LDPC Decoder Bit-Word Correction for Quantized Message Passing

The patent applies parameter changes by transforming the representation of log-likelihood ratio values through correction operations. Specifically, it adjusts the scaling factor and offset parameters when mapping full-precision LLRs to quantized bit-words, enabling better utilization of the limited bit-word levels while maintaining decoding accuracy. This resolves the contradiction by changing the numerical parameters of the message representation rather than increasing bit-word size.

The patent implements preliminary action by pre-calculating and storing correction values in lookup tables before the actual decoding process. These correction values compensate for the quantization effects of limited-bit representations. By performing this correction preparation in advance, the system achieves high decoding performance using compact bit-words without requiring complex real-time calculations during decoding.

The patent changes the parameter representation by introducing correction terms that expand the effective value range of shorter bit-words. Instead of directly using the limited-bit values, the system applies parameter transformations that recover the lost precision, effectively increasing the representable value range while maintaining compact bit-word formats for resource efficiency.

The patent introduces correction lookup tables as intermediary structures that bridge the gap between limited-bit representations and full-precision requirements. These tables serve as mediators that translate compact bit-words into accurate log-likelihood ratio values, preventing information loss while maintaining resource efficiency. The intermediary correction mechanism allows shorter bit-words to convey more information than their raw bit length would suggest.