Explore sea calibration feasibility
Context
After the discovery of a lot of LSM303 devices giving off-nominal results at the DOM integration stage, it was observed compass calibration constants can change between the CLB calibration and its integration in a DOM. Thus, it was proposed to explore in-situ calibration, when the constants are not expected to evolve anymore. This issue is made to share and discuss first results in this direction.
Technical idea
Due to the erratic movement of DOM under the sea, the raw compass data should be distributed on a sphere, for which the center is not on 0. The calibration process is (mainly) correcting this offset, and that's the value we would like to extract from sea data. Fitting the raw data, the first goal is to extract and compare the obtained shift to the calibration constant we have in the DB, for some well calibrated DOM.
To help in this process, few ideas have been considered:
- As a first try, a sphere-fit tentative is available in
km3compass
:calib_self_sphere
(example). However, not too much time should be invested on this module, which is probably a too crude implementation. - Fixing the sphere radius, using other DOMs input as they should experience the same magnetic field.
- Using the accelerometer data in the fit, which should show the same relative displacement than the compass data.
Few points are identified as potential issue:
- The analysis should probably focus LSM303 devices, and not AHRS, as this last have other issues. Also, LSM303 point to 2 different version of the devices: LSM303AGR and LSM303D (see #1 for more details). Both are supposed to have different scaling factor to obtain a field in gauss.
- The longer is the considered period, the larger would be the DOM movement. As the DB sampling is about 1 measurement per minute, the data have to be down-scaled. This down-scaling and its impact are something that should be studied after obtaining a first working approach.