[velocipenguin]

It seems to me that much of the noise inherent to analog accelerometers could be minimized through clever use of shielding and filtering. Team 190's inertial navigation system, while quite impressive, does not appear to be particularly well shielded; the top of that housing appears to be made of Lexan or similar material, which does not provide any electromagnetic shielding. Manufacturers of analog accelerometers usually suggest leaving most of the extraneous copper unetched and grounded on whatever PCB is used, which provides a nice big ground plane surrounding the sensor and its board traces. Performing all of the integration in software would also help somewhat, as op-amp integrators are alarmingly sensitive to noise. Using a simple approximation, such as that suggested elsewhere in this thread, reduces the sample rate enough to filter out quite a bit of the high-frequency noise without compromising the accuracy of the position data. Low-pass filters should be placed on all accelerometer outputs to reduce ripple from the accelerometer's internal clock. Finally, it would be wise to make sure the power supply is VERY well filtered. By paying careful attention to reducing or eliminating noise from external sources, it should be possible to construct an inertial navigation system that is suitably free from error to provide accurate position information throughout the entire match.