Comparison and Verification of Propagation Models Accuracy for Specific Urban Area

POSTER 2015, PRAGUE MAY 14 1 Comparison and Verification of Propagation Models Accuracy for Specific Urban Area Tomáš KOŠŤÁL 1, Martin KOŠŤÁL 2 1 Dept. of Electric Drives and Traction, Czech Technical University, Technická 2, 166 27 Praha, Czech Republic 2 Faculty of Electrical Engineering, Czech Technical University, Technická 2, 166 27 Praha, Czech Republic kostatom@fel.cvut.cz, kostama7@fel.cvut.cz Abstract. This article concerns verification of some of the models used for signal propagation in urban areas. After introduction, which briefly covers the historical evolution of mobile communication, it shortly describes Okumura - Hata, Ikegami and COST231 Walfish prediction models in urban areas and their usage. Furthermore it deals with problematics of practical measurements with a hand-held device and suitable applications for this purpose. Finally it presents a comparison of measurement at 900 MHz and model prediction for Dejvice district of Prague. Keywords Propagation models, Okumura - Hata, Ikegami, COST231 - Walfish, radio wave propagation, cellular network, mobile networks. 1. Introduction Communications and information are nowadays key pillars of modern human society. The new methods of communicating have been always important factors, that influenced the accelerating development of humankind. During past centuries, one can see a continuous search for faster, more robust and more reliable means of communication over long distances. The real breakthrough in means of communication was the usage of electric signal over previous optical or acoustic. The electrical telegraph was made available in the end of the first half of the nineteenth century. It was later followed by invention of telephone in seventieth, but all these progressive means of communications were still limited only to fixed places. There was still a demand to have a kind of communication that would be available literally everywhere without any fixed wiring. These visions became a reality thanks to many researchers and inventors like Hertz, Tesla, Popov, Marconi and others. From this time on, more and more usage of electromagnetic waves for transmitting information could be seen. The first mobile phones used in cars were available in forties. But this was still not a system for wider public. Other significant milestones came in eighties, when first hand-held mobile telephony devices were presented and first mobile networks like Nordic Mobile Telephony went operable. Since then, mobile telecommunications became a standard constituent of everyday life in most parts of the world. This created huge demand for robust and reliable mobile networks. Construction of a modern cellular network is a very complex task. One of the biggest problems is propagation of the signal. In the research history of electromagnetic wave have been discovered, that the radio signal fading can be divided into three basic types which are path loss, shadowing and fast fading [4]. Path loss is given by the physical properties of electromagnetic waves and can be described with a simple equation. Another type, shadowing, is caused by reflection, diffraction and scattering in the propagation environment which consists of hills, trees, buildings, vehicles, people, etc. The last type, fast fading is caused by movement, interference and noise and thus is mostly described by statistical methods. More exact description is not available by the time, because of a great complexity of needed calculations and lack of sufficient computer power. Cells of a modern cellular mobile network are comparatively small hence there is a large amount of them [1]. Antennas are thus mounted to the rooftops of the buildings in most of the cases, so they are not very high over the terrain compared to other means of transmitters such as radio or television. This fact causes, that there are only small possibilities to a have a visual contact between antennas of a transmitter and a receiver. For such micro and macrocells different models have been developed, that will be described in following chapters. 2. Propagation models In this section we will discuss the possible propagation models for Dejvice district, where we have conducted our measurements. It is located in Prague, Czech Republic in wider city centre. The majority of buildings is five to six floors and comprises of compact blocks. It is

2 T. KOŠŤÁL, M. KOŠŤÁL, COMPARISON AND VERIFICATION OF PROPAGATION MODELS ACCURACY FOR SPECIFIC URBAN AREA traditional inter-war period bricks housebuilding. There are not much trees except for main boulevards and other streets are comparatively wide (approximately 20 m). Base Transceiver Stations (BTS) are located on the rooftops in average level of 25 m above ground. 2000MHz for COST231 extension. The complexity of this model and several verifications made it so popular, that in lots of cases it is taken as a reference for comparison with other prediction methods. This model differs for three basic urban types open, suburban and urban. In our case we use the formula for urban areas. The model is accurate from the distance of one kilometre, however we will use it for shorter distance as a reference. The following formula describes the situation in urban area: where L=A B log R E (2) A=69,55 26,16log f c 13,82log h b B=44,9 6,55 log h b E= 1,1 log f c 0,7 h m 1,56 log f c 0,8 where L is attenuation in db over distance R in kilometres and f c is carrier frequency in MHz. Furthermore h b states for base station height and h m for mobile station height in meters. Fig. 1. Netmonitor application Dejvice measured area map with BTS locations. If the space, where the signal propagates, would be without any barriers and obstacles, the total propagation loss would be described by following free space loss formula[4]: L FS =20log 4 π R f c (1) In the real application, the signal has to overcome various landscape objects, vegetation or buildings for which the free space loss could not be used. In addition to the free space path loss we use advanced empirical and physical models. Empirical models are based on measurements of real signal propagation in typical areas [6]. The accuracy of such models is better than the free space path loss, they do not increase calculation difficulty, but market demands more accurate methods. For this reason, physical models were introduced. Their calculation difficulty is still acceptable. These models also need more information about the environment than the empirical ones. In following paragraphs, we will shortly describe the most suitable models for signal propagation in harsh conditions that urban area represents. 2.1 Okumura-Hata Model The most famous propagation model for macrocells [3]. It represents a fully empirical method, based on large scale measurements in Tokyo and its surrounding. It covers frequencies from 150MHz up to 1500MHz or up to 2.2 Ikegami Model The intention of this model is to create an entirely deterministic physical prediction of field strengths at particular street level points [2]. It uses details about heights of buildings and their distance from each other and trace ray paths between transmitter and receiver. The tracing accounts only single reflection from walls. Calculation of a diffraction uses single edge approximation at the building closest to the receiver and a constant value of wall reflection is assumed. The other specific feature of this model is, that it neglects the difference between roof and transmitting antenna heights. Fig. 2. Interpretation of Ikegami model. The reflected and diffracted rays are power summed and we get a following formula: L E =10 log f c 10 log sin Φ + 20 log h 0 h m 10 log w 10 log 1 3 L r 2 5,8 (3) where Φ is angle between a street and direct path to BTS, w is street width and h 0 is a building height. Because this formula describes only the loss behind the last diffraction and reflection from the neighbouring building, we need to add a free space loss, described in equation (1), to get the full total loss. We can then rewrite the equation to get the total loss in Ikegami model as: L=L E L FS (4)

POSTER 2015, PRAGUE MAY 14 3 2.3 Cost231 Walfish-Ikegami Model Enhanced Walfish-Bertoni model includes physical characteristics and empirical corrections for more accurate prediction. The model considers different formulas for line of sight between BTS and mobile, and for non line of sight, when the direct path is obstructed [4]. For our purpose we use only non line of sight version, which is given by L=L FS + L msd + L sd (5) The model uses free space path loss as a basic attenuation level, in equation (1). L msd states for multiple screen diffraction over rooftops. Diffraction to street level is based on Ikegami model and represented by L sd : L sd = 16,9 10 log f c 10 log h 0 h m 2 L Φ w m (6) purposes we have used a free app called Netmonitor. From it's log file we could extract the GPS coordinates and cell ID with signal strength by our script. L(Φ)=4 0,114(Φ 35 ) (7) we use Φ - 90, so we get where L Φ =4 0,069=3,931 (8) L msd =L bsh k a k d log R k f log f c 9 log w (9) L bsh = 18log[1 h b h 0 ] k a =54 k d =18 for h b > h 0, which is our case. (10) We consider our area as medium-sized city, for which we can write k f = 4 0,7 f c 925 1 (11) 3. Measurements We have measured the received signal power with a generic mobile phone receiver. From the measured data we have obtained the propagation loss. Measurements were conducted in Dejvice district, in the neighbourhood of CTU campus. The mobile phone used is Doogee DG550 running Android 4.4.2, advantage of this phone is, that it can provide signal information about neighbouring cells. Surprisingly well known manufactures such as Samsung or LG restrict this feature so monitoring cell cross sections with their devices is impossible. Android API provides complex information about BTS, but only for first SIM card in case of multiple SIM device. In our search for a logging app, we have discovered, that free app with full feature set is almost impossible to find. There are few paid such as G-NetTrack Pro and TEMS MobileInsight, but for our Fig. 3. Netmonitor application interface. UPLINK Mobile station RF power TX antenna gain Peak EIRP Base station RX antenna gain Diversity reception Depolarization loss Low noise amplifier BTS sensitivity Minimum reception level Isotropic path loss DOWNLINK Base station RF power dbm 45,0 A Combiner loss db 2,5 B db 3,0 C TX antenna gain dbi 16,0 D Peak EIRP dbm 55,5 E=A B C+D Mobile station RX antenna gain dbi 0,00 F db 0,00 G MS sensitivity dbm -105,00 H Minimum reception level dbm -105,00 I= F+G+H Isotropic path loss db 160,50 J=E I Tab. 1. Transmitter and receiver parameters [2]. The terrain measurement took us many days, we have covered the Dejvice district substantially densely. We used some tedious trigonometry to count buildings and BTS antennas height. Widths of streets were calculated from air imaging. 3.1 Attenuation calculation dbm 33 A db 0 B dbi 0 C dbm 33 D=A B+C dbi 16 E db 4 F db 0 G db 5 H db 3 I dbm -106 J dbm -128 K = E F+G H+I+J db 161 L=D K For calculation of the attenuation we take a local mean received power and subtract it from the transmitted

4 T. KOŠŤÁL, M. KOŠŤÁL, COMPARISON AND VERIFICATION OF PROPAGATION MODELS ACCURACY FOR SPECIFIC URBAN AREA Fig. 4. Graph of measured data and model predictions. Fig. 5. Graph of prediction models adjusted with offset. power in dbm thus we get attenuation over distance from the base station. Local mean received power is calculated from 20 to 50 samples in approximately 12m 2 area. The mean power is than subtracted from peak equivalent isotropically radiated power (Downlink EIRP) from the BTS, see Tab. 1. Losses and gain at the receiver can be omitted, because it is typically 0 db and 0 dbi respectively. Distance is taken from GPS coordinates. Other parameters for propagation models are in Tab. 2. Street width Building height BTS height 20 m 20 m 25 m 4. Results Minimal measured loss is 110dB and maximal 150dB. All samples are within this range. Loss larger than 150dB could not be measured due to sensitivity limitation and noise. We have done logarithmic fitting on measured data which gave us good logarithmic prediction curve [5]. See comparison of prediction models in Fig. 4. The measurements show pathloss of 110dB and worse. The Okumura-Hata or COST231 Walfish-Ikegami models can be used as a prediction border. The Ikegami model shows a correlation with the fitted logarithmic curve less than 10dB at distances larger than 200m. Fig. 5 shows prediction models with offset, so that they approximate the measured Tab. 2. Average values used for propagation models.

POSTER 2015, PRAGUE MAY 14 5 data more accurately. Correlation after this modification is still not acceptable. 5. Conclusion We have done measurement of GSM 900 MHz band signal coverage in Dejvice district and compared it with commonly used signal propagation models (see Fig. 4). The least accurate one proved to be Okumura-Hata model, but it is not designed for distances shorter than one kilometre. Cost231-Walfish - Ikegami model had similar results. The attenuation is underestimated by factor of 25 db (Fig. 5). Surprisingly well accurate is the Ikegami model where deviation is negligible. For further model prediction in Dejvice district, we suggest to use more physical models based on raytracing like the Ikegami model. About Authors... Tomáš KOŠŤÁL is a graduate student at the Dept. of Electric Drives and Traction since 2014. In his bachelor studies, he studied the Heavy Current Engineering program and continued in his master studies with program Electrical Engineering, Power Engineering and Management. He graduated in 2014 and continues with his doctoral studies at the same department. Currently he is focusing on digital control of semiconductor converters. Martin KOŠŤÁL is an undergraduate student at Faculty of Electrical Engineering, CTU in Prague. In 2014 he studied at Aalto University in Finland. Currently he is focusing on digital communication, multimedia and electronics. References [1] HEINE, G. GSM Networks: Protocols, Terminology, and Implementation. Norwood: Artech House inc., 1999. [2] LEMPIAINEN, J., MANNINEN, M. Radio interface system planning for GSM/GPRS/UMTS. Boston: Kluwer Academic Publishers, c2001, xiv, 278 p. ISBN 0792375165. [3] HATA, M. Empirical formula for propagation loss in land mobile radio services, IEEE Transactions on Vehicular Technology, 29, 317 25, 1980. [4] SAUNDERS, S. R., ARAGON-ZAVALA, A. Antennas and propagation for wireless communication systems. 2nd ed. Hoboken, NJ: J. Wiley & Sons, c2007, xxii, 524 p. ISBN 0470848790. [5] DIAWUO, K., CEMBERBATCH T. Data Fitting to Propagation Model Using Least Square Algorithm: A Case Study in Ghana. International Journal of Engineering Science, Vol. 2, No. 6, June 2013. [6] KLOZAR, L., PROKOPEC, J. Propagation path loss models for mobile communication. In: Proceedings of 21st International Conference Radioelektronika 2011. DOI: 10.1109/radioelek.2011.5936478. [7] RAHNEMA, M. UMTS network planning, optimization, and interoperation with GSM. Hoboken, NJ: IEEE Press, c2008, xviii, 327 p. ISBN 0470823011. [8] http://www.gsmweb.cz/ [9] ABHAYAWARDHANA, V.S., WASSELL,I.J., CROSBY, D., SELLARS, M.P., BROWN, M.G. Comparison of Empirical Propagation Path Loss Models for Fixed Wireless Access Systems. In: 2005 IEEE 61st Vehicular Technology Conference. DOI: 10.1109/vetecs.2005.1543252.