ECE353

Wireless Communication Networks

Lecture 5
Wireless Propagation Channels (Statistical models)
Course Coordinator: Prof. Bassant Abdelhamid
Dr. Moustafa Mohamed
[Link]@[Link]
Spring 2026
 From previous lecture
❑ Deterministic models are studied to predict average received power
➢ Depend on distance between TX & RX
➢ Models as free space, breakpoint, two ray ground reflection, …
❑ Strong variations occur on the received power
❑ Constructive & destructive interference of MPCs cause small scale
fading
➢ Variations in received signal power occur over short distance (one wavelength)
➢ It can be described statistically by local mean of power and variations around it
❑ Shadowing by large objects causes large scale fading
➢ Variations occur on a large scale (Hundred wavelengths)
➢ It can be described statistically by mean of power and statistics of variations
around it
2
 Statistical modelling of channel variations
❑ Calculate the probability that certain channel parameters attain certain
values
❑ The channel parameter we will focus on is the channel gain
➢ It determines received power or field strength
➢ Channel gain is directly proportional to receive power
PRX = channel gain × PTX
PRX|dBm = channel gaindB + PTX|dBm
➢ Channel gain is less than unity (less than zero in logarithmic scale)
➢ If PTX|dBm = 0 ➔ PRX|dBm = channel gaindB

3
 Statistical modelling of channel variations
❑ We will begin with two simple models to gain more intuition about
small scale fading
➢ Time-invariant two-path model
➢ Time variant two-path model
❑ These two models will help us to mathematically understand the effect
of small scale fading
❑ They assume the transmitted signal as unmodulated sinusoidal signal
➢ Narrow band system

❑ Time variant two path model will introduce us to new concept (Doppler
shift)

4
 Time-invariant two-path model
❑ Simplest possible scenario which assumes
➢ Sinusoidal waveform is transmitted between TX and RX separated by distance d
2𝜋
𝐄 𝐭, 𝐝 = 𝐄𝐨 𝐜𝐨𝐬 𝟐𝛑𝐟𝐜 𝐭 − 𝐤 𝐨 𝐝 ko =
𝜆
➢ The instantaneous value of field at t = 0
E = Re{Eo e−jko d }
➢ Now assume two MPCs where the signal is received via two paths
✓ The received superimposed wave at distance r equals
E r = E1 e−jk1r + E2 e−jk2 r
𝑑1 𝑑2
✓ Paths have different runtime 𝜏1 = , 𝜏2 =
𝑐 𝑐

5
 Time-invariant two-path model
❑ Superposition of two waveforms
➢ |E| represents square root of
received power
➢ Constructive & destructive
interference depend on location
➢ If RX at destructive position, the
received signal is the difference
between amplitudes of signals
➢ If RX at constructive position, the
received signal is sum of
amplitudes of both signals
➢ Constructive & destructive area can
be termed mountains & valleys
6
 Time-variant two-path model
❑ Considers variations in the channel with time
➢ Might occurs due to motion of TX or RX or IOs or all of them
➢ This model considers motion of RX only
➢ RX accordingly experience a time varying interference (small scale fading)
➢ Movement of the RX causes a shift in the received frequency which is known as
Doppler shift

7
 Doppler shift
❑ Assume a sinusoidal wave is transmitted in single path
❑ RX is moving away from TX with velocity v in same direction of wave
propagation
E t = Eo cos 2πfc t − k o (do + vt)
v
E t = Eo cos 2π(fc − )t − k o do
λ
v
❑ The frequency of received oscillation decreased by
λ
v
❑ Doppler shift (fd ) = - (It is negative as RX is moving away from TX)
λ

8
 Doppler shift
❑ If RX movement is not aligned in same direction of wave propagation
➢ There is angle (γ) between direction of RX motion and wave propagation
➢ Doppler shift would be predicted by speed of movement in direction of wave
propagation
v
➢ Doppler shift (fd ) = - cos(γ) = - fdmax cos(γ)
λ
➢ Doppler shift represents rate of change of channel
❑ In case of MPCs, each MPC would have different doppler shifts
➢ Assuming two paths, it will look like beats signal

9
 Small scale fading without a dominant component
❑ More generic and practical statistical multipath propagation model
➢ Assume N homogenous plane waves (MPCs)
✓ Created from reflecting/scattering from multiple IOs
✓ No one of them is a dominant component (No LOS exists)

➢ RX moves with velocity v (TX & IOs do not move)

➢ Phases & amplitudes of each MPC are random
➢ Random doppler shift is added for each MPC
❑ Accordingly, the received field strength for unmodulated carrier equals
N

E t = ෍ ai cos[2πfc t − 2πfdmax cos γi t + ∅i ]

i=1

10
 Small scale fading without a dominant component
N

E t = ෍ ai cos[2πfc t − 2πfdmax cos γi t + ∅i ]

i=1

❑ In-phase and quadrature components are equal to

N N

I t = ෍ ai cos[−2πfdmax cos γi t + ∅i ] Q t = ෍ ai sin[−2πfdmax cos γi t + ∅i ]

i=1 i=1

❑ Central limit theorem states that “when superimposing N statistically

independent random variables none of them is dominant the associated
pdf approaches a gaussian random variable with zero mean and
variance 𝜎 2 ”
➢ Accordingly, the pdf of the in-phase and quadrature components are modelled
as gaussian RV with zero mean and variance 𝜎 2
1 −x2
fX x = e2σ2
2πσ 11
 Small scale fading without a dominant component
N N

I t = ෍ ai cos[−2πfdmax cos γi t + ∅i ] Q t = ෍ ai sin[−2πfdmax cos γi t + ∅i ]

i=1 i=1

12
 Small scale fading without a dominant component
❑ We are interested more in amplitude of received signal
❑ Using statistics of real & imaginary parts and Jacobian transformation, it
can be proved that
➢ The phase of received signal has uniform distribution
1
fΨ ψ =
2𝜋
➢ The amplitude of received signal has Rayleigh distribution
−r2
r
fR r = 2 e2σ2 0 <r<∞
σ
❑ This distribution is known as Rayleigh
distribution

13
 Small scale fading without a dominant component
❑ The amplitude of received signal has Rayleigh distribution
−r2
r
fR r = e 2σ2 0 <r<∞
σ2
❑ Rayleigh distribution has the following properties
𝜋
➢ Mean value E r = rҧ = σ
2

➢ Mean square value E r 2 = rഥ2 = 2σ2

➢ Variance = rഥ2 − rҧ 2 = 0.429σ2
➢ Median value = r50 = σ 2ln(2) = 1.18σ
➢ Location of maximum max{fR r } at r = σ
2 −r2
𝑟 r −r2
➢ CDF(r) = ‫׬‬−∞ σ2 e2σ 𝑑𝑟 =1−e 2σ2

14
 −r2
r
❑ fR r = 2 e 2σ2 0 <r<∞
σ
−r2 −r2
𝑟 r
➢ CDF(r) = ‫׬‬−∞ e 2σ2 𝑑𝑟 = 1 − e 2σ2
σ2

15
 Small scale fading without a dominant component
❑ How to calculate Fading margin for Rayleigh distributed fading strength?
➢ The solution is to answer the following question: “Given a minimum receive
power or field strength required for successful communications, how large does
the mean power have to be to ensure that communication is successful in x% of
all situations?”
➢ In other words, how large does the fading margin have to be?
−r2
min
➢ Poutage = Prob(r< rmin ) = CDF(rmin ) = 1 − e 2σ2

➢ Recall that 2σ2 is the average received signal power

✓ Fading Margin in dB (MRayleigh|dB ) = Pavg|dBm − PRxmin |dBm
Pavg
✓ Fading Margin (MRayleigh ) =
PRxmi𝑛
−1
MRayleigh
✓ Poutage = CDF(rmin ) = 1 − e
16
Small scale fading without a dominant component
−r2
min
➢ Poutage = Prob(r< rmin ) = CDF(rmin ) = 1 − e 2σ2

17
 Small scale fading without a dominant component
❑ Example 5.1 (Reference 1 by Andreas F. Molisch)
❑ For a signal with Rayleigh-distributed amplitude, what is the probability
that the received signal power is at least 20, 6, 3 dB below the mean
power.

18
 Small scale fading without a dominant component
❑ Rayleigh distribution is widely used in wireless communications
➢ It is an excellent approximation in a large number of practical scenarios, it is also
confirmed by a multitude of measurements
➢ It describes a worst-case scenario in the sense that there is no dominant signal
component, and thus, there is a large number of fading dips
➢ It depends only on a single parameter (the mean received power)
➢ Mathematical convenience: computations of error probabilities and other
parameters can often be done in closed form when the field strength
distribution is Rayleigh
➢ It is noteworthy that there are scenarios where it is not valid. These can occur,
e.g., in Line Of Sight (LOS) scenarios, and in (ultra) wideband scenarios

19
 Small scale fading with a dominant component
❑ Fading statistics change when a dominant MPC (LoS component) exists
❑ The probability of deep fades is much smaller compared to the Rayleigh-
fading case (without dominant component)
❑ The pdf of the amplitude is computed in a way similar to the derivation
of the Rayleigh distribution
✓ A LoS component with amplitude "A" & zero phase is assumed
✓ Real part has gaussian distribution with mean "A“
✓ Imaginary part has a gaussian distribution with zero mean
❑ Using statistics of real & imaginary parts and Jacobian transformation, it
can be proved that the joint PDF equals
r −(r2 + A2 −2rA cos ψ)
fR,Ψ r, ψ = e 2σ2
2πσ2
20
 Small scale fading with a dominant component
−(r 2 + A2 −2rA cos ψ)
r
fR,Ψ r, ψ = e 2σ2
2πσ 2

❑ PDF of amplitude is calculated by integrating fR,Ψ r, ψ over the phase

➢ It is given by Rice distribution
−(r2 + A2 )
r 𝑟𝐴
fR r = e 2σ2 . Io ( 2 ) 0 <r<∞
σ2 σ

➢ Io is the modified Bessel function of zero order

➢ Mean square value E r 2 = 2σ2 + A2

❑ This distribution is known as Rice distribution

21
 Small scale fading with a dominant component
❑ The ratio of the power in the LOS component to the power in the
A2
diffuse component is called the Rice factor K r =
2σ2
➢ If K r = 0 ➔ Rice distribution becomes Rayleigh distribution
➢ If K r is large ➔ Rice distribution is approximately Gaussian distribution
❑ The stronger the LOS component (High K r ), the rarer the occurrence of deep fading

22
 Large scale fading (Shadowing)
❑ Experimental measurements have shown that the field strength due to
large scale fading, plotted on a logarithmic scale shows a Gaussian
distribution
➢ This distribution has a mean μF|dB & standard deviation σF|dB (expressed in dB)
➢ This distribution is known as lognormal distribution
❑ Poutage = Prob(r< rmin ) = CDF(rmin )
MLarge scale|dB
❑ Poutage = Q
σF|dB

μ𝐅|𝐝𝐁

23
 Large scale fading (Shadowing)
❑ Poutage = Prob(r< rmin ) = CDF(rmin )
MLarge scale|dB
❑ Poutage = Q
σF|dB

μ𝐅|𝐝𝐁

24
 Large & small scale fading
❑ Both large-scale and small-scale fading might occur
➢ Fading margin must consider both effects
❑ One option is to add the fading margin for the Rayleigh distribution and
the fading margin for a lognormal distribution
➢ Simple method
➢ Overestimates the required fading margin

25
 Small scale fading without a dominant component
❑ Example 5.5 (Reference 1 by Andreas F. Molisch)
❑ Consider a channel with σF = 6 dB and μF = 0 dB. Compute the fading
margin for Rayleigh & shadowing effects so that the outage probability
is smaller than 5%.

26
 ∞
1 t2
Q x = න e− 2 dt
2π
x

27
Any Questions ?