I had a thread opened about some accuracy issues with the BQ27510-G3 and worked through the details to obtain a very accuracy golden image with the help of TI support. However, I just had a unit that has a very large capacity discrepancy. You can see the filteredFCC is very different. I recall this happening last time as well and it seems to possibly coincide with us plugging and unplugging the battery. This particular unit ran accurately a week ago when I tested it. It ran for 14hrs+ which is what I calculated and the gauge was accurate. Now it is predicting <10hrs. This is using a golden image that I cycled 10 times to verify accuracy before deploying it to test units.
Here was the sequence of events. The unit was at ~40% SOC and then I unplugged the battery to make some other unrelated hardware changes. I plugged the battery back in and it was suddenly at 4% despite the cell still being at > 3.6V. I plugged it in to charge and surprisingly it was starting to taper the charge current at 40% SOC already. I let it continue to charge and noticed the state of charge jumped from ~70% to 100% because the fuel gauge was so far off when it reached charge termination. So somewhere in here the gauge thinks we lost 1900mAh
We are going to production here in a few weeks and so this is becoming very concerning.