GB1402613A - Detection and correction of errors in information words - Google Patents

Abstract

1402613 Digital data error correction INTERNATIONAL BUSINESS MACHINES CORP 10 July 1972 [10 July 1971] 32158/72 Heading G4A Digital electric apparatus detects and corrects errors in information words read from a data store 1 (e.g. of monolithic semi-conductors), the words each having check bits. Each word has 16 data bits and 6 check bits (e.g. Hamming code) which bits allow the correction of single errors and the detection of double errors. A word is read out of the store and, of a double error is detected, the word is read out again, If the word has only sporadic errors it will probably be correct or only have one, correctible error when read-out again, but, if at least one error is permanent, then the or one of the permanent errors is corrected as described below using circuit KS5 and the sporadic error is later corrected in a known manner by circuit FKS3. Correction of Permanent Error.-In a first step CPU of the processor generates a first data pattern, the first bit being 1 and the rest zero, 6 check bits are generated for these, and the 22 bit word is stored in the location having a double error. Subsequently the stored word is read out and applied to error correction circuit FKS3. If a single error is detected, counter Z9 is incremented and a 1 passes to shift register 10 which is shifted once each checking step. In a second step the CPU generates a 16 bit data pattern where the second bit is 1 and the rest are 0 and the process is repeated. This continues for 16 steps (or however many data bits there are in the word), after which register 10 contains a pattern of 0's and l's (Fig. 4, not shown) used for correction in circuit FKS5 comprising n + k Exclusive OR circuits. A read only store FWS11-FWS16 generates checkbits P from the pattern in SR10 and applies them to correction circuit KS5.