accelerometer or gyroscope

Question

I'm building a camera stabilizing gimbal for use on a drone helicopter. I would like to smooth out some of the turbulance experienced by the copter in windy conditions. What sensor would better serve my needs, an acclerometer or a gyroscope? I am only interested in handling values in the X and Y directions. I also notice there are various "g" sensitivities (1.5g, 6g etc.). Any thoughts on that would help and probably save me time and $$.

Thanks

DAB · Accepted Answer

Hi Barry, I just remembered something else that I did with the camera control that you might want to consider. Use a smoothing filter on the X and y drive to introduce a limited amount of change in the positioning. If you try to keep up with real time, you will have a very shaky video as the camera gets banged about during turbulence or maneuvering. I set up a five degree max per update, which worked out fine for the application I had. Nothing upsets a pilot or operator more than having the video bounce all over the place. Just a thought. Good luck and keep us posted. DAB

DAB · Answer

Hi Barry, Victor is right. You need to firmly establish the vehicles position in space and its attitude with respect of surface coordinants. I once implemented the same problem on a helicopter with an infrared camera. First you have to establish the helicopters position and then fold those 3D parameters back into earth coordinants so that you can establish how the camera field of view is projected from the Helicopters point of view and then back into earth coordinants. So while you only need x and y movement commands, you still need to do the 3D space mapping to find out how you can point the camera at the right spot on the earth. The accelerometer can tell you the acceleration changes, but no the full angular movement. That is why you need the gyroscope. It can establish the yaw, pitch and roll deviations needed for the projection calculations. I implemented these calculations on an old Intel 8085 augmented by an AMD 9511 math coprocessor back in the early 1980's. So a good high end MCU with floating point math capability should be able to handle these changes in real time unless you plan to perform some really violent aerobatics. I hope this helps, DAB

vsluiter · Answer

Barry,

You'll probably need both; the gyroscope measures angular velocity (degrees/second), which is good for maintaining a constant angle. You just need to integrate the angular velocity to get the angular position, and that way you can keep the camera at a constant angle with respect to your copter.

However, that's probably not good enough; the gyros have a small error that adds up while integrating (because integrating boils down to adding up samples), and you probably want to keep the angle constant with respect to the earth, not the copter. That's why people use accelerometers. With an accelerometer you can determine the gravitational vector, i.e., you know where 'down' is. You can use this to correct drift in the integrated gyro angle EXCEPT for rotation around the gravitational vector. That's why people also add magnetometers to correct for that.

The downside of the accelerometer is that it also measures the acceleration of your copter. So if you move quickly, it will have a wrong picture of where 'down' is. Smart filters (Kalman / extended Kalman) are used to neglect acceleration data during fast movements.

To get back to your question: the most versatile solution is to use Invensenses overprized MPU9150 (available on Sparkfun breakout board). It features a 3D accelerometer, gyroscope AND magnetometer. With the default settings you can read out the raw sensor values, but with special firmware (google for Pansenti MPU9150) you can get orientation data out of the device.

To get even more back to your question:

You say you're only interested in 'values' in X and Y directions. But what 'values' do you need? linear displacement? Change in rotation? With respect to what? If you've got that clear you can probably find out easier which sensor fits your basic needs.....