Path Loss Model in Amazonian Border Region for VHF and UHF Television Bands G. Castellanos1 Abstract—The calculation of the received energy is important to determine the interference for systems in frontier zones. The Amazonian region is a tropical rain forest, which propagation impairments are vaguely studied. Measured data concludes that in this region, the traditional propagation models like long-range and vegetation, do not explain the received power. This paper introduces a new propagation path loss model specific for the Amazonian region which includes the long range and foliage impact over the model, based in empirical data collected from the Colombian-Brazil border. Results show that the Amazonian region model obeys the behaviour on tropical forest where lateral wave mechanism predominates over direct-wave or groundreflected one.
1 INTRODUCTION In 2015, the Colombian government through the National Spectrum Agency (ANE), proposes the identification of the 614-698MHz band for IMT services in Region 2 on the Wireless Radiocommunication Conference (WRC15). This proposal was approved and the Colombian government should present bilateral agreement for coexistence between Digital Terrestrial Television (DTT) and LTE services in this band, which includes technical parameters for this services in the frontier with Brazil, Peru and Ecuador. Accordingly, the Colombian School of Engineering (ESCUELA) and ANE developed a project related with the coexistence between DTT and LTE signals in trans frontiers zones under a cooperation agreement. In order to recommend and analyse protection margins and distances, a specific path loss propagation model for the Amazonian region is developed. This paper describes the study and results of a proposed propagation model in the Amazonian border between Colombia and Brazil based in experimental measurements made in the border towns of Leticia (Colombia) and Tabatinga (Brazil) during September and November 2016. The structure of the paper is this: A deep review of long range and vegetation propagation models is described in Section 1. Section 2 describes the measurement procedures for the field study Next, Section 4 presents the proposed model for Amazonian region and validation mechanisms. Finally, Section 5 concludes the paper. 2 STUDY OF PROPAGATION MODELS Radio wave propagation models for UHF and VHF channels are presented in varied number of studies. Phillips et al, presents an extensive survey of wireless
G. Teuta2 propagations models from 1940´s to mid 2010´s and evaluates more that 50 propagation models [1]. They present a taxonomy tree for models and are dividend in two important approaches. The first one called a priori is based in models where previous knowledge is necessary for the definition of the model. In this branch six categories are presented: i) Theoretical/Foundational Models, ii) Basic Models, iii) Terrain Models, iv) Supplementary Models, v) Stochastic Fading Models and vi) Many-Ray Models. The second approach is for models that possibly use measurements for the prediction models and is divided in i) Explicit Mapping, ii) Partition Models, iii) Interactive Heuristic Refinement and iv) Active Learning Geostatistics. This last branch of the taxonomy presented by Phillips et al, considers that the predicted values are not only based on variables known a priori, but that the measurements are highly important and the models describe the procedures to acquire data and the methods to interpolate data for such places where there is not data information available. According to this scheme, we focus in a priori models and some of the most relevant categories for models in vegetation and long-range environments. 2.1
Theoretical and Foundational Models
Free space path loss model is described in the ITUR P525 [2] presented in (1). Based on Maxwell equations, this model defines a lower limit for the path loss. In the equation f is in MHz and d in km. = 32,4 + 20
(
2.2
) = 40 Log( ) − 20 Log(ℎ ) − 20 Log(ℎ ) (2)
Basic Models
Basic models are the top abundant categories. They are based in the computing and correlation of a single path between the transmitter and the receiver accounting variables as Frequency, Antennas heights as primary input variables. Also, these models could cluster into deterministic or stochastic models. The stochastic ones
Faculty of Electronic Engineering, Escuela Colombiana de Ingeniería. Ak 45 #205-59. Bogotá – Colombia. Email:
[email protected], tel: +57 1 6683600. Faculty of Electronic Engineering, Escuela Colombiana de Ingeniería. Ak 45 #205-59. Bogotá – Colombia. Email:
[email protected], tel: +57 1 6683600.
2
( ) (1)
However, if the radio wave propagates near ground, the path loss model that includes the reflection of the ground is the Plane Earth model shown in (2) where d is de distance between antennas, hTx and hRx the antenna elevation of the transmitter and receiver antennas respectively [2], all in meters. The relevance of this model is described in the next section.
________________________________________________________________________________________ 1
( ) + 20
include random variables to cope with channel variation. The vast majority includes extensive field measurements of the received energy which includes all the phenomena that affect the received signal. Examples of these models are Gree-Obaidat, Edwards-Drunkin, deSouza-Lins, TM90, Walfisch-Bertoni, ERCEG, IMT200 and the famous Okumura-Hata, which includes ECC-33, Hata-Davison, COST-231, Extended Hata and Rural- Hata. [1] We are going to focus on OkumuraHata and COST-231 models for this study. The Okumura-Hata model [3], is based on the measurements made by Okumura and depicts the behaviour of propagated signals between 150MHz and 3GHz, 30 and 1000 meters of transmission antenna height and distances up to 100km. The Hata modification based on Okumura´s measurements, is shown in (3). This model works between 150 and 1500MHz, with distances up to 20km, 200 meters up to transmission antenna height and reception antenna height between 1 and 10 meters. Finally, the model presents a correction factor for the height of the reception antenna determined in a(hRx). (
) = 69.55 + 26.16 ( ) − 13.82 (ℎ ) + (44.9 − 6.55 (ℎ ))
2.3
) = 46.3 + 33.9 ( ) – 13.8 +(44.9 − 6.5 (ℎ ))
(ℎ ) – ( ℎ ) ( )+ (4)
Effect of Forest in Propagation Models.
There have been several studies that evaluate the implication of forest, leaf size, and vegetation on propagation models [5] [6] [7]. The majority contemplate the forest as a random mixture of leafs, branches and trunks that leads the signal to suffer from scattering, diffraction, reflection or shadowing, and the reduction of received power. Recommendation ITU-R P.833-9 [5], presents a complex formulation in pro of modelling obstruction by woodlands characterized by: i) Specific Attenuation rate (dB/m) and ii) Maximum total attenuations. The first defines the rate in which energy is loss due the propagation inside the woodland and the second defines how much is the maximum attenuation inside the woodland before the predominant signal arrives for different paths apart for the ones through vegetation. In (5) d is the distance inside the forest, is the specific attenuation in (dB/m) and Am is the maximum attenuation within a particular type of vegetation. =
[1 − exp(− /
]
=
(ℎ ) − ( ) (3)
An extension up to 2GHz of Hata model is the COST 231 [4]. The CM factor which compensate the size of the cities is also introduced by COST, where for suburban and small cities, the value is 0dB, meanwhile for large cities the value is 3dB. The path loss equation of this model is presented in (4). (
Torrico et al in [8], presents s mathematical modelling of trees and leafs as cylinders and discs respectively and calculated the Specific attenuation for diverse polarization and calculate propagation losses relatives to Free Space losses where trees are in the line of sight. According to the recommendation ITU-R P.833-9, for propagation through forest, Song Meng et al [4] describes three propagation mechanisms across vegetation: i) Direct Wave, ii) Lateral Wave, iii) Ground-reflected Wave. Therefore, the received signal is the sum of all three propagation mechanisms, as is described in (5). Fig 1. shows these mechanisms and the relation with urban environments. For VHF and UHF there are four major studies that contemplates the forests and vegetation implications. Based in Rec ITU-R P.833, the following studies are constructed in a simplified equation of losses shown in (6) and (7) where f is in MHz and d is the distance of the receiver antenna inside the forest in meters, is the elevation angle of the directive ray, A, B, C, E and G are empirical found parameters.
(5)
=
+ ∗
(6)
∗( + )
∗
(7)
Weissberger [9] presents a modified exponential falloff model for moderate climate where the propagation considers the Direct Wave principle, rather than Lateral or Ground-reflected wave. This model works for frequencies between 230MHz and 95GHz and is presented in (8). (
)=
1.33 ∗ 0.45 ∗
.
.
14 0
.
< ≤
≤ 400 (8) < 14
The COST 235 [10] model is proposed based on measurements made in range from 9.6 to 57.6GHz but it can be adapted for usage for VHF frequencies. This model could work for woodlands up to 200 meters wide and introduces the impact of the presence of leafs, for different seasons i.e. winter and summer. (
)=
.
26.6 ∗ 15.6 ∗
.
.
.
ℎ ℎ
(9)
Alike, the Fitted ITU (FITU) [11] is based on measurements of 11.2 and 20GHz and is described as follows. (
)=
0.39 ∗ 0.37 ∗
.
. .
.
ℎ
ℎ
(10)
Finally, the Lateral ITU (LITU) [12] model from Nanyang University made in a tropical forest over the monsoon season was carried out with 240 and 700MHz signals, is the most like the Amazonian region, and its measurements are close to the VHF/UHF bands proposed. (
) ≅ 0.48 ∗
.
.
(11)
broadcast antenna. The data collected for the second trip was used for validation of the model and fin tunning of the empirical parameters described in next section. 4 PROPOSED MODEL 4.1 Fig. 1. Propagation Mechanisms in Urban-Forest Eviroments
3 MEASUREMENTS IN FIELD The data was collected in the Amazonian region in the Colombian and Brazilian border towns, of Leticia and Tabatinga that compromises an area of 9.234km2 with nearly 102.000 habitants, which only the 30% live in rural and forest zones. The Amazonian rain forest consists of trees with height up to 40 meters, with no hills or mountains, large swamp and rivers which contributes to the 85% of relative humidity and 25Cº of average temperature. In the Colombian side, there are two broadcast television antennas, on for public channels and the other for private ones. According to ITU recommendations SM 1792 and SM1875, a measurement protocol was established to measure the received average channel power for the TV channels in VHF and UHF. Each point averages 25 measures per channel measured. The established protocol defines the usage of a 3-meter-high directive 14.5dBi gain antenna pointed to the transmitting antenna aid with GPS. The antenna is connected to a Tecktronix MDO-3000 Spectrum Analyser and a TV HD Ranger 2 Television Analyser. A laptop was used to collect data from both devices for future analysis. Fig 2, describes the used setup in the Amazonian region.
Fig. 2. Measurement Setup for In field data collection.
To achieve the objective, two field trips to the Amazonian region were done. The first one obtain 90 point divided in 6 routes around the different antennas following the protocol were antenna angles were compensated to maximize the antenna gain. After data was collected and the modelling was done, a second trip to the area was implemented. In this trip, measurements were done in 60 point only in +-5º angle from the principal lobules in order to avoid antenna gain compensation and just for the public Colombian
Comparisons of Traditional Models with Measured Data.
ICS Telecom software were used to simulate the coverage and propagation for all the channels in the region. Long-range path loss models like Free Space [2], Okumura – Hata and Cost 231 [3] were used to compare the collected data. The measured data of channel 13 compared to traditional models is shown in Fig 3. Result could be grouped based in two tendencies, the first is based in long range models, and the second ones present a higher decadency like the vegetation ones. Though, the collected data present a different behaviour and a new model need to be introduced. 4.2
Path Loss Modelling
Data adjustment was made using the Curve Fitting Toolbox from Matlab. Data was pre-processed to use only the path losses L calculated from the received power, transmitted power and antenna gains. Root Square and Root-mean-square Error were used to validate the fitness. For long-range models, Hata model is better fitted. For vegetation models, Weissberger model outclass the other two models. Due the fact that neither of the evaluates models follows the data trends, we propose a third model combined with part long range and part vegetation model. We defined a relation of this distances as 80%20% based on the urban area and the forest area found in the measurements, respectively. It is important specify that long range models use the distance in kilometres, meanwhile vegetation models use it in meters, hence units should manage accordingly. The three models were evaluated and adjusted using data from the first trip using Matlab toolbox, limiting the parameters of the equations according to literature. In Table I, the best fit evaluation of the models is presented, where the proposed Amazonian model outclass the two traditional models, with a R-Square of 0.3056 and RMSE of 14.596 supporting the hypothesis of a combined behaviour of long-range propagation with a component inside the vegetation. Furthermore, this supports the findings of Tamir [7] [13] where in relative large vegetation deeps, the lateral propagation using diffraction becomes predominant over direct wave inside the vegetation. Our proposed model is presented in (12). ( ) = 45.96 + 33.56 log( ) − 15.5 log(ℎ ) − 1.40 + 9.18 log(0.8 ∗ ) + 0.5 ∗ . ∗ ((0.2) ∗ ∗ 1000) . (12)
Fig. 3. Comparison data with propagation models for 13Ch TABLE I.
BEST FIT FOR PROPOSED MODELS
Proposed Model Long Range. Vegetation Proposed Amazonian
4.3
R-Square 0.2508 0.0003 0.3056
RMSE 14.891 16.637 14.596
Model Validation
The validation of the proposed model is performed using data for the second trip to the Amazonian region. Fig 4, presents the measured validation data compared with the proposed models. The figure shows the performance of the vegetation model that rise the losses sharper than the other models and cannot forecast the data form distances under 800 meters because deviation up to 50dB were found. The Long-range model has a closer conduct related with collected data, but its incremental speed is softer than the plane earth, so for distances larger than 5km could not foresee the path loss. Finally, the proposed Amazonian model fits very close the measured data for distances between 400 meters and 10km. Also, the proposed Amazonian model presents a slightly increment of path loss for deep foliage calculations, where lateral wave predominates. 5 CONCLUSION AND FUTURE WORK We propose a combined model that accounts both for the long range and the vegetation propagation behaviour. Practical measurements show that lateralwave propagation methods are predominant over directwave in large tropical rainforest. The proposed combined model for Amazonian region, obeys these mechanisms tested with the validation data. The model was validated for distance from 400 meters up to 10km in UHF/VHF TV bands according to the measured data. The relation between the vegetation and the long range part should be included in the future analysis of the proposed Amazonian region path loss model. Fine tuning with constant sine-wave carrier using omnidirectional antennas in the Amazonian region is the next step for experimental measurements where the antennas height, carrier frequency and water bodies parameters should be included in the model.
Fig. 4. Comparison from adjusted models with measured data.
Acknowledgments The authors would like to thank Natalia Duarte, Alexander Gordillo and Dayana Matíz from ESCUELA and Vivian Gonzalez, Antonio Alvarez and Martha Suarez from ANE from Colombia, for their help in the accomplishment of the objectives of this project. This work was funded by the National Spectrum Agency-Colombia and Colombian School of Engineering under cooperation agreement 057/2016. References [1] C. Phillips, D. Sicher and D. Grunwald, "A Survey of Wireless Path Loss Prediction and Coverage Mapping Methods," IEEE Communications & tutorals, vol. 15, no. 1, 2013. [2] ITU, "Rec ITU-R P525 - Calculation of free-space attenuation," Geneva, 2016. [3] J. Parsons, The mobile Radio Propagation Channel, New York: Wiley, 2000. [4] COST, COST231 - Digital mobile radio towards future generation systems, Brussels, 1999. [5] ITU, "Rec ITU-R P.833 - Attenuation in Vegetation," 2016. [6] ITU, "Rep ITU-R 236 - Influence of Terrain Irregularities and Vegetation on Tropospherical Propagation," vol. 5, 1986. [7] T. Tamir, "On radiowave propagation in forest enviroment," IEEE Trans on Antenna and Propagation, vol. AP15, Nov 1967. [8] S. Torrico, H. Bertoni and R. Lang, "Modeling Tree Effects on Path Loss in a Residential Enviroment," IEEE Trans on Antennas and Propagation, vol. 46, no. 6, 1998. [9] M. Weissberger, "An initial critical summary of models for predicting the attenuation of radio waves by foliage," Electromagnetic Compatibility Analysis Center, 1981. [10] COST, "COST235 - Radio Propagation effects on nextgeneration fixed service terrestrial telecomunication systems," 1996. [11] M. Al-Nuaimini and R. Stephens, "Measurements and prediction model optimization for signal attenuation in vegetation media at centrimeter wave frecuencies," in Proc in Electrical Eng and Microwave Antenna Propagation, 1998. [12] Y. Song Meng, Y. Hui Lee and B. Chong Ng, "Empirical Near Ground Path Loss Modeling in a Forest at VHF and UHF Bands," IEEE Trans on Antennas and Propagation, vol. 57, no. 5, 2009. [13] T. Tamir, "Radiowaves propagation along mixed paths in forest enviroments," IEEE Trans on Antennas and Propagation, vol. AP25, Jul 1977.
