March 23, 2010 – Tegal president/CEO Thomas Mika explains to SST the company’s reasons for offloading much of its legacy tools business, and where the company is putting its emphasis. Gartner’s Dean Freeman also examines the deal.

Having decided its future lies in DRIE and smaller high-growth markets such as MEMS and 3D packaging, Tegal has sold off its legacy etch and PVD products to OEM Group.

A number of factors, going back several years, led up to this maneuver, according to Thomas Mika, Tegal president/CEO, who answered a Q&A with SST in an e-mail exchange:

A two-fer in 2008.  Mika noted that from 2005-2007 Tegal’s biggest customer, STMicroelectronics, purchased many of the 6500 etch tools for PZT and platinum processes, in its "IPAD" integrated active/passive device. But with concurrent events of late 2008 — the acquisition of the AMMS DRIE technologies, and the start of the industry downturn — legacy thin-film etch and PVD sales "fell off dramatically," Mika said. "Etch basically disappeared," he said. And PVD focused on leading-edge deposition of AlN and other piezoelectric materials, for device development in FBAR and related devices, "was unique to Tegal, but still primarily in development in terms of devices," so the company started a foundry service to help continue PVD sales.

"Had it not been for our acquisition of the AMMS product lines, we would probably be out of business at this point," Mika summed up.

Daring to DRIE.  So since late 2008, Tegal’s systems sales have been primarily in DRIE: "we are now almost 100% dedicated" to it, Mika claims, targeting etch opportunities in MEMS, power devices, and 3D packaging/through-silicon via (TSV) applications. And unlike the old days, no single customer takes up a huge chunk of the company’s business, given the fragmented nature of these new markets the company now targets, Mika says.

Most equipment sales in the past year have been for MEMS, and "perhaps two PVD machines and possibly one etch machine that falls into the legacy category," notes Dean Freeman, research VP at Gartner. With remaining products in deep trench etch, "with what they have left they are mostly a MEMS company" — which, competitively, means going up against SPP/STS, and in certain other small (emerging) markets, like 3D packaging, seeing Lam and AMAT. "It’s possible that we will see the rest of the company now sold into a MEMS specialty company, or that Tegal will continue to niche along with what they have," he suggests.

Getting off the semi recovery/downturn cycle.  "Let’s face it," Mika says: "The semicap industry is recovering…until the next downturn." The industry simply can’t afford to support all equipment suppliers, something with which Mika notes he is in agreement with AMAT top exec Mike Splinter. "It remains to be seen whether the MEMS industry and its participants (both on the buy and the sell side) can mature to the point where the semicap industry did, by truly valuing the technology embodied in the equipment and process know-how of its suppliers," he says. "The jury is out on this." He recalled the gallows-humor joke about the difference between a recession and a depression — whether it’s your neighbor who loses his/her job, or if it’s you.

March 19, 2010 – MEMS companies sold $7B worth of devices in 2009, down about -5% from the previous year, and there were some changes among top-ranked suppliers thanks to comparative growth (or lack of) in some key markets, according to statistics from Yole Développement.

The top 30 companies (out of 250 MEMS developers and manufacturers tracked by Yole) totaled about $5.1B in sales, nearly three-quarters of the total MEMS market. In a sign of the diversifying times, perennial MEMS market leaders Texas Instruments and Hewlett Packard are still 1-2, but they no longer dominate the market like before — they "only" account for 25% of the top 30 total revenues, after being 29% as recently as two years ago.

In 2009, TI wrestled the top spot from HP as sales of its digital light processing chips held up better than did those of HP’s inkjet printheads, notes Yole MEMS analyst Jérôme Mouly, in a statement. (He also notes that perennial ) Top sensor maker Robert Bosch, meanwhile, slipped past STMicroelectonics to claim the No.3 spot, overcoming a sluggish auto market with "impressive results" (125% growth) from its Sensortec unit.

Top 30 worldwide MEMS companies, 2009 revenues in US $M.

Among notable trends in MEMS in 2009:

Inertial sensors/gyroscopes. There’s a battle brewing in this sector, with incumbents STMicroelectronics and Bosch now sharing the spotlight with one upstart, notes Yole CEO Jean-Christophe Eloy. Invensense quickly got its 2- and 3-axis products to market after 2-3 years development (vs. the usual 5-10 cycle for MEMS devices) thanks in part to a five-fold production ramp with foundry partner TSMC — its simplification of capping the MEMS device with the ASIC enables easier packaging as part of the frontend process. The company offers 3-axis single-chip gyro for <$3 as well as various inertial sensor module combinations, and is prepping a 3-axis magnetometer (using its own accelerometer). Look for increased competition in consumer inertial sensors as ST and Bosch counter with their own new products and, respectively, "incredible commercial power" and "incredible technology," notes Eloy.

Automotive shuffles. The industry downturn was particularly harsh on automotive MEMS suppliers (who were already reeling after a harsh 2008). Panasonic benefited by the exit of Systron Donner Automotive — key Systron customer Continental was already shopping for an alternative supplier to competitors’ Bosch products — to rocket up to No.9 with >$200M in sales with its gyroscopes and stabilizers. Denso and Infineon slumped 30%, meanwhile.

Generally speaking, lots of diversity. Other movers and shakers included Dalsa Semiconductor (19%) and Micralyne (7%), whose performance "indicated the growing importance of the MEMS foundry model," according to Yole. (After two years of strong sales, Dalsa is now the top independent MEMS foundry.) Avago (14%) and Kionix (10%, after 70% growth in 2008) flexed muscles in FBAR filters and inertial sensors for consumer markets, respectively. Boehringer (3%) benefitted from growth in drug delivery systems, particularly nebulizers for asthma sufferers. All other leading MEMS companies saw flat or declining sales, some by 30% or more.

New to the Top 30 ranks are Sensata (TI’s former sensor business, which raked in nearly $570M in its March 11 IPO), which morphed into a fabless company to do MEMS designs and integrations. Another spinout, Sensonor (from Infineon), is selling its own inertial sensor modules. And Dalsa also cracked the top 30 in 2009 as customer development projects finally arrive after years in pipelines to enter production at specialty MEMS foundries, Yole notes. Falling out of the top 30 list in 2009 were FormFactor (which was worst on the 2008 rankings with a -51% decline) and Colibrys (which sold off a MEMS facility in late 2009).

Top 30 worldwide MEMS companies, 2009 vs. 2008 growth.

March 17, 2010 – Seiko Instruments Inc. (SII) and the Institute of Microelectronics (IME) of Singapore’s Agency for Science, Technology and Research (A*STAR) say they have developed a high-precision silicon-based MEMS resonator [PDF link] that they say solves a key problem with such devices, to help ensure precise output frequency and enable low-power consumption and miniaturization. They’ve also devices a new wafer-level packaging technology that improves the mechanical strength and lowers fabrication cost.

Benefits of silicon-based MEMS resonators vs. quartz crystals is threefold: integration with CMOS, ideally lower costs, and standardized IC packaging — which will enable not just the use of low-profile packaging, but also the use of standardized IC pick-and-place assembly equipment.

Due to several factors (e.g. thermal expansion and Young’s modulus), silicon-based resonators pose a challenge: resonant frequency fluctuates with temperature changes, generally with coefficient of -40ppm/°C. In other words, for an operating temperature limit of -40°C to 80°C, the resonant frequency will fluctuate ~5000ppm, which makes it very difficult to ensure precise output frequency in the resonator, the companies point out. An electrical circuit can compensate for these changes, but that creates problems in designing low-power consumption and miniaturization, and generates noise in the output.


Figure 1. Microscopic image of the developed silicon resonator. (Source: SII, IME)


Their answer: add a layer of silicon dioxide film to the MEMS resonator, fabricated on a silicon-on-insulator (SOI) substrate, and with a structure that produces residual stress within the resonator (Figure 1).

The new device improves temperature coefficient of resonant frequency fluctuation from -40ppm/°C to +-2ppm/°C, and controls frequency accuracy to a range within max. 500pm at operating temperatures (Figures 2 & 3). This reduces the need for temperature compensation to the minimal, and allows for application in low power consuming, miniaturized, and high-precision resonator.


Figure 2. Thermal properties of commonly used silicon resonator (Source: SII, IME)



Figure 3. Thermal properties of developed silicon resonator (Source: SII, IME)



New wafer-level packaging for vacuum-sealing MEMS devices


As part of their work in MEMS resonators, SII and IME have pushed ahead with a research project begun in 2007: development of a vacuum-sealing process technology for MEMS devices. A vacuum environment (typically anodic bonding) is used for sealing processes for MEMS devices such as resonators with silicon substrates — but is limited to high-temperatures (400°C~500°C)
and voltages (600~1000V) which can cause warping. Moreover, getters need to be added during sealing to remove gases (e.g. oxygen) produced during the vacuum packaging process.

To fix this, SII and IME have come up with a gold-tin eutectic bonding process, that can take place at temperatures low enough to inhibit release of gases during bonding; thus a higher level of vacuum sealing (<26Pa) can be achieved without using getters, and an external electrode can be drawn out laterally without passing through the substrate. This method improves mechanical strength, and also lowers fabrication costs.


Figure 4. Cross-section of vacuum package with gold-tin eutectic bonding. (Source: SII, IME)


The combined MEMS resonator and new wafer-level packing (WLP) technology have been "successfully tested," and SII plans to further evaluate their performance to verify possibility of a miniaturized, low-power-consumption resonator. Application would be in high-frequency resonators (instead of quartz crystals), enabling low-cost, high-mix low-volume production and miniaturized single-chip high-frequency resonators.

March 10, 2010 – Spanish startup Baolabs claims to have come up with a new way to build MEMS devices using standard CMOS manufacturing: put the devices directly in the wafer rather than on top of it, which it says simplifies production and greatly reduces costs.

The company says its "NanoEMS" technology uses existing metal layers in a CMOS wafer to form the MEMS structure, using standard mask techniques. The intermetal dielectric is etched away through the pad openings in the passivation layer using vapor HF — a step that takes less than an hour using existing volume-production equipment, "insignificant compared to the overall production time," it says. The holes are then sealed and the chip is packaged.

The company says it has built MEMS devices using standard 0.18μm/200mm volume CMOS wafers with at least four metal layers and down to 200nm feature sizes, "an order of magnitude smaller than is currently possible with conventional MEMS devices" — while also offering benefits in smaller size, lower power consumption, and faster devices.

"We have solved the challenge of building MEMS in a completely different way," proclaimed Baolab CEO Dave Doyle. Conventional MEMS technologies "are slow, expensive, and require specialist equipment," and must be built either into a recess in the wafer or post-production on top of the wafer.

Metallic mesh created using NanoEMS technology. (Source: Baolab)

The company plans to target various discrete MEMS applications, including RF switches, accelerometers, and "electronic compasses" (i.e., magnetic sensor), as well as other areas that combine several functions in a single chip. In a phone, for example, the NanoEMS technology could replace GaAs FETs for RF switching, or enabling tunable active RF components such as power amplifiers and low noise amplifiers — thus significantly reducing the number of components needed in a phone. Evaluation samples are due later this year for handset designers and manufacturers and for power amplifier and RF frontend module markets, the company says.

Switch contact created using NanoEMS technology. (Source: Baolab)

March 4, 2010 – During the industry downturn, SVTC Technologies — née Cypress Semiconductor’s Silicon Valley Technology Center, spun off in March 2007 with VC/private equity backing and combined with SEMATECH’s ATDF later that year — did what a lot of companies did: focused on tightening its financials and sales/marketing infrastructure.

But in the background a transition was being crafted, which once the recession lifted would help the company step forward into "a phase of much more focused and improved attention to customers, markets, technologies," according to Bert Bruggeman, newly appointed CEO, in an interview with Small Times.

Bruggeman, whose career path has winded from IMEC to Cypress Semiconductor, then a MEMS startup, and eventually SVTC Technologies, defines three key aspects within this new stage of growth:

Market-driven. Traditionally, the company’s facilities pedigree is for more classic CMOS-based technologies, but SVTC has strived to expand this, mainly leveraging MEMS (e.g. deep-silicon etch) — and now the company wants to more broadly leverage in MEMS, he says. The company also plans to widen the reach within its walls for more derivative technology migrations: e.g., device science to biosciences, "to go after in a more structures, organized way."

During the downturn SVTC acquired "quite a few companies as customers, at a rate that we had not been able to do before," Bruggeman noted, and as the industry climate improves he expects not only new business but also the breadth of customer interests to pick up rapidly. "We still see novel ideas, different technologies alongside more classic CMOS," he said, including "half a dozen PV-type engagements" exploring new materials and cell architectures. (Among them is Bloo Solar, with whom it will develop and commercialize "solar brush" thin-film solar module technology, which uses a 3D junction architecture and billions of tiny "bristles" to enable higher efficiency and an alleged 1.5-3× higher power output than current technologies.) He also expects to see (and will welcome) more activity in MEMS, and in life/biosciences applications (e.g. work on different substrates, development of drug delivery or diagnosis applications) which seek "capabilities such as 248nm and 193nm lithography to make next-generation devices."

Meanwhile, SVTC still wants to embrace its core competency in CMOS-based tech. "We believe the silicon platform is so rich — and maybe infinite — in terms of ideas and innovation," Bruggeman said. "We still see new ideas in transistors, in memories, in image sensors — it’s almost limitless. As those ideas come out of 4-in. and 6-in. environments, and need an 8-in. commercial platform, we’ll be there to bring value."

Service expansion. A core of the company’s business model since Day 1, Bruggeman said, has been to do development for customers, or provide the fab space/capabilities so they can do their own development, with pilot-level production (e.g., 1-2 to 4-5 lots/month). With this next growth phase, Bruggeman "wants to expand a little bit more," to expand pilot production though still keeping it low-volume (maybe a few hundred lots/month). That helps customers end up with more characterized, stable technologies and products, which will both help them eventually transition to SVTC’s higher-volume partners (e.g. TSMC — a "strong partner" but Bruggeman admitted that the company is also now "talking to a few other folks"), and better prepare them to get a foothold in their markets, he explained.

Two facets to SVTC’s business helped it in the recent economic upheaval: the financial advantage of "capex avoidance" for customers but still offer production, and develop in a more variable cost structure vs. a fixed cost structure, a key benefit for startups. "Those two very core basic financial values are extremely attractive in the new and changed economy, where people still want to innovate and have ideas," he said.

The company did manage to add significant capabilities in terms of technology during the downturn — though true to its message, these were key but not extravagant capex investments, "upgrades to migrate and make out existing toolbase more flexible," Bruggeman said. These included another 1-2 types of tools for MEMS; finalizing capabilities for amorphous-silicon (a-Si) work; expanded ALD capabilities; and ability to handle more materials (e.g. transparent substrates) and different precursors.

New ways to help customers. With the R&D work of the company and its customers characterized as "little-r and big-D," there is by definition and necessity a level of intimacy and confidence with SVTC and its customers, starting with the technology and product roadmap, which can lead to "very detailed intricacies of development and discoveries," Bruggeman explained. An offshoot of this is "a level of value and confidence and credibility" — which is something SVTC plans to leverage into a new business offering of management consulting. "We have great toolsets and engineers, and a tremendous amount of commercialization management expertise," he said, estimating key SVTC execs have done around two dozen technology platforms and technology transfers. "We want to wrap that up in a much higher offering…[to] bring the next lever of strategic value-add to customers’ executive management teams."

As an example, Bruggeman described a university spin-off that came to SVTC with "very unique technology" that it wanted to scale up to 8-in. wafer silicon processing, but their management was struggling with the methodologies to do that. So on "a very specific request" from that CEO, Bruggeman and a couple of other key SVTC execs began participating in defining the way the company would plan, execute, and monitor the transition from the university to SVTC, and then to an 8-in. platform. "That got us involved at a management level to put together a framework," he said.