Cover Story: New Foundry Friends
MEMS device makers increasingly outsource fabrication
Keeping up with the growing demand for microelectromechanical systems used inside devices such as smartphones will require MEMS makers to make some new fab friends—as in foundries that can fabricate the devices, help drive down costs and speed time to market.
MICROmanufacturing won a 2012 APEX Award for Publication Excellence for this report about the MEMS market.
That’s the trend among a growing number of MEMS device manufacturers. Given the staggering costs associated with building their own fabs, these companies are going fabless—choosing to outsource fabrication of their devices despite having to share intellectual property with a third party.
Though protecting IP remains a stumbling block, many in the industry are convinced that the path to high-volume, low-cost MEMS manufacturing will include more “pure-play” MEMS foundries. And that, in turn, could trigger the adoption of MEMS standards, or at least de facto standards, that some say are required to keep up with the consumer electronic product cycle.
With the global market expected to reach $12 billion for packaged MEMS devices in 2015, the number of fabless MEMS companies will continue to grow over the next 5 years, according to IHS iSuppli, a market research firm based in El Segundo, Calif.
“Fabless semiconductor ?rms in 2010 widened their share of the worldwide MEMS market compared to 4 years ago, holding nearly a fourth of the total MEMS value generated last year,” iSuppli reported in June. Companies outsourcing MEMS fabrication accounted for 23.2 percent of total MEMS revenue in 2010, up from 21.3 percent 4 years ago, according to iSuppli.
A new MEMS piezoresistive, low-pressure sensing die design from All Sensors measures 2mm × 2mm and minimizes position sensitivity because it uses a boss-less structure. Photo courtesy All Sensors.
The statistics include outsourcing by integrated device manufacturers (IDMs)—the firms that traditionally designed and made products in-house. One such IDM, according to iSuppli, is Analog Devices Inc., Norwood, Mass., which up until a couple of years ago produced all its MEMS devices in-house. ADI now outsources the production of its MEMS microphones to the Taiwan Semiconductor Manufacturing Co. (TSMC), an integrated circuit foundry branching out to MEMS.
“The overall increase during the 4-year period, though small, shows that MEMS manufacturing continues to move away from being the sole domain of IDMs,” the market research firm concluded. IHS iSuppli expects the fabless MEMS business share of the market to continue growing over the next 5 years for two main reasons: an influx of innovative MEMS products from fabless startups in that timespan, and mounting pressure from IDM shareholders to keep margins high, which iSuppli suggests will lead to outsourcing part or all of their MEMS production.
“First, a lot of the new companies that come into MEMS with innovative devices for mass markets are fabless startups,” said Jérémie Bouchaud, director and principal analyst for MEMS and sensors at IHS iSuppli. “And there is a much stronger infrastructure for fabless MEMS companies than 5 to 7 years ago, when it was mostly specialty or boutique foundries that could produce your parts.”
Plus, added Bouchaud, much bigger foundry players—including some mainstream complementary metal-oxide semiconductor (CMOS) foundries such as TSMC—have entered the MEMS fabrication market, likely due to the robust forecasts for MEMS.
In fact, Yole Developpement, a market research firm based in Lyon, France, in late September forecast a 15 percent cumulative annual growth rate (CAGR) in dollar value for the MEMS market from 2010 to 2016. Yole expects the market to reach $19.6 billion in 2016.
Compared to a 9 percent CAGR forecast for the IC market, the bigger foundries like TSMC now view the MEMS fabrication segment as more than just an opportunity to fill up excess capacity on their IC fabs.
As a result, Bouchaud said, “we see some IDM companies—especially the public companies that have pressure for profitability—beginning to look at possibly working with these big foundries so they can cut their costs.”
Other factors driving the trend include the $20 million to $50 million experts estimate it would cost to build a MEMS fab, the 2 to 3 years it could take to complete the fab and, last but not least, the fact that venture capital firms behind new MEMS startups like business plans that include some sort of partnership with an outside foundry.
Meeting a nice foundry
Picking the right foundry partner, however, has proven to be a significant challenge. “It’s not like choosing a job shop,” said Alissa M. Fitzgerald, Ph.D., founder of AMFitzgerald & Associates, a MEMS product development and engineering services firm based in Burlingame, Calif. “In mechanical engineering, if you create drawings of whatever you are trying to build, any machine shop knows how to make it.
“With MEMS,” Fitzgerald continued, “the drawing is only part of the design. The other part is the process flow. In MEMS, the device design has implications for the manufacturing process. And the process flow has implications for the device design—the two are inseparable.”
For instance, if a MEMS device maker needs to etch 100µm of silicon, every foundry will do that slightly differently because there are no process standards.
“That’s the reason you can’t throw a MEMS design over the fence as you would with a machine shop,” said Fitzgerald. “The guys on the other side will interpret the design to suit whatever toolset and recipes they have.”
MEMS fabrication process decisions have a direct impact on device design and functionality. When a fabless MEMS device maker takes a prototype to a foundry, the foundry takes all the information and tries to interpret it and set up a process within its existing MEMS fab toolset, Fitzgerald said. Doing so will have unintended consequences on MEMS performance. Then the customer has to test the device, examine it and recommend changes or improvements.
“There has to be a feedback loop between the customer, who can describe the device’s intent, and the foundry,” Fitzgerald said, noting that the tighter that relationship is the better the chances are the product will get to market faster.
From Jim Knutti’s perspective as president of Acuity Inc., a fabless supplier of MEMS-based pressure sensors and other MEMS devices, “there is no other way to go than fabless at this point.
“Back in the 1980s,” Knutti observed, “we had to design our own MEMS fab processes and equipment and modify existing semiconductor equipment. At this point, that’s all available, and all the foundries have access to the same tools. That changes the equation. The competitive advantage doesn’t come from how well you enable the fab process.”
That’s why he doesn’t worry about sharing IP with a foundry. “What differentiates a company like us is how we design the MEMS device, including the topography and modeling.” In short, he said a startup trying to build its own fab today would be like “going out and building your own computer.
“The reality is, most foundries have no use for our design IP,” noted Knutti. “If they got a reputation for releasing or compromising someone’s IP, then I do not think that they would last very long.”
Sharing that sentiment is Dennis Dauenhauer, president of All Sensors, a fabless company specializing in pressure-sensor product development, with an emphasis on low-pressure medical and industrial sensors. Sharing IP is just part of the business, and All Sensors has been fabless since its inception in Morgan Hill, Calif., in 1999.
In fact, its foundry didn’t have the necessary MEMS process initially, but All Sensors worked with them to develop the required process. Before doing so, though, the critical decision the company faced was whether the foundry could meet its needs.
Dauenhauer said that took trust, which is high on the priority list when considering a foundry, particularly from a technical standpoint. A failed foundry partnership can be disastrous because a fabless MEMS device maker can’t take the MEMS manufacturing process when it leaves.
Though lower on the list, geographic proximity can be important. “We use a foundry in Europe, and it’s not easy,” Dauenhauer acknowledged. “We’ve been getting product from them for 2 or 3 years, but still have process-control issues so we have to go there once every other month to make sure they’re doing what they’re supposed to.”
For each fabless MEMS success story, however, there are many more failed attempts. AMFitzgerald has worked with more than 90 companies, Fitzgerald said, “and we’ve seen the pain every one of them goes through. We help customers find foundries, and we know how long that process can take. There are so many good things that can happen in MEMS, but it’s so hard and so expensive. When we see so many failures, it causes us to ask the question, ‘why is it so hard and what can be done to make it easier?’ ”
A ‘wiki’ guide
To help educate the market, the MEMS Industry Group (MIG) has coordinated an online “wiki” to develop a guide for companies seeking to work with a foundry, said Karen Lightman, MIG’s managing director. Introduced in January, “The MEMS Foundry Engagement Guide” is evolving as those in the industry contribute to the guide and help edit it (see sidebar below).
Developed in response to demand from member companies, the guide “is really about being a neutral place to find out what’s going on,” Lightman said. “The reality is it helps everybody if a potential customer comes to a foundry with a better understanding of the expectations.” Plus, the guide helps MEMS device makers learn what questions to ask to determine each foundry’s expertise.
Silex Microsystems Inc., a pure-play Swedish MEMS foundry, is glad to see the guide developing. “Despite … everyone’s best efforts at executing on projects, the fact is that each MEMS project brings with it unique manufacturing challenges,” said Peter Himes, vice president of marketing and strategic alliances at the foundry’s San Francisco office. “Anything that can educate the first-time MEMS developer on what to expect and how to plan their project will be of benefit.”
It’s up to the industry to get the word out, said Theodore Chi, director of marketing and sales at Innovative Micro Technology, a pure-play MEMS foundry in Santa Barbara, Calif. The foundry guide will help people outside the MEMS world understand “that this is a viable technology for providing solutions to problems,” said Chi.
“There are people in the bio industry who don’t know anything about MEMS,” he continued, “and the wiki guide is a good starting point to introduce them to what MEMS is, how a foundry operates and what wafer processing is all about.”
Then there are the customers with decades of CMOS experience—running tens of thousands of wafers a day—who enter the MEMS arena and expect the same kind of pricing to run 10 wafers, Chi noted. The guide “certainly helps communicate that difference.”
IMT offers ultra-high-performance metal via technology (visible in the cross-section of a TSV wafer above) with a 99.996% yield and low resistance. Photo courtesy IMT.
Chi also urges companies to do more than compare the equipment each foundry has because many fabs have similar setups. “There may be some differences, such as whether a foundry [fabricates on a 6" or 8" wafer], but for the most part a comprehensive, turnkey foundry will have all the capabilities,” he said, noting that the critical differences are in staffing, expertise and experience.
In short, he said, find out whether a foundry has built up an arsenal of MEMS process solutions. IMT, for instance, offers its customers ultra-high-performance metal via technology that has been used to produce products with nearly 140,000 hermetic, metal-filled through-silicon-vias (TSVs) per wafer. That’s a critical technology for any MEMS device maker interested in reaching large-scale 3-D integration, according to IMT.
And at Silex, the Swedish foundry developed a TSV technology in the mid-2000s for a customer needing TSV capability for a cellphone application that went into mass production in 2006. Since then, Silex has used the TSV solution on more than 50 other MEMS projects.
In the house
To be sure, not everyone is sold on the fabless model. Dr. Frank Schäfer, who is responsible for product planning and marketing of automotive MEMS sensors at Robert Bosch GmbH, Reutlingen, Germany, said proximity is a key advantage for Bosch, which runs an in-house MEMS fabrication process. “We have the advantage that everything is located right here at one site,” he said, noting that the company’s fab, development and quality-control teams are all located within about 200 meters.
Schäfer said such close proximity makes it easier to develop a new process and to work through problems with the development, manufacturing and quality-control teams. While he understands why startup MEMS companies may want to go fabless, he said when the company reaches a critical volume level it’s better to have your own fab. Only then, he said, will the company have access to all the manufacturing details.
SIL-VIA (above), based on through-silicon-via technology, is capable of producing high-density through-wafer vias with via pitch down to 50µm. Below, the image shows the deep reactive-ion etching (DRIE) process. Images courtesy Silex Microsystems.
On the other hand, said Himes of Silex, future growth will inevitably force MEMS device makers with their own fabs to either outsource for peak demand response or expand their internal capacity. The option chosen will depend on the company’s economics and timeframe.
“Many times,” said Himes, “even if they want to expand internally, the time for new fab expansion is too long to capture the market opportunity.”
Fitzgerald stressed that point, suggesting that the MEMS industry as a whole is missing out on larger markets because it cannot move fast enough, and it will miss out on other markets in the future for the same reason.
“Consumer electronics works on a 9-month and 12-month product cycle,” she said. “Most MEMS companies can’t keep up with that. That doesn’t necessarily mean a new accelerometer is needed every year, but everybody understands time to market. Opportunities are not open forever. If the MEMS industry can’t figure out a way to do things faster, it will be leaving money on the table.”
And that’s not good for any long-term relationship. µ
About the author: Dennis Spaeth is the electronic media editor for MICROmanufacturing and Cutting Tool Engineering magazines and their respective Web sites. Telephone: (847) 714-0176. E-mail: [email]email@example.com[/email].
A guide to foundry engagement
With MEMS device makers moving from an in-house manufacturing model toward a fabless approach, the nuances of MEMS fabrication must be better understood. To that end, MEMS Industry Group began working with our members who represent the MEMS supply chain—device manufacturers, equipment/materials suppliers, foundry partners, software designers, consultants and OEMs—to create a “how-to” guide on working with a MEMS foundry. MIG published this collaborative guide, “The MEMS Foundry Engagement Guide,” earlier this year as an online wiki. It is an evolving document that anyone, including non-MIG members, can access, edit and use as a reference.
The guide helps facilitate the process of selecting a partner and offers practical advice on MEMS manufacturing.
Starting with background information about MEMS foundries, the guide explains how they differ from traditional IC foundries. MEMS foundries tend to specialize in a certain application. MEMS, by definition, covers an array of microdevices across many applications. Because the manufacturing processes—though built on the same basic materials and toolsets as ICs—vary so widely, foundries must specialize to achieve the expertise and efficiency required for low-cost, volume production. Understanding this can help to narrow the field of potential foundry candidates.
MEMS Industry Group's Foundry Engagement Guide. The guide explains why MEMS is so distinct, and it covers the use of unconventional materials, lack of standard processes and MEMS’ often-unique feature sizes, mechanical properties, unit processes and stiction. The guide continues with a self-evaluation checklist to help device manufacturers understand their level of readiness for working with a foundry partner. Device manufacturers must understand and be willing to share many factors, including:
- application and market;
- package type and integration requirements;
- cost targets;
- volume projections; and
- risk evaluation.
Other topics in the guide include Request-for-Quotation requirements and how to match requirements to capabilities.
The guide also includes insight into how foundries select their customers. For example, savvy foundries can often recognize the types of projects and customers that will be successful or not. The guide is available at www.memsindustrygroup.org/foundryguide.
[RIGHT] —Karen Lightman, MIG Managing Director [/RIGHT]
Acuity Inc. (510) 943-6205 www.acuitymicro.com
All Sensors (408) 225-4314 www.allsensors.com
AMFitzgerald & Associates (650) 347-MEMS (6367) www.amfitzgerald.com
IHS iSuppli (310) 524-4007 www.isuppli.com
Innovative Micro Technology (805) 681-2800 www.imtmems.com
MEMS Industry Group (412) 390-1644 www.memsindustrygroup.org
Micralyne (780) 431-4400 www.micralyne.com
Robert Bosch GmbH +49 7121-3535-900 www.bosch-sensortec.com
Silex Microsystems (415) 283-3399 www.silexmicrosystems.com
Yole Developpement +33 4-72-83-01-80 www.yole.fr
Dos and don’ts
Nancy Fares, president and CEO of Micralyne, a pure-play MEMS foundry in Edmonton, Alberta, offered the following dos and don’ts for selecting a foundry partner:
- Base your decision on process, technical capabilities, experience, packaging and test capabilities.
- Find the best combination for cost that matches your company’s quality-control philosophy, view of IP protection and volume goals.
- Find out if the foundry competes or could compete with you.
- Understand and document the process for settling disputes.
- Know your core competency and the IP most critical to protect—and share that with the foundry.
- Don’t micromanage your partner.
Fares, however, doesn’t believe the fabless model is for everyone. MEMS device makers specializing in one device with their own internal fab may have achieved economies of scale that negate the benefits of outsourcing, she noted. [RIGHT] —D. Spaeth[/RIGHT]