Gyroscope questions: building a stable platform

Question

I'm embarking on a summer project and have a few questions. I'm planning to build a 3-axis stable platform for electronics. (Picture a flat surface that stays parallel with respect to the ground while the supports for this platform are free to rotate in any axis)

My questions are:

1. Is it better to use 3 1-axis gyros each located on the center of the axis of rotation or 1 3-axis gyro located in the middle of the platform?

2. Could I accurately compensate for gyro drift by using accelerometers and implementing a kalman filter?

3. I need an accurate and fast gyro for this application. I've been looking through a digikey catalog (http://www.digikey.com/product-search/en/sensors-transducers/gyroscopes/1967243?k=gyro&ColumnSort=1000011&fid=0&pageSize=25) and was wondering what specs I should look for. Do I want a gyro with high sensitivity (mV/°/s) or high range (°/s)?

Thanks Guys!!

Answer 1

1. Doesn't matter a whole lot. However, in consideration of question 2, you might consider getting one "6 DOF" sensor (3 gyros, 3 accelerometers). Integrated sensors probably have better sensor-to-sensor alignment than you can achieve with individual assembly.

2. Yes, but it isn't trivial to get it right, especially in highly dynamic situations. Look up terms such as "coning" and "sculling".

3. Sensitivity is more important than range. Remember, if the system is working correctly, the platform isn't moving at all. You want to catch any deviations (errors) as quickly and accurately as possible. If you had perfectly frictionless gimbals, you wouldn't need an active system at all — the mass of the platform would keep it oriented in space.

Answer 2

Just to be clear, any semiconductor based "gyros" are actually rate gyros, and to get the angle from them you will need to integrate the output from the rate gyro. That means that you will always have a drift component to this integrated output.

The more sensitive the output the better the integral error of subsequent processing. It may not change the actual rate error but you won't introduce additional error in integrating.

Other than putting real gyros in place you're going to have to use sensor fusion and some dead reckoning to even sure a flat platform.

Spacing accelerometer at a span across the platform will allow you to derive a second source of rate change.

A magnetic heading sensor (also possible in 3D) will help with an absolute measurement.

An optical sensor (AKA camera) that fixes on a target and helps correct for drift will also help. For example - a lot of the quad-rotor 'copters use a horizon detector (optical) for feedback and stability control.