NB-LDPC: Absorbing set and importance sampling

The identification of error sources when sub-optimal decoding algorithms are used is an open point both in binary and non-binary LDPC. Here a mathematical definition for absorbing sets in the case of non-binary LDPC is given and several examples are proposed to illustrate it. The definition is consistent with the error patterns resulting from Montecarlo (MC) simulations at high SNR. The absorbing sets patterns, as defined here, are used to predict the error floor of NB-LDPC with the Importance Sampling (IS). The IS and the MC simulation results match up very well when both are available.