Wednesday, March 30, 2011

IMU Calibration Plan

Step 1: Put the IMU into a square frame.

The bridge board is not precisely square, and the gyros are not precisely aligned with the accelerometers, and neither is precisely aligned with the compass. I don't think that this matters. I may be able to measure the orientation of the devices to each other, at least the accelerometers and gyros.

What I do need is a square frame. I have a wooden cube box which is square to my ability to measure it, with a circular bubble level. I have glued the bridge board and its battery into place, such that the switches, buttons, and USB port are all accessible.

Step 2: Warm and Cold.
It is known that the gyro has a temperature-dependent zero offset, and may have a temperature dependent scale as well. This is fine, as the device has a thermometer as well, which I read out with the rotation data. The accelerometer has a thermometer also, but I don't know how temperature sensitive it is.

So, one test will be done in my basement, as close as possible to my computer to make it nice and toasty warm. I'll probably render something during the test to keep things toasty.

Another identical test will be done in my garage with the doors open. This will keep it chillin' out.

Step 3: Accelerometers.
The ideal test site for an accelerometer doesn't exist. The surface of the Earth is great for a 1G signal, but to get a zero-G signal, I need a vacuum column drop chamber with a pillow at the bottom.

What I will do instead is record data with each axis pointing up and down. The surface I put it on won't be completely level, but the box is square, so when it is laid down one way it will be exactly opposing when it is laid down the other way. I will get a +1G and -1G signal on each axis, then average to get the 0G signal.

Step 4: Gyroscopes.
A record turntable would be ideal, but I don't have one. I should check Goodwill tomorrow. What I have instead is a non-driven food turntable. I will mark these such that it is evident when it is exactly at home position. Then I will spin the table up, let it coast a while, and when it slows down, move it to exactly home position. I will spin it both directions around each axis.

Step 5 (for the adventurous): Drop test.
Build a big pile of pillows in the basement. Climb as high as I can, then drop the IMU from the ceiling to the pillows. Try to have it as spin-free as possible, because the accelerometers are not at the center of rotation. The first fraction of a second after drop, before the package has built up any speed and therefore drag, will be the best. All axes will simultaneously go to zero. Repeat in the garage.

No comments:

Post a Comment