Experience with Allan variance method for MEMS gyroscope performance characterization

Abstract

The noise performance of four low cost and one reference classic gyroscope is estimated using the Allan variance (AV). The purpose is to highlight problems connected with this process and to decide which parameters are the most important if we want to choose the best gyro for a particular application. From the AV theory, basic kinds of noise and their characteristic slopes of the AV curve are known. For example, angle random walk has the slope of −1/2 and is caused by the white noise on the gyro output. However, the output of low cost gyro is usually corrupted by the more or less correlated noise. The level of correlation affects the slope of AV curve, so the better sensor has the slope closer to the theoretical value, thus a simpler model can be used. When collecting data for analysis, it is the most important to suppress temperature variations and vibrations. Attention should be paid to choose proper sampling frequency. Also the resolution and the connection of an acquisition card should be considered.