posted on 2009-07-14, 00:00authored byTimothy J. Dysart
Much has been written and predicted about the demise of Moore's law in advancing computing technology. While many of these predictions have fallen by the wayside, the end may be approaching since the 'red brick wall' (areas where scaling progress may end if breakthroughs are not identified) is barely being pushed further into the future. As a result, numerous nanoelectronic devices are being investigated to determine if they can augment or replace silicon-based transistors. Due to a variety of factors, a significant challenge when building a system from any of these devices is how to produce a reliable, predictable system from unreliable components with unpredictable behavior? In this dissertation, the reliability aspects of a specific nanoelectronic architecture -- Quantum-dot Cellular Automata (QCA) -- are explored. QCA is a unique architecture to explore since the same basic device, a simple cell, is used to implement both logic and interconnect components. The device itself can be implemented using different physical properties, which in turn, influences how the signal routing within the QCA circuits and systems should be designed. This dissertation computes the reliability of QCA circuits and systems based on the reliabilities of the underlying components. This is used to identify how reliable the components should be to achieve a target circuit or system reliability and to identify which components are the most critical to circuit reliability. Additionally, various techniques for improving reliability through the use of hardware redundancy are evaluated to determine how a reliable QCA system should be designed. Lastly, various organizations for a specific interconnect component, a straight wire segment, built from a specific device type, are compared to determine how reliable a component can be. Throughout this dissertation, a constant theme is observed: the organization and reliability of the signal routing has a major impact on circuit and system reliability.