Reliability-Ordered DFE for vector systems

In this paper, detectors are considered for a general class of vector systems. A novel detector, the Reliability-Ordered Decision Feedback Equalizer (RO-DFE), is presented in which the ordering of decisions is done in accordance with reliability metrics helping to mitigate the problem of the error propagation inherent to all DFEs. The proposed structure derives the reliability metrics from an initial linear MMSE equalizer. This yields the major advantage that soft decisions from this equalizer can be reused within the DFE resulting in only a minor overall complexity increase. For an important class of channel independent OFDM pre-coders, the equivalence between a hypothetical detector and one operating on an equivalent AWGN channel is established. This is then used to derive a lower bound on the performance of a Maximum Likelihood (ML) detector in terms of Bit Error Rate for QPSK, and in terms of Symbol Error Rate for M-ary QAM over a maximum diversity Rayleigh fading channel. Simulation results are presented showing significant gains of the proposed detection system over conventional ones. The newly derived ML lower bound for QPSK is also compared to a simulated ML implementation and found to be accurate under good SNR conditions.