106 Compaison Between Known Popagation Models Using Least Squaes Tuning Algoithm on 5.8 GHz in Amazon Region Cities Buno S. L. Casto, Mácio R. Pinheio, Gevásio P. S. Cavalcante Fedeal Univesity of Paá (UFPA), Augusto Coea Avenue S/N, ZIPCODE: 66075-900, Belém-Paá-Bazil, e-mails:buno@ufpa.b,macio.pinheio@gmail.com and gevasio@ufpa.b Igo R. Gomes, Oziel de O. Caneio Univesity Cente of Paá (CESUPA), Govenado José Malche Avenue, Belém-Paá-Bazil, e- mails:uiz.igo@gmail.com and ozielcaneio@gmail.com Abstact This pape pesents a pefomance compaison between known popagation Models though least squaes tuning algoithm fo 5.8 GHz fequency band. The studied envionment is based on the 12 cities located in Amazon Region. Afte adjustments and simulations, SUI Model showed the smalle RMS eo and standad deviation when compaed with COST231-Hata and ECC- 33 models. Index Tems 5.8 GHz band, Amazon Region, Linea Least Squaes, Popagation models. I. INTRODUCTION Since the constant incease of the wieless netwoks, studies of signal popagation ae needed to ensue an efficient Pe-Poject Stage in coveage and quality of sevices. This pape pesents a study of the signal popagation in 5.8 GHz on Amazon egion cities. A pefomance compaison between known popagation models is made fo an Amazon Region envionment. The least squaes tuning algoithm has been used to adjust the models to the measuements. It is impotant to emembe that the tems elated to eception and tansmission heights in the models equations have been left unchanged. Although the models adjustments, diffeences in how the models wok with eception and tansmission height have influence in RMS eo and standad deviation which ae the metics adopted in this wok. This pape is oganized as follows. In section II is pesented explanations about the envionment and the data acquisition. In section III a desciption of the popagation models is made. In section IV the least squaes tuning algoithm is pesented. In section V simulations and esults ae shown and finally, section VI shows the conclusions. II. ENVIRONMENT AND DATA ACQUISITION The collected data have been caied out in 12 cities on Paá State at Amazon Region, Bazil. These cities ae known by thei woodland envionments. The vegetation nomally appeas mixed with the esidential and commecial constuctions esulting in a single medium. An example of Amazon egion
107 city is shown in Fig.1. Fig. 1. Aeial view of Santaém city in Paá State, Bazil Diffeent of the taditional measuing campaigns [1]-[2] that ae made with continuous data collection in a mobile unit, this data acquisition has been caied out by taking the punctual RSSI (Received Signal Stength Indicato) in 335 fixed clients installed in 12 cities that have been contemplated with the Digital Inclusion Paá State Govenment Poject named NavegaPaá [3]. The poject consists of WLL (Wieless Local Loop) netwoks installed in the cities, binging boadband access and multimedia sevices. It is inteesting to analyze this collected data because fixed clients have diffeent distances with espect to thei Base Stations and diffeent installation heights. Fom the collected RSSI it can be found the path loss fo each client by using values of tansmission powe, tansmission gain and eception gain. The pocess fo obtaining the distances between the clients and base stations is based on the coodinates that was collected duing the implantation stage of these netwoks. III. PROPAGATION MODELS The popagation models used in this pape ae COST231-Hata, SUI Model and ECC-33 model whose have efeence in some pefomance compaison woks [4]-[5]-[6]. A. Stanfod Univesity Inteim (SUI) Model SUI Model has had in you development the Stanfod Univesity paticipation. Vaiables involved in model pediction pocess ae adopted fo fequencies below 11 GHz. It is inteesting to evaluate model pefomance fo this case because SUI Model employs teain popeties on its equations so the base fo calculating the popagation loss can be accomplished in an non-ideal way diffeent of the fee space equation method.
108 The base of the popagation model and the envionment chaacteization ae epesented by the following equations [7]: d L = A + 10γ log X f + X h d + Fo d > d0 (1) 0 4πd A = 20log 0 (2) λ c γ = a bht + (3) h t f X f = 6log 2000 (4) h X h = 10.8log fo teains type A and B 2 (5) Whee: d - Link distance, m d0 - Initial distance, 100 m λ - Wavelength, m f - Fequency, MHz ht - Tansmitte height, m h - Receive height, m h X h = 20log fo teain type C (6) 2 Paametes, e chosen accoding to Table I: TABLE I. TERRAIN TYPE PARAMETERS Model Paamete Type A Type B Type C a 4.6 4 3.6 b 0.0075 0.0065 0.005 c 12.6 17.1 20 Table I is based on teain types defined in [7]. B. COST 231 - Hata This one is an extension of Okumua-Hata Model. It was made to embace a fequency ange fom 1500 MHz to 2000 MHz. The popagation loss obtained can be calculated though the following equation: ( f ) 13.82 log( hte ) a( he ) + ( 44.9 6.55 log( hte )) ( d ) cm L = 46.3 + 39.9 log log + (7) a ( h ) = ( 1.1log( f ) 0.7) h ( 1.56 f 0.8) e e fo small and medium cities (8) ( h ) 3.2( log( 11.75 )) 2 e h 4. 97 a fo lage cities (9) = e
109 Whee: f - Fequency, MHz d - Link distance, m h - Tansmitte height, m te h - Receive height, m e c - 0 db fo soft and sububan aeas and 3 db fo dense uban aeas m C. ECC-33Model ECC-33 is a model fom Electonic Communication Committee based on analysis in 3.4 and 3.8 GHz band. The path loss is obtained fom de following equations [4]: L = A + A G G (10) fs bm b A fs ( d ) 20log( f ) = 92.4 + 20log + (11) ( d ) + 7.894log( f ) 9.56( log( )) 2 A bm = 20.41+ 9.83log + f (12) G And fo medium city envionments, hb = log ( 13.958 5.8 log( d )) 2 (13) 200 b + Whee: d - Link distance, m f - Fequency, GHz hb - Tansmitte height, m h - Receive height, m ( 42.57 + 13.7 log( f ))( log( h ) 0.585) G (14) = IV. LEAST SQUARES ALGORITHM Due to the diffeent chaacteistics of the envionment whee the models have been made, a tuning poceeding is needed to adjust model paametes to the measued data. Least Squaes (LS) citeion is useful fo linea adjustment cases. In this situation, the algoithm is epesented by the idea of minimizing the sum of the squaes of the diffeences between measued data and pedicted data. These diffeences become an eo function expessed as follows: Whee: - Eo function - Numbe of total used data - Measued data - Pedicted data i= 1 ( ) 2 E = Y i L i (15) n
110 The distance and fequency tems in the models equations wee adjusted by the algoithm, howeve, the tansmission and eception heights tems wee not included in least squaes tuning. Moe details about LS algoithm applied in tuning method ae descibed in [1]-[2]. V. SIMULATIONS AND RESULTS Simulations have been done consideing the mean and specific installation heights of the clients located at the 12 cities in study. The data obtained in the simulations ae shown in Fig. 2-5. Fig. 2. Popagation models pefomance using mean eception heights of the clients Fig. 3. Tuned popagation models pefomance using mean eception heights of the clients
111 Afte simulations, the obtained values of RMS eo (db) and standad deviation (db) fo all thee models ae shown in the Table II, befoe and afte tuning. Models TABLE II. RESULTS FOR MEAN INSTALLATION HEIGHT RMS Eo Befoe Standad Deviation Befoe RMS Eo Afte Standad Deviation Afte SUI 14.66 6.60 6.25 4.47 COST231-Hata 11.29 5.54 6.25 4.47 ECC-33 28.03 7.64 6.24 4.43 Fig. 4. Popagation models pefomance using specific eception heights of the clients Fig. 5. Tuned Popagation models pefomance using specific eception heights of the clients
112 Fo the specific client heights, the obtained esults ae shown in the Table III, fo the RMS eo and standad deviation as well. Models TABLE III. RESULTS FOR SPECIFIC INSTALLATION HEIGHT RMS Eo (Befoe) Standad Deviation (Befoe) RMS Eo Afte Standad Deviation Afte SUI 15.26 6.88 7.22 4.84 COST231-Hata 17.27 9.88 15.51 10.97 ECC-33 31.32 10.67 11.12 6.99 Fom the esults in Table II, it is seem that SUI, COST231-Hata and ECC-33 models, each the same RMS eo (6.2 db) when mean eception height is used in least squaes tuning. In the othe hand, when specific client installation height was used fo tuning pocess, SUI Model obtained the best impovement with a RMS eo of 7.22 db and COST 231-Hata had the wost one equal to 15.51 db. RMS eos have obtained a maximum impovement about 20 db (ECC-33 Model) and a minimum impovement aound 2 db (COST231-Hata Model). The mino standad deviation value belongs to SUI Model. Results ae elevant because RSSI collecting pocess has been pefomed in peculia site-specific clients. Vaiations in models pedictions, fom Fig. 4 and Fig. 5, ae justified because each client has a specific CPE (custome pemises equipment) installation height. RMS eo (RE) and standad deviation (SD) values fo all 12 cities in study ae shown in Table IV. TABLE IV. RESULTS FOR SPECIFIC INSTALLATION HEIGHT Cities SUI Model COST231-Hata Model ECC-33 Model RE SD RE SD RE SD Abaetetuba 7.05 4.24 27.14 10.31 10.95 6.45 Altamia 5.81 3.93 20.26 9.64 10.61 7.61 Bacaena 10.64 6.71 33.48 14.63 11.91 6.86 Itaituba 7.34 4.89 30.26 10.14 10.24 6.42 Jacundá 5.21 3.04 35.27 10.30 7.11 4.02 Maabá 7.14 4.80 26.25 17.77 14.67 9.04 Pacajá 5.25 2.69 33.95 7.62 9.11 5.02 Ruópolis 4.50 2.78 30.45 8.59 7.44 4.70 Santaém 8.86 6.21 22.88 13.18 13.37 7.80 Tailândia 8.08 4.96 37.05 11.22 10.36 6.51 Tucuí 7.87 6.33 24.65 10.70 11.42 7.62 Uuaá 4.46 2.43 31.75 5.34 6.77 3.78
113 VI. CONCLUSION In this pape, a pefomance compaison between COST231-Hata, Stanfod Univesity Inteim (SUI) and ECC-33 models is made fo an Amazon Region envionment. At the final pefomance evaluation, SUI Model has shown a bette behavio than COST231-Hata and ECC-33 Models. Based on the obtained esults, a poposal fo futue woks can conside an adjustment of SUI Model by changing some paametes o adding a tem which is elated to some new envionment featue. It is also foeseen an adjustment in SUI model fo path loss pediction in mobility conditions. Fo such a pupose, measuement campaigns will be caied out. ACKNOWLEDGMENT Authos would like to thank the Paá State Data Pocessing Company (PRODEPA) fo accessing impotant infomation to the wok development. This wok was suppoted by CNPq unde covenant 573939/2008-0(INCT-CSF) and FAPESPA / UFPA / FADESP/SEDECT, Nº. 067/2008. REFERENCES [1] M. Yang, W. Shi, Linea Least Squae Method of Popagation Model Tuning fo 3G Radio Netwok Planning, Fouth Intenational Confeence on Natual Computation ICNC, Jinan, 18-20 Octobe 2008, pp. 150-154. [2] G. R. Palladó, On DVB-H Radio Fequency Planning: Adjustment of a Popagation Model Though Measuement Campaign Results, Maste s Thesis, Depatment of technology and Built Envioment, Univesity of Gävle, Sweden, 2008. [3] NavegaPaá Poject. Available in: http://www.navegapaa.pa.gov.b/ [4] V. S. Abhayawadhana, I.J. Wassell, D. Cosby, M.P. Sellas and M.G. Bown, Compaison of Empiical Popagation Path Loss Models fo Fixed Wieless Access Systems, IEEE 61 st Vehicula Technology Confeence VTC, Stockholm, 30 May-1 June 2005, pp. 73-77. [5] J. Milanovic, S. Rimac-Dje, K. Bejuk, Compaison of Popagation Models Accuacy fo Wimax on 3.5 GHz, 14th IEEE Intenational Confeence on Cicuits and Systems ICECS, Maakech, 11-14 Decembe 2007, pp. 111-114. [6] T. Schwengle, M. Gilbet, Popagation Models at 5.8 GHz - Path Loss & Building Penetation, IEEE Radio and Wieless Confeence RAWCON, Denve, 10-13 Decembe 2000, pp. 119-124. [7] V. Eceg, Channel Models fo Boadband Fixed Wieless Systems, IEEE 802.16 Boadband Wieless Access Woking Goup, Califonia, 2000.