Jinwen Liu

Date of Award

Spring 2020

Document Type

Open Access Dissertation


Electrical Engineering

First Advisor

David Matolak


Next generation communication systems will be designed to be faster, more secure and easier to connect with than current systems. Along with the concept of internet of things (IoT), many more devices will be required to communicate with each other. In the case of aeronautical vehicles and systems, in addition to current navigation and surveillance systems, more data links will be required for multiple applications, such as photography, inspections, and entertainment. Current aviation frequency bands will likely be unable to support all proposed services. Apart from air-to-ground (AG) communication links, airport surface terrestrial links and satellite-to-air links (SA) are also of research interest. Most AG communication systems operate in L-band (960-1164 MHz) and below, with a few operating in C-band (5030-5150 MHz). Bandwidth provided by these frequency bands is limited, and will be unable to meet demands for future applications. Hence higher frequency bands in the millimeter wave range (30-300 GHz) are being actively investigated, aiming to fully utilize the much larger available bandwidths. Since millimeter wave (mmWave) signals behave somewhat different from lower frequency signals in AG, SA and terrestrial links, more work is needed to characterize mmWave channels in terms of tropospheric attenuation, path loss, obstacle attenuation, and wideband multipath fading and Doppler effects.

In this dissertation, we investigate and model the tropospheric attenuation for AG and SA links, and model path loss and obstacle attenuation for terrestrial channels, with focus on aviation applications. Some wideband terrestrial channel measurement and modeling is also included. We utilize the tropospheric attenuation empirical model developed by the International Telecommunications Union (ITU) and quantify the effect of the type of precipitation data input on mmWave channel attenuation. Variability of tropospheric attenuation over the long term is also investigated for rain and cloud attenuation in particular, i.e., we investigate extreme rainy and foggy cases, since mmWave signals are so susceptible to these attenuations. Our findings quantify the differences in tropospheric attenuation model outputs with different precipitation data inputs: we find that differences can be substantial in terms of the percentage of time a given attenuation value is exceeded. Frequencies of 30, 60 and 90 GHz are investigated for terrestrial and short AG links, and frequencies 30 and 45 GHz for AS links, for four different climate types: temperate, subtropical, tropical, and rainforest. Results show that in 1 km terrestrial or AG links, local measured rain data input increases mean rain attenuation by 0.5-2 dB over results when ITU’s regional empirical rain data is input. Fog attenuation may increase by 8 dB at 90 GHz in the same comparison. In AS links, mean rain attenuation increases by 0.8 and 1.1 dB at 30 and 45 GHz, respectively, using local measured data input. Rain attenuation has a larger probability of occurrence at moderate-to-significant rain attenuation values: for example, at 90 GHz, 20 dB rain attenuation occurs at most 0.02% of time with ITU’s input data, but occurs an order of magnitude more often (0.2% of the time) with local measured input data.

For path loss, we employ measurements in several settings, including a small airport building, and compare with ray tracing simulations. Multipath components are simulated via ray tracing software Wireless Insite, to obtain channel impulse responses, from which path loss and delay dispersion (e.g., root-mean-square delay spread (RMS-DS) were estimated. We compare the ray-tracing results with measurements for both narrowband signals and wideband signals of bandwidth 500 MHz. The characterization includes path loss and delay spread, and the mmWave results employ directional antennas. We provide preliminary channel characterization for several indoor channels and an outdoor channel at frequencies of 5, 30 and 90 GHz. Comparing our measured path loss results with free space path loss, mean path loss difference are 2.47, 2.72 and 0.31 dB for 5, 30 and 90 GHz, respectively, in indoor channels. For the widely used “close-in” reference distance path loss model, comparing simulation and measurement in 90 GHz channels, differences in model slope versus distance for simulation and measurement are less than 0.2, and standard deviation of large scale fading is less than 1.8 dB. These differences are less than 0.2 and 2 dB at 30 GHz, and less than 0.4 and 1.8 dB at 5 GHz. For large scale fading, the Generalized Extreme Value (GEV) distribution appears to describe excess path loss the best, instead of the commonly used Gaussian distribution. The Kolmogorov-Smirnov (KS) goodness of fit test statistic for GEV is 3% less than that for the Gaussian, for an example 90 GHz indoor channel. Small scale fading was also investigated for a densely sampled 5 GHz line of sight indoor office channel. The Lognormal distribution was found as the most accurate fit among tested distributions.