On the design of practical reduced-rank DS-CDMA receivers
Francisco R. Rubio, Xavier Mestre · 2006
A class of linear interference-suppression schemes is proposed that maximize the empirical SINR at the output of a DS-CDMA receiver under unknown multiuser interference con- ditions. We build upon well-known reduced-rank MVDR/MMSE solutions based on the Krylov-subspace spanned by the covari- ance matrix of the observations and the spreading sequence of the desired user. Only the effective signature of the intended user is assumed to be known at the receiver side. The optimum coefficients of the resulting polynomial expansion receiver are obtained by approximating the output SINR in the asymptotic regime defined when both the processing gain and the number of observations grow together without bound at the same rate. I. INTRODUCTION Many current mobile radio systems are supported on air- interfaces based on the code-division multiple access (CDMA) scheme, typically in its direct-sequence (DS) implementation. In the last two decades, a huge amount of publications has been dedicated to the theory of multiuser detection dealing with the reception of signals over these channels. In particular, much attention has been given to the linear minimum mean- square-error (MMSE) and decorrelator receivers, as practical alternatives to the highly-complex optimal (finite-alphabet- constrained) maximum likelihood (ML) detector. However, even though they have been proved to perform considerably more efficiently than the conventional single-user matched- filter (MF) receiver, their use tends to be avoided in realistic scenarios. First, the complexity associated with these detection schemes is still prohibitive in situations with a large number of users. On the other hand, the solution following the multiuser detection approach is based on the knowledge of the spreading sequences of all users as well as information about their channels and the background noise level. In some scenarios, like for example in the forward link, it is unrealistic and certainly impractical to consider tracking all this amount of information. In these situations, the detection task is better approached from a point of view of multiple access inter- ference (MAI) suppression (1), (2). The linear interference suppression filter minimizing the mean squared error (MSE) is equivalent to the solution maximizing the output signal- to-interference-plus-noise-ratio (SINR), which in the literature