Toasty Transistors!

[max_torque]

I've been doing some thermal tests for heat disputation on a basic linear NPN transistor voltage dropper.  Using a MJD122 darlington in  Dpak, and measuring the tab temperature with a 1mm thermocouple soldered to the tab, turning up the load, with a constant voltage drop across the transistor, with the device on a pcb of a fixed size and thermally coupled to what is in effect an "infinite" heat sink (ie one that stays pretty much a room temp, because it's massive compared to the power it is dissipating.


Anyway, all was good i got my thermal impedance numbers across a range of loads, and though, hmm, wonder what happens if i give it a bit of stick !   

So, i wind the current up and the tab temp keeps climbing, up past the 150 degC max allowed for the device (at it's junction) up past 200 degC, and up to 220 degC, at which point i'm sinking 15 watts into the poor little thing, which if you believe the datasheet impedance for junction to case at 6.25 degC/Watt must mean the junction is up somewhere above 300degC or twice the datasheet maximum.  And still it keeps hanging on in their, in fact, i left it for around 6 hours like that and it was still working. Eventually i i gave it a bit more stick, and the poor little thing actually un-soldered itself from the pcb, as the solder i had used had a 230 odd degC melting temp.....

At all times the device was inside it's basic SOA for DC operation, although i suspect that graph is SOA at 25degC junction and not 325degC, but even with what i would call a completely gross overload, the device didn't actually fail.  So finally to my question, what actually kills NPN transistors, or perhaps alternatively, how hot to do have to get them to effectively bring their life to single hour long digits??

Is their some form of thermal cycling lifetime, ie bond wires fatiguing that would be reduced by massive short term over temperature for example?

[magic]

Not sure. I once mounted an opamp to my soldering iron's heater, left it at 400~450°C for a few minutes and it didn't work afterwards.
I also desoldered a DPAK MOSFET by leaving its gate floating by accident but it only lasted a few seconds before it fell off. This MOSFET still works (as a simple switch).

[sarahMCML]

Talking of toasty transistors, not far short of 50 years ago I had a T05 can jobby (either a 2N3053 or a 2N2711) get so hot that the top of the can started to bubble! When I removed it from the circuit the collector lead fell off, but it still tested as working. I still have it around in a tin somewhere, and must dig it out and test it again when I move into my new place.

Regards,

Sarah.

[Zero999]

Many components can take a lot more, than specified. I imagine, there's some de-rating going on, to ensure it lasts.

I recently soldered a power LED to a piece of brass bar, fixed to a heatshink and increased the current so much one of the wires unsoldered, even though it didn't fall off the brass bar.

[dzseki]

At higher temperatures the leakage currents increase, that may cause runaway effect, less likely in simple transistors than in complex ICs. Also with elevated temperatures the diffusion rate increases, and the doping profile ulitmately changes over time, altering device behaviour irreversibly.
To my knowledge the +155°C limit temperature is because of the plastic package, as the plastic case has the glass transition point there, over that temperature the case becomes brittle and may (or may not) stress the die mechanically again causing altered device behaviour through piezo effect.

[rfeecs]

Burn in testing to estimate Mean Time to Fail based on temperature is a routine thing manufacturers do to qualify their process.  Here's an example with a plot showing what happens to the lifetime vs junction temperature:
http://www.hp.woodshot.com/hprfhelp/4_downld/products/xrs/at41511_r.pdf