O̶S̶M̶U̶ ̶R̶e̶v̶.̶0̶ (Patreon)

 I've tried various ways of extending the voltage range of the LT1970 which I wanted to build the OSMU around:

• The NI way based on this http://www.ni.com/white-paper/4297/en/ It looked like I could just add equivalent circuitry for negative output voltages, but what I came up with had huge transverse current.
• Photoelectric gate drivers like this https://imgur.com/a/jTHDeaf sort of work, but the only ones I've been able to test have a fast-discharge feature which makes them super unstable
• I am trying current mirrors now https://imgur.com/a/o2131i0 but with those the circuit is almost as large as a fully discrete implementation 

All signs are pointing the way of a fully discrete implementation. It's cheaper, more professional, performance isn't depending on obscure ICs etc. AND everyone I've talked to would prefer that.

What about isolated DACs and ADCs? I wanted to use these isolation amplifiers http://www.ti.com/lit/ds/symlink/amc1301.pdf but I think there are better options. 8 MHz SPI is not too uncommon and even in full duplex mode with 6 or more slaves I think that'd give me sufficient data throughput. Anybody happen to know an integrated SPI 2-ch DAC + 2-ch ADC? Something like http://www.ti.com/lit/ds/symlink/amc7812.pdf but smaller? :)

I've spent a lot of time finding the smallest 10A relays for current range switching: https://www.mouser.de/ProductDetail/Fujitsu/FTR-P6GN012WA?qs=AcmJtBVhzkRfw7wxTo3Y4A== But those are of course horrible, wildly varying contact resistances even within the same part!

All in all: There is almost no component left the way I imagined it in the original design = more PCBs ... yay  🎉