Advanced mobile receivers and downlink channel estimation for 3G UMTS-FDD WCDMA systems

The explosive growth of the mobile Internet is a major driver of wireless telecommunications market today. In future years, the number of online wireless users will exhibit strong progression in every geographic region and devices will tend to support multimedia applications which need big transmission data rates and high quality. Third generation telecommunication systems, like UMTS in Europe, aim at rates approaching 2 Mbits/s in particular cellular environments. This thesis makes several contributions to the area of mobile advanced receivers and downlink channel estimation in the context of the physical layer of UMTS-FDD Wideband DS-CDMA systems. Advanced signal processing techniques are applied to increase performance of the terminal receiver by mitigating the distortion caused by the radio propagation channel and the interference introduced by the multiple access to the wireless system, as well as to improve the downlink channel estimation and/or approximation. The conventional receiver for DS-CDMA communications is the RAKE receiver, which is a filter matched (MF) to the continuous-time operations of pulse shape filtering and channel filtering and to the discrete-time operation of spreading. We propose a restricted class of linear discrete-time receivers for the downlink, they have the same structure as a RAKE receiver, but the channel matched filter gets replaced by a discrete-time equalizer filter that is designed to maximize the Signal-to-Interference-plus-Noise-Ratio (SINR) at the output of the receiver. The complexity of the max-SINR receiver is variable and can possibly be taken to be as low as in the RAKE receiver, like in those max-SINR structures that exploit the nature of the overall channel seen by the receiver, i.e. the convolution between the radio propagation channel and the pulse shaping filter. When a mobile terminal is equipped with multiple sensors, a two dimensions (2D) RAKE receiver, i.e. a spatio-temporal channel MF, is traditionally implemented, but also in this case a max-SINR receiver is conceivable and becomes a spatio-temporal MMSE equalizer. Furthermore, a 2D max-SINR receiver allows better suppression of similarly structured intercell interference. The problem of Intercell Interference Cancellation is also addressed, via the use of the unused spreading codes in the system (the Excess Codes) and of the polynomial expansion symbol- and chip-rate approaches. Base Station Transmit Diversity schemes are also studied and compared for the two linear receiver structures, RAKE and max-SINR receivers. Channel estimation and approximation play an important role, this being a critical issue in achieving accurate user-of-interest detection. The RAKE receiver assumes a sparse/pathwise channel model so that the channel matched filtering gets done pathwise, with delay adjustment and decorrelation per path and maximum-ratio combining of path contributions at the symbol rate. We propose and simulate a number of sparse discrete-time channel approximation algorithms along the lines of Matching Pursuit, of which the Recursive Early-Late (REL) approach appears most promising. We also analyze and simulate the effect of channel estimation on the RAKE output SINR. Pilot-assisted channel estimation operates generally on a slot-by-slot basis, without exploiting the temporal correlation of the channel coefficients of adjacent slots. So we consider the estimation of mobile channels that are modeled as autoregressive processes with a bandwidth commensurate with the Doppler spread. Brute pilot-based FIR channel estimates are then refined by Wiener filtering across slots that performs the optimal compromise between temporal decorrelation due to Doppler spread and slot-wise estimation error. We furthermore propose adaptive filtering techniques to implement the optimal filtering before path extraction.