Autobias amps

[Merlinb]

The second edition of the SS guitar amps book I'm working on briefly discusses that circuit (along with other SS stuff working hand-to-hand with tube amps). Here's a screencap:

I believe there's a typo there.
"When current through the tube increases the base voltage increases as well, which in turn shifts the collector voltage more positive"
It should say that the collector voltage shifts negative. (Otherwise the tube would go into meltdown pretty quick...!)

[teemuk]

I believe there's a typo there.
"When current through the tube increases the base voltage increases as well, which in turn shifts the collector voltage more positive"
It should say that the collector voltage shifts negative. (Otherwise the tube would go into meltdown pretty quick...!)

No, it’s not a typo since this is how the circuit really operates:
The higher the cathode current is, the more positive the collector voltage is.

This was actually one of the major problems of the amp: The circuit works well to a point where one power tube fails. When that happens the auto bias circuit continues its job and - like you say - drifts the power amp into a meltdown.

The second version of YBA-200, the YBA-200-2, abandoned the auto bias and replaced it with an ordinary fixed bias setup, which in the end was more reliable.

[Merlinb]

No, it’s not a typo since this is how the circuit really operates:
The higher the cathode current is, the more positive the collector voltage is.

I think you are mistaken. When the cathode current rises, the base voltage increases pushing it closer to the emitter voltage. This causes the transistor to reduce its current conduction, turning OFF more.
Since the transistor forms the shunt part of a potential divider, if it turns off then the bias voltage (emitter) will go more negative, biasing the valve cooler.

[Steve Conner]

Actually I think teemuk is right... It does work the "wrong way round". Notice the transistor is PNP.

In this case, it looks to me like this is not an auto bias circuit at all, but a ghetto Class-AB2 driver. The idea of the circuit must have been to prevent crossover distortion under overdrive, by absorbing the grid current before it can pump up the coupling capacitor.

[Merlinb]

Actually I think teemuk is right... It does work the "wrong way round". Notice the transistor is PNP.

I can see why you might think that, but you are overlooking the fact that the collector is connected to a NEGATIVE voltage (the bias supply), and the emitter is connected to +4.7V. So making the base a little positive pushes the base towards the emitter, turning the PNP more OFF.

(If it makes it easier, imagine the base could be pushed all the way to +4.7V. The base and emitter would be at the same voltage, so the PNP would definitely be OFF.
Now imagine the base could be pulled down to the collector voltage, the PNP would have to be hard ON)

[Steve Conner]

OK, I take back what I said, teemuk is wrong...

In any case, the way the schematic is drawn makes it unclear. Voltage should always become more positive as you move up the page, so PNP emitters should always point towards the top of the page, and the +4.7V rail should not be alongside the -78V one.

That's one of my pet hates, along with drawing the "bottom" tube of a push-pull pair upside-down.

[loudthud]

I think the Traynor circuit was intended to provide balanced bias current to the toroidal output transformer. The problem with all these analog auto bias circuits is the time constant cannot be made long enough to ignore the current variations that occur in the normal operation of a guitar amp. The digital approach seems like the only way around this. As long as you are going to go that far, you should have a digitally controlled warm-up cycle that sets the bias and enforces a warm-up time. Might as well have a countdown/status display.

[Steve Conner]

Well, if the amp has a toroidal OPT, then that's almost certainly what this circuit will be, some kind of balancing servo.

The problem isn't just that the time constants can't be made long enough. The problem is that those same time constants pump up under overdrive and cause blocking distortion. So if you make them "long enough" you get crappy sound. You need to make them shorter to recover quickly, just like you need to make coupling capacitors smaller in overdriven tube stages.

In a Class-AB circuit, the bias servo's time constant also appears as a low-frequency pole in the NFB loop. (It can't tell between bias and signal current, remember? And a balancing servo is worse, since it modifies the signal current directly.) So if you make it short enough to recover quickly under overdrive, then the pole frequency is high enough to destabilize the NFB. It would be conditionally stable, and start motorboating if you played it loud.

These are probably some reasons why toroidal OPTs aren't used more often in musical instrument service: it's hard to get rid of that DC offset, and an EI core made of plain old transformer iron tolerates it better than a toroid of fancy high-permeability silicon steel.