MU-MIMO Channel Estimation For LTE Uplink

Long term evolution(LTE) technology, based on orthogonal frequency division multiplexing(OFDM)and multiple input multiple output(MIMO), is regarded as the most popular technology in evolution from 3G to 4G. Since LTE Uplink, using single carrier frequency division multiple access(SC-FDMA) scheme, supports multi user MIMO(MU-MIMO) scenario, channel estimation for MU-MIMO is very important. In this paper, based on channel estimation under SISO scenario, channel estimation under MU-MIMO scenario for LTE uplink is studied.Firstly, LTE uplink physical layer simulation platform based on MATLAB is designed. The platform consists of transmitter part, receiver part and channel part, support MU-MIMO scenario and multiple channel models. This platform is used for verification of computation complexity and estimation accuracy for channel estimation, through BLER performance simulation.Secondly, reference signal design and channel estimation technology for SISO scenario are analyzed. The analysis of reference signal consists of introduction of Zadoff-Chu sequence, cyclic time shift, cyclic shift hopping and DMRS pattern. The study of channel estimation technology focus on interpolation estimation, least squares(LS) estimation, transform domain estimation and minimum mean square error(MMSE) estimation. Least squares(LS) estimation is often used as the first step of the joint estimation algorithm, as its principle and realization are relatively simple, but it is affected by the noise a lot. Interpolation estimation, using CFR of reference signal REs to calculate CFR of data REs, needing no other known channel information, consists of linear interpolation estimation, Gauss interpolation estimation and so on. There are two kinds of basic transform domain estimation method used in LTE, which are DFT-based transform domain estimation and DCT-based transform domain estimation. Since DFT-based transform domain estimation can directly use the FFT module in OFDM system, it is used more widely. Because minimum mean square error(MMSE) estimation uses the two order statistic properties of the channel, its performance is better than other estimation methods, whereas its computation complexity is the highest. There are three common MMSE algorithms, which are MMSE algorithm, simplified MMSE algorithm and SVD algorithm.Based on channel estimation algorithm for SISO scenario, reference signal design and channel estimation algorithm for LTE uplink MU-MIMO scenario are investigated. An improvement of existing reference signal is introduced, which uses cyclic shift with maximum distance to decrease the interference between different cyclic shift. It can meet requirement of application of MU-MIMO as well as be compatible to old release. Channel estimation algorithm is a joint estimation algorithm based on DFT-based transform domain estimation, using dynamic noise threshold and frequency domain window function to decrease MSE introduced by CIR energy leakage and frequency domain Gibbs phenomenon. By using the LTE uplink physical layer simulation platform based on MATLAB, simulation result of above improved estimation algorithm is compared with basic DFT-based transform domain estimation algorithm. The simulation curve shows, the improved DFT-based transform domain estimation algorithm has realized its expected function, and the using of dynamic noise threshold and frequency domain window function can increase the accuracy of channel estimation.