Auto Bias boards for class AB

[baudouin0]

I can see how these work. They measure the average cathode current through a resistor and slowly steer the grid voltage until this meets the correct value. This is OK in the class A part of the output curve.

When you get to the class B part then this current becomes quite non linear and there is large potential for the bias to become incorrect. I can see that this can be improved by clipping the cathode voltage (through a series resistor) to about twice that of the DC operating point.

However I still can see it going a bit wrong for large highly asymmetrical audio waveform. I guess they don't occur long enough for the bias to drift very much.

Another method might be to hold the bias voltage when the audio waveform is large (i.e. one valve in cutoff).

Anybody experience of these?

 

[SpreadSpectrum]

I guess they don't occur long enough for the bias to drift very much.

I have experience with clipping autobias circuits for Class AB output stages and this is correct.

The final test I did was to blast real music at ear-bleeding levels and watch what the bias voltage did. It would slowly vary ~100mV on KT88s that were running at ~-35V grid bias. It was very subsonic and low in amplitude. It didn't cause me any concern since I think there are other areas in the amp that are much less perfect, and I love that bias is adjusted perfectly every time I power up the amp. New tubes are plug-and-play.

Now if you were a perfectionist, you could make a microprocessor-based system that adjusts bias only in periods of musical silence and holds whenever signal is present. But I don't think it buys you much over a good clipping autobias circuit.

 

[Osvaldo de Banfield]

You can also try Back Bias. This is done passing trough the cathode resistor not only the final tubes current, but also the previous stages currents. So, the bias approaches fixed bias, mixed with self bias topology. Take a search to the Radiotron's book freely available in the web.

 

[baudouin0]

Cheers I will take a look at the book.

100mV is very stable and much much better than any open loop bias. Also the grid bias resistor can be larger (helps a lot with the driver voltage) as there's much less danger of thermal runaway.