David Bindel, Jason V. Clark, David Garmire, Babak Jamshidi, Shyam Lakshmin, Jiawang Nie Alice Agogino, Zhaojun Bai, James Demmel, Sanjay Govindjee, Ming Gu, Kristofer S.J. Pister

1
David Bindel, Jason V. Clark, David Garmire,
Babak Jamshidi, Shyam Lakshmin, Jiawang NieAlice
Agogino, Zhaojun Bai, James Demmel, Sanjay
Govindjee, Ming Gu, Kristofer S.J. Pister
SUGAR A MEMS Simulation Tool
http//www.sourceforge.net/projects/mems/
Introduction
Design Cycle

• Our goal Be SPICE to the MEMS world
• Provide quick simulations for tight design loop
• Use structural elements like beams and plates,
  circuit elements, and elements that capture
  coupled effects
• Be sufficiently simple to use for brainstorming
• Be a building block for design synthesis and
  other simulation-intensive activities
• Provide framework for measurement comparison to
  validate existing models, derive new models, and
  extract physical parameters

• Users describe devices using text netlists
• Models written in C or Matlab describe netlist
  components
• Global equations assembled from local model
  equations
• Solvers analyze the global equations
• Users invoke solvers from Matlab, Lua, the web,
  or other programs

Netlist
Models
Assembly
Solvers
User Interfaces
MMEMS SUGAR on the Web
SUGAR in action

• Hosted on Berkeley Millennium cluster
• Requires only a browser
• Version 2.0 used in Fall 2001 MEMS course
• Version 3.0 will be used in Fall 2002 MEMS course

• Continuum element discretization might use O(106)
  DOF
• SUGAR model is 10K DOF
• Described with a few parameterized subnets
• Simulated frequency response of micromirror took
  30 sec using model reduction techniques

Measurement feedback
Recent Progress

• SUGAR 3.0
• Initially released 4/5/2002
• Core functionality moved from Matlab to C/C
• Model development
• Almost all SUGAR 2.0 models ported
• Beam model with anisotropic warping effects (in
  testing)
• Plate model with condensation of internal degrees
  of freedom (in testing)
• Numerics
• Model reduction for fast frequency-response
• SuperLU for linear solves

KSJP15

• User interface prototyped and measurement code
  improved (to gt 1 frame/sec)
• Optical measurement facilities provided by
  Mullers Matisse group
• Validate simulations by comparison to measured
  data and improve models