NSF Sponsored Workshop on Structured Design Methods for MEMS

Getting a Clean Separation between Design and Fabrication

Jack Hilibrand and Bernard Chern
Division of Microelectronic Information processing Systems (MIPS)
Directorate for Computer and Information Science and Engineering (CISE)
National Science Foundation
Room 1155, 4201 Wilson Blvd
Arlington, Va. 22230

The heart of the VLSI design methodology, which has been so successful in the past two decades, is the clean separation between the designer and the fabricator. Carver Mead describes the critical choice of an interface to the fabrication community which would permit designers to focus on design and design methodology while not following (or trying to control) the details of the fabrication process:

"The choice of the interface was made to coincide with the most generic definition of wafer fabrication: The final physical wafer would be built up of layers, applied one after the other. Each layer had a 2-dimensional definition. The design rules could be specified by simple geometric relations among the various layers. We specified the result of the operation on the wafer rather than how the result was to be obtained, thus leaving the processing detail to the particular fabricator."

The present practice in MEMS is for the designer to specify and send the mask set to the fabricator. The designer is necessarily involved in the details of the processing so that the mask set design will compensate for any and all divergences between the mask of a layer and the physical structure that is generated by that mask/process combination. A series of iterations enables the designer to bring the masks and the process in line to give the appropriate structures. Thus each mask set is good only for the specific fabricator since adjustments for process idiosyncrasies may not be appropriate for other vendors. If the designer describes the result he is seeking, the fabricators will order masks which make these adjustments and the design documents specifying the resulting object can be given to a variety of vendors for implementation, if desired. All will use different masks but deliver identical products.

It is hard to overstate the importance of this simple change (from describing the mask to describing the object) in the designer/fabricator interface in that it permits and encourages the desired clean separation. The clean separation results in and from fabricators identifying a set of standard processes for which they have detailed knowledge of the trims and adjustments needed in masking. It tends to defocus the designer from the process details to design concepts, design libraries, and design creativity where he can make his optimum contribution. Before VLSI, semiconductor designers did not observe this separation and the result was a highly fragmented design community where the best designs were the result of intense communication between the designer and the in-house model shop which built his parts. When the clean separation was in place, third party vendors were able to serve the design community effectively and design rules defined what could be done within the framework of available standard processes. These standard processes were updated and optimized by the fabricators to reflect progress in the process and equipment technologies and to keep them competitive within the fabrication community (independent of design community involvement). The basis of the clean separation is the existence and use of standard processes known to both the design and fabrication communities. This emphasizes the importance, for MEMS, of moving in the direction of standard processes where it is at all possible.

Return to NSF MEMS Design Workshop Index