Some transistors get extremely hot while audio amplifier on idle

[Guille]

I have to repair an audio amplifier with the following [schematics][/https://drive.google.com/file/d/1472ykcHgJB_G0WCX_M8cZDu5MFYMDcJo/view], the amplifiers thermal protection jumps after you start playing music for a couple minutes, testing the amplifier on idle (switched on, no music) I can feel a couple transistors on the last stage (see attached picture or page 3 on the schematic) geting extremely hot to the touch while the others are just warm, gate resistors seem to measure close to range (0.99k) so now I'm wondering if I can simply change the toasty ones for new ones or should they be matched to the others? Also shouldn't this circuit have some sort of current sharing mechanism to prevent this exact situation? If so where is it?

Thank you all in advanced.

[1audio]

The basic idea is that the FETs negative  thermal characteristic will handle the balancing of the current sharing. As they heat up the threshold voltage increases turning them off. It works OK usually. The easiest way to check the FET's is to measure the gate resistor voltage drop. If there is measurable voltage the fet is toast. That amp has a whole lot of FET's. The lack of a driver circuit with that many outputs will compromise what it can do but is typical implementation.

[magic]

As they heat up, the threshold voltage decreases and turns them on harder.
As they heat up, RDS(on) increases and limits current but only at high currents.

These are lateral FETs, so the crossover point is not too bad at ~80mA according to the datasheet. If they are biased above that point, they will tend to balance. The point would be much higher (over 1A typically) for vertical FETs and those definitely require balancing resistors.
The transistors need to be somewhat matched for this scheme to work, but I have no practical experience to tell if new ones are likely to be close enough.

Maybe bias voltage is too high? Do you have any means of measuring the quiescent current? The circuit doesn't look like it makes it easy, though

[Audiorepair]

You could try replacing the thermal compound/insulators on the ones getting hot, just it case it is a thermal transfer to heatsink issue on particular devices.

[strawberry]

lift R16 R30 and connect to common Source
feed constant voltage for each Gate and measure Drain current matching with bench power supply

[Guille]

Hi, so if I understand correctly this would be a way to test each individual MOSFET and see if it has become unmatched? What would be considered a big deviation?

[strawberry]

when that problematic transistor stands out
idle current could be measured on Drain powersupply wire

[edavid]

The problem is that replacement MOSFETs are almost impossible to find, and very expensive.  Forget about doing any kind of matching.

I would try disconnecting the parts that are overheating, and then see if the amp then works properly at reduced power.

(If it's hard to desolder them, you could try just jumpering gate to source on the hot ones, but that might not work if they are damaged.)

Do you have the factory procedure for setting the bias voltage?

P.S. Here's the schematic link as an actual link: https://drive.google.com/file/d/1472ykcHgJB_G0WCX_M8cZDu5MFYMDcJo

[Guille]

Hi, really I  didn't know I could just desolder the faulty transistors, I guess for each transistor I desolder on the "pull side" I should desolder another on the "push side"?

No, sadly this schematics are all I could find, the brand is local to my country and not too famous.

Is there a general way to adjust the bias current? I'm prett sure it should eb done with potentiometer R22

[magic]

R22 it is. More resistance → more bias.

The problem is measuring how much bias there is at the moment, as there are no resistors in series with the FETs.

[Guille]

I see, and I'm probably saying something stupid here but will the bias current be properly set when point A (between both pairs of diodes) is at 0V relative to ground?

[magic]

Point A appears to be connected to the sources of all FETs and the output. It will always be near ground (with no signal), depending only on offset voltage of the input stage.

The diodes are normally not conducting, their purpose is to limit Vgs of the FETs and thus implement a simple hard limit on output current. Also ESD protection for gates.

The current of the VAS (T8, T10) flows through the bias-setting resistors R22, R23. The total voltage dropped there is applied between N-ch and P-ch gates to bias them into slight conduction. Output voltage ("A") needs not be exactly halfway between N-ch and P-ch gate voltages, because the Vgs for the same source current is likely different between N-ch and P-ch.

[edavid]

Hi, really I  didn't know I could just desolder the faulty transistors, I guess for each transistor I desolder on the "pull side" I should desolder another on the "push side"?

No, I wouldn't bother with that.  How many transistors are overheating?

I see, and I'm probably saying something stupid here but will the bias current be properly set when point A (between both pairs of diodes) is at 0V relative to ground?

No, I think U1 is responsible for keeping the output at 0VDC, independent of bias.

It's best to use a distortion analyzer to set the bias current, especially when so many expensive MOSFETs are at stake.

[Guille]

I don't have that :/ but could you point towards some resources where I can learn how to adjust the bias current? I understand this question is probably out of the scope of this post.

Thank you.

[edavid]

I don't have that :/ but could you point towards some resources where I can learn how to adjust the bias current? I understand this question is probably out of the scope of this post.

It's pretty simple:
1. set the bias to minimum
2. connect low distortion signal source, distortion analyzer, dummy load
3. increase bias until the amp meets its distortion specs

[Audiorepair]

Since you have stated you don't have a distortion analyser:

1. set the bias to minimum.
2. connect signal source and dummy load and scope.
3. increase bias until crossover distortion just disappears.

[Guille]

Thank you, I've also seen this done with a sine wave input  at full gain, setting the bias potentiometer so that the output doesn't chop either of the output peaks.

Is this process dangerous for the amplifier?

[Audiorepair]

Thank you, I've also seen this done with a sine wave input  at full gain, setting the bias potentiometer so that the output doesn't chop either of the output peaks.

Is this process dangerous for the amplifier?

Yes it is, don't do that.


Look, the whole idea of bias is to get rid of crossover distortion.  Period.
This is the VERY small amount of distortion that happens around the zero crossing point part of the waveform.

The chances are, in an amplifier the size you have, you would never even notice it if the bias was zero.
This whole bias thing tends to get thrown totally out of all proportion and importance.


The most important thing in your case is to not have too much bias, as that will cause all the output Mosfets to run hotter than they need to.

So, start with no bias and a very low signal, just enough to produce a volt or so of output into a dummy load.
if it looks a bit squiffy around the crossover point, increase the bias till the sine wave looks good.
Then check with a few more volts of output to see then how it looks.


And that's all there is to it.
You have eliminated the crossover distortion, which is the whole point of biassing.

[strawberry]

design amplifier basics

[CaptDon]

Knowing that we can easily find mosfets with at least 30 amps of capacity each it really seems like that amplifier is of poor design. The debate goes on about source swamping resistors, needed or not with fets, but if I had that many fets in parallel I would chose to use swamping resistors even if they were a mere .1 ohm each. It was common years ago to see up to ten bipolar transistors on each rail, 20 total, but when entering the mosfet years usually two per side was the max. If they wanted more capacity than two 30 amp mosfets per rail could handle they went to Class-D switching and the same pair-per-side could do over 1Kw at 4 ohms per channel with very minimal heat sinking. Also in Class-D they didn't really need to be concerned about load sharing because the parallel fets were driven into full saturation or completely turned off. If you have one or two bad actors in the amplifier you are servicing I would suspect the ones running hot may have a bit of internal leakage either from age or abuse such as overheating from being played full blast in the hot sun at an outdoor gig. The design of your amplifier also tends to be biased at AB1 rather than B which means there will be heat at idle. When set to pure class B there tends to be distortion near zero crossing with fet output stages so they increase the idle current by moving toward class AB1 or AB2.

[edavid]

design amplifier basics

Kind of useless when OP's amplifier doesn't have any source resistors 

[edavid]

Knowing that we can easily find mosfets with at least 30 amps of capacity each it really seems like that amplifier is of poor design.

Those are switching MOSFETs.  It's just about impossible to find any linear MOSFETs any more, but it would be pointless to make them with a 30A rating, because there's no package that can dissipate enough power.

But yes, it does seem that small source resistors would have been a good idea.