ИСТИНА

Recently there have been a lot of research on low-cost LDPC decoders including some variations of bit-flipping algorithm and finite alphabet iterative decoders. Such LDPC decoders have a considerable reduction in hardware complexity compared to the standard min-sum decoders. This makes them a viable alternative to the BCH codes in some flash memory applications, where the IC area and power consumption are of great concern. In this presentation we propose a new class of low-cost error-correction codes for flash memories. They can be described as generalized concatenated codes, where the inner and the outer codes are LDPC codes. We show that these codes with an appropriate choice of the decoding algorithm have a much smaller word error rate compared to the BCH codes of similar rate, throughput and hardware complexity. Besides, unlike the BCH codes the proposed codes can also be used with soft sensing using multiple reads of the flash memory.