3D MIMO-OFDM Channel Estimation

Abstract

In this paper bandwidth efficient pilot design for a MIMO-OFDM downlink is addressed. By exploiting the spatial correlation of signals originating from neighboring transmit antennas, the principle of pilot aided channel estimation (PACE) by two-dimensional (2D) interpolation in time and frequency is extended to the spatial domain; giving rise to three-dimensional (3D) PACE. By invoking the multi-dimensional sampling theorem conditions for regular pilot patterns to attain minimum overhead are derived. Provided sufficient spatial correlation 3D-PACE reduces pilot overhead by 50% compared to conventional 2D-PACE. Unfortunately, spatial correlation may not be universally available: while for outdoor macro-cells, where transmit antennas are typically mounted above rooftop, the assumption of spatially correlated channels is justified, in indoor deployments where spatial correlation tends to be low, spatial interpolation may not be feasible. On the other hand, in indoor environments cell sizes together with mobile velocities are substantially smaller, giving rise to lower Doppler and channel delay spreads. We show that a sophisticated pilot design is able to retain the low pilot overhead of 3D-PACE by exploiting these heterogeneous correlation properties, in the way that high spatial correlation compensates for low correlation in time and/or frequency, and vice versa.