An Improved Markov Random Field Design Approach For Digital Circuits

As the MOSFET dimensions scale down to nanoscale level, the reliability of circuits based on these devices decreases. Therefore, a mechanism has to be devised that can make the nanoscale systems perform reliably using unreliable circuit components. The solution is fault-tolerant circuit design. Markov Random Field (MRF) is an effective approach that achieves fault-tolerance in integrated circuit design. The previous research on this technique suffers from limitations at the design, simulation and implementation levels. As improvements, the MRF fault-tolerance rules have been validated for a practical circuit example. The simulation framework is extended from thermal to a combination of thermal and random telegraph signal noise sources to provide a more rigorous noise environment for the simulation of nanoscale circuits. Moreover, an architecture-level improvement has been proposed in the design of previous MRF gates. The re-designed MRF is termed as Improved-MRF. By simulating various test circuits in Cadence, it is found that Improved-MRF circuits are 400 whereas MRF circuits are only 10 times more noise-tolerant than the CMOS alternatives.

The authors work under the banner of fault-tolerance research group in Universiti Teknologi PETRONAS (UTP). The group is conducting research on various aspects of fault-tolerant circuit design with the support of UTP and MOSTI (Ministry of Science, Technology and Innovation) Malaysia.