AD7682 Bipolar Differential Mode

Hi there,

I'm trying to use an AD7682 to sample a differential sensor signal and am seeing some strange results. If I sample IN0 and IN1 separately as single-ended inputs and subtract the two, I get a nice clean signal. If I try using the bipolar differential option, though, I see a steady-state offset as well as some periodic high-frequency noise on the order of about 10 mV (RMS noise using the single-ended option is typically around 0.5 mV).

Any ideas what could cause something like this?

Updated: Added a chunk of our schematic.