weird thing happens on opamp

[Alex R]

I have in one of those 1-unit studio power amps where they lie everything very flat so you can't get at it. If you wiggle the input XLR, that channel crackles, then turns off, and won't come on again till you reboot the amp.

I had a good poke around but I can't find a bad joint etc. Nor can I get the channel back on again any other way than via the power switch.

I watched the speaker protection relays and all the other relays while this was happening, and they're not latching off when the amp goes off.

The XLR goes into a little preamp board which also has on it the speaker outputs. It has two opamps per channel - 5534s. OK, so after the chopsticking failed, I clipped the Fluke to the chassis and went poking around looking for voltage changes. And I found that when I touched the +ve probe to either the -vDC or the -ve input on one of the opamps, the amp did it again - went off, and wouldn't come back on till rebooted. Just happens on those two terminals.

Any ideas?

[Steve Conner]

Sounds like the op-amp is latching up. Maybe it got hurt by ESD to the input, or some piece of ungrounded apparatus. Whatever might be wrong with it, the cure is probably replacement...

Then again, I have seen bad solder joints that don't respond to chopsticking, so maybe the cure is resoldering of all the relevant joints, and then replacement of the op-amp if that doesn't help.

Is the channel balance and distortion OK when it's working?

[Alex R]

Thanks Steve. The opamp's in a socket so I popped a new one in, and also replaced its neighbour, and remade all the joints in the area. Still the same.

Yes when it's running it performs just like the other channel, I've had them both up to the clipping point.

It's really odd how marked the effect is. Give the XLR a firm wiggle and off it goes; touch the vm to the terminal and the same.

It's a Dynaudio A3 by the way.

[Don Moose]

Is there a chance that, by wiggling the XLR, you're flexing the board enough that a power supply contact is hitting GND and crowbarring?

The key symptom here would be that one or both rails would shut off.

The next option would be a flaky socket that the flexing and probing makes open a pin or two. That seems less likely, though.

[Alex R]

Thank you Don. I will check the rails. No I don't think it's a normal contact problem because no amount of moving things around will bring it back - it's firmly latched off once it's off.

[Alex R]

OK I tried that. Both the opamp rails and the main power rails stay the same in the fail condition. Under a healthy signal the opamps have +/-17v, the main rails +/-79v rising to 82 when the output stops.

Just a touch with the voltmeter probe on the +ve input of the 5534 knocks the amp out, as does quite determined wiggling of the input XLR, which leads to a crackle or two before the dropout happens. Yet it can be driven to undistorted full output before this with no sign of drop-outs. And without fail it comes back when you turn it off and on again.

Things are very tight in there and if it wasn't for the voltmeter effect I'd be sure it was a physical short or bad joint. But the vm thing is odd, as is the latching.

I've swapped in new opamps, even tried a TL071, and the effect is still the same.

Feel free to tell me to recheck something. I can take it!

[Don Moose]

Is +ve a power supply pin, or the positive input?

I'm starting to think it's not the op amp but something with the feedback loop. Grasping at straws here, but what voltmeter are you using - could it be beating up the input/fb loop in a way that makes it latch up?

Since changing chips doesn't fix it, and it takes a power cycle to get it to come back, I'm having trouble seeing it as being the socket - that leaves the input and feedback components.

Hope this helps!

[Alex R]

Thanks Don. It's the +ve input to the opamp. I'm using a Fluke 25.

[Don Moose]

Ugh. This seems pretty out there, but if there's a cap across the op-amp inputs, I could see a jolt from the meter stuffing a charge onto that cap and jamming the op-amp output up against a rail hard enough to latch there.

How does wiggling the XLR do that, though? How does the circuit do that on its own from time to time?

Besides, these are relatively old-school op-amps with microamp bias currents. Not the femtoamp bias currents the modern ones have, so even if the cap took a charge off the meter (or something), it wouldn't hold it very long.

I'm stuck, but reasonably sure the problem is around the op-amp, not the op-amp itself.

I don't have an explanation, but I do admire the problem.

[Alex R]

Good one ain't it?

The symptoms are so strong I just have to get it in the end. Wish I had a schematic though.

I will go around the opamp in the fail condition to see if it's sticking to a rail. Maybe I'll try an analogue meter.

Can you explain that microamp femtoamp thing for me please?

[Don Moose]

Can you explain that microamp femtoamp thing for me please?

I'll take a stab at it. An op-amp input isn't quite high-impedance enough that input currents are zero. Older bipolar (non-FET) op-amps have input bias currents in the microampere range. A 1uF (1e-6) cap charged to 1V, being drained at 1uA would take a certain amount of time to decay below 1/2V (too tired to remember/find the right equation). That same cap and charge being drained at 1fA (1e-12) would take 1,000,000 times longer to decay that far.

Op-amps with JFET inputs tend to have the ultra-low bias currents.

So, the mechanism I'm hung up on is unlikely because a 4558 is a good, low noise, non-JFET op amp. It just won't leave all that charge stuck there.

I'd love to see a schematic, too. I have to think there's something in that op-amp's feedback loop that's going bad. Hey - a loop that opens up is going to force the op-amp to one rail or the other based on what the inputs are doing and stay there until the inputs are reversed/equalized.

The bit that's messing with me is that a power-cycle undoes whatever it is. That doesn't usually fix an iffy solder or failing ... anything but a cap.

Wish this would help.

[Alex R]

Thanks Don, it does help a lot. I can check the opamp outputs to see if they're stuck to a rail. I'll get to it tomorrow or maybe later today.

Thanks for the explanation. Modern opamps would be the TLO series for instance? Because they admit no current they won't drain away a problem voltage.

[Don Moose]

Modern opamps would be the TLO series for instance?

Right - a TLO is a JFET input op-amp, and there are some ultralow-bias versions out there.

Because they admit no current they won't drain away a problem voltage.

That's the way I understand it. Theoretically, that makes them very close to ideal because they present virtually no load. Theory and practice are occasionally orthogonal when something doesn't go right.

Best of luck!