Distributed antenna systems with randomness

In a cellular distributed antenna system (DAS), distributed antenna elements (AEs) are connected to the base station via an offline dedicated link, e.g. fiber optics or line-of-sight RF. Distributed antennas have been recently shown to provide considerable gains in coverage and capacity, at much lower cost than decreasing cell size. Previous studies have neglected the key sources of randomness in such systems, notably (i) random channel effects (fading and shadowing) and (ii) the random quantity and locations of both the mobile users and the AEs. Typically, path loss has been the focus, and the AEs are assumed to be regularly spaced, both of which are significant idealizations. First, we develop an analytical framework that allows random channels to be accommodated. We use this approach to show that selection transmission (using a single AE) is preferable to maximum ratio transmission (which uses all the AEs) in a multicell environment. Interestingly, the opposite is true in an isolated cell. Second, since AEs are placed opportunistically (on tall structures with backhaul access) rather than regularly, we develop a stochastic geometry-inspired approach to determine the outage probability as a function of the number of randomly placed AEs, which we model as a point process. With selection transmission, the outage probability is shown to decrease exponentially with the number of AEs and users. In the most general setup - with multiple distributed antennas and users, and both AE selection and user selection - we show that randomly deployed AEs provide nearly the same performance as regularly spaced AEs.