Phase Noise Estimation Algorithm For MIMO System

As one of the key technologies of the 5th generation mobile communication,massive MIMO technology has received extensive attention.However,massive MIMO requires hundreds of antennas and corresponding crystal oscillators in the base station.The phase noise generated by unstable oscillators will seriously affect the performance of the system.Firstly,the MIMO system and phase noise are modeled,then the MIMO channel with phase noise is estimated by least squares method,and the estimated channel information is used for decision-directed and pilot-symbol-aided phase noise estimation.Secondly,in the decision-directed phase noise estimation method,this paper firstly proposes the method of estimating the phase noise by the unscented Kalman filtering method,and smoothing the estimated phase noise by the RTS smoothing filtering method.We iterate the two processes of data symbol decision and phase noise estimation,and then compare the extended Kalman algorithm,the unscented Kalman algorithm and the unscented Kalman RTS smoothing algorithm,and analyze their performance in different antenna numbers,different phase noise variances and different modulation modes.The experimental results show that when the phase noise is strong,the unscented Kalman filter can obtain better phase noise estimation than the extended Kalman filter,and the unscented Kalman RTS smoothing algorithm has the best estimation effect,but when the phase noise is very strong or the modulation constellation is dense,the hard input unscented Kalman RTS smoothing algorithm may leading to a vicious circle between the data symbol decision and phase noise estimation,which seriously deteriorate the MIMO performance.Thirdly,in the pilot-symbol-aided phase noise estimation method,we periodically insert the pilot into the data symbols and calculate the phase noise at the pilot.Then we estimate the phase noise at the data symbols by using the smoothing filtering method.We derive the Wiener filtering algorithm,and apply the unscented Kalman RTS smoothing algorithm and the unscented Kalman linear smoothing algorithm to track the phase noise with pilot symbols aided and then compare these methods in different ways.The experimental results show that when the phase noise is little strong,the performance of the unscented Kalman RTS smoothing algorithm is close to the unscented Kalman linear smoothing algorithm and they all perform better than the Wiener filtering algorithm.It should be noted that the pilot-aided method is not particularly sensitive to the frequency of pilot insertion.Finally,we focus on the decision-directed Kalman RTS smoothing algorithm and the pilot-assisted unscented Kalman RTS smoothing algorithm to compare and analyze the problems that may occur in these phase noise estimation algorithms.The experimental results show that,if we all use hard input method,the pilot-aided method has a better performance than the decision-directed method,and is very close to the ideal scene of decision-directed method,that is,the perfect decision feedback.