Controlling the Error Floor in LDPC Decoding

TL;DR

This paper investigates how to reduce the error floor in LDPC decoding by balancing message growth within absorption sets, using analysis and importance sampling to verify and optimize performance at ultra-low BER levels.

Contribution

It introduces a method to lower the LDPC error floor by controlling message dynamics in absorption sets, supported by analysis and importance sampling verification.

Findings

Proper message balancing reduces the error floor significantly.

Importance sampling confirms the effectiveness of the proposed method.

Iteration count and message quantization impact ultra-low BER performance.

Abstract

The error floor of LDPC is revisited as an effect of dynamic message behavior in the so-called absorption sets of the code. It is shown that if the signal growth in the absorption sets is properly balanced by the growth of set-external messages, the error floor can be lowered to essentially arbitrarily low levels. Importance sampling techniques are discussed and used to verify the analysis, as well as to discuss the impact of iterations and message quantization on the code performance in the ultra-low BER (error floor) regime.