NSF Sponsored Workshop on Structured Design Methods for MEMS

Structured Design Methods for MEMS

Ramaswamy Mahadevan
MEMS Technology Applications Center
PO Box 12889
Research Triangle Park, NC 27709-2889

Is there a sufficient similarity between MEMS and VLSI systems to borrow from VLSI design techniques?

VLSI systems use a small number of elements (devices) - transistors, capacitors and resistors - to build all other functional blocks and systems. Unfortunately MEMS does not appear to have as small a set of elements that can be utilized to synthesize any generic system functionality with a similar degree of flexibility. It should be possible to show that a moderately large functional design space could be covered using multiple instances and interconnection of elements from a small set of basic electromechanical elements, structured design techniques would be an immediate by product. (Perhaps along the lines of synthesis of mechanisms using four bar linkages?) Currently, MEMS design appears more akin to analog VLSI design; functional blocks like actuators, sensors, mechanical suspensions/springs, bearings, etc. are used in lieu of operational amplifiers, capacitors, resistors, and switches; interaction between blocks has to be taken into account.

Unlike VLSI, in MEMS the 2-D planar topology alone is not sufficient to model the electro-mechanical behavior of the system. It is necessary to incorporate process cross sectional information as well to generate the 3-D structures that will result when using a given mask layout - process combination. This requires the use of suitable device and/or technology models or representations. A few CAD programs do take mask and process descriptions to generate the 3-D structure that would result. However, there is a need to standardize the process description and database formats. There is also a definite need for layout representations and mask pattern generation software that support arcs and freeform structures in an efficient and accurate manner. (This also impacts on the time and cost of generating a mask that contains many arcs or freeform polygons.)

Unlike VLSI, separation of form and function may not be present. For surface micromachined devices, function of a device may be considered to be relatively independent of its position and orientation. However, the relative positions and orientations of devices in a bulk micromachining process do determine the structural shapes and functional behavior of the device.

Like VLSI, MEMS wafer fabrication and design can be considered to be cleanly separated where surface micromachined process are concerned. In fact independence of the fabrication process from the mask pattern is generally the desired process ideal sought. However, in some MEMS processes incorporating electroplating (such as LIGA), the fabricated device structure can depend on the mask pattern and/or fill factor of the mask pattern. (In addition, in MEMS processes such as bulk micromachining the fabricated structure also depends on the orientation and size of the layout pattern; however this can be accounted for in a process model.)

Despite these differences, MEMS can borrow from many of the VLSI design methods and CAD tools. Top-down and bottom-up design, synthesis, device and/or block level detailed simulation, functional or behavioral models and system level simulations are all equally applicable to MEMS. Even a version of LVS is possible for MEMS. The main obstacle hindering the easy use of such a approach is the multiplicity of physical phenomena that may be utilized in a MEMS system requiring (possibly coupled) mechanical, electromagnetic, fluid, and thermal analysis. For detailed simulations (equivalent to SPICE level simulations in VLSI) this can be done using existing FEA or BEA solvers. This requires an open CAD framework with a database format that includes mask layout and process technology information (process cross section and related mechanical properties) that can be easily exported/imported to/from the appropriate existing FEA solver. The alternate approach would be to use an equation based solver to model all the phenomena. SPICE itself can be used for functional or behavioral simulations of MEMS at the system level. There is also a need for simulators that are faster than FEA (and perhaps less accurate) analogous to switch level simulators in VLSI design.

The implementation of the following would help facilitate structured design and virtual prototyping for MEMS.

  1. A layout and process representation that allows efficient representation of arcs and freeform shapes along with process crossectional information, material properties, and perhaps even mesh information.
  2. An integral layout and solid modeling/viewing tool to help designers visualize 3-D structures that will result from the combination of layout and process description. (Existing tools can be adapted.) Visualization of 3-D structures that result from a given layout and process appears to be the biggest obstacle for new MEMS designers.
  3. An open CAD framework and data translators for sharing design information between existing and new MEMS specific CAD tools.
  4. MEMS specific mesh generation and analysis tools that are optimized for structures with thin layers (i.e., where one dimension is much smaller than the others)\@.
  5. Improved simulators for computationally efficient modeling of MEMS at block and system levels.
  6. A CAD tool for the synthesis of mechanisms and/or structural elements.
There are various existing CAD tools that already have some of these features. For example, Stanford's TCAD VIP-3D, MIT's MemBuilder and MEMCAD, and IntelliSense's IntelliCad have layout and process description, 3D model extraction and mesh generation features. IntelliCad does have an integrated framework including process description and modeling, layout, solid model generation, and FEA analysis.

Return to NSF MEMS Design Workshop Index