Abstract: There are several key issues of high resource consumption and poor real-time performance in the existing field programmable gate array (FPGA) platform for multiple-input multiple-output (MIMO) channel emulation. Based on the non-stationary discrete channel model with sum of sinusoids structure, an efficient iterative method is proposed in this paper to generate massive complex exponential signals in real time. By introducing a fixed-point error compensation algorithm and time-division structure, the accuracy of the output channel fading is guaranteed and the hardware resource consumption is further optimized. For single channel emulation, the resource consumption is significantly reduced 6.64% and 3.64% compared with the traditional LUT and CORDIC method respectively. According to 3GPP standard Extended Vehicular A Model (EVA) scenario, measurement results have shown that the statistical properties of the output channel, i.e., power delay profiles (PDP) and Doppler power spectral density (DPSD) are consistent well with the theoretical results. Thus, it can be applied on the real-time performance analysis, design verification of massive channel emulation.