Amp's power MOSFET died. Can I just remove it?

[DaveC]

In my 400W audio power amp, a power MOSFET died, I think. The speaker output
from one of the channels is 5.5v DC, whereas it's supposed to be zero.

Configuration is 3 parallel MOSFETs connecting the output to the plus rail,
and 3 connecting it to the minus rail.Rails are plus and minus 68v, nominal,
no ripple.

Can I just remove the failed plus-side MOSFET from the circuit? If I can do
this, should I also remove one of the minus-side ones from that channel, as
well?

Eventually I'll look into exactly what has failed (MOSFET, driver FET, or
some passive component), but right now I just want to get the amp's output
back to zero and use it right away.

How can I easily determine which of the three MOSFETs has failed?

This amp is a little-known brand (VSP Labs) from a small, long-closed
company, so no schematics or repair documentation is available.

Suggestions?

Thanks,

 

[Mike Harrison]

In my 400W audio power amp, a power MOSFET died, I think. The speaker output
from one of the channels is 5.5v DC, whereas it's supposed to be zero.

Configuration is 3 parallel MOSFETs connecting the output to the plus rail,
and 3 connecting it to the minus rail.Rails are plus and minus 68v, nominal,
no ripple.

Can I just remove the failed plus-side MOSFET from the circuit? If I can do
this, should I also remove one of the minus-side ones from that channel, as
well?

Eventually I'll look into exactly what has failed (MOSFET, driver FET, or
some passive component), but right now I just want to get the amp's output
back to zero and use it right away.

How can I easily determine which of the three MOSFETs has failed?

This amp is a little-known brand (VSP Labs) from a small, long-closed
company, so no schematics or repair documentation is available.

I doubt its a MOSFET as (like most semis.) they almost always fail short, so you would see more than
5V.
By all means try disconnecting though, can't see it would do much harm.

 

[Luhan Monat]

In my 400W audio power amp, a power MOSFET died, I think. The speaker output
from one of the channels is 5.5v DC, whereas it's supposed to be zero.

Dave,

I used to work fixing this kind of equipment (many moons ago). Here's
what I suggest.

Disconnet the unit from the AC power - do not simply turn it off!!!!

You need to disconnect at least 2 of the 3 leads on the mosfets and use
and ohm meter to check for leakage. This will find you the bad part.
Be carefull to keep yourself grounded to the unit's chassis or you may
kill one of the mosfets.

Visually determine if each bank is basically just paralleled - source,
gate, and drains. If so, you can just remove one bad one and run the
unit at reduced power.

Here is the important part: Rig a 40 watt 110 volt light bulb in series
with the AC power feed each time you test the unit after replacing
parts. If the light comes on full, you have other problems to go after.
If it comes on bright but settles down to less than full intensity,
you can then run full (AC) power to test it out.

I used to have one of these wired permenantly into my testbench along
with a switch to bypass the light bulb. Anything I worked on was
plugged in there. Low tech, but very useful.

 

[Walter Harley]

In my 400W audio power amp, a power MOSFET died, I think. The speaker output
from one of the channels is 5.5v DC, whereas it's supposed to be zero.

Configuration is 3 parallel MOSFETs connecting the output to the plus rail,
and 3 connecting it to the minus rail.Rails are plus and minus 68v, nominal,
no ripple.

Can I just remove the failed plus-side MOSFET from the circuit? If I can do
this, should I also remove one of the minus-side ones from that channel, as
well?

I recently repaired a couple of high-power MOSFET amps. Take a look to see
if there are low-value resistors in series with each MOSFET's drain, before
they are paralleled. If the sources and drains are paralleled without any
resistors, then you have to replace all three MOSFETs, with matched units;
otherwise one MOSFET will hog all the current and blow up. (For some
reason, for a period of time, designers had got it into their heads that
MOSFETs were immune to this problem, and so they stopped using the
precautions they had used for BJT's. Unfortunately, this was a mistake.)
The sad news is that to get three matched MOSFETs, you're going to have to
buy a dozen or so and then hand-pick the best-matched ones.

Make sure that the gate resistors are still intact (and the rest of the gate
drive circuitry). Sometimes when a MOSFET fails it takes out the gate
resistor too; and then, when you connect up the replacement without fixing
the gate resistor, the new MOSFET has a dangling gate, meaning it waits a
random period of time and then turns full on, thus blowing itself up again.
Ask me how I know.

You may find that it is hard to get exact replacements to your MOSFETs: some
of the ones used in older audio amp designs are no longer commonly sold.
For one set of MOSFETs, I had excellent results buying equivalents from
Profusion PLC, in the UK (even though I'm in the USA, they were very
inexpensive and low-hassle, and shipped incredibly quickly).

IF you're lucky enough that just one of the MOSFETs is blown, then yes, you
could just remove it and operate at lower power. Indeed, there's probably
not even any point in removing one from the other side.

Note that if you're seeing DC on the output, then either the amp has no
protection circuitry or the protection circuitry is also fried. In neither
case would I feel very comfortable about connecting that amp to any speakers
I owned...

 

[Jeff]

I doubt its a MOSFET as (like most semis.) they almost always fail short, so you would see more than
5V.
By all means try disconnecting though, can't see it would do much harm.

Agree with all
Unless its loaded and driven to full power, then good
luck.
If leaky fet is suspected I would test out of circuit
very rare for them but I have seen a few.
Jeff

 

[Ban]

DaveC wrote:
|| In my 400W audio power amp, a power MOSFET died, I think. The
|| speaker output from one of the channels is 5.5v DC, whereas it's
|| supposed to be zero.
||
|| Configuration is 3 parallel MOSFETs connecting the output to the
|| plus rail, and 3 connecting it to the minus rail.Rails are plus and
|| minus 68v, nominal, no ripple.
||
|| Can I just remove the failed plus-side MOSFET from the circuit? If I
|| can do this, should I also remove one of the minus-side ones from
|| that channel, as well?
||
|| Eventually I'll look into exactly what has failed (MOSFET, driver
|| FET, or some passive component), but right now I just want to get
|| the amp's output back to zero and use it right away.
||
|| How can I easily determine which of the three MOSFETs has failed?
||
|| This amp is a little-known brand (VSP Labs) from a small, long-closed
|| company, so no schematics or repair documentation is available.
||
|| Suggestions?
||

Dave,

what gives you the idea that (only) one of the transistors "died"? Do you
have any measurement gear to verify this?
I would advice you to bring this amp to a musical instruments repair shop,
because your skills do not seem to be adaequate for a repair. :-(
Or let it Rest In Peace.

ciao Ban

 

[[email protected]]

I recently repaired a couple of high-power MOSFET amps. Take a look to see
if there are low-value resistors in series with each MOSFET's drain, before
they are paralleled. If the sources and drains are paralleled without any
resistors, then you have to replace all three MOSFETs, with matched units;
otherwise one MOSFET will hog all the current and blow up. (For some
reason, for a period of time, designers had got it into their heads that
MOSFETs were immune to this problem, and so they stopped using the
precautions they had used for BJT's. Unfortunately, this was a mistake.)
The sad news is that to get three matched MOSFETs, you're going to have to
buy a dozen or so and then hand-pick the best-matched ones.

Would it be practical/possible to add an equalizing resistor in
series with each drain in the event he can't get matched MOSFETs
to allow using an unmatched replacement? Would that be
effective?

 

[Walter Harley]

Would it be practical/possible to add an equalizing resistor in
series with each drain in the event he can't get matched MOSFETs
to allow using an unmatched replacement? Would that be
effective?

It probably depends on the PCB layout, etc. In the amp I was fixing it
didn't seem like I'd be able to do a professional-quality job. If I was
just trying to get something cobbled together to limp through a tour or some
such, I'd give that a shot.

And if I were designing one from scratch I would definitely include them.

 

[DaveC]

what gives you the idea that (only) one of the transistors "died"? Do you

have any measurement gear to verify this?
I would advice you to bring this amp to a musical instruments repair shop,
because your skills do not seem to be adaequate for a repair. :-(
Or let it Rest In Peace.

Well there's the black-and-white options laid bare.

I prefer to *learn* and *strech* my skill set, sometimes. I have much
understanding of electronics, but limited understanding of amplifier
circuits. I think I'll use this amp as a learning device, rather than just
throw $$ at it or throw it at the landfill.

News groups such as this are full of very talented, experienced techs and
engineers, many of whom are willing to help an "amplifier-challenged" tech
such as myself.

 

[DaveC]

Suspect that it is probably not, as I had first assumed, the power MOSFET.

Symptoms:
1. This channel's MOSFETs do not get warm at all while amp is idling with no
input. Other channel MOSFETS get plenty warm under identical conditions.
Presume that bias on MOSFETS for defective channel is incorrect.

2. Amp's output waveform is severely distorted (see below).

Poking around the input circuitry I put scope probe on emitter of 2N4033.

With 0.1v p-p input:
http://home.covad.net/~peninsula/Amp/Wave1.jpg

With 0.15v p-p input:
http://home.covad.net/~peninsula/Amp/Wave2.jpg

While I was watching this, the waveform increased in its distortion over just
a 2-second period to this:
http://home.covad.net/~peninsula/Amp/Wave3.jpg

With 0.25v p-p input:
http://home.covad.net/~peninsula/Amp/Wave4.jpg

Does this look familiar? Is it typical of a common type of failure?

Without schematics (this is a 20-year-old amp from a long-since defunct
company), I'm left to poke around and try to follow the signal path looking
for good signal shape and distorted shape, which I can use to guess at
suspect components.

Any help in getting to the bottom of this would be greatly appreciated.

Thanks,

 

[Jeff]

Suspect that it is probably not, as I had first assumed, the power MOSFET.

Symptoms:
1. This channel's MOSFETs do not get warm at all while amp is idling with no
input. Other channel MOSFETS get plenty warm under identical conditions.
Presume that bias on MOSFETS for defective channel is incorrect.

2. Amp's output waveform is severely distorted (see below).

Poking around the input circuitry I put scope probe on emitter of 2N4033.

With 0.1v p-p input:
http://home.covad.net/~peninsula/Amp/Wave1.jpg

With 0.15v p-p input:
http://home.covad.net/~peninsula/Amp/Wave2.jpg

While I was watching this, the waveform increased in its distortion over just
a 2-second period to this:
http://home.covad.net/~peninsula/Amp/Wave3.jpg

Those are weird waveforms. I would be looking in the feed back circuit
first. I'd suspect an electrolitic cap gone bad, especally after 20 years.

 

[Dario2]

Will you do the same measurements in oscilloscope in DC mode ? Seems a
component is burned, but i need a simple diagram of output stage and DC
snapshots to judge.
Dario2

 

[[email protected]]

Suspect that it is probably not, as I had first assumed, the power MOSFET.

Symptoms:
1. This channel's MOSFETs do not get warm at all while amp is idling with no
input. Other channel MOSFETS get plenty warm under identical conditions.
Presume that bias on MOSFETS for defective channel is incorrect.

2. Amp's output waveform is severely distorted (see below).

Poking around the input circuitry I put scope probe on emitter of 2N4033.

With 0.1v p-p input:
http://home.covad.net/~peninsula/Amp/Wave1.jpg

With 0.15v p-p input:
http://home.covad.net/~peninsula/Amp/Wave2.jpg

While I was watching this, the waveform increased in its distortion over just
a 2-second period to this:
http://home.covad.net/~peninsula/Amp/Wave3.jpg

With 0.25v p-p input:
http://home.covad.net/~peninsula/Amp/Wave4.jpg

Does this look familiar? Is it typical of a common type of failure?

Without schematics (this is a 20-year-old amp from a long-since defunct
company), I'm left to poke around and try to follow the signal path looking
for good signal shape and distorted shape, which I can use to guess at
suspect components.

Any help in getting to the bottom of this would be greatly appreciated.

Thanks,
--
DaveC
[email protected]
This is an invalid return address
Please reply in the news group

Dave - Great pictures! I particularly like the .1 v input
resultant - looks like 1 kHz with "notches". It's almost
a shame to fix it!

You may have thought of this already, but could you
lift the 2N4033 emitter and then drive it into a
power resistor instead of the MOSFETS? I'm thinking
that might isolate the source of the "notch generator".
Right now, its not clear if the Mosfets are pulling
down the signal, making the notch, or the notch is
actually on the input of the 2N4033.

It dies look like the Mosfets are shorting the output
on the emitter whenever it reaches .1 volt. Driving the
emitter into a resistor would eliminate that possibility.

What does the output from the Mosfets look like at
..1 v input? Identical wave but larger amplitude?
If not, and the output from the Mosfets appears
only when the notch appears, then it would appear they
are biased wrong, as you mentioned. Can you compare
bias levels with the other channel?

Someone else mentioned electrolytics - they are famous
for causing audio distortion, as I'm sure you know.
I would love it if you could scope it down to the failing
'lytic - if that is the problem - rather than just
replacing. I kick myself now for the ones I've replaced
for distortion problems - I never looked at them closely
with the scope. Maybe if I had done it as carefully
as you set your scope and generator, I'd have seen
notches, too. What I've seen looks more like your last
picture, and I've never bothered to look at it closely.

I hope you'll keep posting on this - it is very
interesting. I've never run into a waveform like
yours in the .1v picture.

 

[Walter Harley]

Suspect that it is probably not, as I had first assumed, the power MOSFET.

That's what I suspect too, since they usually fail open or short, and you
mentioned a DC offset on the output of several volts, which corresponds to
neither.

Since this is a stereo amp, another option you have is to compare signals
from side to side.

If you can identify the feedback components, then you can open the circuit
(by unsoldering a resistor, e.g.) and run it without global feedback, which
will probably help considerably in identifying the problem.

Some observations on your waves:

First, I think you said this had 68V rails. But in your extremely-clipped
image (0.25v in), it looks like the negative trace is clipping at about 18V.
So it's not anywhere close to clipping in the output stage - this is
happening somewhere earlier.

Second, assuming that the indicated vertical scale is correct (that is,
you're using a 1x probe, or an indicating 10x probe), I notice that the
positive clipping happens at about 1.2v, or two diode drops. That might be
a clue as to what's happening - e.g., when a Darlington configuration
somewhere is finally biased on, it's making the rail collapse. Or some
such.

Third, it's odd that in the high distortion wave (0.25Vpp in), you get all
the way up to 10v before the positive side clips. (Is it possible that the
earlier waves were with a 10x probe and this one is 1x??)

Hard to know more without a schematic. If it were me, I'd take the time to
trace the schematic. Not only will that aid your repair job, it will also
be a good learning experience in itself.

Did this failure happen suddenly, or gradually? That would help tell you
whether to look at electrolytics versus semiconductors.

 

[Luhan Monat]

Suspect that it is probably not, as I had first assumed, the power MOSFET.

Symptoms:
1. This channel's MOSFETs do not get warm at all while amp is idling with no
input. Other channel MOSFETS get plenty warm under identical conditions.
Presume that bias on MOSFETS for defective channel is incorrect.

2. Amp's output waveform is severely distorted (see below).

Poking around the input circuitry I put scope probe on emitter of 2N4033.

With 0.1v p-p input:
http://home.covad.net/~peninsula/Amp/Wave1.jpg

With 0.15v p-p input:
http://home.covad.net/~peninsula/Amp/Wave2.jpg

While I was watching this, the waveform increased in its distortion over just
a 2-second period to this:
http://home.covad.net/~peninsula/Amp/Wave3.jpg

With 0.25v p-p input:
http://home.covad.net/~peninsula/Amp/Wave4.jpg

Does this look familiar? Is it typical of a common type of failure?

Without schematics (this is a 20-year-old amp from a long-since defunct
company), I'm left to poke around and try to follow the signal path looking
for good signal shape and distorted shape, which I can use to guess at
suspect components.

Any help in getting to the bottom of this would be greatly appreciated.

Thanks,

The second image looks familiar. Could be parsitic oscillations at the
positive peak. If the frequency is way higher than the scope can
handle, then it can appear as a notch.

 

[DaveC]

The second image looks familiar. Could be parsitic oscillations at the
positive peak. If the frequency is way higher than the scope can
handle, then it can appear as a notch.

Tek 2465 4-channel, 200MHz bw. I've run the timebase way up and don't see any
HF.

 

[DaveC]

If you can identify the feedback components, then you can open the circuit
(by unsoldering a resistor, e.g.) and run it without global feedback, which
will probably help considerably in identifying the problem.

I'm going to replace all the electrolytics and then see what's what. Then
I'll compare each channel, point by point, and see if that tells me anything.

Some observations on your waves:

First, I think you said this had 68V rails. But in your extremely-clipped
image (0.25v in), it looks like the negative trace is clipping at about 18V.
So it's not anywhere close to clipping in the output stage - this is
happening somewhere earlier.

I may have misled you, in that I was looking at the outputs with the scope in
AC-coupled mode. I presume that all the waveforms are biased by the 5.5 vdc
that I see on the output of the amp with no input.

Second, assuming that the indicated vertical scale is correct (that is,
you're using a 1x probe, or an indicating 10x probe), I notice that the
positive clipping happens at about 1.2v, or two diode drops. That might be
a clue as to what's happening - e.g., when a Darlington configuration
somewhere is finally biased on, it's making the rail collapse. Or some
such.

Third, it's odd that in the high distortion wave (0.25Vpp in), you get all
the way up to 10v before the positive side clips. (Is it possible that the
earlier waves were with a 10x probe and this one is 1x??)

Only a 1x probe.

Hard to know more without a schematic. If it were me, I'd take the time to
trace the schematic. Not only will that aid your repair job, it will also
be a good learning experience in itself.

Only as a last resort.

Did this failure happen suddenly, or gradually? That would help tell you
whether to look at electrolytics versus semiconductors.

I think it was gradually (IIRC). I noticed distortion (this was *years* ago,
btw) and presumed the PS filter caps (huge dudes you can seein this photo:
http://home.covad.net/~peninsula/Amp/Surge1.jpg )

I replaced all the PS filter caps and the distortion remained. It may well be
the smaller electrolytics. I'll know tomorrow.

Thanks,

 

[DaveC]

OK, it's not electrolytic caps. I've replaced all dozen or so of the small
caps and the symptoms are unchanged.

Now I'll isolate each stage by lifting one of the legs of each device and
driving it into a resistor, as someone mentioned earlier in the thread. This
will help to isolate the symptom.

Any suggestions while I do that would be gratefully appreciated.

Thanks,

 

[DaveC]

I may have misled you, in that I was looking at the outputs with the scope
in AC-coupled mode. I presume that all the waveforms are biased by the
5.5 vdc that I see on the output of the amp with no input.

Disregard this statement. The waveforms I posted were taken at the collector
output of an intermediary stage in the circuit. It is not the output stage,
however the output does mirror these waveforms.

It is true that I had the scope in AC-coupled mode, so no DC component was
measured. At my next measurement I'll check the DC value, too.

Thanks,

 

[Ban]

DaveC wrote:
|| OK, it's not electrolytic caps. I've replaced all dozen or so of the
|| small caps and the symptoms are unchanged.
||
|| Now I'll isolate each stage by lifting one of the legs of each
|| device and driving it into a resistor, as someone mentioned earlier
|| in the thread. This will help to isolate the symptom.
||
|| Any suggestions while I do that would be gratefully appreciated.
||
|| Thanks,
|| --
|| DaveC

Dave,
again an unuseful idea, you were on the right track before(with the bias!),
just first measure the gate voltage of both the upper and lower Fets (if
they are FETs at all) and then check the bias voltage pot(s), maybe replace
it, or adjust it to a value of 100-200mA idle current, or better check how
much that is on the other side. To mesure this current take out the fuses of
each side if there are any in the 65V line or just put the ampmeter into
this line. Start with a high range not to blow the fuse in the meter.

Since the feedback from the output will distort the balance of the input
stage, it was not the right place to measure. Go backwards from the output
stage in DC-mode with the scope, and compare each measurement with the same
point of the working side.
Repairing means a bit more than replacing parts, finding the reason of the
failure.
good luck

ciao Ban