Novel Adaptive Equalization Techniques for a Transmit Diversity Scheme

Space-time block coding (STBC) has added a new dimension to broadband wireless communication systems. Applications such as wireless Internet access and multimedia require the transmission of high data rates over frequency selective fading channels. The reliability of the wireless communication system can be increased by using diversity techniques combined with an equalizer at the receiver to eliminate the inter-symbol interference caused by multipath propagation. Generalizing Alamouti's famous STBC method to frequency selective channels, Time Reversal-Space Time Block Coding (TR-STBC) was first introduced in [1] and has since been shown to be an effective transmit diversity scheme [2, 3, 4]. TR-STBC-based schemes are considered promising candidates for indoor transmission [5] as well as for the enhanced data rates of the global evolution (EDGE) system [2, 3]. The optimal equalizer for a TR-STBC-based transceiver is the Maximum Likelihood Sequence Estimator (MSLE), realized using the Viterbi algorithm. Unfortunately, a Viterbi equalizer is difficult to implement in real-time due its exponential increase in complexity with the number of antennas and the length of the channel impulse response. Thus, we consider an adaptive algorithm-based Decision Feedback Equalizer (DFE). Such a DFE requires only linear processing complexity while maintaining good performance. Theoretically, the two output streams of a 2 x 1 TR-STBC decoder are uncoupled in terms of the input signal statistics and uncorrelated in terms of the channel noise statistics. The standard approach to removing the inter-symbol interference from these streams is to use either two parallel independently-adapted Single-Input Single-Output (SISO) equalizers or to use a single Multiple-Input Multiple-Output (MIMO) equalizer. By exploiting the common second-order statistics of the two output streams, we proposea novel hybrid equalizer structure which shares the statistical information between two SISO equalizers while constraining them to have common tap weights. To accommodate various levels of performance versus computational complexity, we propose novel Least Mean Square (LMS), Normalized Least Mean Square (NLMS), and Recursive Least Squares (RLS)-based adaptive algorithms for this new equalizer architecture. We use both statistical analysis and Monte Carlo simulations to characterize the dynamic convergence of these algorithms and to compare our new structure with the conventional uncoupled SISO equalizers and fully-coupled MIMO equalizer. We show that our new equalizer outperforms the other two equalizers using a reduced computational complexity similar to the uncoupled SISO equalizers. As expected, with increasing complexity, we find that the novel RLS-based algorithms converge the fastest followed by the novel NLMS- and LMS-based algorithms. We also consider alternative packet structures and kick-start methods to increase the convergence speed and reliability of the equalizer at realistic complexity. Finally, adding multiple receiver antennas to our system, we extend our equalizer structures and algorithms to the 2 x NR case. Using analysis and simulations, we demonstrate that the added receiver diversity in this case yields even greater reliability