Toilet Bowl Effect after all calibration

PX4 Autopilot
1

Hello everyone,
Recently I had our quadcopter drone tested in an open grass field with full sensor calibration conducted including mag calibration. However, as you can see in log 1) 3D view, the drone suffered TBE as switched to position mode.

  1. TBE sample log
    https://review.px4.io/plot_app?log=7aeaaaca-0466-443a-ab90-3999fba3f06e
    2)Normal case in the same scenario
    https://review.px4.io/plot_app?log=b0c80463-7991-4150-8204-70755f73e902

I’ve quite studied the TBE factors and other similar cases on px4 discuss. I personally think it wasn’t mag’s or IMU’s fault since the following conditions are met:

  1. mag calibration was applied (should’ve solved offset/scale)
  2. roll/pitch/yaw angles fit closely to setpoints
  3. no significant vibration compared to the normal case

I was wondering if it could be the GNSS problem since it rtk wasn’t performing very well that day even in opensky. Sincere gratitude to anyone who may offer suggestions to clarify this issue.

2

Hello! What type of MAG used? Is it 3D? It can also be magnetometer, even if calibration is done. What type of fusion used (param EKF2_MAG_TYPE)?

3

Hello! First of all, great thanks for the reply. We use MAG_3D as the fusion type since it seems to be a good method for solving leveling issue.