GB1363586A - Error detection and correction circuits - Google Patents

Abstract

1363586 Digital data error detection and correction INTERNATIONAL BUSINESS MACHINES CORP 14 June 1972 [26 June 1971] 27710/72 Heading G4A A modified Hamming code is used to detect multiple errors, and to correct single errors, in words, e.g. read from a data store. A modified Hamming code is defined as a Hamming code in which each data bit is related to an odd number of check bits, one half of the data bits in an information word are related to an even number of check bits of one half of the check bits and to an odd number of check bits of the other half of the check bits, the other half of the data bits in an information word are related to an odd number of check bits in said one half of the check bits and to an even number of check bits in said other half of the check bits, and the data bits of each byte are all uniquely related to the same half of the check bits. Each word has 64 data bits Di and eight check bits Cj and the bits Di are passed to check bit generators RBG1 to regenerate the check bits (Gj). The check bits Gj are compared with the check bits Cj in a comparator 2 to produce eight syndrome signals, erroneous bits being indicated by a 1 syndrome. These pass to an error bit location circuit FBL3, to a circuit UGP4 which. finds out if the number of 1 syndromes is odd or even, and to a circuit FKF5 which indicates whether there are any errors. A circuit 6 sends a KBF signal when there is a correctable error (odd number of syndromes with binary 1) and a UKBF signal when the error is uncorrectable (even number of syndromes other than zero with binary 1). If the error is correctable, the data bits Di are corrected in circuit DBM7 to give Di<SP>1</SP>. A circuit 8 then forms a parity bit for each eight bits of Di<SP>1</SP>. In another embodiment (Fig. 3, not shown) a parity bit is shown which enables the corrected parity bits to be produced simultaneously with the corrected data bits.