[WIP] Pr compasscal offdiag

[dagar]

Opening this so we don't lose track of it.

[AlexisTM]

This should be new_report.y += ((zraw_f * _range_scale) - _scale.z_offset) * _scale.z_offdiag;

[dagar]

Thanks @AlexisTM. I left this open as a reminder to review the entire magnetometer calibration.

[mhkabir]

Is this PR in a workable state? I have an application which needs compensation for soft iron distortions, and was wondering if it is safe to carefully start using this.

[dagar]

Is this PR in a workable state?

I doubt it, but we could use this as a starting point.

[mhkabir]

OK, I updated this locally and fixed a few things. Should be in a working state now, but I haven't tested yet.

I also want to switch the application of the calibration into vector/matrix math, something like :

matrix::Vector3f mag(xraw_f * _range_scale,
		yraw_f * _range_scale,
		zraw_f * _range_scale);
// apply offsets
matrix Vector3f offsets(_scale.x_offset,
			_scale.y_offset,
			_scale.z_offset);

mag = mag - offsets;

// apply scale calibration
matrix::Matrix3f scales(_scale.x_scale,	_scale.x_offdiag, _scale.y_offdiag,
        		_scale.x_offdiag, _scale.y_scale, _scale.z_offdiag,
        		_scale.y_offdiag, _scale.z_offdiag, _scale.z_scale);

mag = scales * mag;

versus currently :

mag_report.x = ((xraw_f * _mag_range_scale) - _mag_scale.x_offset) * _mag_scale.x_scale;
mag_report.x += ((yraw_f * _mag_range_scale) - _mag_scale.y_offset) * _mag_scale.x_offdiag;
mag_report.x += ((zraw_f * _mag_range_scale) - _mag_scale.z_offset) * _mag_scale.y_offdiag;
       
mag_report.y = ((yraw_f * _mag_range_scale) - _mag_scale.y_offset) * _mag_scale.y_scale;
mag_report.y += ((xraw_f * _mag_range_scale) - _mag_scale.x_offset) * _mag_scale.x_offdiag;
mag_report.y += ((zraw_f * _mag_range_scale) - _mag_scale.z_offset) * _mag_scale.z_offdiag;
       
mag_report.z = ((zraw_f * _mag_range_scale) - _mag_scale.z_offset) * _mag_scale.z_scale;
mag_report.z += ((xraw_f * _mag_range_scale) - _mag_scale.x_offset) * _mag_scale.y_offdiag;
mag_report.z += ((yraw_f * _mag_range_scale) - _mag_scale.y_offset) * _mag_scale.z_offdiag;

What do you think, @dagar @AlexisTM.

[mhkabir]

PR updated and rebased.

General TODOs for sensor calibration pipeline :

• Some new drivers may need changes for the off-diagonal calibration application
• Horrible boilerplate for applying calibrations in every driver (accel/gyro/mag) should be moved into drv_<sensor>.h files.
• Potentially move application of calibration to matrix/vector math
• We should remove IOCTLs from sensor drivers and replace with uORB "service calls" or similar.

[mhkabir]

@dagar I flew this a few times. Seems fine. Also gets rid of MAG SENSORS INCONSISTENT on a particular vehicle for me. Without the soft iron compensation, my GPS compass and the onboard compasses had some offsets between then.

[dagar]

This ends up being a lot of verbose duplicated code. Why don't we just start using Matrix?

[dagar]

Does anyone know the history here? I'm wondering why it wasn't already done.

[dagar]

@priseborough could you review this as well? Any specific suggestions for testing/validation?

[dagar]

I believe we could add member initializers here instead of doing it manually in every driver's constructor (if it's no longer used in any C code).

[stale]

This issue has been automatically marked as stale because it has not had recent activity. It will be closed if no further activity occurs. Thank you for your contributions.

[stale]

Closing as stale.

[dagar]

Completed in #15235.