Performance of prediction models in suburban/rural residential areas at 860, 2300 and 3500 MHz

Planning a broadband wireless access (BWA) system like LTE or WiMAX in suburban/rural areas highly depends on the accuracy of pathloss predictions. Due to the wide range of possible frequency bands from 800 MHz up to 3.5 GHz, a lot of different prediction models are applicable. To assess the accuracy of different models, we conducted measurements in a rural area in southern Germany at 860, 2300 and 3500 MHz and compared those to the results of predictions by the COST 231 Walfisch-Ikegami, the Extended Hata and the Stanford University Interim (SUI) model. Assessing the accuracy of the models by the mean and the standard deviation of the error, we have identified the model best suited for predictions at each of the measured frequencies.