Developments n Characterzaton of Moble Rado Propagaton May 19, 1999 IEEE Santa Clara Valley Antennas & Propagaton Socety Peter S. Rha, Ph.D. (psrha@sfsu.edu) San Francsco State Unversty 1
Messages to Drve Home wth Nature of Complexty nvolved Almost Impossble for Analytcal or Computer based Solutons Thus, Most are Emprcal Models based Feld Measurements Hard to Generalze and Scale based on Few Measurements Thus, New Measurements needed Whenever Varables Change Statstcally Meanng Data takes Extensve Efforts and Costly Some Understandng of Moble Rado Models What they are: Mostly Fadng and Mult-path Effects Mostly Statstcal n Nature Basc Underlyng Theory s not so Hard Some Apprecaton of Development Progress over the Years Key Contrbutons and Drvers Necessty of Models and Relatons to Technology Evoluton 2
Outlne Rado Propagaton Fundamentals Spectrum General Consderatons Moble Rado Propagaton for Cellular/PCS Objectves Dependences Usage Mult-Dscplne Perspectves Channel Modelng Framework Underlyng Math Models & Realty Assocaton wth Technology Progress Channel Models Key Contrbutons Fundamentals of Path Loss Models for Cellular Engneerng Models for MODEM Engneers Models Attempt for Coherent & Comprehensve Consoldaton of Models An Example of Recent Propagaton Study What s Ahead n Moble Rado Propagaton 3
Frequency Spectrum Freq. Band Desgnaton Servces 3-30 khz VLF Navgaton, sonar 30-300 khz LF Navgatonal Beacons 300-3000 khz MF AM, Martme Rado, Drecton Fndng 3-30 MHz HF Shortwave, amarture rado, Telephone, Telegraph 30-300 MHz VHF TV, FM, Moble Rado, Radar, Ar Traffc 300-3000 MHz UHF TV, Mcrowave Lnks, Radar, Satellte 3-30 GHz SHF Mcrowave Lnks, Satelltes, Radar 30-300 GHz EHF Radar 300-10 7 GHz IR/Optcs Fber Optcal Lnks Desgnaton: V: Very, L: Low, H: Hgh, U: Ultra, S: Super, E: Extremely, F: Frequency Cellular Reverse (MS ) Forward ( MS) 824 825 835 845 846.5 849 A A B A B 1 10 MHz 10 MHz 1.5 2.5 869 870 880 890 891.5 894 45 MHz PCS Reverse (MS ) Forward ( MS) 1850 1865 1870 1885 1890 1895 1910 A D B E F C 15 MHz 5 15 MHz 5 5 15 MHz 1930 1945 1950 1965 1970 1975 1990 80 MHz 4
Fundamentals of Propagaton Classfcaton Antenna Locatons Terrestral,Satellte, Arborne, Propagaton Meda Lower Atmosphere, Surface, Ionosphere, Meteor, Underwater Propagaton Path Obstructons NLOS, LOS, Free Space (Fresnel Zone Clearance) Sgnal Attenuaton Mechansms Spreadng, Reflectve, Dffractve, Absorptve (most, ran) Sgnal Propagaton Mechansms Reflecton, Dffracton, Scatterng, Refracton Polarzaton Vertcal, Cross, Horzontal,, Crcular Terrestral Channel Features Terran, Man-Made Obstacles, Waters, Folage * Those most mportant for Cellular/PCS ndcated n Bold 5
Partculars of Propagaton for Cellular/PCS Modelng Objectves Obtan locaton & tme dependent characterstcs for optmum spectrum utlzaton Sgnal Path Loss Sgnal Imparment Interference Statstcs Modelng Dependences Physcal Envronment Natural and Man-Made Features Sgnal Type Frequency Sgnal BW Polarzaton Technology Analog or Dgtal Modulaton/Codng Multple Access Methods Advancement of Sgnal Processng Technques Usage of Models Cell Plannng: Coverage (Outage), Capacty (Interference) Control Algorthm Desgn: Access, Power Control, Handoff Recever/Transcever Desgn: Modem, Codng, Interleavng, Equalzer, Rake Recever,... 6
Mult-Dscplne Perspectves EM Rado Propagaton Cellular Network Communcatons Problems of Interest Physcs of Propagaton frequency polarzaton antenna type/heght physcal & electrcal propertes of medum Lnk Budget path loss fade margn dversty gan Sgnal Processng fade rate Doppler shft coherence BW & tme # of mult-paths dversty correlaton Solutons Sought Models b/ Phys. Theores Attenuaton Dffracton Reflecton Refracton Scatterng Penetraton Rado Network Desgn cell sze # of cells locatons freq reuse antenna type/heghts Comm. System Desgn Mod/Demod Interleavng FEC Channel Equalzer Rake Recever Codes Goodness of Soluton Analytc./Computer Sol ns smplcty, applcablty & closeness to feld measured data Customer Satsfacton call attempt success rate call drop rate handoff success rate handoff rate Performance Complance Mnmum E b /N o BER FER Dversty Gan Acquston Tme Ultmate Goals Formulaton of Analytcal Models Mnmum Investment Customer Satsfacton Max. Bts/Sec/Hz Shannon s Lmt 7
Taxonomy of Moble Rado Propagaton Propagaton Condtons and Network Deployment Frequency: Cellular, PCS, MMDS, LMDS, Wreless LAN Envronment: Cluttered Cty, Urban, Suburban, Rural, Indoors,... Cell Sze: Macro-, Mcro-, Pco-cell, (Herarchcal) Antenna: Heght, Drectvty, Polarzaton, Tlt, Spacng Coverage: Outdoors, In-Buldng, Subway,... Moblty: Hgh Speed, Cty Drvng, Pedestran, Fxed Models of Man Interest (Statstcs n Nature) Network Plannng Path Loss and Slow (Long-Term) Fadng Dversty Correlaton between Sector Antennas, Sectors, and Neghborng Base Statons Communcaton System Desgn Fast (Short-Term) Fadng, Coherence Tme, Doppler Frequency Mutpath Delay Profle: RMS Delay, Coherence BW, Correlaton Angular Spread: RMS Beamwdth, Correlaton 8
Man Factors to Propagaton Characterzaton Cellular Engneerng Cell Plannng: Sze, Freq. Reuse Coverage Probablty Rado Network Tech. Macro-, Mn-, Pco-cell, Networkng n Herarch. Dynamc Freq. Assgn. Smart Antenna FCC Spectrum Cellular Band (8 to 900 MHz) PCS Band (1.8 to 1.9 GHz) Moble Rado Propagaton Channel Characterzaton Cellular Standards AMPS, TDMA,CDMA 1G, 2G, 3G, 4G... Communcaton Tech. Analog vs. Dgtal Narrow to Wde BW Adv. Sgnal Process. 9
Mathematcal Framework of Models Black Box Approach nput System h(t) H(ω) output nput Lnear Tme-Invarant System output Smple Undergraduate Level Changes n Tme nput Lnear Tme-Varant output Maybe Graduate Level Changes Randomly nput Statstcal Lnear Tme-Varant output Defntely Graduate Level SO, What s the Dg Deal? 10
So, What s the Bg Deal? It comes n myrad of shapes and szes! STV STV STV STV Why Not A General Theory? Why Smulate Methods based on Physcal Theory for Generalzaton don t exst yet, Requres tons of nput data and runs forever Expensve Alternatve probably never wll Make Feld Measurements Return for nvestment not good busness. Subject Not that Well Understood Slow Advancement n Scentfc Knowledge Dong Better n Europe, however. 11
Tech. Evoluton & Detaled Structures needed 1970-1983 AMPS Analog FM 30 KHz Narrow Band Analog Path Loss: Mean, Varance (Shadow Fadng) Doppler Effects: FM Nose (Raylegh fadng) Dversty: Spatal Correlaton 1989-1992 IS-136 Dgtal TDMA 30 KHz Narrow Band Dgtal Delay Spread: Equalzaton up to ~ 42 µsec delay needed. Doppler Effects: Raylegh Fadng, Interleavng, Demodulaton 1990-1993 IS-95 Dgtal CDMA 1.25 MHz Wde Band Dgtal Delay Spread Characterstcs (~ 1 µsec) Hgher Delay Resoluton Needed # of resoluton paths fadng stats and correl. bwtn them how fast they come & go Correlaton btwn Sectors/Cells for Soft HO 1998-2000 cdma200 Dgtal CDMA > 5 MHz Broad Band Dgtal Delay Spread Characterstcs (> 1/4 µsec) Even hgher resoluton needed, # of resol. paths more crtcal, Stats of each path may not be Raylegh anymore 12
Propagaton Scenaro for AMPS & TDMA Forward Lnk Sgnal Recepton Forward Lnk Interference Scenaro for AMPS and TDMA (N=7) r Base Staton MS An Example of Importance n Prop. Character. Interference level strongly dependent on propagaton exponent α: ~1/r α Major Impact on Frequency Reuse Effcency Sx Major Interferers of about equal strength 13
Propagaton Scenaro for CDMA Forward Lnk Sgnal Recepton and Interference Scenaro for CDMA nterference strength sgnal Another Example: Depends on Multpath Condton Due to breakage n orthogonalty between same cell users codes But, Dstance Dependent PL not as Important as n AMPS & TDMA Everybody nterferes wth everybody else 14
Cellular Engneerng: Coverage Relablty Shadow Fadng Log-Normal Model 1 Y( x) = 2πσ e 2 ( x x ) 2 2σ x: PL, x: mean PL, σ:sf Sgma Typcal Objectve for Area Coverage = 90 % Contour Relablty Area Relablty Path Loss Example for Max PL Allowed = 150 db (from Lnk Budget) 50 % P cov (r a ) Fade Margn: F mg 150 db 150-σ db 150-2σ db R a R b R c 84 % 93 % 50 % 84 % 150 P cov (r b ) PL, db 150-F mg 150 R a PL, db 150-σ 150 PL, db P cov (A) = 90 % F mg = 1.6 σ z xmax Pcov( r) = Y( x x( r)) dx 93 % P cov (r c ) r P A a cov ( ) = 2πz 0 Pcov( r) rdr x max : Max. PL allowed 150-2σ 150 PL, db 15
What do Cellular Engneers care about? Path Loss vs. Dstance Model (~ 1/r α ) Cell Coverage Radus: Nose Lmted Area [ α Rc ] Cell Radus: Interference Lmted Area [ α N reuse ] Shadow Fadng Model (Log-Normal wth σ) Cell Radus: Lnk Budget Margn [ σ Rc ] Correlaton over Dstance (Exponental) [? ] (In HO smulaton) Fast Fadng Model (Raylegh Fadng) Mn. Requred S/N (Eb/No) or S/I (Eb/Io) [ Rc, Cap. ] Usually already accounted for n numbers gven by Comm Eng. Doppler Spreadng [typ; md speed (30 km/hr): Eb/No ] Usually already accounted for n numbers gven by Comm Eng. Correlaton over Dstance (Bessel Functon; λ/2 Decorrelaton) Useful for estmatng the nterval (vs. speed) for averagng out fast fadng n the feld measured data for local mean PL analyses Delay Spread Mult-path Detals (Coherence BW & Tme) TDMA Equalzer or CDMA Rake Recever Performance Not usually used, but can provde good area specfc nformaton whch may be accounted for n cell plannng. 16
Evoluton of Channel Models 1st Generaton Analog Tme-Varant Memoryless : tme Path Loss, Fast Fadng: Doppler Freq., Slow Fadng 2nd Generaton TDMA and CDMA... Tme-Dspersve: (tme & delay) + Coherence Bandwdth, Coherence Tme 2G+ and 3G TDMA and CDMA... Horzontal Angular Spread: ( tme, delay & angle ) Beam Profle: effectve beam wdth, correlaton 3G+, 4G and beyond (???)... Vertcal Angular Spread... Polarzaton... Fxed Cellular wth Hgh Frequency and Wde BW Termnal Antenna Drectvty and Pontng Drecton 17
Key Papers n Moble Rado Channel Fundamental Theores S.O. Rce, Mathematcal Analyss of Random Nose,, Statstcal Propertes of a Sne Wave Plus Random Nose, TJ P.A. Bello, Characterzaton of Randomly Tme-Varant Lnear Channels, IEEE T. Comm 1944 1963 Fast Fadng: Theory and Measurements Raylegh & Rce Dstrbutons, GWSSUS Channel M. Nakagam, The m-dstrbuton: A General Formula of Intensty of Rapd Fadng,? R.H. Clarke, A Statstcal Theory of Moble-Rado Recepton, TJ < 1960 1968 Path Loss: Measurement Based W.C. Jakes, Ed. Mcrowave Moble Comm,Wley, 1974 Raylegh Dst., Bessel Correl Nakagam, Jake s Model Y. Okumara, Feld Strength and Its Varablty n VHF and UHF LMR Servce, Rev. Elec. Comm. Lab, Japan Hata, Emprcal Formula ffor Propagaton Loss n LMR, IEEE T. VT. 1968 1980 Path Loss: Physcal Theory Based Hata/Okumura Model (Cost231 Macro-cell Model) F. Ikegam, Propagaton Factors Controllng Mean Feld Strength n Urban Streets, IEEE T. A/P J. Walfsch & H. Berton, A Theoretcal Model of UHF Propagaton n Urban Envronments, IEEE T. A/P 1984 Wdeband Delay: Early Measurements 1988 Walfsch-Ikegam Model (Cost231 Mcro-cell Model) G. L. Turn, et al., A Statstcal model of urban multpath propagaton, IEEE T. VT D.C. Cox., 910 MHz Urban Moble Rado Propagaton: Multpath Caracterscs n NYC, IEEE T. Comm 1972,3 18
Mathematcal Model for 1G Analog Varable: Temporal Model: path loss l(d) h( t) = c( t) = l( d) r( t) s( t) fast n t r(t) slow n t s(t) System Model Frst Order Dst. Raylegh Log-Normal Re{ e j ω c t } Second Order Corr. c(t) Bessel Exponental h(t) Re{ h( t) e j ω c t } F{ } F -1 { } H(f) Doppler Effects 19
Mathematcal Model for 2G Dgtal Varables: Temporal Delay Profle System Model h( τ; t) = c( τ; t) = a ( t) δ ( τ τ ) h( t) = z c( t; τ ) dτ τ f Impulse Response h(τ;t) Coherence Tme AutoCorr: φ h (τ; t) E[h(τ;t) h(τ;t+ t)] t ν F τ m τ δ ( t) e j ω c t Transfer Functon H(f;t) F -1 D(τ; ν) Doppler Functon h(τ;t) Coherence BW AutoCorr: φ H ( f;t) E[H(f;t) H(f+ f;t)] t ν τ f Doppler Spreadng h t e j ω c ( τ; ) t ( τ ) S(f; ν) Scatterng Functon 20
2D Tme Varyng Channel Illustraton Illustraton t 1 t 2 t 3 t n p(τ, t 1 ) p(τ, t 2 ) p(τ, t 3 ) p(τ, t ν ) Seres of Delay Profle Snap shots τ τ τ τ p(τ, t) p(τ, λ) Fourer Transform over t λ t τ z 2 j πλt P( τ, λ) = p( τ, t) e dt λ: Doppler Spectrum τ 21
Mathematcal Model for 2G+ Varables: Temporal Delay Profle Spatal System Model δ ( t) e j ω c t h(τ;θ;t) h t g e j ω c ( τ ) ( θ η) t ( τ ) Industry Accepted Model whch ncludes angular doman Yet to Come h(t).. 1D Channel Model h(τ;t). 2D Channel Model h(τ;θ;t). 3D Channel Model 22
What s Ahead for Moble Rado Propagaton? Smart Antenna Angular Resoluton and Inter-ray Correlaton Fxed Cellular Revst of Moble Channel Models for Termnal Fxed, Hgh Elevaton, Drectonal Wder Bandwdth for 3G and Beyond (5, 10,15 MHz) Fner delay resoluton needed Present nter-ray correlaton models need to be revsted Other Channels not addressed n ths talk Indoors, and Mcro- and Pco-Cells Hgher Frequency 23
Case Study, 1: Fxed Wreless Channel Paper: IEEE J. on SA/Com, March 99 (AT&T-Labs) Measurement Equpment Frequency = 1.9 GHz; Sgnal BW = 8 MHz Tme Resoluton =.125 µsec Measurement setup Fndngs Where: Suburban areas n NJ and Illnos Tx antenna: 65 o Beam Wdth MS Rx antenna: Heght = 3 to 10 m, BW = 32 o BW and Omn Path length:.5 to 2 km Drectonal case: spke-plus-exponental profle Longer delay paths arrve at angles and come through sde lobes - hgher attenuaton for longer delay paths makes sense Power rato between spke to exponental paths ~ K = 8 db RMS tme delay of exponental paths ~ τ o =.2 µsec + K and τ o essentally not correlated Relatvely nsenstve to antenna heghts and path length Omn case: no such structure found 24
Case Study, 2: Theory g( τ; t) = [ A + a ( t)] δ ( τ τ ) = Aδ ( τ τ ) + a ( t) δ ( τ τ ) = g ( τ ) + g ( τ; t) A s a fxed ampltude and a (t) s zero mean complex Gaussan F M Mean Ampltude 0 determnstc E[ g( τ; t)] = E[ g ( τ )] + E[ g ( τ; t)] = E[ Aδ ( τ τ ))] A Mean Power F M E[ g ( τ; t)] = E[ g ( τ )] + E[ g ( τ; t)] = E[ A ] + E[ a ] = A + σ 2 2 2 2 2 2 2 F M Delay Profle P = F H A A 2 2 + σ 2 2 + σ I K F I 2 rms = H P K 2 2 RMS Delay Spread τ τ P τ 25
Case Study, 3: Fndngs and Completeness General Model g( τ; t) = Aδ ( τ τ ) + a ( t) δ ( τ τ ) Drectonal Antenna Case Emprcal Model based on Measurements: Spke-plus-exponental / τ m g( τ; t) = Aδ ( τ τ ) + a ( t) δ ( τ τ ) = Aδ ( τ τ ) + b e δ ( τ τ ) 0 0 0 0 Strong drect arrval path + many lower strength late arrval paths. Strength of late arrval paths decreases exponentally. = 0 Completeness n Characterzaton Dstance dependency? - No, based on.5 to 2 km range measurements Resoluton dependency? - Would not show up f BW < 5 MHz or so - Fner structure whch warrants a dfferent model may exsts. Envronment dependency? Not reported Antenna Beam Wdth Dependency? Not reported. Frequency dependency? - Not reported. Fadng dstrbuton of ndvdual paths? - Not reported. Fadng dstrbuton of combned sgnal? - Not reported. Correlaton between multple paths? - Not reported 26