Would that be the failure mode of a TVS?

[TrickyNekro]

Hi!

I am trying to repair a device that has a its power inputs protected by a TVS, specifically a SMCJ15A guarding a 15V rail. The problem is that the failure is intermediate. The power supply
has an 1A max output at that rail and the fuse, also 1A is not blown. Even with that rail "shorted" the powersupply still works with the other rails, since the other rails provide much more
current and this supply rail is most probably not monitored.

I powered the device, would operate mostly normally for some cycles under load. Then it would fail with the power rail almost shorted to ground ( one diode drop to ground ).
Then powercycling the device would restore sometimes operation, sometimes not.

I feel like I am dealing with a degraded TVS that conducts, heats up, and then shorts while hot. Unfortunately I haven´t done the finger temperature test as the device was
powered in failure mode. Getting the board in and out of the device is also not an easy task and I would need almost an hour so I would rather, for the time being have a discussion.

Is it correct to suspect a failure mode like this from a TVS? In any case I´ll change it, if I find stock :-P, but would like to narrow down before re-installing the board!

Thanks,
Lefteris,
Greece

[TrickyNekro]

It would also make sense that it hasn´t blown a hole through the TVS as 1A and 0.68V, being the drop of the diode when shorted,
is about 0.6W and with the SMC package having a 75C/W thermal resistance to ambient that would be about 45°C above ambient,
which is easily tolerable by diode standards.

[David Hess]

The TVS protects in reverse breakdown, so roughly 15 volts at an amp or 15 watts which will overheat it pretty quickly.

What I have seen in the past is that the power rail for whatever reason will spike or surge causing the TVS to conduct and protect the circuits, but if the TVS is undersized, then it will be damaged and eventually fail, usually with a short.  Spike or surges are usually caused by poor dynamic performance or frequency compensation of the voltage regulator.

[TrickyNekro]

The TVS protects in reverse breakdown, so roughly 15 volts at an amp or 15 watts which will overheat it pretty quickly.

Yeah, that would seem to be the case if it was conducting only 1A, or at least present the resistance to conduct 1A.

But that would have been very momentary. It is probably heavily degraded, warms up and then shorts. But somehow somehow... when it cools down then it is working again.
Thermal expansion much?

Also the TVS seems to be overrated for the rail. There are two 15V rails on the device. One is +15V the other -15V. The +15V can supply 4.3 Amps the -15V can supply 1A from the main power brick.
Both are probably un-regulated as the regulation is done either by measuring core-flux or from the 5V rail that can supply up to 13A. There is a trimmer for the the 5V rail, but that could also be
linear post regulation as the +15V and -15V remain unaffected.

I have re-capped the powersupply brick some weeks ago as the device diagnostics would propose that this would be the problem. After the re-capping I got this problem, so these diodes have been stressed. Also it makes sense that the supply rail with the least current handling capabilities would get the most ripple when overloaded.

It is funny the way that it fails... Since I will change them, I will try to decap them and post a photo. The damage according to an application note from Vishay, should be visible.

Cheers!

[TrickyNekro]

Also note that I was measuring at the power supply, when measuring the shorted rail. Usually the TVS fails short with 0.1V across them according to Vishay.
Add to that the cable resistance and most importantly the fuse resistance and measuring 0.68V at the shorted rail should not be all that strange.

The dirty me, wants to just test the device just by omitting the TVS. The problem is that the device is rather critical and expensive, so I will not do that.
It´s not the diode price that is expensive, it´s the downtime that is expensive. But I can find stock so I´ll be either replacing it with the same
or a similar part number.

[wraper]

Why would TVS be a primary suspect for increased current consumption? It's not a typical failure mode at all. If you do not want to remove it (I don't see why it would endanger the circuit if it's just across the power rail), then put your finger on it and check if it heats up.

It´s not the diode price that is expensive, it´s the downtime that is expensive.

And such discussion on the forum is supposed to be the fastest way to fix it? If you suspect it so much, why don't you desolder it and test it out of the circuit?

[TrickyNekro]

It´s not the diode price that is expensive, it´s the downtime that is expensive.

And such discussion on the forum is supposed to be the fastest way to fix it? If you suspect it so much, why don't you desolder it and test it out of the circuit?

Yes actually, if somebody else had a similar experience. Repair is ruling out suspects after all. And because it is an intermittent fault, which actually are
particularly pesky and notorious to find out, asking if somebody else had such an experience with a TVS, it makes sense to concentrate on that problem.
For all that I know, it could be a linear regulator down the line that is busted or maybe a relay somewhere else.

I have already desoldered and ordered replacements, so don´t worry about that. As I said, since I suspect them, I will be replacing them anyhow and test again
in the device.

I will also be testing the TVS itself. But testing is not all that easy. I have to use the same switching power supply and switch on and off a load and see how it behaves. If it latches closed
at some point then, yes it is that. But I have to set my scope, find the resistors, set up a mosfet with a generator, cooling for the load etc. etc. etc. I lose nothing by asking. If I had a digital
load, which seems to be next on my buy list. I would have done that already.

And if in the end, this is the problem, it get´s written down for others to see for future repairs. Win win for everyone.

The problem is, that if it is just a degraded component, by testing I can cause it to fail open, which might also mislead you to other conclusions so it´s not just black and white mate.
TVSs after all fail either open, closed or degraded. I do suspect that the device is degraded and has formed a melting zone on the silicon, that when it cools down breaks and starts
operating again. Something similar is described in an application note from Vishay so, it is not all that far from reality to suspect.

[sean0118]

Are you sure it's just a TVS and not a crowbar circuit?

[TrickyNekro]

Are you sure it's just a TVS and not a crowbar circuit?

No. Although the board that I got out of the device, is the main power distribution board, so I would guess something like that would be on that board.
On this board I do not see something like that. But sure you never know. I can´t remember seeing a crowbar with a relay, but I will check it out further.
I´ll have to replace the TVS, install the board again and see what comes out of it for the moment.

Meanwhile, I will be testing the TVS. Not much more I can do without losing track of it all.
But you would expect that a crowbar would have blown the fuse. Well if it exists, that is. But I have also seen that before, blown crowbar and
the fuse ( with the correct rating mind you ) not to have blown. If for whatever reason the power supply connected to the board does not supply enough current, this can happen.

[wraper]

Did you actually measure the current when device fails? IMHO you probably have a faulty PSU and fault has nothing to do with current consumption.

[TrickyNekro]

Did you actually measure the current when device fails? IMHO you probably have a faulty PSU and fault has nothing to do with current consumption.

It could be. I just tested the TVS. They seem pretty ok. They clamp around their maximum breakdown voltage and do leak peanuts at their
standoff voltage so they are out of the equation. And I just hoped for an easy repair really... 

It could be the PSU, or a board down the line that has the problem. I haven´t measured the current but the PSU seemed pretty hot to the touch
with the short active. After powercycling the short was solved and the PSU seemed to operate at a more "normal" temperature. Though I won´t
be taking it out of the equation for the moment. The "problem" is that I have recapped the output capacitors of the PSU, so I kinda think it should
be ok, but that could be another can of worms to be honest, it´s not really a complicated beast and it seemed to be hot around the HV switching
transistors which suggest higher loading so most probably it is not the case.

Ah well... I will update with the repair, most probably I´ll have to go dig the other board out of the machine.

[TrickyNekro]

Did you actually measure the current when device fails? IMHO you probably have a faulty PSU and fault has nothing to do with current consumption.

So I have checked the two boards inline that get power from the power supply and... well... it seems like it might just have been an over-temp issue with the PSU.
It seems like it uses the same heatsink for the two diodes, one for the +5V, one for the +15V and one negative regulator for the -15V. The negative rail regulator
is a UPC24A15, from the scarce infos online it seems like it has both over-temperature and over-current protection.

It also seems like 5V rail was set a bit higher than it should, which of course could lead to more power draw from the various boards and higher temperatures.

Testing open air, by lowering the 5V rail I saw an at least 2°C reduction at heatsink temperatures, only power 2 boards out of the 5. I will be re-installing and testing further.

From the other side I still have my scruples because of the nature of the device, it´s not like such devices do not account for temperature variations. But what can I say? I can only test more...

So it´s either over-temp or the 3rd board. Checked every single ceramic, didn´t see anything strange. If it was, it would most probably gone up in flames ( had this happen to me ).
Without loading the other two lines, I can not tell for use. I loaded the -15V rail and I didn´t get a failure.

BTW, if I haven´t mentioned yet, the TVSs are fine and I just re-installed them, don´t change it, if it ain´t broken!