Radio channel modeling: from GSM to LTE and beyond Alain Sibille Telecom ParisTech Comelec / RFM
Outline Introduction: why do we need channel models? Basics Narrow band channels Wideband channels MIMO channels Multi-link channels Perspectives & conclusion Séminaire Comelec, 7 juin 2012 2/56
Introduction: why do we need channel models? Séminaire Comelec, 7 juin 2012 3/56
Why do we need channel models? From the operator s point of view For network planning, BST deployment As inputs to engineering rules & tools From the manufacturer s point of view For performance evaluation For device/equipment design optimization From the researcher s point of view For trying novel network architectures For evaluating novel antenna technologies Séminaire Comelec, 7 juin 2012 4/56
Why do we need channel models? How to proceed? 1. Propagation research done by researchers 2. Extract the substance of the channel physics into something tractable 3. Standardize channel models de-facto or through official bodies: COST, IEEE, ETSI, 3GPP Séminaire Comelec, 7 juin 2012 5/56
Why do we need channel models? Example - back to basics: GSM (operator s need) The main need: to model the attenuation (path loss) vs. distance, for «typical» environments and varying BST height Séminaire Comelec, 7 juin 2012 6/56
Why do we need channel models? Example - back to basics: GSM (operator s need) The main need: to model the mean attenuation (path loss) vs. distance, for «typical» environments and varying BST height Séminaire Comelec, 7 juin 2012 7/56
Why do we need channel models? Example - back to basics: GSM (operator s need) The main need: to model the mean attenuation (path loss) vs. distance, for «typical» environments and varying BST height COST-Hata model for suburban or rural environments: (db) (1500-2000 MHz, 1-20 km) This is typical of an empirical path loss model Séminaire Comelec, 7 juin 2012 8/56
Why do we need channel models? Example: MIMO WLAN - IEEE 802.11n (manufacturer s need) MIMO channels deeply involve the «space variant» characteristics of the channel (and the multi-antenna system) Séminaire Comelec, 7 juin 2012 9/56
Why do we need channel models? Example: LTE / LTE-A (manufacturer & operator s need) A more complicated networking scheme, implying a sophistication of channel models Cell-Edge Beamforming Séminaire Comelec, 7 juin 2012 10/56
User throughput Why do we need channel models? Example: LTE / LTE-A (manufacturer & operator s need) A more complicated networking scheme, implying a sophistication of channel models Relaying Distance to BS Relay Séminaire Comelec, 7 juin 2012 11/56
Why do we need channel models? Example: LTE / LTE-A (manufacturer & operator s need) A more complicated networking scheme, implying a sophistication of channel models Coordinated Multipoint Tx and Rx Séminaire Comelec, 7 juin 2012 12/56
Why do we need channel models? Example: UWB RFID (researcher s game: ongoing) In every wireless communications research project, one needs to make channel measurement campaigns! Séminaire Comelec, 7 juin 2012 13/56
Why do we need channel models? Trend Vehicular Multi-user WINNER LTE IEEE 802.11n MIMO channel Directional channel Hiperlan Body Area Networks IEEE 802.11 IEEE 802.15.6 UMTS IEEE 802.15.4a Ultra Wide Band GSM Impulse response Path loss COST 207 COST 231 COST 259 COST 273 COST 2100 COST IC1004 Séminaire Comelec, 7 juin 2012 14/56
Basics Séminaire Comelec, 7 juin 2012 15/56
Basics Propagation channel Tx signal Rx signal Transmission channel The transmission channel comprises antennas and all objects contributing or hampering propagation between input/output ports The propagation channel excludes the antennas and expresses all wave propagation phenomena between Tx and Rx Both channels are considered to be linear (time variant) filters characterized by their impulse response Séminaire Comelec, 7 juin 2012 16/56
Basics Path 2 LOS (Line Of Sight) Building Direct Path Base Station Path 3 Mobile Terminal Tall Buildings Spherical waves at Tx Planar waves at Rx and in-between These are approximations! (may not be verified ) Séminaire Comelec, 7 juin 2012 17/56
Basics NLOS Direct propagation (LOS path) Specular reflection Diffraction Diffuse scattering Refraction/transmission scatterers Séminaire Comelec, 7 juin 2012 18/56
Basics NLOS The simple multipath based discrete channel model of the impulse response: h( t, ) i A i t i t H( t, F) i A i t exp 2 j F i t Time Delay Time Frequency At baseband, A i (t) is complex valued Séminaire Comelec, 7 juin 2012 19/56
Basics What (may) need be modeled The multipath amplitudes The multipath angles (at Tx/Rx) The multipath delays The time dependence of these parameters (comes from Tx/Rx mobility, or mobility of the environment) For various environments, Tx/Rx distances, locations Séminaire Comelec, 7 juin 2012 20/56
Narrowband channels Séminaire Comelec, 7 juin 2012 21/56
Narrowband channels Received signal: r I t It cos2 F t Qt sin2 F t Assumption: c t SLOS exp jjlos Sn exp jjn t Q LOS/NLOS small scale fading t SLOS exp j LOS Sn exp j n t n n c NLOS LOS Rx S LOS behaves as a deterministic variable (DV) S n and q n, j n behaves as random variables (RV) Central-limit theorem I(t), Q(t) are non-centered, identically distributed Gaussian RV Séminaire Comelec, 7 juin 2012 22/56
Narrowband channels LOS/NLOS small scale fading Statistics: Random part of I, Q: PDF I or Q 1 exp s 2 2 I 2 or 2s Q 2 LOS Std deviation s 0 Envelope r Rice distribution: 0 PDF exp K factor: ratio of deterministic to random mean powers K NLOS Rx I 2 2 r r S0 r S0 2 2 0 2 s 2s S 2s 2 0 2 s Séminaire Comelec, 7 juin 2012 23/56
Statistics: Narrowband channels LOS/NLOS small scale fading Rice distribution: PDF K= - db (Rayleigh) K= 3 db K= 10 db exp I 2 2 r r S0 r S0 2 2 0 2 s K 2s S 2s 2 0 2 s NLOS LOS Rx Séminaire Comelec, 7 juin 2012 24/56
Narrowband channels The three spatial scales of fading Small distance/medium distance/long distance fading (shadowing) Séminaire Comelec, 7 juin 2012 25/56
Long distance fading Narrowband channels Path loss Friis formula (free space): G G 4 Pt Gr Gt 1 Pr LdB 32.4 20log dkm 20log F L Going from 1 GHz to 10 GHz 20 db higher attenuation! Pr P t r t 2 d d, 2 MHz Is this frequency dependence maintained in NLOS channels? Yes! Frequency variations roughly speaking obey F -2 in NLOS Is this distance dependence maintained in NLOS channels? No! The path loss exponent n often exceeds 2 L d d L 0 d d 0 n Séminaire Comelec, 7 juin 2012 26/56
Narrowband channels Path loss Long distance fading Séminaire Comelec, 7 juin 2012 27/56
Narrowband channels Shadowing Medium distance fading What is that? Shadowing! = medium distance fading = macroscopic fading Séminaire Comelec, 7 juin 2012 28/56
Narrowband channels Shadowing Medium distance fading Gaussian in db = lognormal fading Séminaire Comelec, 7 juin 2012 29/56
Wideband channels Séminaire Comelec, 7 juin 2012 30/56
Wideband channels Small scale / short term / wideband fading Constructive/destructive interference between multipaths H( t, F, r) A i i t exp 2 j F t, r i i t, r t,0 i k i r 2F Path 21 Base Station Path 2 Direct Path Path i Path 3 Building k i Constructive/destructive interference Mobile Terminal r Tall Buildings Séminaire Comelec, 7 juin 2012 31/56
Wideband channels Small scale / short term / wideband fading Constructive/destructive interference between multipaths H( t, F, r) A i i t exp 2 j F t, r i Coherence band width i t, r t,0 i k i r 2F Flat (narrowband) fading Selective (wideband) fading Frequency Séminaire Comelec, 7 juin 2012 32/56
Wideband channels Small scale / short term / wideband fading Constructive/destructive interference between multipaths H( t, F, r) A i i t exp 2 j F t, r BW i i t, r t,0 i k i r 2F Time or space frequency Séminaire Comelec, 7 juin 2012 33/56
Wideband channels Frequency Delay (Fourier): impulse response D1/2.BW D rms Delay Séminaire Comelec, 7 juin 2012 34/56
Wideband channels Fourier pairs Time Doppler Delay Frequency Position Wavector Angle Séminaire Comelec, 7 juin 2012 35/56
Wideband channels Fourier pairs x k x k exp x exp jkx sinf jk x f The directional spectrum is obtained from the spatial structure of the signal Séminaire Comelec, 7 juin 2012 36/56
Wideband channels Angle vs. delay Street canyon 900 MHz terminal side Séminaire Comelec, 7 juin 2012 37/56
Wideband channels Angle vs. delay Dunand & Conrat, 2008 Séminaire Comelec, 7 juin 2012 38/56
Wideband channels Angle vs. delay Dunand & Conrat, 2008 Séminaire Comelec, 7 juin 2012 39/56
MIMO channels Séminaire Comelec, 7 juin 2012 40/56
Séminaire Comelec, 7 juin 2012 41/56 MIMO channels The impulse response becomes a matrix! The relative variation of coefficients (with position, frequency, time) dramatically impacts the diversity/spatial multiplexing performance The multipath structure is responsible for this variation ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( 2 1 2 22 21 1 12 11 NM N N M M h h h h h h h h h H
MIMO channels A non physical approach : based on the structure of the MIMO correlation matrix Kronecker approximation: The method lends itself to easy stochastic generation in a software tool: Statistical modeling: define statistical distributions for the coefficients of the correlation matrices 3GPP (LTE), IEEE 802.11n, IEEE 802.16 Séminaire Comelec, 7 juin 2012 42/56
MIMO channels A non physical approach : based on the structure of the MIMO correlation matrix Kronecker approximation: H UΛV min( n R, n T ) i1 H u v i i H i Séminaire Comelec, 7 juin 2012 43/56
MIMO channels Double directional channels for MIMO channel modeling Tx Rx Statistical modeling: define statistical distributions for path amplitudes, path delays, path DoA, path DoD 3GPP, IEEE 802.11n, IEEE 802.16 Séminaire Comelec, 7 juin 2012 44/56
MIMO channels A semi-physical modeling approach : GSCM (Geometry-based Stochastic Channel Modeling) Statistical modeling: define statistical distributions for scatterers positions, characteristics COST 259, 273, 2100 Séminaire Comelec, 7 juin 2012 45/56
MIMO channels What s that for? Example : standardization of OTA (Over The Air) test methods for MIMO terminals Séminaire Comelec, 7 juin 2012 46/56
Multi-link channels Séminaire Comelec, 7 juin 2012 47/56
Multi-link channels Multi-link channels are encountered in future networks Where BS can simultaneously be connected to several users for which terminals may simultaneously be connected to several BS Séminaire Comelec, 7 juin 2012 48/56
Multi-link channels Multi-link channels are encountered in future networks Where BS can simultaneously be connected to several users for which terminals may simultaneously be connected to several BS channel modeling requires proper account of macroscopic spatial correlations Séminaire Comelec, 7 juin 2012 49/56
Perspectives & conclusion Séminaire Comelec, 7 juin 2012 50/56
Perspectives Antennas are part of the radio channel! Propagation channel Tx signal Transmission channel Rx signal # An instrumentation antenna (for channel measurements) A use case Séminaire Comelec, 7 juin 2012 51/56
Perspectives Antennas are part of the radio channel! The super-antenna concept: antenna + user in near field F. Harrysson et al., COST 2100 TD 07-379, Sep. 2007 Séminaire Comelec, 7 juin 2012 52/56
Perspectives Antennas are part of the radio channel! In-body antennas: e ~50 s ~2 S/m! channel channel Embedded antenna Composite antenna-channel problem Séminaire Comelec, 7 juin 2012 53/56
Perspectives A variety of propagation environments for new wireless use cases Séminaire Comelec, 7 juin 2012 54/56
Conclusion The main message: radio channel modeling is a rich subject, combining propagation physics, data processing and a lucid view of systems/networks requirements The increasing sophistication of channel investigations and models stems from the complexification of wireless networks, from picocells to macrocells (even satellites) throughout It seems that researchers in this area are not yet jobless Séminaire Comelec, 7 juin 2012 55/56
Séminaire Comelec, 7 juin 2012 56/56