r: 1 £/3(!!/ !lEGS VTC... 973 In Situ Microwave Reflection Coefficient Measurements for Smooth and Rough Exterior Wall Surfaces I I I i. I Orlando Landron 1, Martin J. Feuerstein 2, and Theodore S. Rappaport 1 [ 1) Mobile and Portable Radio Research Group (MPRG) Bradley Department of Electrical Engineering Virginia Polytechnic Institute & State University Blacksburg, VA 24061 [2] U.S. West Advanced Technologies Wireless Communication Services 4001 Discovery Drive. Suite 320 Boulder. CO 80303 (perpendicular) to the plane of incidence. as shown in Fig.!. The properties of each dielectric at the interface is characterized by its perminivity (E), magnetlc permeability(~) and conductance (0'). The ret1ection coefficients in (!) account only for specular reflection, which occurs for smooth surfaces. When the surface is rough, impinging energy will be scattered in angles other than the specular angle of reflection (i.e. diffuse reflection), thereby reducing energy in the specularly reflected component. The Rayleigh criterion [6] is commonly used w; a test for surface roughness. giving the following critical height (he> of surface protub~rances. Abstract - Microwave reflection coefficient measurements ut 1.9 GHz and 4.0 GHz are given for u variety of typical smooth and rough exterior building surfaces. The measurements are compared to theoretical Fresnel re!lection coetlicients using Gaussian rough surface scattering models when applicable. Individual multipaths are resolved temporally using a spread spectrum sliding correlation system, and spatially using directional antennas. The measurement test cases included walls rn11de ol' limestone blocks, glass, and brick. The results indicate that the Fresnel reflection coetlicients adequately describe the reflective properties of the glass and brick surfaces. The rough stone wall meusurement,s ure shown to be bounded by predictions for a smooth surface model and a Gaussian rough surface model. The results of this work can be used to estimate rellection coellicients in ray tracing algorithms for propagation prediction. h A " = -- (2) l:lcos8.I The height (h) of a given rough surface is defined as the minimum to maximum surface protuberance. A surface is considered smooth if h <he and rough if h >_he. For the case of rough surfaces, Ament [7] has derived -a scattenng loss factor (p ) to account for diminished energy in the specular din:ction of retle~tion given by INTRODUCTION Recent studies on ray tracing techniques for propagation prediction [1-3] have shown promising results 111 predicting channel parameters such as path loss and RMS delay ::.pread. Ray tracing approximates electromagnetic waves as discrctc propagating rays that undergo attenuation, reflection, and scattering phenomena due to the presence of buildings, walls, and other obstructions. Due to a lack of dielectric property data on common building materials, ray tracing techniques often assume a fixed reflection loss for specularly ret1ected rays, regardless of incident angles, material properti~s and surface roughness. The Fresnel reflection formulas can be used to determme the ret1ection loss. provided the dielectric properties of the mteriace are known. To provide accurate building ret1ection models, measurements at 1.9 GHz and 4.0 GHz have been made for a variety of typical smooth and rough surfaces. Using measured dielectric properties [4], th~ empirically measured reflection coefficients are compared with Fresnel ret1ecuon formulas using a Gaussian rough surface scattering model when applicable. The results of these comparisons can be used to develop simple, and empirically accurate, building reflection models for stte-specitic propagation prediction techniques. r (1tO'hcosei)j (3) • A. where 0' is the standard deviation of the surface height about the mean sur1ace height. Equation (3) assumes that the surface heights are Gaussian distributed with negligible sharp edge and shadowing effects. Using p to modify the reflection coefficients ;, referred as the Gaussian rough surfa.:e s.;atlenng model, and is writ ..:n as Ps (f.L) roug h = expl-8 .... =P,rl <fu) roug h = Psrl, Boithias [6] reports that the scattenng Joss factor of agreement with measured results when modilied as r (1t'O'h.:osei')1 Ps = expl-8 - - . I. (3) (4) gives better ! (1tO'hcosei)2'"j . Jl /ol8 - /.. when: / 0 [x] is the muc..lilicJ Ucssd fun.:t1un uf L.eruth urJer. E4ual!Ull~ (3) and (5) are approximately equal when the Bessel function argument becomes small. causing I 0 to approach unity. 1.9 GHz AND 4.0 GHz MEASUREMENT SYSTEMS The 1.9 GHz measurements were performed using a wide band REFLECTION AND SCATTERING MODELS The Fresnel reflection coefficients ( r) relate the tield reflected from an infinite ditilectric slab to the incident field, assuming the Incident field has a planar wavefront with constant phase. Th~ renect!On coefficients are given by r ·1- 11 2 cos8i-11 1 cos8 1 11 2 cos8i +11 1 cos8 1 r 11 = 11 2 cos8 1 -11 1 cos8 1 11 2 cos8 1 -r 11 1.:os8i <I> Where the wave impedances (11) and transmitled wave angh:s (8 1 ) are. complex functions of dielectric material properties, angle of tns cbtdence ( 8. ), and frequency [5]. The parallel (perpendicular) U SC , ' I . npt m (1) refers to the E-field component that is parallel Figure 1 • Geometry tor Fresnel retlection coetlicicnt calculation Resear·::ch:;----------------------sponsored by DARPAIESTO and !he MPRG Industrial Affiliates. 0.7803-1266-X/93/$3.00C 1993IEEE 77 spread spectrum sliding correlation system with a 2.047 chip PN s~uence. Th1s PN sequence was driven by a 230 Mcps local oscillator. giving a system resolution of approximately 8.7 ns. The desired LOS and reflected multipaths were resolved temporally via the sliding PN code and spatially usmg log-periodic antennas with 60u beamw1dths. For better spatial resolution. the 4.0 GHz measurement system used 18.6 dB standard gam horns wllh measured E-plane and H-plane beamwidths of 17.2° and 14.3°, respecllvely. For th1s system. a 32.76 7 chip PN sequt:n.:c: wa:. implemented, which provided superior correlation properties. The 4.0 GHz system used a 240 Mcps sequence clock, giving a system resolution of approximately 8.3 ns. Detailed specif1cations of this system are given in [?J. Due to the superior spatial and temporal resolution of the 4.0 GHz system. most of the measurements were performed at the higher frequency band, where the surface roughness etrects are more pronounced. REFLECTION COEFFICIENT MEASUREMENT The retlection coefficient measurements were performed using a two-measurement technique [8] shown in Fig.2. For each transmmer and receiver location (determined by desired incident and rellected angles and the system temporal resolution). multiple line-of-sight (LOS) and ret1ection measurements were made for power averagmg purposes. The LOS measurement pr,,\·1des a reference fur cumpanson with the rece1ved re11ected !'·· ·wer un the re11e.:t!un measurement. The observed power delay p;·,,file for the reflection measurement generally contains a LOS and ;·;; r1ected· peak, as shown in Fig.3. The observed time delay between ·.:.<: two components (i.e. differential delay) corresponds to the difference in path lengths. The measured LOS power is taken to be the Friis transmission value, while the measured reflected power is taken to be the free space value for the unfolded path length times the reflection loss, or ~ (PR) = ref/ PTGTGR')..." ., 1 1 1fi" (6) (41t)- (dl +d2) Equation (6) implicitly assumes that the rel1ecting boundary is ~'''''''"''''''''''''''''''''''''"''''''''~"'~ LOS Measurement Rel1ection Measurement Figure 2 ·Two-measurement technique for determination rdl~:ction coellici~:nt t..•e":Js...,reo :..·:)5 :J...,O We'·e:!e':l v._.rt•O.:''""S 01 5:c,.,eo •O'· ":.:,:,1 :J ... I • . C ... z 01 ..,6 z .. . 2 o.c.ae ~-0016 1, -o.ueu :S-o.••• ~ -~ :~~ .;< -- ••• (I 1 !\ ~--~co_. ...... \ ,, I' I IiI\ 1 ·- ... :;· ... • • ... o• .. . ••• .. t•••c-...--·•-••01l<jj,., ... ., ... ,""···.-· ........... -·~ • ;,_'~ ... ,...... 1\ I. - ...................... ~~- - (P R ) ref/ 1 J;,J • ' \', \V\~, j,~u~Mf'V'\r"JJW)Jli\.,I\i"~~"~';r~"~,.} !\JlV l Y -O . .l:l6 infinitely large. which is approximately true when the wall dimensions are much larger than distances d 1 and d2 in Fig.2 and the wall surface area is much larger than the illuminated area. Assuming these conditions are satistied. dividing (6) by the Friis transmission value for the LOS measurement and solv.ing for the reflection coefficient yields dt +d.,J. (PR) fl lr\ = - - · . dws Therefore, ·the empirical rel1ection coefficient is determined by the rauo of the reflected and LOS power measurements, weighted by the difference in measurement distances (i.e. accounting for the path loss difference in the two measurements). The sites for the reflection coefficient measurements included a rough stone wall· made of limestone blocks, a smooth metallised glass wall and a brick wall. The chosen sites had relatively homogenous surface characteristics (i.e. free of clutter. windows, doors, etc.), and the variation of surface roughness at these sites ranged from rough stone to smooth glass. Table I lists the estimated surface roughness parameters and lhe measured dielectric properties [4) of the limestone and brick surfaces at 4.0 GHz. The roughness Table 1: Roughness and dielllctric parameters lor the limestone and brick·wull surfaces at 4.0 GHz. · Wall h (em) cr h (em) £r ll, cr (S/m) Limestone 12.7 2.54 7.51 0.95 0.028 Brick 1.27 0.508 4.44 0.99 0.010 parameters for these surfaces are the same for both frequencies, and the dielectric properties are approximately the same as well. The dielectric properties for the glass wall were not measured in [4). MEASUREMENT RESULTS . Figures 4-7 show the results of the measured reflection coefficients at the limestone wall site compared to the predicted values using the Fresnel formulas. Each plotted reflection coefficient results from an average of multiple reflection coefficient measurements (typically 10). The 1.9 GHz measurement results in Figures 4 and 5 are only compared to the Fresnel rel1ect1on formulas for smooth surfaces and the Gaussian rough surfa.::t! model using (3). From these figures. the measured reflection coefi1c1ents are bounded by the smooth and rough surface predictions. In virtually all cases, Figures 4 and 5 show that simply using the Gaussian rough surface model alone gives pessimtstlc predicttons to the actual reflection coeflicient values. F1gures 6 and 7 also show the measured 4.0 GHz rel1ection coeflicients are bounded by the Fresnel predictions for smooth and rough surfaces. Again, the Gaussian rough surface model (with either scattering loss f:~.::turJ tends to give pessimistic predicuons of the measured reflection coeitic1ent vai:.!~s. However, using the scattering loss factor in (5) gives improved comparison results. Figures 4-7 suggest that in the absence oi measured data. an adequate model for rough building surfaces can be easily computed by averaging the Fresnel ret1ection coefficients for smooth and rough surfaces, as a function of incident angle as shown below. f .!.,II fAVG:: [f.!.,ll (S;))AVG:: Figure 3. Example retlected path measurements at 4.0 GHz (time axis has arbitrary zero reference) 78 (7) re (PR) LOS sf.!..II +p 2 ( 1+ = 2 p s) f.!.,ll (&J Mecsured Reflection Coefficients of Rouqn Stone we:· ("e~"t•co• ' - 1 9 .::;: ..... : <;--> P<etoenoicu•or Polorizot•on o.,ter,,o oo•or•zCt•O"'J f - 4 Meosureo Re-tte-ct'c.r·. .:oerf •..:.c-nts ..:>f Rcn... qf"' Stone W.:lll G,...z <;-- > Peroend•O:~.o~'C' Po•or•Zot•on (vert•C:O' onte,.,no oo•o~'•Zot•O.,J •o / .E.: 6 ~ .J.~ ~- ~ .g v 3 g.: .: ",.. .. ,, .... o•'I'-'''Oeo••O" ~·~ .... ., . .,,.,, ... ~ D•ao-•·•'1 r • ~ 136 , • .J :;• .. • ; . .;.o "'C. ... ~'" 'II :;I·· DO' :;I,..••••~ ' : .. : - . -3~ • : ~· , _ ., 2~ Mecsure.:l Rei·ectio,.,. f • 1 1 9 G~""~Z ~oeificie,....!s ; -a ~~ 7 ~: 6 ~.:......,<~P"' 011.011 ... 000" '"''0Cf' ca._,, .., ... ., .. 'l ........... , '- * Gl-1: ;~ ... 1 : :t•c- 0 ---------- ~ouqn ........ - .... - Stone wou -~~~ ....O"' . . .<INIOC:CII•O" '----. I '\1.-••••"'e'"tP•oc;..l'•h•' c. • t••· .. u•• •"''co'• ~::,:~·.o.:,•:;.~" ~o "? !a•." • '"''": • ... , ••., ... •""')" ....... ,., ••• 0 O::ll.,. • C.9!1 DO•-·· .. ~ -~--· . . . · : .. c ... ·~·~· ]v-; I ·h . . . ~·~- ! e ....... ,, .. ,i i ~ 0 ~ / c ~c:: :5 .... e.c ~- § :..o ~ c.s ~ 0.4 "v' 0 .:..!:1 :: .:..8 ~05r------­ .go.J ~~ (oeq'ee~; .... --> Poroul!• ;:.o•o'•Zot•Or"' tnor•:onto• onte,.,no po•or :ot•O.,J ~ Oc. .. q" .IO"e DO'O,.....el•'t "• ·;;: :c- !_ 4~ ... .::.oerq ,:.,,.,q•e Meosureo Reflc:-ction CoeHic•ents oi Stone Wo:: ,."'''•...,a•• -o•• •oco••o .. 'S•e-• ~ ... e ... c -· :·r.•·• ... , . • ., ce. o • c. ~;i. _. • o ~6 "•os .. ••oaco•a " Figure 6- Measured rellection coeflicients lor limestone wall at 4.0 GHz (perpendicular polarization) (--;. Po•olle• coo•o .. •:"O!•on \"O'•::.nlc• ..:nte""'O oo•c• :ot•O"'J c £ ~:: ..:.: 0 -o " ~ 0.1 Figure 4 · l'\'leasured retlection cocfth:ients for limestone wall at 1.9 GHz (perpendicular polarization) /. 0 0 0 ·=- l! ~ ~ . .> §...:0~ 'C.._· "-' ::::; c 0'----------------~~---3G 4~ • o~o.. J.0 Incident At"'qle (degrees) ..: 'J 10 jU ~~ SC 6~ IMC·CI~,t Al"''qle (degrees) 70 sc 90 Figure 5 -l'\'lcasured rdlcction coefficients lt1r limestone wall at 1.9 GHz (parallel polarization) Figure 7 • 1\leasurcd rcllcl:lion cudlicitmts for limcstonl! wall at 4.0 G Hz tparalll!l polarization) In. (lSJ. r .1.u is either the perpendicular or parallel polanzauon rel1ect10n cuc:flkient given in (I) and p s is the s.::attering loss factor in (3). This simple model is compared to each measured rellection ..:uodlkient 111 Figures 4-7. The error statistiCS of these comparisons. including mmunum and maximum errors. average error (EuvgJ and standard deviation (s). are given in Table 2. Table 2 shows that using measurement results fur vertical and horizontal antenna polarizations, respe..:uvely. The dielectric properties of clear, nondoped glass are typically quoted as: E = 5. ll = I and cr = 10" 12 S/ m. Since the glass wall is metallised.rthe conductivity will be much higher than 10" 12 S/m. Bec;tuse the conductivity is unknown. several Fresnel rellection coeflictents with various conductiVIties (10" 12 to 10 S!m) are plotted in Figures l) and 9. A material having a conductivity ut' 1U.O S/m woulJ bo: ..:un~tJered a semH:unductur. sim!lar to seawater or intrinsic germamum [5]. No comparisons with Gaussian rough surface models were m:.~de in this case. ln Fig.8. the measureJ rellection coefficients closely agree with the theoretical Fresnel rel1ection coefticients using cr- 5 Slm. 111di.::ating that it is more r~:tleo.:tive than ordinary. clear glass. For manv of these me<1surements with e. 2! 45 °, the rellection . ' coeflicients are even gre:1ter than that predicted using cr I 0.0 Slm. In Fig.9. the comparisons indicate the measured rellection coefticients are well approx.unated by the Fresnel rellection coefticients usmg a- 2.5 S/m. Althuugh the ex.a.::t diele.:tru; properties of thi.; glass :.~re nut known. Figures 8 and 9 indicate th.:tt the ret1ecuon coeflicielll .:;an be: accurately moJeled using the smooth surf'-l.:c: Fresnd furmuh~' <llld tht: propt:r o..ltt:lc:.::tn.: properties. Figures 10 and r I illu~trate llfe !frick wall ro!t1ectJOn coetlicient results for vertical and hunwntal antenna polarizations. respectively. The smooth surface unJ Uaussian rutigh surface models were compared with the measured results. Fur these cases, there is little Jifference between the s..:attenng lu~~ l<t~oturs of t3J and (5), because of the small standard deviation of height (a h). The measured retlection coefficients of the bnck wall are not well bounded by the Table 2: Error statistics for comparisons of measured rcllcction cocfticients with simple model ot' (~) Freq. (GHzJ Polarization Em in Em ax Eavg s 1.9 Perpendicular 0.002 0.240 0.096 0.072 1.9 Parallel 0.002 0.103 0.057 0.037 4.0 Perpendicular 0.017 0.3!9 0.146 0.096 4.0 Parallel 0.002 0.184 0.097 0.054 rA \'G to approximate the measured retlect!On coefficients of the limestone wall varies in degree of accuracy. In most cases. the: .:rrur estimates indicate that r A VG giv~s a fairly accurate modd t'~r surfaces with wndom roughness features wluch is simplt: to implement and delined by real-world. physic<11 features. In gener<ll. f:\\'(j giv~~ an Improved t!Stimat~ uver f.l.ll (With or WltlWUt the s..:auenng luss fa.::tor). A lea.:;t squares .::urve fit or linear regrc:ss1un .:ould have: been <1pplied to tll~se data to prov1de a more a..:curato:: Jt:,..;npttun. but tlus mudd wuulJ nul bt: rubust 11l that 1t rc:Ljuu·e~ empirical data for different surfaces and matenals. Figures l) and 9 display the glass wall rel1ection cueffic1ent = 79 lvleosurea Retlect•on Coeffic•ents of Br•c~oe. Wolf Meosurec t=<eflect•on Cueffic•ents cf Gloss Wa11 I- 4 1 0 <--> Pe-rQerHj•cuiOr Po•orozotoon ("'e-rt•cc: .onter'\no ~o•o••zot•onJ t - - 0----- - 9 0.6 <.--"' Perc:>eroa.c ... •or Polorozoticl"' ( ... t!r-t•COI on1ero,.,o polor•zot•on) ~09 •Ooa• ..... OQI" o..,<lg'~ e ................ q . . . ,. •• .,c-• ~:;::·.o.-G·:;,"'!,"""''•~••-• k•••• ,, .. , .o-·- o---,; 7 4 CHZ 1() ------_-_--;oe::-_-_~ - : - - -:. : --. -. _., ~09 .);! 0 8 ~0 G...,.z - ":!- - u ~ 0.~ ~Q .. •••• 51 .. GO,.I CO"' It' IG("GI•Q"' •:.-••'i•~<j ...... t l • ~ c '51=..:Q~,Sl.~.,s~Q"';G a: 0 . cr• 0 ' ~0~ tt• 2 g.o2 G·OS$ ~ 0.1 0.0 0 20 '0 30 40 SC ~- - Meos ..... reu 1.0 .~ 0.9 ;;g 0.8 ~0.7 c g u.o u C...,.z c.--;, 6C 80 ,,.,C•Cle,.,t Ang•e (de';'ees) Kei•ect•.::.~. 90 (:.::e••·cze.,ts .:;..r C·oss f- co .. o•·•-•o..q" •w•••c• ---------- - ·a- - - a: 0 0 .• .goJ .. --- . 0 c ~ --- ·~"" "" ~0.5 :; - :.:---· /,/ ~ 0.4 -Q ' 0 . .3 ·a 0.2 ~0.2 ~ 0 ~ ~01 "" 10 20 30 40 ~0 6- lnc:•Oe-Mt Angle (c:!eq•ees 70 80 co sc Figure 9 • Measured reflection coeflicients fur gluss wall ut 4.0 GHz (parallel polarization) predictions for the smooth and Gaussian rough surfaces. in general. This is expected since the roughness of the brick surface is such that the critical heights (he) are greater than the maximum protuberances for many of the incident angles. Because of this, the Rayleigh criterion predicts that the surface appears smooth to the incident radiation and the reflection will be predominantly specular. This is ev1denced m Figures 10 and 11 since most of the measurements fall dose to the predicted Fresnel reflection coeflicients for smooth surfaces. eo 90 I I 01 0 . . . . 0 99 •o .. ca .. t. .. ••oc-• ll"O'o""•'•'' "'• 1 17 c.., ""' • 0 ~0.8 c- -------------. -- o 0 - \0 -L..__ 3~ :7~: 40 50 60 ,,.,c•oent Al"ql~ (deqrees) // 70 eo 90 Figure 11 ·Measured rcllcctiun cuet'ticients l'or brick wall ut 4.0 GHz (parallel polarization) ACKNOWLEDGMENTS The authors would like to acknowledge Michael Keitz for his assistance with the mt:asurement systems, and Amanda Montague. Hao Men·g and Bhushan Rde for their contributions during the experimental data collection. REFERENCES [lI R~ppapon. [21 Symposium on Wir~less Personal Communications. Blacksburg. VA. June 17-19, 1Y':!2.pp.l6.1-16.27. Honcharenko. W.. Benoni. H.L.. Dailing. J.L. Yee. H.D.: "Mecfumism-s CONCLUSION Microwave rellection coefticient measurements were presented at 1.9 GHz and 4.0 GHz for a variety of typical smooth and rough building surfaces. The measurement results were compared to the theoretical Fresnel reflection formulas using Gaussian rough surface scattering models when applicable. The measurement test cases include a very rough wall made of limestone blocks. a glass wall, and a brick wall. The theoretical comparisons show tlwt the Fresnel reflection formulas for smooth surfaces adequately describe the rel1ective properties of the glass and brick walls. The rough stone wall measurements were shown to be bounded by predictwns f~•r a smooth surTace and the Gaussian rough surface. A simple model to predict- empirical reflection coeftic1ents for surfaces with random roughness features was presented. and was shown to adequately predict the measured results. This model is delined by physical roughness ieatures and material properties and can be used to further enhance site-specific propagation prediction models. 70 I ~ ? 0.6 u ,,, ~ o.o• ::~ ... - o.99 3~·o~ac- ••-;;::·. ";::·.~:.;:~~·:~~.~:;···; ,, - .... .. c ~u7 '!,• 8 <••n•o••-•• •••'•u ,._.. . ., ""' WeG • .,•.CIGOiol ~08 a-·--- !IC, .:.o ~0 60 •n:•Ot:"t Anqle (ceqree5J . . . -···· ....... 1.0 ~0.9 -- ·=- ". Mecsure.::: ~C""C:-~ttcn Coefficr.s-nts of 8rzc~<. ...Vall "GHz <--> Po,ou.oe· .:>o•cr•zot•on (norizonto• ontenno oo•or•zotion) v\~P t - - - - Q- •• Figure 10 ·Measured rdl~:ction coefficients for brick wall at 4.0 G Hz tperpcndicular polarization) r ~0.~ 0.0 •o 0 ;:.crone• Polo••ZOI•..J,.., l'"'O"::'Or"'IIOI o.,tennc co•O'" :;;ot•oro; --------- 1o .. • Qli~-J" O"''IJC .. LoG'G ... OIO<O •·••2"~:,..... 0 5,:. t.r • Figure 8 • Measured ret1ection coetlicient.s for glass wull ut 4.0 G Hz (perpendicular polarization) f- 4 9• c. o••••c•"c o<ag..-•1·•• -- --- --- . -~ - T.S.. Scu.ld. S. Y.. Schau bach. K.R. : "Site-Specific Propagt.mon Pretltct/UII fur PCS System Destgn", Virginia Tech's 2ND Governing UHF Propagation 011 Single Floors in Modem Office Buildings", IEEE Trans. on Vehicular Tl'Chnology. vol.41. no.-1. November 1992. pp.-IY6-504. lkegami. F.. Takeuchi. T.. Yoshida. S.: "Theoretical Prediction ofMecuz Field Strength for Urbun Mobile Radio", IEEE Tnu1s. on Antennas and Propagation. vol.39. no.3. March 1991, pp.299-302. [4 1 Sou. C.K .. Landron. 0 .. Feuerstein. M.J. : "Cizaracterizatwn of Electronzagnenc Pruperttcs uf BtHldmg Matenals for Use in SaeSpeci;lc Propagatw11 Predictwn" . . Muh .c and Portable Radio Research Group T~chnical Report# 92-12. VA Tech. 1992. [51 Balanis. C.A. : Ad\·anc~d Engineering Electromagnetics, John Wiley and Sons. Inc .. 1'\.:w Yurk. !989. [61 Boithias, L. :Radio Wa\·~ Propugutiun. McGraw-Hill Inc., New York. 1987. [7] Amenl, W.S. : "Toward a Theory of Reflection by a Rough Surface". Proceedings of the IRE. vol.41. no. I. January 1953. pp.l42·146. [81 Landron, 0. : Microwave Multipath Resolution In Mlcrocellular Environments. Virgmia l~ch Masters Th~is, Augusl 1992. [31 80