Path Loss Measurements for a Non-Line-of-Sight Mobile-to-Mobile Environment

Transcription

1 Path Loss Measurements for a Non-Line-of-Sight Mobile-to-Mobile Environment J. Turkka, M. Renfors Abstract This paper shows results of narrowband path loss measurements in a typical urban and suburban mobile-tomobile radio environment at 900 MHz band. The measurements were made with two omni-directional antennas with a transmitter and a receiver antenna height of 1.5 meters. The results of measurements provide practical values for path loss exponent and standard deviation of shadowing in a nonline-of-sight radio environment. These parameters can be used with a simple power law path loss model to predict reliable communication ranges for future communication systems operating in a mobile-to-mobile environment, such as a relay extended cellular networks, non-line-of-sight vehicle-to-vehicle communications or relay assisted positioning applications. Key words: mobile-to-mobile, path loss measurements, path loss exponent; shadow fading; coverage dimensioning. M I. INTRODUCTION OBILE-TO-MOBILE radio channels can be found in vehicle-to-vehicle communications, ad-hoc networks, sensor networks and relay based cellular networks. During the recent years, statistical properties of mobile-to-mobile channels have been investigated by many researchers concentrating on small-scale characteristic of the mobile-tomobile radio channels [1]-[4]. However, no large-scale path loss models were found, that would be suitable for a long-distance mobile-to-mobile communication system operating in an urban or a suburban environment at 900 MHz band. Propagation of radio waves is a complex phenomenon and usually characterized by reflections from electrically smooth surfaces, refractions between two propagation mediums, diffractions from the edges of buildings and scattering from electrically rough surfaces. Characteristics of a mobile-tomobile radio channel are quite different compared with the traditional base station to mobile radio channel. A traditional base-to-mobile radio environment is often defined to consist of a stationary transmitter which is located above the rooftops and a mobile receiver which is placed near ground. In a built-up environment, the propagation occurs mainly above the rooftops and is later diffracted and reflected to the ground level just before the receiver. In contrast to the base-to-mobile radio environment, a Manuscript received June 13, This work was supported in part by the National Technology Agency of Finland (TEKES) and Nokia Research Center. J. T. Turkka is with the Radio Network Group of Department of Communication Engineering at Tampere University of Technology; P.O. Box: 553 FI-33101, Tampere, Finland; phone: ; fax: ; jussi.turkka@tut.fi. mobile-to-mobile radio environment consists of two nonstationary transceivers and both of them are located close to the ground level. Therefore, the environment is much more dynamic as both of the mobiles can be in motion. Propagation of the signals in mobile-to-mobile environment is characterized by multiple reflections, diffractions and a more challenging scattering environment. The reliability of a base-to-mobile communication channel can be improved with antenna diversity, higher antenna gains, and increased transmission power. However, these improvements are seldom available for mobile-to-mobile communications and therefore, it is important to understand the physical limitations of the mobile-to-mobile channel. In this paper, measurement based path loss exponent and shadowing parameters are derived for a simple power law path loss model. The measurements were carried out in urban and suburban non-line-of-sight environments with low elevation antennas for a transmitter and a receiver. This paper is organized as follows: In Section II, the measurement configuration and scenarios are described. In section III, the empirical propagation model that is tuned with measured data is shown. Measurement results are described in Section IV. In Section V, the coverage ranges of adjusted models are shown. Finally, the conclusions are drawn in Section VI. II. MEASUREMENT METHOD A. Measurement configuration The measurement configuration consisted of a stationary transmitter and a receiver mounted to a car. An omnidirectional quarter-wave dipole antenna with 3 dbi gain was mounted on a tripod and connected to the signal transmitter with a 10 meter feeder cable. A narrowband continuous wave signal with the carrier frequency of MHz was fed to the TX antenna with 30 dbm power. An omnidirectional half-wave dipole antenna with 3 dbi gain was mounted on the roof of the car and a spectrum analyzer was inside the car, taking samples at intervals of one second. The height of the transmitter and the receiver antenna was 1.5 meters which corresponds well to the definition of mobile-to-mobile propagation scenario. The transmitter antenna location was stationary during all measurements, which is not exactly the case for the mobile-to-mobile communications. However, the mobility of the receiver and the transmitter affects mainly the small scale effects and the Doppler spectrum [1]. Therefore, while performing the large scale path loss measurements, the results are reliable

2 Transmitter Frequency TX Power TX Antenna TX Feeder Cable TABLE I MEASUREMENT CONFIGURATION MLJ PCS MHz 30 dbm AV1950 (gain 3 dbi) RG214 (loss 2.4 db) TABLE II MEASUREMENT SCENARIOS Building Density Building Height Kissanmaa 13 % of total area Less than 8 meters Tammela 21% of total area More than 15 meters Hervanta 8% of total area Less than 15 meters Receiver RX Antenna RX Feeder Loss FSL3 Spectrum Analyzer λ/2 omni (3 dbi) 3 db regardless of the stationary transmitter. Table I summarizes the measurement configuration. B. Measurement Environment Measurements were performed during late November 2007 and the measurement campaign consisted of two slightly different propagation environments in Tampere, Finland. The first measurement survey was performed in Kissanmaa district reflecting a typical suburban region. The second measurement survey was performed in Tammela district reflecting a typical urban region. The Kissanmaa district consists of a scarcely built-up environment and small houses with two to three floors, and backyards. The average building height is less than eight meters and wood and brick are the most commonly used building materials. In addition, the region is relatively flat and characterized with parks and forest areas. Approximately 13 % of the area is filled with the buildings. The measurement survey in Kissanmaa consisted of four routes. Three routes were fixed directly away from the transmitter and the purpose of those measurements was to investigate the behavior of the propagation slope and the signal attenuation without significant effect of shadowing. However, this kind of measurements can be vulnerable for guided waves propagation phenomenon, where street structure and orientation help the radio waves to propagate, resulting in too optimistic measurement results. The purpose of the fourth route was to collect samples from spatially different equidistant locations. This is more preferable for illustrating the behavior of the shadowing. Tammela district was chosen to represent a typical urban region which consists of blocks of densely built-up buildings. The average building height is more than 15 meters and buildings may have five to eight floors. In addition, the region is characterized by parks and open parking areas. Approximately 21 % of the area is filled with buildings built of concrete. Table II summarizes the average building density and height statistics for measurement regions extracted from digital map information. In addition, some extra measurements were performed later during May The measurements were performed in Hervanta district (i.e. another suburban district) and the target was to verify the line-of-sight behavior of the mobileto-mobile channel. The measurement area in Hervanta consists of thick forest and open areas. Only 8 % of the area is buildings with an average height of less than 15 meters. III. EMPIRICAL MODEL A simple power law path loss model [5] was chosen for predicting the distance of reliable communication between two mobiles. A modified power law path loss model is given as d PL( d ) = PL( d0 ) + 10n log + c + Xσ, d0 where PL(d) is the mean path loss value, PL(d 0 ) is the measured or predicted reference path loss at distance d 0. The reference path loss value is approximated either using the free space path loss formula or through field measurements at distance d 0 [5]. Variable n is the path loss exponent and it describes how quickly the signal attenuates as a function of distance (i.e. a slope or a steepness of the path loss curve). Variable c is the offset correction factor and it describes the constant offset between the reference model and the measurements. Variable X σ is the random variable describing the shadow fading deviation from the mean path loss value which is present due to the fact, that two spatially different locations having the same transmitter-receiver separation may undergo totally different kinds of radio paths. Therefore, measured signals may differ greatly in their mean predicted signal levels. Field measurements have verified this variation to be environment dependable and distributed lognormally [5]. The path loss model (1) was adjusted to fit to measurement samples by minimizing the squared error between the measured samples and the predicted samples (i.e. least squares criteria). Model was tuned by finding proper values for path loss exponent n and offset correction c. The reference path loss value PL(d) was defined to be 65.5 db based on the free space path loss assumption at 50 meters distance from the transmitter. The 50 meters reference distance is based on prediction of breakpoint distance in micro cellular environments which defines the distance of line-of-sight propagation [6]. This results in a simple two slope model, where path loss of first 50 meters follows free space loss attenuation. After 50 meters, the path loss is assumed to correspond to the measured mobile-tomobile environment characteristics. (1)

3 Fig. 1. A path loss curve for Kissanmaa region. Fig. 2. A path loss curve for Tammela region. IV. MEASUREMENT RESULTS A. Path Loss Analysis This section shows measured path losses for Kissanmaa and Tammela regions. For parameter adjustment and path loss analysis, all samples further than 50 meters away from the transmitter were considered, excluding the samples below the noise floor level of spectrum analyzer. Fig. 1 and Fig. 2 show the path loss curves for Kissanmaa and Tammela regions. The solid curves indicate the adjusted models with the path loss exponent and the offset correction. The dashed lines indicate the reference models without the constant offset c. The light grey markers show the measured path loss values. In Kissanmaa, the measured path loss exponent n is 3.6 and the measured constant offset c is 31.5 db. In Tammela, the measured path loss exponent n is 4.9 and the measured constant offset c is 23.5 db. In both cases, the measured path loss exponents were higher compared to the corresponding macro cellular environments [7]. The behavior was expected based on the previous micro cellular antenna height and peer-to-peer studies [8]-[12]. However, these studies did not provide values for long-distance mobile-to-mobile path loss models. The value of constant offset c was surprisingly high in Kissanmaa and Tammela measurements being nearly 30 db below the assumed free space loss reference point at the 50 meters distance. This might occur due to many reasons. There might be obstruction obstacles and the line-of-sight connection is lost earlier than assumed comparing with the assumed reference point. If there are many scatterers, a clutter loss or a loss due multiple reflections might explain the difference between the reference model and the measured values. Earlier studies showed that unfavorable receiver location in mobile-to-mobile environment can result in db extra losses [13]. Moreover, propagation behind the corners of buildings was observed to cause db attenuation to the path loss values in urban environment [12]. Fig. 3. A path loss curve for Hervanta region. On the other hand, there is already a nearly 20 db loss present in the beginning of the measurement in Kissanmaa and Tammela after first 10 meters. Therefore, it is possible that there are some extra losses because of antenna mismatch, poorly connected cables or broken connectors. For reliable coverage predictions, the effective isotropic radiated power (EIRP) must be known precisely. In practice, the offset is a contribution of all above-mentioned reasons. Fig. 3 shows the later measurements in Hervanta, where the measured path loss exponent n is 5.1 and the measured constant offset c is 6.8 db verifying that LOS criteria is fulfilled. This indicates that absolute value of path loss in Kissanmaa and Tammela might be overestimated. However, the path loss exponent and the shadowing standard deviation can be derived regardless the exact value of EIRP which affects only the reliability of the absolute path loss value (i.e. EIRP does not affect to the steepness of path loss curve or the behavior of shadowing). Also the dual slope behavior of the mobile-to-mobile environment can be observed on Fig. 1 - Fig. 3 in Tammela and Hervanta, the 50 meters assumption for the breakpoint distance was quite accurate but in Kissanmaa, it seems that 30 meters might have been a better assumption.

4 TABLE III PROPAGATION PARAMETERS Kissanmaa Tammela Hervanta Offset correction c 31.5 db 23.5 db 6.8 db Path loss exponent n Shadowing STD 8.2 db 7.7 db 6.1 db TABLE IV COVERAGE RANGE ESTIMATIONS Kissanmaa Tammela Hervanta Fig. 4. A probability distribution of prediction error in Kissanmaa. Fig. 5. A probability distribution of prediction error in Tammela. This indicates that the local scattering environment can significantly affect to the reference distance and therefore, it is important to plan carefully the placement of the transmitter when doing mobile-to-mobile measurements. B. Shadow Fading A standard deviation (STD) of error between the samples and the prediction is a good measure of shadow fading, if the samples are collected in the right manner which smoothes out the fast fading component. If there are no other external errors biasing the results, then only a slow fading is present in the error. Therefore, the probability distribution of the error should be log-normally distributed. This is identified by observing the probability distribution function (PDF) of the prediction error. Fig. 4 and Fig. 5 show probability distribution functions of prediction error in Kissanmaa and Tammela regions. The standard deviation of the shadowing is 8.2 db in Kissanmaa and 7.7 db in Tammela. In Hervanta, the standard deviation of the shadowing is 6.1 db. The solid curves show the normal distribution with standard deviations corresponding to the measurements. The dashed lines show the measured probability distribution functions. Studies indicated that shadow fading in Tammela and Hervanta follows the theoretical log-normal distribution. On the other hand, in Kissanmaa, the curve of measured PDF does not fit well to the theoretical curve and the results are biased. This systematic error can be explained by noting that one of the 90 % location probability 410 m 340 m 665 m 50 % location probability 550 m 420 m 830 m Reference model 3.0 km 1.0 km 900 m measurement routes in Kissanmaa faced relatively better propagation conditions compared with the other measurements in Kissanmaa (i.e. the constant offset c was only 18 db). The measured standard deviations are similar to the results measured at 2 GHz UMTS frequencies in urban and suburban mobile-to-mobile channels [13]. Though, the analysis method was slightly different. V. COVERAGE DIMENSIONING An estimation of the coverage range for a communication system that has a maximum allowable path loss of 130 db can be done based on the tuned path loss model. A slow fading margin of 5 db is incorporated with the maximum allowable path loss for providing more reliable coverage range estimation. The slow fading margin correspond 90 % area location probability in case of 8 db shadowing standard deviation and path loss exponent of 4 [7]. Table III shows propagation parameters and Table IV shows coverage ranges in urban and suburban mobile-to-mobile environment operating at 900 MHz band according to the adjusted path loss models. On Table IV, the coverage range estimations in all three cases are for an outdoor-to-outdoor radio channel without any building penetration losses. In the first case, the maximum allowable path loss was 130 db incorporated with the planning margins. If the shadow fading margin is included to the dimensioning, then coverage range of 90 % area location probability corresponds to 410 meters in Kissanmaa, 340 meters in Tammela and 665 meters in Hervanta. The coverage ranges can be either observed on Fig. 1 - Fig. 3 or derived from (1) with proper propagation parameters. Without the 5 db planning margin, the coverage range estimation corresponds to 50 % point location probability at the edge of the coverage range. In this case, the range estimation is 550 meters in Kissanmaa, 420 meters in Tammela and 830 meters in Hervanta. However, if the offset correction c is omitted from the solution and it is assumed that the propagation would follow the reference model, then the coverage range estimations for 90 % area location probability would correspond to three kilometers in Kissanmaa, one kilometer in Tammela and 900 meters in Hervanta. This would be the case, if the measured signal levels were not biased and EIRP was exactly 30 dbm

5 as assumed. However, in practice, the local radio propagation conditions, obstructing obstacles and imperfect transceiver location affect the propagation greatly, causing offset to the empirical model. Therefore, the reference model can only be used as an upper bound when approximating the possible communication ranges. VI. CONCLUSIONS AND DISCUSSION This paper provides practical values for path loss models which can be used in planning of forthcoming mobile-tomobile wireless systems. Coverage dimensioning is essential and important part of wireless system design, verifying whether or not the designed system is capable to meet the given coverage requirements, therefore a proper propagation model is need for path loss predictions. The presented results verified that the path loss exponent for a typical suburban and urban environment is higher for a mobile-to-mobile environment compared with the base-to-mobile environment. In practice, the value of the path loss exponent can be assumed to be between 4 and 5 in suburban and urban environments at 900 MHz band. Even higher path loss exponents can be assumed for 2 GHz and 5 GHz mobile-tomobile frequency bands [14]. The standard deviation of shadowing was also investigated and it was found to be 6-8 db for a longdistance mobile-to-mobile environment. Similar values were observed in macro cells operating at 900 MHz band but over longer distances [15]. In addition, it was observed that the local propagation conditions affected quite much to the propagation of individual propagation paths, when comparing standard deviations of different Kissanmaa routes with the overall standard deviation in Kissanmaa. Even if the surroundings for two different propagation paths are similar, there can be significant differences in path loss values. Therefore, it is advisable to use many transmitter locations when characterizing mobile-to-mobile propagation environments with measurements. By doing so, the influence of near-tx surroundings and unfavorable transmitter location can be mitigated from the measurement results. A big difference in coverage ranges between the reference models and the adjusted models was also observed. This emphasizes the importance of using accurate propagation models for range dependable systems and applications operating in mobile-to-mobile environment. The difference between base-to-mobile and mobile-to-mobile environments is a consequence of two facts. Firstly, likelihood of the lineof-sight connection between a transmitter and a receiver is reduced when both of the antennas are placed close to ground level. Secondly, propagation in mobile-to-mobile environment is characterized with more challenging and dynamic propagation conditions. One target of the future studies is to understand how mobile-to-mobile radio environments should be characterized (e.g. is the average building density and height a good measure of the environment or should topographic variations be emphasized more?). Also, more measurement surveys are planned for verifying the behavior of the offset term c and developing better methods for measurements, data analysis, and propagation model tuning. ACKNOWLEDGMENT Authors would like to thank European Communication Engineering (ECE) Ltd. for providing measurement equipment, Mr. Teemu Pesu, Mr. Toni Levanen and Mr. Jukka Talvitie for guidance and help with the measurements, Elisa Oyj Finland for allowing the measurement campaign and former FM-kartta for providing the digital map. REFERENCES [1] R.S. Akki and F. Haber, A statistical model for mobile-to-mobile land communication channel, IEEE Trans. Veh. Technol., vol. VT- 35, no. 1, pp. 2-7, Feb [2] R.S. Akki, Statistical properties of mobile-to-mobile land communication channels, IEEE Trans. Veh. Technol., vol. 43, no. 4, pp , Nov [3] J. M.G. Linnartz, and R. F. Diesta, Evaluation of Radio and Networks, Tech. Rep. UCB-IT-PRR-96-16, University of California, Berkeley, PATH Research, California, [4] Kovács Z. I., Radio Channel Characterization for Private Mobile Radio Systems Mobile-to-Mobile Radio Link Investigations, Ph.D. Dissertation, 2002 Sep., Aalborg University, Denmark. [5] Rappaport, S. T., Wireless communications principles and practice, prentice-hall, 2002, chapter 4, pp [6] S. C. M. Perera, A. G. Williams and G. B. Rowe, Prediction of breakpoint distance in microcellular environments, IEEE Electronic Letters 8 th July 1999, Vol. 35 No. 14. [7] J. Lempiäinen, and M. Manninen, Radio Interface System Planning for GSM/GPRS/UMTS, Kluwer Academic Publishers, 2001, chapter 5, pp. 18, [8] P. E. Morgensen, P. Eggers, C. Jensen and J. B. Andersen, Urban Area Radio Propagation Measurements at 955 and 1845 MHz for Small and Micro Cells, IEEE GLOBECOM 91, [9] K. L. Blackard, M. J. Feuerstein, T. S. Rappaport, S. Y. Seidel S and H. H. Xia, Path loss and delay spread models as functions of antenna height for microcellular system design", IEEE 42nd Vehicular Technology Conference, [10] K. Sohrabi, B. Manriquez, and G. Pottie, Near-ground Wideband Channel Measurements in MHz. Proc. 49th IEEE Conf. on Vehicular Technology VTC, 1999, Vol. 1. [11] G. G. Joshi, C. B. Dietrich Jr., C. R. Anderson, W. G. Newhall, W. A. Davis, J. Isaacs and G. Barnett, Near-ground channel measurements over line-of-sight and forested paths, IEEE Proc. Microw. Antennas Propag., Vol. 152, No.6, December [12] F. Lotse and A. Wejke, Propagation Measurements for microcells in central stocholm, IEEE [13] K. Konstantinou, S. Kang, and C. Tzaras, A Measurement-based Model for Mobile-to-Mobile UMTS links, Vehicular Technology Conference, VTC2007-Spring. IEEE 65th. [14] Z. Wang, E. K. Tameh, A. R. Nix, Statistical Peer-to-Peer Channel Models for Outdoor Urban Environments at 2GHz and 5GHz, IEEE [15] S. R. Saunders, Antennas and Propagation for wireless communications systems, John Wiley & Sons, 2005, chapter 9, pp