Message Passing Algorithms for Phase Noise Tracking Using Tikhonov Mixtures

TL;DR

This paper introduces a low complexity iterative decoding algorithm for strong phase noise channels using Tikhonov mixtures, with a novel mixture reduction and clustering method that improves performance over existing algorithms.

Contribution

The paper presents a new mixture reduction algorithm with proven bounds and a clustering method for Tikhonov mixtures, enhancing phase noise tracking efficiency.

Findings

The proposed algorithm outperforms state-of-the-art low complexity methods.

The Tikhonov mixture approach is equivalent to tracking multiple phase trajectories.

Complexity is reduced by limiting the number of mixture components.

Abstract

In this work, a new low complexity iterative algorithm for decoding data transmitted over strong phase noise channels is presented. The algorithm is based on the Sum & Product Algorithm (SPA) with phase noise messages modeled as Tikhonov mixtures. Since mixture based Bayesian inference such as SPA, creates an exponential increase in mixture order for consecutive messages, mixture reduction is necessary. We propose a low complexity mixture reduction algorithm which finds a reduced order mixture whose dissimilarity metric is mathematically proven to be upper bounded by a given threshold. As part of the mixture reduction, a new method for optimal clustering provides the closest circular distribution, in Kullback Leibler sense, to any circular mixture. We further show a method for limiting the number of tracked components and further complexity reduction approaches. We show simulation results and complexity analysis for the proposed algorithm and show better performance than other state of the art low complexity algorithms. We show that the Tikhonov mixture approximation of SPA messages is equivalent to the tracking of multiple phase trajectories, or also can be looked as smart multiple phase locked loops (PLL). When the number of components is limited to one the result is similar to a smart PLL.