The opportunities for microelectromechanical systems (MEMS) are wide open and the technology is established. While there are no one-size-fits-all foundry solutions, customers seeking volume capability can take their pick of several competitive facilities. The key is finding a partner with the knowledge, experience and up-to-date infrastructure that can help get the product to market at the cost point most appealing to major OEMs, while ensuring manufacturability and quality.
Over the last 10 years the MEMS industry has spawned a wealth of applications. Several of these applications, including the much-cited examples of automotive engine controls, airbag accelerometers and inkjet printer heads, have moved into high-volume commercial production, thereby validating the basic MEMS technology, materials and processes used. Indeed, the substantiation and commercialization of these devices prompted a flurry of MEMS-related telecom acquisitions around the year 2000, generating a host of MEMS startups seeking to address a range of applications. Many of the startups included MEMS foundries anticipating volume manufacturing business from the emerging telecom sector.
By the end of 2003, however, most companies involved in optical MEMS had folded and six of the dozen independent foundries had closed their doors. So what went wrong? While in some cases the designs (especially in the telecom sector) were far too complex to be reliable given the manufacturing capability of the time, the failed foundries had a few traits in common.
As MEMS engineers are aware, MEMS devices depend on a wide range of processes and an enormous number of variables. The high-volume devices already on the market had been commercialized only after years of process refinement and test. Research and development teams typically need extensive amounts of time to reach conclusions about the physics of the processes and materials. Devices manufactured using a single material such as polysilicon may require marked variations in process depending on the source of the polysilicon and the deposition method.
The failed commercial foundries tried to address multiple product opportunities across multiple markets and multiple processes, with the result that their R&D time and even prototype-manufacturing variables increased exponentially. For smaller foundries, this was the kiss of death.
At the same time, these independent foundries, many of which had developed from academic R&D backgrounds, had little experience with high volumes, and the typical infrastructure and optimizations required for mass manufacture. The prevailing attitudes were based on the hope that if you found a process "they will come." Little attention was paid to how components would fit into existing designs and what the customer was willing to pay for it. In the case of RF MEMS, some large-volume consumer equipment manufacturers were seeking MEMS-based components at a price range of 10 cents in volumes of 10 million or more. This figure even included packaging and test. Clearly, the requirements were beyond the range of smaller fabs, dooming the broad-based business model.
In assessing the independent foundries that survived, we can begin to see the formation of a foundry business model that may predominate at least until the industry can lay a stronger foundation with standard material definitions and some form of semistandard processes. In particular, while the remaining designers and manufacturers have learned their lesson and have made strides in design-for-manufacturability and reliability, these qualities alone are no longer enough to persuade volume OEMs to make what is perceived as a frightening jump into replacement technology, even for simple discrete components.
These integrators do of course perceive the benefit of the technology and are intent on offering their customers significant value. The question is not "if," since MEMS is an increasingly proven technology, but "when." But since 2000 these mainstream equipment suppliers many of whom are virtually giving away technology such as standard-cell phones have become accustomed to squeezing their component suppliers to a bare minimum to derive better margins.
Given such stringent requirements, independent foundries have a critical role to play in making innovative MEMS products technically and financially attractive to the mainstream market. This role, while not easy, is a natural fit for foundries, since batch-produced MEMS can facilitate size reduction and cost savings at the same time. However, achieving size and unit cost goals in a timely and cost-effective manner is a solution that not all foundries can offer.
First of all, to achieve this goal fabs must have a few robust, established processes that they know inside and out through experience and solid statistical process control. While the drawback is that this may mean some specialization in particular kinds of devices, the major benefit is that these processes will provide a baseline recipe from which to start.
Having such proven processes means that, first, foundries know their cost base - in materials, tools and personnel and that they can offer a certain amount of consistency that shortens time-to-market. These teams know the properties of materials and base techniques. They are familiar with deposition methods, grain size, how the materials will respond to the introduction of other materials, what factors may cause structural stress, fatigue or other failure modes within the typical process parameters.
Next, it's important for established independent foundries to have expert staff and equipment operators on hand whose combined knowledge base can add significant value in reducing device cost, size and power consumption. In some cases the fab will have its own design-for-manufacture team, and in all cases the fab should offer some form of continuous improvement through total quality management.
Using such a staff, a manufacturer can, for example, shrink a standard MEMS-based pressure sensor die from 2 x 2 to 0.65 x 0.65 mm, to fit more than 25,000 die in a single six-inch wafer for major cost reductions to the customer. It may also be able to integrate a pressure sensor and DSP on the same small die for a high-performance device aimed at the automotive and medical markets. It is important to seek out experienced teams. The millions of units of experience such teams have can offer major insights into improvement methods, from how to customize equipment for better etch results to cheaper alternatives in wafer-level bonding for hermeticity.
Further, a good independent foundry will have invested in the needed infrastructure for cost reduction. Obvious examples include six-inch wafer cassette handling, which is now the minimum standard for achieving higher volumes in MEMS. The latest equipment can also offer a significant advantage once the team has gained familiarity with it. Stepper technology, track handling and advanced etching control provide improved MEMS device tolerances at lower cost while also enabling more complex geometries. The newest deep-reactive ion-etching tools offer unparalleled vertical-depth etching and smooth sidewalls for consistent enhanced performance of devices.
Finally, a superior foundry, over years of production of related devices, will typically have developed its own value-added intellectual property that may offer the opportunity for significant enhancement of a customer's device. Such IP might include integration of additional passive components for greater space savings to combining MEMS devices with improved microelectronics for faster and more accurate performance. The design services team can assist the fabless developer in assessing affordable and reliable options for volume production.
Jim Knutti (firstname.lastname@example.org) is president and chief executive officer of Silicon Microstructures Inc. (Milpitas, Calif.).