Abstract:
Basic redundancy information is non volatily stored in a reserved area (that is an area of the array that is not addressable by the user of the device) of the addressable area of the array and is copied on volatile storage supports at every power-on of the memory device. The unpredictable though statistically inevitable presence of fail array elements also in such a reserved area of the memory array that would corrupt the basic redundancy information as established during the test-on wafer (EWS) phase of the fabrication process and thus increasing the number of rejects, lowering the yield of the fabrication process, is effectively overcome by writing the basic redundancy data in the reserved area of the array with an ECC technique, using a certain error correction code that may be chosen among so-called majority codes 3, 5, 7, 15 and the like or Hamming code for 1, 2, 3 or more errors, in function of the fail probability of a memory cell as determined by the testing on wafer of the devices during fabrication (fail probability of the specific fabrication process used). A significant area saving is achieved compared to the use of fuse arrays and other known approaches.
Abstract:
The invention relates to a method and system for correcting errors in multilevel memories, both of the NAND and of the NOR type. The method provides the use of a BCH correction code made parallel by means of a coding and decoding architecture allowing the latency limits of prior art sequential solutions to be overcome. Two possible solutions are shown. The parallelism being used for blocks C, 1 and 3 can be chosen in order to optimise the system performances in terms of latency and device area.
Abstract:
A method for making error corrections on digital information coded as symbol sequences ( x ), for example digital information stored in electronic memory systems or transmitted from and to these systems is described, providing the transmission of sequences ( x ) incorporating a portion of error corrector code allowing the sequence ( x ) which is more probably the original transmitted through the calculation of an error syndrome using a parity matrix to be restored when received. Advantageously according to the invention, the error code incorporated in the original sequence ( x ) belongs to a non Boolean group.
Abstract:
A reading method for a memory device with error-correcting encoding envisages the steps of: carrying out a first reading of a plurality of memory locations (A 0 , A 1 , ..., A LS-1 ) to generate a first recovered string (S 1 ), and performing a first decoding attempt using the first recovered string (S 1 ). When the first decoding attempt fails, the memory locations are read at least one second time, and at least one second recovered string (S 2 -S N ) is generated. On the basis of a comparison between the first recovered string (S 1 ) and the second recovered string (S 2 -S N ), a modified string (S M ) is generated, in which erasures (X) are located, and at least one second decoding attempt is carried out using the modified string (S M ).
Abstract:
The invention relates to a method and system for correcting errors in multilevel memories using binary BCH codes. The number of errors is estimated by analyzing the syndrome components (5). If the number of estimated errors is one, then simple decoding for a Hamming code is performed. Otherwise, conventional decoding of the BCH code is carried out (2,3). This avoids the computation of the error locator polynomial and its roots in the presence of only one error and, thus, reduces the average decoding complexity.