A Study of 2D Non-Stationary Massive MIMO Channels by Transformation of Delay and Angular Power Spectral Densities

In this paper, we propose a transformation method to model space-time-variant (STV) two-dimensional non-stationary wideband massive multiple-input multiple-output (MIMO) channels. This method enables us to obtain the STV joint probability density function of the time of arrival and angle of arrival (AOA) at any time instant and antenna element of the array from a predefined configuration of the scatterers. In addition, we introduce a simplified channel modeling approach based on STV parameters of the AOA distribution and demonstrate that key statistical properties of massive MIMO channels, such as the STV temporal autocorrelation function and Doppler power spectral density, can be derived in closed forms. As examples of applications, we study multiple array-variant properties of three widely-used geometry-based stochastic models (GBSMs): the Unified Disk, Ellipse, and Gaussian scattering models. Furthermore, we present numerical and simulation results of the statistical properties of these three GBSMs and compare them with those obtained using the conventional spherical wavefront approach. We point out possible limitations of the studied channel models to properly represent massive MIMO channels.