Defect oriented test of inertial microsystems

Microsystems consist of transducers (sensor and actuator devices) that are integrated with electronic circuits to act as real-time interfaces between the largely non-electrical real world and electrical computational systems. The development and deployment of new microsystems greatly depends on advances in design, test, and packaging. Our work has been focused largely in the area of microsystem test, which can account for a significant fraction ( e.g., one-third) of the total manufacturing cost of the product.; In this dissertation, we have addressed several areas of microsystem test, including design-for-manufacturability (DFM) and built-in self-test (BIST), that have not been previously dealt with adequately. Since knowledge of the impact of failure sources is necessary to develop efficient test strategies, we have considered, in detail, the effect of spot defects caused by failure sources like particle contaminants on the behavior of inertial microsystems (e.g., resonator and accelerometer). We relied on finite element analysis (FEA) to extract specifications of defective microstructures for the purpose of defect classification. The information acquired through defect classification has allowed us to integrate standard, specification-based test methods into a closed-loop DFM flow that has been demonstrated to generate microsystem layout designs more resistant to catastrophic defects caused by particle contaminants. Therefore, we have shown how knowledge of misbehavior caused by defects can be effectively used to improve yield.; Since a microsystem device can be affected by multiple failure sources, we have extended our defect investigation to include the interaction of multiple defects on a single device. Our analysis of multiple defects has shown the possibility of two defects canceling each other's effects and preventing the manifestation of any misbehavior, a phenomenon called misbehavior masking. We have shown how misbehavior masking can lead to defective devices that evade standard, specification-based tests, yet later fail in the field. In our quest to achieve faster defect analysis, we have found the use of high-level simulation techniques, specifically, behavioral modeling, to be much more effective (>50X faster compared to FEA) at tolerable loss in accuracy (<10%).; We have used fast behavioral defect simulation to further investigate misbehavior masking and developed a BIST technique that is capable of detecting masked defects by indicating the presence of device asymmetry. Our BIST approach exploits the symmetrical characteristics of microstructures by creating and sampling signals from symmetrically-located regions of the microstructure. Increasing observability in this way reveals more about the structural deformation resulting from local manufacturing variations and spot defects. The differential nature of the BIST allows it to be independent of calibration, which in turn makes it suitable for manufacturing test, a feature lacked by prior BIST methods. (Abstract shortened by UMI.)...