Efficient implementation of linear minimum mean square error code division multiple access receivers via recursive filtering

TDMA and CDMA based multiple-access communications systems use guard intervals in time and guard bands in frequency to prevent interference between users, i.e., multiple access interference (MAI). As an alternative, CDMA systems improve spectral efficiency by providing better frequency reuse since adjacent cells often reuse the same frequency allocation. However, CDMA systems are much more prone to intracellular interference and are consequently prone to the MAI problem. “Multiuser” CDMA receiver structures attempt to remove intracellular interference.; An optimal mufti-user detection approach has been developed by Sergio Verdú. However, Verdú's receiver has a complexity that grows exponentially in the number of users, rendering it impractical. At the other extreme is the conventional receiver. The conventional receiver, which consists of a bank of matched filters, is the optimal receiver structure in the absence of MAI. The performance of the conventional receiver suffers in the presence of significant amounts of MAI. Therefore, there is a need for suboptimal receiver structures that do not grow exponentially in complexity while still providing some level of MAI reduction.; Specifically, in this dissertation we investigated the application of adaptive minimum mean square error (MMSE) IIR filtering to the problem of MAI mitigation in direct sequence spread spectrum (DSSS) communications utilizing short signature sequences. A linear noted lattice filter (NLF) configuration consisting of pole and tap parameters was used to minimize the MSE between the output of the NLF and the transmitted data bit for a particular user. There are challenges associated with the natural MSE criterion. One specific challenge, multimodality of the MSE surface, is addressed by appropriate selection of the initial values for the pole parameters.; It was shown that under the presence of maximal MAI conditions the NLF implementation of the MMSE CDMA receiver provided an efficiency advantage compared to an implementation utilizing finite impulse response (FIR) filters. Furthermore, it is well known that MMSE receiver structures exhibit some inherent resistance to the MAI problem. The NLF implementation was also shown to be even more resistant to the near/far problem as compared to FIR MMSE receiver structures proposed in the literature.