Can MotionMC v2.9.0 calibrate with constrained Alt-Azimuth motion only?

[fpier]

We are using MotionMC v2.9.0 on STM32F407 with an LIS2MDL magnetometer and LSM6DSV16X IMU.

Our application is an astronomical Alt-Azimuth mount, therefore the sensor can rotate around azimuth and altitude axes, but cannot be freely tumbled around all three axes.

MotionMC is updated approximately every 10 ms (100 Hz), with LIS2MDL magnetic data converted to µT.

During a test session we performed multiple azimuth sweeps at different elevations. After approximately 54,000 MotionMC updates the observed raw magnetic coverage was approximately:

X span: 533 mG
Y span: 738 mG
Z span: 616 mG

Despite this relatively wide 3-axis magnetic coverage, MotionMC remained at quality 0 (UNKNOWN) for the entire session. Hard-iron bias remained {0,0,0} and the soft-iron matrix remained identity.

With unrestricted manual 3D tumbling of the same hardware, MotionMC is able to reach GOOD calibration.

We would like to understand whether MotionMC requires unrestricted 3-axis tumbling, or whether a two-degree-of-freedom Alt-Azimuth trajectory can theoretically provide sufficient excitation.

In particular:

• Is true 3-axis tumbling required by MotionMC?
• Are there minimum orientation/coverage requirements for calibration convergence?
• Can highly correlated samples at 100 Hz affect convergence?
• Is continuous motion preferred over stationary samples at discrete orientations?
• Are there configuration parameters suitable for constrained-motion applications?
• For a 2-DOF system, would ST recommend using a custom hard/soft-iron calibration and then applying the resulting bias/matrix to the compass processing?

We can provide serial logs, the exact MotionMC integration code and the test trajectory if useful.
The same firmware integration and the same hardware reach GOOD calibration when the device is freely tumbled in 3D.