A method for electrical calibration of MEMS accelerometers through multivariate regression

This paper is the follow-up of the research we conduct to find an alternative way to measure the sensitivity of capacitive MEMS accelerometers. The goal is to be able to perform production test and calibration using only an electrical test bench, i.e. without applying acceleration. In the previously published study, we introduced three parameters that can be measured through electrical means and that are related to the accelerometer sensitivity through analytical expressions. However some restrictive assumptions were made on the conditioning electronic of the accelerometer. In this paper, the design for test capability needed to implement the measurement is discussed and improvement of the method is presented. It is based on a new set of electrical parameters that can be more easily measured and a MARS regression algorithm to find the relation between these parameters and the sensitivity. The method is evaluated through Monte Carlo simulations.