LDPC Codes Explained
A practical introduction to low-density parity-check codes, sparse parity-check matrices, Tanner graphs and iterative decoding.
Low-density parity-check (LDPC) codes are linear error-correcting codes defined by sparse parity-check matrices. Their importance comes from the combination of strong coding performance and decoders that can exploit the sparse graph structure in a highly parallel way. LDPC codes are used or studied across communications, storage and quantum error correction.
Core idea
The parity-check matrix H describes which code bits participate in which parity constraints. Because H is sparse, it can be represented as a Tanner graph with variable nodes for code bits and check nodes for parity constraints. Iterative decoders exchange messages along graph edges. Sum-product and min-sum are common examples. Their behaviour depends not only on the underlying code but also on the chosen matrix representation, scheduling, quantisation and stopping rules.
Engineering relevance
Two LDPC constructions with similar rate and nominal distance can behave differently under a practical decoder. Short cycles, trapping sets, absorbing sets, pseudocodewords and message saturation can dominate the error floor. Hardware implementation adds another layer: memory bandwidth, edge storage, routing, parallelism and update schedule can determine throughput and area.
Key points
- Sparse parity checks enable iterative message-passing decoders.
- Tanner-graph structure influences convergence and rare failure modes.
- The printed parity-check matrix is a representation of the code; equivalent representations can decode differently.
- Waterfall performance and error-floor behaviour should both be measured.
Continue exploring.
This page explains established concepts and engineering principles. Proprietary JobMettle methods and internal ThinkQuick technology are not disclosed.