Absorbing sets are combinatorial structures in a Tanner graph that have been shown to characterize iterative decoder failure, and particularly error floor behavior, of LDPC codes. In this paper, we examine the connection between absorbing sets and the syndromes of their support vectors. Using this framework, we provide a new characterization of fully absorbing sets, which have been considered the most harmful for iterative decoders. We also show how the sets of absorbing set support vectors appear as translates of codewords in subspaces of the code. These techniques are used to derive new search methods for absorbing sets.
展开▼
