Abstract
Due to supply voltage reduction and process variations effects, the error free margin for dynamic voltage scaling has been drastically reduced. Thus, supply-Voltage Over Scaling (VOS) has emerged as an efficient means to achieve ultra-low energy efficient systems, that tradeoff energy efficiency and reliability. Recently, N-Modular Redundancy (NRM) has been used as an effective fault-tolerant design method in which N copies of an unreliable block are employed and a voting strategy such as majority, median or mean is deployed to select an output. This paper presents a novel NMR voting algorithm based on the maximum a posteriori (MAP) and the statistics of output bit-failure rate. The design of Distributed Arithmetic (DA) filters is employed to demonstrate the effectiveness of the proposed technique and show the trade-offs between the robustness of the system and the power savings. Simulation results show significant enhancement in terms of output signal to noise (error) ratio and reliability as compared to other NMR fault tolerant techniques. A case study of a low-pass DA filter employed in a simple communication system shows that up to 12% power savings could be achieved when compared to the conventional NMR while satisfying the system bit-error performance.