 |
Oct. 28, 2003 – If you put a bunch of MEMS experts in a room and ask them what are the industry’s biggest problems, they will probably cite the telecom meltdown, the contraction in the broader technology market and the reluctance of venture capitalists to support unproven technology.
 |
Most likely a few also will mention packaging. Unlike the external forces battering the industry, packaging is a less recognized maelstrom within. Industry insiders say packaging is usually responsible for at least 60 percent of the cost of a MEMS device, and sometimes as much as 85 percent. Nevertheless, it suffers from a chronic lack of attention.
 |
“It doesn’t get the grant dollars,” said Dave Monk, sensor development engineering manager for Motorola Inc.‘s sensor division.
“Another issue is the talent pool,” said John O’Connor, package development manager for Digital Light Processing at Texas Instruments Inc. “It takes a good two to four quarters to train people.”
“Everyone thinks it’s still an integrated circuit,” said Bob Mehalso, a packaging consultant with Microtec Associates. “It just turns out that packaging a MEMS device and packaging an integrated circuit is totally different. And it surprises me how many people don’t understand that.”
Mehalso is on the technical advisory board of Ardesta LLC, the parent company of Small Times Media.
These and other experts say packaging is hands down the biggest challenge for commercializing more MEMS devices. The technical difficulties involved in packaging can be considerable. But even more problematic is an industry wide mindset that views packaging as secondary — if it is considered at all. The experts say that attitude is beginning to change, though, as the industry matures.
Mehalso told the story of a micropump that never made it to market. “They spent a few million dollars on it,” he said, “but it was just not designed from a microsystems perspective.” The design was a miniaturized version of a macroscale process that didn’t address packaging until the end of the design cycle.
By the time engineers realized they couldn’t economically package and scale up the device, Mehalso recalled, it was too late. “The only way to cut the price down was to start over,” he said.
MEMS veterans say that’s not an isolated story. With startup funding hard to come by, Mehalso added, getting it right the first time is even more important than it was before.
“You either design all these things together the first time or you will do it the second time,” said Joe Giachino, industrial liaison at the Wireless Integrated MicroSystems Engineering Research Center at the University of Michigan.
That is, he added, “if you get a second chance.”
The attitude shift experts say the industry needs has a number of facets: It includes factoring in that MEMS devices are categorically different from ICs despite sharing some fundamental processing technologies. It acknowledges that each type of MEMS device interacts with the world in a different way, and demands custom or at least semi-custom solutions.
It also understands that packaging considerations must be part of the design process from day one, especially since MEMS devices can be damaged by the very processes used to create them. It recognizes that more than one packaging solution may be needed: to protect the device during manufacturing and to shield it in its working environment. Finally, it requires that packaging be taught in academic programs as an integral part of making MEMS devices.
The proof is in the package
At Motorola’s sensor division in Tempe, Ariz., Monk and Packaging Manager Dave Chapman recently grappled with packaging conundrums while developing a new pressure sensor for tire pressure applications.
“The particular challenge with the pressure sensor is that it’s a device that’s open to the environment,” said Monk.
Within a tire, he said, there are oils, lubricating grease and other materials that could cripple the sensor. To keep the junk out but let the pressure in, the team initially tried gel, a solution used decades before in the development of disposable blood pressure sensors.
It led to an interesting phenomenon: The centrifugal energy the gel acquired from the tire’s rotation put an additional force on the pressure sensor. Essentially, said Chapman, “The added mass to the diaphragm made it an accelerometer.”
It’s the kind of thing you’d want to know about long before you completed a design, he said. While the effect might ultimately allow Motorola to develop a combination pressure sensor and accelerometer, the more immediate need was for a tire pressure sensor that worked.
Back at the drawing board, the team developed an alternative: a laminated, polymer filter. Like the gel, it let in air pressure while filtering out contaminants. However, the filter didn’t add significant mass to the sensor module.
According to Monk and Chapman, a healthy respect for the challenges of packaging is critical for developing MEMS products. At the extreme are devices like manifold absolute pressure sensors, which Motorola puts through acid exposure tests. But, they said, even fully enclosed MEMS devices such as inertial sensors need to be packaged so that sensing elements aren’t affected by everyday environmental stresses: a car’s vibration, for example.
In addition, said Monk, the trend in the industry toward ever more sensitive devices — hyper-sensitive inertial sensors for cell phones with a tilt-based user interface, for example — exacerbates the problem. “As you go to a more and more sensitive device, it becomes more sensitive to packaging stress.”
Shedding some light on the matter
O’Connor, the package development manager for Digital Light Processing at Texas Instruments, joined the DLP group four years ago to work on thermal management problems. Within weeks he was recruited to work on packaging problems for the digital micromirror device (DMD), an array of MEMS mirrors used for projection applications.
“The DMD is a complex MEMS device with 500,000 to 2 million moving structures,” he said. Providing a reliable and stable operating environment is, “a significant challenge.”
That’s especially true as TI ramps up volume. From 1997 to 2000, O’Connor said, the device was in low volume production. The company began including it in portable business projectors, where it enabled a new generation of lighter, brighter products. More recently, the company deployed the DMD in home theater projectors, digital televisions and large venue and cinema projectors, necessarily increasing its volumes.
TI’s decision to blend processing and packaging steps together is one reason the DMD is successful, MEMS experts say. After micromirrors are created on a wafer, a protective layer is applied and the wafer is partially sawed. Later, after the chips are tested, the wafer is broken along the scoring lines created by the saw.
To be closer to its Asian customers and save money, TI recently began outsourcing DMD packaging to Amkor Technology Inc., a West Chester, Pa., packaging specialist with Asian manufacturing facilities in a number of countries.
“The bottom line is we offered them lower cost,” said Michael Steidl, Amkor’s senior vice president of advanced product development.
Although Amkor provides packaging, test and assembly services for ICs, optical devices, MEMS and more, there’s a difference when it comes to MEMS, Steidl said. For starters, “you can’t use a cheap package.”
However, he added, there are certain applications where MEMS can be adapted to more standardized types of packaging solutions. That can save companies a lot of money. For example, “In the wireless world, they all of a sudden want low cost,” Steidl said. “That screams for standard packaging.”
However, standard packages generally only accommodate MEMS sensors that are capped with a protective cover during wafer processing. Such caps also protect the sensing elements from the sawing process. Devices like RF components or inertial sensors — those that don’t require a hole or window through which their sensor elements are exposed to the world — are more likely to be packaged this way.
Regardless of whether or not a cap is used, “Packaging can become pretty expensive if they don’t think about it right up front,” Steidl said.
That isn’t anything new to some in the industry. “It was back in the Ice Age when they were chasing woolly mammoths,” said Gene Burk, now sales manager for MEMS products for DALSA Semiconductor and then an engineer for Fairchild Semiconductor Corp.
He was asked to improve the performance and reduce the cost of a pressure sensor. His solution: redo the entire design with an eye toward a different package.
In order to replace an “elaborate and expensive” manual solution that required hand assembling a package around a sensing element (a diaphragm), he sandwiched two wafers together, one with a micromachined diaphragm and the other with a cavity.
The biggest challenge involved getting pressure onto the diaphragm without destroying it. Exposing it directly to air would let in corrosives and water. So Burk put a gel in the cavity that let pressure through but kept particles and moisture out. The techniques were later used in developing the disposable blood pressure sensor.
In the past, packaging wasn’t emphasized in educational programs. Many research teams won grants for developing devices but not packaging. Their students learned to appreciate the intricacies of MEMS, but were incapable of meeting industry’s growing need for MEMS packaging.
To that end, the National Science Foundation established an engineering research center (ERC) at the Georgia Institute of Technology in 1994 devoted to microsystems packaging research. Today, the interdisciplinary Microsystems Packaging Research Center comprises 30 faculty and more than 300 students, as well as industrial participation.
The Packaging Research Center takes an approach called system-on-package, where researchers attempt to put an entire microsystem — which could include a microprocessor, communications components, actuators and sensors — onto a single package.
“You design the system and its package simultaneously,” said Farrokh Ayazi, a MEMS researcher and a member of the packaging center. “Co-design is the term that we use.”
By designing systems and packages in tandem, Ayazi said, the group aims to cut the overall costs. Co-design makes it possible to use less expensive standard packages. It could also make it possible to put multiple MEMS components inside the same package, cutting costs by making the components share the same housing.
However, he cautioned, it’s not easy. The low-cost packages, for instance, are not hermetic; other approaches are required for systems that need an airtight seal.
Furthermore, the leads that make the electronic connection between a microsystem and whatever it is linked to can also compromise a seal. Ayazi said the center is investigating ways to manufacture devices with leads that extend from the bottom of the substrate.
Although a MEMS device might function perfectly well in the controlled environment in which it was created, “Everyone in industry wants to know what the performance is after it’s packaged,” Ayazi said. Only then is the device a viable product.
Ken Wise, director of the Wireless Integrated MicroSystems ERC at the University of Michigan, said the university’s graduate program is pursuing a similar concept. A current environmental monitors project includes development work on a modular package with individual system elements that can be interchanged with one another.
The move toward complete microsystems also reflects a change in mind-set, according to Wise. The integrated approach incorporates packaging into the design from the very beginning. As these packaging trends continue, he said, the distinction between sensor design and sensor packaging begins to erode.
Motorola’s Monk put it this way: “The bottom line is that, as we learn more and more, we’re considering packaging and transducer design to be very close, if not synonymous.”
Coming Wednesday: Partnering makes the difference for packaging, testing and assembly.