MEMS commercialization: What’s taking so long?

Micro-electromechanical Systems (MEMS) began their development vis–vis the discovery of silicon's piezoresistive effect by Charles Smith of Bell Labs in 1954. Less than a decade earlier, in the same Bell Labs in New Jersey, John Bardeen and team discovered the transistor—the basis of today's semiconductor industry.

Though MEMS and semiconductor ICs share processing similarities, their business issues vary because in the 50-plus years since their discoveries, the IC market has grown to more than $220 billion, whereas the MEMS market reportedly still totals less than $10 billion. Why the discrepancy?

Most products take time and resources to transition from discovery to full commercialization. It wasn't until 1990 that full commercialization of the first MEMS pressure sensor was achieved, based on the piezoresistive effect. The adoption of accelerometers in air bags in the early 1990s, like the adoption of pressure sensors in automobiles for electronic engine control in the 1970s, was a major success.

Numerous MEMS devices have been commercialized in the past few years—as MEMS have dropped in cost, we've seen the integration of MEMS microphones, accelerometers, displays and RF devices into portable electronics (handsets, for example) and games (Wii), for instance.

Each year we evaluate the 14 most critical success factors for the MEMS industry, and create a “MEMS Commercialization Report Card” grading each factor from “A” to “D.”

Below are key findings from the 2007 report, based on interviews with more than 50 MEMS suppliers, users and instrastructure providers worldwide. (Results of the 2008 study will be published in early 2009.)

The overall 2007 report card grade was “B-,” but it's the individual-factor grades that tell the critical story. MEMS have clearly turned the corner with their integration into a large variety of industrial and consumer products and ultimately will be integrated into many products that touch our lives every day.

MEMS infrastructure development has improved dramatically from its C+grade in 1998 to its A-.

Infrastructure development has taken a path similar to that of the semiconductor industry, where the majority of companies (and virtually all MEMS startups over the past five years) have adopted the fabless or fab-lite model.

This is a result of the availability of more that 60 foundries worldwide that provide a broad spectrum of services, from wafer fab (Asia Pacific Microsystems and Colibrys, for example), specialized equipment (EVG, Suss Microtec), packaging (Engent, Infotonics) and software development tools (Microcosm, SoftMEMS).

There's no paucity of resources to support time and cost-effective MEMS development.

A major barrier to MEMS commercialization has been the lack of design for manufacturing and test strategies adopted by the technology suppliers. Traditionally, MEMS development has focused on the MEMS device—the accelerometer, pressure sensor and so on—rather than on the overall solution, which typically incorporates an Application Specific Integrated Circuit (ASIC).

This is partly because MEMS designers have considered MEMS devices only in context of their Ph.D. theses, not in context of real-world packaging or integration. The ASIC costs two to three times the cost of the MEMS, ande has at least two to three times the mask layers and four to five times the surface area of silicon.

Often, the MEMS function could be implemented in a surface layer of a monolithic device, but its function is considered standalone by its designer. Most recently, wafer level and other chip-stacking strategies have been used to minimize solution footprint. Many new MEMS designs also incorporate networking functions—both wired and wireless—which can be realized in a separate chip or integrated into the ASIC.

A major challenge is that MEMS must work in hostile environments. Unlike their semiconductor counterparts (PC boards, for instance) that typically perform in benign environments, MEMS work in nasty surroundings—under automobile hoods, for instance, where they're subject to temperature, shock, vibration, and EMI/RFI and other variables.

MEMS' ability to survive in such environments depends on the package, which must be robust yet inexpensive. Also, the package can not mechanically affect the MEMS to the point at which it imparts electrical shifts of offsets to the device.

The overall cost of the packaging and test of a MEMS-based solution typically accounts for approximately 60 to 70 percent of the product's total cost. As such, MEMS solutions must be co-engineered from day one of the design cycle for system design optimization.

Determining packaging and test strategies after chip design and selection of partitioning strategies based on which functionalities will reside in the MEMS chip or the ASIC can create the need for costlier, less reliable solutions.

The poster child for MEMS solutions is the tire pressure monitoring system (TPMS). Many companies produce these sensor application modules, which contain numerous components, including MEMS pressure sensor, temperature sensor, motion sensor/accelerometer and an ASIC containing A/D conversion, battery management, signal conditioning/compensation and perhaps even the Tx/Rx chip to send signals to and from the passenger compartment display.

And don't forget the software! Supplier strategies on integration and partitioning vary, but they all create MEMS-based solutions.
This type of “system think” will be mandatory for the continued success of MEMS commercialization.

As we enter Q4 2008, some trends emerge:

R&D will be downgraded because of budget tightening in federal and commercial organizations.

Investment in MEMS startups will be curtailed due to the weak economy.

MEMS-related marketing budgets will be downgraded.

On the brighter side, DFM&T will continue to make significant progress, thanks largely to lessons learned from large-scale production of MEMS for consumer products.

Roger H. Grace is president of Roger Grace Associates (

0 comments on “MEMS commercialization: What’s taking so long?

Leave a Reply

This site uses Akismet to reduce spam. Learn how your comment data is processed.