Proper linear MOSFET for HV load

I use a source resistor and an opamp per fet. Power all the opamps from one cheap dc/dc converter. The opamp offset will be small, so you can use a small current sense resistor. Because the control is so tight - we are synthesizing almost perfect fets - it makes classAB biasing essentially perfect, with no pots or Vgs-th sensitivity.

Sometimes I also compute junction temperature in real time and use that as the shutdown.

Lots of switchmode fets get wimpy in the upper-right of their SOAR graph. Keeping the voltage low is good there.

Not much, in the extreme range specified. Similarly this complicates sensing, pretty much requiring an opamp per FET, otherwise the source resistor drop is too high (maybe even too much for the BJT method suggested elsewhere, too).

Tim

Nah, the resistors are there all the time, just changing them as required for setting (e.g., changing the total resistance to emulate a constant current characteristic, or whatnot).

Yeah, this has the advantage of needing only one dump resistor, and the downside of being significantly more noisy. It's hard to filter anything with such a huge impedance range; the best you can do is target the middle, and tell the end user not to expect CC/CV/resistive/whatever performance at frequencies beyond so-and-so.

I've done it for a limiting fuse/switch before, which works very nicely (mind, I used a TVS stack for more consistent voltage, and more compact energy handling), but the noise is a bit more tolerable there (it's only during fault conditions).

Tim

For linear amps, we just tested a bunch to

Ohkay, those are indeed destroyed! Do you have any data you can share about how they compared to their datasheet SOA?

Jon

Oh I got a dandy forya. I'll Spice it. (and I did it like 20 years ago on paper)

It was a long time ago, so I can't find the data. We did decide to use IXTH11P50 and IRFPS37N50A.

Most of the electronic loads I've used have been to test power supply transient performance, etc. Sudden jumps in load resistance wouldn't be a good idea.

Errr...

So if we set it near a fair-number-of-bits rollover, it can do both automatically!!

But seriously, any switched load worth its salt, switches slowly enough (say with MOS SSRs or just transistors), and fills in the gaps with a linear sink that's fast enough to do so.

They also tend to be unary (aka thermometer code) so you don't get nasty code crossings.

Or at least... that's how I designed mine. I can't speak for others.

Tim