search

kriswiner / BMX-055

22 stars 9 forks source link

Question about degraded result when By is up #5

Open hww opened 4 years ago

hww commented 4 years ago

The filter works now in all cases, except when the sensor is turned so that By looks up and Bz is on the horizontal plane. The graph shows this sequence: 1) With Bz looking up, I rotate the sensor 90 degrees back and forth around Bz axis. It works well. 2) I turn the sensor so By points up and Bz is on the horizontal plane. 3) With the sensor as in (2) and keeping Bz in the horizontal plane, I rotate the sensor 90 degrees back and forth around By. The filter's result is degraded. bmx055_issue

kriswiner commented 4 years ago

I think this is a problem of yaw degeneracy, i.e., which direction is South when you are standing on the North pole?

On Tue, May 26, 2020 at 5:59 AM Valery notifications@github.com wrote:

The filter works now in all cases, except when the sensor is turned so that By looks up and Bz is on the horizontal plane. The graph shows this sequence:

  1. With Bz looking up, I rotate the sensor 90 degrees back and forth around Bz axis. It works well.
  2. I turn the sensor so By points up and Bz is on the horizontal plane.
  3. With the sensor as in (2) and keeping Bz in the horizontal plane, I rotate the sensor 90 degrees back and forth around By. The filter's result is degraded. [image: bmx055_issue] https://user-images.githubusercontent.com/773541/82903543-c729a880-9f69-11ea-9d82-8d7cc02548a6.png

— You are receiving this because you are subscribed to this thread. Reply to this email directly, view it on GitHub https://github.com/kriswiner/BMX-055/issues/5, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABTDLKRDML44WFB4HFSIM7LRTO4KRANCNFSM4NKJDASA .

hww commented 4 years ago

Thank You,

It turned out the problem was bad magnetometer calibration. After calibrating the sensor properly the problem disappeared.

Two questions: Is the Madgwick filter suitable for motion capture applications, i.e., can it track relatively quick rotations and settle down when motion pauses? About the test conditions for fastcompaccelBMX055: must the sensor be in any particular orientation?.

kriswiner commented 4 years ago

1) depends on your MCU speed.

2) yes, generally z-axis parallel to gravity and motionless.

On Tue, May 26, 2020 at 11:33 AM Valery notifications@github.com wrote:

Thank You,

It turned out the problem was bad magnetometer calibration. After calibrating the sensor properly the problem disappeared.

Two questions: Is the Madgwick filter suitable for motion capture applications, i.e., can it track relatively quick rotations and settle down when motion pauses? About the test conditions for fastcompaccelBMX055: must the sensor be in any particular orientation?.

— You are receiving this because you commented. Reply to this email directly, view it on GitHub https://github.com/kriswiner/BMX-055/issues/5#issuecomment-634201294, or unsubscribe https://github.com/notifications/unsubscribe-auth/ABTDLKURCSUI3QT3DAG3AILRTQDPTANCNFSM4NKJDASA .