Today's experiment - I added a BD135 wired as diode as per #135 - This makes quite a difference to the thermal stabilty vs bias. I tested this (not exhaustively) as Panson did elsewhere in the thread - I ran a 1khz sine wave into a dummy 8R load at ~50w or so, getting the heatsink warm, and then withdrawing the signal and watching what happens the bias. The bias started off with a heatsink ~18degC at ~ 90mV ( measured emitter to emitter, so 0r44, and this is high, more on this in a bit) let it rise to 40degC and within a minute or two of shorting the input, the bias drops right back down to ~75mV. As the heatsink cools back down, the bias settles out at the original 90mV. This is repeatable over 3 tests on 2 channels, and seems really stable. I think the fact that the bias measurement when the heatsink is hotter is lower than when cold is safe?
The fly in the ointment here is that the lowest (without changing anything else on the board) I can set the bias on the output stage is just under 90mV. As far as I can see what this means is that I need to fit a resistor somewhere, most likely R12 or R13 to allow for adjustment with the added diode.
I think this is a good advance on the instability I was seeing up until now. I just want to see how the temperatures fare out now in real life use. If I can figure out the bias range adjustment, then all the better. Any obs very welcome.
Right, so I have a better handle on this now, but don't have a tried and tested stable solution yet.
Basically, the diodes alone with a single adjustment pot are not good enough to give stability at higher bias levels. The problem here (to these ears, and this will annoy some people) is that the amp clearly sounds better at higher bias levels through the TT outputs, even if 22mA is the lowest distortion. Give them more like 100mA each and 50degC and the amp is a different beast. However at these levels, the diodes/pot arrangement doesn't seem to work very well. For example, at 19degC heatsinks, bias at 50mA, and as the heatsinks heat up this rises to 100mA or so. So if you adjust when its all hot, then all is OK enough, but I don't know what would happen if the heatsinks went hotter - I haven't been able to get them higher than 52degC. So I have no reason to think that in certain conditions thermal runaway would indeed happen.
The good news of course is that the amp sounds great but I would like a little more safety. The experiments continue....
Basically, the diodes alone with a single adjustment pot are not good enough to give stability at higher bias levels. The problem here (to these ears, and this will annoy some people) is that the amp clearly sounds better at higher bias levels through the TT outputs, even if 22mA is the lowest distortion. Give them more like 100mA each and 50degC and the amp is a different beast. However at these levels, the diodes/pot arrangement doesn't seem to work very well. For example, at 19degC heatsinks, bias at 50mA, and as the heatsinks heat up this rises to 100mA or so. So if you adjust when its all hot, then all is OK enough, but I don't know what would happen if the heatsinks went hotter - I haven't been able to get them higher than 52degC. So I have no reason to think that in certain conditions thermal runaway would indeed happen.
The good news of course is that the amp sounds great but I would like a little more safety. The experiments continue....
I feel like I am constantly putting my ignorance on show here. I have to admit I had zero experience with class AB outputs before going back to look at this thread - years ago when I put the amp together first it was just an assembly job and that was it.
Anyway, with bias at the recommended 22mA or so (I settled around 11mV across emitter to emitter) and some testing back and forth, the biasing seems stable. I did the same test as before - started with 11mV, 20degC heatsink, then ran it to get the heat up to 40-something degC (bias up at 500mV) and then withdrew the signal. Bias drops back to about 8mV and then slowly recovers to 11mV.
From first turn on of the amp, the bias takes about 45mins to come up to 11mV but once there it stays there. When I had that biased much harder, the bias kept creeping up slowly over time and if you left the amp on by accident, undoubtedly it would keep going.
So I have learned that the diodes only work for a limited current range - swamp that current and you lose compensation. Of course thats obvious to you all, but took me a bit to figure it out. Perhaps with a regular Vbe transistor this might work OK. The amp sounds great, but definitely there is something extra to the sound when its biased harder.
Anyway, with bias at the recommended 22mA or so (I settled around 11mV across emitter to emitter) and some testing back and forth, the biasing seems stable. I did the same test as before - started with 11mV, 20degC heatsink, then ran it to get the heat up to 40-something degC (bias up at 500mV) and then withdrew the signal. Bias drops back to about 8mV and then slowly recovers to 11mV.
From first turn on of the amp, the bias takes about 45mins to come up to 11mV but once there it stays there. When I had that biased much harder, the bias kept creeping up slowly over time and if you left the amp on by accident, undoubtedly it would keep going.
So I have learned that the diodes only work for a limited current range - swamp that current and you lose compensation. Of course thats obvious to you all, but took me a bit to figure it out. Perhaps with a regular Vbe transistor this might work OK. The amp sounds great, but definitely there is something extra to the sound when its biased harder.