Replies to This Discussion

DeveloperPermalink Reply by William Premerlani on September 20, 2010 at 6:51pm

Dave,

In the general case, you need a yaw reference to correct for the Z gyro drift, such as a GPS or a magnetometer or both.

You can use either Kalman or DCM to somewhat reduce the gyro drift if the IMU is tumbling, but you cannot drive the drift to zero, and the technique does not work if the IMU is not tumbling.

Best regards,
Bill

• ▶ Reply

Permalink Reply by Dave on September 21, 2010 at 4:03am

"You can use either Kalman or DCM to somewhat reduce the gyro drift if the IMU is tumbling, but you cannot drive the drift to zero, and the technique does not work if the IMU is not tumbling."

Hi Bill,
I don t understand...
do you mean: "if the IMU is not moving about the Z axis?!?!?"

Theoretically doens t matter if the imu is tumbling or not and the magnetometer can register difference only if the imu is moving along Z axis
AM I right about that?!?

Thank you again!
All the best
Dave

• ▶ Reply

DeveloperPermalink Reply by William Premerlani on September 21, 2010 at 12:31pm

Hi Dave,

Let me clarify.

Lets suppose that you have an IMU with only gyros and acclerometers and that the IMU is level and stationary. If you are using the DCM algorithm with a PI feedback controller, it will achieve a perfect lock for roll and pitch. That is, the integrators will eventually exactly compute the slight offsets in the X and the Y axis gyros. However, there is no reference to adjust the Z axis gyro, so there will be a residual offset, and the direction cosine values that correspond to the projection of the X and Y earth frame axes into the body frame will slowly drift. The projection of the earth frame Z axis into the body frame will lock.

So, leave the IMU that way, flat and level, for a long time, there will be close to perfect cancellation of offsets for the X and Y gyros.

Now suppose that you roll or pitch the IMU by 90 degrees!!! Then, the body Z axis will no longer align with the earth Z axis, and the PI drift compensation mechanism will go to work on the Z gyro!!! Generally, the process of compensating the Z gyro will go faster than the rate at which the X and Y gyros start to drift again.

So, if you slowly perform a random tumble of the IMU, you will attain perfect compensation of all three gyros, without needing a magnetometer or a GPS.

However, this not mean that you don't need a GPS or a magnetometer if you are in a body that is tumbling. Although the average net drift for all three axes will be zero, there will be a net yaw angle error, because there is not any sort of yaw angle reference, only yaw rate compensation.

Best regards,
Bill

• ▶ Reply