Extraction of reliability indicators from large-scale, highly variable timing data of MEMS switches

In this paper we present a method for monitoring and processing large-scale highly variable and nonlinear reliability data for MEMS RF Switches. The data is generated by measuring the switch actuation dynamics and a combination of statistical methods are applied to extract the switch closure/opening time. The signal processing is performed in a 2-step approach encompassing basic parametric and nonparametric statistical methods to correctly categorize the obtained datasets, supplemented with a mean-first derivative algorithm to extract the reliability indicators from the filtered data. The presented procedure proves to be highly accurate generating very low error in dataset classification. The acquired results give an insight into the evolution of switch health throughout its whole operation cycle and can lead to further understanding of failure mechanisms in micromechanical structures.