Greetings,
How horribly sounding will 809's in class B push-pull be with feedback or should I be asking how hard would it be to implement feedback in a class B push-pull amp.?
What was called Class B (-9V G1, 20 mA/tube idle, 3.7W dissipation on first grids) when 809 was designed now is called class AB2, I believe.
How horribly sounding will 809's in class B push-pull be with feedback or should I be asking how hard would it be to implement feedback in a class B push-pull amp.?
Yes! The 809 data sheet (has anyone else looked at it today?) shows "Class B" as having 40 mA zero signal plate current; that's 20mA per tube. This is today's definition of class AB. The 3W grid power is at max signal. Grid power is very small at idle.
And it's a great example of the old assumption that class B amps are driven into grid current. Notice thay don't call it "Class B2" but there is grid current.
Today we would call it class AB2 with small overlap.
Crossover distortion:
The turn-off glitch occurs whenever one side of a switching push-pull amp turns off. Even in AB, where there is overlap, there is still a turn-off glitch. The glitch does seem to be smaller if one side turns off after the other has turned on, but it's still there and is an effect of OPT leakage inductance.
Build it!
I would not worry about the definition of class B, built the 809 amp with whatever idle current and positive grid drive makes sense for your chosen load line, and tweak it from there, just like any other amp. The only thing these tubes need vs. a "regular audio triode" is provision for high plate resistance and grid current drive (MOSFETs are our friends). Treat it like a pentode and apply local feedback. It can probably sound amazing and be really pretty with the bright filament and all...
Cheers,
Michael
Then there is "magic" Schading, which uses current sensing Schades and a differential driver to subtract them. This forces the tubes to make a smooth current handover. No abrupt cutoff is possible then.
Thanks for the info Michael. Would you have any schematics of circuits I could try?
Thanks,
Ray
Re: Michael
"The turn-off glitch occurs whenever one side of a switching push-pull amp turns off. Even in AB, where there is overlap, there is still a turn-off glitch. The glitch does seem to be smaller if one side turns off after the other has turned on, but it's still there and is an effect of OPT leakage inductance."
I think there may be a couple of ways to fix that leakage glitch. (other than Mac OT or Circlotron or Twin Coupled schemes) Seeing as how the leakage L causes the plate voltage to swing up wildly positive after the tube cuts off. If one added "nominal gain" (0 dB NFdbk, ie, Closed Loop gain set to match the tube's normal gain into the load) Schade plate feedback resistor(s), the tube would be held on just enough to absorb the glitch (ie, correct tube output voltage maintained). Of course a regular Schade NFdbk should work too.
Then there is "magic" Schading, which uses current sensing Schades and a differential driver to subtract them. This forces the tubes to make a smooth current handover. No abrupt cutoff is possible then. The crossover region is widened from what the bias and intrinsic tube curves overlap would normally obtain.
And it's a great example of the old assumption that class B amps are driven into grid current. Notice thay don't call it "Class B2" but there is grid current.
The turn-off glitch occurs whenever one side of a switching push-pull amp turns off. Even in AB, where there is overlap, there is still a turn-off glitch. The glitch does seem to be smaller if one side turns off after the other has turned on, but it's still there and is an effect of OPT leakage inductance.
Looking back at the SS version in Broskies schematic, imagine the input signal doing a slow ramp from -Max thru crossover to +Max, and visualize what the driver transistors are seeing as feedback on their emitters. For a class A stage, the feedback would be a smooth signal thru the whole ramp, but for a class AB stage the feedback does a smooth ramp up until the crossover region is reached. It then speeds up thru the crossover region, then returns to the normal smooth ramp. This exagerates the crossover error for the feedback, which then tries to smooth it out. Crossover region ends up widened and smoothed. 20 dB (100 X) of improvement in crossover dist. according to Broskies simulation. I've found at least 4 different patents so far on this in SS variants. Edmond Stuart is also using a variation on this in his TMC ppm dist. type SS amps over on the SS forum.
Wavebourn:
"I would go much easier and more straightforward: LTP output stage. "
But that would mean going class A then, OK for Winter I guess.
Um.., you mean those little fuzzies toward the peaks?
Look more like some small transient parasitic HF oscillations, but the trace is too unfocused to tell for sure.