Amp temperature imbalance

[LearningSomethingToday]

I am working on an amp that has 8 output transistors per channel. The 8 output devices are on two separate heat sinks with two complimentary pairs on each heat sink. The two heat sinks are tightly sandwiched to form a single channel module. The amp has two of these sandwiches to form a stereo pair. One of the sandwiches has a heat problem.

I have one of the four heat sink halves getting uncomfortably warm. All the output devices on that heat sink test ok with a meter. The two heatsinks for each module only have thermal feedback on one half. The half that is getting warm has no feedback and I feel will eventually go into thermal runaway.

There is a 1 ohm, 5 watt ceramic resistor on each output transistor. I wanted to see what was occurring across each of these resistors and I get readings all over the place. I read 15mv, 15mv, 53mv, 15mv across the resistors on one half heat sink and 9mv, 56mv, 9mv, 50mv on the other.

I am assuming this would be the bias voltages. Shouldn't these numbers be in close proximity to each other?
From anothers perspective, what am I looking at here?

This amp does have a DC balance adjustment but does not have a bias adjustment. It uses a thermal compensator to adjust bias. I would think that the compensator would make all the output devices read similar bias voltages so I am scrathing my head.


Thanks

 

[kellys_eye]

Unless all the transistors are 'matched' then you WILL see different voltages across the emitter resistors - that's what they're there for!

Similarly, because the transistors are NOT matched the heat dissipation will vary. If you want a 'balanced' heat distribution then you would have to match all the transistors accurately, including their thermal performance characteristics. Impractical at best so it's a problem you live with.

 

[LearningSomethingToday]

Unless all the transistors are 'matched' then you WILL see different voltages across the emitter resistors - that's what they're there for!

Thank you for clarifying this. I have heard the term matching transistors in the past in audio forums but never really understood what that meant. So I imagine ideally transistors are biased just slightly above turning them on to keep them from producing switching /crossover noise but it would be difficult to keep all eight transistors on at the same level if they are not matched. Is that simplified view correct? Is it typical to have wide margins between devices as I am seeing?

I am still sure there is a problem somewhere in my circuit. The heat sink I am talking about is magnitudes different in temperature from the others. Even between the two heat sinks on the same channel (LEFT) there is dangerous difference in temperature. If I leave it idle for a 1/2 hour one half of the channel is warm and the other cannot be touched for more than a few seconds. Also to clarify there are are two heat sinks per channel. The output devices are split up on two heat sinks due to space constraints. This two channel amp has four heat sinks in total. The two halves are of each channel are joined at the main board via cables. Both halves "should" be seeing the same power rails, audio input, bias from main board etc.

 

[kellys_eye]

You may have a shorted transistor on the 'hot' heat sink side. Check the Vbe (voltage between base and emitter) of all the transistors on the hot side and see if there are any well outside the 0.6V(ish) range.

 

[LearningSomethingToday]

You may have a shorted transistor on the 'hot' heat sink side. Check the Vbe (voltage between base and emitter) of all the transistors on the hot side and see if there are any well outside the 0.6V(ish) range.

I checked the heat sink that runs hot and I read values like .592v, .589v, .550v, .555v. That half of the heat sink has the two drivers on it too. Those read .568v and .613v. I don't know of any of this is unusual and I would appreciate you input.

Then I decided to check the output devices on the cool half to compare. Strangely enough one of the transistors reads 0v vbe.
Would 0v indicate a shorted device? An open device? This amp has a pretty large power supply. I would imagine a shorted transistor would let some smoke out in short order.
Could the cool half be driving the hot condition on the other half?

Thank you for your help.

 

[kellys_eye]

Would 0v indicate a shorted device?

Yes.

A push-pull output stage would normally have 0V at the 'centre point' and it may be that the shorted part is causing the output stage to try to compensate for the 'pulling' of the supply to the other rail.

Change the dud transistor.

 

[LearningSomethingToday]

Yes. Change the dud transistor.

Thank you all for your help. I now have the amp in a more reasonable state. All portions of the amp seem to be warming up within a few degrees of each other if I shoot the heat sinks with a thermal pointer. Left channel was 117F degrees and the right was 112F. Now that they are all hot I am wondering if they are too hot while sitting there idle. I would like to ask one additional question regarding bias. This amp only has pots for DC offset. The bias is set by a fixed resistor (a pot was in the PCB design but eliminated by the manufacturer). The thermal tracking is tracked by 3 diodes in series and a resistor network. (pin 3+4 on bias board; attached JPG)

I want to get a handle on how to measure bias on an amp like this. There are no formal service docs and I only have a schematic. Googling this reveals different measurement methods for different amp designs. Some procedures measure mA and dial a pot in. Some measure millivolts and dial a pot in. I am grappling with the idea of why there are different approaches to the same type of transistor topology.

I am attaching a PDF of the schematic. There are 3 test pins on each channel which are not shown on the schematic. The outer pins connect to the emitters of complimentary transistors. I imagine I would measure millivolts there. I currently measure @ 80 to 90 millivolts there now. The center pin is basically the summed output of all transistors (speaker output) and I am guessing it is used to measure DC offset with respect to gnd.

There is about .6v on each base lead at idle. If I blow heat on the bias diode the voltage slowly drops. I also see the current I measured above move downward. So I know the thermal tracking is doing something. The resistor network R218 and 219 in line with bias diodes and I am thinking tweaking those values would get the bias to the correct level but I need to understand approximately what the bias voltage should be and where to measure it at.

Hope I did not go into over-information here. Thanks for any help