Fault-tolerance techniques for hybrid CMOS/nanoarchitecture

Abstract

We propose two fault tolerance techniques for hybrid CMOS/nano architecture implementing logic functions as Look-Up Tables. We compare the efficiency of the proposed techniques with recently reported methods that use single coding schemes in tolerating high fault rates in nanoscale fabrics. Both proposed techniques are based on error correcting codes to tackle different fault rates. In the first technique, we implement a combined two dimensional coding scheme using Hamming and BCH codes to address fault rates greater than 5%. In the second technique, Hamming coding is complemented with Bad Line Exclusion technique to tolerate fault rates higher than the first proposed technique (up to 20%). We have also estimated the improvement that can be achieved in the circuit reliability in the presence of Don't Care Conditions. The area, latency and energy costs of the proposed techniques were also estimated in the CMOS domain.