Novel pre-processing/post-processing for enhanced performance of time-varying, multi-user MIMO-OFDM systems

Chad C Lau, Purdue University

Abstract

Within the general framework of Orthogonal Frequency Division Multiplexing (OFDM), specific challenges arise when facing scenarios such as time-varying channels, multi-user co-channel interference, and MIMO data separation. This research addresses each of these three topics. The first section considers the performance and complexity of a MMSE estimator based on matrix inversion when applied to equalization for OFDM in a time-varying, Doppler-spread multipath environment. Reduced complexity is achieved via a Chebyshev polynomial-based preconditioner applied to the iterative method of conjugate gradients (CG); this preconditioner is shown to substantially reduce the number of iterations required to closely approximate the solution obtained with the computationally complex MMSE estimator based on matrix inversion. The second section deals with equalization and interference cancellation for a multi-user MIMO-OFDM system. A major issued addressed is how one estimates the channel in the presence of co-channel interference even with training symbols. This research proposes a direct-training, per-tone technique which exploits a novel insight that the equalizing weight vectors lie in a low-dimensional, data-independent subspace. MMSE estimation is employed but dimensionality conditions are dictated by Zero-Forcing analysis. The third section proposes a new space-time signaling scheme for MIMO-OFDM using complementary filters derived from the rows of the DFT matrix. The autocorrelation properties of the complementary filters allows multiple complex data signals at the transmitter with an arbitrary number of antennas to be perfectly separated and reconstructed at the receiver without prior channel knowledge while achieving full-rate transmission.

Degree

Ph.D.

Advisors

Zoltowski, Purdue University.

Subject Area

Electrical engineering

COinS