have questions on MEMS gyros + magnetometers... / General Science and Electronics / Forums

Forums

4hv.org :: Forums :: General Science and Electronics

« Previous topic | Next topic »

have questions on MEMS gyros + magnetometers...

1 2 3 next

Move Thread

LAN_403

Patrick

Mon Jun 25 2012, 06:49AM

Registered Member #2431 Joined: Tue Oct 13 2009, 09:47PM
Location: Chico, CA. USA
Posts: 5639

Please ive researched this first in other places and i have some physics knowledge and formal education, but some of these questions i have, are a real punch to my brain.

Question: how is a magnetometer (3 axis) used to minimise/mitagate the drift rate of a gyro ( 3 axis too) of course i have a 3 axis accelerometer too...

is the magnetometer used to re-origin the gyro every so often to prevent cumulative errors due to the drift rate?
If so, how is this done?

the rate of drift is measured in degree second per second (d/s/s) right?

my brain hurts from all IMU/INU math and psyhics.

Steve Conner

Mon Jun 25 2012, 10:24AM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

The way I understand it, a gyro measures angular acceleration, and an inertial navigation system integrates twice to get the angular position. The double integration is very susceptible to drift. (An old-fashioned gyrocompass essentially did the double integration mechanically in the gyro itself, but the same drift issues apply.)

In the yaw axis, the angular position is the same thing as a compass bearing. A magnetometer is a compass, so it also gives a compass bearing that can be used to cross-check the one from the gyro. If they disagree, the integrators in the gyro can be reset to make them agree.

I don't know how this works in the other two axes, since the Earth's magnetic field isn't normally used for navigation in those axes. That's where the 3 axis accelerometer helps, it can sense the Earth's gravitational field and tell you which way is down. That can be used to reset the pitch and roll integrators.

You can think of the gyro and accelerometer working together to give you 6 axes of acceleration data: 3 translational and 3 rotational. This data can then be used for both stability control and navigation.

Patrick

Mon Jun 25 2012, 05:28PM

Registered Member #2431 Joined: Tue Oct 13 2009, 09:47PM
Location: Chico, CA. USA
Posts: 5639

Steve Conner wrote ...

A magnetometer is a compass, so it also gives a compass bearing that can be used to cross-check the one from the gyro. If they disagree, the integrators in the gyro can be reset to make them agree.

ok im going to investigate this further with C code, any other oppinons and explanations are still appreciated...

Patrick

Tue Jun 26 2012, 05:49PM

Registered Member #2431 Joined: Tue Oct 13 2009, 09:47PM
Location: Chico, CA. USA
Posts: 5639

ok, im going to consider strobing the gyro back to a zero rate value since my machine wont be continuosly in rotation in any single axis, therefore the magnometer can be used to reset the error out of the gyro on a timed baisis... I think.

any thoughts on how to do this in C?

Electroholic

Wed Jun 27 2012, 12:40PM

Registered Member #191 Joined: Fri Feb 17 2006, 02:01AM
Location: Esbjerg Denmark
Posts: 720

The standard way is to use a kalman filter, it will tie all the 9dof readings together. You should be able to find imu falman filter codes on the net.

Steve Conner

Wed Jun 27 2012, 01:29PM

Registered Member #30 Joined: Fri Feb 03 2006, 10:52AM
Location: Glasgow, Scotland
Posts: 6706

Sure enough, a quick google search for "imu kalman filter" found this:

Patrick

Wed Jun 27 2012, 05:45PM

Registered Member #2431 Joined: Tue Oct 13 2009, 09:47PM
Location: Chico, CA. USA
Posts: 5639

You guys save me once again, i wasnt even going to look at the kalman types until i had the pid + imu partially complete.

that source was exactly what i needed!!! i should probably startposting on the diydrones forum too. though 4hv is my favorite and i feel like im cheating on a girlfriend>>

Carbon_Rod

Sat Jun 30 2012, 01:35AM

Registered Member #65 Joined: Thu Feb 09 2006, 06:43AM
Location:
Posts: 1155

@Patrick
Make sure to audit an advanced linear algebra, and inverse kinematics course if these programs are not part of your curriculum. These are key to solving numerous problems, and will clearly explain how PID is related to Kalman filters.

Remember that the filter has an initial "learning" moment when state is first initiated, and actuators should be "inhibited" during this brief time (do not constantly "reset" your gyro, as "angular rate" data does not work that way). Additionally, you will need to "tune" the filters to match the moment-of-inertia for your craft.

Tip: separate your navigational and guidance sub-systems, and things will get a little easier.

Good luck at the competition,
Rod

Patrick

Sat Jun 30 2012, 03:42AM

Registered Member #2431 Joined: Tue Oct 13 2009, 09:47PM
Location: Chico, CA. USA
Posts: 5639

Carbon_Rod wrote ...

@Patrick
Make sure to audit an advanced linear algebra, and inverse kinematics course if these programs are not part of your curriculum. These are key to solving numerous problems, and will clearly explain how PID is related to Kalman filters.

I was afraid someone would say this!

Carbon_Rod wrote ...

Remember that the filter has an initial "learning" moment when state is first initiated, and actuators should be "inhibited" during this brief time (do not constantly "reset" your gyro, as "angular rate" data does not work that way). Additionally, you will need to "tune" the filters to match the moment-of-inertia for your craft.

Yes im following the advice from a expert in PIDs at the moment...and i wont strobe the gyro as i previously conjuectured.

Carbon_Rod wrote ...

Tip: separate your navigational and guidance sub-systems, and things will get a little easier.

Yes, i had already made this realization...

Ok so my question remains: how do i prevent the gyros from accummulating the drift error?
Obviously it can be done, my tilt-rotor from KKmulticopter does it...

tilt-rotor (twin-rotor) single board, uses only a 3 DOF gyro, no accelerometers no mags.

EDIT:

Source:

This article may explain things well, not sure yet.

Im making two assumptions:
first: rotational saturation is not a problem with solid state gyros.
second: the gyro drift isnt linear over dozens of seconds, (becuase i cant be that lucky)

Thought: when dealing with angular rates, absolute/instantaneous values are not stable. And therefore only fast changing rates are of value. steep slopes on the graph are useful, shallow slopes are closer to drift...

Carbon_Rod

Sat Jun 30 2012, 05:44AM

Registered Member #65 Joined: Thu Feb 09 2006, 06:43AM
Location:
Posts: 1155

As Steve stated...
there are many links for 9DOF IMU Kalman filters:

I would wager one of your chips is not part of the gyro, and is an accelerometer for a 5 or 6 DOF unit... Who knows, it may indeed be polling multiple gyros like a sync joke or something...

Essentially the predicted-error increases as the gyro drifts.

1 2 3 next

Moderator(s): Chris Russell, Noelle, Alex, Tesladownunder, Dave Marshall, Dave Billington, Bjørn, Steve Conner, Wolfram, Kizmo, Mads Barnkob