On Improvement of Channel Estimation for the Uplink of Large Scale MU-MIMO using DMRS

TDD large scale MIMO recently has emerged as a groundbreaking technology which enables many antennas can be equipped at a Base Station without overhead for downlink pilot. In such systems, the uplink channel estimation plays a critical role since it is needed for both MU-MIMO detection and precoding weight calculation. To facilitate the implementation, the channel estimation should be as low complex as possible. In addition, to support multiusers, the channel estimation should be robust to long delay spread. Previous methods usually consists of FFT/IFFT transformation and windows for separating channel impulse response of multiusers. Another method estimates channel transfer function of each user directly in frequency domain by using a simple sliding window, however, this method is susceptible to long delay spread. To remove FFT/IFFT transformation, and mitigate long delay spread, we propose a method for channel estimation in frequency domain using raised-cosine based filter. Simulation results shown that the proposal outperforms previous methods. Under 3.7 μs delay spread of ITU-B channel, the proposal achieves bit error ratio of 10^-3 with 2dB additional carrier to noise ratio compared to perfect channel estimation. To further validate the proposal, experiments are conducted with 16 antennas at base station and 4 single antenna users using 16, 64, and 256 QAM.