The use of microelectromechanical systems (MEMS) has stirred up a great deal of interest in many markets. Just how important is MEMS technology? According to Gordon Moore, chairman emeritus of Intel Inc., Santa Clara, Calif., "MEMS is a really intriguing technology, and I believe it will have a significant impact in the next century." In fact, many engineers believe that MEMS devices will have as profound an effect on our everyday lives in the next decade as the microchip had in the previous one. Design teams at Xerox, Webster, New York; Texas Instruments, Dallas, Tex.; and many small start-ups are already researching potential applications that range from rudderless aircraft to buildings that self-adjust during an earthquake.
MEMS are tiny mechanical systems—sensors, motors, nozzles, valves, and others—that fit onto the surface of computer chips. They're created using the same semiconductor technology as ICs. For example, to create a MEMS pressure transducer, most of the surface material in a defined area of a silicon wafer is etched away. The result is a transparent diaphragm that can be as thin as a single micron. Resistors are then embedded into the surface of this diaphragm and used to translate the slightest movement of the membrane into a voltage.
Although this all sounds pretty futuristic, don't dismiss MEMS as a "blue-sky" technology. Engineering teams see the technology becoming more mainstream and highly applicable for enhancing today's consumer electronics equipment (see "MEMS Hits The Big Time,"). In fact, any device that requires motion sensitivity is a candidate for MEMS, including computer mice, camcorders, and virtual-reality headsets.
Challenges Of MEMS
The increasing attention paid to applications, coupled with the technology's cost effectiveness has brought about rapid progress in MEMS. But for all the benefits it offers, there are still formidable hurdles that need to be addressed before it can become truly mainstreamed.
One major obstacle is the lack of a smooth communications link between the mechanical and electronic worlds. Engineering teams charged with the design of integrated microelectromechanical systems are particularly conscious of this gap, and, in many cases, are still constrained by an "over-the-wall" approach to integration. Typically, each group within the team handles only the tools traditionally associated with its single discipline, throwing the project over the wall to the next group when its portion of the design is completed. With no common interface, such a separatist approach can result in catastrophe when prototype testing reveals a design flaw requiring additional iterations. One group may fix the error, but what will the repercussions be on the rest of the design, and who will fix them?
For example, none of the information derived from using 3D field solvers on MEMS structures information can be automatically transferred to an IC design tool. Microsystems engineers painstakingly identify material properties and boundary conditions to build a mesh so that the field solver can run a 3D finite-element analysis. The tool then predicts the amount of stress and strain in the structures, structural movement, or any other possible effects. But the information is useless without a means to incorporate the data into an electronic design format that other tools can exploit.
The urgency of developing a completely integrated solution for MEMS design can be demonstrated by looking at automotive airbag system designs. Clearly the airbag design team must consist of engineers across different disciplines. One group requires a finite element method (FEM) simulator to design the MEMS device, while another needs an electrical simulator to design the circuitry. Without an integrated approach, each group must spend precious days just translating data from the other group, a task that adds no value to the design, and that almost inevitably introduces errors into the process.
A secondary hurdle is enabling engineering teams to make full use of pre-existing intellectual property (IP) in MEMS design. The ability to smoothly integrate cores into a system-on-silicon architecture provides designers with both the latest functionality and a major productivity gain that can catapult the product to market months ahead of the competition. Until now, designers had to create MEMS by pushing polygons, and understanding the fine details about the target fabrication process. Obviously, this approach demanded exceptional engineering skills, and expanded design schedules and budgets.
To bring MEMS into the mainstream, it is absolutely essential that the difficulties surrounding MEMS design be minimized. One thing is certain: If electronics engineers can use MEMS devices in their system design, without excessive complexity, then market growth will be significantly intensified. And, the potential of that market growth is enormous. According to research from Ernst & Young Entrepreneurs Conseil, Paris, France, in 1996, the MEMS market was $12 billion for devices and $34 billion for systems. The same firm estimates that by the year 2002, the market will have grown to $34 billion for devices and $96 billion for systems.
Next-Generation MEMS CAD
A market with this kind of growth potential justifies the development of new approaches and tools for MEMS design. A new generation of design tools that combine aspects of electronic design automation (EDA) with the mechanical, thermal, and fluidic computer-aided design (CAD), is needed. The development of an integrated EDA/CAD solution for MEMS structures, offering a continuous top-to-bottom design flow, would bring the benefits of MEMS to the entire design industry. Specialists and non-specialists across different disciplines will then be able to leverage a common design language, as well as design reuse, as they work toward the realization of sophisticated miniature systems, at costs much lower than previously thought possible.
Dr. Dirk Beernaert, principal scientific officer of the Brussels-based European Commission responsible for microsystems in the ESPRIT research and development program, sums up the challenges facing the industry today. "What's needed is not only the combination of the different disciplines required to create a microsystem, but also a full-system-oriented approach, so that MEMS design can be a true subsystem within the overall design. The use of the right design methodology will greatly enhance reusability and efficiency while reducing cost and risk—still considered the main barriers to wider introduction of these technologies in many applications."
In developing the best computer-aided design environment for MEMS, engineering teams will require a number of capabilities. High on the list is the ability to create both monolithic (single-chip) and hybrid (multichip, multitechnology) MEMS designs. Hybrid design has been around for some time, and such environments are readily available. The monolithic approach, however, aims for nothing short of complete cofabrication of electronic and non-electronic functions. To this end, existing microelectronics lines are being extended and adapted to allow MEMS production.
This monolithic environment must allow a continuous design flow that provides benefits for both the non-specialized system-level designer, as well as the device designer. The system level designer needs an architectural-level view to create a new MEMS system from at least two different technology areas; micromechanical and electronic. The device designer, on the other hand, possessing expertise in his technology area, needs a vehicle to pass on his or her knowledge.
The environment must enable the device designer to design modules; simulate them; and pass on the knowledge in the form of characterized, standard-cell libraries. The system-level designer can then take advantage of the multilevel information contained in those libraries, such as layout information, behavioral models, and FEM models. By assembling the desired cells, the designer can create and simulate a system-level design. The resulting system is then handed over to the back-end team for chip-level design work. Once final layout is produced, both the system- and device-level engineers can review the features of the MEMS design.
The proliferation of MEMS products and potential applications has inspired the EDA industry to smooth the bumps posed by the integration of the electronic and mechanical worlds. Coupling mechanical and other solutions, such as electrostatic and thermal simulation, has been a particular problem. Some companies have achieved a measure of success in addressing this problem by using point tools on an ad hoc basis, such as those from Ansys, Canonsburg, Pa.
In an effort to bring MEMS out of the research arena and into mainstream design, the U.S. Pentagon's Defense Advanced Research Projects Agency (DARPA) has been actively promoting the development of a composite CAD program for MEMS. To help jump-start the EDA development in this area, this program funds on-chip MEMS design.
This focus by DARPA has already helped stimulate progress in the EDA sector. Several companies now offer MEMS design-tool support. For example, Mentor Graphics, Wilsonville, Ore., in partnership with MEMSCAP, Grenoble, France—a spin-off of Grenoble-based Circuits Multi-Projects (CMP), and an affiliate of the French TIMA research institute—has recently announced a set of MEMS Engineering Kits (Fig. 1). In addition, Mentor Graphics and MEMSCAP, together with MCNC, Research Triangle Park, N.C., were recently awarded a grant from DARPA for the development of an MCNC foundry-specific engineering kit.
There are several key features a designer should look for when choosing a design kit. First, it must include the basics. Parameterization from behavioral models all the way down to layout generation is an absolute must. The kit should support bulk micromachined structures such as infrared sensors, piezoresistive mechanical devices, and accelerometers, as well as surface-micromachined elements such as capacitive sensors and electrostatic actuators. Both monolithic MEMS design and hybrid design on-chip with analog and digital circuitry should be supported. Furthermore, schematic entry must be complemented with access to analog behavioral modeling and layout generation for structures ranging from bridges and cantilevers, to infrared sensors and accelerometers, etching simulation, cross-sectional viewing, and extended design rules for MEMS layout (Fig. 2). Interfaces to finite-element solvers are necessary to evaluate the 3D models created in the design process.
Another essential requirement is a continuous design flow from front- to back-end. By seamlessly combining the electromechanical and IC design environments, the kit will eliminate the confusion and inefficiency of going back and forth between the two. As a result, a MEMS design can flow seamlessly from structural model to layout, with multidisciplinary teams including mechanical, electrical, component, and system teams, working together in the same environment and directly sharing results. And, if there are interfaces to specific foundries, a fast track from design to market is almost a sure bet. The Mentor-MEMSCAP kits, for example, will support several foundries including MCNC; Bosch, Reutlingen, Germany; Austria Mikro Systems, Graz, Austria; and SensoNor, Horton, Norway.
"One of the biggest limitations to continued growth of the MEMS user base is the fact that with existing design systems you have to be a rectangle-pusher." At least that's how Karen W. Markus, director of the MEMS technology Applications Center at MCNC sees it. She continues, "The limited tools that are available to help a MEMS designer require that person to have at least a basic idea of CAD layout and the process being used."
Markus contends that a vast, untapped market exists in systems designers, people who think in terms of black boxes that capture all the behavioral and intrinsic information. "The engineering design kits currently under development by MCNC and its partners will provide these black boxes and the supporting design framework to the systems design community, opening the door for their participation in MEMS foundry activities. This has the potential to spark an additional level of interest and product introduction beyond the current frenzy," says Markus.
To move MEMS up the design chain, support for design IP is necessary. This is especially important for designers new to MEMS. Intellectual property provides more general-purpose ready-made pieces to help jump-start the design. As a result, instead of laboriously creating a device by pushing polygons, the designer can use MEMS IP to significantly streamline the design effort. This makes it easier for those new to this type of design to efficiently produce a viable MEMS application. The designer can still zero in on crucial areas at the device level, and work with field solvers. However, once the designer is satisfied, lower-level information can be automatically translated into an analog HDL, and integrated with other IP at the system level.
Dr. Beernaert of the European Commission emphasizes the pressing need for reusability, "Currently, every new MEMS application is a new process, a new device design that starts from scratch. This is very inefficient, very expensive, and is asking for mistakes to be made. Having interfaces in the form of IP to specific foundries, and qualified, debugged MEMS building blocks will be critical for faster development times, and for increasing the chances of immediate success."
MEMS: On The Verge
With the ability to support any combination of electrical and micromechanical devices on a single chip, MEMS opens an endless horizon of tantalizing possibilities for future designs. But to fully mine the staggering potential of MEMS tomorrow, design tools and methodologies must be devised today that support that vision. With the combined effort of DARPA and the EDA industry focusing on MEMS, significant effort has already made in that direction. Such progress ensures that the tremendous promise that MEMS holds for next-generation IC-based applications will, in fact, become a reality.