Size and array shape for massive MIMO

With massive multi-input multi-output, it may be the case that large numbers of antennas are closely packed to fit in some available space. Here, channel correlations become important and it is of interest to investigate the space requirements of different array shapes. We focus on uniform square and linear arrays and consider a range of correlation models. We show that the benefits of two-dimensional arrays are dependent on the type of correlation. When the correlation decays slowly over small antenna separations then square arrays can be far more compact than linear arrays or they can offer substantial sum rate enhancements. When the correlation decays more quickly, then the main benefit is compactness.