I'm reworking a power gate and I'd like to include a small "discharge" transistor to basically tie the cut-off power rail to ground. But how can I make sure that transistor won't be disintegrated when it shorts the capacitance of the module to ground ... Is there some checks, maximum energy it can absorb or such ?

EMIR (ElectroMigration and IR drop) and SHE (SelfHEating) simulations. As far as I know, open source tools don't have those features. Just keep the metal tracks large enough. Maybe simulate the maximum current of your switch, as it will limit the current it can discharge. You can size it accordingly by using lower aspect ratios, if speed is not a problem.

Yeah, I made the FET long and narrow to limit the current. It ends up at a few hundred uA at the moment, but yeah, I have no idea how fast heat would go aways ...

1 mA per 1 um metal width. If your transistor diffusions are so small that they can't handle it, perhaps you should resize it. Anyway, if the current spike is too short, it shouldn't be a problem. Maybe the readthedocs page has more info about it

Usually a bulk MOSFET is very tough to overheat. The I_ds is scaled by W, and larger W is more current is larger size. I haven’t really seen a situation where a bulk MOSFET can be overheated. Different are FinFET (the small silicon fins embedded in SiO2 have a self-heating issue), and BJT (since a small NPN can still carry lots of current due to I_cs = I0 * exp(V_be / V_T).

@Harald Pretl Thanks. In the end I went with something probably massively overkill :

There will definitely be no overheat at this current level 🙂

Devices in this generation can support about 1E-3A/um of device width. so as Harald said. Not a concern.

@Andrew Wright Thanks, good to have an order of magnitude in my head, I'll make sure to stay well below 🙂 For metal the tlef had some info but not for the transistor themselves so I wasn't sure what to expect.