Implementation & Performance Analysis of a Non-codebook Based MU-MIMO System with Dynamic Precoding for TDD LTE-Advanced

Abstract

This dissertation presents a non-codebook based Multiple User Multiple Input Multiple Output (MU-MIMO) test-bed system for Time Division Duplex (TDD) Long Term Evolution-Advanced (LTE-A). The system was implemented to verify the performance of a dynamic precoding procedure that selects the most appropriate precoder among a given set of precoding techniques. Using two parameters, the propagation path gain and the condition number of the channel matrix, the proposed MU-MIMO system can adaptively switch the precoding software to guarantee that the upper bound Bit Error Rate (BER) is maintained with a minimum computational burden. This dynamic precoding technique is applied to GPU modems, which comply with 3GPP 36 series release 10 for parts other than the Dynamic precoding block. This dissertation also presents a novel procedure for accurately calibrating the multiple Radio Frequency (RF) paths of the multiple antennas and RF transceivers of the MU-MIMO system. In addition, a synchronization method for time synchronization between eNB and terminal and synchronization between threads constituting the system was presented. From various experimental measurements obtained from the implemented test-bed system, the upper bound BER, which was arbitrarily set to 10-3 in this dissertation, can be maintained with the simplest precoder, Zero Forcing (ZF), unless the propagation path gain becomes less than 0.25. It was also found in the experimental tests that, as the distance between 2 handsets becomes shorter than 1 meter, which causes the condition number of the channel matrix to be larger than 17, the precoder should be switched from Lattice Reduction (LR) to Tomlinson-Harashima Precoding (THP) when the minimum distance between base-station and each handset maintained is set to at least 6 meters. In conclusion, the dynamic precoding presented in this dissertation maintains the MU-MIMO performance of small cell eNB in various environments. The dynamic precoding procedure can expand the cover eNB and enable efficient management, reducing unnecessary user selection and scheduling burn caused by performance degradation.