Approximate-min' Constraint Node Updating for Ldpc Code Decoding

This work inrmduces a technique for updaring messages rhar originare ar rke consrrainr nodes of bi-partite graphs in Low-Density Parity-Check codes. The rechnique computes only WO ourgoing magnitudes ar each consrrainr node and exhibirs no measurable performance loss as compared ro exact belief propagation which compures a unique magnirude for each departing edge fmm a given consrrainr node. The technique eliminares rke need for memory based table look-up in rhe consrrainr node processing and hns been implemented using only shifr, add, and comparison operarions. Finite wordlength resulrsfor a 'worsf-case' code indicare rhar 6 bir quamizalion yields performance rhar is similar IO rhar pmvided by full $oaring poinr compurarion.