By Karen Lightman, managing director, MEMS Industry Group
While micro-electromechanical systems (MEMS) industry leaders such as STMicroelectronics, Texas Instruments, Hewlett Packard, Robert Bosch and Kionix rely on their captive fabs to meet volume production, the movement toward fabless and fab-lite models continues to gain ground. InvenSense, for example, has always been a fabless company, and even powerhouse Analog Devices uses a hybrid approach, choosing internal and external foundries to produce MEMS die for inertial sensor products. Despite the advantages of having a captive fab, which supporters say includes greater control over both capacity and intellectual property (IP), the primary disadvantage—cost—has spurred the use of third-party foundries.
In addition to cost savings, companies work with third-party MEMS foundries for a variety of reasons. They may want to prove a design, prototype a design that is already proven, or mass-produce a MEMS device. With a multitude of options, choosing a MEMS foundry is not a simple decision.
Pure-play MEMS foundries, such as Silex Microsystems, Micralyne, Teledyne DALSA, Asia Pacific Microsystems, Innovative Micro Technology (IMT) and Tronics do not offer design services but they do offer volume production. Partially captive foundries offer another alternative. They will fabricate MEMS die for outside customers when there is excess fab capacity.
Companies such as A.M. Fitzgerald & Associates, Nanoshift and SVTC specialize in design and rapid prototyping, and also consult with their clients to find the perfect foundry partner. Dr. Carolyn White of A.M. Fitzgerald & Associates, who specializes in design, analysis and fabrication for her firm, explains why foundry selection is so critical in the MEMS industry: “In the IC world, you can potentially have a device ready to ship to customers in 18 months—but not in the MEMS world. Even with an existing prototype (with a proven process flow), it takes time to choose a foundry partner and bring the process flow into production. Foundries first do an initial prototype run, then move to pilot production and finally go to full-volume production. That takes at least a year and a half. Total time to market can be five years and could cost in the range of US$10 million for a new device using the fabless model.”
White says that a foundry partner with the right experience can help companies to overcome common technical and logistical challenges—such as coupled physics, moving parts, environmental exposure, and test and packaging challenges. MEMS also presents design challenges that foundries cannot meet alone, according to White. With few formal standards, diverse tool sets and foundry-specific design rules not yet available for existing simulation packages, companies need good design and process engineers to work with the foundry throughout the process.