Novel method for improving performances of normalized MS decoder using WIMAX code

LDPC code is a powerful error correcting code used in many communication systems. Efficient decoding algorithm and hardware implementation of LDPC code is an important issue. The paper proposes an improved version of the Normalized Min-Sum (NMS) decoding algorithm named “Girth-Aware NMS”. The principle of the GA-NMS is to determine “offline” the optimal normalisation factor of the extrinsic message using the structure property of the Tanner Graph associated to the LDPC matrix. The idea can be expressed simply: a check node implied in one, or more, cycle of minimum size (the so-called Girth) in the Tanner Graph is more subjected to self-confirmation than other check nodes. Thus, the extrinsic messages generated by this type of check nodes should be more aggressively “normalized” (or attenuated) than other check nodes. In the context of the IEEE P802.16e standard (WIFI standard), simulation results of the GA-NMS shows both improved decoding performance and reduced average number of decoding iterations.