Efficient noise detection in sensor systems with limited computing resources

Modern smart sensors require the integration of signal processing facilities to provide fault tolerance and self-checking. For high volume production, optimized controllers are employed that typically have only limited computing resources. In this paper, we investigate several algorithms devised to assess the noise in a closed-loop measurement channel under the assumption that the stimulus can be held constant during noise measurement. Starting from the definition of standard deviation, we propose several modifications and obtain an easy-to-implement algorithm relying entirely on addition and shift operations. Numerical experiments verify the practicability of the approach. The proposed algorithm has already been successfully implemented in a capacitive angular speed sensor system for automotive applications.