Use of digit-serial computation to create fault-tolerant designs

It is shown how digit-serial computation may be used to provide a degree of fault-tolerance in linear computations in cases where parallelism cannot be used, because of excessive cost. By splitting a parallel input stream into several bit-serial streams and using one extra stream for error checking, it is possible to detect intermittent and permanent faults in the circuit. The proposed method is useful for digital implementation of filters and other linear circuits in cases where throughput requirements preclude the use of parallelism. By using this method, it is possible to achieve fault-detection with a modest increase in hardware