Do TVS Diodes (1.5K400CA) ever fail open

[cybermaus]

I have a Ultrasonic driverboard (well, 3 of them) that repeatedly blow their FET's (well, OK, two of them blew them twice. But these are $14 FET's, so I do not want to do this too often)

Originally I figured the TVS Diode was still OK, because it did not measure short, and from what I read, TVS's usually fail short?
And I do not have easy equipment to really test a 400V TVS Diode. I probably need to rig something, maybe with a microwave transformer, a energy limiting resistor, and a variac.

But how common is it for TVS Diode's to fail open?
Because here I thus appear to have 2 or 3 of them.

[coromonadalix]

some of them  may spit their insides  loll  leaving a black mark  on the pcb  like a rocket output burn

oh  and they can short themselves  loll baaaaaad   loll    its the way they do their protection mode.

and sadly  some bust open, it depend on the energy they had to manage and dissipate

we did do some tests up to 30kv  on some systems, had to beef up the parts wattage, add gaz tube discharges  etc ...

sometimes removing them is easy and other cases not and pcb traces comes with them ...

[DavidAlfa]

Very rarely, they usually clamp the voltage, then short out if overloaded.
If the peak current is brutal they can completely blow apart, but you'll clearly notice that "something" happened, definitely won't look nice and shiny 

[cybermaus]

Yeah, that's what I would have expected. Either shorted or visibly burst. But these look pristine.
Yet the repeated FET failure (combined with still healthy snubber caps) make me suspect them.

[cybermaus]

Alas, a quickly thrown together test setup shows the voltage is nicely clamped at ~450V, in both directions.
Which I guess is a good value for a 400V TVS diode.

I'll have to find another culprit.

[Kjelt]

I'll have to find another culprit.

Don't know if it is relevant or not but is your max operating voltage of your fets at least 10-15% lower than the factory specified max. voltage ?
If not, read about "cosmic radiation effects".
I thought it was a joke 10 years ago, but it is not and am glad the industry in the meantime picked it up.

https://www.infineon.com/dgdl/Infineon-MOSFET_CoolMOS_CFD7A_Cosmic_Radiation_Assessment-ApplicationNotes-v01_00-EN.pdf?fileId=5546d4627a0b0c7b017a1415c2ef1185&da=t

[cybermaus]

Well, it's not something I am designing, I am repairing. The component choice are a given from the manufacturer

They are IXFK44N50P FET's, which are listed as 500V, and I got them from Mouser, so presumable up to spec.
With 400V TVS diodes, clamping at 450V does mean they are within those 10% guidelines.

[Neomys Sapiens]

According to the Failure Mode/Mechanism Distributions (RAC 1991) - no! 100% Short.

[bdunham7]

Although in theory any component will eventually fail 'open', those will be shorted unless they are physically damaged or, as I've seen a few times, completely missing because they've been blown off the board.

[cybermaus]

Although in theory any component will eventually fail 'open', those will be shorted unless they are physically damaged or, as I've seen a few times, completely missing because they've been blown off the board.

Right, like I said: Either short, or visibly damaged (blown clean of the board was an option in my mind when I wrote that)

According to the Failure Mode/Mechanism Distributions (RAC 1991) - no! 100% Short.

Interesting document, a few remarks/questions:
- Do you have a searchable / OCR copy of the document?
- Or better yet, a more recent version (while I remember 1991 as if it was yesterday, it is in fact quite not so)
- When I read it , it shows 35% vs 45% (zener regulator is the closest match I found, page 2-20. But again, a searchable copy may be better)

[Neomys Sapiens]

According to the Failure Mode/Mechanism Distributions (RAC 1991) - no! 100% Short.

Interesting document, a few remarks/questions:
- Do you have a searchable / OCR copy of the document?
- Or better yet, a more recent version (while I remember 1991 as if it was yesterday, it is in fact quite not so)
- When I read it , it shows 35% vs 45% (zener regulator is the closest match I found, page 2-20. But again, a searchable copy may be better)

Unfortunately, this is the only version that I'm aware of. And I work a lot with such data. When NTIS says 'reproduced from best available copy', this is generally correct.
But on page 3-62 you can find an entry specific for Transzorb diodes.

[Tomorokoshi]

Without knowing the manufacturer of the TVS I looked at the Littelfuse and ST datasheets.

Here is the problem with TVS ratings: you don't really know what voltage you are going to get.

From the ST datasheet:
https://www.digikey.com/en/products/detail/stmicroelectronics/1-5KE400CA/634672

To calculate VCL max versus IPPappli: VCLmax = VCL - RD x (IPP - IPPappli) where IPP appli is the surge current in the application

Where RD = 11 ohms.

That MOSFET is a fairly hefty part. What is the inductance of the transducer? Any idea of how much current is going through it? What happens when it switches off? What is the impedance from the MOSFET to the TVS? What is the impedance from the transducer to the MOSFET?

Any pictures of the board and schematics?

The isn't much margin from the TVS to the maximum ratings of the MOSFET.

[coppercone2]

idk about tvs diodes but I have seen normal diodes like zener or other that develop a hairline fracture in the middle and look fine but are open from mains exposure and if you poke them with a tool they seperate down the middle

I would assume if its built the same way then why not

wonder if i still have that pCB i will look tommorow

I got pissed looking for the problem but then when I put a light on it I noticed the crack line shimmer, measures open with DMM and then I poke it and boom it bends open smoothly with leads still attached to the board and no burn marks

I would say counting on.. obviously visible 'blast' damage is not going to have you covered for open diodes.

[David_AVD]

Are the FETs on the mains primary side like an SMPS ?