Abstract
The performance of a belief propagation decoder for low-density parity-check codes is limited by the presence of trapping sets in the code’s graph. This leads to an error floor at higher signal-to-noise ratios. We propose the use of an augmented decoder which operates by iteratively decoding on a set of graphs which have a subset of repeated check nodes. We compare the augmented decoder to other modified belief propagation decoders that have been presented in the literature. We show that for all the codes considered, the augmented decoder yields the best frame error rate in the error floor region.
Highlights
Low-density parity-check (LDPC) codes provide good error correction performance when a belief propagation (BP) decoder is used [1]
We show that in terms of error floor frame error rate (FER), the augmented decoder outperforms other methods of trapping set mitigation presented in the literature
Variable nodes will converge in different orders depending on which checks have been duplicated, this can potentially lead to avoiding a trapping set that occurs when using standard BP
Summary
Low-density parity-check (LDPC) codes provide good error correction performance when a belief propagation (BP) decoder is used [1]. BP computes the marginal probabilities of the transmitted bits with reasonable accuracy even though the graph representation of the code contains loops These loops can lead to trapping sets which degrade decoder performance, causing an error floor at high signal-to-noise ratios (SNR) [2]. We apply the augmented decoder to a suite of regular and irregular LDPC codes, demonstrating that it is effective at mitigating error floors due to trapping sets. The message sent from check node cj to variable node vi is: Rj,i = 2 arctanh ⎝ These values can be used to make a new estimate at the transmitted codeword: xi. The error floor can be lowered by mitigating trapping sets; this is the aim of the augmented decoder
Sign-in/Register to view highlights & summary 
Published Version (
Free)
Talk to us
Join us for a 30 min session where you can share your feedback and ask us any queries you have
Schedule a call