Low complexity polynomial receivers for downlink CDMA

In downlink CDMA, linear multiuser receivers such as the MMSE receiver are often difficult to implement, especially when the spreading factor is large. A trade-off consists in approximating the inverse of the received signal covariance matrix by a polynomial in this matrix. In this paper, a simple algorithm for computing the coefficients of this polynomial for large spreading factors is proposed. The situation where the channel is frequency selective and the users have possibly different powers is considered. Code matrices are modeled as isometric random matrices or as random matrices with i.i.d. elements and the so called Free Probability theory is used.