Statistical optimal design of microelectromechanical system (MEMS)
A three-step technique for MEMS quality optimization is demonstrated. It exploits the relative merits of its constituent optimization components. Analog Devices' ADXL50 accelerometer was the test device with microstructure dimensions serving as design parameters. Manufacturing design rules [7] and sensor performance requirements served as design constraints. A static model relating input acceleration to sensed voltage was used, neglecting sensor and signal conditioning dynamics. The trimming yield of the ADXL50, with sensitivity as the design target, was improved by 37%. We show that the three-step method enables the search for an optimal design in a semi-automatic manner by facilitating user interaction. Our final goal is to enable the generation of MEMS mask layout while ensuring robust designs with minimum sensitivity to fabrication process variations.
