Autocalibration of MEMS accelerometers

In this paper a new approach of an auto calibration method for micromechanical sensors is proposed. In particular, recalibration of acceleration sensors without any additional laboratory equipment is considered. If the device is stationary, the proposed procedure exploits the fact that the output vector of the acceleration sensor should match the gravity acceleration. The calibration method computes the scale factors and the bias components of the unbalanced acceleration sensor. These parameters are computed through nonlinear optimization. The applied optimization method is a nonlinear parameter estimator based on the Unscented Transformation. This methodology uses the robust statistical linearization instead of the common analytical linearization. In addition, the applied methodology minimizes the amount of temporarily stored measurement data which are mandatory to launch the recalibration algorithm. Reducing the amount of temporarily stored data is equivalent to reducing the memory space and the power required for the algorithm. An effective method for rejecting disturbance acceleration is also included in order to apply user generated data for the recalibration. First the calibration method is evaluated through simulations and second with real data generated by an acceleration sensor. The simulation results show that the algorithm estimates the offset and sensitivity parameters more precisely than the uncertainty introduced through the measurement noise.