What causes transistors to fail?

[pucho812]

This is an odd one.
Have a couple discrete opamps. At output in normal working order they have 0vdc at output. Out of nowhere I get high vdc at output, very close to one of the voltage supply rails of the bipolar supply feeding it. If I take the discrete opamp apart and test every component nothing is failed or shorted. All transistors checkout in working order. When I put it all back together and install the discrete opamps back in. They work just fine. That is until they don’t again. I have check everything I can think of as I don’t have a schematic. But all appears ok. No extra hot transistors or anything.
Hard to pin point an issue where no components exhibit failure and all solder joints were reflowed.

 

[Khron]

Feedback resistor and/or associated traces questionable? And does it end up slamming to the same supply rail, or randomly to either one?

 

[Brian Roth]

This is an odd one.
Have a couple discrete opamps. At output in normal working order they have 0vdc at output. Out of nowhere I get high vdc at output, very close to one of the voltage supply rails of the bipolar supply feeding it. If I take the discrete opamp apart and test every component nothing is failed or shorted. All transistors checkout in working order. When I put it all back together and install the discrete opamps back in. They work just fine. That is until they don’t again. I have check everything I can think of as I don’t have a schematic. But all appears ok. No extra hot transistors or anything.
Hard to pin point an issue where no components exhibit failure and all solder joints were reflowed.

 

[pucho812]

Feedback resistor and/or associated traces questionable? And does it end up slamming to the same supply rail, or randomly to either one?

Traces are good. Feedback resistor seems fine. Always slams close to negative vdc at output

 

[zamproject]

Happen to me more than once...

Last was just few days ago, hybrid design on a summing/master amp on a 24V with VCC/2 reference
The ref was dancing a lot with obvious crack and white noise at output
Everything measure fine including all transistors junction when pulled out, but with one of the transistor out (BC179) everything gone and stable supply...
Replaced it, all good.

Now that I think about, I suspect it always happen with (old) TO18 package ?!? as junction measure fine in diode test I suppose it's related to surrounding substrate which is deteriorating after many decades, as the fact that one leg (collectors ?) is usually hooked to the metal can, offering a large contact to the substrate then to other junctions ends. If isolation is deteriorating, some parasitic parallel resistance may cause issue...

Just a guess...

 

[pucho812]

It’s a better guess than I have. I ran the unit with discrete in question for a day without issues. So it went back into service only to have the dc thing comeback, luckily the unit has installed an auto mute so when dc is present in the audio path it mutes the output.

All transistors are to92 package.

 

[JohnRoberts]

Sometimes you can force a fault with spray cold, or using an iron to heat up an active device.
==
Can you probe circuit node voltages after it goes wonky? That should steer you toward what device or component is misbehaving.

JR

 

[pucho812]

Sometimes you can force a fault with spray cold, or using an iron to heat up an active device.
==
Can you probe circuit node voltages after it goes wonky? That should steer you toward what device or component is misbehaving.

JR

I can. That’s how I first found the dc causing the issue on the discrete opamps.
as the unit didn’t bench fail I can only assume that in the rack caused a thermal problem causing the discrete opamps to fault.

 

[zamproject]

All transistors are to92 package.

I suppose plastic package can also be porous and deteriorate by ageing

I can only assume that in the rack caused a thermal problem

What I describe above it certainly correlated with run time and heat, as malfunction observation increase few minutes after power on

 

[dale116dot7]

Old parts do deteriorate, the metals deposited on transistors and ICs migrate with the voltage fields applied, and sometimes just on their own even without power going through them. When they run hot it happens faster. Also there may be some surge that weakens the part, that is why there are commonly back to back protection diodes across the inputs of op-amp ICs. Sometimes it takes a few weeks between when a part is damaged and when it actually decides to fail convincingly. ESD is similar, if it doesn’t fail immediately, it’ll run for maybe a week or two then quit.

 

[MagnetoSound]

Couldn’t it just be an intermittently leaky cap upsetting the bias?

That’s a lot more common than intermittently bad transistors, and difficult to test for when powered down.

 

[Cqwet Dbdfte]

TO-92's are epoxy, unlike TO-18, TO-5's are metal hermetic. Glass passivation is used to protect silicon in epoxy part, but old parts may not have that. Old gold wire bonding sometimes used heated gold, leading to the "purple plague", ultrasonic bonding solved that.
Epoxy is mildly hygroscopic and does not offer 100% protection, but is good enough for most uses.

 

[JohnRoberts]

A common fault with DOAs is overheating due to poor quiescent or operating current control. The symptom of failing after it warms up for several minutes kind of fits that scenario.

Of course this is just a WAG. YMMV

JR

 

[Voyager10]

What is the DC offset between the + and - inputs when the output is at the -ve rail?

Obviously should be close to zero (millivolts) if it's all working properly. If the + input is significantly negative with respect to the - input, the op-amp is doing its job and the external circuit is at fault. If + is positive w.r.t. -, the opamp itself is faulty.

 

[pucho812]

VDC at input + and input - is .003
So it’s basically nothing. It’s at the output pin that I measure almost to the v-voltage. Opamp is fed from +/- 30vdc

 

[Voyager10]

OK, if that is measured at the same time the output is at the -ve rail, I would expect the '-' input to be also very much negative (with the '+' input near ground, do I assume?).

What's the feedback network between output and the '-' input? Is a resistor becoming disconnected (or a pin loose in a socket)?

 

[JohnRoberts]

VDC at input + and input - is .003
So it’s basically nothing. It’s at the output pin that I measure almost to the v-voltage. Opamp is fed from +/- 30vdc

I think I asked before if you could probe circuit nodes.
===
An op amp could have 100 dB of DC gain so the question he was asking what is the voltage between + and - inputs. 0.003 times 80 dB could be 30V. Of course DOAs are not known for DC precision.

JR

 

[pucho812]

I think I asked before if you could probe circuit nodes.
===
An op amp could have 100 dB of DC gain so the question he was asking what is the voltage between + and - inputs. 0.003 times 80 dB could be 30V. Of course DOAs are not known for DC precision.

JR

I’ll check. Ideally in a working circuit 1vac input signal will yield 1vac output signal.

 

[JohnRoberts]

I’ll check. Ideally in a working circuit 1vac input signal will yield 1vac output signal.

but it isn't working correctly with the output pegged to one rail.

If you start measuring different circuit nodes compared to a working unit, while the output is pegged to the rail, that could reveal what inside is not behaving.

JR

 

[pucho812]

spoke too soon