Time series prediction of short term scintillations of Ku band satellite links in Sri Lanka

Scintillation effect is a dominant atmospheric impairment among the various contributing factors for low fade margin communication systems operating on high frequency bands such as Ku. Empirical analysis of scintillations can be done using signal strength measurements which can be treated as a time series prediction problem in forecasting the short-term scintillation effect. This paper discusses the time series based statistical prediction analysis of tropospheric scintillation effect. The experiment was performed under clear sky conditions where the other attenuation contributions such as rain, dense clouds is minimum. The experiment was setup at the University of Moratuwa, Sri Lanka using a Ku band satellite downlink operating on a moderately high elevation angle of 45 degrees. The time series analysis was performed using linear, non-linear, fuzzy and neural prediction techniques. The results show that though the error of prediction is low for a fuzzy predictor, the neural and non-linear predictors are better in analysing the statistical properties of the predictor results.