EXIT Chart Analysis of Expectation Propagation-Based Iterative Detection and Decoding

This paper proposes a novel framework for extrinsic information transfer (EXIT) chart analysis to examine the convergence behavior of iterative detection and decoding (IDD) algorithms based on expectation propagation (EP), specifically addressing the extrinsic value exchange mechanism through moment matching (MM). An essential element of IDD algorithm design is the mitigation of self-noise feedback by extrinsic value exchange between the symbol detector and the channel decoder across iterations. This study seeks to compare the extrinsic value exchange mechanisms of established turbo equalization and emerging EP-based IDD, both theoretically and numerically, to elucidate their differences. Turbo equalization utilizes the extrinsic log-likelihood ratio (LLR), whereas EP-based IDD functions in the symbol domain through MM. Extensive simulations of multiple-input multiple-output (MIMO) signal detection indicate that EP-based IDD provides superior bit error rate (BER) performance compared to turbo equalization, especially in demanding scenarios involving higher modulation orders and increased spatial multiplexing loads. To provide theoretical support for these findings, we extend the classical EXIT analysis to include symbol-domain operations. The proposed framework analytically demonstrates that the enhanced detection performance of EP-based IDD arises from the exchange of extrinsic values in the symbol domain, in contrast to conventional turbo equalization, which functions in the LLR domain.