“Turbo Transceivers for MIMO Wireless Communications”

Abstract The discovery of the Turbo codes has driven research on the creation of new signal detection concepts that can, in general, be referred to as the Turbo approach. Recently, this approach has made a drastic change in creating signal detection techniques and algorithms such as equalization of inter-symbol interference (ISI) experienced by broadband single carrier signaling over mobile radio channels [1]. A goal of this lecture is to provide the participants with broad views and knowledge of the Turbo concept-based Multiple-Input Multiple-Output (MIMO) signal transmission techniques by explaining how the techniques were developed in various applications, and how they improves performances. 1. Turbo Equalization with Soft Cancellation and MMSE Filtering It has long been believed that the computational complexity needed for the ISI equalization in broadband single carrier signaling is prohibitive, belief of which this lecture shall overturn. The recent advance in signal processing techniques has created a new flexible and robust iterative (Turbo) signal detection framework based on soft cancellation and minimum mean squared error (MMSE) filtering [2], [3], which is referred to as SC/MMSE in this lecture. This lecture is started by briefly introducing the SC/MMSE technique as a general signal processing structure, and it applies the technique to the equalization of channels suffering from severe ISI. This lecture then focuses on MIMO signal detection techniques using SC/MMSE, of which resulting device is called the SC/MMSE MIMO Turbo equalizer [4], [5]. Figure 1 shows a block diagram of the SC/MMSE MIMO Turbo Equalizer for single carrier signaling. The SC/MMSE MIMO Turbo detector performs joint channel estimation, multiple stream signal detection, and decoding of channel codes, all in an iterative manner. The SC/MMSE MIMO Turbo equalizer forms soft replicas of CCI from other users’ signals and ISI on the desired signal to be detected using a priori Log-Likelihood Ratios (LLRs) output by each user’s Soft-Input Soft-Output (SISO) channel decoder. The soft CCI and ISI replicas are subtracted from the received signal vector, of which process is refereed to as soft cancellation. MMSE filtering that follows the soft cancellation aims at suppressing the residual interference components remaining at the output of the soft canceller. A posteriori LLR value for each data stream is calculated from the MMSE filter output, and after de-interleaving, it is brought to each user’s SISO decoder. The SISO decoders update the LLR values user-by-user, and then feed them back to the equalizer part, where soft estimates of CCI and ISI are formed using the updated LLR. The entire process is repeated. Because of the Turbo principle, iterations achieve performance gain. It is shown in this lecture that the SC/MMSE MIMO Turbo equalizer can asymptotically achieve the equivalent performance to the maximum likelihood sequence estimation (MLSE) single user detector, even though the SC/MMSE equalizer’s complexity is a cubic order of the total path number in the equalizer’s coverage. The lecture then introduces several approximation methods [6],[7] of SC/MMSE to further reduce the complexity, which do not cause any significant performance losses. Some key approximations finally allow the Proceedings of ISAP’04, Sendai, JAPAN