EJERS, European Journal of Engineering Research and Science Vol. 2, No. 11, November 2017

Development of Propagation Path Loss Prediction Model for Mobile Communications Network Deployment in Osogbo, Nigeria

Hammed Lasisi, Yinusa A. Adediran, and Anjolaoluwa A. Ayodele

 for short propagation paths. Abstract—Path loss, a major parameter in the analysis and iii. Semi-deterministic model: This is based on both the design of the link budget of a telecommunication system, could empirical models and deterministic aspects [1]. be explained as the reduction in power density of an electromagnetic wave as it travels through space, over a II. PATH LOSS MODELS distance. Path loss prediction models are therefore vital tools in cell planning, cell parameter estimation, frequency assignments Six path loss models are considered in the developmental and interference evaluation. This paper reports on the process of the new model. They are: free space path loss development of a path loss prediction model that describes the model, Okumura-Hata model, COST 231 Hata model, signal attenuation between transmitting and receiving antennas Ericsson model, Egli model and the Lee model. as a function of the propagation distance and other parameters for Osogbo, Nigeria. The model is extensively useful for A. Free Space Path Loss Model (FSPL) conducting feasibility studies for signal prediction, coverage optimization and interference analysis during the initial phase Free space path loss model is the simplest way to model of network planning in the study area and other areas with path loss as all hindrances that might affect signal similar environmental and propagation characteristics. propagation are neglected. Path loss in free space defines how much strength of the signal is lost during propagation Index Terms—Interference; Path Loss; Frequency from transmitter to receiver across a free space devoid of Assignment; Cell Planning; Link Budget. vegetation. FSPL is dependent on frequency and distance, and is given as [2]. I. INTRODUCTION In mobile communication, path loss models are necessary 푃퐿퐹푆퐿푃 = 32.45 + 20푙표푔10푑 + 20푙표푔10푓 (1) for proper planning, interference estimation, frequency assignments, and cell parameter estimation which are basic where PLFSPL is free space path loss (in dB), f is frequency for network planning process as well as Location Based (MHz), d is distance between transmitter and receiver (m). Service (LBS) techniques that are not based on Global B. Okumura-Hata Model Positioning System (GPS) system. Estimation of the path loss is usually termed prediction. Exact prediction is Okumura Hata model is based upon an extensive series of possible only for simpler cases, such as free space survey and measurements made in and around Tokyo city in propagation or the flat earth model. For practical cases the 1968 between 150MHz and1500MHz. Depending on the path loss is calculated using a variety of approximations area of interest, Okumura Hata path loss approximations such as: could be stated as [3]. i. Empirical model: It is derived from in-depth field measurements, that is, measured and averaged losses Urban area: 퐿(푑퐵) = 퐴 + 퐵푙표푔10푅 − 퐸 (2) along typical classes of radio links. It is efficient and Suburban area: 퐿(푑퐵) = 퐴 + 퐵푙표푔 푅 − 퐶 (3) simple to use. The most commonly used of this 10

method are Okumura-Hata model, the COST Hata Open area: 퐿(푑퐵) = 퐴 + 퐵푙표푔 푅 − 퐷 (4) model and the Lee model. 10

ii. Deterministic model: This model is derived from and 퐴 = 69.5 + 26.16푙표푔10푓 − 13.82푙표푔10퐻푏 (5) depend on the detailed and accurate description of all objects in the propagation space. This method is 퐵 = 44.9 − 6.55푙표푔10퐻푏 (6) expected to produce more accurate and reliable prediction of the path loss than the empirical method; 푓 퐶 = 2(푙표푔 ( ))2 (7) however, it is significantly more expensive in 10 28 computational effort. Thus, it is used predominantly 2 퐷 = 4.78(푙표푔10푓) + 18.33푙표푔10푓 + 40.94 (8)

Published on November 6, 2017. 퐸 = 3.2(푙표푔 (11.75퐻 )2 − 4.97 for large cities 푓 > 300MHz (9) H. Lasisi is a lecturer at Osun State University, Osogbo. He completed 10 푚 his BS, MS, and PhD Degree at University of Ilorin. 2 Y. A Adediran is a Professor at University of Ilorin in the Department of 퐸 = 8.29(푙표푔10(1.54퐻푚) − 1.1 for large cities 푓 < 300MHz (10) Electrical and Electronic Engineering. A.A Ayodele is a first class graduate of Osun State University in 퐸 = (1.1푙표푔 푓 − 0.7)퐻 − (1.56푙표푔 푓 − 0.8)for medium to small cities (11) Electrical and Electronic Engineering. 10 푚 10

DOI: http://dx.doi.org/10.24018/ejers.2017.2.11.370 13 EJERS, European Journal of Engineering Research and Science Vol. 2, No. 11, November 2017

height (m). The default values of these parameters a0, a1, a2 where L (dB) is Path loss, A, B, C, D and E are the antenna and a3 for different terrains are given in Table I. correction facto. R is the radial distance between the TABLE I: VALUE OF 푎 ,푎 , 푎 AND 푎 PARAMETERS FOR ERICSSON MODEL transmitter and the receiver. 퐻푏 and 퐻푚 are the base station 0 1 2 3 antenna and mobile antenna heights respectively, and f is the [11] Environment 푎 푎 푎 푎 frequency of transmission (MHz). The model is valid for the 0 1 2 3 Urban 36.20 30.20 12.00 0.1 carrier frequency range 150 MHz ≤ f ≤ 1500 MHz, base Suburban 43.20 68.93 12.00 0.1 station height in the range 30 m ≤ Hb ≤ 200 m, and mobile Rural 45.95 100.60 12.00 0.1 station height in the range 1m ≤ Hm ≤ 10m [3]. C. COST 231-Hata Model E. The Lee Model Committee 231 of the European Cooperation in the field Reference [4] proposed this model in 1982. In a very of Scientific and Technical Research (COST) in 1993 short time it became widely popular among researchers and extended the Hata model frequencies of interest. The model, system engineers mainly because the parameters of the which was renamed COST 231-Hata model, is applicable in model can be easily adjusted to the local environment by the frequency range of 500MHz to 2000MHz and can additional field calibration. In the beginning, the Lee model predict the median path loss for the distance of up to 20 km was developed for use at 900 MHz and has two modes: area- between the transmitter and the receiver with transmitter to-area and point-to-point. Built as two different models, antenna height of 30 m to 200 m, and receiver antenna this model includes an adjustment factor that can be used to height of 1 m to 10m [5]. COST 231-Hata model can be make the model more flexible to different regions of used to calculate path loss in three different environments; propagation. Lee model is given as [5]. urban, suburban and rural (flat) areas. It can be expressed as [6]. Scenario 1:

푃퐿퐶푀 = 46.3 + 33.9푙표푔10푓 − 13.84푙표푔10퐻푏 − 푎퐻푚 푓 Urban area, 푃퐿 = 123.77 + 30.5 log10 푑 + 10 푛 log10( ) − + (44.9 − 6.55푙표푔 퐻 )푙표푔 푑 − 퐶 900 10 푏 10 푚 α (17) (12) 0

Scenario 2: where P is the path loss, d is the distance between Lcm transmitting and receiving antennas (km), f is frequency 푓 Surburban, 푃퐿 = 99.86 + 38.4 log10 푑 + 10 푛 log10( ) − α0 (MHz), Hb is the transmitter antenna height (m), aHm is the 900 (18) antenna height correction factor. The parameter Cm has different values for different environments. It is given as 0dB for rural as well as suburban and 3dB for urban areas. Scenario 3:

The parameter aHm (mobile antenna height correction Rural area, 푃퐿 = 86.12 + 43.5 log10 푑 + factor) is defined in urban areas as 푓 10 푛 log ( ) − α (19) 10 900 0 2 푎퐻푚 = 3.20푙표푔10(11.75퐻푟) − 4.81 (13) Where 훼0 is the correction factor which accounts for the The value for aHm in suburban and rural (flat) areas is base station and mobile station antenna heights, transmit given as powers and antenna gain and it is given as

푎퐻푚 = (1.11푙표푔10푓 − 0.7)퐻푟 − (1.5푙표푔10푓 − 0.8) (14) 훼0 = 훼1 + 훼2 + 훼3 + 훼4 + 훼5 (20)

ℎ푏 2 Hr is the receiver antenna height in metres. 훼 = ( ) (21) 1 30.48 D. Ericsson Model ℎ 훼 = ( 푚) 푘 (22) The Ericsson model software from Ericsson Company 2 3 can be used to predict propagation path loss. Network 푃 planning engineers use it. This model also builds on the 훼 = ( 푡) 2 (23) 3 10 modified Okumura-Hata model to allow room for changes in parameters according to the propagation environment. 퐺 훼 = 푏 (24) Path loss according to this model is given as (Jalel etal., 4 4 2013) 훼5 = 퐺푚 (25)

푃퐿 = 푎0 + 푎1푙표푔1표푑 + 푎2푙표푔10퐻푏 + 푎3푙표푔10(퐻푏 + 푑) − 2 PL is the path loss, n is the path loss exponent chosen to be 3.2(푙표푔10(11.78퐻푟)) + 푔(푓) (15) 3, f is frequency, d is distance, hb is base station antenna 2 height, hm is mobile station antenna height, Pt is transmitter 푔(푓) = 44.49푙표푔10푓 − 4.78(푙표푔10푓) (16) power, Gb transmitter antenna gain, Gm is mobile antenna gain. where PL is the path loss, and f is frequency (MHz), Hb is transmission antenna height (m), Hr is receiver antenna

DOI: http://dx.doi.org/10.24018/ejers.2017.2.11.370 14 EJERS, European Journal of Engineering Research and Science Vol. 2, No. 11, November 2017

F. Egli Model The Egli model is a terrain model for radio frequency Putting (31) into (30) gives propagation. It predicts the total path loss for point-to-point link (line of sight transmission). Typically, it is suitable for 퐸푚𝑖푛 = 푃푇(푑퐵) − 푃퐿 − 퐴푒 + 145.8 (32) cellular communication scenarios where one antenna is fixed and another is mobile in the frequency range of 3 MHz But, to 3 GHz and where the transmission has to go over an 퐶 2 1.64( ) irregular terrain. Egli model is expressed as [3]. 푓 퐴 = 퐺 + 10 log ( ) (33) 푒 4휋 20 log 푓 + 푃 + 76.3, ℎ ≤ 10 푃퐿 = { 10 0 푟 (26) 20 log10 푓 + 푃0 + 83.9, ℎ푟 > 10 Substituting (33) into (32) gives

푃 = 40 log 푑 − 20 log 퐻 − 10 log 퐻 (27) 푐 2 0 10 10 푏 10 푟 1.64( ) 푃 (푑퐵) = 푃 (푑퐵푚) − 퐺 + 10 log ( 푓 ) + 145.8 − 퐸 (34) 퐿 푇 4휋 푚𝑖푛 where H is the height of the base station antenna, H is the b r height of the mobile station antenna in meters, d is the distance from base station antenna to mobile station antenna B. Models Performance Testing Criteria in meters, and f is the frequency of transmission in The performance of considered models were weighed in megahertz (MHz). comparison with the experimental path loss using different statistical evaluation criteria such as Sum Square Error III. METHODOLOGY (SSE), Average Absolute Relative Error (AARE) and Nash- The method used in collecting data is the Drive Test (DT) Sutcliffe efficiency coefficient (E). technique and the resources required are: C. Sum Square Error (SSE) i. Vehicle to convey other mounted DT equipment around the test area at a recommended speed of 8 -10 This describes the total deviation of the calculated path km/hr. loss values (XC) from the existing path loss models output ii. Laptop to serve as interface between the TEMs values (XE). Sum Square Error is given as [5].

software and TEMs mobile station and record the 푁 2 measured data such as base station ID, base station 푆푆퐸 = ∑𝑖=1(푋퐶 − 푋퐸) (35)

antenna height, frequency of transmission, received where N is the total number of data points predicted, X is signal level (RLx), signal quality index (SQI), signal- E the existing model values of path loss and XC is the to-interference ratio (C/I), number of calls attempt, calculated value of path loss. The model with the lowest number of calls blocked, number of dropped calls as SSE is considered to be the most relevant model to the study well as handover information during the drive test. area. iii. Drive Test Software (TEMS, Version 10.0.4 used) interfaces with TEMs mobile station for data D. Average Absolute Relative Error (AARE) measurement. This is the average of the relative errors in the prediction iv. TEMS Mobile Station (MS) interfaces with base of a particular variable and it is expressed as a percentage. station through RF for data measurement. Lower values of average absolute relative error (AARE) v. Global Positioning System (GPS) for positioning indicate better model performance. It can be computed as measurement during the DT. follows [5]. A. Calculation of Path Loss from Measured Data 1 푁 푋퐶−푋퐸 퐴퐴푅퐸 = ∑𝑖=1 | | 푋100 (36) Let Emin = Equivalent minimum field strength, Ae = 푁 푋퐸 Effective antenna aperture, G = Antenna gain, c = signal speed, f= signal frequency, Ф = Minimum power flux min E. Nash-Sutcliffe Efficiency Coefficient (E) density at receiving place, PL = path Loss, PT = BTS power The Nash-Sutcliffe efficiency coefficient is used to transmitted, PR = Mobile Station Power received. Then, given [3]. describe the accuracy of model outputs in relation to calculated data. A value of E equal to 1 depicts a perfect match between calculated path loss model and existing path 퐸푚𝑖푛 = Φ푚𝑖푛 + 120 + 10 log 120휋 = Φ푚𝑖푛 + 145.8 (28) loss model outputs; therefore, the closer the model efficiency is to unity (+1) the more accurate the model [5]. E Φ푚𝑖푛 = 푃푠푚𝑖푛 − 퐴푒 (29) is computed as follows: Putting (29) into (28), 2 ∑(푋퐶−푋퐸) 퐸 = 1 − 2 (37) ∑(푋퐸−푋̅̅̅퐸̅) 퐸푚𝑖푛 = 푃푅(푑퐵푚) − 퐴푒 + 145.8 (30)

̅̅̅̅ But, where 푋퐸 is the average value of the path loss given by a particular existing model. 푃퐿 = 푃푇(푑퐵) − 푃푅(푑퐵) (31)

DOI: http://dx.doi.org/10.24018/ejers.2017.2.11.370 15 EJERS, European Journal of Engineering Research and Science Vol. 2, No. 11, November 2017

IV. RESULTS AND DISCUSSION (MHz), Hb is transmitter antenna height (m), aHm is antenna The field strength data collected during the drive test height correction factor The Parameter Cm has different were converted into path loss and were subjected to values for different environments. It is given as 0dB for statistical performance evaluation criteria. A path loss rural as well as suburban and 3dB for urban areas. prediction model was therefore developed. Results are as The parameter aHm (mobile antenna height correction presented Fig. 1 to 6 factor) is defined in urban areas as

2 170 푎퐻푚 = 3.20푙표푔10(11.75퐻푟) − 4.81 (40)

150 Hr is the receiver antenna height in metres. Its coverage is for frequencies between 1500 MHz and 2000 MHz, mobile 130 station antenna height between 1m and 10 m and base station antenna height between 30 m and 200 m. 110 1 90 0,5

Path Path loss (dB) 0 70 -0,5 -1 50 -1,5 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1 -2 Distance (km) -2,5 Calculated path loss (dB) Okumura hata model (dB) -3

Egli model (dB) COST-231 Hata model (dB) Efficiency Sutcliffe -3,5

Coefficient (E) Coefficient - Ericsson model (dB) Free space model (dB) -4 -4,5

Lee Model (dB) Nash -5 Fig. 1. Carrier to Interference ratio -5,5 -6 40 Fig. 3. SSE for the considered models and experimental path loss 35 30 25 30 20 25 15 10 20 (Unitless) 5 0 15

0 0,2 0,4 0,6 0,8 1 AARE) Carrier to Interference ratio Interferenceto Carrier Distance (km) 10 C/I Best (A) C/I Worst (A) 5 C/I Best (C) C/I Worst ( C)

Fig. 2. Plots of the values for experimental and considered path loss models Error Relative Absolute Average over distance 0 Okumura Egli Cost-231 Ericsson Free Lee Hata space From the graph, the equation to fit the graphs is Models

푦 = 1.0124푥 − 2.9287 (38) Fig. 4. AARE for the considered models and experimental path loss where, y is the required modified path loss and x is COST 9000 231 Hata model for path loss. However, the regression coefficient, R2 of the graphs is 8000 calculated to be R² = 0.9995. This indicates a good 7000 relationship between the graphs because the closer the 6000 regression ratio to unity the better. 5000 Substituting (12), that is, COST 123 Hata path loss model 4000 for x in the curve fitting equation ((38)) and simplify gives a 3000 modified path loss prediction model for mobile communication for the studied area as 2000 (SSE) Error Square Sum 1000 푃퐿퐶푀푀 = 44.58 + 34.32푙표푔10푓 − 14.01푙표푔10퐻푏 − 1.01푎퐻푚 + 0 (45.46 − 6.63푙표푔10퐻푏)푙표푔10푑 − 1.01퐶푚 (39) Okumura Egli Cost-231 Ericsson Free Lee Hata space where PLcmm is the modified path loss prediction model for Models Osogbo, Nigeria. d the is distance between transmitting and Fig. 5. E for the considered models and experimental path loss receiving antennas (km), f is frequency of transmission

DOI: http://dx.doi.org/10.24018/ejers.2017.2.11.370 16 EJERS, European Journal of Engineering Research and Science Vol. 2, No. 11, November 2017

REFERENCES Calculated Path Loss Versus COST 231-Hata Model 130 [1] Abidoye, L. K and Das, D. B, (2014). Artificial Neural Network Modelling of Two Phase Flow in Porous Media. Journal of Hydro 120 Informatics, doi:10.2166/hydro.2014.079. [2] Abhayawardhana V.S., Wassel I.J., Crosby D., Sellers M.P., and 110 Brown M.G., (2005). Comparison of Empirical Propagation Path loss models for fixed wireless access systems. 61th IEEE Technology 100 y = 1,0124x - 2,9287 Conference, Stockholm, pp. 73-77, 2005. R² = 0,9995 [3] Chhaya Dalela., (2012). Propagation Path loss Modeling for 90 Deployed WiMAX Network. International Journal of Emerging Technology and Advanced Engineering. Vol.2, Issue 8, pp. 172-176. 80 [4] Danladi T.A., Lawa A.U., and Aderinola M., (2013).Studies on Effects of Building Internal Pattern on Downlink Mobile Phone

Path Loss (dB) (Calculatd) (dB) Loss Path 70 Signal Strengths and Power Loss. The International Journal of Engineering and Science (IJES). Vol.2, Issue 12, pp. 24-30. [5] Egli J.J., (1957). Radio Propagation above 40 MC over Irregular 60 Terrain. Proceedings of the IRE. Vol.45, Issue 10, pp. 1383–1391. 90 100 110 120 130 Path Loss (dB) (Cost-231 Hata Model) [6] Goldsmith A., (2005). Wireless Communication. Cambridge University Press, New York. Calculated Path Loss Versus Hata Model [7] Isabona Joseph., Konyeha. C. C., Chinule. C. Bright., and Isaiah Linear (Calculated Path Loss Versus Hata Model) Gregory Peter., (2013). Radio Field Strength Propagation Data and Fig. 6. Curve-fitting of experimental path loss and COST 231 Hata path Pathloss calculation Methods in UMTS Network. Advances in loss Physics Theories and Applications, Vol.21, 2013. [8] Jalel Chebi., Ali K. Lawas., and Rafiigul Islam M.D., (2013). Comparison between Measured and Predicted Path Loss for Mobile V. CONCLUSION AND RECOMMENDATION Communication in Malaysia. World Applied Sciences Journal (Mathematical Applications in Engineering). Vol.21, pp.123-128. Propagation path loss prediction model has been [9] Josip Milanovic, Rimac-Drlje S., and Bejuk K., (2007). Comparison developed for the study area using drive test technique in of propagation model accuracy for WiMAX on 3.5GHz. 14th IEEE International conference on electronic circuits and systems, Morocco, conjunction with six existing models. The model indeed pp. 111-114. works very fine because there exists a high degree of [10] Mohammad Shahajahan and Abdulla Hes-Shafi A.Q.M., (2009). correlation between its predicted path loss and experimental Analysis of Propagation Models for WiMAX at 3.5GHz. MSc, values when the model was verified. The method is Blekinge Institute of Technology, Karlskrona, Sweden. [11] Segun I.P., and Olasunkanmi F.O., (2014). Empirical Path Loss therefore recommended because it is less tedious and less Models for GSM Network Deployment in Makurdi, Nigeria. expensive compared to deterministic method which requires International Refereed Journal of Engineering and Science (IRJES). detailed and accurate description of all objects in the Vol.3, Issue6, pp.85-94. propagation space, and more expensive in computational Lasisi Hammed is a lecturer and researcher in the effort. The model produces relatively accurate results. The department of Electrical and Electronics Engineering, model is extensively useful and highly recommended for Osun state University, Osogbo, Nigeria. He holds doctorate (Ph.D) and Master’s degrees in Electrical conducting feasibility studies for signal prediction, coverage and Electronics Engineering from University of optimization and interference analysis during the initial Ilorin, Nigeria. His area of specialization is phase of network planning in the study area and other areas Telecommunication Engineering. He is a registered member of the Council for the Regulation of with similar environmental and propagation characteristics. Engineering in Nigeria (COREN). He can be contacted on [email protected]. Personal hobbies will be deleted from the biography. Author’s formal APPENDIX photo

PATH LOSS VALUES FOR EXPERIMENTAL AND SELECTED MODELS COST Experi Okumur Eric Free Dist Egli -231 Lee mental a hata sson space ance model Hata Mode path loss model mode model (km) (dB) model l (dB) (dB) (dB) l (dB) (dB) (dB) 64.83 92.01 77. 72.2 62.58 0.1 90 82.7500 00 48 2114 834 28 76.87 102.6 98. 78.3 74.14 0.2 101 93.3537 12 185 0058 040 23 83.91 108.8 110 81.8 80.90 0.3 107 99.5565 49 213 .1698 258 43 103.957 88.91 113.2 118 84.3 85.70 0.4 111 5 24 223 .8003 246 19 107.371 92.78 116.6 125 86.2 89.42 0.5 115 1 88 359 .4946 628 32 110.160 95.95 119.4 130 87.8 92.46 0.6 118 3 61 250 .9643 464 38 112.518 98.63 121.7 135 89.1 95.03 0.7 120 5 40 832 .5888 853 46 114.561 100.9 123.8 139 90.3 97.26 0.8 122 2 536 260 .5948 452 15 116.363 102.9 125.6 143 91.3 99.22 0.9 124 1 997 278 .1283 682 57 117.974 104.8 127.2 146 92.2 100.9 1.0 126 9 300 396 .2891 834 828

DOI: http://dx.doi.org/10.24018/ejers.2017.2.11.370 17