IEEE INTERNATIONAL SYMPOSIUM ON BROADBAND MULTIMEDIA SYSTEMS AND BROADCASTING 2014 1 Studying Digital Terrestrial TV coverage Pablo Flores Guridi, Member, IEEE, Andrés Gómez Caram, Agustín Labandera, Gonzalo Marín and María Simon, Senior Member, IEEE Facultad de Ingeniería, Universidad de la República, Uruguay. Email: {pablof, agomezca, labandera, gmarin, maria}@fing.edu.uy Abstract This article presents the development of numerical models to simulate the propagation of the TV signal, and a set of measurements of the electromagnetic signal to adjust the models. These tools have been devised to help in the deployment of digital television, ISDB-Tb standard, in Uruguay. The aim is a country coverage as wide as possible. A good estimation of the signal propagation shall be used to choose the best places for transmitters, the transmission modes and to give guidelines for reception antennas installation. The implemented models, which are well known, are briefly described. The measuring method is more thoroughly described as there is no established procedure and the spectrum analyzer s settings are discussed. Spatial validation, using neighbouring points is proposed. The measuring method is validated both by its coincidence with the Friis formula for free space propagation, when this condition arises, and by the consistency between neighbouring points. Okumura-Hata and Recommendation ITU-R P.1546-4 models are well suited for coverage prediction in Montevideo using certain parameters that are justified in the article. The statistical analysis of the set of data issued from the campaign is presented and discussed. Index Terms Propagation, coverage, channel modeling and simulation. I. INTRODUCTION URUGUAY has adopted in 2011 the ISDB-Tb digital television standard for open free diffusion, as did the majority of the South American region. The public TV broadcaster is already emitting a pilot signal, and the frequency assignation for commercial, public and community channels was already defined. However, the experience about propagation of the open TV signal is yet recent and scarce in the region. The geographical and demographic conditions are very different from Japan, and also are the building materials. A good estimation of the signal propagation, regarding the received power, interferences and noise immunity, shall be used to choose the best places for transmitters, the transmission modes, and to give guidelines for receiving antennas installation. To achieve this goal, a software simulation package was developed, based on previous work which was first intended for GSM signals. It includes different well documented models, suitable for broadcasting signals in the UHF band, in particular Okumura-Hata and Recommendation ITU-R P.1546-4. Due to the fact that Okumura-Hata does not contemplate terrain characteristics, its prediction is simpler than that from ITU-R P.1546-4. It is appropriate for a rough view of signal levels in the city, in contrast with ITU-R P.1546, This work was funded by Agencia Nacional de Investigación e Innovación (ANII), Uruguay. Official website: http://www.anii.org.uy. which returns a more detailed output, expressed out in less regular patterns. In order to register the ISDB-Tb signal levels in Montevideo and to compare its values with the models prediction, a measurement campaign was planned and carried out. The received signal power was measured using a standard handheld spectrum analyzers. Several bibliography suggest using different types of detectors. Its selection regarding accuracy and robustness is discussed in this article. The Uruguayan authorities have established 51 dbµv/m as the electric field strength value that a channel must achieve in the border of its coverage area. Its laboratories are testing the receivers to be authorized, whose sensitivity is also specified. The relationship between those values is a key factor when it comes to planning the emitter s location, height and power. In Section II the numerical models and its software implementation are presented. The campaing is described in Section III, in which the measurement methods are also discussed. A protocol was designed to improve the reliability of the collected data, based in the consistency between neighbouring sites, which is also described in Section III. The collected data are analyzed in Section IV, paying attention to coincidences and discrepancies between models and measurements. A very reasonable fitting with known and adjusted models is obtained, which proves that the software simulation package is correctly implemented. In Section V, broad guidelines for this kind of evaluation are given. II. NUMERICAL MODELS AND SOFTWARE PACKAGE A. Free-Space Path Loss Consider a signal transmitted through free space to a receiver located at a distance d from the transmitter. Assume the are no obstacles nearby to cause reflection, diffraction or scattering; so the signal propagates along a straight line between them: the channel model associated with this transmission is called line-of-sight (LOS) [1]. The ratio between the power transmitted and received is given by the simplest form of the Friis transmission equation [2], P r P t = G t.g r [ ] 2 λ, (1) 4πd where G t and G r are the transmitter and receiver antenna field radiation patterns in the LOS direction. This equation shows how the attenuation of the signal increases with the square
IEEE INTERNATIONAL SYMPOSIUM ON BROADBAND MULTIMEDIA SYSTEMS AND BROADCASTING 2014 2 of the distance d and decreases with the square of the carrier wavelength λ. However, most broadcasting systems operate in complex propagation environments that can not be modelled by freespace path loss. Several path loss models have been developed throughout the years to predict signal levels in rural, urban and suburban areas. All of them differ in how they approach shadowing, fading, diffraction, reflection and scattering of the signal. In this paper only two of them will be discussed, with an special focus on Recommendation ITU-R P.1546-4. B. Empirical and Semi-Empirical Path Loss Models These models are mainly based on empirical measurements performed at different distances from the transmitter, for different frequencies and environments, usually significantly different from those in which the models are going to be applied. Therefore, models must be analyzed and tuned for different regions and environments, which implies the need of taking measurements for verifying, adjusting or even discarding each model. Anyway, its original versions can be used as guidelines. 1) Okumura Model: This model was developed empirically in 1970s by Okumura [3] with data measured throughout the city of Tokyo, Japan. This model was developed to work with frequencies ranging 150 M Hz to 1500 M Hz, transmitting antennas height between 30 m and 100 m, mobile station antennas between 1 m and 10 m, and link distances greater than 1 km and lower than 100 km [1]. It was first conceived for analogical cellular networks. 2) Hata Model: The Hata [4] model for urban areas, also known as Okumura-Hata, is an empirical formulation of the graphical path loss data provided by Okumura. It takes as input the environment (urban, suburban or rural areas), the frequency of the carrier, the distance between transmitter and receiver, the height of the base station antenna and the height of the mobile reception antenna. Because of its simplicity, it is natural to choose Okumura- Hata as prime model to get a first approach of the propagation of the TV signal. Anyway, since terrain data are not considered, several factors as terrain clearance angle corrections or tropospheric scattering are shelved. 3) Recommendation P.1546-4: This recommendation provides a method for point-to-area propagation predictions for terrestrial services in the frequency range 30 MHz to 3000 M Hz. It is intended for use on tropospheric radio circuits over land paths, sea paths and/or mixed land-sea paths between 1 1000 km length for effective transmitting antenna heights less than 3000 m. The method is based on interpolation/extrapolation from empirically derived field-strength curves as functions of distance, antenna height, frequency and time and locations percentages. The calculation procedure also includes corrections to the results obtained from this interpolation/extrapolation to account for terrain clearance and terminal clutter obstructions [5]. The Recommendation ITU-R P.1546-4 takes as an input a surrounding representative parameter that can be either urban, dense urban or suburban area. This parameter must be set for every environment in which the prediction is done, so empirical data should be collected and analyzed. C. Implemented Software Specific software was implemented in order to compare and eventually adjust the above propagation models with digital terrestrial TV signal actual propagation. It is fed with geographical information and calculates the field strength or the received power in every desired area according to both the transmitting and receiving antennas characteristics, which can use either default or custom radiation patterns, and the transmission parameters, such as the transmitter effective radiated power (ERP) or the propagation path and surrounding characteristics, for each path loss model. This software was used to predict the emission of the national television network signal field strength in some specific locations in Montevideo, Uruguay. Free-space path loss model was used, then, also Okumura-Hata and Recommendation ITU-R P.1546-4. These models performance was then compared with data collected during a measurement campaign, described in Section III. Finally, the model which best fits the city s conditions was selected and the best parameters for it environment were chosen. A. Planning III. MEASUREMENT CAMPAIGN In order to acquire the ISDB-Tb signal levels in Montevideo and to compare its values with the models predictions, a measurement campaign was planned and carried out. The measured signal was the emission of the national television network (Televisión Nacional Uruguay: TNU). The measurement points were chosen according to distance and position relative to the transmitting antenna, as well as terrain and urban characteristics, in order to sweep the whole urban area of Montevideo. For that purpose, 7 radials were traced according the cardinal points. Depending on the radial lenght, a set locations were chosen, as it is shown in Figure 1. The measurement of TV electromagnetic field was better in open places; dense urban zones give very variable results as a consequence of reflection and diffraction effects. The received signal power was measured using a standard handheld spectrum analyzer that had been compared with very accurate instruments. Measurement of OFDM signals presents many difficulties because of its high peak factor. RMS, sample and min/max envelope detectors, usually found in spectrum analyzers, throw different results. Several bibliography suggest using different types of detectors, but after extensive laboratory studies, RMS was chosen as the adequate detector. B. Measuring method The measurement of RF signals is a highly complex task due to the many variations and effects of stochastic nature influencing the signal. Given this, it was considered that was not enough to make a single measurement for each location, but would be useful to have several measurements to compare between each other and then reach a unique representative
IEEE INTERNATIONAL SYMPOSIUM ON BROADBAND MULTIMEDIA SYSTEMS AND BROADCASTING 2014 3 Fig. 1. Measurement points for urban and suburban areas of Montevideo. Source: Montevideo 34 520 35.1400 S and 56 110 12.0800 W. Google Earth. October 13, 2013. value. Very similar values indicate a reliable measure. In this way it would be possible to be free of destructive effects such as multipath fading or reflections, which can affect a specific point but can not always affect the others. This also presents advantages when comparing with predictions made with ITU-R P.1546-4. It was noted that predictions made using reception height values of a few meters can present considerable differences in signal power in places just a few meters away. This makes it possible to detect these signal drops in a location. The procedure used was baptized Method of the Four Corners. It involves taking a measurement at each corner of the block to which it belongs the location chosen. This method gave highly satisfactory results. C. Results Several weeks were required to carry out the measurement campaign. A total of 88 measurements were taken in 26 different locations. For each measure, the geographic coordinates and the signal level were obtained. In most cases measures were taken at the four points planned, except in those where was not a defined block or it had no four corners. In these cases, two or three measurements were taken depending to the case. IV. DATA P ROCESSING AND A NALYSIS As a result of the measurement campaign, 88 points were measured from a total of 26 locations. Before doing any further analysis, it was necessary to remove outliers from the data. A. Measurements Selection Criteria In order to achieve this, quantitative and qualitative criteria based on a series of characteristics of the measured signal were developed. Some of the aspects taken into account were: 1) Channel spectrum shape. 2) Location characteristics: foliage, buildings, structures and traffic. 3) Signal level and dispersion among measurements in the same location. 4) Line of sight with the transmitting antenna. After applying these criteria to the dataset, two locations were discarded, obtaining a preliminary group of twenty four locations. Two locations with line of sight (LOS) to the transmitting antenna were chosen in order to validate the measurement procedure. Location SO3 is situated on top of Cerro de Montevideo (the city s highest hill) and shows an almost theoretical free space loss condition. The similarity between the link budget considering free space loss and the power measured proved the measuring method to be valid. Similarly, location SO2 is situated near the bay and, despite being in front of the transmitting antenna, is slightly obstructed by trees, so the result differs from the calculated value. Both locations had to be removed from the propagation models analysis because they do not meet the model s hypothesis. Hence, a final subgroup of twenty two locations was created. In Figure 2 the comparison between these measured values and the Friis transmission equation applied to the locations can be seen. B. Comparison Among Measurements and Predictions When the results from the different models and the measured points were compared, several issues had to be pointed out. 1) Free space loss represents an upper limit for received signal strengh. 2) From the different ITU-R P.1546-4 types of environment, suburban should be used. 3) Results considering ITU-R P.1546 considering 50% and 90% of the time are very similar.
IEEE INTERNATIONAL SYMPOSIUM ON BROADBAND MULTIMEDIA SYSTEMS AND BROADCASTING 2014 4 ONO6 SO1 NE4 S4 NO6 E4 SE1 NO1 NE1 S5 E1 S3 NO2 SE3 NE5 SO6 E2 SO5 ONO5 E6 NO3 NO4-10 -20 Prx(dBm) -30-40 -50-60 Measurement Free Space Loss Okumura-Hata P.1546 90/50 Suburban -70-80 Location Fig. 2. Comparison between measurements and models for the 22 locations subgroup. 4) Okumura-Hata does not consider the effect of big slopes in the terrain, while ITU-R P.1546-4 does but overestimates its effects (as seen in locations SO5 and SO6). 5) Locations near the transmitting antenna such as NO6 may be affected by multipath constructive interference, resulting in high signal power. In order to improve Okumura-Hata s performance, a new set of parameters adjusted to this dataset was proposed. It was done according to Lee s formula [6] [7]: A + B log d (2) Iterative cross-validation was used to estimate the parameters with linear least square method. The calculated mean error and the standard deviation of the error for the proposed models are shown in Table I. Model Zone Mean error Error σ 2 (db) (db) ITU-R P.1546-4 50% time Suburban 0.73 9.59 50% locations ITU-R P.1546-4 90% time Suburban 0.24 9.64 50% locations Standard OH Urban 1.12 9.3 Adjusted OH - 2.77 6.99 TABLE I COMPARISON BETWEEN RESULTS OBTAINED WITH THE NEW ADJUSTED MODEL, OKUMURA-HATA AND ITU-R P.1546-4. As seen in the table, the proposed values do not present a significant improvement in the perfomance of the model as it improves the deviation of the error but shows bigger mean error. In addition, both ITU-R P.1546-4 perform better than Okumura-Hata in mean error and similarly in error deviation. V. CONCLUSIONS Even though some references advise to use the spectrum analyzers with its min/max or sample values detectors, RMS measurement has proven to be more stable and expressive regarding the power density of OFDM signals. The implemented mathematical models fit with the measured data in three typical cases: when there is direct vision between transmitter and receptor, the Friss formula gives the good prediction; in medium reception zones both Okumura- Hata and ITU-R P.1546 perform well; and when the signal is weak, obstacles are present and diffraction is important, ITU- R P.1546 gives the better prediction, being in the pessimistic (and then conservative) side. The parameters to be selected in ITU-R P.1546 are, at least for Montevideo and very probably for Uruguay, sub urban type and terrestrial path. ACKNOWLEDGMENT The authors would like to thank the collaboration of UR- SEC, the Uruguayan Regulatory Body, for their participation in the measurement campaign and for calibration of instruments. REFERENCES [1] A. Goldsmith, Wireless Communications, Cambridge University Press, 2005. [2] H. T. Friis, A Note On A Simple Transmission Formula, Proc. IRE, vol. 34, p. 254, 1946. [3] Okumura, Yoshihisa, et al. Field strength and its variability in VHF and UHF land-mobile radio service. Rev. Elec. Commun. Lab 16.9 (1968): 825-73. [4] Hata, Masaharu. Empirical formula for propagation loss in land mobile radio services. Vehicular Technology, IEEE Transactions on 29.3 (1980): 317-325. [5] International Telecommunication Union, Recommendation ITU-R P.1546-4, 2009.
IEEE INTERNATIONAL SYMPOSIUM ON BROADBAND MULTIMEDIA SYSTEMS AND BROADCASTING 2014 5 [6] W. C. Y. Lee, Lee s Model a condensed version shown in Appendix II, written by the Propagation Ad Hoc Committee of IEEE Vehicular Technology Society appeared in a special issue of IEEE Transactions on Vehicular Technology, February 1988. pp. 68-70. [7] W. C. Y. Lee Mobile Cellular Telecommunications Systems, McGraw Hill Co., Chapter 4. Pablo Flores Guridi has obtained his Electrical Engineering degree in 2012 at Universidad de la República, Uruguay. He is currently working in his master thesis which is related with the work described in this article. His main academic interest is digital television. Andrés Gómez Caram is a student of Electrical Engineering in the Universidad de la República, Uruguay. He participated in the current work during his degree final project. Agustín Labandera is a student of Electrical Engineering in the Universidad de la República, Uruguay. He participated in the current work during his degree final project. Gonzalo Marín is a student of Electrical Engineering in the Universidad de la República, Uruguay. He participated in the current work during his degree final project. María Simon is Electrical Engineer and Full Professor at Uruguay s public university, Universidad de la República, in the field of Telecommunications. Her academic interests, in the field of Telecommunications, include Information Theory, Signal Coding and Data Networks, especially traffic studies. She is working also in Digital Television. She has been Minister of Education and Culture between 2008 and 2011 and President of the Board of ANTEL (National Administration of Telecommunications, public company) between 2005 and 2007. From 1998 to 2005 she has been Dean of the School of Engineering, Universidad de la República.