low cost isolation amplifier

Really? all you people with your helpful suggestions! LOL.

I interpreted the original message as

Look what I found on the WWW and posted under a sarcastic subject line.

Don't know the ATTiny but assuming they can do PWM at 20 MHz at 10 bits this would mean somewhat > 50us, not as bad as I intuitively would have assumed really. Filtering that will take its toll though, probably making it impractical (being after low cost). I was thinking more of a some "normal" DAC which is cheap enough and an MCU with an ADC driving it via some optocoupler. But like I said I don't know if there are any DACs that cheap, John is after low cost for this.

We'll untimately want to acquire voltage and current from an isolated power supply or dummy load channel, and push that into an FPGA, so the deliverable is data. We would close some loops in the FPGA, like current and power limiting, and do waveform acquisition as a bonus feature, so I'd like at least 100K e/i pairs per second.

It could just about be done with a uP on the high side, using its ADC. Move the current limits and such up there. If we could get delivery on uPs!

We use the ADUM7703 isolated ADC in another product, and they are great, just a tad expensive for a high channel count system.

I guess we could consider some tacky home-made PWM thing.

And, don't forget power; if you want a half a megavolt of isolation for power, an ultrasonic transducer, half a meter of ceramic rod, and a second 'pickup' transducer can do the trick. That was for a van de Graaff apparatus...

My solution - back in 1975 - was a three-winding transformer. Two of the windings were made with fine wire, and used only for voltage feedback. I drove the high current winding to get the voltage I wanted to send on the feedback winding, and monitored the current I was pushing in. The isolated output was feed into a sample-and-hold amplifier, which got switched to "hold" whenever the drive current got too high and was being pulled back fast to an equal and opposite current in the other direction.

The "hold" period was only a couple of milliseconds (which was fine for process control signals) and the whole set-up was cheap and remarkably accurate.

A few years later, I shifted a twelve bit word across an isolation barrier as Manchester-2 serial string, using a small feritte toroid stapled to the board with six U-shaped bits of wire - three were the drive windings and the other three the sense windings. I used a comparator with hysteresis as the receiver. That was pretty cheap too. Sigma-delta A/D converters could give you a 20-bit word these days.