Massive MIMO systems have a special importance among suggested technologies for physical layer of 5G cellular communications which achieves various advantages employing a large number of antennas. However, in case of real implementation, the large number of antennas causes high hardware and energy costs and makes the implementation very difficult. Thus a solution is necessary for reducing costs and energy consumption. One of the features of Massive MIMO systems is robustness to hardware impairments, so recently it has been proposed to reduce the resolution of ADCs in RF chains to decrease costs and hardware complexity. Especially, a mix of ADCs with different bit widths for different antennas can achieve higher energy efficiency and better performance compared to other solutions such as totally one-bit sampling. In this thesis after a brief introduction of Massive MIMO, previous works on architectures consist of a mix of ADCs with different bit widths or Mixed-ADC has been reviewed. Then two-level and multi-level Mixed-ADC architectures have been modeled in a spatially correlated system with a switch between antennas and ADCs and capacity of each model has been derived. Based on derived relations optimization problems for allocating ADCs to antennas in Mixed-ADC architecture has been formulated. Different solutions have been proposed to deal with problems. Finally using simulation, the results of proposed methods has been compared with conventional methods and the significant effectiveness of proposed optimizations has been shown. Key words: Massive MIMO, Analog to Digital Convertor, Mixed-ADC, Spatial correlation, Optimizatio
