burning mosfets


2002-08-10 4:21 pm
I've built a AB class mosfet output amp using two pairs of IRF640/9640 as output devices.
The strange thing is that the amp has worked, and suddenly it stopped working and smoke appeared in the output mosfets, and they didn't had any shortcircuit beetwen GS or DS or GD!
What can be happening, and how can I solve it.

PS: I have built this amplifier before but with less power and only one pair at the output something like three weeks ago, and it is working fine, the only changes are some different resistor values the aditional pair and some zener protection diodes.

Need lots of help.
Diogo ,
you say that it ' was working , then it stopped '

and then the magic smoke leaked out.....

what were you running the amp into when it failed, and how hard were you driving it?

what is the voltage output of the psu ?
IRF640's are pretty tough, but FET's don't cope with overvoltages.

you mention that you added extra output fets, did you use degenerative resistors to 'balance' the current ? and did you match the fets in each rail?

if the output stages weren't matched then one (or two) device(s) would be dissipating more power than the rest -> pop!:xeye:

the only other failure modes that I can think of offhand would be due to oscillation at high frequecies ( which in my experience tends to smoke filter components and driver stages ) or possibly gate punch-through leading to output stage breakdown.

did the driver stage survive the failure, and did you have any evidence of instability?



2002-08-10 4:21 pm
It was stable, no doubt that. The problem was only on the output, the rest of the circuit is ok. I used 0R10 source resistors.
I don't know how to match mosfet, becouse I'm only a begginer so I used different dissipators for them. Could it be a mistake in the zener's protection value?


2002-08-10 4:21 pm


  • diogomosfet2.gif
    14.2 KB · Views: 613

Nelson Pass

The one and only
Paid Member
2001-03-29 12:38 am
I don't see any Gate protection for your IRF610 coming
off the diff pair. When you clip the amp, the gate can
be driven to supply rail values, and will often break at
40 volts or so. Maybe a nice 10-15 V Zener from Gate to
V- would be a good idea.


Burning mosfets

olá Diogo ;)

the way that the pot (10k Ohm)
that change the bias setting in your schematic
is implemented, is dangerous...if for a malfunction of the pot the wipper make no contact with resistitive part of the pot...
the output mosfets will see a lot (maybe too much) bias across they!

Maybe can be that ,that have kiled them!


Jorge Santos

PS:Um abraço deste lado do Atlantico!


2002-08-10 4:21 pm
But I wasn't adjusting the trimpot when it has burned!
Even so do you think it might have been that?
I'll change it anyway, the actual board has melted....:hot:

What else you guys think I could do to improve this amplifier?
And what do you think about the IRFs I used.
They are cheap but the sound was great.
What other cheap mosfets I could use to my next projects, the IRFPs and 2SKs are too much expensive for me, I need to be sure that I'll not blow them to spend money....

And there is any other possibility?

Nelson Pass

The one and only
Paid Member
2001-03-29 12:38 am
Yeah, always put the bias pot Base to Emitter. Why not
try using a Mosfet like an IRF610 or so for the bias device?
That way the gain will only need to be about 2 to get
the correct value, instead of 12, and the bias adjustment
will be a lot less sensitive.



Paid Member
2002-09-25 11:01 am
I see another possible issue here:

The addition of the second pair of transistors is what is causing the problem. Remove them first. Maybe they're 'overloading' the bias stage. I would change from the BC 546 and implement probably a Darlington pair, to maybe add to what Nelson said, and increase the sensitivity and enhance the ability of the bias stage to drive two output stages.

The other is the transistors themselves. They should be rated to take full rail voltage, not just one half of it, in an ideal situation. When one half of the output is locked into conduction and will show full voltage across the other half.