Calibration state in IMU_PLUS

edited November 2018 in JavaScript

(editing the whole post as I discovered the error in the original post)

I just realized that only the NDOF mode has calibration state, while the IMU_PLUS does not work (reading the signal will break connection with sensor).

1) Isn't it a bug given that IMU_PLUS still uses acc/gyro to get the sensor orientation?

I see that (after calibrating) if I disconnect using the mw_debug_disconnect, next reconnection the calibration is still at high accuracy, but with mw_debug_reset is not.

2) Is there a possibility to cache/store that calibration so after a reset or power on can be reused or should we avoid at all costs that the sensor runs out of battery and/or needs a reset?
3) If the sensor is kept powered on, let's say one week, moving in different environments, how does it affect to sensor fusion calibration? I am talking about magnetic interferences.

As a side question regarding calibration for "production mode":

4) How may "commercial" products (IMU-based motion capture suits) may be doing their calibration? I don't see any "move each of the 17 IMUs provided with the suit 0-45-90-135-etc... degrees in an axis and perform an 8 in the air" in their documentation.

Comments

  • edited November 2018

    (deleted post, moved content to original post for clarity)

  • @aribes said:
    1) Isn't it a bug given that IMU_PLUS still uses acc/gyro to get the sensor orientation?

    https://mbientlab.com/community/discussion/2664/firmware-v1-4-2#latest

    I see that (after calibrating) if I disconnect using the mw_debug_disconnect, next reconnection the calibration is still at high accuracy, but with mw_debug_reset is not.

    2) Is there a possibility to cache/store that calibration so after a reset or power on can be reused or should we avoid at all costs that the sensor runs out of battery and/or needs a reset?

    That is a feature in the works.

    3) If the sensor is kept powered on, let's say one week, moving in different environments, how does it affect to sensor fusion calibration? I am talking about magnetic interferences.

    Don't use the magnetometer if you are worried about magnetic interference.

    As a side question regarding calibration for "production mode":

    4) How may "commercial" products (IMU-based motion capture suits) may be doing their calibration? I don't see any "move each of the 17 IMUs provided with the suit 0-45-90-135-etc... degrees in an axis and perform an 8 in the air" in their documentation.

    Don't know nor is that relevant if they are not based on Bosch sensors / software. For our boards, follow the recommend Bosch motions.

  • @Eric said:

    @aribes said:
    1) Isn't it a bug given that IMU_PLUS still uses acc/gyro to get the sensor orientation?

    https://mbientlab.com/community/discussion/2664/firmware-v1-4-2#latest

    Thanks, we had 1.4.1 installed. I will update our boards.

    [...]

    3) If the sensor is kept powered on, let's say one week, moving in different environments, how does it affect to sensor fusion calibration? I am talking about magnetic interferences.

    Don't use the magnetometer if you are worried about magnetic interference.

    Without the magnetometer, orientation is not in a global frame of reference. How is the frame of reference computed when using IMU_PLUS?

    We show the orientation in a 3d cube (using javascript). If using the magnetometer (NDOF), after randomly moving the sensor and then keeping it still, the cube also stops but then drifts for about 1-2 seconds updating its orientation. But this does not happen with IMU_PLUS (which is the reason why we prefer that mode).

    Could it be the case that the sensor fusion algorithm takes "some time" to incorporate the magnetometer (heading) readings into the computed orientation? How does the sensor fusion accelerometer/gyroscope data range affects it?

    [...]

  • @aribes said:

    3) If the sensor is kept powered on, let's say one week, moving in different environments, how does it affect to sensor fusion calibration? I am talking about magnetic interferences.

    Don't use the magnetometer if you are worried about magnetic interference.

    Without the magnetometer, orientation is not in a global frame of reference. How is the frame of reference computed when using IMU_PLUS?

    Orientation is relative from the board's initial position.

    We show the orientation in a 3d cube (using javascript). If using the magnetometer (NDOF), after randomly moving the sensor and then keeping it still, the cube also stops but then drifts for about 1-2 seconds updating its orientation. But this does not happen with IMU_PLUS (which is the reason why we prefer that mode).

    Could it be the case that the sensor fusion algorithm takes "some time" to incorporate the magnetometer (heading) readings into the computed orientation? How does the sensor fusion accelerometer/gyroscope data range affects it?

    That is expected. The magnetometer is used to correct saturation and is only sampled at 25Hz.

This discussion has been closed.