A Method Of MUD Based On Self-tuning Step Size LMS Algorithm In CDMA System

Multi-Access Interference(MAI),multi-path fading and near-far effect are the predominant factors that determine the performance of receivers in Code Division Multiple Access(CDMA) system, but MAI is not inevitable. MUD can can reduce MAI effectively, relax near-far effect dramatically, improve the system performance obviously and increase the system capacity. Conventional techniques of MUD need many apriori knowledge(address code of each users, relative time delay ,power and parameters of channels etc.),these parameters are difficult to get in time-varing and multi-path fading channels. In recent years, many research activities have been focused on blind adaptive multiuser detection, because it doesn't need training sequence,but spreading code and transmitting delay of desired users,furthermore it can track time-varing mobile communicatin environment.The main works of this paper can be summarized as follows:(1) CDMA systems were introduced, theory of spreading frequency communication and characters of spreading sequence were analyzed and pointed out essential reason of MAI and MUD is the key technique of 3G . Conventional MUD and blind adaptive MUD techniques were summarized and analyzed, as well as their advantages and disadvantages were listed.(2) Beginning of representative form of linear MUD, the essential of blind adaptive multiuser detection was discussed. Making cost function ,subjected some linear terms, maximum/minimum, was their goal and optimal solutions to tap-weight were resolved. The problem and its advantage and disadvatage about MOE blind adaptive multiuser detection in synchronous,no-fading and one path channel were analyzed: Minimum output energy detector has character of global convergence ,but its stable residual mean square error was large,it converged slowly and it was hard to get optimal step size.(3) A method of MUD based self-tuning step size LMS(Least Mean Square ) algorithm was proposed in multi-path fading channel and its deduction was given. This algorithm can overcome the defect of the constant step size LMS algorithm...