Soft detection and decoding in wideband CDMA systems

Abstract

A major shift is taking place in the world of telecommunications towards a communications environment where a range of new data services will be available for mobile users. This shift is already visible in several areas of wireless communications, including cellular systems, wireless LANs, and satellite systems. The provision of flexible high-quality wireless data services requires a new approach on both the radio interface specification and the design and the implementation of the various transceiver algorithms. On the other hand, when the processing power available in the receivers increases, more complex receiver algorithms become feasible.

The general problem addressed in this thesis is the application of soft detection and decoding algorithms in the wideband code division multiple access (WCDMA) receivers, both in the base stations and in the mobile terminals, so that good performance is achieved but that the computational complexity remains acceptable. In particular, two applications of soft detection and soft decoding are studied: coded multiuser detection in the CDMA base station and improved RAKE-based reception employing soft detection in the mobile terminal.

For coded multiuser detection, we propose a novel receiver structure that utilizes the decoding information for multiuser detection. We analyze the performance and derive lower bounds for the capacity of interference cancellation CDMA receivers when using channel coding to improve the reliability of tentative decisions.

For soft decision and decoding techniques in the CDMA downlink, we propose a modified maximal ratio combining (MRC) scheme that is more suitable for RAKE receivers in WCDMA mobile terminals than the conventional MRC scheme. We also introduce an improved soft-output RAKE detector that is especially suitable for low spreading gains and high-order modulation schemes. Finally we analyze the gain obtained through the use of Brennan's MRC scheme and our modified MRC scheme.

Throughout this thesis Bayesian networks are utilized to develop algorithms for soft detection and decoding problems. This approach originates from the initial stages of this research, where Bayesian networks and algorithms using such graphical models (e.g. the so-called sum-product algorithm) were used to identify new receiver algorithms. In the end, this viewpoint may not be easily noticeable in the final form of the algorithms, mainly because the practical efficiency considerations forced us to select simplified variants of the algorithms. However, this viewpoint is important to emphasize the underlying connection between the apparently different soft detection and decision algorithms described in this thesis.