current and thus 

 introduces some phase shift, but this is permissible because the shift in the 

 three coupling circuits is not yet approaching 60 degrees. 



Without going into the matter too deeply it may be said that the effect of 

 correcting circuits such as these is to reduce the amplifier pass-band. There- 

 fore to apply successfully feedback to an existing amplifier it is helpful if 

 the amplifier has a useful gain over a band which is wider than is finally required. 

 This is the reason for the dictum that efforts to salvage a bad ampHfier with 

 negative feedback may be disappointing. 



If the amplifier is a.c. coupled there will be another frequency — a very low 

 one — at which oscillations can take place, due to phase advance in the coupling 

 capacitance. To deal with this we connect across the coupling capacitance a 

 correcting network comprising R^ and C^ {Figure 11.4) producing the high- 

 Reactance of 

 C2 must be 

 negligible 



/?2 



h 



;^i 



Figure 11.4 



pass circuit described in Graph 12. At the oscillation frequency R^ is com- 

 parable with the reactance of C, and C^ must be sufficiently large for its 

 own reactance to be small compared with the resistance of R^. The effect 

 is to reduce the phase shift without increasing the loop gain. An alternative 

 approach to the stability problem is to make B deliberately frequency con- 

 scious, so that B is substantially real within the pass-band but complex with- 

 out it and of such a nature as to introduce compensating phase shifts and/or 

 reduce the loop gain at the frequencies where the amplifier would otherwise 

 tend to oscillate. 



The conclusion to be drawn from all this is that negative feedback over a 

 number of stages is quite comphcated and should not hghtly be undertaken. 

 It is, however, perfectly easy to apply round a single stage because the phase 

 shift in the coupling cannot reach 180 degrees, and a similar effect to overall 

 feedback may be had by a number of small feedback loops, as in Figure 11.5. 

 Let us see how this may be done. 



Suppose we have a triode voltage amphfier, automatically biased, and 

 inadvertently leave the cathode bypass capacitance out {Figure 11.6). The 

 load seen by the valve is any resistance through which the anode current 

 passes, and across which it causes voltage fluctuations. Therefore Rj^ is now 

 as much part of the load as Rj^. Now the input to the valve is the grid-cathode 

 vokage and this is evidently dVin — ^Vr^, but ^Vr^ is part of the total 



166 



