I agree that your second example illustrates your point more clearly and the fact that you successfully built it that way and it works is also a strong point for you.
I followed your advice and simulated the base to collector shunt and the capacitor bypass of the base resistor using the simulation setup that mimics my EF3. Note this is a simulation of just the EF3, without any amplifier. And I also simulated no base resistor for the driver at all (also no caps of course). I'm a bit inconclusive regarding the results.
It is evident that the bypass capacitor reduces phase shift over the B-C shunt. Even better is no base stopper at all, which seems a bad advice however.
In the example I simulated, the 10R base stopper and the 10R+1nF B-C snubber do not significantly reduce the gain peak compared to no base stopper, but significantly worsen phase shift. The bypass cap on the other hand side even increases the gain peak with moderate impact on phase shift compared to no base stopper.
Transient response looks a bit messy for all three options.
I found a good paper about preventing emitter follower oscillation in another thread:
Prevent Emitter Follower Oscillation - Chessman and Sokal
I followed your advice and simulated the base to collector shunt and the capacitor bypass of the base resistor using the simulation setup that mimics my EF3. Note this is a simulation of just the EF3, without any amplifier. And I also simulated no base resistor for the driver at all (also no caps of course). I'm a bit inconclusive regarding the results.
It is evident that the bypass capacitor reduces phase shift over the B-C shunt. Even better is no base stopper at all, which seems a bad advice however.
In the example I simulated, the 10R base stopper and the 10R+1nF B-C snubber do not significantly reduce the gain peak compared to no base stopper, but significantly worsen phase shift. The bypass cap on the other hand side even increases the gain peak with moderate impact on phase shift compared to no base stopper.
Transient response looks a bit messy for all three options.
I found a good paper about preventing emitter follower oscillation in another thread:
Prevent Emitter Follower Oscillation - Chessman and Sokal
Attachments
I examined 3ef in a global loop, open-loop triple is not very useful for me. I'm using fast drivers 2sc4793/2SA1837 (100MHz, 20pf) with 22-47R stoppers.
Important to have low signal impedance at HF (VAS output imp), < 100R. If higher, use zobel 22R + 22pF to the ground at the 3ef input. See post #54
Important to have low signal impedance at HF (VAS output imp), < 100R. If higher, use zobel 22R + 22pF to the ground at the 3ef input. See post #54
Last edited:
This is also my preferred approach, even if slight contrary to the thread title.
My view is the feedback around a non-linear amplifier simply moves (multiplies) the distortion, creating a somewhat dead, small sound.
There was a suggestion by Thorsten Loech as i recall, about the hazards of including the OP Transformer (tube amps) in the overall feedback loop because:
1. At the frequency extremes - where feedback was really needed, the open loop gain was poor anyway.
2. The delay through the transformer with all its parasitics means the signal being fed back was a bit 'out of date'.
I was a bit sceptical if an OPT, left to it's own devices, could give a good, tight, controlled sound, so I tried it out on a Class A amp, loading up driver tube and power tube with feedback, and just letting the OPT take care of itself.
The result was total stability, and a very open, clear sound, that completely sold me on the topology, so I left it like that.
For solid state, I've heard of a series of very short loops being used - 2 transistors or so, rarely, but they did appear to also have a very open sound. My theory of why this 'open' and natural (realistic) sound was a character is this:
Feedback only really works on linear systems, non-linear systems tend to simply move it. By allowing non-linearities to build up over - say - 6 gain stages, and then to wrap a big (and now slow) loop around it, it tends to move a lot of non-linearity at once, in a rather clumsy way.
Anyway, just my 2c, back to the thread
(And Happy New year!)