MRF Reinforcer: A Probabilistic Element for Space Redundancy in Nanoscale Circuits

Shrinking devices to the nanoscale, increasing integration densities, and reducing of voltage levels down to the thermal limit, all conspire to produce faulty systems. Frequent occurrence of both soft faults due to noise and signal coupling, and hard faults due to process variations and defects can be expected. As error-free logic states can no longer be guaranteed, one possible solution is a paradigm shift to a fault-tolerant probabilistic framework. The probabilistic approach is based on the theory of Markov random fields (MRF), which guarantees the optimality of probabilistic reasoning and is extensible to a large number of logic variables interacting locally in small, connected signal nets. This article builds on the recent demonstration of significant noise immunity of MRF based logic elements and introduces a new redundancy element that can handle soft errors and single event upsets in sequential logic. The MRF probabilistic model is mapped onto CMOS circuitry, using feedback be...

• Type: Journal
• Year: 2006
• Where: MICRO
• Authors: Kundan Nepal, R. Iris Bahar, Joseph L. Mundy, William R. Patterson, Alexander Zaslavsky
• Keywords: Hardware, MICRO 2006, MRF Probabilistic Model, Probabilistic, Redundancy Elements