Power prediction on broadband channels

The fading envelope in mobile communications causes rapid variations of the receiving conditions. Fast radio resource allocation and planning require accurate predictions of the changing received power. To form a prediction of the fading power, the individual taps of the channel are predicted and their squared magnitude are summed. The dynamics of the taps can be modeled as subsampled autoregressive or ARMA-processes, leading to the different linear predictors studied. The prediction performance is evaluated on measured mobile radio channels at 45 locations with a bandwidth of 6.4 MHz and Doppler frequencies in the range 30-110 Hz. With the proposed predictors the normalized mean square error for the power prediction is below one percent for prediction ranges up to 2.5 ms, for a majority of the measured channels.