Turbo Equalization and EP: A Comparative Analysis of Extrinsic Value Exchange Mechanisms

The key to designing iterative detection and decoding (IDD) algorithms is an extrinsic value exchange mechanism that suppresses self-noise feedback during iterative processing between the symbol detector and channel decoder. This pa-per provides a novel and fair comparison of two IDD algorithms-turbo equalization and expectation propagation (EP), focusing on their extrinsic value exchange mechanisms from both theoretical and numerical simulation perspectives. While turbo equalization generates the extrinsic value in the log-likelihood ratio (LLR) domain by subtracting the detector output LLR from the decoder output LLR, EP-based IDD generates the extrinsic value in the symbol domain via moment matching (MM). Numerical studies demonstrate that EP-based IDD out-performs turbo equalization in terms of bit error rate (BER) performance in multiple-input multiple-output (MIMO) signal detection scenarios, particularly when higher-order modulation is applied. To investigate this phenomenon, we propose an extrinsic information transfer (EXIT) analysis tailored for examining the iterative behavior of EP-based IDD.