4 Radio Propagation - Fading

Lecture 4 Radio Wave Propagation: Fading and Multipath
Personal & Mobile Communication Spring 2008 Dr Daniel So
Lecture Aims
Understand the theory of multipath fading channel Know how to calculate the parameters for fading channel Learn different types of multipath fading channel Some common fading channel models
Backgrounds (I)
Large-scale propagation (Lecture 2)
Predicts mean received signal strength at large Tx-Rx distance
Hundreds or thousands of meters
Path loss, shadowing etc Importance

Proper site planning
Small-scale propagation
Characterize the rapid fluctuations over short distance or time Fading Importance
Proper receiver design to handle the rapid fluctuations
Personal & Mobile Communication Spring 2008 Dr Daniel So 3
Backgrounds (II)
Factors Affecting Fading (I)

multipath LOS pulses pulses
signal at sender
Multipath propagation
Signal arrives at Rx through different paths
Reflection, Diffraction, Scattering
Paths could arrive with different gains, phase, & delays Small dist variation can have large amplitude variation
e.g. 2 paths with perfect reflector (Plan earth model) ~ ~ d At 900MHz, 0-to-0 within 30cm ETOT = 2 ELOS sin
Factors Affecting Fading (II)

Speed of mobile/surrounding objects
The mobile can be in motion and the environment can also be varying (cars, pedestrians, etc)
Induces Doppler shift
Doppler Shift
The change of frequency due to movement
Phase change
1 v fd = = cos Frequency change 2 t
= 2l = 2vt cos
v = speed of mobile, = carrier wavelength +/-ve when moving towards/away the wave
Factors Affecting Fading (III)

Signal transmission bandwidth
If the signal bandwidth is wider than the channel bandwidth, the received signal will be distorted Consider a 2-ray model
Second ray arrive at one symbol period later than the first ray
h( ) = ( ) + 2 ( + Ts ) H ( f ) = 1 + 2 e j 2fTs
Multipath Channel Impulse Response (I)

The multipath channel can be modelled as a filter
A summation of all multipath Each multipath contain its gain, phase and delay Varies with time and distance
If the mobile is moving, distance is also related to time d = vt
Two time-related variables

t is the time variation due to motion is the time variation due to multipath delay Excess delay - relative delay compared to the first arriving path
h p (t , ) = i (t , ) exp[ j i (t , )] ( i (t ))
i(t) = i-th path excess delay at time t N = total number of arriving paths
i =0
N 1
Multipath Channel Impulse Response (II)
Multipath Channel Impulse Response (III)

Received signal y (t , ) = x(t ) h p (t , )
x(t) = passband transmitted signal hp(t,) = channel impulse response due to motion and excess delay
Baseband equivalent representation

Signal processing is done in baseband Need to obtain a baseband representation of the channel hb(t,) Consider the transmitted passband signal x(t)
Let s(t) be the baseband signal with s(t) = sI(t) + jsQ(t)
x(t ) = sI (t ) cos(2f c t ) sQ (t )sin (2f c t ) = Re s(t )e j 2f ct
10
Multipath Channel Impulse Response (IV)

Similarly, baseband received signal r(t)
y (t ) = Re r (t )e j 2f ct
Baseband equivalent channel impulse response h(t,)

h p (t , ) = Re h(t , )e j 2f ct
r (t ) =
Hence
1 s(t ) h(t , ) 2
come from down-conversion from passband to baseband To retrieve the in-phase component, multiply with carrier x(t ) cos(2f c t ) = s I (t ) cos 2 (2f c t ) sQ (t ) cos(2f ct )sin (2f c t )
1 1 s I (t )(1 + cos(4f ct )) sQ (t )sin (4f c t ) 2 2 1 = s I (t ) 2 =
11
After LPF:
Multipath Channel Impulse Response (V)

Discrete-time baseband impulse response
Divide multipath delay into discrete segments called excess delay bins i-th excess delay = i = i i = {0,K, L 1}
= delay bin width; L = maximum resolvable delay path
h(t , ) = ai (t , ) exp[ j (2f c i (t ) + i (t , ))] ( i (t ))

i =0
N 1
= a (t , k ) exp[ j (t , k )] ( k )
k =0
L 1
i(t,) = i-th path random phase shift at time t a(t,k) = baseband real amplitude of k-th bin at time t (t,k) = phase shift due of k-th bin at time t
Multipath Channel Impulse Response (VI)
13
Tap Delay Line Model

If the excess delay bin equals to the symbol period, the channel can be modelled as a tap delay-line filter
Each tap delay is exactly 1 symbol period
st
a 0 e j 0
ISI
z-1
a1e
j1
z-1
z-1
a L 2 e j L 2
z-1
aL 1e j L1
nt rt st = baseband signal, nt = noise, rt = received baseband signal

Problems with Multipath Channel

Several paths arrive within one delay bin ()
These paths cannot be resolved (NL) Different gains and phase will be combined Constructive and destructive interference can occur Fading
The received signal power changes rapidly from bin to bin When a symbol is in deep fade (all bins in that symbol period have small values), it cannot be detected correctly Diversity technique required
If the excess delay is longer than one symbol period

Inter-symbol interference (ISI) occurs Equalization techniques required
WSSUS Channel
Assume channel is wide-sense stationary (WSS)
The autocorrelation function is dependent on the time difference
i.e. Autocorrelation function is the same at any time
E h(t1 , 1 )h* (t 2 , 2 ) = Rh (t1 t 2 ; 1 , 2 )
Assume all paths are uncorrelated

Uncorrelated Scattering (US)
Rh (t1 t 2 ; 1 , 2 ) = Rh (t1 t 2 ; 1 ) ( 1 2 )
Rh=0 unless 1=2
Channel Autocorrelation Function

~ Let ak (t , ) = a(t , k )exp[ j (t , k )] Rh (t1 t 2 ; 1 , 2 ) = E [h(t1 , 1 )h* (t 2 , 2 )]
L 1 L 1 ~ ~ = E ak (t1 , 1 ) ( 1 k ) ai* (t 2 , 2 ) ( 2 i ) i =0 k =0
~ ~ = E ak (t1 , 1 )ai* (t 2 , 2 ) ( 1 k ) ( 2 i )
k =0 i =0 L 1
L 1 L 1
= Ra (t1 , t 2 ; 1 ) ( 1 k ) ( 1 2 )
k =0
~ ~ where Ra (t1 , t2 ; 1 ) = E [ak (t1 , 1 )ai* (t 2 , 1 )] is the autocorrelation function of the discretised channel bin
Note that 1=2, i.e. uncorrelated among different bins
Power Delay Profile

Characterise the power distribution against the excess delays
Average of |h(t,)|2 over time t
1 P( ) = lim T T
L 1
h(t , ) dt = Rh (0; )
2
For discrete-time model

P( ) = a(t , k ) ( k ) = Pk ( k )
2
L 1
k =0
k =0
Pk = power at k-th delay bin
Common delay profiles

Uniform: Pk being constant over k Exponential: Pk = cexp(-k/c)
Time Dispersion Parameters (I)

Determined from the power delay profile
Treat the power delay profile as a probability mass function Calculate the mean, second moment, and standard deviation for it
Mean excess delay

2 a = E [ ] = prob(k ) k = k 2 k k k ak k
Second moment
= E [ 2 ]
___ 2
2 a = prob(k ) k2 = k 2 k2 k k ak k
RMS delay spread Maximum excess delay (X dB) = X - 0

= 2
___ 2
0 = first arriving signal delay X = max excess delay within X dB of the strongest path
19
Time Dispersion Parameters (II)
20
Time Dispersion Parameters (III)
21
Coherence Bandwidth
Coherence bandwidth Bc
Frequencies separated by less than this bandwidth will have their fades highly correlated
Flat frequency spectrum within Bc
Signals will be affected differently with the frequency separation goes beyond Bc Frequency correlation higher than 0.9 and 0.5
Rh ( f1 f 2 ; ) > 0.9
Bc 1 50
Rh ( f1 f 2 ; ) > 0.5
Bc
1 5
RMS delay spread , Coherence bandwidth Frequency correlation , Coherence bandwidth

Doppler Spread and Coherence Time

Parameters to describe the time varying nature of a channel Doppler spread BD
Measure of spectral broadening due to time variation BD = 2fd-max fd-max = max Doppler shift = v/
Coherence Time Tc
Time duration that the fading parameters remain fairly constant Coherence time for correlation above 0.5:
Rh (t ;0) = Ra (t ) > 0.5
Tc 9 16f d max
Mobile speed , Doppler spread , Coherence time

Types of Fading (Delay Spread)

Flat fading
Signal BW BS < Coherence BW BC Delay spread < Symbol period TS A deep fade can degrade the system performance significantly
24
Types of Fading (Delay Spread)

Frequency selective fading
BS > BC , > TS ISI occurs => equalization technique required With proper equalization, frequency diversity can be achieved
25
Types of Fading (Doppler Spread)

Fast fading (Time selective fading)
TS > Coherence time TC , BS < Doppler spread BD Channel impulse response changes within the symbol duration Occurs for very low data rates With packetised transmission, fast fading is now commonly referred to rapid channel changes within one packet or frame With proper system design, time diversity can be obtained
26
Types of Fading (Doppler Spread)

Slow fading
TS << TC , BS >> BD Channel changes at a rate much slower than the symbol duration Very common, especially in high data rate applications
GSM900 at 200km/h, TC 1ms, User frame = 576.92 s
Quasi-static slow fading

Channel is static within a frame but varies independently from frame to frame Used in simulation to provide an averaged performance over many channel realisations
Types of Fading
Fast/slow and frequency flat/selective fading is not mutually exclusive
28
Common Channel Models - Rayleigh

Consider the channel gain at k-th bin with Nk arriving paths
~ a k = a k e j k =
N k 1 i =0
k ,i
j k ,i
N k 1 i =0
I k ,i
Q Q + jak ,i = akI + jak
I and Q is the in-phase and quadrature phase component of channel gain
Assumptions
Infinite arrival paths at the same time All paths have zero mean and similar variance (i.e. no dominant path) All path gains are statistically independent
By central limit theorem, the I and Q are Gaussian distributed

Rayleigh distribution = envelope of the sum of 2 quadrature Gaussian source (x, y)
Common Channel Models - Rayleigh

Rayleigh fading
A commonly used model for no line-of-sight (N-LOS) channels Q
Q ~ ak = akI + jak = ak e j k
ak = a + a
I2 k
Q2 k
k = tan 1
ak I ak
I & Q component akI and akQ is Gaussian distributed N(0, 2) Magnitude ak is Rayleigh distributed Phase k is uniformly distributed over 2
Rayleigh distribution pdf

2 ak ak exp 2 0 ak 2 p(ak ) = 2 ak < 0 0
E[ak2] = 22 = average channel power

Rayleigh Fading Channel Simulator
31
Common Channel Models - Rician

Rician fading
A channel with a dominant path and numerous weak multipath
i.e. with line-of-sight path
Channel fading statistics is Ricean distributed

When the dominant component fades away, the statistics degenerates to Rayleigh
PDF
r r 2 + A2 Ar I0 2 exp 2 p(r ) = 2 2 0 A 0, r 0 r<0
A = peak amplitude of the dominant signal I0(.) = zero-order Bessel function of the first kind Often described by K = A2/22
32
Common Channel Models

Example of Rayleigh and Ricean distribution
33
Common Channel Models Clarke Model (I)

Assuming all rays arrive in horizontal direction and at the same time
Channel gain with N arriving paths
~ a = ai e j i
i =0 N 1
When the mobile is moving, each ray experience different Doppler shifts
~ a (t ) = ai e j (2f i t + i )
i =0 N 1
f i = f D max cos i
where fi and i is the Doppler shift and direction of travel for path i fD-max is the maximum Doppler shift
Clarke Model (II)

Consider the channel autocorrelation function
Rh (t1 t 2 ; 1 , 2 ) = Ra (t1 , t 2 ; 1 ) ( 1 k ) ( 1 2 )
k =0 L 1
As all paths arrive at the same time (i= k i, k)

can be removed and L=1
Let t=t1, t2=t1+t
Rh (t ; ) = Ra (t )
N 1 N 1 ~ (t )a * (t + t ) = E a e j (2f it +i ) a e j (2f k (t + t )+ k ) ~ Ra (t ) = E a i i k =0 i =0
= E a e
i =0 N 1 i =0
N 1
2 j 2f i t i
] = E[a ]E[e
N 1 i =0 2 i
j 2f i t
= E ai2 E e j 2f Dmax t cos i

[ ][
]
35
Clarke Model (III)

Computing the expectation for the second term
Assuming the angle of arrival is uniformly distributed [-, ]
1 Ra (t ) = E a 2 i =0
2 i N 1 i =0
N 1
[ ]
e j 2f Dmax t cos i d i
= E ai2 I 0 (2f D max t ) = Pav I 0 (2f D max t )

where I0(x) is the zeroth order Bessel function of the first kind 1 I 0 ( x ) = e jx cos d 0 Pav is the average channel power
[ ]
Pav = i =0 E ai2
N 1
[ ]
36
Clarke Model (IV)

Doppler Spectrum
The power spectral density (PSD) is the Fourier transform of the autocorrelation function
S h ( f ) = {Rh (t ; )} = {Pav I 0 (2f D max t )}

Pav = 1 ( f f D max )2 0 f < f D max f > f D max
Significance
Convolved with signal spectrum

Spectrum will be smeared This is why Doppler spread BD = 2fd-max Receiver must be able to handle this widened bandwidth
Infinity at fD-max because of uniform arrival assumption

Simulating Doppler Spread
38
Summary
Multipath channel impulse response Parameters of multipath channel
Time dispersion parameters: mean, rms delay spread, max excess delay Coherence bandwidth Coherence time and Doppler spread
Different types of fading

Frequency flat/selective fading Fast/slow fading
Common fading channel models

Rayleigh/Ricean fading Clarke Model
Tutorial Question
Consider the provided power delay profile
Calculate the mean excess delay, rms delay spread, and the max excess delay (10dB) for the power delay profile provided Estimate the 50% coherence bandwidth of the channel Would this channel be suitable for AMPS (30kHz) or GSM (200kHz) service without the use of an equalizer?
If you are travelling at 100km/h and the system carrier freq is 900MHz
What is the maximum Doppler shift? What is the 50% coherence time? What is Doppler Spread?
Tutorial Question Solution

Max excess delay (10dB) = 5s (1)(5) + (0.1)(1) + (0.1)(2) + (0.01)(0) = 4.38s Mean excess delay =
(0.01 + 0.1 + 0.1 + 1)
Second moment RMS delay spread Coherence bandwidth
(1)(5)2 + (0.1)(1)2 + (0.1)(2)2 + (0.01)(0)2 = (0.01 + 0.1 + 0.1 + 1)

2
= 21.07 s 2
= 21.07 4.382 = 1.37 s
Bc
1 5
1 = 146kHz 5 1.37 10 6
AMPS do not need an equalizer (30kHz BW) but GSM does (200kHz BW)
Tutorial Question Solution

f = 900MHz v = 100km/hr = 100*1000/3600m/s = 27.778m/s Maximum Doppler shift
f D max vf 27.778 900 10 6 = = = = 83.333Hz c 3 108 v
50% Coherence time

Tc = 9 = 2.149ms 16f D max
Doppler Spread BD = 2*fD-max = 166.67Hz
42

4 Radio Propagation - Fading

Transféré par

Informations du document

Description originale:

Copyright

Formats disponibles

Partager ce document

Partager ou intégrer le document

Options de partage

Avez-vous trouvé ce document utile ?

Ce contenu est-il inapproprié ?

Droits d'auteur :

Formats disponibles

4 Radio Propagation - Fading

Transféré par

Droits d'auteur :

Formats disponibles

Lecture 4 Radio Wave Propagation: Fading and Multipath

Personal & Mobile Communication Spring 2008 Dr Daniel So

Personal & Mobile Communication Spring 2008 Dr Daniel So

Path loss, shadowing etc Importance

Personal & Mobile Communication Spring 2008 Dr Daniel So

Factors Affecting Fading (I)

Factors Affecting Fading (II)

Factors Affecting Fading (III)

Personal & Mobile Communication Spring 2008 Dr Daniel So

Multipath Channel Impulse Response (I)

Two time-related variables

Multipath Channel Impulse Response (II)

Personal & Mobile Communication Spring 2008 Dr Daniel So

Multipath Channel Impulse Response (III)

Baseband equivalent representation

Multipath Channel Impulse Response (IV)

Baseband equivalent channel impulse response h(t,)

Personal & Mobile Communication Spring 2008 Dr Daniel So

Multipath Channel Impulse Response (V)

h(t , ) = ai (t , ) exp[ j (2f c i (t ) + i (t , ))] ( i (t ))

Multipath Channel Impulse Response (VI)

Personal & Mobile Communication Spring 2008 Dr Daniel So

Tap Delay Line Model

nt rt st = baseband signal, nt = noise, rt = received baseband signal

Problems with Multipath Channel

If the excess delay is longer than one symbol period

E h(t1 , 1 )h* (t 2 , 2 ) = Rh (t1 t 2 ; 1 , 2 )

Assume all paths are uncorrelated

Channel Autocorrelation Function

Power Delay Profile

For discrete-time model

Pk = power at k-th delay bin

Common delay profiles

Time Dispersion Parameters (I)

Mean excess delay

RMS delay spread Maximum excess delay (X dB) = X - 0

Personal & Mobile Communication Spring 2008 Dr Daniel So

Time Dispersion Parameters (II)

Personal & Mobile Communication Spring 2008 Dr Daniel So

Time Dispersion Parameters (III)

Personal & Mobile Communication Spring 2008 Dr Daniel So

RMS delay spread , Coherence bandwidth Frequency correlation , Coherence bandwidth

Doppler Spread and Coherence Time

Mobile speed , Doppler spread , Coherence time

Types of Fading (Delay Spread)

Personal & Mobile Communication Spring 2008 Dr Daniel So

Types of Fading (Delay Spread)

Personal & Mobile Communication Spring 2008 Dr Daniel So

Types of Fading (Doppler Spread)

Personal & Mobile Communication Spring 2008 Dr Daniel So

Types of Fading (Doppler Spread)

Quasi-static slow fading

Personal & Mobile Communication Spring 2008 Dr Daniel So

Common Channel Models - Rayleigh

Q Q + jak ,i = akI + jak

I and Q is the in-phase and quadrature phase component of channel gain

By central limit theorem, the I and Q are Gaussian distributed

Common Channel Models - Rayleigh

Rayleigh distribution pdf

E[ak2] = 22 = average channel power

Rayleigh Fading Channel Simulator

Personal & Mobile Communication Spring 2008 Dr Daniel So

Common Channel Models - Rician