A Study On Blind MUD Algorithms In DS-CDMA Systems

Multi-Access Interference (MAI) is the primary factor that will determine the performance and capacity of Code-Division Multiple-Access (CDMA) system. To avoid or even eliminate the influence of MAI, Multi-User Detection (MUD) that helps the system perform better and enlarge the capacity of the system can be adopted. However, MUD does not work without using the array of training before coefficient is locked, what is more, constant sending of array of training will cause big waste of spectral resources. Therefore, people turn to the study of the blind self-adaptation that does not need training. In recent years, the study on blind multi-user detection has aroused great attention around the world. This thesis puts focus on blindMUD algorithms combining theory analysis and through simulations based on the current MUD technologies. Following are the main contributions of this thesis:(1) Make analysis and simulations on Signal Interference Rate ( SIR ), Bite Error Rate(BER), minimal output energy and anti-interference performance in different channel environment of LMS algorithm, RLS algorithm and Kalman algorithm of the blind multi-user measures. The computer simulation shows that RLS algorithm and Kalman algorithm are of great convergence performance, stable SIR and accurate tracking in timing channels.(2) As to the problem that MOE detection that of blind MUD LMS algorithm will convergence the overall situation, but stable remaining residue square errors are big and the ideal MMSE detection will not be acquired, author proposes an improving type of blind MUD LMS algorithm and makes analysis and simulation of its SIR, BER and anti-interference performance. The simulation of the computer indicates thatboth the BER performance and the convergence performance of this algorithm are superior to LMS algorithm. The improving blind MUD LMS algorithm keeps MOE's advantage of the convergence of the overall situation and the ideal stable state of SIR output. What is more, this algorithm can availably avoid MAI.(3) The convergence speed of RLS algorithm is quite higher than that of LMS algorithm, but the SIR of RLS algorithm is not stable as well as Kalman algorithm's. An improving type blind MUD RLS algorithm is proposed in this thesis and analysis, study and simulation of its SIR, BER and anti-inference performance are made. The simulation of the computer shows that it is superior to RLS algorithm in the aspect of stability and BER and it works almost as well as Kalman algorithm.