Adaptive blind multiuser DS-CDMA downlink equaliser

Introduction: In a DS-CDMA downlink scenario, transmission over a dispersive channel destroys the mutual orthogonality of the codes which are used to multiplex the various users in the system. As a result, the received and code-demultiplexed user signals are subject not only to inter-symbol interference (ISI) owing to channel dispersion but also to multiple access interference (MAI). Consequently the conventional DS-CDMA code-matched filter receiver suffers severe performance degradation, thus motivating the need for better detection strategies. There have been extensive research efforts to introduce a reliable multiuser detector [1], apt to alleviate the effect of both MAI and ISI. While the maximum likelihood (ML) estimation based detector offers the best possible performance [1], its unrealistic complexity renders it unsuitable for downlink applications owing to handset constraints. Alternatively, sub-optimal schemes such as the MMSE multiuser equaliser offer moderate complexity and generally good performance [2]. Recovering several users at the same time in multiuser equalisation exploits more knowledge of the system and has been performed blindly using a constant modulus (CM) criterion [3]. However, in [3, 4] additional orthogonality constraints and mutual decorrelation of the recovered user sequence are required. In this Letter, we derive, based on the definition of a signal model in the following Section, a suitable CM cost function and a stochastic gradient algorithm which is structurally similar to the multiple error filtered-X LMS algorithm in [5]. The addition mutual decorrelation between the various decoded user signals as required in [3] and [4] can be neglected owing to the code-filtering.