We have a product here that recently has started having higher fallout in our production test, Z Axis only. The Z Axis voltage measurement is out of our test limits, and it appears like it's getting worse recently the more we build.
I've taken some measurements of all three outputs (X, Y, Z) at 0g, and with a 3.3 volt supply, I see 1.62, 1.65, and 1.84 volts respectively. So the X and Y are almost dead on the theoretical value of Vcc/2, which would be 1.65 volts.
The Z axis at 1.84 volts appears to be still in spec, but I was surprised that it is not tracking the other two axis, and I bet if I can find some older boards and do the same test, its shifted higher over time.
I recently moved our test limits to account for the apparent shift, but it appears I didn't widen them enough after we have collected more manufacturing data and seen more fallout. To be clear, our test limit was shifted based on 5 years worth of data, and now I'm now seeing more outages, so they apparently weren't widened far enough.
By looking at the data sheet, it appears that Analog Devices allows for a wider 0g range on the Z Axis than they do for the X and Y (although the "typ" is 1.5 for all three at 3 volts), so should I consider what I see normal, and I need to change the limits even more?