Increasing bias of 140W class AB amplifier for larger class A region?

[mike7877]

One of my auxiliary amps is an Arcam A32 140WPC class AB amplifier.

I believe its class A operation is about normal, extending to something around 0.05-0.1W output into 4-8 ohms (ie. the point at which both transistors begin to be used for output).

Beyond running a little warmer, is there any danger to increasing the bias for 1-2 watts output?

Not the A32, but a lower power version of it.

The A32 has an even larger heatsink, and, like the model pictured above, the holes in the case line up with the fins of the heatsink... so, effectively, the heatsink is just sitting in an open room (the unit doesn't even begin to feel warm until pretty heavy use)

 

[MaxwellsEq]

Without knowing the circuit and intent of the designer (such as open-loop gain, feedback design), it's tricky to know if there's a danger to this. Assuming it's push-pull, in general, more bias leads to both transistors being on longer, so small increases in bias can have bigger impact than you would expect. Similarly, there's more standing current in resistors and driver circuitry and the PSU will be handling more current than normal. So the heat doesn't just come from the output devices - it's distributed throughout the case.

If you don't mind risking it, the best approach would be to measure before and after and then increase the bias in small steps, whilst measuring and checking temperatures.

 

[mike7877]

Without knowing the circuit and intent of the designer (such as open-loop gain, feedback design), it's tricky to know if there's a danger to this. Assuming it's push-pull, in general, more bias leads to both transistors being on longer, so small increases in bias can have bigger impact than you would expect. Similarly, there's more standing current in resistors and driver circuitry and the PSU will be handling more current than normal. So the heat doesn't just come from the output devices - it's distributed throughout the case.

If you don't mind risking it, the best approach would be to measure before and after and then increase the bias in small steps, whilst measuring and checking temperatures.

Cool, I thought it would probably just be about thermals, especially for an increase to only 1-2W.

I've got a decent IR thermometer which can measure components down the size of a 1/4W resistor, so it should be good! I'll keep an eye on those components to make sure nothing cooks at idle. I know the transistors will fine because I've tested the amp continuously where it draws the most power at the least efficiency, and temps stabilized in the 60s.

Whatever voltage I add to the bias is what is taken away from the new peak voltage output, right - because 0 is effectively moved to one side by that amount? (while the voltage is also added to the other side, the waveform needs to be able to be symmetrical)

 

[AnalogSteph]

You better be able to verify that your plan actually does performance any good - because that is not exactly a given. With higher than optimum bias, you may see an improvement in the Class A region but worsening beyond that. For an EF type output stage, you're generally looking at optimum bias for 13-26 mV across each emitter resistor. It is common to err on the low side because output transistor junctions heat up a lot faster than the bias spreader can react.

What is certain is that higher bias reduces transistor SOA and increases the risk of things going wrong should the output accidentally get shorted.

 

[sergeauckland]

I remember reading an article showing distortion plotted against bias current that clearly showed an optimum point, which is where I expect your amplifier is biased. Both above and below that, distortion was worse, (albeit mitigated by the amplifier's overall feedback) with the added problem of reduced SOA as Steph mentioned above for increased bias.

I don't see any benefit in adjusting the bias outside the range the manufacturers recommend. It won't sound better, and may not last as long.

S

 

[mike7877]

You better be able to verify that your plan actually does performance any good - because that is not exactly a given. With higher than optimum bias, you may see an improvement in the Class A region but worsening beyond that. For an EF type output stage, you're generally looking at optimum bias for 13-26 mV across each emitter resistor. It is common to err on the low side because output transistor junctions heat up a lot faster than the bias spreader can react.

What is certain is that higher bias reduces transistor SOA and increases the risk of things going wrong should the output accidentally get shorted.

I have an interface with sufficient resolution to measure performance before and after.

I won't be leaving it like this permanently, most likely.
edit; So I'll measure it before hand too, see what it's at to start so I can put it back. I thought 7mV was recommended (in memory from a project before)- those were older transistors though, maybe they were different. edit2: in any case, the thing is around 15 years old and it's never been serviced afaik, so it's probably due. The service manual isn't [readily] available online, so I'll go by your recommendation


The amp has over-current protection - a relay disconnects power (the same one that delays output for the second immediately following powerup). I'll be extra aware though

 

[mike7877]

I remember reading an article showing distortion plotted against bias current that clearly showed an optimum point, which is where I expect your amplifier is biased. Both above and below that, distortion was worse, (albeit mitigated by the amplifier's overall feedback) with the added problem of reduced SOA as Steph mentioned above for increased bias.

I don't see any benefit in adjusting the bias outside the range the manufacturers recommend. It won't sound better, and may not last as long.

S

I remember reading about an optimal point for AB operation too, but that shouldn't be relevant in my case because I'm not going to be using the amp in AB mode- just low power A. It might occasionally slip into AB, but I only intend to run it up to a watt or two max, so there will be no optimal crossover point because there won't be crossover! lol