I have a resistance I'd like to measure, and my main choices seem to be:
A noninverting op-amp with a fixed reference as the input and the sensor in series with the feedback resistor and ground, Vout = Vref*(1+Rf/Rs). The problem here is the reference voltage eats about a bit of ADC resolution. I can live with it, but it's not ideal. Cutting the reference down with a trimmed voltage divider is an option, too.
The circuit recommended by the sensor manufacturer is a inverting op amp, with a negative voltage through the sensor into the input, Vout = -Vin*Rf/Rs. The trick here is the negative voltage.
There are a ton of threads on getting a negative voltage to drive an op amp, but I'm not doing that here. This is a single supply mcp600x, with a negative input.
I'm short on space, but I can have a charge pump running off the AVR's PWM, using two diodes and two caps. This isn't regulated, but do I care, if its only juicing the inputs to some op amps? The datasheet says 19 pA typical input bias current, but no max. I'd be using eight channels. The sensor resistance would vary between several kohms and seveal Mohms.
I can choose an op amp with more detail in the datasheet, too. They aren't ordered yet :)
Some of the charge pump circuits I'm stumbled across (the more authoritative ones, generally) run the PWM through a FET and that charges the caps. When would this be useful, other than to switch a higher voltage, which I don't need to do, and what's a good FET for the job?
I did find a web page with a neat H-bridge circuit and an AVR that used RFD3055s: http://www.dharmanitech.com/2008.... This example looked just right, since it's breadboard friendly and doesn't use a driver.