Queueing theory approach to rain fade analysis at microwave and millimeter bands in tropical Africa.

Abstract

With an overwhelming demand of larger bandwidth required for high capacity data with content-rich services ranging from high-speed video streaming to multimedia content, there is a continuous need to migrate to higher microwave bands, particularly beyond the regular Ku and Ka bands (between 11 - 40 GHz). The presence of precipitation at these microwave and millimeter bands (3-300 GHz) generally induce rain fade, which is a constraint to network providers intending to achieve optimal service delivery, at acceptable signal to noise ratios (SNRs). In practice, fade countermeasures – static or dynamic – are necessary to combat the consequences of chronic fluctuations of rainfall resulting in signal deterioration and impairment over communication links. However, the implementation of dynamic fade countermeasures is systematically tied upon the available Channel State Information (CSI), which is often timevariant relative to the occurrence of precipitation events. Time-variation of rainfall events are perceptible in measurable rainfall microstructural parameters which vary intensely in space and time. These spatio-temporal variations yield the generation of observable random patterns of signal attenuation during rain events, often in a stochastic manner. To this end, researchers have emphasized on understanding the underlying behaviour of generic rainfall microstructural parameters such as rainfall rate, rainfall Drop Size Distribution (DSD) and radar reflectivity. Therefore, the investigation of these stochastic properties of rainfall processes is primary in the determination of recognisable patterns of rainfall rate and other microstructures. This thesis introduces the queueing theory approach via the Markov Chain technique to investigate the time-varying characteristics of the rainfall process from distrometer data in subtropical and equatorial Africa. Rainfall data obtained from these two climatic locations, at one minute integration time, were processed from sites in Durban, South Africa and Butare, Rwanda, over a specified measurement period. Initial investigation and comparison of rainfall microstructures undertaken at both sites clearly show key differences in their probability distribution profiles at Stratiform-Convective (SC) bounds. The underlying queue discipline of rainfall spikes and their queue metrics are determined and appraised for system performance using rainfall time series database. The results show rain spike generation processes vividly exhibit a First-Come, First- Served (FCFS) semi-Markovian distributed traffic of M/Ek/s discipline, with a varying degree of servers, for different rainfall regimes. Comparison of queue statistics results over different rainfall regimes at the two locations reveal significant differences in their queue metrics and performances. The knowledge obtained from the queue statistics and SC probability analysis are further employed in the determination and classification of rainfall cells, rainfall growth models and path attenuation prediction. The results are compared and validated with data collected from a 6.73km, 19.5 GHz terrestrial link in Durban.