autocalibrating sparkfun 9DOF

I am very interested in the new 9DOF board on sparkfun (http://www.sparkfun.com/commerce/produc … ts_id=9623)

I am thinking I could get 3 boards, and mount each on the side of a cube rotated 45 degrees on the face. This would give me 9 axes for each sensor but no axes would be parallel. The advantage to having extra axes is threefold:

  1. You get lower noise by roughly 1/sqrt(n) so effectively 9 axes gives half the noise of 3.

  2. You can tell if a sensor or board fails quite easily, and with enough redundancy, continue operation.

  3. You can calculate the sensor scale factors and misalignments for all the sensors because the system is overdetermined. This is not possible with only 3 axes per sensor unless you have a special calibration procedure, and possibly special equipment for this.

It might even be economical to group boards to calibrate them, then store the calibration matrices on the boards, and separate them for use.

Any comments on this? Anyone who has these boards already, what kind of results are you getting?