Adaptive reduced -rank linear equalization for the forward link in wideband DS -CDMA

Samina Chowdhury, Purdue University


This thesis deals with synchronous Direct-Sequence CDMA transmission using orthogonal channel codes in frequency selective multipath, motivated by the forward link in 3G CDMA systems. The chip-level MMSE estimate of the (multi-user) synchronous sum signal transmitted by the base, followed by a correlate and sum, has been shown to perform very well in heavily loaded systems, compared to the conventional RAKE receiver. We present a reduced-rank, low-complexity approximation of the full-rank, chip-level minimum mean-squared error (MMSE) based on the multistage nested Wiener filter (MSNWF), introduced by Goldstein and Reed. We show that, with perfect channel knowledge, only a small number of stages is needed to achieve near full-rank BER performance over a practical SNR range. We then implement a symbol-level adaptation that trains the equalizer coefficients on the code-multiplexed pilot symbols after despreading. Adaptive MSNWF operating in a low rank subspace exhibits very good convergence characteristics with low sample-support and yields error rates comparable to RLS, and significantly superior to RAKE. An important advantage of the MSNWF is that it can be implemented in a “lattice-structure” which involves much less computation than RLS. ^ We also introduce structured MMSE equalizers that exploit estimates of the multipath arrival times and the underlying channel structure. Due to the sparseness of wideband CDMA multipath channels, the channel vector lies approximately in a subspace spanned by a small number of columns of the pulse shaping filter convolution matrix. The data vector is projected onto this low-rank subspace to construct a structured equalizer. We demonstrate that the performance of these structured low-rank equalizers is much superior to unstructured equalizers in terms of convergence speed and error rates. We then extend our investigation of adaptive MMSE equalizers to time-varying channels, and show that the structured low-rank MMSE equalizer is capable of adequately tracking time-variations in the fading channel for normal vehicular speeds. ^




Major Professor: Michael D. Zoltowski, Purdue University.

Subject Area

Engineering, Electronics and Electrical

Off-Campus Purdue Users:
To access this dissertation, please log in to our
proxy server