Authors: L. Nguyen, H.J. Lee, M.A. Maher and H. von Sosen
Affilation: Tanner Research, Inc., United States
Pages: 274 - 27
Keywords: system level simulation, methodology, modeling, optimization, MEMS
A Methodology for System Level Simulation, Modeling and Optimization of MEMS Devices Categories: Co-simulation and Optimization System, Multi-level Modeling Linh Nguyen, Hee Jung Lee, Mary Ann Maher, and Harald von Sosen Tanner Research, Inc. VLSI systems designers can begin their design process with a behavioral description of a system, then proceed to system simulation, optimization, netlist generation, and physical design. MEMS designers typically begin with device design, this process is often followed by several fabrication iterations for device characterization, and the rest of the system-the electronics and packaging-are layered on top of the device in successive design iterations. This methodology results in long design cycles and is very costly. A more structured approach to MEMS CAD would permit the designer to specify system level performance goals and then propagate those goals down to a synthesis tool that generates physical layout for the devices. In this paper, we report on new optimization, modeling, simulation and design methods for MEMS systems. We have extended our VLSI circuit simulation tools to include new capabilities for hierarchical optimization and modeling of MEMS and electronics. This work follows that first proposed by Fedder and Mukherjee . Our optimizer provides flexibility in the choice of system goals and can consider a general set of design and process parameters. A functional modeling interface was added to our ODE system solver, allowing us to model a system's behavior hierarchically as a composition of components. The functional modeling interface can be used to model any MEMS or electronic component as an n-terminal device with through and across variables. MEMS behavioral models can thus be treated as functions of geometry, process parameters, and boundary conditions. The functional models themselves can be of a more general forrn than those found in VLSI design and can include non-linearities. The optimization tool is used to control the ODE system solver, it relates the system level goals down to the component level models. The goal functions, the behavioral models of components and models of the electronics are then solved as one system. Goal functions may be based on simulation results in time domain or AC analysis and user specified functions may also be used. We demonstrate the design of a sample system using these new methods which have been incorporated into a commercial design tool suite for the analysis, design, and optimization of MEMS systems . Our example system is a Q-controlled resonant filter  and it is composed of two comb drives, a plate, a folded-spring structure, and electronic control circuitry. With the resonant frequency set as the system goal, the optimizer sought a set of device parameters (in this case, beam length and width for the folded spring and length and width for the plate) which would produce that resonant frequency. To verify our approach we compared the resonant frequencies of several fabricated resonant filters with values predicted by our optimization tool. A comparison of predicted versus measured resonant frequencies for a variety of device parameters is shown in Figure 1.
Nanotech Conference Proceedings are now published in the TechConnect Briefs