Practical algorithm-based fault tolerant DFT system implementation on a hypercube multiprocessor

Most practical fast DFT systems with non-trivial N number sample points are generally implemented by only several processing units. Efficient time-sharing these few units and including low cost fault detection and reconfiguration is an important design issue for highly dependable DFT systems. The authors discuss achieving low cost fault detection for DFT system by scheduling the limited processing elements on a hypercube. To simplify the discussion, only radix-2 DFT systems are discussed, i.e., N will be always assumed as a power of 2