We used ADXL327 in one of our products. We use it calculated product Tilt. It is powered by 3V(comes from a reference chip followed by a buffer). and the three output interface to a internal ADC of dsPIC without buffer. Because we don't need more accuracy, so no calibration is used. 1.5V are supposed for 0G. Gtotal is calculated based on that. Gtotal is about 0.9-1.1. for prototype (about 30-50 board produced). We rev up the board with other changes. But for new boards, Gtotal is changed with the board tilt. we found the bias of Z is not 1.5V. It is range from 1.4 to 1.63V.
We swapped the sensor for one board between first batch and second batch. Old sensor (was in first batch) had the similar Z output initial. We solder again and showed a better peformance as expected. New sensor (was in second batch) showed bad peformance in old boards even with a deadbug mounting (tried to rule out soldering issue).
The typical output of Z is 1.5V and its min is 1.2 and max is 1.8V in page 3 of datasheet. But Figure 5 show most of them range from 1.46 to 1.52V. So it looks like normal in my case, 1.63V of Z at 0g. (is that really normal?)
The question is why soldering change the Z bias output in my case.