Development and experiment of linear and non-linear precoding on a real-time multiuser-mimo testbed with limited CSI feedback

This paper presents the development and experimentation of linear precoding (LP) and non-linear precoding (NLP) on a real-time multiuser-multiple input multiple output (MU-MIMO) testbed subject to limited channel state information (CSI) feedback. LP techniques such as zero-forcing and minimum mean square error precoding have already been experimented and the related results are available in the literature. However, they are not capacity-achieving. NLP techniques such as vector perturbation and Tomlinson-Harashima precoding (THP), on the other hand, achieve in theory the capacity of MU-MIMO, but such claims have not been verified in practice, to date. One reason is that theoretical NLP works rely on the assumption of full and perfect CSI at the transmitter as NLP techniques are sensitive to CSI error. The merit of this work is to experimentally confirm (through hardware implementation and pre-field test experimentation) the feasibility of NLP subject to limited CSI feedback using the fading simulator. Experimental results reveal that the testbed achieves less than 2 dB and 3 dB degradation of the block error rate compared to the floating point computer simulations in LP and NLP, respectively.