A steepest descent algorithm is used to update the adaptive weights of a two-stage synchronous Code-Division Multiple-Access (CDMA) receiver that was proposed recently. An issue of the adaptive CDMA system - the convergence and stability property of the receiver is investigated in this thesis.

This adaptive synchronous CDMA receiver uses a decorrelator at the first stage and adopts a neural network which acts as an interference canceler at the second stage. It can achieve near-optimum performance. Furthermore, its computational complexity is just a square function of the number of users. The only requirement is the knowledge of the users' signature sequences.

The analysis shows that the algorithm for the adaptive weights is convergent and straightforward in implementation. The guaranteed fast convergence of the receiver weights and the tractable theoretical analysis on it, as revealed in this thesis, make this adaptive receiver a promising approach for wireless communications.