| dc.description.abstract | Microelectromechanical systems (MEMSs) consist of moving mechanical microparts often integrated with electronics and optics that may be used for sensing or actuating purposes. MEMS and micro-opto-electromechanical system (MOEMS) packaging requirements vary widely with application, but they generally involve protecting the device from the damaging effects of the environment, such as moisture and dust. Reliability is often not considered as a design factor during product development. Rather, reliability is assessed using life tests, accelerated tests, and other techniques after a product has completed its development cycle. The goal of design for reliability (DfR) is to be proactive by introducing reliability early in product development so that concerns are identified and assessed at every stage, from the conception to obsolescence. In this paper, we present a framework for applying DfR principles to MOEMS packaging. Such an approach is desirable for several reasons. First, it reduces the cost and time for product development by departing from the “build-test-rebuild” approach. Second, it provides better understanding of the process input-output relationships, so the practitioner is better able to make informed design decisions. Lastly, this can lead to enhanced product performance, reliability, and reduced cost. To demonstrate the use of DfR in MOEMS packaging, we present a case study involving carrier level packaging of a MOEMS switch. The reliability requirements for this device are stringent, namely, a shelf life of 25years or more, requiring hermetic sealing through the use of metal seals and no organic compounds inside the package. Simulation and experiments are used systematically in order to guide the package design and process windows, ensuring that the device passes MIL-STD reliability tests. The packaging processes include fluxless die-to-carrier attachment, optical fiber-to-carrier attachment, and hermetic sealing. Results show that our packaging approach can determine adequate process windows using only a small number of reliability experiments. | |
