NSF Sponsored Workshop on Structured Design Methods for MEMS

VLSI and MEMS, VLSI vs. MEMS


Ted J. Hubbard
Department of Mechanical Engineering
Technical University of Nova Scotia
Halifax, Nova Scotia
hubbardt@tuns.ca

Micro-electro-mechanical Systems (MEMS) is an emerging technology with great potential. Commonly known as micro machines, MEMS refers to mechanisms and circuits manufactured with feature sizes of several microns, allowing for large volume production and low unit cost.

Although the fabrication methods for MEMS are derived from the VLSI (very large scale integration) microelectronics industry, MEMS is an interdisciplinary field combining electrical engineering, mechanical engineering, materials science, physics and chemistry.

As MEMS moves from individual devices to systems with many interacting components, mechanical design will increase in importance. Macro or "real world" machine design has developed a large body of expertise in synthesising complex mechanisms. These methods and formalisms are essential tools in the growth of MEMS.

This position paper will examine VLSI and MEMS, VLSI vs. MEMS; that is both the similarities and the differences between the two.

There are many lessons to be learned from VLSI that are applicable to MEMS. VLSI is perhaps the most highly developed form of structured design. A designer is able to specify a high level description of a device without having to know the details of the fabrication. In fact this clean separation of design and fabrication means that a designer would prefer not to know the details of the fabrication process. The structured and separated design methodology of VLSI led to revolutionary reductions in design and prototyping costs. Can a similar advance be made in the field of MEMS, especially given that MEMS fabrication technology is derived from VLSI techniques? The question is particularly relevant since unlike VLSI, some macro mechanical research such as concurrent engineering seeks to remove the separation between design and fabrication.

There are specific differences between VLSI and MEMS that will require modifications to VLSI design methodologies for use in MEMS design. VLSI design is based on a set of conservative design rules, so that the designer can be confident that the process variations will not significantly alter the performance of fabricated devices. Designers sacrifice performance for design methodology. MEMS fabrication technologies are developed from VLSI fabrication technology, but they tend to "push the envelope" in order to produce desired mechanical properties. Thus there may be a smaller performance margin available for sacrifice. The more critically the operation depends on precise process specification, the less clean the separation between design and fabrication.

MEMS has traditionally been compared to digital VLSI (in this paper the term VLSI is used to mean digital VLSI), perhaps a closer match is analog VLSI. Like analog VLSI, MEMS has "piggy-backed" on VLSI technology. Analog VLSI design depends on a detailed knowledge of fabrication technology and there is a much less of a separation between design and fabrication. As in analog VLSI design, MEMS design sometimes involves breaking the VLSI design rules. Analog VLSI and MEMS rely much more on custom devices rather than the standard libraries of digital VLSI.

VLSI geometry tends to be Manhattan (rectangular). The macro mechanical world, shapes are inherently curvilinear, a multitude of shapes are needed. Rectangular shapes and sharp corners are usually avoided, if only to reduce stress concentrations. For macro mechanical devices form is intricately linked to function. The shape of a mechanism, (a cam for example) critically effects not only its efficiency, but its function itself. This stronger link will lessen the possible separation between design and fabrication. Related to form is the three dimensional nature of mechanical devices. The function of a mechanical device also depends on its three dimensional orientation and the orientation of its neighbors. The connections between devices is three dimensional, compared to the scalar behavior of electrical current. Thus the three dimensional routing of power (fluid, thermal, mechanical, and electrical) needs to be considered.

VLSI devices can be manufactured from a relatively small set of primitives, a small library can be quite rich in content. Macro mechanical devices have a much larger set of required primitives; the minimum size of library for rich content may be quite large. The move from custom designs to standardized libraries may be delayed by incomplete libraries. MEMS devices which have large numbers of devices tend to be arrays of identical arrays, i.e., mirror arrays. This approach is closer to VLSI with its memory arrays, than to macro mechanical systems where many different devices are interconnected in one system. There may be two causes: small libraries and difficulty interconnecting different parts.

In summary, MEMS is often seen as bridging the gap between VLSI and macro mechanics. It offers an opportunity to apply VLSI structured design methods to mechanical systems. However, it is important to stop considering MEMS as peripherals to electronics. Their fundamental mechanical nature will necessitate a reformulation of existing VLSI design methodologies. VLSI CAD tools cannot simply be repackaged and used for MEMS, the experience of the macro mechanical research community should be incorporated. Having said that, VLSI structured design tools represent the foundation upon which MEMS structured design should be built. With a firm design foundation, MEMS can realize the same growth as VLSI.


Return to NSF MEMS Design Workshop Index