We changed an existing design using an AD7734 (AD part) with an ADS1255. Based on data sheet information this should give us 20 bits stable at the 1Khz sampling we were using, along with better temperature stability and some other positive consequences.
It took quite a bit of effort to actually configure the part and get results - the data sheet while "accurate" can be interpreted in different ways especially when other parts do some of the SPI differently. However results arrived but did not seem terribly stable. It could be argued that we had about 18 bits clean but even that seemed to drift.
Further investigation has shown that when operating at 2.5sps we get a significant drift with time in the raw data reading when connecting input to reference (2.5V). Uncalibrated the ADC is giving a result in the expected region, however the last 4 bits are unstable AND the output drifts the equivalent of 8 bits over approx 5 minutes, up and down. The reference is 10ppm stable and doesn't change over this time (measured on external 6.5 digit DMM). However given the ratiometric nature of the measurement reference drift should be negated.
We have tried several chips all with similar results. I can accept that the layout is a little compromised through implementation necessity however the AD7734 when connected input to reference gave a rock steady output with stable temperature in a similar environment.
All suggestions would be appreciated.
