Schottky diode failure?

[elekorsi]

I was cleaning and organizing my workshop and found a 37kW VFD which i replaced with new one at the customer some time ago. I wanted to throw it out, but then remembered that it had some strange, untypical failure. I connected it to the mains and the problem was that it was constantly falling into undervoltage alarm, but the DC bus voltage is OK when measured with DMM (530-540V). Quick look into the monitoring parameters, and i could see that something is wrong with the DC bus voltage measurment, as it was measuring somewhere around 330V. It just triggered something in me and i just had to take a closer look at it... After all, it's nice to have one at home in case someone needs it urgently in the middle of the night...
Visual inspection showed nothing special, taping the board and components made no change. Then i started with cool/heat method, and immediatly after spraying the board, meassured voltage jumped to around 450V, heating with heat gun and it went back down. So i quite quickly located the macro location and proceeded to locate the micro location with freeze spray and clean soldering tip to locate the component. I found that its the schottky diode that is causing problems, so i first made sure that there is no cold joint. After some heat/cool cycles (maybe 4-5) everything magically started to work as it should, the measurment in the monitoring parameter is bang on with that one on the DMM 

So, long story short, my question is:
-is it possible for the diode to fail in that way, and then magically recover after some heat treatment?

Diode type: double schottky diode BAR43SFILM in SOT23 package

[Terry Bites]

Its more likely to be an intermittent solder joint. Resoslder the diode- it costs nothing.

[Ian.M]

Why chance it unless inspection under magnification confirms the dry joint?  If not, replace the ****er, either with anything similar you have in stock or put it on your next order.

[bdunham7]

-is it possible for the diode to fail in that way, and then magically recover after some heat treatment?

Yes, it actually is.  Toasting or reflowing circuit boards sometimes works as a fix, and not always due to soldering issues.  In those cases, the fix is often short-lived.  Replace it and you are likely done with the repair, the only other likely issue would be with the PCB itself.

[james_s]

I've seen a similar failure with diodes a couple of times, I suspect it's a microscopic crack in the die or a fault with the bonding wire. Thermal cycles can cause the connection to close but it's likely to fail again and some inopportune moment. Since a diode is cheap I would just replace it and be done. Then sell the unit as refurbished if you don't need it.

[T3sl4co1l]

Crack, sure, or tin whisker is another possibility.

Or leaky flux or other mildly conductive residue, which can form tin tracks by electrolysis, particularly in humid climes.  I would guess a VFD might be made with conformal coating, which slows that sort of thing, but maybe doesn't perfectly prevent it.

If it's external to the component, removal, cleaning, soldering, and cleaning again, should be good for a long while.  If it's internal, do all that while also installing a brand new component (you know, from a reputable distributor and all that).

Tim

[elekorsi]

Well, forgot to mention, of course i resoldered the diode as soon as i noticed the heat/cool behaviour, as cracked solder joint was the first thing on my mind... but even after that, there was no change, then i heated it and cooled it down couple more times to confirm the exact location and the thing just started to work... I desoldered it and according to diode test it is OK on the bench.

I did quite some repairs on the industrial electronics with a high rate of succes, with such failures mostly on electrolytic caps and tantals  (2000s era PCBs) and cold/cracked joints... But never saw a diode fail in such way, that's why i am asking if its possible 
Microscopic crack in the dye seems like the most probable reason...

And of course that diode will be replaced at next order...

[Ian.M]

SOT-23 packages are commonly internally wire bonded.  Surely you have encountered the 'classic' intermittent wire bond failure?  Its most noticeable in LEDs where it can turn a simple LED + DC power source into a slow 'blinker' thermal oscillator - as the die heats up the plastic encapsulation expands slightly more than the bondout wire, pulling the broken wire bond away from the die till its cooled off enough to make contact again.