I could averaging the values I measure and that would get rid of most of the noise, no need for a kalman filter! I would still have the thermal noise but I am assuming here it would be rather small.
You need to be a member of diydrones to add comments!
One source of error is vibration - if you want to average multiple gyros - better to use different makes with different operating frequencies - otherwise, they could all react the same way to a harmonic.
The drift that a Kalman (or other kind of complementary filter) eliminates isn't just from noise.
Try sitting in the passenger seat of a car, reading the speedometer once per second, and using that information to work out the number of inches you've traveled. You'll be reasonably close, but not exact. How close you are will depend on how accurate your numbers are, how often you take samples, and how much accelerating and decelerating the car is doing.
The Kalman filter deals with this problem by using a real world measurement that's noisy or infrequent, like the road signs that tell you how far you've gone. Using that information you adjust the estimate you've been making with the short-term numbers (the speedometer values) to correct for the accumulating errors.
One way of eliminating erroneous accelerometer measurements is to throw out any measurements that deviate from one "g" (9.81 m/s^2, 32.2 ft/s^2) by some margin and requiring that measurements be within that range for a given period of time.
Noel Hughes
Replies
Try sitting in the passenger seat of a car, reading the speedometer once per second, and using that information to work out the number of inches you've traveled. You'll be reasonably close, but not exact. How close you are will depend on how accurate your numbers are, how often you take samples, and how much accelerating and decelerating the car is doing.
The Kalman filter deals with this problem by using a real world measurement that's noisy or infrequent, like the road signs that tell you how far you've gone. Using that information you adjust the estimate you've been making with the short-term numbers (the speedometer values) to correct for the accumulating errors.
Jason
Noel Hughes