Linearizing an electrostatically driven MEMS speaker by applying pre-distortion

The market for tablets, laptops and mobile devices is increasing rapidly. Product designs trend to thinner housings at concurrently higher energy consumption, which plays a major role for battery-powered devices. Microelectromechanical (MEMS) speakers, fabricated in complementary metal oxide semiconductor (CMOS) compatible technology, merge energy efficient driving technology with cost economical fabrication processes. A recently developed finite element (FE) model for the precise computation of the non-linear MEMS speaker is presented. In addition, dynamic linearization is achieved by applying feedback linearization to the local model network (LMN) as a data-driven dynamic model of the MEMS speaker to obtain a dynamic feedforward control (FFC) for pre-distortion. In the FE model, we fully take into account geometric non-linearity, contact in the snap-in mode, non-linear electrostatic coupling forces as well as the moving of charged thin membrane-plate structures (electrodes) in an electric field. Furthermore, since we concentrate on MEMS loudspeaker applications, quality of sound plays a major role, where an undistorted and linear system defines the challenge. In the process of MEMS speaker design, both, the FE model and the FFC using LMN, can be used to optimize the MEMS speaker towards sound pressure level (SPL) and total harmonic distortion (THD).