Abstract

Since the discovery of the significant gain in capacity provided by multiple antennas over fading channels, tremendous research and development efforts in academia and industry have been invested to multiple-input-multiple-output (MIMO) technology. However, difficulties still exist in the design of flexible MIMO transmission schemes. The overall goal of our study is to develop efficient turbo MIMO transceivers that are capacity-achieving and yet with reasonable complexity. First, we study the design of serially concatenated MIMO transmitters. For simpler design and flexible rate-versus-performance tradeoff, conventional encoders are used before a linear space-time modulator. A joint iterative receiver based on the turbo principle is assumed that precludes the use of Tarokh's design criteria for such a concatenated system. Extending the extrinsic information transfer (EXIT) charts to MIMO systems, design criteria that concern both the data rate and error performance are developed for the inner space-time (ST) modulator. Based on the new design criteria, an optimal space-time linear dispersion modulation scheme is presented. In addition, the tradeoff between constellation size and symbol rate for a given data rate is discussed. Simulation results are provided to verify the new design criteria and to demonstrate the merits of the proposed coded space-time modulation. For such a turbo MIMO transmitter, the abundance of highly developed outer decoders allows us to focus on the study on the inner multi-user detection (MUD). To start, a new parallel interference cancellation (PIC) MUD scheme is proposed. This detector can be used as a stand-alone multi-user detector in a non-iterative MIMO receiver. The new scheme employs a novel nonlinear minimum mean square error (MMSE) estimator as the soft decision device. The nonlinear estimator exploits the knowledge of symbol alphabet for refined estimates with smaller mean square error (MSE). Simulation results demonstrate that the proposed detector significantly outperforms the conventional detectors with comparable complexity. The nonlinear MMSE estimator is further extended to develop the soft-output MUD for iterative turbo MIMO receivers. The nonlinear estimation makes use of a priori information from the outer decoder and preliminary estimates gleaned from channel observation as well. With the new nonlinear MMSE estimator, a generalized SIC-MMSE scheme is proposed. Depending on the methods for generating preliminary estimates, two generalized SIC-MMSE detectors are proposed: a 2-staged SIC-MMSE and a recursive SIC-MMSE detectors. Analytical study and simulations are carried out to demonstrate the merits of the two proposed detectors. Last, tradeoffs among the components of the transmitter that affect error performance at the receiver are discussed to provide design guidelines for practice