Generalized RAKE receivers for the CDMA downlink

Code Division Multiple Access has been chosen as the air interface for the worldwide third generation (3G) cellular standards. Interference in CDMA systems is a well known problem and, interference suppression and multiuser detection techniques in such systems have therefore attracted much interest in the recent years. While the main focus has been on the uplink, recently the problem of low complexity downlink interference suppression has gained attention. One of the reasons for this is the fact that 3G standards promise very high data rates in the downlink.;Due to physical limitations at the mobile station, only receivers that require low implementation complexity can be viable. This thesis addresses the problem of interference suppression in the CDMA downlink using RAKE receivers. We first describe the CDMA downlink system and explain why simple RAKE receivers (that have complexity almost as low as that of conventional RAKE receivers) with appropriately chosen taps can indeed suppress interference effectively. Prior work in this area exploits the fact that the users' spreading codes at the base station are orthogonal and uses RAKE receivers that equalize the downlink channel. We show that such an approach is suboptimal in principle, and derive RAKE receivers based on new design criteria that yield significantly better performance when orthogonal short codes are used at the base station.;We then extend these principles to practical, orthogonal long code systems where the spreading codes vary every symbol period and prove that channel equalization turns out to be the optimal approach when a constraint that the same RAKE weights be used for all symbol periods, is imposed (due to complexity reasons). We then suggest a very simple way to implement such a receiver in practical systems. We also consider a quasi-orthogonal random code model for the spreading code and show that interestingly, the optimal receiver turns out to be the same as that for long code systems that uses orthogonal codes. Simulation results are provided to illustrate the performance difference between various receivers for each spreading code model. Suggestions for future work are also included.