CAD tools for MEMS should be developed that are easy to use for novices, electronic circuit designers, and MEMS experts. MEMS "experts" abilities are usually focused on sub-areas of MEMS and they too will greatly benefit from basic CAD tools. There is room for a hierarchy of representations and models of MEMS to satisfy the needs of a spectrum of users. Features of easy to use MEMS CAD tools include:
Many design issues govern the CAD representation of MEMS and VLSI frameworks. To see if a VLSI framework or design methodology is extensible to the MEMS field, the differences and similarities between the two must be explored. In VLSI, the major part of the design occurs in a single primary energy domain with interactions described primarily by electrical quantities i and v. In MEMS, the components of the system work in multiple energy domains - electrical, mechanical, thermal, radiant, chemical, magnetic, acoustic and fluidic - with coupling between them. Thus, any CAD framework for MEMS must support representation and analysis in multiple energy domains. In VLSI, a 2D layout abstraction is supported whereas in MEMS a 3D representation of structures is needed by the designer--especially for analysis. Process level design is also necessary in the MEMS arena and requires a separate design/simulation framework. Unlike VLSI design, there currently isn't a "clean separation" between the designer and the fabrication process for all MEMS processes. In addition, digital VLSI design typically has many higher levels of abstraction that are not currently used in MEMS design.
However, there are still many similarities between MEMS and VLSI design methods. The multiple levels and views of MEMS components and system used by designers are similar to those in VLSI. Different levels of abstraction are used for analysis spanning FEA/BEM to lumped element macro-models to functional models. MEMS bears a closer resemblance to MMIC or analog VLSI design where "custom" layout is common. There is a similar need to meet specifications and optimize layout to satisfy design and process constraints.
In order to provide suitable simulations of function to support MEMS design, appropriate design levels and representations are necessary. In addition, sufficient generality must be present to support multiple energy domain systems.
Some ideas can be extracted from the VLSI paradigm to form a design framework for MEMS. VLSI design evolved into a design hierarchy of different levels, as illustrated in Figure 5 (Matrix of VLSI Design Representations.). High to low levels of abstraction in VLSI include system, register, logic, circuit, device, and materials. At each level, there exists three different views: behavioral, structural, and physical. The behavioral view represents how a design works, the structural view represents how elements are connected together in a design, and the physical view represents how designs are implemented.
The structure of the VLSI design matrix may be applied with modification to formation of a MEMS design hierarchy. The concept of behavioral, structural, and physical views as equally important representations of each hierarchical level can be directly borrowed from the VLSI domain. Levels in our vision of a MEMS design hierarchy are somewhat parallel to those in VLSI, however there are some important differences. An initial partitioning of design levels from the top down includes system, module (sub-system), circuit, device, and material levels. This matrix of MEMS design representations is illustrated in Figure 6 (Matrix of MEMS Design Representations.) for a prototypical MEMS system. Pictorial illustrations at each level and view are taken from a relatively mature MEMS application, a torsional micro-mirror. The micro-mirror example couples together several energy domains including mechanical, electrical, fluidic, and optical. The example micro-mirror device is a micromechanical plate suspended by two simple beams and electrostatically deflected by a pair of parallel-plate actuators. Individual micro-mirror devices may be incorporated into larger arrays, and these arrays may be part of a larger system.
The system level deals with high-level architecture issues such as component placement, component interaction, upper-level packaging, and data bus routing. Systems could be implemented as a set of discrete parts, as a hybrid package, or as a monolithic microsystem. This highest level of abstraction in MEMS is not well understood and will be better defined as individual MEMS devices and processes mature. A "strawhorse" concept at this level is illustrated in Figure 6 (Matrix of MEMS Design Representations.). The physical view is a drawing of the packaged system that specifies the physical placement and size of its constitutive parts. The structural view specifies the high-level component interconnections, which includes coupling between multiple energy domains. The behavioral view includes many kinds of microsystem analyses, yet undefined, dealing with topological, architectural, and specification issues.
The module level deals with sets of components that combine in a sub-system to satisfy an individual function. The specific example in Figure 6 (Matrix of MEMS Design Representations.) is a micro-mirror array that provides a display function. The physical view displays the layout between micro-mirror devices, the structural view identifies the multi-domain interconnect between devices, and the behavioral view includes system analysis of the array. The specific needs for tools at this level may overlap with the requirements from the system and circuit levels. Unification of the various energy domains is one target area that could significantly impact usefulness of MEMS representations at both the module and system level. Such unification would enforce energy conservation of the behavioral representations, which is of fundamental importance.
The circuit level involves representation of individual sensor and/or actuator components. The physical view is device layout, which could be a 2-D, 2.5-D, or 3-D representation. The structural view is envisioned as a multi-domain MEMS schematic, specifying the connectivity between lumped-parameter MEMS elements. For the torsional micro-mirror example, the schematic representation specifies the connectivity of several MEMS lumped-parameter elements: a plate mass which acts as the mirror, two beam springs, and two electrostatic actuators. A schematic view at the circuit level does not presently exist for MEMS. Such a view may be an important intermediate representation for MEMS synthesis and analysis tools. The behavioral representation includes multi-energy-domain mixed-signal simulation in both the time-domain and frequency-domain.
The device level primarily deals with numerical representations of the MEMS device. The physical view is a full 3-D model of the device. The structural view is a discretized version of the 3-D model, with enough detail to extract the essential dynamic and static modes of the device. The behavioral view primarily involves numerical simulation, such as finite-element and boundary element analysis, with coupling between energy domains. Research needs include fast numerical analysis of large multi-domain problems. Making numerical analysis tools "MEMS-friendly" is important in order to improve accessibility of this technology.
The lowest level in the hierarchy is the material level. The distinction between the physical and structural levels is still unclear at this level. These views include a 3-D process view of the device showing connection between different materials. The behavioral view deals with process simulation and modeling. The simulations would include materials constitutive relationships at all relevant energy domains, such as predicting strain gradient in a material.
Several capabilities are required in the design representation to support movement between the different views and levels of hierarchy. This list is not meant to be comprehensive, but suggests some of the critical information required in the database which supports the design hierarchy. Some capabilities required of the design representation include:
Rapid, efficient, accurate, analysis of many alternatives during the design process is crucial to developing high-quality, robust devices and systems. Developing a widely-used, extensible framework for design analysis, perhaps building on the approach used in VLSI analysis (switch-level simulations, netlist extraction, etc.), will significantly shorten the time, and reduce the cost, of MEMS design.