A low complexity MCMC algorithm for MIMO system with bias technique

Abstract

Markov Chain Monte Carlo (MCMC) simulator is proposed as promising method for multiple-input multiple-out (MIMO) system, which provides a good tradeoff between performance and hardware complexity. But the conventional MCMC algorithm needs to calculate conditional distributions in every sampling, where nonlinear exponent calculations are involved and the values of conditional log likelihood ratio (CLLR) have large dynamic range. It also suffer stalling problem at high signal-to-noise (SNR) regimes. In this paper, we propose a low complexity MCMC algorithm based on Max-Log updating. Samples are directly updated in log-domain with small dynamic range of CLLR. A bias technique is also proposed to remedy the stalling issue. Results show that the proposed MCMC detector can reduce 50% complexity with 2 dB gains at high SNR regimes compared with the conventional enhanced MCMC detector. The enhanced MCMC algorithm also outperforms minimum mean square error based on parallel interference cancellation (MMSE-PIC) by 2dB performance gains with 10% less complexity.