Classification Codes for Soft Information Generation from Hard Flash Reads

In this paper, a tensor product-based classification code (CC) is proposed to generate coarse soft information based on the hard flash reads for use in soft information-utilizing error correction codes (ECCs), particularly low-density parity check (LDPC) codes. Analysis and simulation results have shown that with a small ECC overhead, CC helps the LDPC code decoder achieve better performance compared to a hard-input case. Furthermore, the proposed code has very simple encoding and decoding algorithms; therefore, it can be easily implemented in hardware to aid existing ECC systems.