Spectrally Efficient Frame-Format-Aided Turbo Equalization With Channel Estimation

Chained turbo equalization (CHATUE) has been recently recognized as a low-complexity frequency-domain turbo equalization technique that eliminates the necessity of transmitting the cyclic prefix (CP), and hence allows for spectrally efficient signaling in wireless communications. However, two issues arise from the original version of CHATUE (referred to as CHATUE1) as a consequence of eliminating the CP, which are the noise enhancement and the latency due to the time-concatenated structure. This paper proposes a new version of CHATUE (referred to as CHATUE2) to solve the noise enhancement problem. CHATUE2 retrieves the circulant structure of the channel matrix, originally inherent within the CP transmission, by utilizing composite replica signals that combine the received signals and the soft reference signals replicated from the log-likelihood ratio (LLR) fed back from the decoder. For this purpose, this paper determines the optimal combining ratio based on the minimum-mean-square-error (MMSE) criterion. CHATUE2 is hence able to achieve an improvement in bit-error-rate (BER) performance over CHATUE1. In addition, this paper provides a solution to solving the latency problem by making a practical assumption on the training sequence (TR) transmission, which is required to perform channel estimation generally, in practical systems. Furthermore, this paper proposes a new channel estimation technique, i.e., chained turbo estimation (CHATES), which improves the spectrum efficiency and asymptotically achieves the Cramér-Rao bound (CRB). CHATES assumes that the TR length is exactly equal to the channel impulse response (CIR) length, although the conventional technique requires twice as long or even longer TR lengths. Numerical results show that CHATUE2 with CHATES achieves a gain of 1 dB over conventional turbo equalization with a CP at a BER of 10^-5 in realistic propagation scenarios represented by channel sounding measurement data and in model-based - requency-selective fading channels.