Path-Loss characteristics of urban wireless channels

Wireless channel data was collected in Cambridge, Massachusetts for diverse propagation environments over distances ranging from tens of meters to several kilometers using mobile 2.4-GHz transmitters and receivers. The 20-MHz bandwidth signals from eight individually movable van-top antennas were Nyquist sampled simultaneously with 12-bit accuracy. Although path-loss variance for any given link length within single residential/urban neighborhoods was large, single streets typically exhibited path-loss, L(dB) = -10 log₁₀(P _r/P_t) ≅ 10αlog₁₀r ± C, where P is the received or transmitted power, r the link-length, the street-dependent path-loss coefficient, and C the loss incurred at street intersections. Measurements yielded α ≅ 1.5±3.2β±0.27 for 2<α<5;β is the fraction of the street length having a building gap on either side. Experiments over links as short as 100 meters indicate a 10-dB advantage in estimating path loss for this model compared to optimal linear estimators based on link length alone. Measured air-to-ground links were well modeled by α =2 for the elevated LOS path, and by stochastic log-normal attenuation for the ground-level scattering environment. These models permit path-loss predictions based on readily accessible environmental parameters, and lead to efficient nodal placement strategies for full urban coverage.