Adaptive Mitigation of Radiation-Induced Errors and TDDB in Reconfigurable Logic Fabrics

Self-reliance capabilities of mission-critical systems gain importance as technology scaling and logic capacity of SRAM-based reconfigurable devices increase. The Sustainable Modular Adaptive Redundancy Technique (SMART) is evaluated to optimize the reliability, availability, and energy efficiency of reconfigurable logic devices with a given area footprint. A Monte Carlo driven Continuous Markov Time Chain (CMTC) simulation is conducted to assess availability using runtime adaptation with SMART in comparison to conventional design-time static Triple Modular Redundancy (TMR) techniques. In harsh environments, adaptive redundancy is shown to improve system availability under lengthy repair times, and to a more significant degree under rapid recovery times. When compared to TMR, adaptive redundancy achieves power savings ranging from 22% to 29%, at a reduced area cost ranging from 17% to 24%, while maintaining comparable levels of availability.