Assessment of Error Bounds for Path Loss Prediction Models for TV White Space Usage in Ekiti State, Nigeria

Abstract

Path-loss propagation models are useful in radio communications for the prediction of signal's coverage area, link's design and power budget. They are equally used for radio channel characterization to accurately predict television coverage, interference analysis and ensure coexistence between the primary service providers and secondary users (through frequency re-use). One of the challenges with the application of a predictive path loss model for any environment other than the one it was developed for is the issue of high prediction errors. This is due to their high dependence on environmental complexity and terrain. In this paper, we investigated the error bounds of eight empirical path loss models to evaluate their reliability of predicting path losses on the UHF band in Ekiti State, South West Nigeria. Measurement of the signal strength for the UHF channel 41 (631.25MHz), Television Broadcasting Station at Ado-Ekiti, was carried out via major routes spanning through the Urban and Suburban areas of the State, using the station as reference. The signal strength values were converted to path losses and compared with predictions of eight selected models. The prediction error, relative error, root mean square error (RMSE), spread corrected mean square error (SC-RMSE), skewness and the normalized error probability density function metrics was calculated to determine the error bound which was used to validate the best predictive model for the routes under consideration. The results of this investigation show that no single model gives an accurate prediction consistently based on the evaluating metrics. However, the Electronic Communication Committee (ECC) 33 model provides better values for the overall metrics considered with RMSE values of 8.48 dB and 9.62 dB (between it and measured values) for Ekiti Suburban and Urban routes respectively. Therefore, optimizing ECC 33 model will bring the RMSE values to the standard acceptable range for both sub -urban and urban routes. The significance of this finding is that ECC 33 model has the least prediction error compared to other selected models and by extension the closest value to the measured values. This validates it to be suitable for the prediction of path losses on the UHF band over the study area.

Statistical tuning of cost 231 Hata model in deployed 1800mhz GSM networks for a rural environment

“…The success of the mobile communication industry can be traced to the development of propagation models [10]. In this research work, the ECC-33 model and COST 231 Hata model were used to compare the path loss values obtained from measurement.…”

“…The ECC empirical formulation is a path loss prediction model developed by the Electronic Communication Committee (ECC) based on the original measurement data from Okumura model in Tokyo, Japan [10]. The ECC group extrapolated the original data by Okumura and modified its assumptions.…”

“…The ECC group extrapolated the original data by Okumura and modified its assumptions. This model is valid for the frequency band of 700MHz to 3500MHz [10]. This model presented the Path loss ( ( ) ) equation based on four factors in [11] as given in equation (1):…”

Application of Levenberg-Marguardt Algorithm for Prime Radio Propagation Wave Attenuation Modelling in Typical Urban, Suburban and Rural Terrains

“…There exist ample propagation loss models in literature (e.g, Ikegami model [7], Hata model [8], Okumura model [9], Walfisch and Bertoni model [10], COST 231 Hata model [11], etc), which have been developed for point-to-point and point-to-area predictive analysis and simulation of signal attenuation losses in wireless communication systems. However, these models' efficacies are typically valid only for a target propagation environment with similar antenna arrangement and specific frequency range [12][13][14][15][16][17][18][19]. One key way to improve on signal propagation loss predictive modelling process is by means of numerical optimisation algorithms based modeling.…”

“…The Gauss-Newton algorithm, is a distinctive numeric minimization algorithm with an iterative application procedure [19]. To apply it, we commence by to providing some initial guess values for the parameter vector of interest  , The Gauss-Newton algorithm only converges iteratively to the desired global minimum if the chosen initial guess values are to some degree near the final solution.…”