An improved technique for estimating accelerometer linearity from multipoint data

The advantages and limitations of the static multipoint test are considered, and a test and data reduction procedure which avoids some of the difficulties of the traditional treatment is described. A simple linear model is initially assumed, and these values are subtracted from the indicated acceleration. The resulting differences are then examined in order to estimate the form of the model equation. Conventional least-squares fitting techniques are utilized to estimate the final coefficient values. Plots are provided for a number of common nonlinearities and, as examples, actual test data taken from a total of six accelerometers representing five different designs are analyzed, utilizing both analog and pulse rebalancing.