A signal processing method for improving the accuracy of MEMS gyroscopes

A technique is proposed whereby MEMS gyroscopes can be used to make accurate angular rate measurements. In this technique, signal processing techniques are used to combine the relatively inaccurate outputs of a large number of MEMS gyroscopes into a single accurate output without requiring an excessive amount of computation. First, averaging is used to combine sets of gyroscope outputs so that the effect of additive noises and noise due to parameter uncertainties is reduced. The several outputs thus formed are then combined into a single output by a dynamic element matching technique which essentially combines the noises due to random bias, white noise, and noise due to parameter uncertainties into a single measurement noise. The single output from the dynamic element matching process is sent to an extended Kalman filter for final processing. Each step of the process decreases the noise magnitudes or their effect on the accuracy of the gyroscopes; and therefore for an appropriate number of gyroscopes, a desired accuracy should be achievable. The system designer should also be able to control such additional design factors such as multiple signal `fan-in' and reliability. The technique presented in this paper is new and additional work is on-going to determine its practical value.