| ??? 07/30/09 13:54 Read: times |
#168032 - Centrifugal acceleration Responding to: ???'s previous message |
Kai said:
Well, additionally measuring the centrifugal acceleration isn't probably of much interest, so take only one ADXL202, put it at the center of mass, read out the x and y signal and compare with the g vector to find out the roll and pitch tilt angles. The ADXL202 mounted at the center of gravity is immune against centrifugal accelerations anyway. Farshid said:
I will try to think about this more. I don't quite understand how with putting one accelerometer at the center of a flying robot (if this was what you meant)it would be possible to neglect the upward/downward acceleration and proceed to getting the angle directly. Maybe I am oonly considering 2 dimensions. I didn't say "upward/downward acceleration" but "centrifugal acceleration"! Have a look again at the scheme: 
Assume the plane is flying without changing its speed and flight direction and the roll and pitch tilt angles are zero. Then, the two "green" acceleration sensors only feel the gravitational acceleration g, but because the component of g in parallel to the sensor axes is zero, the output signal is also zero. Now assume the plane is rolling. Then the component of g in parallel to the sensor axes is no longer zero but increasing and the output signal is also increasing. Speaking in pictures the reference masses inside the sensors begin to move to their "-" terminals. But that's not the only acceleration the reference masses feel. Due to the rolling the reference masses are forced to move on a circular path, which corresponds to a centripetal acceleration acting on the spring frame, which the reference masses respond with a centrifugal acceleration of opposite sign. By other words the rolling also makes the reference masses to move away from the roll axis. So, in the one green acceleration sensor the reference mass is moving to the "+" terminal and in the other to the "-" terminal causing identical signal components but of opposite sign. And because of adding the signals of both roll tilt sensors the impact of centrifugal accelerations will cancel. Make no mistake, this cancelling only works if the sensors are mounted symmetrically arround the roll axis and only with this rolling. Upwards and downwards accelerations which act on the whole plane (center of gravity) do not cancel by this scheme. If you take only one sensor and put it directly at the center of gravity, i.e. directly on the roll axis, then the sensor is automatically immune to this kind of centrifugal accelerations, because the distance to the roll axis is zero then. But, again, no immunity to upwards and downwards accelerations acting on the whole plane. Farshid said:
Perhaps so, I think I kinda remember other quad rotor projects using more than one accelerometers along Gyros. Have to go through them again more carefully to see the reasons for that. What you will mostly find are schemes measuring the acceleration of two dimensions (X/Y), which can be done by one ADXL202, but not using two ADXL202. There are some schemes using two ADXL202, but these measure the tilt angles of X/Y/Z, which the authors of your link do not. Kai |
