Coordinate frame description conflict between imu_cpp.h and online documentation

[KernelKlinq]

There appears to be an inconsistency with the coordinate definitions for the IMU. The X and Y appear to be switched for accelerometer readings. I am currently trying to make a new orientation definition that allows the MadFlight FC3 to be mounted vertically. The AHRS code seems to have pitch and roll reversed when the orientation from the imu_cpp.h is implemented, even though the NED and gyro orientations seem to make sense for NED.

The imu_cpp.h file has the following comment: (x and y appear to be incorrectly switched in the text below)

Body frame is NED:
x-axis North(front) - gyro-x roll right is positive - accelerometer-x vehicle right side down position is +1 G
y-axis East(right) - gyro-y pitch up is positive - accelerometer-y vehicle nose down postion is +1 G
z-axis Down - gyro-z yaw right is positive - accelerometer-z vehicle level position is +1 G

The following paragraph appears in Madflight online web documents and gives a different description of the orientation test, with nose-down reading on X instead of Y.

Gyro/Accelerometer Orientation (IMU)
The imu_align parameter sets the sensor orientation.
The label is yaw / roll (in that order) needed to rotate the sensor from its normal position to its mounted position.
The normal sensor position is NED (North East Down), i.e. x-axis points forward (N), y-axis points right (E), z-axis points down (D).
Use CLI pacc to display the IMU accelerometer outputs.
Holding the quad horizontal should give ax:0 ay:0 az:1, for example: ax:-0.02 ay:-0.00 az:+1.00
Holding the quad nose down should give ax:1 ay:0 az:0, for example: ax:+0.96 ay:-0.04 az:-0.07
Holding right side down should give ax:0 ay:1 az:0, for example: ax:+0.05 ay:+1.00 az:+0.06
If not, adjust the parameter imu_align and re-upload until this matches.

Does anyone know if the AHRS code is wrong?

[qqqlab]

The comment in imu_cpp.h was incorrect, thanks for reporting this. It has been corrected in master now.

AHRS should work correctly for the 8 possible imu_align options with the FC mounted horizontally.

If you want to mount the FC vertically you'll have to edit the update() method in imu_cpp.h to output NED coordinates. Or add the 16 possible vertical orientations to the imu_align enum and add corresponding code to the update() method. (Note: you probably also need to update the mag code.)

[KernelKlinq]

Thank you for verifying - I think that AHRS outputs are working now and both my mag compass and my accel-gyro are all reading what the rest of the Madflight code expects. I am attempting a very unconventional airframe (coaxial bicopter with suspended mass on a two-axis servo gimbal for pitch and roll) and I mounted the board vertical, pins down, components rearward, silkscreen forward.

I added new options to the gizmos in the cfg file and added the following case in imu_cpp.h:

case Cfg::imu_align_enum::mf_COAXFLOP : { float tmp; tmp=ax; ax=az; ay=ay; az=-tmp; tmp=gx; gx=gz; gy=gy; gz=-tmp; tmp=mx; mx=mz; my=my; mz=-tmp; } break;

I added the following case to mag.cpp:

case Cfg::mag_align_enum::mf_COAXFLOP : { float tmp; tmp=x; x=z; z=y; y=tmp; } break;

[qqqlab]

Hope you get it to work, sounds like a nice new year's project. Nice that you're is using this repo for its intended purpose! Would be great if you post some photos here.