Comparison of Partial CSI Encoding Methods in Multi-User MIMO Systems

Abstract

In this paper, we discuss different algorithms that can be used to encode channel state information (CSI) in realistic multi-user multiple-input multiple-output (MIMO) systems where there are only few users experiencing similar propagation conditions and the mobile user receivers do not necessarily have the same number of receive antennas. We divide systems with CSI encoding into four classes: time-division multiplexing (TDM) with and without linear pre-coding, and multiple user scheduling with and without linear pre-coding. The practical aspects such as system's complexity and approaches for transmitting the CSI feedback and rate information from the mobile receivers to the base station are discussed and compared for different bit rates in the feedback link. We show that significant increases of the mean throughput of the multi-user scheduling systems demand much higher feedback link bit rates than TDM solutions. We also demonstrate that, while optimum, the non-linear precoding systems may introduce unacceptable degree of complexity into the base station design while linear pre-coding offers a very good trade-off between performance and complexity.

Influence of CSI Feedback Errors on Capacity of Linear Multi-User MIMO Systems

“…After selecting the best matching VQ entries, their indices are encoded using concatenation of error-detecting and errorcorrecting codes and transmitted to the base station. If the mobile user is scheduled by the base station, its received signal is demodulated using some form of linear processing based on the SVD results [7], [8].…”

“…Once retrieved, the indices are then used to select entries from two independent codebooksB and S [7], [8]. The first codebook consists of power-independent n T × n T matrices of eigenmodes, mutually orthogonal in case of single-user transmission and non-orthogonal in case of multi-user transmission.…”

“…As shown in [6], [8], the preferred approach in TDM MIMO is to let a receiver directly request quantized matricesB(I B ) andŜ(I S ) based on estimated H k . Having the received indices from all K users, the base station estimates the capacity of their respective channels and, in the next transmission epoch, activates the user with the highest one.…”

“…The second method uses the CSI feedback link to transmit the information about the downlink channel from the downlink receiver back to the downlink transmitter. The problem has attracted attention of the scientific community and papers like [2]- [8] provide solutions for different kinds of systems requiring explicit CSI feedback.…”

Untitled

“…[11]- [14] with channel estimation and imperfect feedback), but the basic principles appear in different forms from different disciplines (information theory, signal processing, communications theory and techniques, adaptive antennas, etc.). However, there are practical shortcomings with each type of optimized LPD design.…”