Abstract-2-step maximum likelihood block signal detection employing QR decomposition and M-algorithm (QRM-MLBD) can significantly improve the bit error rate (BER) performance of single-carrier (SC) transmission while reducing the computational complexity compared to maximum likelihood detection (MLD). In 2-step QRM-MLBD, unreliable symbol candidates are removed by performing minimum mean square error based frequency-domain equalization (MMSE-FDE) prior to QRM-MLBD. However, a large number M of surviving paths is still required in the M-algorithm to achieve a BER performance close to the matched filter (MF) bound. In this paper, to remedy this problem, we introduce a detection ordering to 2-step QRM-MLBD. We will show by computer simulation that the use of detection ordering can achieve the same BER performance as 2-step QRM-MLBD while reducing the computational complexity.
Keywords-component; Single-carrier, near maximum likelihood detection, MMSE-FDE, QR decomposition, M-algorithm
I.INTRODUCTION The broadband wireless channel is composed of many propagation paths with different time delays and therefore, the channel becomes severely frequency-selective [1]. The minimum mean square error based frequency-domain equalization (MMSE-FDE) can improve the bit error rate (BER) performance of single-carrier (SC) transmission with a low computational complexity [2,3]. However, a big performance gap from the matched filter (MF) bound [4] still exists due to the presence of the residual inter-symbol interference (ISI) after FDE [5]. Although the maximum likelihood detection (MLD) [6] is the optimum signal detection, its computational complexity is extremely high.Maximum likelihood block signal detection employing QR decomposition and M-algorithm (QRM-MLBD) [7] was proposed for SC transmission over a frequency-selective fading channel [8,9]. QRM-MLBD can achieve a BER performance close to the MF bound with less complexity than MLD. However, its computational complexity is still high. In order to reduce the computational complexity, we proposed 2-step QRM-MLBD [10,11] which discards unreliable symbol candidates by performing MMSE-FDE prior to QRM-MLBD. However, a large number M of surviving paths is required in the M-algorithm to achieve a BER performance close to the MF bound. In this paper, to remedy this problem, we introduce the detection ordering. The idea of detection ordering was proposed for multiple-input multiple-output (MIMO) system based on the observation that channels have different conditions [12,13]. However, in SC block transmission, the