ECE6604 PERSONAL & MOBILE COMMUNICATIONS. Week 4. Envelope Correlation Space-time Correlation

Similar documents
ECE6604 PERSONAL & MOBILE COMMUNICATIONS. Week 3. Flat Fading Channels Envelope Distribution Autocorrelation of a Random Process

EE6604 Personal & Mobile Communications. Week 5. Level Crossing Rate and Average Fade Duration. Statistical Channel Modeling.

Chapter 2 Propagation Modeling

On the Utility of the Circular Ring Model for Wideband MIMO Channels

The Impact of Beamwidth on Temporal Channel Variation in Vehicular Channels and its Implications

Mobile Radio Communications

The Impact of Beamwidth on Temporal Channel Variation in Vehicular Channels and its Implications

Perfect Modeling and Simulation of Measured Spatio-Temporal Wireless Channels

Lecture 7: Wireless Channels and Diversity Advanced Digital Communications (EQ2410) 1

A Wideband Space-Time MIMO Channel Simulator Based on the Geometrical One-Ring Model

Emmanouel T. Michailidis Athanasios G. Kanatas

Chapter 5 Random Variables and Processes

EE303: Communication Systems

Introduction to Probability and Stochastic Processes I

EE401: Advanced Communication Theory

Lecture 6: Modeling of MIMO Channels Theoretical Foundations of Wireless Communications 1. Overview. CommTh/EES/KTH

Multiple Antennas. Mats Bengtsson, Björn Ottersten. Channel characterization and modeling 1 September 8, Signal KTH Research Focus

Problem Sheet 1 Examples of Random Processes

Spectral Analysis of Random Processes

Lecture 2. Fading Channel

ON THE SIMULATIONS OF CORRELATED NAKAGAMI-M FADING CHANNELS USING SUM OF-SINUSOIDS METHOD

A VECTOR CHANNEL MODEL WITH STOCHASTIC FADING SIMULATION

Indoor MIMO Channels: A Parametric Correlation Model and Experimental Results

5 Analog carrier modulation with noise

Digital Band-pass Modulation PROF. MICHAEL TSAI 2011/11/10

Lecture 6: Modeling of MIMO Channels Theoretical Foundations of Wireless Communications 1

Parametric Characterization and Estimation of Dispersive Multi-Path Components with SAGE in Radio Propagation Channel

SOLUTIONS TO HOMEWORK ASSIGNMENT 1

Module 4. Signal Representation and Baseband Processing. Version 2 ECE IIT, Kharagpur

Digital Communications

A GENERALISED (M, N R ) MIMO RAYLEIGH CHANNEL MODEL FOR NON- ISOTROPIC SCATTERER DISTRIBUTIONS

Communication Systems Lecture 21, 22. Dong In Kim School of Information & Comm. Eng. Sungkyunkwan University

Classes of sum-of-cisoids processes and their statistics for the modeling and simulation of mobile fading channels

A TIME-VARYING MIMO CHANNEL MODEL: THEORY AND MEASUREMENTS

Power Spectral Density of Digital Modulation Schemes

EAS 305 Random Processes Viewgraph 1 of 10. Random Processes

Es e j4φ +4N n. 16 KE s /N 0. σ 2ˆφ4 1 γ s. p(φ e )= exp 1 ( 2πσ φ b cos N 2 φ e 0

A MIMO MOBILE-TO-MOBILE CHANNEL MODEL: PART I THE REFERENCE MODEL

7 The Waveform Channel

Design and Simulation of Narrowband Indoor Radio Propagation Channels Under LOS and NLOS Propagation Conditions

Point-to-Point versus Mobile Wireless Communication Channels

EE6604 Personal & Mobile Communications. Week 12a. Power Spectrum of Digitally Modulated Signals

5.9 Power Spectral Density Gaussian Process 5.10 Noise 5.11 Narrowband Noise

Fourier spectra of radially periodic images with a non-symmetric radial period

EE6604 Personal & Mobile Communications. Week 13. Multi-antenna Techniques

Stochastic Processes. M. Sami Fadali Professor of Electrical Engineering University of Nevada, Reno

A Non-Stationary MIMO Vehicle-to-Vehicle Channel Model Based on the Geometrical T-Junction Model Ali Chelli and Matthias Pätzold

Computational Data Analysis!

Parameter Estimation

Ranging detection algorithm for indoor UWB channels

EE4601 Communication Systems

s o (t) = S(f)H(f; t)e j2πft df,

PROPAGATION PARAMETER ESTIMATION IN MIMO SYSTEMS USING MIXTURE OF ANGULAR DISTRIBUTIONS MODEL

Solution to Homework 1

Microphone-Array Signal Processing

ENSC327 Communications Systems 2: Fourier Representations. Jie Liang School of Engineering Science Simon Fraser University

Lecture 8 ELE 301: Signals and Systems

Engineering Notebook

Classes of Sum-of-Sinusoids Rayleigh Fading Channel Simulators and Their Stationary and Ergodic Properties

Wireless Communications

This is a Gaussian probability centered around m = 0 (the most probable and mean position is the origin) and the mean square displacement m 2 = n,or

Signals and Systems: Part 2

SET 1. (1) Solve for x: (a) e 2x = 5 3x

On Coding for Orthogonal Frequency Division Multiplexing Systems

On capacity of multi-antenna wireless channels: Effects of antenna separation and spatial correlation

Signals and Spectra - Review

THE number of MIMO channel models available in the

ELEG 3143 Probability & Stochastic Process Ch. 6 Stochastic Process

A Non-Stationary Mobile-to-Mobile Channel Model Allowing for Velocity and Trajectory Variations of the Mobile Stations

Figure 3.1 Effect on frequency spectrum of increasing period T 0. Consider the amplitude spectrum of a periodic waveform as shown in Figure 3.2.

On the Average Crossing Rates in Selection Diversity

Iit Istituto di Informatica e Telematica

2.1 Introduction 2.22 The Fourier Transform 2.3 Properties of The Fourier Transform 2.4 The Inverse Relationship between Time and Frequency 2.

Comparative analysis of statistical models for the simulation of Rayleigh faded cellular channels

Envelope PDF in Multipath Fading Channels with Random Number of Paths and Nonuniform Phase Distributions

ESTIMATION OF VELOCITY IN UNDERWATER WIRELESS CHANNELS

Analog Electronics 2 ICS905

Chapter 5 Notes. 5.1 Using Fundamental Identities

Trig Identities, Solving Trig Equations Answer Section

Winter 2014 Practice Final 3/21/14 Student ID

Solutions: Homework 7

Direct-Sequence Spread-Spectrum

SENSOR ERROR MODEL FOR A UNIFORM LINEAR ARRAY. Aditya Gadre, Michael Roan, Daniel Stilwell. acas

SIMULATION of mobile radio channels is commonly used

2016 Spring: The Final Exam of Digital Communications

Implementation of a Space-Time-Channel-Filter

Fading Statistical description of the wireless channel

Carrier Transmission. The transmitted signal is y(t) = k a kh(t kt ). What is the bandwidth? More generally, what is its Fourier transform?

Detecting Parametric Signals in Noise Having Exactly Known Pdf/Pmf

Spherical trigonometry

III. Spherical Waves and Radiation

Appendix A PROBABILITY AND RANDOM SIGNALS. A.1 Probability

ENGIN 211, Engineering Math. Fourier Series and Transform

Fourier Analysis and Power Spectral Density

EE401: Advanced Communication Theory

Ph.D. Defense, Aalborg University, 23 January Multiple-Input Multiple-Output Fading Channel Models and Their Capacity

Parametric Signal Modeling and Linear Prediction Theory 1. Discrete-time Stochastic Processes

We are IntechOpen, the world s leading publisher of Open Access books Built by scientists, for scientists. International authors and editors

SIMULATION OF DOUBLY-SELECTIVE COMPOUND K FADING CHANNELS FOR MOBILE-TO-MOBILE COMMUNICATIONS. Yuan Liu, Jian Zhang, and Yahong Rosa Zheng

Radiative Processes in Astrophysics. Lecture 9 Nov 13 (Wed.), 2013 (last updated Nov. 13) Kwang-Il Seon UST / KASI

Transcription:

ECE6604 PERSONAL & MOBILE COMMUNICATIONS Week 4 Envelope Correlation Space-time Correlation 1

Autocorrelation of a Bandpass Random Process Consider again the received band-pass random process r(t) = g I (t)cos2πf c t g Q (t)sin2πf c t where g I (t) = N n=1 g Q (t) = N n=1 C n cosφ n (t) C n sinφ n (t) Assuming that r(t) is wide-sense stationary, the autocorrelation of r(t) is φ rr (τ) = E[r(t)r(t+τ)] = E[g I (t)g I (t+τ)]cos2πf c τ +E[g Q (t)g I (t+τ)]sin2πf c τ = φ gi g I (τ)cos2πf c τ φ gi g Q (τ)sin2πf c τ where E[ ] is the ensemble average operator, and φ gi g I (τ) = E[g I (t)g I (t+τ)] φ gi g Q (τ) = E[g I (t)g Q (t+τ)]. Note that the wide-sense stationarity of r(t) imposes the condition φ gi g I (τ) = φ gq g Q (τ) φ gi g Q (τ) = φ gq g I (τ). 2

Auto- and Cross-correlation of Quadrature Components The phases φ n (t) are statistically independent random variables at any time t, uniformly distributed over the interval [ π, π). The azimuth angles of arrival, θ n are all independent due to the random placement of scatterers. Also, in the limit N, the discrete azimuth angles of arrival θ n can be replaced by a continuous random variable θ having the probability density function p(θ). By using the above properties, the auto- and cross-correlation functions can be obtained as follows: φ gi g I (τ) = φ gq g Q (τ) = lim N E τ,θ,φ [g I(t)g I (t+τ)] = Ω p 2 E θ[cos(2πf m τ cosθ)] φ gi g Q (τ) = φ gq g I (τ) = lim N E τ,θ,φ [g I(t)g Q (t+τ)] = Ω p 2 E θ[sin(2πf m τ cosθ)] τ = (τ 1,τ 2,...,τ N ) θ = (θ 1,θ 2,...,θ N ) φ = (φ 1,φ 2,...,φ N ) and Ω p is the total received envelope power. Ω p = E[g 2 I (t)]+e[g2 Q (t)] = N Cn 2 n=1 3

2-D Isotropic Scattering Evaluation of the expectations for the auto- and cross-correlation functions requires the azimuth distribution of arriving plane waves p(θ), and the receiver antenna gain pattern G(θ), as a function of the azimuth angle θ. With 2-D isotropic scattering, the plane waves are confined to the x y plane and arrive uniformly distributed angle of incidence, i.e., p(θ) = 1 2π, π θ π With 2-D isotropic scattering and an isotropic receiver antenna with gain G(θ) = 1,θ [ π, π), the auto- and cross-correlation functions become φ gi g I (τ) = = Ω p 2 J 0(2πf m τ) φ gi g Q (τ) = 0 where J 0 (x) = 1 π π cos(xcosθ)dθ 0 is the zero-order Bessel function of the first kind. 4

1.0 0.8 Autocorrelation, φ ri r I (τ)/(ω p /2) 0.6 0.4 0.2 0.0-0.2-0.4-0.6-0.8-1.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Time Delay, f m τ Normalized autocorrelation function of the quadrature components of the received complex envelope with 2-D isotropic scattering and an isotropic receiver antenna. 5

Doppler Spectrum The autocorrelation function and power spectral density (psd) are Fourier transform pairs. S gg (f) = φ gg(τ)e j2πfτ dτ φ gg (τ) = S gg(f)e j2πfτ df The autocorrelation of the received complex envelope g(t) = g I (t)+jg Q (t) is φ gg (τ) = 1 2 E[g (t)g(t+τ)] = φ gi g I (τ)+jφ gi g Q (τ) The Fourier transform of φ gg (τ) gives the Doppler psd S gg (f) = S gi g I (f)+js gi g Q (f). Sometimes S gg (f) is just called the Doppler spectrum. 6

Bandpass Doppler Spectrum We can also relate the power spectrum of the complex envelope g(t) to that of the band-pass process r(t). We have φ rr (τ) = Re [ φ gg (τ)e j2πf cτ ]. By using the identity Re[z] = z +z 2 and the property φ gg (τ) = φ gg( τ), it follows that the band-pass Doppler psd is S rr (f) = 1 2 [S gg(f f c )+S gg ( f f c )]. Since φ gg (τ) = φ gg( τ), the Doppler spectrum S gg (f) is always a real-valued function of frequency, but not necessarily even. However, the band-pass Doppler spectrum S rr (f) is always real-valued and even. 7

Isotropic Scattering For 2-D isotropic scattering, the psd and cross psd of g I (t) and g Q (t) are S gi g I (f) = F[φ gi g I (τ)] = S gi g Q (f) = 0 Ω p 1 2πf m 1 ( fm f )2, f f m 0, otherwise 10.0 Power Spectrum, S gi g I (f)/(ω p /2πf m ) (db) 5.0 0.0 1.0 0.5 0.0 0.5 1.0 Normalized Doppler Frequency, f/f m Normalized psd of the quadrature components of the received complex envelope with 2-D isotropic scattering and an isotropic receiver antenna. Sometimes this is called the CLASSICAL Doppler power spectrum. 8

Non-isotropic Scattering Rician Fading Suppose that the propagation environment consisting of a strong specular component plus a scatter component. The azimuth distribution p(θ) might have the form p(θ) = 1 K +1ˆp(θ)+ K K +1 δ(θ θ 0) where ˆp(θ) is the continuous AoA distribution of the scatter component, θ 0 is the AoA of the specular component, and K is the ratio of the received specular to scattered power. One such scattering environment, assumes that the scatter component exhibits 2-D isotropic scattering, i.e., ˆp(θ) = 1/(2π),θ [ π,π). The correlation functions φ gi g I (τ) and φ gi g Q (τ) are φ gi g I (τ)= 1 Ω p K +1 φ gi g Q (τ)= K K +1 2 J 0(2πf m τ)+ K K +1 Ω p 2 sin(2πf mτ cosθ 0 ). Ω p 2 cos(2πf mτ cosθ 0 ) 9

0.7 0.6 0.5 p(θ) 0.4 0.3 K=1 θ 0 = π/2 0.2 0.1 0 4 π 2 0 2 4π Arrival Angle, θ Plot of p(θ) vs. θ with 2-D isotropic scattering plus a LoS or specular component arriving at angle θ 0 = π/2. 10

The azimuth distribution p(θ) = 1 K +1ˆp(θ)+ K K +1 δ(θ θ 0) yields a complex envelope having a Doppler spectrum of the form S gg (f) = 1 K +1 Sc gg(f)+ K K +1 Sd gg(f) (1) where Sgg d (f) is the discrete portion of the Doppler spectrum due to the specular component and Sgg(f) c is the continuous portion of the Doppler spectrum due to the scatter component. For the case when ˆp(θ) = 1/(2π),θ [ π,π], the power spectrum of g(t) = g I (t)+jg Q (t) is S gg (f) = 1 K+1 Ω p 2πf m 1 + K K+1 1 (f/fm ) 2 Ω p 2 δ(f f mcosθ 0 ) 0 f f m 0 otherwise Note the discrete tone at frequency f c +f m cosθ 0 due to the line-of-sight or specular component arriving from angle θ 0.. 11

Non-isotropic scattering Other Cases Sometimes the azimuth distribution p(θ) may not be uniform, a condition commonly called non-isotropic scattering. Several distributions have been suggested to model non-isotropic scattering. Once possibility is the Gaussian distribution p(θ) = 1 exp 2πσS (θ µ)2 2σ 2 S where µ is the mean AoA, and σ S is the rms AoA spread. Another possibility is the von Mises distribution 1 p(θ) = 2πI 0 (k) exp[kcos(θ µ)], where θ [ π,π), I 0 ( ) is the zeroth-order modified Bessel function of the first kind, µ [ π,π) is the mean AoA, and k controls the spread of scatterers around the mean. 12

1.4 1.2 1 k=2 k=5 k=10 0.8 p(θ) 0.6 0.4 0.2 0 4 π 2 0 2 π 4 Arrival Angle, θ Plot of p(θ) vs. θ for the von Mises distribution with a mean angle-to-arrival µ = π/2. 13

Calculating the Doppler Spectrum The Doppler spectrum can be derived by using a different approach that is sometimes very useful because it can avoid the need to evaluate integrals. The Doppler spectrum can be expressed as S gg (f) df = Ω p (G(θ)p(θ) + G( θ)p( θ)) dθ. 2 The Doppler frequency associated with the incident plane wave arriving at angle θ is f = f m cos(θ), and, hence, Therefore, where df = f m sin(θ)dθ = f 2 m f 2 dθ. S gg (f) = Ω p/2 f 2 m f2(g(θ)p(θ) + G( θ)p( θ)), θ = cos 1 (f/f m ). Hence, if p(θ) and G(θ) are known, the Doppler spectrum can be easily calculated. For example, with 2-D isotropic scattering and an isotropic antenna G(θ)p(θ) = 1/(2π). 14

Space-time Correlation Functions Many mobile radio systems use antenna diversity, where spatially separated receiver antennas provide multiple faded replicas of the same information bearing signal. The spatial decorrelation of the channel tell us the required spatial separation between antenna elements so that they will be sufficiently decorrelated. Sometimes it is desirable to simultaneously characterize both the spatial and temporal correlation characteristics of the channel, e,g, when using space-time coding. This can be described by the space-time correlation function. To obtain the spatial or space-time correlation functions, we must specify some kind of radio scattering geometry. 15

Spatial Correlation at the Mobile Station ( n) S M n n2 n1 ( n ) M (1) A M M M v O B (2) A M O M M D R Single-ring scattering model for NLoS propagation on the forward link of a cellular system. The MS is surrounded by a scattering ring of radius R and is at distance D from the BS, where R D. 16

Model Parameters O B : base station location O M : mobile station location D: LoS distance from base station to mobile station R: scattering radius γ M : mobile station moving direction w.r.t x-axis v: mobile station speed θ M mobile station array orientation w.r.t. x-axis A (i) M: location of ith mobile station antenna element δ M : distance between mobile station antenna elements S (n) M : location of nth scatterer. α (n) M : angle of arrival from the nth scatterer. ǫ n : distance O B S (n) M. ǫ ni : distance S (n) M A (i) M. 17

Received Complex Envelope The channel from O B to A (q) M has the complex envelope g q (t) = N n=1 C n e jφ n j2π(ǫ n +ǫ nq )/λ c e j2πf mtcos(α (n) M γ M), q = 1,2 where ǫ n and ǫ nq denote the distances O B S (n) M and S (n) M A (q) M, q = 1,2, respectively, and φ n is a uniform random phase on the interval [ π,π). From the Law of Cosines, the distances ǫ n and ǫ nq can be expressed as a function of the angle-of-arrival α (n) M as follows: ǫ 2 n = D 2 +R 2 +2DRcosα (n) M Note sign change since the angle is π α (n) M ǫ 2 nq = [(1.5 q)δ M ] 2 +R 2 2(1.5 q)δ M Rcos(α (n) M θ M ),q = 1,2. Assuming that R/D 1 (local scattering), δ M R and 1±x 1±x/2 for small x, we have Hence, ǫ n D +Rcosα (n) M ǫ nq R (1.5 q)δ M cos(α (n) M θ M ),q = 1,2. g q (t) = N n=1 ( C n e jφ n j2π D+Rcosα (n) M +R (1.5 q)δ M cos(α (n) ) M θ M) /λ c e j2πf mtcos(α (n) M γ M),q = 1,2. 18

Space-time Correlation Function Thespace-timecorrelationfunctionbetween thetwo complex fadedenvelopes g 1 (t)andg 2 (t) is φ g1,g 2 (δ M,τ) = 1 2 E[g 1(t)g 2 (t+τ)] The space-time correlation function between g 1 (t) and g 2 (t) can be written as φ g1,g 2 (δ M,τ) = Ω p 2N N n=1 E [ e j2π(δ M/λ c )cos(α (n) M θ M) e j2πf mτ cos(α (n) M γ M) ]. Since the number of scatters is infinite, the discrete angles-of-arrival α (n) M can be replaced with a continuous random variable α M with probability density function p(α M ). Hence, the space-time correlation function becomes φ g1,g 2 (δ M,τ) = Ω p 2 2π where a = 2πf m τ and b = 2πδ M /λ c. 0 e jbcos(α M θ M ) e jacos(α M γ M ) p(α M )dα M. 19

2-D Isotropic Scattering Forthecaseof2-Disotropicscatteringwithanisotropicreceiveantennas,p(α M ) = 1/(2π), π α M π, and the space-time correlation function becomes φ g1,g 2 (δ M,τ) = Ω ( ) p 2 J 0 a 2 +b 2 2abcos(θ M γ M ) Thespatialandtemporalcorrelationfunctionscanbeobtainedbysettingτ = 0andδ M = 0, respectively. This gives Finally, we note that φ g1,g 2 (δ M ) = φ g1,g 2 (δ M,0) = Ω p 2 J 0(2πδ M /λ c ) φ gg (τ) = φ g1,g 2 (0,τ) = Ω p 2 J 0(2πf m τ) f m τ = v τ = δ M λ c λ c For this scattering environment, the normalized time f m τ is equivalent to the normalized distance δ M /λ c. The antenna branches are uncorrelated if they are separated by δ M 0.5λ c.. 20

1.0 0.8 Autocorrelation, φ ri r I (τ)/(ω p /2) 0.6 0.4 0.2 0.0-0.2-0.4-0.6-0.8-1.0 0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 Time Delay, f m τ Temporal and spatial correlation functions at the MS with 2-D isotropic scattering and an isotropic receiver antenna. Note that f m τ = δ M /λ c. 21

Spatial Correlation at the Base Station v ( m) S M m1 (1) A B ( m) M M m ( m) B B O M m2 O B B (2) A B R D Single-ring scattering model for NLoS propagation on the reverse link of a cellular system. The MS is surrounded by a scattering ring of radius R and is at distance D from the BS, where R D. 22

Model Parameters O B : base station location O M : mobile station location D: LoS distance from base station to mobile station R: scattering radius γ M : mobile station moving direction w.r.t x-axis v: mobile station speed θ B base station array orientation w.r.t. x-axis A (i) B : location of ith base station antenna element δ B : distance between mobile station antenna elements S (m) M : location of mth scatterer. α (m) M : angle of departure to the nth scatterer. ǫ m : distance S (m) M O B. ǫ mi : distance S (m) M A (i) B. 23

Received Complex Envelope The channel from O M to A (q) B has the complex envelope g q (t) = N m=1 C m e jφ m j2π(r+ǫ mq )/λ c e j2πf mtcos(α (m) M γ M), q = 1,2 (2) where ǫ mq denote the distance S (m) M A (q) B, q = 1,2, and φ m is a uniform random phase on ( π,π]. To proceed further, we need to express ǫ mq as a function of α (m) ApplyingtheLawofCosinestothetriangle S (m) M O B A (q) B, thedistanceǫ mq canbeexpressed as a function of the angle θ (m) B θ B as follows: ǫ 2 mq = [(1.5 q)δ B ] 2 +ǫ 2 m 2(1.5 q)δ B ǫ m cos(θ (m) B θ B ),q = 1,2. (3) where ǫ m is the distance S (m) M O B. By applying the Law of Sines to the triangle O M S (m) M O B we obtain following identity ǫ m sinα (m) M = sin ( R π θ (m) B D ) = ( sin π α (m) ( M π θ (m) B M. )). 24

Received Complex Envelope Since the angle π θ (m) B is small, we can apply the small angle approximations sinx x and cosx 1 for small x, to the second equality in the above identity. This gives or R (π θ (m) B ) sin It follows that the cosine term in (2) becomes ( D π α (m) M (π θ (m) B ) (R/D)sin(π α (m) M ). cos(θ (m) B θ B ) = cos(π θ B (π θ (m) B )) = cos(π θ B )cos(π θ (m) B )+sin(π θ B )sin(π θ (m) B ) cos(π θ B )+sin(π θ B )(R/D)sin(π α (m) M ) = cos(θ B )+(R/D)sin(θ B )sin(α (m) M ) (4) Using the approximation in (4) in (2), along with δ B /ǫ m 1, gives ǫ 2 mq ǫ2 m 1 2(1.5 q) δ B ǫ m ) [ (R/D)sin(θ B )sin(α (m) M ) cos(θ B )]. 25

Received Complex Envelope Applying the approximation 1±x 1±x/2 for small x, we have [ ǫ mq ǫ m (1.5 q)δ B (R/D)sin(θ B )sin(α (m) M ) cos(θ B ) ]. (5) Applying the Law of Cosines to the triangle O M S (m) M O B we have ǫ 2 m = D2 +R 2 2DRcos(α (m) M ) ] D 2 [ 1 2(R/D)cos(α (m) M ) and again using the approximation 1±x 1±x/2 for small x, we have ǫ m D Rcos(α (m) M ) (6), Finally, using (5) in (4) gives ǫ mq D Rcos(α (m) M ) (1.5 q)δ B [ (R/D)sin(θ B )sin(α (m) M ) cos(θ B ) ]. (7) Substituting (7) into (2) gives the result g q (t) = N C m e jφ m+j2πf m tcos(α (m) M γ M) m=1 ( e j2π R+D Rcos(α (m) M ) (1.5 q)δ B [ (R/D)sin(θ B )sin(α (m) ]) M ) cos(θ B) /λ c, which no longer depends on the ǫ mq and is a function of the angle of departure α (m) M. (8) 26

Space-time Correlation Function Thespace-timecorrelationfunctionbetween thetwo complex fadedenvelopes g 1 (t)andg 2 (t) at the BS is once again given by φ g1,g 2 (δ B,τ) = 1 2 E[g 1(t)g 2 (t+τ)] Using (8), the space-time correlation function between g 1 (t) and g 2 (t) can be written as φ g1,g 2 (δ B,τ) = Ω p 2N N m=1 E [ [ e j2π(δ B/λ c ) (R/D)sin(θ B )sin(α (m) ] M ) cos(θ B) e j2πf mτ cos(α (m) M γ M) ]. Since the number of scatters around the MS is infinite, the discrete angles-of-departure α (m) M can be replaced with a continuous random variable α M with probability density function p(α M ). Hence, the space-time correlation function becomes. φ g1,g 2 (δ B,τ) = Ω p 2 π where a = 2πf m τ and b = 2πδ B /λ c. π e jacos(α M γ M ) e jb[(r/d)sin(θ B)sin(α M ) cos(θ B )] p(α M )dα M, 27

2-D Isotropic Scattering For the case of 2-D isotropic scattering with an isotropic MS transmit antenna, p(α M ) = 1/(2π), π α M π, and the space-time correlation function becomes φ g1,g 2 (δ B,τ) = Ω p 2 e jbcos(θ B) J 0 ( a 2 +b 2 (R/D) 2 sin 2 (θ B ) 2ab(R/D)sin(θ B )sin(γ M ) ). The spatial and temporal correlation functions can be obtained by setting τ = 0 and and δ B = 0, respectively. The temporal correlation function φ gg (τ) = φ g1,g 2 (0,τ) = Ω p 2 J 0(2πf m τ) which matches our result for the received signal at a mobile station. The spatial correlation function is φ g1,g 2 (δ B ) = φ g1,g 2 (δ B,0) = Ω p 2 e jbcos(θb) J 0 (b(r/d)sin(θ B )) = Ω p 2 e jbcos(θb) ( ) J 0 (2πδ B /λ c )(R/D)sin(θ B ) Observe that a much greater spatial separation is required to achieve a given degree of envelope decorrelation at the BS as compared to the MS. This can be readily seen by the term R/D 1 in the argument of the Bessel function. 28

Normalized Spatial Crosscovariane, φ g1 g 2 (δ B ) 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 φ B =π/2 φ B =π/3 φ B =π/6 0 0 5 10 15 20 25 30 BS Antenna Separation, δ B /λ Envelope crosscorrelation magnitude at the base station for different base station antenna orientation angles, θ B ; D = 3000 m, R = 60 m. Broadside base station antennas have the lowest crosscorrelation. 29

Normalized Spatial Crosscovariance, φ g1,g 2 (δ B ) 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 R=30 R=60 R=90 0 0 5 10 15 20 25 30 BS Antenna Separation, δ B /λ c Envelope crosscorrelation magnitude at the base station for θ B = π/3 and various scattering radii, R; D = 3000 m. Smaller scattering radii will result in larger a crosscorrelations. 30