Hermitian Splitting Precoding (HSP) for Wireless MIMO Systems: All The Channel Capacity, Half The Signaling Cost

摘要

In wireless MIMO systems, capacity-achieving precoding schemes, be they linear (e.g. Singular-Value-Decomposition (SVD) based transmit beamforming) or not (e.g. Tomlinson-Harashima Precoding, THP) necessitate full channel state information (CSI) at the transmitter, thereby requiring the entire channel gain matrix to be fed back from the receiver to the transmitter (in what is known as Closed-Loop MIMO). Contrarily, schemes that do not require such a feedback (Open-Loop MIMO) tend to perform poorly in MIMO systems when the number of antennas is large. One such Open-Loop scheme is QR-based Decision Feedback Equalizer (QR-DFE). In the MIMO Digital Subscriber Lines (DSL) literature, QR-DFE is known to be optimal whenever the channel matrix is diagonally-dominant, a property that holds true in coupled DSL cables. Alas, unlike the MIMO DSL channel, the wireless MIMO channel does not exhibit diagonal dominance. In this work, we propose a joint transmit-and-receive precoding scheme whereby the MIMO channel matrix is transformed into a diagonally-dominant form, thereby making QR-DFE optimal. The proposed transmit precoder is based upon a skew-Hermitian matrix, which has to be fed back from the receiver. Owing to the symmetric nature of such matrix, the CSI feedback signaling cost of the proposed scheme is nearly half the cost of conventional capacity-achieving Closed-Loop schemes.