The Study Of Massive Mimo Based Channel Estimation And Diversity Multiplexing Tradeoff

Massive MIMO is a technology studied by researchers during three years. Massive MIMO system works in a TDD mode, and with the help of huge number of communication antennas, it can achieves spatial diversity gain and multiplexing gain, thus improves the spectrum efficiency of the system.However, in order to make the biggest use of Massive MIMO, we have to make sure that enough channel state information is obtained before signal being transmitted, that’s because only with the acknowledge of CSI will we do the power allocation and antenna selection. Meanwhile, diversity gain increases by sacrificing the multiplexing gain, and vice versa, this feature makes it very important to analyze the balance of these two gains.In this paper, we discussed the channel estimation problem of Massive MIMO, and proposed an integrated Bayesian estimation method that combined uplink measure and downlink measure. This method considers that the source of pilot pollution for uplink and downlink is totally different, thus the reciprocity is no longer valid, so we take a combination of both direction measurement. As a channel will stay static during a coherent interval, the channel vector between the target base station(or interfering base station) and the target terminal will also keep unchanged. By the use of this characteristic and Bayesian channel estimation, we can get the channel vector for both target and interfering base station. This method will greatly reduce the impact of pilot pollution, and the channel vector we get can contain information of interference which will provide help for future’s base station cooperation. Besides, this paper discussed the diversity-multiplexing tradeoff (DMT) in limited SNR scene. We get the outage probability of the system by using the ergodic capacity distribution of a particular channel model, and then the DMT curve under different multiplexing gain definition. This result reveals that the DMT distribution under limited SNR converges slower than the ideal curve,which means it needs a SNR offset。That characteristic inspires us the performance of a massive MIMO system should not just be judged by the diversity gain and multiplexing gain, SNR offset should also be considered. Simulation results (by MATLAB) prove the correctness of these conclusions.