Audio power amp has large DC offset in output - SOLVED!

[hummusdude]

I built this power amp design that I found online and it has been working very well for a few months. Recently it developed a - 18 VDC bias on the output, as well as 1.36 Vppk, 120 Hz ripple. The rectifier diodes in the power supply get hot quickly, and the current through the 8 ohm resistive load is 1.7 Amps DC. So a ton of power in the DC offset.

My first thought was to look at the power supply output since the issue is in both channels, because the ripple is 120 Hz, and because of  the sawtooth shape of the ripple. All that seems like a filter cap problem. But the power supply is putting out a clean -27 VDC with virtual no ripple, which is to spec for the design.

So it would seem the problem is in the power amp boards. Since it's in both sides I have to wonder if I shorted something while working on it that caused a spike in current to the LM 3886 chips. I've seen op amps fail in this way...where the output gets pegged to a rail or something similar. Or perhaps I should look at the caps on the power amp boards? But if it's not the power supply it's seems really odd that the ripple would be at 120 Hz.

I'm working on attaching a schematic for one of the power amp channels, a schematic of the power supply and a screen shot from my scope taken with the output attached to an 8 ohm resistive load. I keep getting an error message when I attach the files even though the format is okay and the total size is less than 1MB.

[hummusdude]

Here's the schematic for the power supply.

[bdunham7]

But the power supply is putting out a clean -27 VDC with virtual no ripple, which is to spec for the design.

Is this tested with the load connected and putting out the fault current?  What does the positive supply measure under the fault condition?

[hummusdude]

Here's the scope shot of the output waveform.

[fcb]

The zener in the MUTE circuit? Can you try it without, in-case you are putting it in some weird half-way state.  Schematic on p7 of the datasheet looks like you could end up in some middle bias area?

How does it perform without a load attached? What if you put a very light loadf on the o/p?

[hummusdude]

Yes, I tested with 8 ohm resistive load. The positive supply measures 25.2 VDC during fault condition. Negative supply is at -22.2 VDC. Would that be sag due to the high current issue?

[bdunham7]

Yes, the sag and the sawtooth ripple are probably simply a result of the supply being overloaded.  Something has biased your amplifiers all the way negative.  Are the inputs (audio and signal ground both) completely disconnected?

EDIT:  Can you scope the power supply rails with the load connected (fault condition)?

[hummusdude]

With no load attached output is -27VDC and +27VDC.

I would have to de-solder the zener to get it out of the circuit. I'd like to rule out everything else before I pull the boards and do that.

about the zener...I just noticed that it is forward biased in the schematic, but I thought zener's had to be reverse biased to work as a voltage regulator. The datasheet says pin 8 needs 0.5mA to turn off the mute function. Not sure why the designer thought a zener would be useful here. With R5 at 10k it seems like V- would have to sag to - 5VDC before the current to pin 8 would be insufficient to keep the chip turned on.

[hummusdude]

yes, both are disconnected.

[bdunham7]

I would have to de-solder the zener to get it out of the circuit. I'd like to rule out everything else before I pull the boards and do that.

You can just bypass the zeners for test purposes.  I'm wondering if you've had a rectifier fail open (or even short) in your power supply.  Perhaps check those next.

[madires]

Looking at the schematics I don't see any common ground for the positive and negative rail in the power supply section. Same for signal ground and power ground in the amp section???

[hummusdude]

I would have to de-solder the zener to get it out of the circuit. I'd like to rule out everything else before I pull the boards and do that.

You can just bypass the zeners for test purposes.  I'm wondering if you've had a rectifier fail open (or even short) in your power supply.  Perhaps check those next.

Oh of course. I can just jump the leads of the zener to remove it from the circuit and see if that changes anything.

As for the rectifier diodes, not sure how that might produce the symptom. Can you explain?

[hummusdude]

The zener in the MUTE circuit? Can you try it without, in-case you are putting it in some weird half-way state.  Schematic on p7 of the datasheet looks like you could end up in some middle bias area?

How does it perform without a load attached? What if you put a very light loadf on the o/p?

By load I assume you mean the load on the output, yes?

EDIT: Just for clarity, I measured the output of the power supply with no load attached to the output of the amp. I guessed it would be okay to do this since there is no signal on the input. The output of the power supply was a clean +- 27 VDC. The symptom shows up when I connect an 8 ohm resistive load to the output of the amp, also with 6 ohm speakers (that's how I discovered the problem). I could try 4 ohms on one channel but I don't have any other high-wattage resistors handy to go lower.

[bdunham7]

As for the rectifier diodes, not sure how that might produce the symptom. Can you explain?

Under no load you will get a nice clean 27VDC even with a blown (open) rectifier, since you'll still have half-wave rectification.  Under load you'll see problems.  That should be easy to check. And by 'light load" I think fcp means something like 100 ohms--see what you get for an output that way with no inputs connected.  It seems very strange that both channels would do this if it isn't power supply related.  You may have a strange combination of issues, like a negative power supply failure causing the amps to go in and out of mute mode.

[floobydust]

I would suspect an open ground, there's at least four per channel that have to be done right. I include reverse diodes across the PSU rails, in case one of the two fuses blows or one rail goes reverse polarity (on power up or down) so it doesn't damage the IC's.
A long-term big DC offset will damage the NFB capacitor C4, you can put a diode(s) across it to protect it or use a bi-polar part.

[hummusdude]

As for the rectifier diodes, not sure how that might produce the symptom. Can you explain?

Under no load you will get a nice clean 27VDC even with a blown (open) rectifier, since you'll still have half-wave rectification.  Under load you'll see problems.  That should be easy to check. And by 'light load" I think fcp means something like 100 ohms--see what you get for an output that way with no inputs connected.  It seems very strange that both channels would do this if it isn't power supply related.  You may have a strange combination of issues, like a negative power supply failure causing the amps to go in and out of mute mode.

That makes sense. If the output were half-wave rectification would that give a 60 Hz ripple?

I tested the ouput of the amps with a 250 ohm load and got some strange behavioiur...see the attached images. Interesting that the right and left channels both have high frequency oscillations but amplitude and freq vary. With this load seeing some negative offset but not as much as under higher load.

[fcb]

You’ve got a Zobel network on the o/p and nothing looks wildly out on the feedback network etc.

The HPF cap on the feedback network looks a bit big, but probably not enough to be an issue.

The HF oscillation could be down to layout OR could be your mute circuit bouncing around - did you try linking it straight to -ve?

You keep talking about your PSU as a possible source - you are overthinking things.  If you can get a +v, -v and 0V from it and you have a bit of local capacitance on the LM3886 you’ll be fine at this stage.

[hummusdude]

You’ve got a Zobel network on the o/p and nothing looks wildly out on the feedback network etc.

The HPF cap on the feedback network looks a bit big, but probably not enough to be an issue.

The HF oscillation could be down to layout OR could be your mute circuit bouncing around - did you try linking it straight to -ve?

You keep talking about your PSU as a possible source - you are overthinking things.  If you can get a +v, -v and 0V from it and you have a bit of local capacitance on the LM3886 you’ll be fine at this stage.

Well, there were several things pointing to that in the beginning. Mostly I'm just responding to others asking about the power supply.

Here's a scope shot with the zener jumped. I left the 10k resistor in series to limit current to 2.7 mA on the mute pin. To me it looks less like oscillation and more like a cap being charged and discharged. The time constant for the soft mute circuit (R5 and C5) is 1 sec so if it's related to that it's not jumping out at me.

I don't think the layout is the issue because it has been working fine for a few months and I have not changed the layout.

[hummusdude]

Looking at the schematics I don't see any common ground for the positive and negative rail in the power supply section. Same for signal ground and power ground in the amp section???

Here's the layout of the power supply. I think this addresses your question.

[madires]

Yes, it does!

[hummusdude]

This is just odd enough that I wanted to run this by the forum. In order to eliminate the power supply as a possible source of the problem once and for all, I connected V- directly to an 8 ohm load and turned it on long enough to get a scope shot. With the horizontal scale = 10 ns, I get 800 mV of ripple in a square wave?? I'm about 99% certain that this must be an artifact of my scope...siglent SDS 1104X-E (100 MHz, 1GSa/s). Perhaps it is quantizing the ripple because the frequency is too high for the front end?

[bdunham7]

I've no idea what you are trying to see in that scope shot at 10ns/div.  Everything you are seeing is an artifact of how you have set it up and it doesn't indicate anything useful.

The amplifier section should reject almost any conceivable noise in the power supply (OK--not at 100MHz, but....), so if all of the grounds are indeed securely connected together and you have anything reasonable on V+ and V-, the amplifier should not be going all the way negative nor oscillating.

The 120Hz ripple you saw is what you would always see if you overload a full-wave linear power supply like this.  The only reason any of us were suggesting looking at the power supply was in case the amplifier mute circuit or some other part was not properly designed or constructed or had failed and was unstable enough that it might be affected by a power supply deviation that would ordinarily not affect a properly set up amplifier.  In my case, the only reason to go down that road at all was because you stated that the amplifier had worked properly for a while and then failed in both channels for some unknown reason.

If you have relatively clean and balanced power under no load and small load conditions, but you still have that oscillation or beating, the problem is not the power supply.  I would start checking all of the components, but really I think the most likely cause here is that you have somehow blown up your LM3886 chips.  They're supposedly well protected and you don't have a lot of voltage, so I'm not sure exactly how that would happen.  The big mystery would seem to be why the amplifier feedback circuit does not respond to it going full negative.  Since it doesn't do this under no load, it probably isn't a shorted output.

You might start by connecting more channels of your scope right to the LM3886--to both inputs, the output and the mute pin.  Seeing those simultaneously (at a slower timebase please!) would be very helpful.  Also, if you can, measure R1, R2, R3 and R4 in circuit and check their related capacitors, especially C4.

[xavier60]

Also use only one ground reference point for all measurement. The 0V junction of the large filter capacitors is a good point.
Voltage readings of all IC pins while in the fault condition might be useful.

[floobydust]

LM3886 amplifiers kits do fall into audiophool territory, some ripoff parts out there like the filter capacitors being hollow and tiny crap parts inside, as fakes.
If there is unexplained mains ripple on the rails, or the amplifier is oscillating, make sure the rail capacitors are what they claim to be. Next to check is the layout and grounding.

Otherwise, If OP's outputs are stuck at the rail, then both IC's are dead. Looking at nsec sweeps will not bring them back and I've seen Rigols add nonsense to their traces especially at the noise floor, so don't bother with the microscope.

Measure the DC voltages at each IC pin and post them here.

[andy3055]

Check the voltages directly on the pins 1,5 for +ve and pin4 for -ve. They should be identical in the number, if you know what I mean. If so, you might be having a failed LM3886.

In case you need, here is the link to the data sheet: https://www.ti.com/lit/ds/symlink/lm3886.pdf