I'm using the ADS131E06 to measure <2kHz signals to a required "noise-free" resolution of 125uV at the input (considering all gain drift errors, but not reference or converter noise), with ~8kSPS output data rate (decimation ratio 128), PGA gain=2, and a REF5040A external reference. By my calculations, this is equivalent to 15 bit [noise-free] resolution. The ADS131E06 datasheet shows a buffered external reference circuit, but considering my requirements, I would like to use just the REF5040A with a 1uF input bypass, 1uF NR to GND, and 47uF tantalum output cap. There would also be a 10uF || 0.1uF between REFP and REFN at the converter input (still close to the reference output cap). The expected ambient temp range is -40 to +85C.
I'm hoping the large tantalum output cap will reduce overall noise and gain peaking, and this and the limited bandwidth will help counter the [unspecified] increased noise at 85C, resulting in noise inside the converter that will be <1 LSB.
Do you think this is an adequate solution?
For background, I'll be synchronously sampling 1024 points (8 waveform cycles) before doing an FFT to derive fundamental and harmonic magnitudes, so this should help to average out any remaining noise effects.
