Calibration of Propagation Model for Indoor Tunisian Environment

Transcription

1 SETIT 25 3r International Conference: Sciences Of Electronic, Technologies Of Information An Telecommunications March 27-3, 25 TUNISIA Calibration of Propagation Moel for Inoor Tunisian Environment Mohame AYADI*, Sami TABBANE**, Zia BELHADJ*** Aress of Mohame AYADI, Sami TABBANE & Zia BELHADJ Abstract: This paper presents the result of path loss measurements for an example of micro-cellular environment, which is the PALMARIUM commercial centre locate at Habib Bourguiba at 9 MHz ban. Measurements are taken with fixe transmitter with note book computer attache to allow ata an position logging in ile moe. Base station height of 3 m is teste with mobile receiver height of,7 m to emulate a typical micro-cellular scenario. Obtaine results are teste an compare with an Inoor empirical moel. Using linear regression metho, we have trie to etermine the value of ifferent parameters intervene in the moel like Wall attenuation factor, beak point istance an path loss exponent. Obtaine calibrate moel can be use in future Inoor cellular network planning. Keywors: Calibration, Cheung moel, Inoor propagation, linear regression metho. Introuction Inoor preiction of propagation is becoming useful to mobile telephone operator that provies services in large cities. There, many subscribers are peestrians emaning that coverage shoul be provie within builing, shopping areas, an airport an train stations. Combine this with the extremely height ensity of subscriber within these cities (making the raio spectrum resources very scare, mobile operators are invite to maximise their network capacity an hence performing accurate inoor propagation preiction. The accuracy of the fiel strength preiction epens particularly on the applie raio wave propagation moel. In literature we can fin many ifferent propagation moels for various environments, like urban an rural terrain. In the beginning, free space propagation moel has been propose. The receive fiel strength epens only on the istance between the transmitter, the receiver an the path loss exponent. The approximation can be rapily compute from one equation (Rappaport 22. A more realistic moel was propose by Okumura. It is essentially base on computing the free space path loss, then aing or subtracting correction factor to take into account the ifferent morphological features of terrain. The moel was mae applicable by Hata, who has publishe empirical values of the correction factors (Formella 2. Many alternative approaches have been propose such as: ray tracing an raiosity moel (Ayai &. Ta. 22. Although those moels are qualifie exact ones an give results, which are more precise than empirical moel, they require etaile site specific information about the particular builing. Moreover, large computation time on personal computer can not interact with algorithms for

2 SETIT25 optimizing base station location (Ayai &. Ta. 22. For these reasons, the use of ray tracing an raiosity is restricte to particular situation waiting for their optimization an amelioration. In this paper, we propose to use a precise empirical moel that has been evelope by (Cheung, & Sau997. We calibrate it for Tunisian inoor environment by etermining its optimal parameters using moern linear regression metho. In the secon section, we present the previous one work an the necessary backgroun for the use of our moel. In section 3 we escribe the environment, where measurements are taken, use tools, an obtaine results before concluing. 2 Previous Work Let enote the channel path loss, Pt the peak transmitter power, Gt the transmitter antenna gain an Gr the receiver gain. is given by (Ceung &Sau 997, Rappaort22. =Pt (Bm+Gt (Bi+Gr (Bi-Pr (Bm ( The most frequently moel inicates that mean path loss increases exponentially with istance (Rappaport 22. ( n (2 Where n is the mean path loss exponent, which inicates how fast path loss increases with istance. is a reference istance an is the transmitterreceiver separation istance. When we plotte on log-log scale, this power low relation ship is a straight line. Absolute mean path loss in ecibel (B is efine as the path-loss in ecibel from the transmitter to reference istance plus aitional path moss escribe by: ( B = Pl( + nlog( (3 ( is ue to free space propagation from the transmitter to m reference ( =m from which all measure path loss value refers (Missaoui & Ta 2. Pl ( (4 = m = 2log( 4π λ ( is equal to 3,52 (B at 9 Mhz. 2.. Inoor Log-normal Shaowing The moel in equation (3 oes not consier the fact that the surrouning environmental clutter may be vastly ifferent at two various locations having the sale transmitter receiver separation. This leas to measure signal, which are vastly ifferent than the average value preicte by equation (3. Measurements have shown that at any value of, the path loss ( at a particular location is ranom an istribute log-normally about the mean istance epenant value (Missaoui & Ta 998. ( B = Pl( + nlog( +χσ (5 Where χ σ a zero-mean Gaussian is istribute ranom variable (B with stanar eviation σ (B. The reference istance, the path loss exponent an the stanar eviation are statistically eterminate for any arbitrary location having a specific transmitter receiver location. In practice the values of n an σ are compute from measure ata, using a linear regression metho such that the ifference between the measure an estimate path loss is minimise in mean square error. 2.2 Attenuation factor moel Although log-normal shaowing is the most use one, it suffers from the lack of accuracy since it oesn t take account for specific effects in builing. As alternative, another moel has been propose by (Cheung, & Sau997. which takes account of aitional attenuation cause by walls an floor. n ( B = log( + WAF+ FAF (6 P Where P an are the numbers of walls an floors between the transmitter an the receiver, WAF an FAF are respectively Wall Attenuation Factor, an Floor Attenuation Factor. Although this propose moel performs well in certain circumstances, it suffers from limitation. It oes not inclue propagation effect such as istance epenence of path loss exponent, angle epenence WAF, FAF, reflection an iffraction. As result, preiction accuracy can be poor in certain part of builing an especially at large istances from transmitter (Cheung, & Sau Distance epenence of path loss It is evient that propagation loss as a function of istance has two istinct regions. In the first region, which is within 5-2 m of the transmitter, propagation loss is similar to that occurring in free space (Cheung, & Sau997, because at a very near istance from the antenna, obstruction such as wall an floor o not interact significantly with propagation waves. At a further istance in the next region, the propagation loss increases as the electromagnetic waves become obstructe by ceiling or walls of the room in the builing. The istance at which this

3 SETIT25 transition in propagation loss occurs is referre to the break point (Cheung, & Sau997. This break point phenomenon is incorporate into (6 by moifying the equation to (Cheung, & Sau 997: ( B = log( n U + n + P ( log( log( * U( + WAF+ FAF (7 Where, n an n2 are path loss exponent either sie of break point an U(. is unit step function efine as: if < U ( = if >= The aitional parameters, n,n2 introuce can be obtaine from the use of Fresnel zones(cheung & sau 997. By consiering the size of the first Fresnel zone a istance from the transmitter an etermining at what istance it will become obstructe the peak point can be calculate. The iameter of the first Fresnel zone at a istance from the transmitter is approximately given by. Z f = (λ /2. Consequently in a corrior of with of.5 m the first Fresnel zone will be obstructe at istances greater than 7.5 m for frequencies aroun 9Mhz (Cheung, & Sau Angle epenence of attenuation When electromagnetic raiation is obstructe on a wall or floor obliquely, less power will be transmitte, than it woul occur at normal incience. By incorporating the angle epenence attenuation through floor an walls, equation (7 will be (Cheung, & Sau997: = n + ( B log( U( log( P * U( + WAF( p/cosθ p + FAF( q/cosθ f Where WAF(p an FAF(q are the value of attenuation factor at normal incience an the θ p an θ q are the angles, which are respectively between the p th wall an q th floor an the straight line path joining the transmitter to the receiver. 2.5 Description of the environment. + log( Propagation preiction using the new moel has been one in Palmarium shopping centre, which is locate at Habib BOURGUIBA Street in Tunisia. The PALMARUIM is constructe in 997. It consists of three floors of approximately 876 square metre. The construction of the builing is a concrete block an a glass wall. The floor to floor height is 3m with suspene ceiling containing an air conitioning. To perform propagation site specific, information was obtaine from builings plan, which are treate n n2 n2 with AutoCAD an aequate evelope application, we convert plans into specific format that store all information relative to walls co-orinates, total number of walls, nature of wall material, etc... To compare our preiction to measurements, 387 samples are taken on many targets with a uniform istance between samples at 9 MHz with the transmitter in the same location as for the preiction The following figure gives the ifferent targets of measurements, base station location in PALMARUIM centre. Figure: PALMARUIM centre The results of measurement are presente in the following graph. Figure 2: Measurements versus istance 3 Calibration of The Moel 3. Calibration of the Log-normal shaowing moel Applying the moel given by (5, the problem is to etermine the optimal values of path loss exponent, anχ σ. To solve this problem linear regression metho is use. Equation (5 can be written: ( B = A+ Blog( +χσ (9

4 SETIT25 A an B are the constant to etermine. Their values are consiere constant within given frequency an antenna height. : is the transmitter receiver in m. Let i enote the path loss given by the moel at istance i an mes enote the measure value of path loss at the same istance i. For one sample (9 can be written (John 998: (, Log( i i * ( = ( B A For N samples the system is: =M*Y Where: i=... n ( Log Log ( 2 = Y= B A Log( N M ( For each istance i the error between measurements an preicte value is given by: e i= i ( Ã+ B* Log( i ( For N samples the equation system can be e= mes written: (2 where: M*Y = e e e... 2 en an Y A B = Optimal parameters à an B are foun by minimising (John 998. e t e ete= ( mes M* Y t *( mes M* Y (2 ( e t e Y if we write = we obtain M t*( M* Y mes opt Optimal value of Y is given by: = Y = ( M tm M tmes (3 As mentione, optimal value of Y are obtaine by solving the equation (3. Applying this system equation to measurement, we have foun Ã= 28,5. This means that the exponent path loss is equal to n=2,8 an B=37,37. Having n=2,8, which is between 2 an 4, is an expectable result. It can be explaine by the interior walls of this centre, which are mae of glass. This make the attenuation factor less. Figure 3: The foun moel The stanar eviation χ σ which is suppose a zero-mean Gaussian istribute ranom variable, is about 4,47 B. 3.2 Calibration of Cheung Moel Cheung moel is completely escribe by (8. To make the new moel applicable an simple to use, some hypotheses shoul be aopte. As mentione the path loss exponent n is use for istance less than the break point istance. At these istances, propagation is similar to free space. So n shoul be about 2. once antenna effects are remove. The break point istance in a corrior of with.5 m is about 7.5 m for frequencies aroun 9Mhz. In a room, where there is more clear space, this istance is approximately 2 m. To keep our moel simple to use we propose to take break point istance value between these limits an use m for the break point in all situations. Since measurements are taken in the same floor, the floor attenuation factor FAF is consiere equal to an oesn t intervene in equation The problem refers back to fin the optimal value of the wall attenuation factor (FAF an the path loss exponent n2. For a istance greater than the break point istance, (8 can be written: Where ( B = A+ Blog( + C log( A= n B= n2 C = p* WAF (4 P is the number of wall crosse by the straight line joining transmitter to receiver. Applying the linear regression metho as efine in the previous section, obtaine results show that n2 have the value of 2,4 an WAF is about 7,6 B. The stanar eviation of the error between preicte an measurements value is about 8 B.

5 SETIT25 As we can notice, the stanar eviation of the Cheug moel is less than of the log normal shaowing moel. This result can be interprete by the fact that the new moel takes account of supplement attenuation, introuce by the wall an the floor. 4. Conclusion In this paper we have presente an calibrate two empirical moels for preiction of propagation inoor environment. The first one is the log-normal shaowing: in this moel, path loss is a function of transmitter- receiver istance, path loss exponent an ranom variable that represents the istribution of supplement attenuation introuce by shaowing. Using linear regression metho to calibrate the moel for Tunisian environment, we have foun that path loss exponent is equal to 2,8 an the stanar eviation of error is equal to4,47. The secon moel is the Cheung moel, it takes account of many supplements attenuations cause by walls an floor, break point an path loss exponent for istance less or greater than break point istance. The calibration of this moel has shown that the path loss exponent for greater istance than the break point is about 2,4 an the stanar eviation is about 8 B. This result can be interprete by the fact that the new moel takes account of supplement attenuation introuce by the wall an the floor. The calibrate moel can be use for future inoor cellular network planning as airport, commercial centre. This stuy can be etaile in orer to foun the exact value of wall attenuation factor in each type of material. Once one, results given by the new moel can be more accurate. Nabeul- Hammamet, Tunisia, April 998, pp6-65. John O. R., «Applie Regression Analysis, A Research Tool», Secon Eition, Springer Eition998. H. Leib, V. Bohassian, M.Kimpe, Ray tracing for inoor raio channel estimation, IEEE Transactions on Antenna Propagation, Vol 42, Mars993. Francois Sillon et Claue Peuch, Raiosity an Global illumination Morgan Kaufmann Publisher 995. K. W. Cheung, J. H.M.Sau an R. D. R. Murch, «A new empirical moel for inoor Propagation»,, IEEE Transactions on Vehicular Technologies, September 997. T.S. Rappaport, Wireless Communication Principles an practice Prentice Hall Eition, Eition 22. References M. Ayai, S. Tabbane Comparison of raiosity an ray tracing in micro-cellular urbain environment Algotel 22, Mai 22 M. T. Missaoui et S. Tabbane, «Moélisation e la Propagation Raio ans l Environnement e l Afrique u Nor : Application aux Réseaux Cellulaires», INPTAMSE,Rabat 9-2 Mars 2. pp Arno Formella, «raio tracer A tool for Deterministic Simulation of Wave Propagation» Universitat es Saarlanes germany. URL : formella 27//. M. T. Missaoui an S. Tabbane,«Moelling Raio Propagation in the 9 MHz Frequency Ban in a North African Urban Environment», CESA'98, IEEEIMACS Multiconférence