So yesterday I was working on the schematic (in KiCad) for the main connections, and came across a few important difficulties that hadn't come up earlier when I was doing a rough sketch on paper. The thing was, earlier on I'd just said "the voltage needs clamping, but that's easily done with some diodes – worry about details later". When I was actually doing the schematics, I suddenly had to think about those details. And the voltage measurement interface that I'd thought was pretty easy is suddenly much harder.

The first detail is that if the input simply comes through 10 M ohm as I was planning, then small signals have such a small current available that leakage and bias currents become a major problem.

The second detail is that diodes don't make good clamps. When the current is low (e.g. a small voltage across a large resistance), the voltage drop is also low. I had previously looked at some diode data sheets, and decided that there seemed to be an asymptote at about 0.5 V that was being approached – instead I should have read up on some basic diode theory, where at low enough currents (low enough that the resistance doesn't show much affect, so often several tens of milliamps) the diode forward voltage has an exponential relationship to the diode current. Great, that won't work, and especially not when you consider the above problem...

So I'm going for a two-part solution (if it works, and unless I come up with something else): voltages less than 1.5 V go through the resistance circuit so that they've got a low input impedance (but they go straight into an op-amp buffer, so the actual input impedance is actually significantly higher than 10 M – the low actual resistance only matters when the voltage clamping comes into affect); and using low-leakage MOSFETs triggered by comparators for clamping.

The biggest issue I can see at a glance is that I can't use the voltage measurement in the resistance circuit when actually measuring resistances, since it would be measuring the current instead. Better would be to put it in its own circuit, but that would mean a second high-voltage solid-state relay, and they're not exactly cheap...