Originally Posted by clolson
Does anyone have any feedback on how well the sparkfun imu works? I haven't seen one running, but I've held one in my hands. I was a bit surprised that the z axis board was really loose. It seemed like movement within the unit itself could account for +/- 5-10 degrees in angular error. The whole thing had a less-than-solid feel. I assume a person would have to do some extra work to make sure all the bits are locked down relative to themselves so there isn't any flapping or movement within the unit. What is a little vibration going to do to it? Also, if the sensors aren't oriented exactly orthogonal to each other, there is going to have to be some sort of calibration that's done in software to get good results. I know higher end units will do this sort of calibration at the factory and compensate for assembly variation with some high end matrix mathematics. I don't want to sound too negative on the sparkfun unit here since haven't seen it running, but holding one in my hand, I'm slightly sceptical. I think you probably get what you pay for (and the sparkfun unit could be a little overpriced in my speculation.) But again, I haven't seen it actually running so I'm curious what real world results people are getting with it.
I am working on an IMU-based autopilot project with my neighbor, and we purchased the Sparkfun 6-DOF IMU (the cheap one). We were also very wary of the attachment of the vertical boards, but they are well connected and (relatively) sturdy. The sensors are noisy and can have a pretty big bias. We've got some simple gyro integration software running on it now (no compensation using gravity) and we can generate 10-20 degrees of error in a minute of hand-driven movement. You do get what you pay for.