Two-stage iterative decoding algorithms for a class of cyclic codes

This paper presents a class of iteratively decodable cyclic codes. Codes in this class have large minimum distance; however, their Tanner graphs contain many short cycles of length 4. With the conventional iterative decoding based on belief propagation, these short cycles significantly degrade the error performance of the codes. To avoid the degrading effect of these short cycles in performance, two-stage iterative decoding algorithms are devised. Cyclic codes have encoding advantage over other linear block codes. Encoding of a cyclic code in systematic form can be implemented with a single feedback shift-register.