I have an ADS131M08 ADC that I am using to measure voltage in various positions in my industrial EV battery (all positions are above the highest cell, but at different points on busbars, before or after fuses/contactors, etc) relative to the battery's ground. I have managed through software and hardware filtering to reduce the noise down to an acceptable level, but another issue has arisen: When I plug in the battery charger (which is a noisy beast that outputs 600A), the output values that the ADC outputs tend to drop by a few percentage points. After the current ramps up, the ADC will suddenly start reporting that the busbar voltage is closer to 34V, rather than the 36V that I measure with a DMM.
Below is what the input noise filtering looks like. I chose an ADC with differential inputs as there are other shunt measurements I need to make which require a differential. After some experimentation, I decided to tie the ADC's analog power supply ground to battery ground, as it gave much better performance and lead to noise cancellation across different measurement points (it seemed to put all of the noise on each channel in phase with each other). I've used such large capacitors because I'm trying to measure only DC, and they've done a great job with noise suppression.
This attenuation is present on all of the ADC's channels, even ones that are at 0V due to open contactors. And it's even present on all of these channels after the charge stops, the charger is disconnected, and those channels (that were 0V during the charge) go back up to the regular battery voltage. It is definitely something systemic in the battery or the ADC, and not specific to any individual channel. My only explanation is that it's something to do with the ADC's internal Vref (I did not attach an external one) shifting up by a few percent, and all of the measurement values appearing attenuated as a result. The ADC goes back to normal after I disconnect it from the battery voltage or refresh power to the board it's on.
