A study of two equalization techniques for sparse multipath channels

Sparse multipath channels are wireless links commonly found in communication systems such as terrestrial broadcasting, underwater acoustic and cellular land mobile. Their impulse responses are characterized by a few significant multipath terms that are widely separated in time. With high speed transmission, the length of a sampled sparse channel can reach hundreds of symbol intervals, although the majority of taps in the sampled channel are near zero-valued. Classical equalizers thus become too complex for tackling these channels, as their complexity is either a linear or an exponential function of the sampled channel length. In this thesis, two low-complexity techniques are suggested for solving this problem.; First, nonuniformly spaced tapped-delay-line (TDL) equalizers (NU-Es) are examined. Analytical expressions that explicitly indicate the tap values and tap positions of Infinite-length, symbol-spaced (T-spaced) TDL equalizers for sparse multipath channels are derived, and simple design rules for allocating taps to finite-length, T- and T/2-spaced, minimum mean square error (MMSE) NU-Es are formulated based on the derived results. This design-rule-based method demonstrates a better trade-off between accuracy and efficiency than existing tap allocation schemes. The resultant NU-Es also achieve a lower overall computational complexity than conventional, uniformly spaced TDL equalizers (U-Es) of the same span for both directly adaptive and channel-estimate-based implementations. The approach can be extended to design NU-Es in receive diversity systems.; The second solution is a turbo equalizer that utilizes a parallel-trellis framework for both its maximum a posteriori (MAP) equalizer and decoder. The framework allows a sparse multipath channel to be equalized with a bank of low-state trellises. With prefiltering, a broad range of sparse multipath channels can be tackled, including nonminimum-phase ones. Analogously, the framework is suitable for a class of convolutional and turbo codes with sparse generator polynomials. Use of these codes results in a low-latency, memory-efficient turbo equalizer, as the encoding structure of the codes partially integrates the interleaving operation, thus simplifies the interleaver design. Two generalizations of the parallel-trellis turbo equalizers to multiple-input-multiple-output (MIMO) systems have also been considered.