I am looking forward to using the IMU brick in a motion control application.  What is the positional accuracy using an IMU brick using Quaternion math?  I realize that the 32 bits of data is more than I need, but am more interested in the IMU's ability to find itself in 3D space from a set starting point.

hi,

 

i think the critical point is more or less the time resolution you need. The longer you have to integrate the better your accuracy will be.

We want to fly a camera over a football pitch using the IMU to show camera orientation (pan, tilt and roll) and x/y/z position.  We will have as much time as needed to set up and the camera will be stationary for may seconds between 'shots'.  My question is can I trust the IMU positional data or should I stay with counting steps on the stepper motors to estimate the position. 

nice application,

 

i think some seconds should be enough to obtain an accurate position. But you will get only pan, tilt and roll.

Getting x,y,z will be much easier with other methods. 

Counting the stepper motor steps will definitely have a better accuracy.

 

The problem is that the timing of the sending of the USB package depends on the scheduler of your operating system. So instead of a message every ms you might get a message in 0.8ms and the next one in 1.3ms and the next one in 0.9ms and so on.

 

There is nothing we can do about that as long as the programming is done on one of the modern operating systems (linux/macos/windows).

 

If you want to calculate your postion with the IMU data, these timing differences will destroy your accuracy. The calculations would have to be _very_ precise.

 

In theory we could do the position calculations on the Brick itself. I already looked at that and tried some things. The problem is: If you use the IMU Brick together with two demanding bricklets (e.g. IO 16), we are already at the absolute limit of calculations we can do with the 64Mhz.

Thanks borg - I suspected as much.  I may still use the IMU as a back up and for calibration.