Fundamental Limitations of Large Antenna Arrays for Millimeter Wave Systems

Most studies on millimeter wave systems assume that (an asymptotically) large number of antenna elements can be deployed at either the base-station or the user equipment (UE) ends. With a typical analog beam codebook of array steering vectors, a consequence of this assumption is an initial beam acquisition process that has an onerous beam scanning latency and increased power consumption. One way out of this predicament is to use broadened beams with significantly enhanced beamwidths at both ends. With this background, this paper argues that practical considerations such as chip area, cost and complexity of the fabrication and calibration processes motivate the use of only a finite precision in terms of phase shifters and amplitude control. As a result, beam weights that rely on precise phase control in realizing the objective of beam broadening endure a dramatic degradation in array gain performance as antenna dimensions increase. These observations motivate the choice of more moderate antenna dimensions (as a function of the realizable phase and amplitude precision) as optimal from a millimeter wave transmissions standpoint.