The rapid demand for broadband wireless access with fast multimedia services initiated a vast research on the development of new wireless systems that will provide high spectral efficiencies and data rates. A potential candidate for future generation wireless systems is multi-carrier code division multiple access (MC-CDMA). To achieve higher user capacities and increase the system data rate, various multiple-input multiple-output (MIMO) technologies such as spatial multiplexing and spatial diversity techniques have been proposed recently and combined with MC-CDMA.This research proposes a chip level coded ordered successive spatial and multiuser interference cancellation (OSSMIC) receiver for downlink MIMO MC-CDMA systems. As the conventional chip level OSIC receiver [1] is unable to overcome multiple access interference (MAI) and performs poorly in multiuser scenarios, the proposed receiver cancels both spatial and multiuser interference by requiring only the knowledge of the desired user's spreading sequence. Simulation results show that the proposed receiver not only performs better than the existing linear detectors [2] but also outperforms both the chip and symbol level OSIC receivers. In this work we also compare the error rate performance between our proposed system and MIMO orthogonal frequency division multiple access (MIMO OFDMA) system and we justify the comparisons with a pairwise error probability (PEP) analysis. MIMO MC-CDMA demonstrates a better performance over MIMO OFDMA under low system loads whereas in high system loads, MIMO OFDMA outperforms MIMO MC-CDMA. However if all users' spreading sequences are used at the desired user receiver, MIMO MC-CDMA performs better than MIMO OFDMA at all system loads.In the second part of this work, user grouping algorithms are proposed to provide power minimisation in grouped MC-CDMA and space-time block code (STBC) MC-CDMA systems. When the allocation is performed without a fair data rate requirement, the optimal solution to the minimisation problem is provided. However when some fairness is considered, the optimal solution requires high computational complexity and hence we solve this problem by proposing two suboptimal algorithms. Simulation results illustrate a significantly reduced power consumption in comparison with other techniques.