It's been my experience that local feedback (plate-to-plate) actually decreases the driver stage gain, not the output stage gain. The output stage still requires the same amount of drive voltage and it's power output should remain appox equal to open loop
.
Yeah, I don't know what I was thinking yesterday. I am 1200 miles from home and was recovering from food poisioning.
The gain of the circuitry inside the feedback loop should not change as the feedback is applied. The gain reduction takes place at the summing junction where the feedback is applied. This assumes ideal conditions. In a silicon opamp with 100+ db of open loop gain and a linear summing junction things are close enough to ideal for these assumptions to apply.
In a tube amp where open loop gain is limited and the summing junction is not ideal the point where the feedback gets applied may not be a single point. The LTP summing junction does a good job if the signal and feedback are applied to opposing grids. The usual single gain stage with signal applied to the grid and feedback applied to the cathode may see some modulation of the input impedance if grid current is approached. These schemes are however close enough so that we can assume that most of the gain reduction occurs in the input stage where the summing takes place.
In the typical Schade feedback circuit the summing junction is somewhat spread across two stages, both of which can be non linear. Assuming the usual resistor from the plate of the output tube to the plate of the driver, there is a voltage divider that determines the feedback ratio. The "Schade" resistor forms a divider against the parallel combination of the resistor from the output tube grid to the bias supply, the driver plate load resistor, the Rp of the driver tube, and the grid impedance of the output tube.
The two fixed resistors are constant and known. The Rp of the driver is somewhat constant and fairly high in a pentode. The Rp of a triode is fairly low and varies with signal level in a triode. This is why many experts state that Schade feedback will not work with a triode driver. The grid impedance of the output tube should be near infinite until the grid is driven positive, then it approaches zero.
Note:
I have built several working amps with Schade feedback applied to a triode driver despite being told that it won't work. Schade will also work in triode LTP drivers in P-P amps. I made a really nice sounding 3 stage amp with cross coupled "Schade" applied across two stages.
In the ideal world (pentode driver) the summing junction is the plate load resistor and the feedback resistor. In reality the feedback voltage appearing at the plate of the driver tube will affect the operation of the driver tube itself. The output tube will have some grid current, but the effect is minimal until near zero bias is reached. In the positive grid region the feedback and the drive voltage stare to get shunted to ground. In the red board I have observed up to about +20 volts on the grid without undue distortion. Yhis is dependent on choice of driver tube. The best choice so fat is the 6GU5 hexode!
This makes an exact determination of just where the summing junction is under all conditiona nearly impossible. I am sure it will give the simulator fits. I am assuming that LTspice is the simulator here? I have used it and it will simulate the Simple SE perfectly. Many of my other circuits will drive it nuts. I have found it useless whenever positive grid operation or screen drive is attempted. I would think it might have a problem with Schade applied to to a triode driver too. I know it's not the simulator, it is the fact that most tube models were not meant to simulate grid current, screen drive, or simulated George testing.
Theories:
It is possible to isolate the output tube grid effect by using a mosfet follower to feed the grid all of the current it can eat. It is also possible to use a mosfet follower to isolate the feedback summing junction from the driver tube. This would make triode driver use easy. It is possible to apply the feedback to G3 of a dual control pentode like the 6AS6 or a mixer tube like the 5915 or 6BE6. It is possible to apply the feedback to G2 of a conventional pentode. My experience with mixing mosfets and tubes in LTspice has been a mixed bag. I am sure that the dual control stuff won't simulate even if a model can be found. To test my theories a breadboard is under construction. I call it the Grand Unified Theory breadboard or GUT board. Who knows when I will finish it though.
If you are using LT spice and care to share your ASC and INC's I (and others) may have time to play with them in the next few days.