Bridging he Gap between linear and non-linear precoding in small- and large-scale MIMO downlinks

We propose a performance-complexity tradeoff for vector perturbation (VP) precoding in the downlink of multiuser multiple input multiple output (MU-MIMO) systems, that bridges the gap between linear and non-linear channel-inversion type precoding. To this end, we introduce a VP design that applies a performance-determined threshold to the desired norm of the preceded signal, to reduce the search through candidate perturbation vectors by the sphere encoder. Once this threshold is met, the search for the perturbation vectors finishes thus saving significant complexity at the transmitter. Towards future largescale MIMO systems, we test t he proposed tradeoff in systems with increasing transmit antennas is fixed physical space. While it is known that linear precoding becomes near-optimal in the massive MIMO region when ignoring transmit correlation, a) we show that t here still remains a performance gap between linear and non-linear precoding once realistic antenna array deployment is considered and b) we utilize the proposed scalable approach to bridge this gap. The presented analysis and results show that t he sum rates achieved are within 90% of those for conventional VP at less than half the computational complexity.