Relative Signal Strength Coverage Optimization in Indoor and Outdoor Wireless LAN Environments

Transcription

1 Relatve Sgnal Strength Coverage Optmzaton n and Wreless LAN Envronments Moses E. Ekpenyong, UduakObong M. Umoren Department of Computer Scence Unversty of Uyo, PMB. 17, 53, Uyo, NIGERIA and Udoma J. Bassey Department of Mathematcs and Statstcs Unversty of Uyo, PMB. 17, 53, Uyo, NIGERIA ABSTRACT Fadng and obstacles consttute maor threats to effectve qualty of servce (QoS) delvery n wreless local area network (WLAN) envronments. In ths contrbuton, we nvestgate the sgnal qualty of ndoor and outdoor WLANs over a defned coverage area. We present expermental analyss of case studes that wll be useful for further research and valdate the system s performance n practce. Usng an optmzed form of the pathloss models, a smulaton of the system s carred out over short and extended coverage. Smulaton results show that sgnal qualty could be effectvely managed to mprove the system s performance for both ndoor and outdoor envronments n the presence of fadng and other envronmental factors. Keywords: receved sgnal power, multpath fadng, pathloss, lnk qualty, network coverage. 1. INTRODUCTION One maor challenge encountered n the feld of wreless communcatons s the ablty to deal wth the phenomenon called fadng. A useful approach to tacklng ths problem s to mplement predcton models, takng nto account the transmtter, recever, dstance and sgnal propagaton parameters of the study envronment. Understandng the propagaton characterstcs of WLANs s essental for effectve network deployment, as ths offers network operators a clue of the coverage capacty of the access ponts based on ther locatons and possbly elmnatng the need for ste surveys. As such, the strength, range and coverage area of an access pont are mostly affected by ts postonng n reference to the envronment [1]. Obstacles do also mpede sgnal propagaton n WLAN envronments. Ther presence could reflect, refract, dffract, scatter or absorb sgnals [-3]. The rate of sgnal defect n ths case wll largely depend on the type and constructon of the obstacle(s). In a wreless LAN, apart from the frequency spectrum, the receved sgnal strength (RSS) determnes the lnk or connecton qualty of the network system. In today s technology, the sgnal strength of a WLAN can be boosted usng sgnal amplfers, thus extendng the reach of the sgnal. Apart from amplfers, antennas and access ponts also contrbute to effectve sgnal propagaton. However, the strength of a sgnal s hndered by the dstance between the transmttng and recevng devces. Therefore, the sgnal reach of a wreless LAN largely depends on ts archtectural desgn. One key to solvng the problem of weak recepton s to enhance the sgnal strength and reduce the nose level. However, hardware devces such as cordless phones and mcrowaves are common culprts of ncreased nose levels. Wreless access ponts have bult-n WF antenna that emts sgnal unformly n all drectons over a dstance range of 5-3 feet (7-91 meters). Antennas that exhbt such behavour are called omn-drectonal antennas. An omndrectonal antenna produces constant feld strength n azmuth (horzontal), but can have a drectonal radaton pattern n elevaton (vertcal). It dffers from the sotropc antenna, whch produces constant feld strength n both azmuth and elevaton. The sgnal transmtted by omn-drectonal antennas can be weakened over long dstances due to nterference and because the antenna s sgnalng power radates n all drectons. A condton referred to as attenuaton descrbes the loss n sgnal strength as sgnal travel farther from ths devce. A ground plane (an electrcally conductng surface that serves as the near-feld reflecton pont for the antenna) could be constructed to help boost the sgnal. Materals used for constructng ground planes are large metal sheet, wres or rods. The conductors should be desgned to radate from the base of the antenna. Fxng dead spots n a buldng can be acheved by nstallng an AP wth two extended antennas. A repeater could also be used to extend the sgnal coverage. Repeaters are used to rebroadcast sgnals from the current wreless APs to clents. But these devces halves the mum bandwdth because both receve and transmt frames are processed usng the same rado frequency (RF) channel, whch effectvely doubles the number of transmtted frames. The deployment of Voce over Internet Protocol (VoIP) and TCP/IP-based applcatons on WLAN consttutes a maor deployment dffculty due to the qualty of servce demands by the network [] for several reasons: 9

2 () () () (v) The varablty of WLAN QoS parameters (bandwdth, packet loss, delay and tter) s hgh n realstc systems. Ths condton leads to nstablty wthn the network and sgnfcantly reduces the applcaton performance. Contenton s lkely to occur when two or more servces share same communcaton channel. Multmeda applcatons such as VoIP or vdeo streamng requre tmely servcng of the traffc. Ths s a challengng task n WLANs, even when QoS enforcement s mplemented, as most exstng QoS mechansms concentrate on bandwdth provsonng. Roamng between access ponts, a typcal WLAN event, ntroduces communcaton gaps that may cause servce nterruptons, an undesrable state for real-tme applcatons. Consequently, a detaled analyss s requred to nvestgate WLAN applcatons deployment, especally when consderng ther use n envronments wth specfc requrements such as safety/msson-crtcal systems or dsaster management. Ths paper therefore adopts an expermental approach to optmzng the sgnal qualty of ndoor and outdoor WLANs, subect to network coverage constrants between the transmttng and recevng devces. A startng pont for ths paper s to conduct a study of exstng WLAN envronments (wth network applcatons) to understand the basc propertes of the system and ther nteracton wthn the network. To acheve a hgh degree of realsm, we shall combne the observatons n real lfe wth the analytcal model and arrve at an optmzed model adapted to the study envronment and sutable for easy predcton of the system. The feld measurements are obtaned from exstng ndoor and outdoor WLANs wthn Uyo (urban) Metropols n Akwa Ibom State, Ngera. The essence of data collecton s to enable us predct the system wth some degree of confdence.. RELATED RESEARCH Due to growng nterest n moble data applcatons, In recent years, WLANs have been successfully deployed worldwde and wll probably be used to nterconnect cellular systems for the provson of hgh speed data servces n the next generaton (fourth generaton) wreless networks, where handovers between WLAN cells and cellular cells are possble [5]. Several researches have been carred out n the area of wreless LAN. In [], the sgnal strength of.11n WLAN equpment s studed. They foresee an upgrade n the performance of the equpment, as well as predct the network coverage as an extenson to the new standard usng a mathematcal model. In [7], expermental data that valdates the use of the log dstance path-loss model for dealng wth sgnal attenuaton s presented. Measurements were performed usng off-the-shelf IEEE.11b hardware at dstance varyng from 1 to 5 meters. Also, the model was used to predct the sgnal strength wthn and around a standard offce envronment. In [], studes of the WLAN and Bluetooth pconet range nterference are presented. The authors propose an nterference range model that allows a user to determne an acceptable range of nterference n a gven envronment. In [9], the RSS s used to estmate the poston of a moble node nsde a buldng (ndoor). In [5], measurements of the RSS of WLAN beacons n an ndoor envronment are presented and from theoretcal analyss, Gamma random varables are used to model the RSS varaton from the average power. Hs results show that the theoretcal model correlates well wth measured data. In [], a study of Georga Insttute of Technology has been carred out usng drect-ray, solo path-loss exponent, adapted to the Sedel- Rappaport propagaton model wth.ghz, 1.11g outdoor WF network deployment. The standard devaton of the predcton error for ther proposed model s approxmately 5.5dB on the average, whch agree wth other path-loss models n the outdoor doman [11]. Lechty [1] outlnes the acheved predcton accuracy of the propagaton model used n [], by measurng and analysng an establshed. GHz,.11g outdoor WF network deployed on the campus (.e. the Georga Insttute of Technology). The proposed model performs wth accuracy compared to other models and offers a smple desgn wth a strong predctve model for network plannng and deployment. One maor aspect of assessment that research lterature has gnored, s accountng for low and peak network loads and a relatve analyss and modelng of both ndoor and outdoor envronments. We therefore study these WLAN envronments wth the goal of optmzng the receved sgnal qualty from same dstances away from the access ponts of these networks and then derve optmzed performance models for effectvely assessng the network s QoS. 3. EXPERIMENTAL SETUP AND ANALYSIS OF EXISTING INDOOR AND OUTDOOR WLAN Expermental Setup The experments descrbed n ths paper were conducted n typcal offce envronments. The setup n Fgure 1 shows the locaton of exstng masts and brdges n the Unversty of Uyo. But the WLAN buldng (envronment) we are dealng wth s ndcated usng an arrow.

3 Average Average envronments. The data measurements were done at the opposte sde of the buldng where there was free-space for measurements. Durng data collecton, we classfed the measurements nto lnk status (sgnal and nose) and data packets transfer (data sent and receved). Daly measurements of these parameters were taken for a perod of two weeks at three dfferent ntervals (mornng, afternoon and evenng) - to accommodate low and hgh traffc. Fgure 1. Expermental setup for the ndoor wreless LAN Fgure shows the expermental setup for the outdoor WLAN. The WLAN buldng s also ndcated wth an arrow. Fgure. Expermental setup for the outdoor wreless LAN The requred data were captured usng a Dell wreless LAN card utlty analyzer/montor (software). Ths card s controlled wth specal DELL drver nstalled on a clent laptop whch permts the collecton of RF montorng mode measurements. In ths mode, the card s prohbted from assocatng tself wth any access pont (AP), but nstead scans avalable Wreless Fdelty (WF) channels and dsplay measurements of the RSS of the WF selected, ndcatng the RSS for each AP t samples. Wth ths method, t s possble to measure all APs whose RSS s wthn the dynamc range of the card at any gven pont. The ndoor envronment operated at a frequency of.ghz and ts sgnal was transmtted va a Lnksys Wreless-G access pont (WRT5G) wth.11g wreless network standard and compatble wth.11b standard wth mum data transfer rate of 5Mbps. The outdoor sgnal transmtted usng three sector beam formng antennas at 9 o, each facng dfferent drectons wth connected amplfers. Its transmttng frequency also operated at.ghz. Both transmttng routers were set to nfrastructure mode. The data were obtaned over a dstance of m and m n steps of 1m and m respectvely, from the access ponts n both envronments. The reason for choosng same measurement lengths s that both case studes had same operatng capacty and the necessty of obtanng balanced readngs from both Analyss of expermental data In Fgure 1, we observed a sharp reducton n sgnal qualty between m and m from the access pont of the Student care café (WLAN ndoor) envronment. After the m mark, the sgnal stablzed tll the m mark. Ths sudden drop n sgnal qualty could be attrbuted to the huge actvty and frequent moble obstacles (movement of persons) wthn the envronment, as the locaton of the WLAN s at the entrance of the Unversty of Uyo and close to the hostels. Also at Webcenta (outdoor WLAN) envronment, the network experenced sgnal degradaton between 3m-m dstance, as observed n Fgure. Ths could be due to the non-lne of sght notced durng feld measurements or readngs, as the mast of the WLAN was obstructed by the offce buldng, thus obscurng sgnal transmsson. Average readngs were compled for both envronments. These readngs wll be used as predctors n the optmzed pathloss model and wll enable us predct the performance of both systems Perod (days) Fgure 1. Analyss of observed sgnal and nose data for Student-care café (ndoor) WLAN envronment at varous dstances for the study perod Perod (days) Sgnal data1 Nose data Sgnal Nose Fgure. Analyss of observed sgnal and nose data for Webcenta (outdoor) WLAN envronment at varous dstances for the study perod 11

4 Expermental Pathloss Models Pathloss models are used to predct the coverage area and sgnal propagaton of the envronment. Presented n Fgure 3 s a graph comparng the computed mean pathloss of the ndoor and outdoor WLAN envronments. Tradtonally, pathloss s computed usng the relaton: PL = Pt S (1) p where P L s the pathloss P t s the transmtter power S p s the sgnal power Extractng the average pathloss for both ndoor and outdoor envronments usng equaton (1), and plottng these losses aganst dstance, yelds Fgure 3. Transmtter powers obtaned from the feld - 15dB for ndoor WLAN and db for outdoor WLAN, were used to compute the respectve pathlosses. We observed that both ndoor and outdoor wreless LANs ntersected (experenced same pathloss) at (3m,.dB), but on the average, the degree of pathloss was hgher n the ndoor WLAN. Generally, the pathloss tend to follow same pattern n both envronments. For the purpose of further research, the computed pathlosses for both envronments are presented n the Appendx. Mean pathloss (db) y = 11.5ln(x) +.15 R² =.9 y = 3.ln(x) R² =.5 Log. () Log. () Standard devaton (db) y = -.71Ln(x) R =. y = -.11Ln(x) R =.3 Log. ( ) Log. () 1 (m) Fgure. Comparatve analyss of standard devaton vs. dstance for ndoor and outdoor WLANs The average standard devaton was obtaned as 1.59 for ndoor envronment and 1.19 n outdoor envronment, whch confrmed the effcency of the outdoor WLAN over ndoor WLAN.. OPTIMIZATION MODEL Notaton The followng notatons are used to descrbe the optmzaton model: a, = 1, L, N : Access Pont (AP) r, = 1, L, M : User/recever pl ( a, r ): Pathloss from user r to AP a pl : mum tolerable pathloss We assume here that a has unknown coordnates, whle the coordnates of users r are assumed to be known and users are dstrbuted n desgn area accordng to the desgn specfcaton.. 1 (m) Fgure 3. Comparatve analyss of mean pathloss vs. dstance for ndoor and outdoor WLANs From the trend lne equaton, we construct a smple predctve pathloss model wth log dstance component as: PL = α + nlog( d) () where α represents envronmental constant n s the propagaton or pathloss exponent d represents the dstance Note that n s the slope of the ftted curve. Therefore the pathloss exponents for the ndoor and outdoor WLANs are 1.15 and.3 respectvely. Studyng the varablty n both envronments, we observed that both WLANs ntally showed hgh varablty. Ths could be attrbuted to sgnal attenuaton and nature of obstacles surroundng the envronments. Model Descrpton In ths paper, the obectve functon s based on pathlosses. The pathloss for each recever should satsfy the followng condton: pl a r pl 1, L, (3) mn = 1,..., N ( ) M, So, we state our frst constrant from equaton (1), that pathloss s evaluated aganst the mum tolerable pathloss pl. Ths ensures that at each recever s locaton, the sgnal qualty does not fall below a certan threshold. The threshold value, pl can be computed by subtractng the recever s threshold ( R th ) from the transmtter power P t, thus: pl = Pt R () th Equaton () can be expressed as an equalty form, thus: + ( (5) mn pl a, r ) pl = = 1,..., N 1

5 where ( γ ) + = ( γ,) Therefore, a feasble soluton s possble ff: M + mn (, ) () pl a r pl = = 1 = 1,..., N Pathloss Model Followng from Equaton (), the pathloss functon at a dstance d ( a, r ), above the reference dstance ( d ) s gven by: ( a, r ) d ( )[ ] ( )[ ] (7) pl a, = + r db pl d db log d pl s equal to the free space pathloss Thus, ( ) d wth respect to d [1, 13-15], d s typcally 1m, m or 1km, dependng on the envronment [1]. The RF path between the transmtter and the recever s nfluenced by the separatng dstance and the nature of obstacles (doors, walls, furnture, people, etc) scattered around the envronment. Includng the loss caused by these obstacles, we rewrte equaton (7) as: d ( ( a ) n, r )[ ] = ( )[ ] + () pl a, r db pl d db log + obs l d = 1 where obs represents the obstacle and l represents loss n db, created by the obstacle. Observe that equaton () s dscontnuous because of the presence of obstacles. Log Normal Shadowng Equaton () does not consder the dfference n receved sgnal power for same transmttng dstances. Ths dfference s caused by dfferent envronmental dsorders whch may dffer at two dfferent locatons havng same dstance from the transmtter. Equaton () provdes the expected mean sgnal strength when the separated dstance s d. The actual receved sgnal strength may surround ths mean value. Practcal measurements show that the dstance d at any gven locaton s random and log-normally dstrbuted about the mean dstance value. Consderng ths, equaton () becomes: d ( ( a, r ) pl a, r )[ db] = pl( d )[ db] + log + (9) n =1 where obs l + X σ X σ s a zero mean Gaussan dstrbuted random varable measured n db wth standard devaton σ, also measured n db (.e. X σ ~ N(,1) or M = e ( t ) Xσ x / d ). Ths varaton or loss n sgnal strength caused by blockage or absorpton n envronments from ponts of equal dstance to the transmtter s referred to as shadow fadng. The Gaussan random varable s added to the pathloss to compensate for unpredctable shadowng. Equaton (9) s sutable for modelng pathlosses under dverse envronmental condtons. Next, we adapt equaton (9) to our case study envronments. Frst, we account for the pathloss exponents n both envronments, by ntroducng the parameter n, representng the pathloss or propagaton exponent. Substtutng ths nto equaton (9) yelds: d( a, r ) pl( a, r )[ db] = pl( d )[ db] + n log + () n =1 obs l + X σ We wll smulate equaton () usng observed data means as predctors to the optmzed model, for short and extended dstances. 5. MODEL SIMULATION AND DISCUSSION OF RESULTS Expermental Data Analyss In ths secton, we analyze the expermental data under the followng QoS parameters: () Sgnal and nose levels: In both case studes (ndoor and outdoor envronments), the commercal nature of the envronments caused nstablty n nose levels. Ths s seen n Fgures 1 and respectvely. In Fgure 1, sgnfcant spkes were notced n the ndoor WLAN envronment on days 5, and 9 at the th, 5 th and th meter dstances, respectvely. In the outdoor envronment, there was dsrupton at the 3m mark. These defects could be attrbuted to dsturbances by the wnd and the presence of obstacles. Generally, the sgnal strength deterorated as the recever went farther from the transmtter. () Sgnal-to-Nose Rato (SNR): The.11b/g WLAN performance s observed to deterorate wth ncreasng dstance and obstacles. In Fgure 5, the nose level ncreased n the ndoor envronment as the recever went farther from the transmtter. Ths occurred between m-m, after whch the SNR stablzed. The outdoor WLAN experenced more SNR dstortons compared to the ndoor WLAN, because of the ncreased network capacty (number of users the network can support). Average SNR d Duraton (days) StudentCafe() Wencenta() Fgure 5. Comparatve analyss of average SNR n ndoor and outdoor WLANs () Average data sent: Apart from the th and 7 th day when the average data sent drastcally dropped

6 due to system downtme, the readngs tend to dffer slghtly from each other n the ndoor envronment. In Fgure, the average data sent s generally hgher over all dstances n the ndoor envronment than n the outdoor envronment. Average data sent Perod (days) Webcenta () Student-care () Fgure. Comparatve analyss of the average data sent n both ndoor and outdoor WLANs (v) Average data receved: In Fgure 7, a drop n receved data rate was notced on the th day n the outdoor WLAN. But n general, data receved tend to stablze over the study perod n both envronments. Average data receved Duraton (days) 15 Webcenta() StudentCafe() Fgure 7. Comparatve analyss of the average data receved n both ndoor and outdoor envronments Model Smulaton The optmzed pathloss model s expected to reveal the true behavour of the real system. In order valdate the model, data already acqured from the exstng (real) system were used as nput to a program smulatng the model, and coded n MATLAB. These data are shown n Table 1. Table 1. Smulaton parameters/nput data Parameter Value envronmental constant.15db envronmental constant 7.5dB Pathloss exponent for ndoor 1.15 Envronment (n) Pathloss exponent for outdoor.3 Envronment (n) (d) 1m-m, m-m Standard devaton (σ ) 1.59 for ndoor, Zero mean Gaussan random numbers ( X ) σ for outdoor N(,1.59) for ndoor, N(,1.19) for outdoor The model was then smulated over short (1m-m) and long (m-m) dstances to ease the study of the overall effect of pathloss on both systems.. DISCUSSION OF RESULTS In ndoor envronments, t has been observed that the sgnal strength s mpared by complex propagaton defects such as reflectons, refractons and multpath effects [17]. Ths accounts for the wde varaton n sgnal strength across the envronment, and s akn to our observaton n Fgure 9. In Fgure, we observed that for short dstance range (1m- m), ndoor WLAN experenced lesser pathloss compared to outdoor WLAN. We can attrbute ths trend to the fact that the senstvty of ndoor equpment tends to be stronger around the vcnty where the rado s nstalled to complement for certan propagaton defects. Smlar plots, but wth extended coverage range (m-m) s shown n Fgure 9. Here, the performance of ndoor WLAN deterorates faster as the dstance ncreases, compared to outdoor WLAN. Ths calls for a check of the senstvty level of transmttng devces, as well as proper network plannng technques to mprove the expected qualty of servce. Comparng Fgure 9 to Fgure 3, we notce that the smulated/predcted performance (Fgure 9) for ndoor envronment s hgher than the emprcal plot (Fgure 3). The wde devaton reveals the absence of fadng predcton n the exstng system or the nablty of the ndoor devce to transmt beyond a certan dstance. As expected, both plots follow same trend, as the outdoor WLAN outperformed ndoor WLAN. Important mplcatons can be drawn from the smulaton. It s observed that the optmzed model resulted n less varance assocated wth the poston and range plot, and thus yeldng better localzaton. The optmzed pathloss models can also reduce the cost of actve ste surveys. Ths s acheved n ths paper by usng parameters relatng to the study envronment. Wth the smulated models, more producton runs could be made to predct the sgnal attenuaton between the transmtter and recever at dfferent sgnal propagaton (dstance) range, wth the ncluson of parameters that concerns the terran profle, and ts surface features. We have observed that usng feld measurements of the study envronment as model predctors s advantageous because t accounts for the envronmental factors, regardless of separate parameter recognton [1]. Pathloss(dB) Graph of Pathloss vs (m) Fgure. A graph of Pathloss vs. dstance, for ndoor and outdoor WLANs 1

7 Pathloss(m) Graph of Pathloss vs (m) Fgure 9. A graph of pathloss vs. dstance, for ndoor and outdoor WLANs From the results we dscovered that most WLAN operators are most lkely to ms-predct the network performance, thus gvng room to wrong problem detecton. 7. CONCLUSION AND FUTURE WORKS We have modeled the sgnal strength of ndoor and outdoor WLAN envronments, by studyng the propagaton characterstcs of both envronments, consderng fadng effects as well as obstacles. In tacklng these ssues, we took measurements of QoS parameters over a defned dstance and compute the pathloss for both envronments. An optmzed model that predcts the sgnal qualty of WLANs over varous dstances was then derved from the experment. Expermental data obtaned from the feld were used to valdate the models. In both envronments, we could predct the dstance at whch wreless sgnal qualty receved was optmal. We observed that the lnk qualty (SNR) degrades wth dstance and other envronmental factors. The fadng phenomenon was also approached wth the am of profferng a practcal soluton. Ths approach was accomplshed by montorng the sgnal qualty coverage (at varous dstances). As an outlook, we shall nvestgate ndoor localzaton and the problem of nterference. 7. REFERENCES [1] M, Iskander and Z. Yun, Propagaton Predcton Models for Wreless Communcaton Systems, IEEE Transactons on Mcrowave Theory and Technques, Vol. 5, No. 3,, pp [] D. Hucaby, CCNP BCMSN Offcal Exam Certfcaton Gude. Fourth Edton, CISCO Press, 7. [3] G. Durgn, T. Rappaport and H. Xu, Measurements and Models for Rado Pathloss and Penetraton Loss n and Around Homes and Trees at 5.5 GHZ, IEEE Transactons on Communcatons, Vol., No. 11, 199, pp [] R. Beuran, L.T. Nguyen, K. T. Latt, J. Nakata, and Y. Shnoda, QOMET: A Versatle WLAN Emulator, In Proceedngs of The 1st IEEE Internatonal Conference on Advanced Informaton Networkng and Applcatons (AINA 7), Nagara Falls, Ontaro, Canada, pp , 7. [5] W. Zhang, Modelng of WLAN Beacon Sgnal Strength Measured n an Envronment. In Proceedngs of the Internatonal Conference on Wreless Networks, ICWN 3, Las Vegas, Navada, USA, pp. 7-75, 3. [] J.N. Daves, V. Grout and R. Pckng, Predcton of Wreless Network Sgnal Strength Wthn a Buldng. In Proceedngs of the 7 th Internatonal Network Conference, Unversty of Plymouth, pp ,. [7] D.B. Fara, D. B., Modelng Sgnal Attenuaton n IEEE.11 Wreless LANs Vol. 1. Techncal Report TR-KP-11, Kw Proect, Stanford Unversty, 5. [] P. O. Bobbe, and A.L. Yussf, Modelng and Smulaton of IEEE.11 Wreless-LAN and Bluetooth Pconet Range Interference. In Proceedngs of IASTED CIII Conferences, St. Thomas, Vrgn Islands, pp. -9,. [9] K. Sayrafan-Pour and D. Kaspar, A Novel Modelbased Postonng Usng Sgnal Strength. In Proceedngs of 1 th Annual IEEE Internatonal Symposum on Personal, and Moble Rado Communcatons (PIMRC 7), Athens, pp. 1-5, 7. [] P. Petrus, J.H. Reed, and J.S. Rapport, Geometrcally Based Statstcal Channel Model for Macrocellular Moble Envronments. In Proceedngs of IEEE Global Telecommuncaton Conference, pp , 199. [11] S.Y. Elchakw, Propagaton Predcton and Measurement on Wreless LAN.GHz Applcatons at Unverst Teknolog, Malaysa, M.Sc. Thess. Unverst Teknolog, Malaysa,. [1] L.C. Lechty, Path Loss Measurements and Model Analyss of a. GHz Wreless Network n an Envronment. M.Sc. Thess, Georga Insttute of Technology, 7. [13] K. Pahlavan and K. Prashant, Prncples of Wreless Networks: A Unfed Approach. Prentce Hall, New Jersey, USA,. [1] M. Panwan., A.L. Abbott and T.S. Rappaport, Interactve Computaton of Coverage Regons for Wreless Communcaton n Multfloored Envronments, IEEE Journal on Selected Areas n Communcatons, Vol. 1, 199, pp. -3. [15] R. Morrow, Wreless Network Coexstence. McGraw Hll, USA,. [1] J. B. Anderson, T.S. Rappaport and S. Toshda, Propagaton Measurements and Models for Wreless Communcaton Channels. IEEE Communcatons Magazne, pp. -9, [17] M. Hassan-Al and K. Pahlavan, A New Statstcal Model for Ste-specfc Rado Propagaton Predcton Based on Geometrc Optcs and Geometrc Probablty, IEEE Transactons on Wreless Communcaton, Vol. 1, No. 1,, pp [1] V. Gupta, S.C. Sharma and M.C. Bansal, Frnge Area Path Loss Correcton Factor For Wreless Communcaton, Internatonal Journal on Recent Trends n Engneerng, Vol. 1, No., pp Receved: July 1. Accepted: December 1. 15