The AWA B92 chassis I'm repairing is now complete. The first question is, what's the ideal quiescent current for the 2N2955 output transistors? The left channel's conducting about 7mA (it varies a bit, which is a worry), and the right channel's not conducting at all.

I would expect the base to draw 7mA The transistor is good for 15A PNP. Check wiring.

http://pdf.datasheetcatalogue.com/datasheet/unisonic/2N2955.pdf

But the quiescent current should be as low as possible 7mA sounds about right.

If you can, use a scope and adjust for lowest current without crossover distortion. using a very low level signal.

Remember my tip about running it without the output transistors? It should still work at very low power. Get it working properly like that and then add the output transistors. A 24 volt lamp in series with the DC to the amp while testing will save you burning anything out.

OK, the right channel's working now, it was a bad transistor socket. But both channels tend to run out of control after a few minutes, with the current going up to 200mA. I used the three silicon diodes for bias of the drivers as you suggested Ian, but now I've changed it back to two diodes plus a 1K trimmer so I can keep a possible runaway under control. At the moment the problem is the BC547 NPN drivers overheating badly. They get too hot to touch. Even if I changed the NPN to the original AS147 or BC147 they'd still be hot. The PNP drivers are staying cool, though. Just to see, I changed one output pair from silicon 2N2955 to germanium OC26, and the corresponding NPN driver stopped overheating. So perhaps I should go with germanium for output as the amp had originally?

I have also run the circuit without the output transistors, and both channels have near enough to the right voltages and are a close match. The sound quality seems OK too.

(This post corrected, wrong drivers were identified.)

Not sure why without having a good look at the circuit but if the OC26s work and you have them, I'd go with that.

Maybe you 2n2955 are a bit low in beta. The spec for 2n2955 and 2n3055 is pretty loose in this regard.

Alternatively, put heatsink flags on the drivers.

Another thought...

If the 2N2955 collector is O/C that would cause a hot PNP driver.

The flange of the 2N2955s is thinner than that of the OC26s. Maybe the screws aren't making reliable contact.

No, the screw contact is good. Actually it was the NPN drivers overheating, I misread the circuit and made a mistake in my post. Corrected now. I think it's the extra bias requirement for silicon transistors that's overheating the NPNs. If I wind the trimpot back they cool down, but that turns off the output transistors.

I'd like to go with the OC26s but I've only got two, plus an odd AD143. So if anyone's got a spare germanium PNP TO3, it would help me get this job finished. Otherwise I'll have to get parts from Europe, because the local equivalent is the NTE104 which costs $68 per each.

Would matched pairs of NOS AD149's be any good?

https://www.kevinchant.com/germanium-output-transistors.html

Perfect! They're very like the OC26's, so I'd only need one pair.

You might still have a problem if the beta is low, but yes, AD149s should work.

I’ve got a couple of 2N301’s you can have, same gain.
They are not matched but test OK.
Could possibly find some more.
JJ

They should work, lower Ft tho...

Circuit uploaded to Post 4.

Put the 2N2955s back in and fit 39 ohm or 47 ohm in series with the base of TR304, i.e. the lower output transistor.

Set the bias pot so you get 0.05 volts across R334 with no signal.

That should get you a result.

All working properly now, with OC26's on one channel and 2N301's on the other. The last step was replacing some badly drifted carbon composition resistors in the tone circuit to get both channels approximately equal in volume and quality. Thanks for the help, everyone.