Help on over-heating Op-Amp LM2904

[Rat_Patrol]

I have a 20 year old industrial controller I'm trying to repair. It is responsible for running high voltage solenoids.

When just sitting idle, the LM2904 will run about 15c hotter than a reference board (luckily, I have access to a working controller), and the diode that comes off the DC/DC converter that supplies the VCC is also running warm in comparison, confirming the higher current.

If I replace the LM2904 with a new one, it will work just sitting at an idle (powering some micros that run the whole thing), but as soon as I have it operate some solenoids, it seems to damage the LM2904. I replace it, its GTG again until I ask it to run solenoids.

The chip seems to be responsible for the 80v boost converter circuitry for the solenoids, and also the 5v that runs the micros and other ICs, as that is what I loose when the chip fails and that is what I have traced it out to.

Link to the LM2904 datasheet
https://www.onsemi.com/pub/Collateral/LM358-D.PDF

Thoughts?

[george.b]

Hard to say without schematics.

[Rat_Patrol]

Wish I had schematics.

[fzabkar]

I'm intrigued as to how an op amp can be involved in power supply generation. :-?

Could we see a photo of the PCB?

[Rat_Patrol]

Unfortunately no photo, boss would probably fire me if I started putting photos of the board out there, as we had to sign some rock solid NDAs since a lot of what we do is sub-contract work for other companies.

Here is what I've mapped out of the IC's duties so far. It appears the other half does something similar with the 80v supply.

[TerminalJack505]

My suspicion would be that the cap sitting between the MOSFET's source and the op amp's inverting input is causing the voltage at the input to be pulled beyond the op amp's power supply rails.  I don't think that part is susceptible to latch-up but it can source current from its inputs when they are below the negative supply.  The maximum allowed current is 50mA.

See note 6 on page 4 of the datasheet.

[fzabkar]

My suspicion would be that the cap sitting between the MOSFET's source and the op amp's inverting input is causing the voltage at the input to be pulled beyond the op amp's power supply rails.

But there is a relatively big 10uF capacitor at that point. Perhaps it has dried out?

[TerminalJack505]

My suspicion would be that the cap sitting between the MOSFET's source and the op amp's inverting input is causing the voltage at the input to be pulled beyond the op amp's power supply rails.

But there is a relatively big 10uF capacitor at that point. Perhaps it has dried out?

Good point.  The equipment is 20 years old so that sounds reasonable.

[Rat_Patrol]

I've already replaced the 10uF cap, it was electrolytic, no joy.

I've actually been replacing these components one-at-a-time all day, still keep overheating and popping.

[Rat_Patrol]

Also was able to map out the other half of the that op-amp today.

I guess it doesn't do anything directly with the 80v converter circuit, but once it fails, it drops the VCC to about 2.8v and that must affect the 80v circuitry, which is apparently controlled by a LM2901. Haven't mapped that out yet, but I don't think its the culprit.

[fzabkar]

Would it help to measure the resistances between Ground and each of V-, V+ and Vout, and between Vcc and each of the same pins? You could then compare the results against a good op amp from the same batch. This might point you to the source of the damage.

[Rat_Patrol]

You know, looking at that first pic, it is not unlike a convoluted voltage regulator, doesn't it?

Since VCC is 5 volts, and I know the sensors out to the compressor are supposed to be 5v, is it not just regulating the 24v source (Yeah, its 24v DC, it's for controlling a mobile hydraulic crusher), it appears to just be making a 5v reference for the sensors.

Or am I looking at this wrong?

So maybe half of this op-amp is for making a 5v sensor reference voltage, the other half I'm not completely sure of yet.

[TerminalJack505]

I can't quite make sense out of that first schematic.  Is the node that runs between the MOSFET's source and the 47k resistor in parallel with the 0.01uF cap actually GND?  It doesn't seem correct the way it is shown.

Have you checked (with a scope) the output of that op amp to see if it is oscillating?  The 0.001uF cap between the output and the inverting input may be bad in that case.

[Rat_Patrol]

It should all be accurate. GNDD is ground. There are 2 grounds: regular battery(-) ground and just a static grounding type ground (mounting screws "pads" have a ground circuit that grounds to the case).

[Rat_Patrol]

Out of Q? (FQP19N20) pin 3 is 5 volts.

[fzabkar]

IC304A in the second diagram is attempting to maintain its V- input at 2.5V, ie the same as the voltage at V+, which is set by the 2.5V precision reference IC (LT1009ID). Since V- is derived from Vcc via a 1:1 potential divider (1K:1K), then IC304A is operating as a 5V regulator for Vcc.

The regulator's error voltage drives TP1005 via Q1000 and Q1001. I don't understand why Q1001 would be powered from the same Vcc voltage that it is trying to control. :-?

The regulator's error voltage drives the Gate of Q1000 via TP1004. Therefore Q1000 must be the pass transistor for the linear regulator. Its Drain pin (TP1005) must be the raw DC input, while the Source pin is the regulated 5V output (Vcc).

Which voltage powers the op amp? Is it powered directly from the battery (24V)?

[TerminalJack505]

You know, looking at that first pic, it is not unlike a convoluted voltage regulator, doesn't it?

Since VCC is 5 volts, and I know the sensors out to the compressor are supposed to be 5v, is it not just regulating the 24v source (Yeah, its 24v DC, it's for controlling a mobile hydraulic crusher), it appears to just be making a 5v reference for the sensors.

Or am I looking at this wrong?

So maybe half of this op-amp is for making a 5v sensor reference voltage, the other half I'm not completely sure of yet.

Yes, this seems right.

[fzabkar]

Is Q1001 drawn correctly? ISTM that it should be a P-channel MOSFET, otherwise its body diode would be shunting the raw DC supply in forward biased mode. Edit: Even if it were P-channel, I still don't see the point of connecting its D & S directly across a supply rail, unless it is intended as a crowbar. :-?

[fzabkar]

You know, looking at that first pic, it is not unlike a convoluted voltage regulator, doesn't it?

Since VCC is 5 volts, and I know the sensors out to the compressor are supposed to be 5v, is it not just regulating the 24v source (Yeah, its 24v DC, it's for controlling a mobile hydraulic crusher), it appears to just be making a 5v reference for the sensors.

Or am I looking at this wrong?

So maybe half of this op-amp is for making a 5v sensor reference voltage, the other half I'm not completely sure of yet.

Yes, this seems right.

I agree.

The op amp is comparing Vcc against V_sensor, so the target voltage for V_sensor must be 5V. The error voltage drives the Gate of FQP19N20 (the pass transistor) via the PNP/NPN pair.

The 1.8 ohm resistor is a current sensor which turns on the upper PNP transistor when the current reaches 333mA (= 600mV / 1.8 ohms). This then turns on the NPN transistor, which in turn reduces the gate drive to the pass transistor.

[Rat_Patrol]

Yes, all N channel MOSFETS

[fzabkar]

Would it make sense, for troubleshooting purposes, to add Schottky rectifiers between ground and each of the V- and V+ inputs, and between those same inputs and the op amp's supply pin? This would help to eliminate latchup as a failure mode.

You might also like to consider powering the op amp from clean DC source to eliminate that as a point of failure, if only for troubleshooting purposes.

[fzabkar]

Yes, all N channel MOSFETS

But then there would be a forward biased body diode across the supply. That can't be right.

https://www.st.com/content/ccc/resource/technical/document/datasheet/66/f2/4e/c4/ee/dc/45/e4/CD00002848.pdf/files/CD00002848.pdf/jcr:content/translations/en.CD00002848.pdf

[TerminalJack505]

I can't really say what the problem might be.  Here are my suggestions...

• Check that resistor on the output of the op amp in your first schematic.
• Check the cap going from the output to the inverting input.
• Put a scope on each of the op amp outputs and check to see if they are oscillating.  So far as I can tell from your schematics, they shouldn't.
• Put a scope on each of the inputs and make sure they don't exceed the rails of the op amp.
• Test one of the failed ICs out of circuit to see if just one of the op amps is bad.  If so then focus on the parts of the circuit that affect that op amp.

[SMdude]

Perhaps there is nothing wrong with that section of the circuit at all as when you replace the opamp it works nicely and regulates the voltage just fine.
Then when you engage the 80v circuit, perhaps there is a voltage transient on the opamps supply that damages the opamp. The only way to catch this is with a storage scope monitoring the opamp supply. Perhaps there is a bad diode somewhere?
Given the high resistance of the resistors at the opamp output and inputs, I don't think it would be likely that the opamp is being damaged on its input/output pins.

[duak]

What happens to the LM2904s?  Do any of the inputs or outputs short to GND or Vcc or go open?

Here's a link to some design considerations for the LM2904: http://www.ti.com/lit/an/sloa277/sloa277.pdf

Two things come up on a quick perusal:

p.4 -1.4: Vcc max might be as low as 26 V - is this exceeded?

p.8: if the inputs are brought much below 0 V, the output can experience phase reversal.  I don't see how that can happen here unless V_SENSOR goes way negative.