Unique techniques for a reliable SBNR architecture

Abstract

An arithmetic error code is presented whose properties are derived from a union of redundancy and the signed binary number representation (SBNR) system. The main advantage of the SBNR error code is the limited distance an error propagates from a faulty gate. The SBNR error code requires duplication of an entire full adder, with an optional portion being triplicated. The decision hardware contains the circuitry needed to select the correct answer. It based its decision on evaluation of the communication digits in conjunction with the two possible answers. One disadvantage of the SBNR code is in the method of detection. If an identical error exists in both of the original derivations, an incorrect result will be given for that particular digit. The outstanding feature of this error code is the possible number of errors that can be corrected.<>