You're pretty close. The reason for calibrating by turning in a circle or waving in a figure 8, is that in a perfect world, the readings would form a perfect circle or sphere.
In real life, of course, there are internal deviations due to speakers, circuit traces, radios, etc. This deviation needs to be figured out on a per-device basis, although model tests can come up with a general baseline to start from.
View attachment 439867
As someone who's had his own share of wacky bugs, I'd love to know what caused this one.
As I said before, it could be a unit conversion error, or a communications error between the M7 and A7 ( 32 bit variable vs. 64 bit variable?), or perhaps a code mistake using the default model calibration constants instead of individual calibration values.
Often code works in the lab because the developer uses one device, which is reset quite often. Errors don't show up until you begin mass field testing, which in this case obviously failed to notice the problem.
