An update regarding discontinuity on quaternion data: We always warm start our BHI160B by uploading previously saved acc/gyro biases to the sensor and ensuring that the acc/gyro accuracies both go to level 3 before any test. Recently, we also tried to perform a 6-face calibration after the warm start and in this situation discontinuity in Game Rotation Vector quaternion data is slightly less frequent but still exists which leads to more than 30 deg error in Euler angles. We tried to match the time when discontinuities happen with changes in acc/gyro biases or other events but we cannot see any associations. We perform our tests in a way that we hold the sensor in hand and swing our arm to +- 60 deg. We also have a platform which has a little bit of vibration and the same movements make discontinuity even more frequent. Any advice to solve this problem is highly appreciated.
... View more