Using a BSP100 mosfet and several times because of equipment installation errors the mosfet drain get +12V battery applied and so when it switches to ground it dies. The thing is, gate is shorted to drain but drain-source is open circuit. I thought that’s a bit unusual. Seen it happen several times. Bonding wire failure?
I used to use these a lot and had trouble with failures on some prototypes, eventually turned out to be ESD damage during assembly!
As most today low Rds MOSFETs datasheet have line with: "current rating limited by package"
Bond wires are weak link, they can blow even sooner than chip is busted, but their melting will kill it as it produce hot spot
Step 1. Die overheats.
Step 2. Die melts, becomes 3-way short. Gate oxide is disturbed, junctions are melted together.
Step 3. Fault current ensues. Melt spot becomes vaporization spot; die surface and bondwires evaporate into plasma.
Step 4. Supersonic plasma ball builds pressure until the package is fractured, releasing a sharp "CRACK" into the surrounding air.
It would seem your example went through at least steps 1-3, and had little gate current to vaporize that bondwire, leaving a 2-way short.
Drain will almost never become open, because it's the tab the die backside is glued onto. source bondwires will vaporize instead. Gate may become disconnected due to the cracked package if step 4 is reached, or remain connected after all.
Note that during steps 3 and 4, the transistor becomes an all-way short, exposing gate and source circuits to some fraction of drain circuit voltage. This can destroy e.g. gate drivers, even behind a coupling transformer, even if the device ends up a three-way open. Transients are not to be underestimated!
I've seen a lot of "silent" deaths in my work. Steps 1-3 are followed, but there isn't enough energy available (or the device doesn't fail in a low enough resistance mode) to blow open the resulting short. A bit annoying for fault finding, but, it doesn't leave shrapnel which is nice, right?
Tim
I've seen a lot of "silent" deaths in my work. Steps 1-3 are followed, but there isn't enough energy available (or the device doesn't fail in a low enough resistance mode) to blow open the resulting short. A bit annoying for fault finding, but, it doesn't leave shrapnel which is nice, right?
I think I prefer the shrapnel over more complicated debugging
I've seen blown bond wires a couple times but it has been rare for me.
I'm all too familiar with shorted mosfets causing a cascading failure. I had a washing machine motor controller that blew up every IC on the board including the ASIC running the whole show. It's the reason I'm not really a huge fan of non-isolated gate drive ICs.
I know some people even using failed IGBT modules with parallel chips for hobby use
Most times is only one chip shorted and bond wires can be blown by big capacitor
So rest of module can be used at appropriate lower powers
I once blew the silicone potting gel out of a 1700V 1200A IGBT module. It was very loud and messy. (I was glad to be behind a blast shield and wearing earmuffs for that testing.)
ABB I think went through a lot of trouble to develop special IGBT packages that were guaranteed to fail short. I think they moved away from bond wires to outright compressions springs.
What would be the benefit of failing short? I would have thought that failing open would limit the damage, especially in a half bridge leg.
Circlotron makes a great point - is a short better than an open? Would it be better for the drivetrain in an auto to lock up or free wheel when something fails.
In the days before T0-247 (and similar) packages I was developing a PWM servo using TO-3 MOSFETs. Something went wrong and an internal arc was struck between the case and the source pin or bonding wire. It burnt a 1mm hole in the steel case before shutting down. I didn't do a post mortem on the part (an IRF340) so I don't know if it was a manfacturing defect or maybe just a bonding wire failing and springing up to the case.
Yes. Sorry.
The ABB application was in High Voltage DC power converters, where you connect a whole heap of 6.5kV (or thereabouts) in series to make a composite switch rated at 150 kV at a thousand amps or more. An open circuit would be Bad News here because it would arc like nobody’s business. So instead they use more switches than necessary and accept some short circuit failures.
The converters include special gate drive and snubber systems which distribute the off voltage evenly across each switch.
I've seen blown bond wires a couple times but it has been rare for me.
I'm all too familiar with shorted mosfets causing a cascading failure. I had a washing machine motor controller that blew up every IC on the board including the ASIC running the whole show. It's the reason I'm not really a huge fan of non-isolated gate drive ICs.
I got a good one for you, yesterday -- I made a control board using a PWM chip (internal 5V reference), and the inverter with a Hall effect sensor for power measurement. Well, in one test I grenaded the inverter; as it happens, the transistors failed silently (stage 3 failure), and fault current flowed from the filter caps into the transistors through the sensor. The sensor is the only part that exploded! Conveniently, said sensor was powered by the PWM chip's 5V reference (the circuit is 12V otherwise), and it happened to take out both!
Tim
I once blew the silicone potting gel out of a 1700V 1200A IGBT module. It was very loud and messy. (I was glad to be behind a blast shield and wearing earmuffs for that testing.)
That's something I have always liked about power electronics, when something goes wrong it tends to be pretty exciting! Unfortunately it also tends to be expensive.
Many's the time at my current job where I've wished for a nice clean 'bang' type fault, instead of the elusive intermittent bastard type fault.
On the other hand, I don't have to fill out safety paperwork, and the last thing to make clouds of smoke was the kitchen toaster.
If only the nice clean ‘bang’ you speak of was limited to the power device instead of snaking it’s way through the board.