The question occurred to me, what really is wrong with RoHS power devices (transistors / etc.)?
Stuff like diodes, BJTs and SCRs are a no-brainer, although an SCR might be triggered if a whisker gets to the gate. Obviously, burning the whisker costs a transient, which is no good if you need a perfectly stable signal for instance. Fair enough for those applications. But what about burly power circuits (rectifiers, motor drivers, etc.), and what about insulated gates?
Assuming you have a stiff gate drive (which, well, you ought to!), what hazard are tin whiskers, really? The distance across a TO-247's pins will set destruction probably in the decades range, and when they get there, some current might flow C-E, blowing that whisker instantly from available C-E load current (>>1A). But a G-C short will demand current into the gate drive, which is usually coupled with a resistor. Any ballpark data on whisker resistivity and fusing time for various current pulses? I heard
30mA somewhere, and have no idea about time, though I would guess anywhere from microseconds to tens of miliseconds.Analysis: Let there be a whisker a very short distance away from the gate (or from the gate approaching the collector). For the last however many years this product has been operating, the gap has been great enough that, under all circuit conditions thus encountered, breakdown voltage has not been exceeded, despite millions of cycles of operation. But one day, the whisker grows just barely enough that, subject to power line variations, ambient ionizing radiation, humidity, etc., the air gap breaks down on the next cycle. Now, it occurs to me that breakdown need not be destructive; it could be a very small corona discharge, which would tend to heat up one side more (electron bombardment), perhaps even regulating the length of the whisker. Or, the corona current might be sufficient to fuse it without destructive currents flowing.
Sound about right so far?
What if the whisker does arc over, or make physical contact? (More likely in lower voltage MOSFETs where the whole glow-corona-arc-spark discharge thing doesn't work.) Then, current in proportion to the drain/collector voltage and whisker resistance flows into the gate, which may or may not be destroyed outright (sparks can be on the order of nanoseconds, and if the whisker turns to plasma, there isn't much keeping the gate and drain voltages apart). If it survives, then the gate drive starts pulling current back out. But the drain is now dropping (at a rate of tens of ns, depending), so there isn't as much -- if anything -- to push against. But with it down, it will tend to level out, in a time corresponding to *waves hands wildly*, various R, L and C components in the system. G and D voltages stabilize around the threshold voltage (acting in "diode strapped" mode), and now the drive has something to fight against. Probably a fraction of a microsecond after the whisker hit, the gate drive is able to get ahold of the gate again, and things start returning to normal, depending on how long the whisker takes to melt/vaporize/deionize, which could take miliseconds.
LOL, it just occured to me also that whiskers could resonate at the operating frequency! Wouldn't that be insane if your circuit ran at the characteristic frequency of a whisker bobbing precariously close to sensitive points? I mean, 600V on an IGBT collector is a good bit of electric force for something in the micron range.
Tim
-- Deep Fryer: A very philosophical monk. Website @