Introduction to

Fading Channels, part 2

Dr. Essam Sourour
Alexandria University, Faculty of
Engineering, Dept. Of Electrical
Engineering
1

Small Scale Fading

Local reflections cause multipath
Each path has a random gain, with random
magnitude and random phase
j n
Each gain is represented in baseband as n e
Receiver, and/or reflectors, may be moving

Building 2

Small Scale Rayleigh Fading

Assume a group of paths with small relative delay
Net effect is one path with random gain and phase

j n

R e

According to Central Limit Theorem, the net gain

Rej is complex Gaussian with zero mean
The envelope R is Rayleigh distributed and the
phase is uniform [0, 2]
Building 2

Rayleigh Fading
n e j n x n jy n
x n n cos n

and

y n n sin n

The net gain is the sum of all closely delayed paths:

R e j n e j n x n jy n g g r e al jg imag
n

Each of greal and gimag is the sum of many independent

random variables
Hence greal and gimag are independent and Gaussian
with zero mean and variance 2 each
Fading gain g = greal + j gimag is complex Gaussian with
zero mean and variance 22 (sum of two variances)
4

Rayleigh Distribution
From probability theory we know:
2
2
R g g real
g imag
R is Rayleigh Distributed

g tan 1 g real g imag is Uniformly Distributed

If s t is transmitted, then r t Rs t e j will be received
Received amplitude follows Rayleigh distribution
2
2
pr

r exp r 2

, r 0

Received power follows Exponential distribution

p y

exp y 2 2
2 2

, y 0

Received phase follows Uniform distribution

p

1
, 0 2
2
5

Amplitude, Rayleigh Distribution

pr

r exp r 2 2 2

, r 0

Power, Exponential Distribution

p y

exp y 2 2
2

, y 0

Fading of 16 QAM signal

No fading

Faded signals with random amplitude and phase

Signal has a higher probability of being week

For example, to receive the 16-QAM signal
we must estimate and compensate for the
amplitude and phase
8

Effect of Mobility

Fading gain changes with time

g(t)=greal(t) + j gimag(t)

Fading change rate depends on the

v v
f

maximum Doppler frequency

c
Coherence time << 1/fD

Example: fc=1GHz, v=100 km/h gives:

fD = 92.6 Hz , Coherence time << 10.8 ms

Fading Example for R

10

Statistical Properties 1
Complex fading gain g(t)

g t g r e al t j g imag t

The two parts greal(t) and gimag(t) are zero

mean
E g t E g
t 0
real

imag

The two parts greal(t) and gimag(t) are

statistically independent
E g real t g imag t t 0, for any t

11

Statistical Properties 2
Fading gain is correlated over time
Usually Jakes model is used in mobile comm.
Autocorrelation function given by
A t E g t g t t 2 J 2 f t
*

Jo() is the Bessel function of order zero

Ag(t) indicates how much the gain is correlated
with itself after delay t
Power spectral density of fading is the FT of the
autocorrelation function
Sg f

2 2

f D

1 f f D

f fD
12

Statistical Properties 3
Usually the fading gain is normalized to unity
power, i.e, 2=1/2
A g t J 0 2 f D t

fD t

Sg f

1
f D

1
1 f f D

f/fD

f fD

13

Rician Fading Channel

If the channel also includes a LOS component we
get Rician fading
Fading gain is now
g

g rician

6 44Rayleigh
7 4 48
g real j g imag S
,
g real
j g imag

greal is Gaussian with mean S and variance 2

The envelope R is Rician distributed (see Proakis
chapter 2)
2
2
R Rician

g g
,
real

imag

Rician Fading Channel

The channel amplitude R is Rician
r 2 S 2
r
rS
I 0 2
p r 2 exp
2

, r 0

I0 = modified Bessel function of order zero

Now, when s(t) is transmitted
r t g Rayleigh s t S s t

Power ratio K=LOS/faded=S2/(22)

If K=0 we are back to Rayleigh fading
As K increases, more power to LOS
15

Rician PDF

K=1, 2, 3
16

Large Delays Effect

Channel may consists of groups of delays (echoes)
Each group is composed of many closely delayed
paths
Maximum Delay Spread: Delay between first and last
Typically few microseconds outdoor and less than
hundred of nanoseconds indoor
Channel with large delay spread is an FIR filter:
L 1

h t g l t t l
l 0

17

Power Delay Profile

Power of the multipath decay as delay
increases according to power delay profile
Each path gl has a variance v l 2 l2
Example, exponential profile v l v o exp l L
Example, uniform profile v l v o , for l L10,..., L 1
v l 1

Typically, fading is normalized

L 1
L 1
l 0
Mean delay spread: l v l v l
l 0
l 0
L 1
L 1
RMS delay spread
2
RMS l v l v l
l 0

l 0

18

Time and Delay Picture

Channel may have many
resolvable paths
Each path at a certain delay
Each path changes with
time, t, and has its delay,
L 1

h t , g l t t l
l 0

Autocorrelation function:

A t , E h * t , h t t ,

Scattering function: twice Fourier

Transform of the Autocorrelation function,
over t and
19

Simulating Classical Fading Model

Jakes model

20

Simulating Classical Fading Model

Assume a mobile station in the middle of
4N reflectors
Reflections with equal amplitude but
different Doppler
Doppler from path with incident angle n is
fn=fM cos(n) , fM is the maximum Doppler
Reflectors have different propagation delay
around the circle
21

Classical Fading Model

After some mathematical manipulations, the gain
of the path hk(t):
2
hk t
N

n
W
k cos n j sin n cos nt n ,k
n 1

N 16 (a good choice)
f c carrier frequency in Hz

n M cos n
n
N
random angle in (0, 2 )

v = vehicle speed

c = speed of light = 3 * 108 m/s

n ,k

M 2 v c f c
n

Wkn Walch code chip n in code # k

n 0.5

2 N
22

Classical Fading Model

With L resolvable multipath, the channel model
L 1
L 1
is given by
h t v l hl t t l
l 0

v
l 0

The gains vl select the desire delay profile

They are normalize the total channel power to 1

23

Walch Codes of length 16

n

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

-1

1
24

Fading References
1. Classical Model: W. C. Jakes, editor,
Microwave Mobile Communications,
New York, Wiley 1974
2. Modifications: P. Dent, G. E. Bottomley,
and T. Croft, Jakes fading model
revisited, Electronic letters, vol. 29, pp.
1162-1163, June 1993
3. Good reference: Chapter on Fading
channels in Digital Communications by
Bernard Sklar
25

Effects on Signal
Small Scale Fading

Slow

Fast

Small Scale Fading

Selective

Non-selective

Fast: Channel changes

within symbol. Tc<Ts

Selective: Delay Spread >

symbol time Ts

Slow: Channel constant

during more than symbol
time. Tc>Ts

Non-Selective: Delay
Spread < symbol time Ts

Speed and Selectivity are independent issues

26

Definitions
Coherence time = 1/max doppler = 1/f D
Coherence bandwidth = 1/max delay
spread
Slow fading: Symbol time < coherence time
Non-selective fading: Signal bandwidth <
coherence bandwidth
Fast fading and selective fading are the
opposite
27

Effects on Signal, cont.

Fast Fading:
Due to high speed
High distortion to the received
signal

Slow Fading:

s(t)

Terminal may fall in a fading

g(t)
null for long time
Worse performance
g(t)

time
Fast Fading

time
Slow Fading
time

28

Effects on Signal, cont.

Selective Fading:
Due to high Delay Spread w.r.t
symbol duration
Channel is random FIR filter

Signal spectrum

Non-Selective Fading:
Delay Spread << symbol
duration
Channel is one tap

Selective

Channel gain

Channel gain

frequency

frequency

Non-Selective

frequency

29

Fading Counter Measures

Receiver Antenna Diversity
Transmitter Antenna Diversity
Transmitter and Receiver Antenna
Diversity (MIMO Systems)
Rake Receiver
Channel Equalization
Channel Coding

30

Receiver Antenna Diversity

Receiver may have two or more
antennas
Two main types:
Antenna Selection:
Select stronger antenna signal.
Best for slow, non-selective fading

Antenna Combining:
Optimally combine signal of antennas (MRC)
More complexity & better performance
31

Maximal Ratio Combining

so

ho*

ho

|h0|2 so

so ho

h1

h1*

so h1

|h0|2 so + |h1 |2 so

|h1|2 so
Maximal Ratio Combining
Receiver Diversity

32

Transmit Diversity
Two antennas are used in Tx
Two successive symbols are pre-coded as
shown
Need two orthogonal sources for two
channels estimation

33

Tx & Rx Diversity (MIMO)

Same as Tx Diversity, but with two Rxs

We have 4 channels, h0, h1, h2 and h3
Each receiver combines as before
The two receivers are then combined

s0 then -s 1*

ho
h1

s1 then s0*

h3

h2

Combine

Combine

so ( |h0|2 + |h1|2 + |h2|2 + |h3|2 )

s1 ( |h0|2 + |h1|2 + |h2|2 + |h3|2 )

34

Rake Receiver

Used for Direct Sequence Spread Spectrum Systems

Multipath diversity = multipath is advantageous
One finger (correlator) per path
Each finger synchronized to one path
Finger outputs combined (MRC)

. dt

Tb

c h ip
p ( t)

c 1 (t)

C a rrie r

d 1 ( t)

. dt

Tb

p ( t- )

c 1 (t- )

2
35

Rake Receiver, Cont.

Need to estimate channel gain for each
path
Rake Receiver performs Maximal Ratio
Combining
Number of fingers = number of paths
(ideally)
Small inter-path interference

36

Channel Equalization
Equalizers attempt to compensate for
channel fading effects
Linear Equalizer: FIR filter with adaptive
tap weights
Adaptation to minimize some criteria
Most famous: Least Mean Square (LMS)
Other criteria: Recursive Least Squares,
Kalman Filter, etc.

37

Linear Equalizer
y0

w0

y1

Z-1

w1

y2

Z-1

w2

Z-1

Z-1

wN-2

wN-1

yN-1

+
e

Threshold
+
data

LMS: wj(n+1)=wj(n) e*(n) yj(n)

38

Summary
Fading Types:
Large Scale: Distance + Shadowing
Small Scale: Fast or Slow & Flat or Selective

Counter Measures:
Diversity Types
Rake
Equalization

39