UC San Diego

As CMOS technology advances to the nanometer scale, semiconductor industry is enjoying the ever-increasing capability of integrating more and more devices and elements on a single die. Meanwhile, the reliability of the integrated circuit (IC) product is being severely challenged, as many previously negligible noise effects are becoming more prominent, causing significant performance and reliability degradations of nanometer integrated circuits. In particular, radiation-induced transient error is quickly evolving to a serious limiting factor in the circuit reliability. Unfortunately, it has not been sufficiently and successfully addressed in previous technology generations, especially in cost- sensitive mainstream applications. Due to tight design constraint, budget and application requirement, traditional redundancy-based techniques that have been exploited in space and mission-critical applications are no longer applicable. There is an urgent need for cost- effective techniques, methodologies and flows to facilitate the development of reliable IC products. This dissertation is dedicated to the quest for solutions in the analysis, design and optimization of highly error- tolerant nanometer circuit systems. As will be elaborated and demonstrated throughout the entire dissertation, all developed techniques and methodologies share distinguished characteristics of being novel, accurate, economical, practical and scalable, as compared to other existing works. Together they form a unified and automated reliability optimization framework that will enrich the legacy of the IC design industry