Tweet
Originally posted by bob p
View Post
But in the general case of a class AB amp, the DC is not constant. "Bias" is the DC current in both sides of the circuit in the absence of signal. When you impose signal, the signal turns one tube more on and one of them more off. At some point, the "off" tube shuts down entirely and the "on" tube carries the rest of that half-cycle. There is a complex half-wave-rectified signal running through the "on" tube until the signal shuts it down on the other polarity of signal, where the other tube then has its own complex half-wave rectified signal.
The DC current from the power supply for all but class A biasing is a constant(-ish) DC at idle, and increases with increasing signal. That's one reason that AB biasing is even used: it makes better use of both the power supply capability and the power dissipation of the output tubes. So the trick in figuring out a bias level in the face of signal swings is to somehow subtract out the signal-caused current from the static bias current. Worse, what you're getting from the preamp is a signal voltage, and what you're driving is a non-resistive speaker load. So the output current (which is what you're trying to control with bias) varies in a way that the input voltage doesn't necessarily represent. Simple subtraction and scaling don't necessarily get you to where you want to be. Even if it did, you're doing the delicate operation of subtracting two quantities that are nearly equal to sniff out the difference you're trying to control. It's an error-prone process, just on that basis.
And you want the DC bias to be always active, because simply biasing the tubes at no signal isn't what bias is for. It's for making the dynamic changeover from one signal polarity to the other happen in as smooth and non-objectionable a way as you can make it, in the face of all possible kinds of signal traversing both the top and bottom edges of the handover between polarities.
So the task is to compute what the handover current is for wildly varying currents in the output stages, not just to separate AC from DC. It's a special DC buried in that mess. There are several ways to attack that. I tried a few of them that I could find from research or dream up on my own. Some worked, some didn't.
At some point, having a couple of schemes work, I took a reality moment to think about what my objectives were. After all, I was after draining the swamp, not collecting alligator hides. Bias creep is real, and it gets worse with worse preparation of output tubes, as the residual positive gas ions poison things. It can't be dealt with easily if you have to haul the amp to a tech to get a re-bias. The obvious solution is to make the amp bias itself, right?
That's a good approach, if you can do it well, and not make the amp otherwise less usable. A flakey self bias circuit is worse than the old model of haul-it-to-the tech. My cut on it was that if you can make rebiasing so simple that any guitarist that's sober enough to tell colors apart do it between songs in a set, you can actually allow the guitarist to keep his own tubes running, even in the face of bias creep.
It's an alternate approach. Simply making bias happen on its own, with no user intervention, would be good. But it has to really, really work; even then, this forum will be entertaining posts about how to mod or bypass the autobiasing circuit to get the auto-amps back to fixed bias.
IMO getting a handle on the bias problem is exactly what needs to be done -- especially when the tubes we have available today are all substandard, as the problem can only get worse.
Comment