I am trying to understand this IMU calibration video.


In the video the operator gets accelerometer data when pointing each axis in the gravity direction. He then has an algorithm at minute 6:34.

I don't know where this algorithm comes from.

My understanding of how you get an accelerometer bias is you add the output when an axis is pointing in the gravity direction to the output when that axis is pointing antiparallel to gravity and then you divide by 2. Where is he getting this algorithm?


1 Answer 1


Let b, the bias error of a sensor.

If you take two measurements of a known value (gravitational acceleration in this case), in opposite directions, say m1 and m2, without any bias towards any direction the following is true:

1) m1 + m2 = 0.

However, if there's a directional bias error, the sensor readings become:

2) (m1 + b) + (m2 + b) = 2b.

Hope this gives some explanation.

  • $\begingroup$ I understand this. But this is not what he is doing. He has some M matrix (see video) and I'm wondering where that comes from $\endgroup$
    – rielt12
    Aug 24, 2017 at 17:33
  • $\begingroup$ Ok, In the video, the results of three sensors for each axis facing down is written down as a matrix (9 readings for positive placement and 9 more for negative ones) And they are finding cross-coupling errors as well, which can be interpreted as an orientation caused reading shift in each axis. The basic idea is expanded to three dimensions, from what I understand. $\endgroup$ Aug 24, 2017 at 18:03
  • $\begingroup$ This process is quite related to 3 d rotation and conversion from one coordinate frame to another one. (System dynamics topic). $\endgroup$ Aug 24, 2017 at 18:05

Your Answer

By clicking “Post Your Answer”, you agree to our terms of service and acknowledge you have read our privacy policy.

Not the answer you're looking for? Browse other questions tagged or ask your own question.