clivee1
Member
It would be best to use the accelerometer for high-frequency height control and use the wand for low-frequency height control. The accelerometer will be subject to bias, scale factor, and alignment errors that will result in unbounded (ie, very, very large) errors if the acceleration is integrated to get vertical velocity and height. However feeding back wand position to the height estimate will stabilize the estimate against these errors. At the same time, the accelerometer will respond faster to a change in height than will the wand, making it most suitable for adding lead to the height control.
This is what I do. I do not amplify the accelerometer input to provide a phase advance, but do use the vertical velocity in the height control loop to achieve the same thing through a controlled gain setting.
Attached pic show data log from moment of instability, and a full 18kt crash. At the moment I have the height coupled to the wand quite tightly (0.5 sec) but this still filters a significant amount of chop that would feed through to the servo.
Scalling:
Height (mm)
Vvel (mm/S)
Vacc (mm/S/S)
The height trace follows the wand with no phase lag, but with reduced chop.
As my confidence grows I am sure that I will be able to further decouple and fly in inertial space.
This is what I do. I do not amplify the accelerometer input to provide a phase advance, but do use the vertical velocity in the height control loop to achieve the same thing through a controlled gain setting.
Attached pic show data log from moment of instability, and a full 18kt crash. At the moment I have the height coupled to the wand quite tightly (0.5 sec) but this still filters a significant amount of chop that would feed through to the servo.
Scalling:
Height (mm)
Vvel (mm/S)
Vacc (mm/S/S)
The height trace follows the wand with no phase lag, but with reduced chop.
As my confidence grows I am sure that I will be able to further decouple and fly in inertial space.
