Statistical Modeling of Integrated Sensors for Automotive Applications

In this paper we show a way to solve the need for improved, faster and more reliable development of integrated sensor systems. This can only be achieved when production and calibration uncertainties are considered already in design phase. In our approach, we employ statistical methods base on known or estimated variances and correlations of system parameters. This allows us to model the system completely (under the assumption of Gaussian distributions), by using a reduced parameter set which only contains the aforementioned parameters and their covariance matrix. Since the proposed methodology is especially valuable for the automotive industry, a case-study is presented where we applied this methodology to an integrated automotive magnetic angle sensor system. We can thus show how effectively and reliably this methodology solves statistical requirement needs already in an early development-phase.