Receiver and parameter estimation techniques for quasi-synchronous reverse link multicarrier CDMA system

Abstract

The performance of a multicarrier code division multiple access (MC-CDMA) system is limited by frequency offsets, insufficient cyclic prefix condition, and multiple access interference (MAI). In a quasi-synchronous reverse link, a receiver is unable to equalize the frequency offsets and the insufficient cyclic prefix due to the independence of each user's channel and frequency offset. As a consequence, conventional detectors cannot give accurate symbol decision even in a single user system, which is free from MAI. The viable approach to deal with such complicated situation is to integrate the MAI suppression with the compensation of frequency offsets and insufficient cyclic prefix. Unfortunately, the existing techniques do not consider all of these problems. They only focus on the system with MAI and frequency offsets or the system with MAI and insufficient cyclic prefix. This dissertation proposes a multiuser detection with intersymbol interference cancellation (MUD-ISIC), which surpasses the existing techniques by the capability to jointly overcome all the concerned problems. Apart from this feature, the MUD-ISIC, which is able to adopt non-linear detection techniques, is more flexible than the existing techniques, which only accept a linear minimum mean square error detector. This dissertation does not only introduce the new receiver but it also proposes practical parameter estimation techniques, including the estimators for transformed signature sequences, ISI generation sequences, and noise variance, to realize the MUD-ISIC. Since the estimation of channel impulse response and frequency offsets has extremely high complexity and is unlikely to be able to implement, the required parameters are directly estimated. From the mean square error analyses, the training sequences, which achieve the minimum mean square error, are proposed. The proposed parameter estimation techniques enhance the attractiveness of this dissertation over other existing works, which only concentrate on the receiver design without clarifying its feasibility. The simulation results demonstrate several relevant merits of the proposed techniques over the other known techniques. The results also confirm the benefits of the optimum training sequences. According to the simulation results, the lowest bit error rate (BER) is achieved in the system with insufficient cyclic prefix, which contrasts to the conventional system design where the sufficient cyclic prefix condition is normally imposed. The MUD-ISIC does not only gain better BER but it also has a potential to offer the higher symbol rate.