This thesis investigates the performance of a multi-user multiple-input single- output (MISO) system considering maximum ratio transmission (MRT) downlink precoding. The transmitted signal from the base station (BS) to each user is as- sumed to experience the double scattering channel. We adopt the minimum-mean- square-error (MMSE) channel estimator for the proposed model. Within this setting, we are interested in deriving tight approximations of the ergodic rate assuming the number of BS antennas, users, and scatterers grow large with the same pace. Under the special multi-keyhole channels, these deterministic equivalents are expressed in more simplified closed-form expressions. The simplified expressions reveal that unlike the standard Rayleigh channel in which the SINR grows as as O(N/k), the SINR associated with a multi-keyhole channel scales as O(S/N). This particularly shows that the K reaped gains of the large-scale MIMO over double scattering channels do not linearly increase with the number of antennas and are limited by the number of scatterers. We further show that the derived asymptotic results match the simulation results closely under moderate system dimensions and provide some useful insights into the interplay between N, K and S.

In this thesis, multiuser receivers for code division multiple-access (CDMA) downlink are studied under frequency selective fading channel conditions. The receivers investigated in this thesis attempt to estimate desired symbol as a linear combination of chip-rate sampled received signal sequence. A common matrix-vector representation of signals, which is similar to the model given by Paulraj et. al. is constructed in order to analyze the receivers studied in this thesis. Two receivers already well known in the literature are introduced and derived by using the common signal model. One of the receivers uses traditional matched filter and the other uses symbol-level linear minimum mean square error (MMSE) estimation. The receiver that uses traditional matched filter, also known as the conventional RAKE receiver, benefits from time diversity by combining the signal energy from multiple paths. The conventional RAKE receiver is optimal when multiple-access interference (MAI) is absent. Linear MMSE based receivers are known to suppress MAI and to be more robust to noise enhancement. The optimal symbol-level linear MMSE based receiver requires inversion of large matrices whose size is determined by either number of active users or spreading factor. These two parameters can be quite large in many practical systems and hence the computational load of this receiver can be a problem. In this thesis, two alternative low-complexity receivers, which are chip-level linear MMSE equalizer proposed by Krauss et. al. and interference-suppressing RAKE receiver proposed by Paulraj et. al., are compared with the linear full-rank MMSE based receiver and with the conventional RAKE receiver in terms of bit-error-rate performance. Various simulations are performed to evaluate the performance of the receivers and the parameters affecting the receiver performance are investigated.