On-Chip Diagnosis for Aging and Early-Life Failure Detection

Carnegie Mellon University, Pittsburgh PA

Investigators

Abstract

Despite the best efforts of the semiconductor industry, integrated electronic systems are not inherently robust because they are "born" (manufactured) with a unique "personality" that stems from the industry's inability to precisely fabricate their underlying circuits, and to write software that controls the resulting uncertainty. An integrated system personality is further shaped by its environment (e.g., temperature, noise, etc.) and usage (i.e., the frequency and type of applications executed), and since both can fluctuate over time, so can the system personality. Systems also "grow old" and degrade predictably due to various wear-out mechanisms but also unexpectedly due to early-life failure sources. These "nature and nurture" influences make it extremely difficult to design a robust integrated system that will operate optimally for all possible personalities. This research achieves integrated-system robustness by judiciously testing and diagnosing the system for any faults through the development of a generalized, scalable on-chip circuitry that tests and diagnoses any system without significant power and performance penalty. Success of this research effort will help ensure that integrated-system manufacturers can produce robust systems, in the most cost-effective manner, despite the large number and variety of failure sources that negatively impact system reliability. Also, our need for continued development of STEM (science, technology, engineering and mathematics) education in under-represented groups will be aided by this effort via involvement of a diverse group of graduate and undergraduate researchers.