Channel model

Outline 
Outline
• Introduction to Channel Modeling
• Representation of a Channel by Filter
• Stochastic/Statistical Channel Modeling Considerations
• Wideband‐Time Dispersive‐Channel Modeling 
C id
Considerations
i
• Rayleigh Fading Model
• Rician fading model
f d d l
• Nakagami Fading Model
• Comparing Rayleigh, Rician and Nakagami Fading models
™ Channels may be time varying or static. (Multipath effect makes
the channel time varying and depending upon the constructive or
d t ti interference,
destructive i t f th quality
the lit off the
th received
i d signal
i l will
ill
vary).
™ Effect of mobility
mobilit is that,
that channel varies
aries with
ith user’s
ser’s location and
time, which results in rapid fluctuations of received power. Less
variations will be observed, the slower you move.
™ Channels may be time dispersive or non dispersive. (due to
dispersion pulse spreading will be observed, which will result in
ISI
S effect)
ff )
™ Channels may be linear or nonlinear.
™ All channels act as a low pass filter under certain conditions as
they show pulse spreading effect.
™ Channel may be fast fading or slow fading, frequency selective
or flat fading.
 Two main Channel model
Two main Channel model
• Outdoor Propagation model 
• Indoor Propagation model
Indoor Propagation model
• There are many constraints to model a channel.
• There may be different environment for different situation
and
d different
diff t weather.
th
• There may be different objects in a room, or different
surrounding conditions.
conditions
• Hence, it is very difficult to predict the number of reflected
rays and whether the constructive interference (add in phase)
will occur or destructive.
• Because of this, channel model is based on probability and its
behavior is represented by Probability Density Function
(PDF).
 Channel response over fading and 
nonfading environment 
• In non‐fade channel if the C/I / ratio is increased slightly, there will
be considerable drop in bit errors.
• In a fading channel the received signal is very weak and many bit
error occurs. This phenomenon remains present, even if the
(average) signal‐to‐noise ratio is large.
 The effects due to multipath reception
The effects due to multipath reception
Application Effect

Fast moving user: rapid fluctuations of the signal amplitude and phase

Wideband (digital) signal: dispersion and intersymbol interference

Analog television signal: "ghost" images (shifted slightly to the right)

MultiCarrier signal: different attenuation at different (sub

(sub-)carriers
)carriers and at
different locations
Stationary user of a narrowband good reception at some locations and frequencies; poor
system:
t reception
ti att other
th locations
l ti andd frequencies.
f i
Satellite positioning system: strong delayed reflections may cause a severe
miscalculation of the distance between user and satellite.
This can result in a wrong estimate of the position.
• Multipath
Multipath reception scenario shown in (a)  can 
reception scenario shown in (a) can
be modeled as summation of signal and (b) its 
phase delayed versions
h d l d i
( )
(a)                                        (b)
( )
• Although channel fading is experienced as an unpredictable
and stochastic phenomenon to the user’s device or the
system planner,
planner powerful models have been developed that
can accurately predict average system performance.
• Countermeasures can be used to avoid system failure even if
the channel exhibits fades at particular frequencies of
particular locations.
• Such countermeasures are mostly used at the receiver side
and applied for channel modeling.
 Representation of a digital Channel by Filter
• Wireless channel for mobile radio communication is a challenging
medium that requires careful design for reliable transmission.
• We
W can consider,
id
– Stationary channel or
– Time varying
y g channel
• On stationary channel, if an impulse is transmitted, then at the Rx,
multiple delayed versions of impulses will be received at different
instants off the
h time due
d to multipath
l h with
h reduced
d d amplitudes.
l d
• They are just like delayed samples and hence, difference equation with
coefficient values can be used to represent this concept,
concept which is the
equation for an FIR filter.
 r(n) = w0s(n)+w1s(n‐1)+w2s(n‐2)+‐‐‐‐‐‐‐
• In this
h equation the
h coefficients
ff w’s
’ are the
h complex
l channel
h l
impulse response or fading coefficients.
• For
F the
th time
ti varying
i channels,
h l this
thi whole
h l scenarioi becomes
b
dynamic.
• The equation forms the basis for an equalizer design.
design
• The channel model for time varying channel can be
approximated by Auto Regressive (AR) model as shown.
shown
h(n+1) =w1 h(n) + w2 h(n‐1) +…. wn h(n‐N) + w(n)
Where, w
w’ss are the coefficient value to estimate the channel.
• The concept will be better highlighted in the topic of adaptive
channel equalization and channel estimation where weights
are to be adjusted.
• In general, an autoregressive (AR) model is a type of random
process which is often used to model and predict various
types of natural and social phenomena.
• In our case, the received time domain signal r(n) is a function
off the
th transmitted
t itt d signal
i l s(n),
( ) the
th estimated
ti t d channel
h l transfer
t f
function he(n) and AWGN w(n). It can be expressed as

r(n) = s(n)*he(n) + w(n) 
• Where,
h ‘*’ denote
d convolution
l
Indoor propagation channel 
Indoor propagation channel
model
 Stochastic/Statistical Channel 
Modeling Considerations
• IIn principle,
i i l three
h diff
different d
domains
i determine
d i the
h radio
di
signal transmission:
– Physical conditions selected or operation scenario
– Dispersion phenomena of wave propagation and
– Transceiver characteristics.

• Dispersion
p in frequency,
q y time, direction and p
polarization is a
crucial aspect of radio communication.
• Basically these modeling considerations are applicable to
I d
Indoor propagation
ti channel
h l model
d l (narrow
( b d channel
band h l
model)
• We have to distinguish between multipath propagation where
short‐term or small‐scale fluctuations (fast fading) and long‐
term or large‐scale
large scale fluctuations comprising gradual and
sudden changes of path parameters mainly because of
movements either of the mobile terminals ((MT)) or reflectors
and scatters, respectively.

• Transceiver characteristics contributing to the SRCM

(Stochastic Radio Channel Model) are the parameters which
describe the MT movement and the antenna configuration
with its radiation pattern and diversity properties.
Effects to be considered for stochastic 
radio channel model 
• Concluding from this table, the consideration of all aspects
leads to a rather high computational complexity of the SRCM.
• The dedication of the SRCM to a certain class of operation
scenario,
i i.e.
i th delivery
the d li off indoor
i d services
i with
ith a given
i
maximum bit rate (bandwidth) at a prescribed frequency
range is a first step toward a simplified model approach.
range, approach
• Moreover, the identification of a few types of environment
allows a remarkable complexity
p y reduction.
• Thus, four important indoor situations can be chosen:
– small rooms,
– large rooms,
– factory halls or
– corridors
• The allocated frequency range plays a key role
w.r.t the p
path loss, while the room dimensions
influence directly the delay spread.
• The evaluation of future generations of indoor
systems equipped with smart antennas and
intended forf unrestricted mobileb l operation
will require full complexity channel modeling.
• A SRCM has been developed in order to simulate realistic
channel impulse responses (CIR) according to a wide range of
possible physical situations within a given category of
environments.
• If the channel is estimated on the basis of channel statistics it
is called blind method of channel estimation. The complexity
of this method is very high.
• Among the different algorithms for the calculation of the
SRCM parameters from the measured data the Space‐Altering
Generalized Expectation‐Maximization (SAGE) algorithm
reveals to be a powerful tool.
• The SAGE algorithm will be a promising signal processing
scheme for the next generation of wireless communication
systems using smart antennas.
• some processing stages for the algorithm in which fading and
AWGN can be added artificially to measure the channel
effects is shown in figure.
 Wideband‐Time Dispersive‐Channel 
Modeling Considerations
• Most of the latest communication systems are digital and
wideband, like CDMA and OFDM .
• At the
th same time,
ti wireless
i l channel
h l is
i a dielectric
di l t i medium
di andd
hence the refractive index of the channel, phase velocity,
wave and media propagation constants etc. etc are the key
parameters.
• So over a wide rangeg of frequencies
q in the transmission
bandwidth, frequency dependent performance is obtained.
• Wide band channel Modeling may be done using
measurements mainly on two areas
– Collection of results from extensive measurement operation.
– From the measurements the channel models are derived, which should
fulfill the following two criteria
• They ey must
us be ssimple
p e eenough
oug to
o aallow
o aan aanalytical
a y ca cocomputation
pu a o o of
basic system performance.
• They must be very close to the physical reality; in other words the
performance computed by these models must be close to the
performance measured in actually existing mobile radio channels.
• These requirements
q are contradictory,
y, so models of different
complexity and accuracy have been developed.
 Small Scale Fading models
Small‐Scale Fading models
• “Small‐scale (fast) fading” …
– quick amplitude variations of the received signal due to / 
constructive / destructive interference of multipath

Three common models
• Rayleigh fading
Rayleigh fading
• Rician‐distributed fading (K‐factor)
• Nakagami 
Rayleigh Fading Model
Rayleigh Fading Model 
 Rayleigh Fading Model 
y g g
• The model was first p proposed
p in a comment p
paper
p written byy
Lord Rayleigh in 1889.
• A mobile radio signal r(t) can be artificially characterized by
t
two components t m(t)(t) andd r0(t) based
b d on natural
t l physical
h i l
phenomena.
r (t) = m(t)r0(t)
• The component m(t) is called local mean, long‐term fading, or
lognormal fading and its variation is due to the terrain contour
between the base station and the mobile unit.
• The factor r0(t) is called multipath fading, short‐term fading,
or Rayleigh fading and its variation is due to the waves
reflected from the surrounding buildings and other structures.
A mobile radio signal fading representation. (a) A mobile signal fading. (b) A short‐
term signal fading.
• Rayleigh
y g fadingg is also called multipath
p fadingg in the
mobile radio environment. When these multipath
waves bounce back and forth due to the buildings g
and houses, they form many standing‐wave pairs in
space.
p
• When a mobile unit is standing still, its receiver only
receives a signal strength at that spot,
spot so a constant
signal is observed.
• When a mobile unit is moving,
moving the fading structure
of the wave in the space is received. It is a multipath
fading.
fading
 • In mobile radio channels, the Rayleigh distribution is
commonl used
commonly sed to describe the statistical time varying
ar ing nature
nat re
of the received envelope of a flat fading signal, or the
envelope of an individual multipath component.

• Signal amplitude (in dB) Vs time

f an antenna moving
for i at constant
velocity exhibiting Rayleigh fading.
• Deep fades occur occasionally.
occasionally
• Although fading is a random
process, deep fades have a
tendency to occur approximately
every half a wavelength of motion.
• The Rayleigh distribution is a good model for channel
propagation when there is no strong line of sight path from
transmitter to receiver.
receiver
– Ex. This can be used to represent the channel conditions seen on a
busy street in a city, where the base station is hidden behind a building
several blocks away and the arriving signal is bouncing off many
scattering objects in the local area.

• The PDF (probability density

function )of Rayleigh
distribution shown.
Rayleigh fading distribution
In mobile radio channels,
channels the Rayleigh
distribution is commonly used to describe the
statistical
t ti ti l time
ti varying
i nature
t off the
th received
i d
envelope of an individual multipath component.
Multipath Reflections due to local objects arrive at stationary receiver
• nth path has electric field strength En & relative phase θn
• complex phasor of
complex phasor of N signal reflections given by
signal reflections given by

N
Ẽ = 
= ∑ En exp( jθ n )
n =1
• Ẽ is a RV representing effects of multipath channel 

small changes in path length Æ large changes in phase

2π ∆x = difference in path length
∆φ =  Δx ∆φ = change in phase
λ

e.g.  fc = 2.4GHz  (λ ≈ 0.125m) Æ ∆φ = 50.26 ∆x

∆x ∆φ
0.10 m 5.03 rad (288o)
5 53 rad (317o)
0 11 m 5.53
0.11
0.12 m 6.03 rad (345o)
2/14/2011 32
Central Limit Theorem: for large N Æ Ẽ becomes Gaussian distribution
Central Limit Theorem: for large N Æ becomes Gaussian distribution

lim ∑ E (cos((θ
N →∞ n =1
n n ) + j sin((θ n ) ) = Zr +jZi

where Zr and Zi are real Gaussian random variables

Consider mean of  n component of Ẽ , denoted as   E

of nth component of Ẽ denoted as Enexp(jθn)

E[Enexp(jθn)]   = E[En]   E[exp(jθn)]  

⎛ 2π 2π
⎞
= E[En]  ⎜ ∫ cos(θ n )dθ n + j ∫ sin (θ n )dθ n ⎟
⎜ ⎟
⎝0 0 ⎠
=  E[En] (0+0) = 0

Ẽ Ẽ is a 0‐mean Gaussian RV
thus E[Ẽ] = 0  and  Ẽ
2/14/2011 33
Consider variance (power) in Ẽ given by mean square value 

⎡ N N
⎤
E[|Ẽ|2]] =  E ⎢∑ E n exp ( jθ n )∑ E m exp (− jθ m )⎥
⎣ n =1 m =1 ⎦
⎡N N ⎤
= E ⎢ ∑∑ E n E m exp ( j (θ n − θ m ))⎥
(equals 0 for all n ≠ m)
(equals  0 for all n ≠
⎣ n =1 m =1 ⎦

⎡ N 2⎤
= E ⎢ ∑ E n ⎥ = P0
⎣ n =1 ⎦
• θn‐ θm is the difference of 2 random phases = random phase
• by symmetry Æ power is equally distributed between real & imaginary
parts of Ẽ
parts of Ẽ
• P0 = average receive power

2/14/2011 34
Rayleigh PDF given by

⎧ r ⎛ r2 ⎞ 0 ≤ r ≤ ∞
p⎜⎜ − 2 ⎟⎟
⎪ 2 exp
p(r) = 
⎨σ ⎝ 2σ ⎠
⎪0 r < 0 
⎩

σ = rms value of received voltage before envelope detection
σ 2 = time average power of received signal before envelope detection

Received Signal Envelope Voltage 

p(r)

[ σ] =  0.6065
p(r)) = Pr[r = 
p(
σ

2/14/2011
0    σ 2σ 3σ 4σ 5σ r 35
Rayleigh CDF
• probability that received signal’s envelope doesn’t exceed specific value, R

R
⎛ R2 ⎞
P(R) = Pr[r ≤ R] = 
∫0 p( r )dr = 1 − exp⎜⎜⎝ − 2σ 2 ⎟⎟⎠

Mean Value of Rayleigh Distributed Signal

Mean Value of Rayleigh Distributed Signal 

∞
π
rmean = E[r] 
E[r] = 
∫ rp(r )dr = σ
0
2
= 1.2533σ

Variance of Rayleigh Distribution represents ac power in signal envelope  
∞
π
σ2r = E[r2] – E2[r] =
∫r p(r )dr − σ
2 2

0
2
⎛ π⎞
= σ 2⎜ 2 − ⎟ = 0.4292σ
2
2/14/2011 ⎝ 2⎠ 36
Envelope’s rms Value is the square root of mean square or 
∞

∫ p( r )dr
r d
2
σr =  E[ r 2 ] =
0

= = 1 414σ
2σ 2 = 2σ = 1.414

σ = rms voltage of original signal prior to envelope detection

M di V l of r
Median Value f is found by
i f db

rmedian
1
d =
∫ p(r )dr
0 2

rmedian = 1.177σ

2/14/2011 37
Conceptually, each location corresponds to different set of {φn}
• deep fades >20dB (R < 0.1Rrms) occur only about 1% 
• wide variation in received signal strength due to local reflections
• if received signal were measured from a number of stationary locations
Æ power measurements would show Rayleigh distribution
t ld h R l i h di t ib ti

Rayleigh Distribution Graph
1

0.1
R)
Pr (r < R

0.01

10‐3

10‐4
‐40   ‐30   ‐20   ‐10     0     10
amplitude 20log10(R/Rrms)
2/14/2011 38
Rayleigh Fading Signal
median differ by only 0.07635σ (0.55dB)
• mean & median differ by only 0.07635

• in practice ‐ median is often used

‐ fading data usually consists of field measurements & a particular 
distribution cannot be assumed 
‐ use of median allows easy comparison of different fading 
distributions with widely varying means

2/14/2011 39
variations in Rayleigh PDF with change in 
y g g Variations in Rayleigh CDF with 
Variations in Rayleigh CDF with
standard deviation change in standard deviation
 Probability density function
Probability density function
• In probability theory, a probability density function (pdf), or
density of a continuous random variable is a function that
describes the relative likelihood for this random variable to
occur at a given point.
• The probability for the random variable to fall within a
particular region is given by the integral of this variable’s
density over the region.
• The probability density function is nonnegative everywhere,
and its integral over the entire space is equal to one.
 Cumulative distribution function
Cumulative distribution function
• In probability theory and statistics, the
cumulative distribution function ((CDF),
) or jjust
distribution function, describes the probability
that a real
real‐valued
valued random variable X with a
given probability distribution will be found at
a value less than or equal to x.
x
• The max spread fd is determined by the speed of the vehicle and is
only experienced by the spectral components arriving on paths
directly in front of (max positive) or directly behind (max negative)
the motion of the receiver.
• the most spectral density s(f) gives proportionate relationship,
which correlate to the rays that are coming directly in front of and
directlyy behind the motion of the antenna.

⎧ ⎫
⎪ ⎪
⎪⎪ 1 ⎪⎪
S( f ) ∝ ⎨ ⎬ f − fc < fd
⎪ ⎪
2
⎛ f − fc ⎞
⎪ df 1 − ⎜⎜ ⎟⎟ ⎪
⎩⎪ ⎝ fd ⎠ ⎪⎭

• Where, fc = center frequency, fd = max spread, f = actual received

frequency / transmitted frequency
• we assumed d a unity
it local
l l mean power.
• In theory, these edges are infinitely high, but for wireless
design verification purposes, the cutoff is typically 6 dB
between power at fc and power at fc ± fd
 Multiple Rayleigh Fading Signals
Multiple Rayleigh‐Fading Signals
• In wireless system,
y , typically
yp y interference from multiplep
transmitters is experienced. Each signal may experience
multipath fading and shadowing.
• Cumulating multiple Rayleigh‐fading signals requires
investigation of the nature of the signals contributing to the
i t f
interference.
• Consider signal behavior during an observation interval of
duration T which is short compared to the rate of channel
fading.
• However,
However modulation can affect amplitudes and phases
during T.
 Two extreme cases are distinguished
Two extreme cases are distinguished
1. Coherent (or phasor) cumulation:
– This occurs if, during T , the phase fluctuations caused by
the modulating signals are sufficiently small, and the
carrier frequencies of the signals are exactly equal.
– The joint signal behaves as a Rayleigh phasor, with
Gaussian in‐phase and quadrature phase components.
– Th
There i
instantaneous power isi exponentially
i ll distributed.
di ib d
– The local‐mean power is equal to the sum of the local‐
mean powers of the individual signals as shown in Fig.Fig
Phasor diagram of a set of scattered waves resulting in a Rayleigh‐fading envelope
• Consider the transmitted signal has the form
s(t)= cos (ωct+ψ)
• Consideringg the Doppler
pp effect on the received signal
g s’(t)
( ) which
can be expressed as

N
S ' (t ) = ∑c n
cos( 2π f t + ψ + φ + 2π Δ f t)
c n n
n =1

Angle of arrival
l f l
amplitude 

v
The Doppler shift of each wave is
• v is the speed of the receiving antenna.
i h d f h i i
Δ
n
= cos
λ
f α n

• αn relative to the direction of the motion of the antenna. 
• An inphase‐quadrature representation of the form 
s’(t) = I(t) cos ωct – Q(t) cos ωct
the in‐phase component including Doppler effect can be 
written as
N 2π vff c t
I (t ) = ∑ cn cos( cos α n + ψ + φ n)
n =1 c
and
d quadrature
d phase
h component

N 2π vf c t
Q (t ) = ∑ cn sin( cos α n + ψ + φ n)
n =1 c
• The maximum Doppler shift fd occurs for a wave coming from
the opposite direction as the direction the antenna is moving
to It has a frequency shift with fc is the carrier frequency and
to.
c the velocity of light.
v
fd =
c
f c
• Coherent addition can occur only if phase modulation with a
very small deviation is applied, or if observation interval is
taken short with respect to the rate of modulation.
modulation
• Let's consider a stationary user, thus v = 0.
• An inphase and quadrature representation reduces to

N
I (t ) = ∑c
n =1
n co s(ψ + φ n )
N
Q (t ) = ∑c
n =1
n sin (ψ + φ n )
• Conclusion:
– both the inphase and quadrature component, I(t) and Q(t) can be
interpreted as the sum of many (independent) small contributions.
contributions
– Each contribution is due to a particular reflection, with its own
amplitude cn and phase ϕn.
– For sufficiently large N, the Central Limit Theorem says that the
inphase and quadrature components tend to a Gaussian distribution
of their amplitude.
p
– I(t) and Q(t) appear to be independent and identically distributed.
 Central limit theorem
Central limit theorem
• The Gaussian distribution plays a central role in statistics
because of a mathematical relationship known as the Central
Limit Theorem.
Theorem
• The central limit theorem says that if your samples are large
enough the distribution of means will follow a Gaussian
enough,
distribution even if the population is not Gaussian.
 Gaussian distribution
Gaussian distribution
• In a Statistical tests analyze a particular set of data to make
more general conclusions. There are several approaches to
doing this,
this but the most common is based on assuming that
data in the population have a certain distribution. The
distribution used most commonlyy byy far is the bell‐shaped p
Gaussian distribution, also called the Normal distribution.
• The Gaussian distribution extends from negative infinity to
positive infinity.
2 Incoherent (or power) addition:
2. Incoherent (or power) addition:
– If the phases of each of the individual signals substantially
fluctuate due to mutually independent modulation, the
signals add incoherently.
incoherently
– The interference power experienced during the
observation interval is the power sum of the individual
signals.
– With incoherent cumulation,, the jjoint interference signal
g
behaves as a band‐limited Gaussian noise source if the
number of components is sufficiently large.
– Moreover, any fade of one of the signals is likely to be
hidden by the other interfering signals.
• Multiple Incoherent Rayleigh‐Fading Signals with
Equal Mean‐Power
– If the interference is caused by the power sum of n
Rayleigh‐fading signals with identical local‐mean power,
th PDF off the
the th joint
j i t interference
i t f power is
i the
th n‐thth
convolution of the exponential distribution of the power of
an individual interfering signal.
signal
– The PDF of the joint interference power caused by
interferingg signals
g with different local‐mean p
powers can be
approximated by a gamma distribution.
PDF of Received Signal Amplitude
PDF of Received Signal Amplitude
• In order to obtain the PDF of the signal g amplitude
p of a
Rayleigh fading signal, we observe the random processes of
the inphase and quadrature components, I(t) and Q(t) ,
respectively, at one particular instant t0.
• For sufficiently large N, the Central Limit Theorem says that
I(t0) and Q(t0) at a particular
partic lar instant t0 are Gaussian
Ga ssian random
variables, each with variance σ2.
• In a flat fading channel,
channel the received signal s’(t) is given by the
transmitted signal s(t), multiplied by a time varying
( ) and a noise contribution w(t)
attenuation α(t) ()
s’(t)= α(t)s(t) + w(t)
• The time function of attenuation variation is difficult to
represent mathematically as it is the matter of probability and
depends upon channel environment.
• For such cases it is better to represent the probability
distribution function for α which is usually follows a Rayleigh
di t ib ti
distribution.
 / 2 )* exp[‐α2/ 2σ
pdf(α)= (α/σ / 2] 0 < α < ∞
• Where, σ is the variance of the underlying Gaussian process.
• The condition for this to be valid is that there are many
statistically independent scatterers and no single scatterer
makes a dominant contribution.
contribution

• If the set of reflected waves are dominated by one strong

component, Rician fading is a more appropriate model.
 Summary for Rayleigh fading distribution:
Summary for Rayleigh fading distribution:

• The probability that the envelope of the received signal does not
exceed a specified (threshold) value J is given by the corresponding
CDF
2
J J
P(J)= ∫0 p(α )dα = 1 − exp(− 2 )
2σ
π
• The mean value of Rayleigh distribution is given by E[α] = σ = 1.2533σ
2

• The variance of Rayleigh distribution is given by σα2=E[α2]‐ E2[α] =

0.4292σ2
• The RMS value of the envelope is i.e. where σ is the standard
i e , where,
deviation and median value of α is 1.117σ.
Rician fading model
• Th
The Rician
Ri i fading
f di model d l is
i similar
i il to that
h for
f Rayleigh
R l i h fading,
f di
except that in Rician fading a strong dominant component is
present ((line‐of‐sight
p g wave). )
• Advance Rician models also consider that
– The dominant wave can be a phasor sum of two or more dominant
signals,
i l e.g. theh line‐of‐sight,
li f i h plus
l a ground d reflection.
fl i Thi combined
This bi d
signal is then mostly treated as a deterministic (fully predictable)
process, and
– The dominant wave can also be subject to shadow attenuation. This is
a popular assumption in the modeling of satellite channels.
• Besides the dominant component,
p , the mobile antenna
receives a large number of reflected and scattered waves as
shown in following Fig.
Rician PDF for σ =1  Rician CDF for σ =0.25
• Rayleigh fading is considered a worst case scenario.
• In rural environments, where the multipath profile includes a
few reflected ppaths combined with a strongg line of sight
g p path,
the spectral power follows a Rician distribution.
• The angle of arrival of the direct ray, as well as the ratio of the
power between the direct ray and the multipath rays,
determine effect of the energy from the direct path on the
normal Rayleigh model.
model
• Observing this effect in the frequency domain, what we see is
a spike in power corresponding to the frequency shift
attributed to the direct ray.
• Ex
– iin the
h GSM specifications,
ifi i the
h anglel off arrival
i l off the
h direct
di path
h is
i set
to 45 degrees. The cosine of the angle to see how much statistical
weight we give to that particular path.
– In this case it is 0.7.
– Thus we have a spike in received signal power at 0.7 times the max
frequency shift.
shift
• The Rician model also allows for setting the ratio of power
between the direct path and the Rayleigh path.
• This ratio is called the K factor.
• The Rician K K‐factor
factor is defined as the ratio of signal power in
dominant component over the (local‐mean) scattered power.
• When K = 0,, the channel is Rayleigh
y g
• when K = ∞, the channel is AWGN
• The variation of Doppler spectrum in terms of spectral density
s(f) is given by
Rician Fading Distributions
if signal has dominant stationary component (LOS path) Æ small scale 
if signal has dominant stationary component (LOS path)  Æ small scale
fading envelope has Rician Distribution
• random MPCs arrive at different angles & are superimposed on dominant 
signal
• envelope detector output yields Æ DC component +  random  MPCs

Rician distribution is result of dominant signal arriving with weaker MPCs 
• as dominant component fades Æ Rician distribution degenerates to 
Rayleigh
y g distribution
• composite signal resembles noise signal with envelope that is Rayleigh 
• same as in the case of sine wave detection in thermal noise 

2/14/2011 68
Rician PDF
⎧ r ⎛ r 2 + A2 ⎞ ⎛ Ar ⎞
⎪ 2 exp⎜⎜ − ⎟I
2 ⎟ 0 ⎜ 2 ⎟
A ≥ 0 , r ≥ 0 
p(r) =  ⎨σ ⎝ 2σ ⎠ ⎝ σ ⎠
⎪0 r < 0
⎩
A = peak amplitude of dominant signal
Ι0(•) = modified Bessel function of the 1st kind & zero order
Rician Distribution often described in terms of Rician Factor K
• K = ratio of deterministic signal power to multipath variance

K = 
A2
2σ 2
⎛ A2 ⎞
K (dB) =  10 log⎜⎜ 2 ⎟⎟dB
⎝ 2σ ⎠
p
• as dominate path attenuates Æ A ggrows small 
• Rician distribution degenerates to Rayleigh

2/14/2011 69
• the
th joint
j i t PDF off amplitude
lit d and
d phase
h ψ is
i
pdfr, ψ = (r/2πσ2 )* exp[(‐r2+A2‐2rAcos(ψ)/ 2σ2]
• An
A alternative
l i amplitude
li d distribution
di ib i which
hi h has
h gained
i d
popularity especially for the evaluation of measurements, is
Nakagami m m‐distribution
distribution.
 Rician Fading

Pr (r < R)
1
K = 0 (Rayleigh)
0.1 K=5dB
K=10dB
0.01 K=13dB

10‐33

10‐4
‐40
40   ‐30
30   ‐20
20   ‐10
10     0     10
0 10
amplitude 20log10(R/Rrms)

probability of deep fades reduced as K
b bili fd f d d d grows

2/14/2011 71
 Multiple Rician Fading Signals
Multiple Rician Fading Signals
• A sinusoidal signal s(t) = cos wct received over a Rician
multipath channel can be expressed as
’(t) = A cos wct + ΣNn=1 α(n)
s’(t) ( ) cos (w
( ct + φn)

Where
• A is the amplitude of the line‐of‐sight component
• α(n) is the amplitude of the n‐th
n th reflected wave
• φn is the phase of the n‐th reflected wave
• n = 1 .. N identify
id tif the
th reflected,
fl t d scattered
tt d waves.
Examples of Rician fading are found in
• Microcellular channels
• Vehicle to Vehicle communication
• Indoor propagation
• Satellite channels
Signal Outages in Fading Channels
Signal Outages in Fading Channels
• The mobile Rayleigh or Rician radio channel is characterized
by rapidly changing channel characteristics. As typically a
certain minimum (threshold) signal level is needed for
acceptable communication performance, the received signal
will experience
p periods of sufficient signal
p g strength
g or "non‐
fade intervals"
– insufficient signal strength or "fades"
– During fades the user experiences a signal outage.
• For a large fade margin F >>1, the probability that the
instantaneous power drops below the noise threshold z tends
to (1 + K ) exp(− K )
outage probability = F
Where, F = the local mean power ‐ the threshold z.
• Fade margin is the ratio of the average received power over
some threshold power, needed for reliable communication.
• Outage event: set of channel realizations that cannot support
reliable transmission at a rate R . Equivalently, is the set of
channel realizations with an associated capacity less than a
transmit rate R.
 PDF of Signal Amplitude
PDF of Signal Amplitude
• The derivation is similar to the derivation for Rayleigh fading.
• In order to obtain the probability density of the signal
lit d α we observe
amplitude b th random
the d processes I(t) and
d Q(t) att
one particular instant t0. If the number of scattered waves is
sufficiently large,
large and are independent and identically
distributed, the central limit theorem says that I(t0) and Q(t0)
are Gaussian distributed, but, due to the deterministic
dominant term, no longer zero mean.
Nakagami Fading Model 
• Alternative amplitude distribution, Nakagami m‐distribution

m
2 ⎛ m ⎞ 2m−1 ⎡ m 2 ⎤
f (α ) = ⎜⎜ ⎟⎟ α exp⎢− α ⎥
Γ(m) ⎝ Ω ⎠ ⎣ Ω ⎦
Ω2
α ≥, m ≥ 1/ 2, m = 2
• Γ = Gamma function (α − Ω)2
• m =  parameter used to adjust the PDF to data samples
• For m = 1 Nakagami reduces to Rayleigh dist
For m = 1, Nakagami reduces to Rayleigh dist. 
• The parameter Ω is the mean square value  Ω = E[α2].  
• Nakagami can be converted to Rice by changing m. 

• http://www.vosesoftware.com/vosesoftware/ModelRiskHelp.htm
• Nakagami and Rice distributions are quite similar and each
can be approximately converted to the other for m ≥ 1:

(K +1) m −m
2 2

m= K=
m− m −m
2
(2 K + 1)
• The distribution of the signal power can be used to find
probabilities on signal outages.
– If the envelope is Nakagami distributed,
distributed the corresponding
instantaneous power is gamma distributed.
– The parameter m is called the 'shape factor' of the Nakagami or the
gamma distribution.
– In the special case m = 1, Rayleigh fading is recovered, with an
exponentially distributed instantaneous power
– For m > 1, the fluctuations of the signal strength reduce compared to
Rayleigh fading.
• In current wireless communication, the main role of the Nakagami
model
d l can be
b summarized
i d as follows
f ll or we can say When
Wh does
d
Nakagami Fading occur?
1 It describes the amplitude of received signal after maximum ratio
1.
diversity combining .
– After k‐branch maximum ratio combining (MRC) with Rayleigh‐fading signals,
the
h resulting
l i signal
i l isi Nakagami
N k i with
i h m = k.
k MRC combining
bi i off m‐Nakagami
N k i
fading signals in k branches gives a Nakagami signal with shape factor mk.

2. The sum of multiple independent and identically distributed (iid)

Rayleigh‐fading signals has Nakagami distributed signal amplitude.
– This is particularly relevant to model interference from multiple sources in a
cellular system.
3. The Nakagami distribution matches some empirical data better
than
h other
h models
d l
4. Nakagami fading occurs for multipath scattering with relatively large
delay‐time
delay time spreads,
spreads with different clusters of reflected waves.
waves
– Within any one cluster, the phases of individual reflected waves are
random, but the delay times are approximately equal for all waves. As a
result
lt the
th envelope
l off eachh cumulated
l t d cluster
l t signal
i l is i Rayleigh
R l i h
distributed.
– The average time delay is assumed to differ significantly between clusters.
If the delay times also significantly exceed the bit time of a digital link, the
different clusters produce serious Intersymbol interference, so the
multipath self‐interference then approximates the case of co‐channel
interference by multiple incoherent Rayleigh‐fading signals.
5 Th
5. The Rician
Ri i and d the
h Nakagami
N k i model
d l behave
b h approximately
i l
equivalently near their mean value.

– This observation has been used in many recent papers to advocate the
Nakagami model as an approximation for situations where a Rician
model would be more appropriate.
– While this may be accurate for the main body of the probability
density, it becomes highly inaccurate for the tails. As bit errors or
outages mainly occur during deep fades,
fades these performance measures
are mainly determined by the tail of the probability density function
(for probability to receive a low power).

• http://www.vosesoftware.com/vosesoftware/ModelRiskHelp.
htm
 Assignment 
Assignment
• Explain following terms
– Energy and Power Signal
– Signal to Noise Ratio
Signal to Noise Ratio
– Channel Bandwidth
– Signal Bandwidth 
g
– Bit rate 
– symbol rate
• Explain Different types of noises in wireless systems
• Prepare a Comparison table of  Rayleigh, Rician and 
Nakagami Fading models
 summary
• Modeling Concept
• Results: Dynamic shadowing is correlated 
y g
between links!
 Small Scale Fading
Small‐Scale Fading
• “Small‐scale (fast) fading” …
– quick amplitude variations of the received signal due to 
constructive/destructive interference of multipath
constructive/destructive interference of multipath
• Three common models
– Rayleigh fading
y g g
– Rician fading
– Nakagami fading 
Optimal Signal Detection in AWGN 
Optimal Signal Detection in AWGN
LTI Channel Using Matched Filter
g
 AWGN
• In reality, transmission is always corrupted by noise whatever
may be
b theh type off channel
h l assumed.
d
• The usual mathematical model of the radio channel is the
Additi White
Additive Whit Gaussian
G i Noise
N i channel.
h l
– It is a very good model for the physical reality as long as the thermal
noise at the receiver is the onlyy source of disturbance.
• The AWGN channel model can be characterized as follows:
– The noise is additive. The noise w(t) is an additive random disturbance
of the useful signal s(t), that is, the receive signal is given by
r(t) = s(t) + w(t)
• The noise is white, i.e., the power spectral density is flat so the
autocorrelation of the noise in time domain is zero for anyy non‐zero time
offset.
• The noise samples have a Gaussian distribution. Gaussian probability
density function with variance σ2 is given by

1 − ( x − m ) / 2σ 2
2
p( x) = e
2π σ 2
 Matched filter
Matched filter
• Aim: it correlates a known signal,
g , or template,
p , with an
unknown signal to detect the presence of the template
in the unknown signal.
g
• It is equivalent to convolving the unknown signal with
a conjugated time
time‐reversed
reversed version of the template.
• The matched filter is the optimal linear filter for
maximizing the signal to noise ratio (SNR) in the
presence of additive stochastic noise.
Matched filter
Matched filter
• It can be shown to be the optimal detector if
– The channel produces AWGN,
– The channel is LTI, and
– Exact time reference is available, the signal amplitude as a
f ti off time
function ti i precisely
is i l known.
k
• If the noise is non‐white,
– the
h matchedh d filter
fil can still
ill be
b applied.
li d
– In this case, one can pre‐filter the incoming signal, to make the
noise component white.
white This is called a whitening filter.
filter
• If the channel is dispersive,
– the matched filter concept can still be used,
used
– but one must multiply the incoming signal with a locally
ggenerated copy
py of the expected
p waveform after transmission
over the channel.
• That is, the receiver must estimate the channel impulse
response and apply this to the reference signal waveform. The
incoming signal is correlated with a reference waveform,
waveform
which is dispersed in the same manner as the channel
p
disperses the radio signal.
g
• A complication is that such dispersion causes inter symbol
interference. Theoretically, it is no longer optimum to detect
the received symbols one by one.