Enhancing MEMS Design Using Statistical Process Information

Micro electro-mechanical systems (MEMS) and devices are taking an increasingly active role in products ranging from automotive airbag sensors to micromirrors used in projectors. Even with the increasing use of MEMS devices, they are still largely overshadowed by their VLSI cousins. MEMS designers typically use VLSI-tuned processes to develop and even produce MEMS devices. This practice leads to a number of manufacturing tolerance related issues because the design rules used for VLSI are not always appropriate for MEMS devices. Both VLSI and MEMS devices are susceptible to mechanical variation but they have different trade-offs between dimensional control and performance. This thesis focuses on identifying and developing a methodology to tune MEMS devices to be robust to process variation and hence increase manufacturability. This goal is addressed on two fronts. First, a flexible, object oriented, process representation called the Semiconductor Process Representation (SPR) has been implemented which provides process and device designers a single representation to share fabrication process information. The use of views enables each user to see the data in a domain-specific fashion. A prototype editor has been developed to browse and modify the SPR data structure, with particular emphasis here on process variation information. SPR provides a single unified data model for design, simulation, and fabrication of MEMS and VLSI devices. Next, a methodology has been developed which reduces the sensitivity of devices to process variation. The sensitivity of minimum geometry structures to process variation has been identified as a key starting point for device enhancement. By identifying and optimizing critical geometries, a device can be efficiently optimized to be much more robust to geometric process variation. A parameterized resonator synthesis tool has been enhanced to include and test these geometric optimizations. Finally, a complete distributed system including an SPR server, editor and resonator synthesis tool has been demonstrated. This prototype system enables designers to create and fabricate robust devices based on actual process statistics. Thesis Supervisor: Duane S. Boning Title: Associate Professor of Electrical Engineering and Computer Science. Thesis Supervisor: Donald E. Troxel Title: Professor of Electrical Engineering and Computer Science.