March 2007
Philippe Azoley, Julien Bon, Peter Pfluger
MEMS technologies developed during the '90s have been successfully applied in various industries. Well-educated engineers successfully design and simulate their MEMS devices based on well-known system requirements. However, the subsequent search for one or two appropriate foundries that will "just run the wafers" turns out to be more difficult than anticipated and often ends in failures and redesigns.
MEMS fabs generally have a well-characterized but limited offering of process blocks available under SPC. New MEMS devices generally need process fine-tuning, because suboptimal design-process combinations yield devices that are not cost-effective. The question then arises: how to validate and optimize the design with its associated process adaptations?
Knowing the Probability and Consequences of Failure
Controls and tests done by foundries that are limited to typical in-line and on-wafer tests of dies are insufficient. They must be augmented by functional - that is, application-relevant - testing of devices following assembly and packaging in order to reach high overall yields.
If further inputs from the MEMS integrator's production process are added, then a complete Failure Mode and Effects Analysis (FMEA) methodology can be set-up. FMEA covering the entire production loop is compulsory because most failure modes of new MEMS devices are unknown when they come out of R&D. This process also facilitates the communication between the custom MEMS manufacturer and the QC and engineering departments of the MEMS integrator (See Diagram).
click image to enlarge
The Benefits of FMEA
Proper FMEA encompassing the manufacturer loop AND the integrator loop allows their teams to:
- Define and follow the right indicators (which are a priori unknown)
- Test the right parameters in the right way (addressing a major cost issue)
- Focus and fine-tune process controls
- Anticipate application problems
- Obtain inputs for improvements and new product generations
Rigorous FMEA produces several key benefits to users, including:
- Lowering the MEMS component price over time, while preserving fair margins for both the integrator and the MEMS manufacturer, and
- Optimizing use of the MEMS device in its application and improving its characteristics
In one example of Tronics' application of such methodology, the overall production yield (front-end x back-end) of an extreme-performance, vacuum-packaged MEMS-on-SOI inertial device increased from 43% after its production start in early 2003, to 55% in 2004, 67% in 2005 and over 80% in 2006. The device was systematically analyzed according to 16 identified failure modes (8 in front-end fab processes at wafer level and 8 in back-end processes at the individual component level).
In summary, traditional subcontractor relations or one-sided client-foundry interaction almost inevitably leads to suboptimal products and design or economic failures. Our extensive experience in productization of new MEMS devices underscores the necessity of feeding back application-relevant data and tests results into the MEMS design optimization, process fine-tuning and process-control procedures.
