UC Santa Barbara

To adapt with current trends in wireless data consumption, providers need to increase capacity by orders of magnitude. Millimeter (mm) wave communication has the potential for delivering this boost through unprecedented levels of spatial reuse. High non-line-of-sight propagation loss at small wavelengths produces a highly localized interference pattern, enabling dense deployment of mobile access points each serving tiny "picocells", 10's of meters in diameter. On the other hand, small wavelengths allow packing massive antenna arrays in reasonable form factors that can be mounted on street fixtures such as lampposts, and are able to form very narrow beams that provide significant spatial isolation. We first consider the problem of backhaul support for the dense deployment of picocellular access points. As wired backhauling is not scalable, we propose a multihop mesh network of high-bandwidth directional point-to-point links at mmWave or THz frequencies to connect picocellular access points to the existing wired gateways, and develop a framework for joint optimization of routing and link scheduling for maximizing backhaul throughput. Scaling array sizes, on the other hand, poses serious challenges both in signal processing and in hardware implementation. As arrays grow large, their beams becomes narrow and accurate user tracking is required to maintain a beamformed link. Current mmWave hardware offers only analog RF beamforming capacity, and most systems do not maintain phase coherence from one packet to another. We therefore propose a noncoherent compressive channel estimation framework with logarithmic overhead scaling that relies solely on magnitude measurements. As technology matures, the realization of fully digital beamforming with one RF chain per antenna element will enable the synthesis of multiple simultaneous beams on a single frontend. Hardware impairments are often more serious for wideband communication at higher frequencies, and a systematic framework for quantifying their impact is crucial. We develop such a framework for analyzing the effect of phase noise on mmWave multi-user massive MIMO, scaled using a tiled architecture. Our analysis provides a cross-layer design tool which can be employed by hardware designers to determine allowable masks for the phase noise power spectral density for different circuit components.