I have a lot of SSRs that I want to measure the voltage drop across, both ON and OFF. I want maybe 25 millivolt accuracy with a common-mode voltage up to +-72.
Each SSR has two 1M precision pickoff resistors and a mux that dumps the resulting currents into a diffamp, then an ADC.
Even buying 0.1% resistors, the common-mode error stackup is big.
So this is the fix: we use a two-opamp diffamp where the upper amp is a current inverter into the lower amp.
The lower output represents the voltage across the SSR. But the upper amp output tells us what the common-mode voltage is at any instant. So if we digitize the pair simultaneously, and have a cal table, we can take out the residual common-mode error and the opamp offsets. We can generate the cal table in manufacturing test, load it into the FPGA, and do the math in hardware. The Susumu thinfilm resistors are very stable over time, so we should be OK.