Thermal instability in MOSFETs

[wg_ski]

A few years ago, when looking up data sheets for mosfets specifically for class D amps I noticed that a lot of newer types have a region of thermal instability at high (or even not very high) Vds that is not unlike BJT second breakdown. I figured that it was just a consequence of making them “better faster cheaper”, and that optimizing for low Qg was probably responsible. No big deal - they’re intended for switching, right? But I looked at data sheets recently for some devices I had traditionally used for lower-speed switching duties - and found their recent data sheets showed the same thermal instability regions in their SOA graphs. Went back And looked at the PDFs I downloaded years ago, and they show the SOA to be strictly TJ limited - like you expect a mosfet to be. What gives? Is it something that manufacturers recently realized whe their newer flavors started having problems when used as linear amplifiers, has it always been this way, or has something fundamentally changed in the way they process them? Is an older IRFP2907 or IRF1405 better than a new one? Similar Fairchild types still show the old SOA graphs - are they also susceptible to this issue? Newer types i’m Using at high voltage as linear amps (tube voltage regulators and source followers) still show TJ-limited SOA on the data sheets - but can this be trusted? Or is it something you just have to watch type by type like you do bipolars? I wouldn’t be as concerned if newer data sheets on older types weren’t showing poor thermal performance.

 

[davidsrsb]

New production of older types are still using modern processes and cell structures, so not in detail the same as the original parts.
My suspicion is that the really early Siliconix VFETs were nearly as SOA robust as laterals, but almost anything after that has some SOA issues.

 

[sesebe]

"Second breakdown" is present on most of the MOS-FET-s but is not documented by suppliers.

 

[davidsrsb]

The original VFETs 40 years ago like the VN66 were a single cell, fairly high Rds and had a low temperature coefficient
something I have noticed is that recent generation parts have a very high threshold temperature coefficient, so they get runaway.
This is a problem with class AB amplifier designs like the SKA