Hi!
I would like to point out that we are talking about two different possible sources of distortion here:
Sensor calibration
YJBeetle is referring to the sensor calibration, which is performed once (or a few) times in the lifetime of a KCeptor.
There's a way to visualize the calibrated magnetic ellipsoid with the Chordata Blender add-on. A properly calibrated magnetomer should give a sphere that looks like the one shown by YJBeetle (non deformed and centered at the origin)
plug a Vector node to the output of the notochord node
And use the --raw flag on the notochord. You will see an arrow which points in the direction of the selected vector. Note that the Vector node has the same object as target than the Test cube node (KCeptor_R1 in this picture), this will give you a visualization that is oriented in world coordinates. Rotate the sensor in all directions and you should see a sphere being created by the tip of this vector.
Note also that the magnetic vector in world coordinates should point roughly in the same direction while you move the sensor, since the magnetic flow at a certain point in space should be constant (given no external disturbances or long term magnetic drift is taken into account). This vector has a vertical component that is bigger than the horizontal component in higher latitudes, either north or south. In the north it points mainly downwards and in the south it should point upwards. This is somewhat counter-intuitive.
In-pose calibration and real-time disturbances
The current calibration relies on magnetic data to rotate the heading of the sensor. A small room full of electronic objects and probably metallic structure would probably create significant disturbances on the magnetic field. If you perform this calibration in such an environment you will probably obtain a heading offset that increases in the lower parts of your body (normally closer to electronic devices or the metallic structure on the floor). cubesky it would be great if you can test the suit in an environment which you know is completely clean of local magnetic disturbances. A park or open space is normally the easiest way to go, you just need a laptop and a cellphone to create a wifi hotspot.
As I said before we are currently working in the implementation of the new pose calibration algorithm in the blender add-on. We should be able to release a first version soon. This new algorithm development is already finished and has given us a much greater results with captures recorded in magnetically disturbed environments, take a look here.
So it would be great to test it in your place once it's available. But in the meantime performing the test I described with the current algorithm can help us diagnose the source of your current problem.
Of course, after the calibration the disturbances present in the room still disturb the capture. The cleaner the space where the capture is done the better. We will be working in the magnetic disturbance compensation step during this first part of the year. There's already one in the notochord but is disabled by default because it created some artifacts under some circumstances.
