Wireless and Mobile Networking:

Facts, Statistics, and Trends

Nitin Sharma
SS ZG520
WMC
 Overview

Wireless: History
Wireless Infrastructure Hype Cycle 2013
Wireless Speed Trends (Moores Law)
Global Mobile Data Forecast [Cisco]
Trends
 Wireless: History
1880: Hertz discovered electromagnetic waves
1898: First commercial radio data service
1921: First Mobile Radio: Wireless dispatch system for
Detroit Police
1946: First Mobile Telephone Service: In St. Louis by
AT&T. Half-duplex Push to talk.
1970: First Cellular Phone Service: AT&T Chicago
1971: First Wireless Data Network: Aloha at University
of Hawaii
1990: First Commercial WLAN Product AT&T WaveLAN
1997: First WLAN Standard - IEEE 802.11 2-Mbps
Gartners hype cycle
Life Cycle of Technologies
Wireless Networking Infrastructure
Hype Cycle 2013
 Wireless Innovations
5G: Beyond 4G. Expected in 2020. 100X LTE
Cognitive Radio: Find unused channels and use them
802.11ah: Low-speed coordinated communication for M2M
TeraHz Waves: Sub-millimeter waves. 1 mm to 0.1mm
wavelength. 0.3 to 3THz. Between Radio and light
802.11ad: WiGig. Gigabit Wireless
Smart Antennas: Antenna arrays that can orient towards
direction of arrival
LTE-Advanced: Next generation of LTE. Real 4G. 1 Gbps
802.11ac: 500Mbps-1 Gbps WiFi
WiFi Direct: Point-to-Point WiFi without access point
802.11u: Authentication for 802.11 hotspots
 Wireless Innovations (Cont)
Small Cells: 10m to 2km. Includes Micro cells, Pico cells,
Femto cells
802.22: Wireless regional area network using white
spaces in TV channels
Super WiFi: Long-distance internet access using TV
white spaces
TD-LTE: LTE using time-division duplexing rather than
frequency division duplexing
ZigBee: Trade name for 802.15.4 personal area
networks. Like WiFi for 802.11
802.11r: Fast Base Station transition
LTE: Long-Term Evolution. 3.9G
802.11n: WiFi with multiple antennas
802.11k: Discover the best AP before transition
 Wireless Speed Trends

Doubling every 18 months Moores Law

Ref: G. Fettweis, The limits of 4G and how to design a new 5G Phy, http://www.ieee-
ctw.org/2013/slides/Fettweis.pdf
 Global Mobile Data Forecast [Cisco]
Global Mobile data grew 70% in 2012
885 Peta bytes per month in 2012 from 75 Peta
bytes per month in 2000
Mobile video traffic is more than 50% of the
mobile traffic
Connection speeds doubled 526 kbps in 2012 up
from 248 kbps in 2011
4G
Note: The next 12 slides are all from Cisco VNI
Ref: Cisco Visual Networking Index: Global Mobile Data Traffic Forecast Update
http://www.cisco.com/en/US/solutions/collateral/ns341/ns525/ns537/ns705/ns827/white_paper_c11-
520862.html
 Global Mobile Data Forecast [Cisco]
1. Global IP Traffic: 5X in last 5 years, 3X in next 5 years =>
23% Compound Annual Growth Rate (CAGR)
2. Busy hour traffic growing faster: 3.4X in next years
3. Mobile data traffic will increase 10X in 5 years
4. Business traffic will increase by 2X in 5 years
5. Metro traffic (2/3) and long haul (1/3)
Metro traffic will grow faster than long-haul traffic
Due to Content Delivery Networks (CDNs)
6. CDNs will carry a higher fraction of total traffic (63% in
2019 vs. 39% in 2014)
7. Over half-of all IP traffic will originate from non-PC devices
 Global Mobile Data Forecast [Cisco]
8. Traffic from wireless devices will exceed those
from wired
9. Number of devices will be 3X the population
size
10. Video traffic will be 80% of consumer traffic
11. Million minutes of video will cross Internet
per second
12. IP Traffic: fastest growth rate in the Middle
East and Africa followed by Asia Pacific
 Growth in IP Traffic

3X in 5 years
 Number of Devices

Devices = 3X population
Large # of M2M connections
Exponential Growth in Mobile Data
Highest Growth in Asia Pacific
Most Traffic due to Smart Phones
Majority of Traffic due to Mobile
Video
Offload from Cellular to Fixed
 Total IP Traffic

2/3 of all traffic will originate from non-PC

sources
 IoT Growth Areas

Highest growth in connected homes

3G and 4G on the Rise
2G vs. 3G vs. 4G Traffic Volume
 Top 1% Users

In January 2010, top 1% users generated 52% of

traffic
 Cellular vs. WiFi

4 times more WiFi

 Machine to Machine Modules

Security, health, Sensors Internet of things

Machine to Machine Traffic
Mobile Devices with IPv6
 100% Mobile Cellular Penetration
\\\\

96% globally, 128% in developed countries, 89%

in developing countries
Ref: ITU, The world in 2013: ICT Facts and Figures,
http://www.itu.int/en/ITUD/Statistics/Pages/facts/default.aspx
 Mobile Penetration

CIS = Commonwealth of Independent

States
Ref: ITU, The world in 2013: ICT Facts and Figures, http://www.itu.int/en/ITU-D/Statistics/Pages/facts/default.aspx
Growth of Mobile Broadband
 Trend: Access is All Wireless
Wireless, in the form of WiFi, started in 1999.First it
was an option.
Now it is standard in all computing devices
Most of the access (end user connectivity) is
wireless
Mobile phones, tablets have multiple wireless
technologies: 2G, 3G, 4G, WiFi, Bluetooth, NFC but
no wired connectivity
Trend: End user Computing is all Mobile
2007: Apple introduced iPhone
68-83% of people in the United States have
smart phones. 56% worldwide.
Internet is now available to masses
A large fraction of population uses smart
phones and tablets as their sole computing,
communication, entertainment device
Operating systems: Android, iOS, and Windows
All about mobiles
 Trend: Most Revenue in Wireless
US Wireless industry is valued at $195.5 billion
> publishing, agriculture, hotels and lodging, air
transportation, and motion picture and recording.
Wireless industry directly/indirectly provides
more than 2.6% of all US employment
Wireless revenue is expanding
Mobile Advertising Revenue
 Trend: Distributed App Development
Desktop Era: Large software packages office, Photoshop,
, each costing $100-$1000 per seat
Mobile Era: $0.99 apps or free apps for desktop
publishing, image manipulation,
Millions of App developers versus a large software
company
Personal Apps to Enterprise Apps
2 Billion Apps with 50 billion downloads
70 Apps stores
Mobile presence critical - Apps for Facebook, New York
Times, ...
Trend: Cloud Computing and Storage
High-Speed Wireless
Remote computing and storage
Ideal for mobile devices
Cloud storage:
Google Drive, iCloud, Sky Drive, Drop Box
Cloud Computing:
Google Docs, Office 360,
 Internet of Things
Only 1% of things around us is connected. Refrigerator, car,
washing machine, heater, a/c, garage door, should all be
connected but are not.
From 10 Billion today to 50 Billion in 2020 Should include
processes, data, things, and people.
$14 Trillion over 10 years
Third in the list of top 10 strategic technologies by Gartner (After
Mobile devices, Mobile Apps, but before Clouds, )
a.k.a. Internet of Everything by Cisco
Smarter Planet by IBM
Industrial Internet by GE
Cyber-Physical Systems (CPS)
Internet of European Things (more popular in Europe)
Ref: J. Bradley, The Internet of Everything: Creating Better Experiences in Unimaginable Ways, Nov 21, 2013, http://blogs.cisco.com/ioe/the-
internet-of-everything-creating-better-experiences-in-unimaginable-ways/#more-131793
Ref: Gartner Identifies Top 10 Strategic Technologies, http://www.cioinsight.com/it-news-trends/gartner-identifies-top-10-strategic-
technologies.html
 IEEE Communications Society Digital
Library: Top 10 Downloads (Nov 2013)
Performance Analysis of Macrodiversity MIMO Systems with MMSE and
ZF Receivers in Flat Rayleigh Fading
Are we ready for SDN? Implementation challenges for software-defined
networks
A survey on sensor networks
A simple transmit diversity technique for wireless communications
Performance analysis of the IEEE 802.11 distributed coordination function
Cognitive radio: brain-empowered wireless communications
A Survey of Defense Mechanisms Against Distributed Denial of Service
(DDoS) Flooding Attacks
Improving network management with software defined networking
A survey of spectrum sensing algorithms for cognitive radio applications
Network virtualization and software defined networking for cloud
computing: a survey
Observation: 7 wireless, 2 SDN, 1 Security.
Time Spent on Mobiles
Time to Reach 50M Users
 Summary: Wireless and Mobile
Trends
WiFi has grown worldwide in just 15 years
5G, Cognitive radio, M2M, TeraHz, Smart
Antennas, LTE Advanced are topics for active
research.
Wireless speed growth is following Moore's Law
Mobile subscriptions are approaching world
population
Most of the traffic is video, growth in asia pacific
 Acronyms
AP Access Point
CIO Chief Information Officer
CIS Commonwealth of Independent
CMO Chief Marketing Officer
CPS Cyber-Physical Systems
DDoS Distributed Denial of Service
DSL Digital Subscriber Line
GB Giga Byte
GE General Electric
GHz Giga Hertz
Hz Hertz
ICT Information and Communications Technologies
IEEE Institution of Electrical and Electronic Engineers
iOS iPhone Operating System
IPTS Institute for Prospective Technological Studies
IPv6 Internet Protocol Version 6
 Acronyms (Cont)
ITU International Telecommunications Union
KISDI Korea Information Society Development Institute
LTE Long-Term Evolution
MIMO Multiple Input Multiple Output
MMSE Minimum Mean Squared Error
NFC Near Field Communications
NGO Non-Governmental Organization
OFDM Orthogonal Frequency Division Multiplexing
RAN Regional Area Networks
RFID Radio Frequency Identification
SDN Software-defined networks
SSD Solid-state Storage Drive
TD-LTE Time-Division Duplixing Long-Term Evolution
TeraHz 1012 Hertz
THz Tera Hertz
TV Television
 Acronyms (Cont)
US United States
USB Universal Serial Bus
VNI Visual Networking Index
WiFi Wireless Fidelity
WiGig Gigabit Wireless
WLAN Wireless Local Area Network
WPAN Wireless Personal Area Network
ZigBee Trade name for 802.15.4
Chapter 2

INTRODUCTION TO SIGNALS
Definition
A Signal: is a function that specifies how a
specific variable changes versus an independent
variable such as time. Usually represented as an
X-Y plot.
Classification of Signals (1/4)
Analog vs. Digital signals:
Analog signals are signals with magnitudes that
may take any value in a specific range.

Digital signals have amplitudes that take only a

finite number of values.
Classification of Signals (2/4)
Continuous-time vs. discrete-time:
Continuous-time signals have their magnitudes
defined for all values of t. They may be analog
or digital.

Discrete-time signals have their magnitudes

defined at specific instants of time only. They
may be analog or digital.
Classification of Signals (3/4)
Periodic vs. aperiodic signals:
Periodic signals are signals constructed from a
shape that repeats itself regularly after a
specific amount of time T0, that is:
f(t) = f(t+nT0) for all integer n

Aperiodic signals do not repeat regularly.

Energy and Power

Ef =
2
| f (t ) | dt

T /2
1
Pf = lim
T T
T / 2
| f (t ) |2 dt
T +t0
1
= | f (t ) |
2
PPeriodic f dt
T t0
Classification of Signals (4/4)
Energy Signals: an energy signal is a signal
with finite energy and zero average power
(0 E < , P = 0)
Power Signals: a power signal is a signal with
infinite energy but finite average power
(0 < P < , E ).
More on Energy and Power Signals
A signal cannot be both an energy and power signal.
A signal may be neither energy nor power signal.
All periodic signals are power signals (but not all non
periodic signals are energy signals).
Any signal f that has limited amplitude (|f| < ) and is time
limited (f = 0 for |t |> t0) is an energy signal.
The square root of the average power of a power signal is
called the RMS value.
 Evaluate E and P and determine the type of the signal

=a (t ) 3sin(2 t ), < t <

1 1
1
= Pa =
2 2
| a (t ) | dt | 3sin(2 t ) | dt
=Ea =
| 3sin(2 t ) | dt 10
2 2
| a (t ) | dt 0

1
1
= 9 [1 cos(4 t ) ] dt
1
= 9 [1 cos(4 t )]dt 0
2
01

2 1
1

= 9 dt 9 cos(4 t )dt
1 2 20
= 9 dt 9 cos(4 t )dt
0
1
2 9 9

= sin(4 t )
= J 2 48 0
9
It is a Power signal = W
2
 Evaluate E and P and determine the type of the signal

b (t ) 5e 2|t | , < t <

=
T /2 T /2
1 1

2
= = 5e 2|t |
=
| b (t ) | dt 5e
2 2|t | 2
P lim | b (t ) |2
dt lim dt
Eb dt b
T T T T
T / 2 T / 2

0 T /2
0 1 1
=
= 25 e dt + 25 e dt + e 4t dt
4t
4t 4t 25 lim e dt 25 lim
T T T T
T / 2 0
0
25 1 4t 0 25 1 4t T / 2
=
25 4t 0 25 =
e + e 4t
lim e + lim e
4 T T T / 2 4 T
T 0
4 4 0
25 1 25 1 2T
=
25 25 50
+ = J = lim
1 e 2T

+ lim e 1
4 T T 4 T T
4 4 4
=0+0 =0

It is an energy signal
Basic Signal Operations (1/4)
Time Shifting: given the signal f(t), the signal
f(tt ) is a time-shifted version of f(t) that is
0

shifted to the left if t is positive and to the

0

right if t is negative.
0
Basic Signal Operations (2/4)
Magnitude Shifting:
Given the signal f(t), the signal c +f(t) is a
magnitude-shifted version of f(t) that is shifted
up if c is positive and shifted down if c is
negative.
Basic Signal Operations (3/4)
Time Scaling and Time Inversion: Given f(t),
the signal f(at) is a time-scaled version of f(t),
where a is a constant, such that f(at) is an
expanded version of f(t) if 0<|a|<1, and f(at) is
a compressed version of f(t) if |a|>1.
If a is negative, the signal f(at) is also a time-
inverted version of f(t).
Basic Signal Operations (4/4)
Magnitude Scaling and Mag. Inversion:
Given f(t), the signal bf(t) is a magnitude-
scaled version of f(t), where b is a constant,
such that bf(t) is an attenuated version of f(t)
if 0<|b|<1, and bf(t) is an amplified version of
f(t) if |b|>1.
If b is negative, the signal bf(t) is also a
magnitude-flipped version of f(t).
 Given f(t), sketch 43f(2t6)
f(t)

-2 6
-1
Unit Impulse Function (Dirac delta function)
Graphical Definition:
The rectangular pulse (t)
shape approaches the unit 1/
impulse function as
approaches 0 (notice that 0
the area under the curve
is always equal to 1). t

/2 /2
Unit Impulse Function (Dirac delta function)
Mathematical Definition:
The unit impulse function (t) satisfies the following
conditions:

1. (t) = 0 if t 0,

2.
(t )dt = 1

Properties of Delta Function
f(t)(t) = f(0)(t)

f (t ) (t =
t 0 ) f (t 0 ) (t t 0 )

f (t ) (t )dt = f (0)

f (t ) (t T )dt = f (T ) (t t )dt = f (t ) (t t )dt = f (t )

0 0

0 0

t
0, t < 0 du (t )
( )d = = u (t ) therefore, = (t )
1, t 0 dt
Trigonometric Fourier Series
A signal g(t) in the interval
t t t +T can be represented by
1 1 0

g (t ) = a0 + an cos(n0t ) + bn sin( n0t ) t1 t t1 + T0
n =1
t1 + T0 t1 + T0
1 2
a0 =
T0 g (t )dt an =
T0 g (t ) cos(n t )dt
t1
0
t1
t1 + T0
2
bn =
T0 g (t ) sin(n t )dt
t1
0 T0 = 2 / 0
 Or, in the compact form

g (t ) = C0 + Cn cos(n0t + n ) t1 t t1 + T0
n =1

bn
C0 = a0 ; Cn = a + b ;
2
n
2
n n = tan
1

an
If g(t) is even then bn = 0 for all n
If g(t) is odd then an=0 for all n.
Remarks on Fourier Series (FS) Representations
The frequency 0= 2/T0 is called the fundamental
frequency and the multiple of this frequency n0 is
called the nth harmonic.
FS of g(t) is equal to g(t) over the interval t1 t
t1+T0 only.
The FS for all t is a periodic function of period T0
in which the segment of g(t) over the interval t1 t
t1+T0 repeats periodically.
If the function g(t) itself is periodic with period T0
then the FS represents g(t) for all t.
Exponential Fourier Series

g (t ) = D e
n =
n
jn 0 t
= D0 + D e
n =
n
jn 0 t
t1 t t1 + T0
( n 0)
1
Dn =
T0 T0
g (t )e jn 0 t dt
Dn is related to Cn and n as
1
| Dn |=| D n |= Cn Dn = Dn = n
2
| Dn | is called the amplitude spectrum of the signal.
Dn is called the phase spectrum of the signal.
They provide a frequency-domain representation of the signal.
Parsevals Theorem
Let g(t) be a periodic signal. The power of g(t) is
equal to the sum of the powers of its Fourier
Components.

g (t ) = C0 + Cn cos(n0t + n )
1
Pg = C0 +
2
C n2
n =1
2 n =1

g (t ) =
n =
Dn e jn0t
Pg = Dn
2

n =

D
2
For g(t) real, D n = Dn Pg = D02 +2 n
n =1
Chapter 3

ANALYSIS AND TRANSMISSION OF

SIGNALS
From Fourier Series to Fourier Transform
We found how to evaluate the frequency spectrum of
periodic functions of period T0 or aperiodic functions
over an interval of T0.
The representation is in the form of discrete harmonics
of frequencies n0 = 2n/ T0
What is the spectrum of aperiodic function for all t, i.e.
T0 infinity?
In this case the spectrum is a continuous function of ,
and is called the Fourier Transform (FT) of the time
function.

2
Definition and Properties of FT

j t
G ( ) = F [ g (t )] = g (t ) e dt

G() is, in general, a complex signal that has both magnitude and phase

G ( ) = G ( ) e j G ( )
If g(t) is REAL then |G()| is even and G() is odd. (Conjugate Symmetry)
G*() = G()
|G*()| = |G()| = |G()|
G(-) = G()
If g(t) is REAL and EVEN then G() is PURE REAL
If g(t) is REAL and ODD, then G() is PURE IMAGINARY

3
 FT of Impulse Function

F [ (t )] = (t )e jt dt = (t )e j ( 0 ) dt = (t )dt = 1

(t) 1

4
 FT of the Gate Function
0 | t |> / 2
t
rect = 1 / 2 | t |= / 2

1 | t |< / 2
/2 /2
1 j 2 j

t t
j t 1 j t
e dt = e dt = e e 2
j t
F rect = rect e =
/ 2
j / 2
j
j 2
j
e e 2 sin( / 2)
= =
/ 2 2j / 2

The function sin(x)/x is called sinc(x)
rect(t/) sinc(/2)

5
Inverse Fourier Transform (IFT)

j t
G ( ) = F [ g (t )] = g (t ) e dt


1

j t
g (t ) = F [G ( )] =
1
G ( ) e d
2

g(t) G()

6
Basic Signal Operations in the Frequency Domain

Magnitude Scaling:
kg(t) kG()
Time Scaling
g(at) |1/a|G(/a)
Time Shifting
g(t t ) G()e-jt
0 0

7
More on Fourier Transform Operation
Linearity: g (t)+g (t) G ()+G ()
1 2 1 2

Symmetry Property: G(t) 2g()

Frequency Shifting: g(t) e j t G( )
0 0

Time Differentiation dg(t)/dt jG()

dng(t)/dtn (j)nG()
t
1
Time Integration g ( )d G (0) ( ) + j G ( )

8
Proof of Time Scaling Property

j t
F [ g (at )] = g ( at ) e dt

Let = at d = a dt
if a > 0: =
1 1
j ( at ) j ( ) 1

1 1
F[ g (at )] = g (at )e a d (at ) = g ( )e a d = G
a a a a

if a < 0: = -
1 1
j ( at ) j ( ) 1

1 1
F[ g (at )] = g (at )e a d (at ) = g ( )e a d = G
a a a a

1
g (at ) G
a a
9
Proof of Time Differentiation Property
dg (t ) d
F = F {g (t )}
dt dt
d 1

G( )e d
jt
=F
dt 2
1

dg (t )
F =F
d
{ }
dt G ( )e d
j t

dt 2
1

j t
=F jG ( )e d
2
[ ]
= F F -1{ jG ( )} = jG ( )

10
 FT of a Cosine Function
cos(0t) = [ej t + e-j t]
0 0

(t) 1
1 2 () (Symmetry Property)
1 ej t 2 (0) (Frequency Shifting)
0

cos(0t) [ (0) + (+0)] (Linearity)

Similarly
sin(0t) (/j) [ (0) - (+0)]

11
 Multiplying by a Sinusoid

Find F [g(t)cos(ot) ]

Using the frequency shifting property,

g(t)cos(ot) (1/2)[G( 0) + G( +0)]

g(t)sin(0t) (1/j2 )[G( 0) G( +0)]

12
 Find FT of the signal below
z2(t)
5

t +1
4
z 2 (t ) = 2 cos ( 6 t ) rect
2 4
1
t
-5 -4 -3 1 4 7

-2

1 4 ( 6 ) j ( 6 )( 1) 4 ( + 6 ) j ( + 6 )( 1)
Z 2 ( )
= (2)4sinc e + (2)4sinc e
2 2 2
= 4sinc ( 2 ( 6 ) ) e j ( 6 ) + 4sinc ( 2 ( + 6 ) ) e j ( + 6 )

13
Convolution
Time convolution

g (t ) * f (t ) = g ( ) * f (t )d = f ( ) * g (t )d

g (t ) * f (t ) G ( ) F ( )

Frequency Convolution

G ( ) * F ( ) = G ( s ) * F ( s )ds = F (s) * G( s)ds

1
g (t ) f (t ) G ( ) * F ( )
2
14
 Multiplying by a Sinusoid
(Alternative Solution)

Find F [g(t)cos(ot) ]
Using the convolution property,
g(t)cos(ot) (1/2)[G()*{ [ (0)+ (0) ]
Note that
G()* (0) 2[g(t) (1/2)ej t ]G (0)
0

That is G()* (0) = G (0)

g(t)cos(ot) (1/2)[G( 0) + G( +0)]

15
Time-Frequency Duality
For any relationship between g(t) and G(), there exists a
dual relationship obtained by interchanging the roles in the
original relationship with possibly minor modifications
Shifting:
g(t t ) G()e-jt and g(t) ej t G( 0)
0 0 0

Convolution
1
g (t ) * f (t ) G ( ) F ( ) g (t ) f (t ) G ( ) * F ( )
2
Scaling
g(at) |1/a|G(/a)

16
Linear Systems
A system is frequently described by its impulse response
h(t).
The FT of h(t) is called the transfer function H().
The input-output relation of a linear system is given by:
y(t) = g(t)*h(t)
In frequency domain
Y() = G() H()
|Y()|ej ()=|G()||()|ej[ ()+ ()]
Y G H

17
Distortionless Communication System (1/2)
If the output signal of a specific communication
system is an amplified/attenuated and delayed
version form of the input signal, than the
system is a distortionless communication
system.
For such system y(t) = kg(t-td).
How do we describe h(t) and H() of a
distortionless system?

18
Distortionless Communication System (2/2)
h(t) = k (t-td)
H() = ke-jt
d

|H()|=k; H()=-td
For a distortionless system, the amplitude
spectrum must be flat, and the phase spectrum
must be a linear function of .
Note that the phase must be linear with
frequency in order for the delay of various
frequency components be constant
19
Filters
A filter is a systems that passes certain
frequencies and block others.
Because they operate on frequencies, they are
easier to visualize in the frequency domain.
Therefore, filters are usually represented by
their transfer functions H().

20
Low-Pass Filters

HLPF()

-W1 W1

Bandwidth W1
W1 also called the cutoff frequency

21
High-Pass Filters

HHPF()

-W2 W2

Bandwidth not defined

Filter defined by its cutoff frequency W2

22
Band-Pass Filters
HBPF()

-W2 -W1 W1 W2

BPF is characterized by two frequencies, the lower

cutoff frequency W1, and the upper cutoff frequency W2
The bandwidth of the filter is BW = W1 W2

23
Practical Filters (1/2)
The frequency responses for the three types of filters shown above
are those of ideal filters. There is an extremely sharp transition
between the passbands and stopbands.
The sharpness of the transition between passband and stopband is
determined by something called the ORDER of the filter.
The order of the filter is generally determined by the number of
reactive components (capacitors and inductors) used to build the
filter.
A zeroorder filter is a flat filter that allows all signals to pass.
A firstorder filter has very slow transition between the passband
and stopband.
A secondorder filter has a steeper transition, and so on.

24
Practical Filters (2/2)
Ideal (infinite order) LPF
HLPF()
3rd order LPF

2nd order LPF

1st order LPF

-W1 W1

An ideal filter has an infinite order which makes it unrealizable.

Also, an ideal filter would result in an infinite amount of delay.

25
Chapter 3

The Cellular Concept - System Design

Fundamentals
I. Introduction

Goals of a Cellular System

High capacity
Large coverage area
Efficient use of limited spectrum
Large coverage area - Bell system in New York City
had early mobile radio
Single Tx, high power, and tall tower
Low cost
Large coverage area - Bell system in New York City had 12
simultaneous channels for 1000 square miles
Small # users
Poor spectrum utilization
What are possible ways we could increase the number
of channels available in a cellular system?
2
 Cellular concept
Frequency reuse pattern

3
 Cells labeled with the same letter use the same
group of channels.
Cell Cluster: group of N cells using complete set of
available channels
Many base stations, lower power, and shorter
towers
Small coverage areas called cells
Each cell allocated a % of the total number of
available channels
Nearby (adjacent) cells assigned different channel
groups
to prevent interference between neighboring base
stations and mobile users
4
 Same frequency channels may be reused by cells a
reasonable distance away
reused many times as long as interference between same
channel (co-channel) cells is < acceptable level
As frequency reuse # possible simultaneous
users # subscribers but system cost (more
towers)
To increase number of users without increasing radio
frequency allocation, reduce cell sizes (more base
stations) # possible simultaneous users
The cellular concept allows all mobiles to be
manufactured to use the same set of freqencies
*** A fixed # of channels serves a large # of users
by reusing channels in a coverage area ***
5
GSM Cellular Architecture

6
 Mobile Station (MS)
MS is the users handset and has two parts
Mobile Equipment
Radio equipment
User interface
Processing capability and memory required for
various tasks
Call signalling
Encryption
SMS
Equipment IMEI number

7
Subscriber Identity Module

A small smart card

Encryption codes needed to identify the
subscriber
Subscriber IMSI number
Subscribers own information (telephone
directory)
Third party applications (banking etc.)
Can also be used in other systems besides
GSM, e.g., some WLAN access points
accept SIM based user authentication
8
Base Station Subsystem
Transcending Rate and Adaptation Unit (TRAU)
Performs coding between the 64kbps PCM coding used
in the backbone network and the 13 kbps coding used
for the Mobile Station (MS)
Base Station Controller (BSC)
Controls the channel (time slot) allocation implemented
by the BTSes
Manages the handovers within BSS area
Knows which mobile stations are within the cell and
informs the MSC/VLR about this
Base Transceiver System (BTS)
Controls several transmitters
Each transmitter has 8 time slots, some used for
signaling, on a specific frequency

9
Network and Switching Subsystem
The backbone of a GSM network is a telephone network with
additional cellular network capabilities
Mobile Switching Center (MSC)
An typical telephony exchange (ISDN exchange) which supports
mobile communications
Visitor Location Register (VLR)
A database, part of the MSC
Contains the location of the active Mobile Stations
Gateway Mobile Switching Center (GMSC)
Links the system to PSTN and other operators
Home Location Register (HLR)
Contain subscriber information, including authentication
information in Authentication Center (AuC)
Equipment Identity Register (EIR)
International Mobile Station Equipment Identity (IMEI) codes for
e.g., blacklisting stolen phones

10
HLR
One database per operator
Contains all the permanent subscriber information
MSISDN (Mobile Subscriber ISDN number) is the telephone
number of the subscriber
International Mobile Subscriber Identity (IMSI) is a 15 digit
code used to identify the subscriber
It incorporates a country code and operator code
IMSI code is used to link the MSISDN number to the
subscribers SIM (Subscriber Identity Module)
Charging information
Services available to the customer
Also the subscribers present Location Area Code, which
refers to the MSC, which can connect to the MS.

11
 Authentication Center (AuC) used in the security
data management for the authentication of
subscribers.
Equipment Identity Register (EIR) used to
maintain a list of legitimate, fraudulent, or faulty
MSs. optional in GSM network, and is not used
generally.
HLR (Home Location Register) a database to
store and management permanent data of
subscribers
VLR (Visitor Location Register) a database to
store temporary information about subscribers
needed by MSC in order to service visiting
subscribers
12
II. Frequency Reuse/Planning

Design process of selecting & allocating

channel groups of cellular base stations
Two competing/conflicting objectives:
1) maximize frequency reuse in specified area
2) minimize interference between cells

13
 Cells
base station antennas designed to cover specific cell
area
hexagonal cell shape assumed for planning
simple model for easy analysis circles leave gaps
actual cell footprint is amorphous (no specific shape)
where Tx successfully serves mobile unit
base station location
cell center omni-directional antenna (360 coverage)
not necessarily in the exact center (can be up to R/4
from the ideal location)

14
 Choices of Hexagonal Cell Geometry
Factors
Equal area
No overlap between cells
Choices

For a given R, A3 provides maximum coverage area.

By using hexagon geometry, the fewest number of cells
covers
a given geographic region.

15
 cell corners sectored or directional antennas
on 3 corners with 120 coverage.
very commom
Note that what is defined as a corner is
somewhat flexible a sectored antenna covers
120 of a hexagonal cell.
So one can define a cell as having three antennas
in the center or antennas at 3 corners.

16
III. System Capacity

S : total # of duplex channels available for use

in a given area; determined by:
amount of allocated spectrum
channel BW modulation format and/or standard
specs. (e.g. AMPS)
k : number of channels for each cell (k < S)
N : cluster size # of cells forming cluster
S=kN

17
 M : # of times a cluster is replicated over a
geographic coverage area
System Capacity = Total # Duplex Channels = C
C=MS=MkN
(assuming exactly MN cells will cover the area)
If cluster size (N) is reduced and the geographic area
for each cell is kept constant:
The geographic area covered by each cluster is smaller, so
M must to cover the entire coverage area (more clusters
needed).
S remains constant.
So C
The smallest possible value of N is desirable to maximize
system capacity.
18
 Cluster size N determines:
distance between co-channel cells (D)
level of co-channel interference
A mobile or base station can only tolerate so much
interference from other cells using the same
frequency and maintain sufficient quality.
large N large D low interference but small
M and low C !
Tradeoff in quality and cluster size.
The larger the capacity for a given geographic area,
the poorer the quality.

19
 Frequency reuse factor = 1 / N
each frequency is reused every N cells
each cell assigned k S / N
N cells/cluster
connect without gaps
specific values are required for hexagonal geometry
N = i2 + i j + j2 where i, j 1
Typical N values 3, 4, 7, 12; (i, j) = (1,1), (2,0),
(2,1), (2,2)

20
 To find the nearest co-channel neighbors of a particular cell
(1) Move i cells along any chain of hexagons, then (2)
turn 60 degrees and move j cells.

21
22
23
Frequency Reuse Notation

NSK frequency reuse pattern

N=Number of cells per cluster
S= Number of sectors in a cell
K = Number of frequency allocations per cell
1X3X3

24
Frequency Reuse Notation (Cont)

25
Fractional Frequency Reuse

Users close to the BS use all frequency

subchannels
Users at the cell boundary use only a fraction of
available subchannels

26
IV. Channel Assignment Strategies
Goal is to minimize interference & maximize use of
capacity
lower interference allows smaller N to be used greater
frequency reuse larger C
Two main strategies: Fixed or Dynamic
Fixed
each cell allocated a pre-determined set of voice channels
calls within cell only served by unused cell channels
all channels used blocked call no service
several variations
MSC allows cell to borrow a VC (that is to say, a FVC/RVC
pair) from an adjacent cell
donor cell must have an available VC to give

27
 Dynamic
channels NOT allocated permanently
call request goes to serving base station goes
to MSC
MSC allocates channel on the fly
allocation strategy considers:
likelihood of future call blocking in the cell
reuse distance (interference potential with other cells
that are using the same frequency)
channel frequency
All frequencies in a market are available to be used

28
 Advantage: reduces call blocking (that is to say,
it increases the trunking capacity), and
increases voice quality
Disadvantage: increases storage &
computational load @ MSC
requires real-time data from entire network related
to:
channel occupancy
traffic distribution
Radio Signal Strength Indications (RSSI's) from all
channels

29
V. Handoff Strategies

Handoff: when a mobile unit moves from one

cell to another while a call is in progress, the
MSC must transfer (handoff) the call to a new
channel belonging to a new base station
new voice and control channel frequencies
very important task often given higher priority
than new call
It is worse to drop an in-progress call than to deny a
new one

30
 Minimum useable signal level
lowest acceptable voice quality
call is dropped if below this level
specified by system designers
typical values -90 to -100 dBm

31
 Quick review: Decibels

S = Signal power in Watts

Power of a signal in decibels (dBW) is Psignal = 10 log10(S)
Remember dB is used for ratios (like S/N)
dBW is used for Watts

dBm = dB for power in milliwatts = 10 log10(S x 103)

dBm = 10 log10(S) + 10 log10(103) = dBW + 30
-90 dBm = 10 log10(S x 103)
10-9 = S x 103
S = 10-12 Watts = 10-9 milliwatts
-90 dBm = -120 dBW

Signal-to-noise ratio:
N = Noise power in Watts
S/N = 10 log10(S/N) dB (unitless raio)
32
 choose a (handoff threshold) > (minimum
useable signal level)
so there is time to switch channels before level
becomes too low
as mobile moves away from base station and
toward another base station

33
34
 Handoff Margin
= Phandoff threshold - Pminimum usable signal dB
carefully selected
too large unnecessary handoff MSC loaded down
too small not enough time to transfer call dropped!
A dropped handoff can be caused by two factors
not enough time to perform handoff
delay by MSC in assigning handoff
high traffic conditions and high computational load on MSC
can cause excessive delay by the MSC
no channels available in new cell

35
 Handoff Decision
signal level decreases due to
signal fading dont handoff
mobile moving away from base station handoff
must monitor received signal strength over a period
of time moving average
time allowed to complete handoff depends on
mobile speed
large negative received signal strength (RSS) slope
high speed quick handoff
statistics of the fading signal are important to
making appropriate handoff decisions Chapters
4 and 5
36
 1st Generation Cellular (Analog FM AMPS)
Received signal strength (RSS) of RVC measured
at base station & monitored by MSC
A spare Rx in base station (locator Rx) monitors
RSS of RVC's in neighboring cells
Tells Mobile Switching Center about these mobiles and
their channels
Locator Rx can see if signal to this base station is
significantly better than to the host base station
MSC monitors RSS from all base stations &
decides on handoff

37
 2nd Generation Cellular w/ digital TDMA (GSM,
IS-136)
Mobile Assisted HandOffs (MAHO)
important advancement
The mobile measures the RSS of the FCCs from
adjacent base stations & reports back to serving base
station
if Rx power from new base station > Rx power from
serving (current) base station by pre-determined
margin for a long enough time period handoff
initiated by MSC

38
 MSC no longer monitors RSS of all channels
reduces computational load considerably
enables much more rapid and efficient handoffs
imperceptible to user

39
 A mobile may move into a different system
controlled by a different MSC
Called an intersystem handoff
What issues would be involved here?

Prioritizing Handoffs
Issue: Perceived Grade of Service (GOS) service
quality as viewed by users
quality in terms of dropped or blocked calls (not
voice quality)
assign higher priority to handoff vs. new call request
a dropped call is more annoying than an occasional
blocked call
40
 Guard Channels
% of total available cell channels exclusively set
aside for handoff requests
makes fewer channels available for new call
requests
a good strategy is dynamic channel allocation (not
fixed)
adjust number of guard channels as needed by demand
so channels are not wasted in cells with low traffic

41
 Queuing Handoff Requests
use time delay between handoff threshold and
minimum useable signal level to place a blocked
handoff request in queue
a handoff request can "keep trying" during that time
period, instead of having a single block/no block
decision
prioritize requests (based on mobile speed) and
handoff as needed
calls will still be dropped if time period expires

42
VI. Practical Handoff Considerations

Problems occur because of a large range of

mobile velocities
pedestrian vs. vehicle user
Small cell sizes and/or micro-cells larger #
handoffs
MSC load is heavy when high speed users are
passed between very small cells

43
 Umbrella Cells
Fig. 3.4, pg. 67
use different antenna heights and Tx power levels to
provide large and small cell coverage
multiple antennas & Tx can be co-located at single
location if necessary (saves on obtaining new tower
licenses)
large cell high speed traffic fewer handoffs
small cell low speed traffic
example areas: interstate highway passing thru
urban center, office park, or nearby shopping mall

44
45
 Cell Dragging
low speed user w/ line of sight to base station (very strong
signal)
strong signal changing slowly
user moves into the area of an adjacent cell without handoff
causes interference with adjacent cells and other cells
Remember: handoffs help all users, not just the one which is
handed off.
If this mobile is closer to a reused channel interference
for the other user using the same frequency
So this mobile needs to hand off anyway, so other users
benefit because that mobile stays far away from them.

46
 Typical handoff parameters
Analog cellular (1st generation)
threshold margin 6 to 12 dB
total time to complete handoff 8 to 10 sec
Digital cellular (2nd generation)
total time to complete handoff 1 to 2 sec
lower necessary threshold margin 0 to 6 dB
enabled by mobile assisted handoff

47
 benefits of small handoff time
greater flexibility in handling high/low speed
users
queuing handoffs & prioritizing
more time to rescue calls needing urgent
handoff
fewer dropped calls GOS increased
can make decisions based on a wide range of
metrics other than signal strength
such as also measure interference levels
can have a multidimensional algorithm for
making decisions
48
 Soft vs. Hard Handoffs
Hard handoff: different radio channels assigned
when moving from cell to cell
all analog (AMPS) & digital TDMA systems (IS-136,
GSM, etc.)
Many spread spectrum users share the same
frequency in every cell
CDMA IS-95
Since a mobile uses the same frequency in every cell, it
can also be assigned the same code for multiple cells
when it is near the boundary of multiple cells.
The MSC simultaneously monitors reverse link signal
at several base stations

49
 MSC dynamically decides which signal is best
and then listens to that one
Soft Handoff
passes data from that base station on to the PSTN
This choice of best signal can keep changing.
Mobile user does nothing for handoffs except
just transmit, MSC does all the work
Advantage unique to CDMA systems
As long as there are enough codes available.

50
VII. Co-Channel Interference

Interference is the limiting factor in

performance of all cellular radio systems
What are the sources of interference for a
mobile receiver?
Interference is in both
voice channels
control channels
Two major types of system-generated
interference:
1) Co-Channel Interference (CCI)
2) Adjacent Channel Interference (ACI)

51
 First we look at CCI
Frequency Reuse
Many cells in a given coverage area use the same
set of channel frequencies to increase system
capacity (C)
Co-channel cells cells that share the same set of
frequencies
VC & CC traffic in co-channel cells is an
interfering source to mobiles in Several different
cells

52
 Possible Solutions?
1) Increase base station Tx power to improve radio
signal reception? __
this will also increase interference from co-channel
cells by the same amount
no net improvement
2) Separate co-channel cells by some minimum
distance to provide sufficient isolation from
propagation of radio signals?
if all cell sizes, transmit powers, and coverage patterns
same co-channel interference is independent of
Tx power

53
 co-channel interference depends on:
R : cell radius
D : distance to base station of nearest co-channel cell
if D / R then spatial separation relative to cell
coverage area
improved isolation from co-channel RF energy
Q = D / R : co-channel reuse ratio
hexagonal cells Q = D/R = 3N

54
 Fundamental tradeoff in cellular system design:
small Q small cluster size more frequency
reuse larger system capacity great
But also: small Q small cell separation
increased co-channel interference (CCI) reduced
voice quality not so great
Tradeoff: Capacity vs. Voice Quality

55
 Signal to Interference ratio S / I, ____________

S : desired signal power

Ii : interference power from ith co-channel cell
io : # of co-channel interfering cells

56
 Approximation with some assumptions

Di : distance from ith interferer to mobile

Rx power @ mobile ( Di ) n

57
 n : path loss exponent
free space or line of sight (LOS) (no obstruction)
n=2
urban cellular n = 2 to 4, signal decays faster
with distance away from the base station
having the same n throughout the coverage area
means radio propagation properties are roughly the
same everywhere
if base stations have equal Tx power and n is the
same throughout coverage area (not always true)
then the above equation (Eq. 3.8) can be used.

58
 Now if we consider only the first layer (or tier)
of co-channel cells
assume only these provide significant interference
And assume interfering base stations are
equidistant from the desired base station (all at
distance D) then

59
 What determines acceptable S / I ?
voice quality subjective testing
AMPS S / I 18 dB (assumes path loss exponent
n = 4)
Solving (3.9) for N

Most reasonable assumption is io : # of co-channel

interfering cells = 6
N = 7 (very common choice for AMPS)
60
 Many assumptions involved in (3.9) :
same Tx power
hexagonal geometry
n same throughout area
Di D (all interfering cells are equidistant from the
base station receiver)
optimistic result in many cases
propagation tools are used to calculate S / I when
assumptions arent valid

61
 S / I is usually the worst case when a mobile is at the
cell edge
low signal power from its own base station & high
interference power from other cells
more accurate approximations are necessary in those cases

S R 4
=
I 2( D R ) 4 + 2( D + R ) 4 + 2 D 4

62
N =7 and S / I 17 dB

63
 Eq. (3.5), (3.8), and (3.9) are (S / I) for forward link
only, i.e. the cochannel base Tx interfering with
desired base station transmission to mobile unit
so this considers interference @ the mobile unit
What about reverse link co-channel interference?
less important because signals from mobile antennas (near
the ground) dont propagate as well as those from tall base
station antennas
obstructions near ground level significantly attenuate mobile
energy in direction of base station Rx
also weaker because mobile Tx power is variable base
stations regulate transmit power of mobiles to be no larger
than necessary

64
I. Adjacent Channel Interference

Two major types of system-generated

interference:
1) Co-Channel Interference (CCI) discussed in last
lecture
2) Adjacent Channel Interference (ACI)
Adjacent Channel Interference (ACI)
Imperfect Rx filters allow energy from adjacent
channels to leak into the passband of other
channels

65
 desired filter response

actual filter response

66
 This affects both forward & reverse links
Forward Link base-to-mobile
interference @ mobile Rx from a ______ Tx
(another mobile or another base station that is not
the one the mobile is listening to) when mobile Rx
is ___ away from base station.
signal from base station is weak and others are
somewhat strong.
Reverse Link mobile-to-base
interference @ base station Rx from nearby mobile
Tx when desired mobile Tx is far away from base
station
67
 Near/Far Effect
interfering source is near some Rx when desired
source is far away
ACI is primarily from mobiles in the same cell
some cell-to-cell ACI does occur as well but a
secondary source
Control of ACI
dont allocate channels within a given cell from a
contiguous band of frequencies
for example, use channels 1, 4, 7, and 10 for a cell.
no channels next to each other

68
 maximize channel separation
separation of as many as N channel bandwidths
some schemes also seek to minimize ACI from
neighboring cells by not assigning adjacent
channels in neighboring cells

69
70
 Originally 666 channels, then 10 MHz of
spectrum was added
666+166 = 832 channels
395 VC plus 21 CC per service provider
(providers A & B)
395*2 = 790, plus 42 control channels
Provider A is a company that has not
traditionally provided telephone service
Provider B is a traditional wireline operator
21 VC groups with 19 channels/group
at least 21 channel separation for each group
71
 for N = 7 3 VC groups/cell
For example, choose groups 1A, 1B, and 1C for a
cell so channels 1, 8, 15, 22, 29, 36, etc. are used.
57 channels/cell
at least 7 channel separation for each cell group
to have high quality on control channels, 21 cell
reuse is used for CCs
instead of reusing a CC every 7 cells, as for VCs,
reuse every 21 cells (after every three clusters)
greater distance between control channels, so less
CCI
72
 use high quality filters in base stations
better filters are possible in base stations since they
are not constrained by physical size and power as
much as in the mobile Rx
makes reverse link ACI less of a concern than
forward link ACI
also true because of power control (discussed below)
choice of modulation schemes
different modulation schemes provide less or more
energy outside their passband.

73
 Power Control
technique to minimize ACI
base station & MSC constantly monitor mobile
received signal strength
mobile Tx power varied (controlled) so that
smallest Tx power necessary for a quality reverse
link signal is used (lower power for the closer the
mobile is to the base station)
also helps battery life on mobile

74
 dramatically improves adjacent channel S / I
ratio, since mobiles in other cells only transmit
at high enough power as transmitter controls
(not at full power)
most beneficial for ACI on reverse link
will see later that this is especially important for
CDMA systems

75
III. Trunking & Grade of Service (GOS)

Trunked radio system: radio system where a

large # of users share a pool of channels
channel allocated on demand & returned to channel
pool upon call termination
exploit statistical (random) behavior of users so that
fixed # of channels can accommodate large # of
users
Trade-off between the number of available channels
that are provided and the likelihood of a particular user
finding no channels available during the busy hour of
the day.

76
 trunking theory is used by telephone companies to
allocate limited # of voice circuits for large # of
telephone lines
efficient use of equipment resources savings
disadvantage is that some probability exists that
mobile user will be denied access to a channel
blocked call : access denied blocked call cleared
delayed call : access delayed by call being put into
holding queue for specified amount of time

77
 GOS : measure of the ability of user access to a
trunked system during the _______ hour
specified as probability (Pr) that call is blocked or
delayed
designed to handle the busiest hour typically
______
Erlang : unitless measure of traffic intensity
e.g. 0.5 erlangs = 1 channel occupied 30 minutes
during 1 hour
Table 3.3, pg. 78 trunking theory definitions

78
 Offered Traffic Intensity (A)
Offered? not necessarily carried by system
(some is blocked or delayed)
each user Au=H Erlangs (also called in queueing
theory)
= traffic intensity (average arrival rate of new calls,
in new requests per time unit, say calls/min).
H = average duration of a call (also called 1/ in
queueing theory)
system with U users A = UAu = UH Erlangs
capacity = maximum carried traffic = C Erlangs =
(equal to total # of available channels that are busy
all the time) 79
 Erlang B formula
Calls are either admitted or blocked

A = total offered traffic

C = # channels in trunking pool (e.g. a cell)
AMPS designed for GOS of 2%
blocked call cleared (denied) BCC

80
 capacities to support various GOS values

Note that twice the capacity can support much more than
twice the load (twice the number of Erlangs).
81
 Erlang C formulas
blocked call delayed BCD put into holding
queue
GOS is probability that a call will still be blocked
even if it spends time in a queue and waits for up to
t seconds
equations (3.17) to (3.19) in book

82
 Graphical form of Erlang B formulas

83
 Graphical form of Erlang C formulas

84
 Example: Find how many users can be
supported in a cell containing 50 channels for a
2% GOS (Blocked Calls Cleared) if the average
user calls twice/hr with an average call duration
of 5 minutes.
What is the corresponding C from the figure?

What is A (Traffic Intensity) from the figure?

So, how many users can be supported?

85
 Trunking Efficiency
measure of the # of users supported by a specific
configuration of fixed channels, efficiency in terms
of users per available channel = U / C
Table 3.4, pg. 79 assume 1% GOS
Assume Au = 0.2
1 group of 20 channels:

2 groups of 10 channels, with equal number of users

per group:

86
 the allocation of channel groups can
substantially change the # of users supported by
trunked system
The larger the trunking pool, the better the trunking
efficiency.
as trunking pool size then trunking efficiency

What is the relationship between trunking pool size,
trunking efficiency, received signal quality, and
cluster size?
As cluster size decreases
87
 Note: Trunking efficiency is an issue both in
FDMA/TDMA systems and in CDMA systems
(where the capacity limit is the number of
possible codes and the interference levels).

88
89
90
91
92
93
94
95
96
97
98
IV. Improving Cellular System Capacity

A cellular design eventually (hopefully!)

becomes insufficient to support the growing
number of users.
Need to provide more channels per unit coverage
area
Would like to have orderly growth
Would like to upgrade the system instead of rebuild
Would like to use existing towers as much as
possible

99
 Cell Splitting
subdivide congested cell into several smaller
cells
increases number of times channels are reused
in an area
must decrease antenna height & Tx power
so smaller coverage per cell results
and the co-channel interference level is held
constant

100
 each smaller cell keeps same # of channels as
the larger cell, since each new smaller cell uses
the same number of frequencies
this means that we keep that same cluster size
capacity because channel reuse per unit area
smaller cells micro-cells

101
 Illustration is for towers at the corners

102
 advantages include:
only needed for cells that reach max. capacity not
all cells
implement when Pr [blocked call] > acceptable GOS
system capacity can gradually expand as demand
disadvantages include:
# handoffs/unit area increases
umbrella cell for high velocity traffic may be needed
more base stations $$ for real estate, towers, etc.

103
 complicated design process
new base stations use lower power and antenna
height
What about existing base stations?
If kept at the same power, they would overpower new
microcells.
If reduced in power, they would not cover their own
cells.
One solution: Use separate groups of channels.
One group at the original power and another group at
the lower power.
New microcells only use lower power channels.
As load growth continues, more and more channels are
moved to lower power.
104
105
106
 Sectoring
cell splitting keeps D / R unchanged (same
cluster size and CCI) but increases frequency
reuse/area
alternate way to capacity is to _____ CCI
(increase S / I ratio)

107
 replace omni-directional antennas at base station
with several directional antennas
3 sectors 3 120 antennas
6 sectors 6 60 antennas

108
 cell channels broken down into sectored groups
CCI reduced because only some of neighboring co-
channel cells radiate energy in direction of main cell
center cell labeled "5" has all co-channel cells
illustrated
only 2 co-channel cells will interfere if all are using
120 sectoring
only 1 co-channel cell would interfere when using
60 sectoring
If the S/I was 17 dB for N = 7 and n = 4, what is the
S / I now with 120 sectoring?
24.2 dB

109
110
 How is capacity increased?
sectoring only improves S/I which increases voice
quality, beyond what is really necessary
by reducing CCI, the cell system designer can choose
smaller cluster size (N ) for acceptable voice quality
smaller N greater frequency reuse larger system
capacity

What would the system capacity, Cnew, now be when

using 120 sectoring, as compared to the old capacity,
Cold ?

111
112
113
 much less costly than cell splitting
only require more antennas @ base station vs.
multiple new base stations for cell splitting
primary disadvantage is that the available
channels in a cell are subdivided into sectored
groups
trunked channel pool , therefore trunking
efficiency
There are more channels per cell, because of
smaller cluster sizes, but those channels are broken
into sectors.

114
 other disadvantages:
must design network coverage with sectoring
decided in advance
cant effectively use sectoring to increase capacity
after setting cluster size N
cant be used to gradually expand capacity as
traffic like cell splitting
More Handoffs
More antenna, more cost

115
 Next topic: Mobile Radio Propagation - Large-
scale path loss, small-scale fading, and
multipath
Free space propagation loss
Reflections
2-ray model
Diffraction
Fading
Multipath

116
IV. Improving Cellular System Capacity

A cellular design eventually (hopefully!)

becomes insufficient to support the growing
number of users.
Need to provide more channels per unit coverage
area
Would like to have orderly growth
Would like to upgrade the system instead of rebuild
Would like to use existing towers as much as
possible

1
 Cell Splitting
subdivide congested cell into several smaller
cells
increases number of times channels are reused
in an area
must decrease antenna height & Tx power
so smaller coverage per cell results
and the co-channel interference level is held
constant

2
 each smaller cell keeps same # of channels as
the larger cell, since each new smaller cell uses
the same number of frequencies
this means that we keep that same cluster size
capacity because channel reuse per unit area
smaller cells micro-cells

3
 Illustration is for towers at the corners

4
 advantages include:
only needed for cells that reach max. capacity not
all cells
implement when Pr [blocked call] > acceptable GOS
system capacity can gradually expand as demand
disadvantages include:
# handoffs/unit area increases
umbrella cell for high velocity traffic may be needed
more base stations $$ for real estate, towers, etc.

5
 complicated design process
new base stations use lower power and antenna
height
What about existing base stations?
If kept at the same power, they would overpower new
microcells.
If reduced in power, they would not cover their own
cells.
One solution: Use separate groups of channels.
One group at the original power and another group at
the lower power.
New microcells only use lower power channels.
As load growth continues, more and more channels are
moved to lower power.
6
7
8
 Sectoring
cell splitting keeps D / R unchanged (same
cluster size and CCI) but increases frequency
reuse/area
alternate way to capacity is to _____ CCI
(increase S / I ratio)

9
 replace omni-directional antennas at base station
with several directional antennas
3 sectors 3 120 antennas
6 sectors 6 60 antennas

10
 cell channels broken down into sectored groups
CCI reduced because only some of neighboring co-
channel cells radiate energy in direction of main cell
center cell labeled "5" has all co-channel cells
illustrated
only 2 co-channel cells will interfere if all are using
120 sectoring
only 1 co-channel cell would interfere when using
60 sectoring
If the S/I was 17 dB for N = 7 and n = 4, what is the
S / I now with 120 sectoring?
24.2 dB

11
12
 How is capacity increased?
sectoring only improves S/I which increases voice
quality, beyond what is really necessary
by reducing CCI, the cell system designer can choose
smaller cluster size (N ) for acceptable voice quality
smaller N greater frequency reuse larger system
capacity

What would the system capacity, Cnew, now be when

using 120 sectoring, as compared to the old capacity,
Cold ?

13
14
15
 much less costly than cell splitting
only require more antennas @ base station vs.
multiple new base stations for cell splitting
primary disadvantage is that the available
channels in a cell are subdivided into sectored
groups
trunked channel pool , therefore trunking
efficiency
There are more channels per cell, because of
smaller cluster sizes, but those channels are broken
into sectors.

16
 other disadvantages:
must design network coverage with sectoring
decided in advance
cant effectively use sectoring to increase capacity
after setting cluster size N
cant be used to gradually expand capacity as
traffic like cell splitting
More Handoffs
More antenna, more cost

17
 Next topic: Mobile Radio Propagation - Large-
scale path loss, small-scale fading, and
multipath
Free space propagation loss
Reflections
2-ray model
Diffraction
Fading
Multipath

18
Chapter 4: Large-Scale Path Loss

Chapter 4 Mobile Radio Propagation:

Large-Scale Path Loss
 This lecture: Electromagnetic propagation
properties and hindrances.

What are reasons why wireless signals are hard

to send and receive?

2
I. Problems Unique to Wireless (not wired) systems:

Paths can vary from simple line-of-sight to ones

that are severely obstructed by buildings,
mountains, and foliage.
Radio channels are extremely random and
difficult to analyze.
Interference from other service providers
out-of-band non-linear Tx emissions

3
Wireless Channel Modeling
Three main mechanisms of electromagnetic
wave propagation are:
1) Reflection
2) Diffraction
3) Scattering

4
Reflection, Diffraction & Scattering

5
Wireless Channel Modeling...
Due to these EM wave propagation
mechanisms, radio propagation can be roughly
described by three nearly independent
phenomenon :

6
 Interference from other users (same network)
CCI due to frequency reuse
ACI due to Tx/Rx design limitations & large #
users sharing finite BW
Shadowing
Obstructions to line-of-sight paths cause areas of
weak received signal strength

7
 Fading
When no clear line-of-sight path exists, signals are
received that are reflections off obstructions and
diffractions around obstructions
Multipath signals can be received that interfere with
each other
Fixed Wireless Channel random & unpredictable
must be characterized in a statistical fashion
field measurements often needed to characterize radio
channel performance

8
 ** The Mobile Radio Channel (MRC) has
unique problems that limit performance **
A mobile Rx in motion influences rates of
fading
the faster a mobile moves, the more quickly
characteristics change

9
II. Radio Signal Propagation

10
 The smoothed line is the average signal
strength. The actual is the more jagged line.
Actual received signal strength can vary by
more than 20 dB over a few centimeters.
The average signal strength decays with
distance from the transmitter, and depends on
terrain and obstructions.

11
 Two basic goals of propagation modeling:
1) Predict magnitude and rate (speed) of received
signal strength fluctuations over short
distances/time durations
short typically a few wavelengths () or
seconds
at 1 Ghz, = c/f = 3x108 / 1x109 = 0.3 meters
received signal strength can vary drastically by 30
to 40 dB

12
 small-scale fluctuations called _____ (Chapter 5)
caused by received signal coming from a sum of
many signals coming together at a receiver
multiple signals come from reflections and
scattering
these signals can destructively add together by being
out-of-phase

13
2) Predict average received signal strength for
given Tx/Rx separation
characterize received signal strength over distances
from 20 m to 20 km
Large-scale radio wave propagation model models
needed to estimate coverage area of base station
in general, large scale path loss decays gradually
with distance from the transmitter
will also be affected by geographical features like
hills and buildings

14
Propagation Models
Large scale propagation models
- To predict the average received signal strength over
large T-R separation distances (several hundreds or
thousands of meters).
Typically, the local average received power is computed
by averaging signal measurements over a measurement
track of 5 to 40 wavelengths.
Small scale propagation models (or fading models)
To characterize the rapid fluctuations of the receiver
signal strength over very short distances ( a few
wavelengths) or short time durations (on the order of
seconds).
In small scale fading, the received signal power may
vary by as much as three to four orders of magnitude
(30 to 40 dB).

15
 Free-Space Signal Propagation
clear, unobstructed line-of-sight path satellite and
fixed microwave
Friis transmission formula Rx power (Pr) vs. T-R
separation (d)

16
 where
Pt = Tx power (W)
G = Tx or Rx antenna gain (unitless)
relative to isotropic source (ideal antenna which
radiates power uniformly in all directions)
in the __________ of an antenna (beyond a few meters)
Effective Isotropic Radiated Power (EIRP)
EIRP = PtGt
Represents the max. radiated power available
from a Tx in the direction of max. antenna gain,
as compare to an isotropic radiator

17
 = wavelength = c / f (m). A term is related to
antenna gain.
So, as frequency increases, what happens to the
propagation characteristics?

L = system losses (antennas, transmission lines

between equipment and antennas, atmosphere, etc.)
unitless
L = 1 for zero loss
L > 1 in general

18
 d = T-R separation distance (m)
Signal fades in proportion to d2
We can view signal strength as related to the
density of the signal across a large sphere.
This is the surface area of a sphere with radius d.
So, a term in the denominator is related to distance
and density of surface area across a sphere.

19
 Path Loss (PL) in dB:

20
 d2 power law relationship
Pr decreases at rate of proportional to d2
Pr decreases at rate of 20 dB/decade (for line-of-
sight, even worse for other cases)
For example, path loses 20 dB from 100 m to 1 km
Comes from the d2 relationship for surface area.

Note: Negative loss = gain

21
 Example:
Path loss can be computed in terms of a link budget
calculation.
Compute path loss as a sum of dB terms for the
following:
Unity gain transmission antenna.
Unity gain receiving antenna.
No system losses
Carrier frequency of 3 GHz
Distance = 2000 meters

22
 Close in reference point (do) is used in large-scale models

do : known received power reference point - typically 100 m or

1 km for outdoor systems and 1 m for indoor systems
df : far-field distance of antenna, we will always work problems
in the far-field

2D2
df = df D df

D: the largest physical linear dimension of antenna

23
 Reference Point Example:
Given the following system characteristics for large-
scale propagation, find the reference distance do.
Received power at do = 20 W
Received power at 5 km = 13 dBm

Using Watts:

Using dBm:

24
III. Reflections

There are three basic propagation mechanisms

in addition to line-of-sight paths
Reflection - Waves bouncing off of objects of large
dimensions
Diffraction - Waves bending around sharp edges of
objects
Scattering - Waves traveling through a medium
with small objects in it (foliage, street signs, lamp
posts, etc.) or reflecting off rough surfaces

25
 Reflection occurs when RF energy is incident upon
a boundary between two materials (e.g air/ground)
with different electrical characteristics
Permittivity
Permeability
Conductance
Reflecting surface must be large relative to of RF
energy
Reflecting surface must be smooth relative to of
RF energy
specular reflection

26
 What are important reflecting surfaces for
mobile radio?

Fresnel reflection coefficient

describes the magnitude of reflected RF energy
depends upon material properties, polarization, &
angle of incidence

27
IV. Ground Reflection (2-Ray) Model
Good for systems that use tall towers (over 50 m
tall)
Good for line-of-sight microcell systems in urban
environments

28
 ETOT is the electric field that results from a combination of a
direct line-of-sight path and a ground reflected path

is the amplitude of the electric field at distance d

c = 2fc where fc is the carrier frequency of the signal
Notice at different distances d the wave is at a different phase
because of the form similar to

29
 For the direct path let d = d ; for the reflected path
d = d then

for large TR separation : i goes to 0 (angle of incidence

to the ground of the reflected wave) and
= 1
Phase difference can occur depending on the phase
difference between direct and reflected E fields
The phase difference is due to Path difference ,
= d d, between

30
 From two triangles with sides d and (ht + hr) or (ht hr)

31
 can be expanded using a Taylor series
expansion

32
 which works well for d >> (ht + hr), which means
and are small

33
 the phase difference between the two arriving
signals is

E0 d 0
ETOT (t ) = 2 sin
d 2
2 hr ht
< 0.3 rad
2 d
E0 d 0 2 hr ht k
ETOT (t ) 2 2 V/m
d d d

34
 For d0=100meter, E0=1, fc=1 GHz, ht=50 meters, hr=1.5 meters, at t=0

35
 note that the magnitude is with respect to a
reference of E0=1 at d0=100 meters, so near 100
meters the signal can be stronger than E0=1
the second ray adds in energy that would have been
lost otherwise
for large distances it can be shown
that

36
37
V. Diffraction

RF energy can propagate:

around the curved surface of the Earth
beyond the line-of-sight horizon
Behind obstructions
Although EM field strength decays rapidly as
Rx moves deeper into shadowed or
obstructed (OBS) region
The diffraction field often has sufficient
strength to produce a useful signal

38
 Huygens principle says points on a wavefront
can be considered sources for additional
wavelets.

39
 The wavefront on top of an obstruction generates
secondary (weaker) waves.

40
 The difference between the direct path and
diffracted path, call excess path length

Fresnel-Kirchoff diffraction parameter

The corresponding phase difference

41
42
 The excess total path length traversed by a ray
passing through each circle is n/2

43
 The diffraction gain due to the presence of a knife
edge, as compared the the free space E-field

44
45
46
47
48
49
50
51
VI. Scattering
Received signal strength is often stronger than that
predicted by reflection/diffraction models alone
The EM wave incident upon a rough or complex
surface is scattered in many directions and provides
more energy at a receiver
energy that would have been absorbed is instead reflected to
the Rx.
Scattering is caused by trees, lamp posts, towers, etc.
flat surface EM reflection (one direction)
rough surface EM scattering (many directions)

52
53
VII. Path Loss Models

We wish to predict large scale coverage using

analytical and empirical (field data) methods
It has been repeatedly measured and found that
Pr @ Rx decreases logarithmically with
distance

PL (d) = (d / do )n where n : path loss exponent or

PL (dB) = PL (do ) + 10 n log (d / do )

54
 bar means the average of many PL values at a
given value of d (T-R sep.)
n depends on the propagation environment
typical values based on measured data

55
 At any specific d the measured values vary
drastically because of variations in the
surrounding environment (obstructed vs. line-
of-sight, scattering, reflections, etc.)

Some models can be used to describe a

situation generally, but specific circumstances
may need to be considered with detailed
analysis and measurements.

56
 Log-Normal Shadowing
PL (d) = PL (do ) + 10 n log (d / do ) + X
describes how the path loss at any specific location may vary
from the average value

has a the large-scale path loss component we have already

seen plus a random amount X.

57
 X : zero mean Gaussian random variable, a bell curve
is the standard deviation that provides the second
parameter for the distribution
takes into account received signal strength variations
due to shadowing
measurements verify this distribution
n & are computed from measured data for different
area types
any other path loss models are given in your book.
That correlate field measurements with models for different
types of environments.

58
59
 HW-3
3-16, 3-17, 4-4, 4-14, 4-23(a)-(d)

60
Lecture 6 Fading

Chapter 5 Mobile Radio Propagation:

Small-Scale Fading and Multipath

1
Last lecture

Large scale propagation properties of wireless

systems - slowly varying properties that depend
primarily on the distance between Tx and Rx.
Free space path loss
Power decay with respect to a reference point
The two-ray model
General characterization of systems using the path
loss exponent.
Diffraction
Scattering
This lecture: Rapidly changing signal
characteristics primarily caused by movement
and multipath.
2
I. Fading

Fading: rapid fluctuations of received signal strength

over short time intervals and/or travel distances
Caused by interference from multiple copies of Tx
signal arriving @ Rx at slightly different times
Three most important effects:
1. Rapid changes in signal strengths over small travel
distances or short time periods.
2. Changes in the frequency of signals.
3. Multiple signals arriving a different times. When added
together at the antenna, signals are spread out in time.
This can cause a smearing of the signal and interference
between bits that are received.

3
 Fading signals occur due to reflections from
ground & surrounding buildings (clutter) as
well as scattered signals from trees, people,
towers, etc.
often an LOS path is not available so the first
multipath signal arrival is probably the desired
signal (the one which traveled the shortest distance)
allows service even when Rx is severely obstructed
by surrounding clutter

4
 Even stationary Tx/Rx wireless links can
experience fading due to the motion of objects
(cars, people, trees, etc.) in surrounding
environment off of which come the reflections
Multipath signals have randomly distributed
amplitudes, phases, & direction of arrival
vector summation of (A ) @ Rx of multipath
leads to constructive/destructive interference as
mobile Rx moves in space with respect to time

5
 received signal strength can vary by Small-scale fading
over distances of a few meter (about 7 cm at 1 GHz)!
This is a variation between, say, 1 mW and 10-6 mW.
If a user stops at a deeply faded point, the signal quality
can be quite bad.
However, even if a user stops, others around may still
be moving and can change the fading characteristics.
And if we have another antenna, say only 7 to 10 cm
separated from the other antenna, that signal could be
good.
This is called making use of ________ which we
will study in Chapter 7.

6
 fading occurs around received signal strength predicted
from large-scale path loss models

7
II. Physical Factors Influencing Fading in Mobile Radio Channel (MRC)

1) Multipath Propagation
# and strength of multipath signals
time delay of signal arrival
large path length differences large differences in
delay between signals
urban area w/ many buildings distributed over large
spatial scale
large # of strong multipath signals with only a few
having a large time delay
suburb with nearby office park or shopping mall
moderate # of strong multipath signals with small to
moderate delay times
rural few multipath signals (LOS + ground
reflection)
8
2) Speed of Mobile
relative motion between base station & mobile
causes random frequency modulation due to
Doppler shift (fd)
Different multipath components may have different
frequency shifts.
3) Speed of Surrounding Objects
also influence Doppler shifts on multipath signals
dominates small-scale fading if speed of objects >
mobile speed
otherwise ignored

9
4) Tx signal bandwidth (Bs)
The mobile radio channel (MRC) is modeled as
filter w/ specific bandwidth (BW)
The relationship between the signal BW & the
MRC BW will affect fading rates and distortion,
and so will determine:
a) if small-scale fading is significant
b) if time distortion of signal leads to inter-symbol
interference (ISI)
An MRC can cause distortion/ISI or small-scale
fading, or both.
But typically one or the other.

10
Doppler Shift
motion causes frequency modulation due to Doppler
shift (fd)

v : velocity (m/s)
: wavelength (m)
: angle between
mobile direction
and arrival direction of RF energy
+ shift mobile moving toward S
shift mobile moving away from S
11
 Two Doppler shifts to consider above
1. The Doppler shift of the signal when it is received at
the car.
2. The Doppler shift of the signal when it bounces off
the car and is received somewhere else.
Multipath signals will have different fds for
constant v because of random arrival directions!!

12
 Example 5.1, page 180
Carrier frequency = 1850 MHz
Vehicle moving 60 mph
Compute frequency deviation in the following
situations.
(a) Moving directly toward the transmitter

(b) Moving perpendicular to the transmitter

13
 Note: What matters with Doppler shift is not
the absolute frequency, but the shift in
frequency relative to the bandwidth of a
channel.
For example: A shift of 166 Hz may be significant
for a channel with a 1 kHz bandwidth.
In general, low bit rate (low bandwidth) channels
are affected by Doppler shift.

14
 Wireless Communication Channels

From Wireless Communications Edfors, Molisch, Tufvesson

Communications over wireless channels suffer from multi-path propagation

Multi-path channels are usually frequency selective
OFDM supports high data rate communications over frequency selective
channels
 Multi-Path Propagation Modeling

Power

Multi-Path
Components

0 1 2 Time

Multi-path results from reflection, diffraction, and scattering off environment

surroundings
Note: The figure above demonstrates the roles of reflection and scattering only on multi-
path
 Multi-Path Propagation Modeling

Power

Multi-Path
Components

0 1 2 Time

As the mobile receiver (i.e. car) moves in the environment, the strength
of each multi-path component varies
 Multi-Path Propagation Modeling

Power

Multi-Path
Components

0 1 2 Time

As the mobile receiver (i.e. car) moves in the environment, the strength
of each multi-path component varies
ISI

19
IV. Multipath Channel Parameters

Derived from multipath power delay profiles

(Eq. 5-18)
P (k) : relative power amplitudes of multipath
signals (absolute measurements are not needed)
Relative to the first detectable signal arriving at
the Rx at 0
use ensemble average of many profiles in a
small localized area typically 2 6 m spacing
of measurements to obtain average small-
scale response

20
21
 Time Dispersion Parameters
excess delay : all values computed relative to the
time of first signal arrival o

mean excess delay

RMS delay spread

where Avg( 2) is the same computation as above as
used for except that
A simple way to explain this is the range of time
within which most of the delayed signals arrive

22
 outdoor channel ~ on the order of microseconds
indoor channel ~ on the order of nanoseconds
23
 maximum excess delay ( X): the largest time where the
multipath power levels are still within X dB of the
maximum power level
worst case delay value
depends very much on the choice of the noise threshold

24
 and provide a measure of propagation delay
of interfering signals

They give an indication of how time smearing

might occur for the signal.
A small is desired.
The noise threshold is used to differentiate between
received multipath components and thermal noise

25
26
27
 Coherence BW (Bc) and Delay Spread ( )
This image cannot currently be displayed.

The Fourier Transform of multipath delay shows

frequency (spectral) characteristics of the MRC
Bc : statistical measure of frequency range where MRC
response is flat
MRC response is flat = passes all frequencies with
equal gain & linear phase
amplitudes of different frequency components are
correlated
if two sinusoids have frequency separation greater
than Bc, they are affected quite differently by the
channel

28
 amplitude correlation multipath signals have
close to the same amplitude if they are then
out-of-phase they have significant destructive
interference with each other (deep fades)
so a flat fading channel is both good and bad
Good: The MRC is like a bandpass filter and
passes signals without major attenuation
from the channel.
Bad: Deep fading can occur.

29
 so the coherence bandwidth is the range
of frequencies over which two frequency
components have a strong potential for
amplitude correlation. (quote from
textbook)

30
 estimates
0.9 correlation Bc 1 / 50 (signals are 90%
correlated with each other)
0.5 correlation Bc 1 / 5 Which has a larger
bandwidth and why?

specific channels require detailed analysis for a

particular transmitted signal these are just rough
estimates

31
 A channel that is not a flat fading channel is
called frequency selective fading because
different frequencies within a signal are
attenuated differently by the MRC.

Note: The definition of flat or frequency selective

fading is defined with respect to the bandwidth of
the signal that is being transmitted.

32
 Bc and are related quantities that
characterize time-varying nature of the MRC
for multipath interference from frequency &
time domain perspectives

33
34
35
 these parameters do NOT characterize the time-varying
nature of the MRC due to the mobility of the mobile
and/or surrounding objects

that is to say, Bc and characterize the statics, (how

multipath signals are formed from scattering/reflections and
travel different distances)
Bc and do not characterize the mobility of the Tx or Rx.

36
 Doppler Spread (BD) & Coherence Time (Tc)

BD : measure of spectral broadening of the Tx

signal caused by motion i.e., Doppler shift
BD = max Doppler shift = fmax = vmax /
In what direction does movement occur to create this
worst case?
if Tx signal bandwidth (Bs) is large such that Bs >> BD
then effects of Doppler spread are NOT important so
Doppler spread is only important for low bps (data rate)
applications (e.g. paging)

37
 Tc : statistical measure of the time interval over
which MRC impulse response remains
invariant amplitude & phase of multipath
signals constant

Coherence Time (Tc) = passes all received signals

with virtually the same characteristics because the
channel has not changed
time duration over which two received signals have
a strong potential for amplitude correlation

38
 Two signals arriving with a time separation
greater than Tc are affected differently by the
channel, since the channel has changed within
the time interval
For digital communications coherence time and
Doppler spread are related by

9 0.423
=Tc =
16 f m2 fm

39
V. Types of Small-Scale Fading
Fading can be caused by two independent MRC
propagation mechanisms:
1) time dispersion multipath delay (Bc , )
2) frequency dispersion Doppler spread (BD , Tc)
Important digital Tx signal parameters symbol
period & signal BW

40
 A pulse can be more than two levels, however,
so each period would be called a "symbol
period".
We send 0 (say +1 Volt) or 1 (say -1 Volt) one
bit per symbol
Or we could send 10 (+3 Volts) or 00 (+1 Volt) or
01 (-1 Volt) or 11 (-3 Volts) two bits per
symbol

41
illustrates types of small-scale fading

42
 This image cannot currently be displayed.

1) Fading due to Multipath Delay

AFlat Fading Bs << Bc or Ts >>
Ts 10
signal fits easily within the bandwidth of the channel
channel BW >> signal BW

most commonly occurring type of fading

43
 spectral properties of Tx signal are preserved
signal is called a narrowband channel, since the
bandwidth of the signal is narrow with respect to the
channel bandwidth
signal is not distorted

What does Ts >> mean??

all multipath signals arrive at mobile Rx during 1 symbol
period
Little intersymbol interference occurs (no multipath
components arrive late to interfere with the next symbol)

44
45
 flat fading is generally considered desirable
Even though fading in amplitude occurs, the signal
is not distorted
Forward link can increase mobile Rx gain
(automatic gain control)
Reverse link can increase mobile Tx power
(power control)
Can use diversity techniques (described in a later
lecture)

46
B) Frequency Selective Fading Bs > Bc or Ts <

Ts 10
Bs > Bc certain frequency components of the signal
are attenuated much more than others

47
48
 Ts < delayed versions of Tx signal arrive
during different symbol periods
e.g. receiving an LOS 1 & multipath 0 (from
prior symbol!)
This results in intersymbol interference (ISI)
Undesirable

it is very difficult to predict mobile Rx

performance with frequency selective channels

49
 But for high bandwidth applications, channels with
likely be frequency selective
a new modulation approach has been developed to
combat this.
Called OFDM

One aspect of OFDM is that it separates a

wideband signal into many smaller narrowband
signals
Then adaptively adjusts the power of each narrowband
signal to fit the characteristics of the channel at that
frequency.
Results in much improvement over other wideband
transmission approaches (like CDMA).

50
 OFDM is used in the new 802.11g 54 Mbps
standard for WLANs in the 2.4 GHz band.
Previously it was thought 54 Mbps could only be
obtained at 5.8 GHz using CDMA, but 5.8 GHz
signals attenuate much more quickly.
Signals are split using signal FFT, break into
pieces in the frequency domain, use inverse FFT to
create individual signals from each piece, then
transmit.

51
2) Fading due to Doppler Spread
Caused by motion of Tx and Rx and reflection
sources.
A) Fast Fading Bs < BD or Ts > Tc
Bs < BD
Doppler shifts significantly alter spectral BW of TX
signal
signal spreading
Ts > Tc
MRC changes within 1 symbol period
rapid amplitude fluctuations
uncommon in most digital communication systems
52
B) Slow Fading Ts << Tc or Bs >> BD
MRC constant over many symbol periods
slow amplitude fluctuations
for v = 60 mph @ fc = 2 GHz BD = 178 Hz
Bs 2 kHz >> BD
Bs almost always >> BD for most applications

** NOTE: Typically use a factor of 10 to

designate >> **

53
54
Lecture 7: Modulation I

Chapter 6 Modulation
Techniques for Mobile Radio

1
II. Digital Modulation

Better performance and more cost effective

than analog modulation methods (AM, FM, etc.)
Used in modern cellular systems
Advancements in VLSI, DSP, etc. have made
digital solutions practical and affordable

2
 Performance advantages:
1) Resistant to noise, fading, & interference
2) Can combine multiple information types (voice, data,
& video) in a single transmission channel
3) Improved security (e.g., encryption) deters phone
cloning + eavesdropping
4) Error coding is used to detect/correct transmission
errors
5) Signal conditioning can be used to combat hostile
MRC environment
6) Can implement mod/dem functions using DSP
software (instead of hardware circuits).
3
 Choice of digital modulation scheme
Many types of digital modulation methods subtle
differences
Performance factors to consider
1) low Bit Error Rate (BER) at low S/N
2) resistance to interference (ACI & CCI) & multipath
fading
3) occupying a minimum amount of BW
4) easy and cheap to implement in mobile unit
5) efficient use of battery power in mobile unit

4
 No existing modulation scheme simultaneously
satisfies all of these requirements well.
Each one is better in some areas with tradeoffs
of being worse in others.

5
 Power Efficiency p : ability of a modulation
technique to preserve the quality of digital messages at
low power levels (low SNR)
Specified as Eb / No @ some BER (e.g. 10-5) where Eb :
energy/bit and No : noise power/bit
Tradeoff between fidelity and signal power
BER as Eb / No

6
 Bandwidth Efficiency B : ability of a modulation
technique to accommodate data in a limited BW
R
B = bps/Hz R : data rate B: RF BW
B

Tradeoff between data rate and occupied BW

as R , then BW
For a digital signal :

7
 each pulse or symbol having m finite states
represents n = log2 m bits/symbol
e.g. m = 0 or 1 (2 states) 1 bit/symbol (binary)
e.g. m = 0, 1, 2, 3, 4, 5, 6, or 7 (8 states) 3
bits/symbol

8
 Implementation example: A system is changed
from binary to 2-ary.
Before: "0" = - 1 Volt, "1" = 1 Volt
Now
"0" = - 1 Volt, "1" = - 0.33 volts, "2" = 0.33 Volts, "3" = 1 Volt

What would be the new data rate compared to the old data
rate if the symbol period where kept constant?

In general, called M-ary keying

9
 Maximum BW efficiency Shannons Theorem
Most famous result in communication theory.

where

B : RF BW
C : channel capacity (bps) of real data (not retransmissions
or errors)
To produce error-free transmission, some of the bit rate
will be taken up using retransmissions or extra bits for
error control purposes.
As noise power N increases, the bit rate would still be the
same, but max Bmax decreases.

10
 So

note that C B (expected) but also C S / N

an increase in signal power translates to an increase in
channel capacity
lower bit error rates from higher power more real
data
large S / N easier to differentiate between multiple
signal states (m) in one symbol n
max B is fundamental limit that cannot be
max

achieved in practice

11
 People try to find schemes that correct for errors.

People are starting to refer to certain types of codes

as capacity approaching codes, since they say
they are getting close to obtaining Cmax.
More on this in the chapter on error control.

12
13
14
 Fundamental tradeoff between B and p (in general)
If B improves then p deteriorates (or vice versa)
May need to waste more power to get a better data rate.
May need to use less power (to save on battery life) at the
expense of a lower data rate.
p vs. B is not the only consideration.
Use other factors to evaluate complexity, resistance to
MRC impairments, etc.

15
 Bandwidth Specifications
Many definitions depending on application all use
Power Spectral Density (PSD) of modulated bandpass
signal

W (f)2
SW ( f ) = lim T
T T

Many signals (like square pulses) have some power at all

frequencies.

16
 B : half-power (-3 dB) BW
B : null-to-null BW
B : absolute BW
range where PSD > 0
FCC definition of occupied BW BW contains
99% of signal power
17
III. Geometric Representation of Modulation Signal

Geometric Representation of Modulation Signals -

Constellation Diagrams
Graphical representation of complex ( A & )
digital modulation types
Provide insight into modulation performance
Modulation set, S, with M possible signals

Binary modulation M = 2 each signal = 1 bit of

information
M-ary modulation M > 2 each signal > 1 bit of
information
18
 Example: Binary Phase Shift Keying (BPSK)

19
 Phase change between bits Phase shifts of 180
for each bit.
Note that this can also be viewed as AM with +/-
amplitude changes

Dimension of the vector space is the # of basis

signals required to represent S.

20
 Plot amplitude & phase of S in vector space :

21
 Constellation diagram properties :
1) Distance between signals is related to differences in
modulation waveforms
Large distance sparse easy to discriminate
good BER @ low SNR (Eb / No )
From above, noise of -2 added to would make
the received signal look like s2(t) error.
From , noise of > - would make the result
closer to - and would make the decoder choose
s2(t) error.
Above example is Power Efficient (related to density
with respect to # states/N)
22
2) Occupied BW as # signal states
If we can represent more bits per symbol, then we
need less BW for a given data rate.
Small separation dense more signal
states/symbol more information/Hz !!
Bandwidth Efficient

23
IV. Linear Modulation Methods
In linear modulation techniques, the amplitude of the
transmitted signal varies linearly with the modulating
digital signal.
Performance is evaluated with respect to Eb / No

24
BPSK
BPSK Binary Phase Shift Keying

25
 Phase transitions force carrier amplitude to
change from + to .
Amplitude varies in time

26
BPSK RF signal BW

Null-to-null RF BW = 2 Rb = 2 / Tb
90% BW = 1.6 Rb for rectangular pulses
27
 Probability of Bit Error is proportional to the
distance between the closest points in the
constellation.
A simple upper bound can be found using the
assumption that noise is additive, white, and
Gaussian.

d is distance between nearest constellation points.

28
 Q(x) is the Q-function, the area under a normalized Gaussian
function (also called a Normal curve or a bell curve)

1 y2 / 2
Q( z ) = e dy
z 2

Appendix F, Fig. F.1

Fig. F.2, plot of Q-function
Tabulated values in Table F.1.

Here

29
 Demodulation in Rx
Requires reference of Tx signal in order to properly
determine phase
carrier must be transmitted along with signal
Called Synchronous or Coherent detection
complex & costly Rx circuitry
good BER performance for low SNR power efficient

30
31
DPSK

DPSK Differential Phase Shift Keying

Non-coherent Rx can be used
easy & cheap to build
no need for coherent reference signal from Tx
Bit information determined by transition between
two phase states
incoming bit = 1 signal phase stays the same as
previous bit
incoming bit = 0 phase switches state

32
 If {mk} is the message, the output {dk} is as shown
below.

can also be described in modulo-2 arithmetic d=

k mk d k 1
Same BW properties as BPSK, uses same amount of spectrum
Non-coherent detection all that is needed is to compare
phases between successive bits, not in reference to a Tx phase.
power efficiency is 3 dB worse than coherent BPSK (higher
power in Eb / No is required for the same BER)
33
34
 QPSK

QPSK Quadrature Phase Shift Keying

Four different phase states in one symbol period
Two bits of information in each symbol
Phase: 0 /2 3/2 possible phase values
Symbol: 00 01 11 10

35
 Note that we choose binary representations so
an error between two adjacent points in the
constellation only results in a single bit error
For example, decoding a phase to be instead of
/2 will result in a "11" when it should have been
"01", only one bit in error.

36
 Constant amplitude with four different phases
remembering the trig. identity

37
38
 Now we have two basis functions
Es = 2 Eb since 2 bits are transmitted per symbol
I = in-phase component from sI(t).
Q = quadrature component that is sQ(t).

39
QPSK RF Signal BW

null-to-null RF BW = Rb = 2RS (2 bits / one symbol time) = 2 / Ts

double the BW efficiency of BPSK or twice the data rate in same
signal BW
40
 BER is once again related to the distance
between constellation points.

d is distance between nearest constellation points.

41
42
 How does BER performance compare to BPSK?

Why? same # of states per number of basis

functions for both BPSK and QPSK (2 states per
one function or 4 states per 2 functions)
same power efficiency
(same BER at specified Eb / No)
twice the bandwidth efficiency
(sending 2 bits instead of 1)

43
 QPSK Transmission and Detection Techniques

44
45
OQPSK
Offset QPSK
The occasional phase shift of radians can cause the
signal envelope to pass through zero for just in
instant.
Any kind of hard limiting or nonlinear amplification
of the zero-crossings brings back the filtered
sidelobes
since the fidelity of the signal at small voltage levels is
lost in transmission.
OQPSK ensures there are fewer baseband signal
transitions applied to the RF amplifier,
helps eliminate spectrum regrowth after amplification.

46
 Example above: First symbol (00) at 0, and the
next symbol (11) is at 180. Notice the signal going
through zero at 2 microseconds.
This causes problems.

47
 Using an offset approach: First symbol (00) at 0,
then an intermediate symbol at (10) at 90, then the
next full symbol (11) at 180.
The intermediate symbol is used halfway through
the symbol period.
It corresponds to allowing the first bit of the symbol
to change halfway through the symbol period.
The figure below does have phase changes more
often, but no extra transitions through zero.
IS-95 uses OQPSK, so it is one of the major
modulation schemes used.

48
49
 In QPSK signaling, the bit transitions of the
even and odd bit streams occur at the same time
instants.
but in OQPSK signaling, the even and odd bit
Streams, mI(t) and mQ(t), are offset in their
relative alignment by one bit period (half-
symbol period)

50
 the maximum phase shift of the transmitted signal
at any given time is limited to 90o

51
 The spectrum of an OQPSK signal is identical
to that of a QPSK signal, hence both signals
occupy the same bandwidth

52
/4 QPSK
/4 QPSK
The /4 shifted QPSK modulation is a quadrature
phase shift keying technique
offers a compromise between OQPSK and QPSK in
terms of the allowed maximum phase transitions.
It may be demodulated in a coherent or noncoherent
fashion.
greatly simplifies receiver design.
In /4 QPSK, the maximum phase change is limited
to 135o
in the presence of multipath spread and fading, /4
QPSK performs better than OQPSK

53
54
55
56
Lecture 9: Modulation

Chapter 6 Modulation
Techniques for Mobile Radio

1
2
Baseband vs Passband Transmission
Baseband signals:
Voice (0-4kHz)
TV (0-6 MHz)
A signal may be sent in
its baseband format
when a dedicated wired
channel is available.
Otherwise, it must be
converted to passband.
Modulation: What and Why?
The process of shifting the baseband signal to
passband range is called Modulation.
The process of shifting the passband signal to
baseband frequency range is called
Demodulation.
Reasons for modulation:
Simultaneous transmission of several signals
Practical Design of Antennas
Exchange of power and bandwidth
Types of (Carrier) Modulation
In modulation, one characteristic of a signal
(generally a sinusoidal wave) known as the
carrier is changed based on the information
signal that we wish to transmit (modulating
signal).
That could be the amplitude, phase, or frequency,
which result in Amplitude modulation (AM),
Phase modulation (PM), or Frequency
modulation (FM). The last two are combined as
Angle Modulation
Types of Amplitude Modulation (AM)
Double Sideband with carrier (we will call it AM):
This is the most widely used type of AM modulation.
In fact, all radio channels in the AM band use this type
of modulation.
Double Sideband Suppressed Carrier (DSBSC):
This is the same as the AM modulation above but
without the carrier.
Single Sideband (SSB): In this modulation, only half
of the signal of the DSBSC is used.
Vestigial Sideband (VSB): This is a modification of
the SSB to ease the generation and reception of the
signal.
Double Sideband Suppressed Carrier (DSBSC)
Assume that we have a message signal m(t) with
bandwidth 2 B rad/s (or B Hz). m(t) M().

Let c(t) be a carrier signal, c(t) = cos(ct), c >> 2B

gDSBSC (t) = m(t)cos(ct)

(1/2) [M( c) + M( + c)].

m(t) X gDSBSC(t)

c(t)

DSBSC Modulator (transmitter)

Time and Frequency Representation of DSBSC
Modulation Process
DSBSC Demodulation
e(t) HLPF()
gDSBSC(t) X BW = 2B
f(t)

e (t)=gDSBSC(t)cos(ct)
c(t)

= m(t)cos2(ct) DSBSC Demodulator (receiver)

= (1/2) m(t) [1 + cos(2ct)]

= (1/2) m(t) + (1/2) m(t) cos(2 ct)
E() (1/2) M() + (1/4) [M( 2 c) + M( + 2 c)].
The output signal f(t) of the LPF will be
f (t) = (1/2) m(t) (1/2) M().
Time and Frequency Representation of DSBSC
Demodulation Process
Demodulation of DSBSC
The modulator circuits can be used for demodulation, but
replacing the BPF by a LPF of bandwidth B Hz.
The receiver must generate a carrier frequency in phase
and frequency synchronization with the incoming carrier.
This type of demodulation is therefore called coherent
demodulation (or detection).
e(t) HLPF()
gDSBSC(t) X BW = 2B
f(t)

c(t)

DSBSC Demodulator (receiver)

From DSBSC to DSBWC (AM)
Carrier recovery circuits, which are required for
the operation of coherent demodulation, are
sophisticated and could be quite costly.
If we can let m(t) be the envelope of the
modulated signal, then a much simpler circuit,
the envelope detector, can be used for
demodulation (non-coherent demodulation).
How can we make m(t) be the envelope of the
modulated signal?
Definition of AM
Shift m(t) by some DC value A
such that A+m(t) 0. Or A mpeak

g AM (t ) = [ A + m(t )] cos(C t )
= A cos(C t ) + m(t ) cos(C t )
Called DSBWC. Here will refer to
it as Full AM, or simply AM
Modulation index = mp /A.
01
Spectrum of AM

g AM (t ) A[ ( C ) + ( + C )] + [M ( C ) + M ( + C )]
1
2
The Buy and Price of AM
Buy: Simplicity in demodulation.
Price: Waste in Power
gAM(t) = Acosct + m(t) cosct
Carrier Power Pc = A2/2 (carries no information)
Sideband Power Ps = Pm/2 (useful)
Power efficiency = = Ps/(Pc + Ps)= Pm/(A2 +Pm)
 Tone Modulation
m(t) = Bcos(mt)
g(t)=[A+ Bcos(mt)] cosct = A[1+ cos(mt)] cosct
= (B2/2)/(B2/2 + A2) = 2/(2+2)
Under best conditions, =1 max =1/3 =33%
For = 0.5, = 11.11%
For practical signals, < 25%

? Would you use AM or DSBSC?

Single-Side Band (SSB) Modulation
DSBSC (as well as AM) occupies double the
bandwidth of the baseband signal, although the two
sides carry the same information.
Why not send only one side, the upper or the lower?
Modulation: similar to DSBSC. Only change the
settings of the BPF (center frequency, bandwidth).
Demodulation: similar to DSBSC (coherent)
SSB Representation
How would we
represent the SSB signal
in the time domain?
gUSB(t) = ?
gLSB(t) = ?
Time-Domain Representation of SSB (1/2)
M() = M+() + M-()
Let m+(t)M+() and m-(t)M-()
Then: m(t) = m+(t) + m-(t) [linearity]
Because M+(), M-() are not even
m+(t), m-(t) are complex.
Since their sum is real they must be
conjugates.
m+(t) = [m(t) + j mh(t)]
m-(t) = [m(t) - j mh(t)]
What is mh(t) ?
Time-Domain Representation of SSB (2/2)
M() = M+() + M-()
M+() = M()u(); M-() = M()u(-)
sgn()=2u() -1 u()= + sgn(); u(-) = - sgn()
M+() = [ M() + M()sgn()]
M-() = [M() - M()sgn()]
Comparing to:
m+(t) = [m(t) + j mh(t)] [M() + j Mh()]
m-(t) = [m(t) - j mh(t)] [M() - j Mh()]
We find
Mh() = - j M()sgn() where mh(t)Mh()
Hilbert Transform
mh(t) is known as the Hilbert Transform (HT) of m(t).
The transfer function of this transform is given by:
H() = -j sgn() |H()| = 1

H() = jsgn()

()=
(
/2
) )
sgn(
j /2

It is basically a /2 phase shifter
/2
 Hilbert Transform of cos(ct)
cos(ct) [( c) + ( + c)]
HT[cos(ct)] -j sgn() [( c) + ( + c)]
= j sgn() [( c) ( + c)]
= j [( c) + ( + c)]
= j [( + c) - ( - c)] sin(ct)

Which is expected since:

cos(ct-/2) = sin(ct)
Time-Domain Operation for Hilbert
Transformation
For Hilbert Transformation H() = -j sgn().
What is h(t)?
sgn(t) 2/(j) [From FT table]
2/(jt) 2 sgn(-) [symmetry]
1/( t) -j sgn()
Since Mh() = - j M()sgn() = H() M()
1
Then mh (t ) = * m(t )
t

1 m( )
= d
t
Finally
gUSB (t ) = m+ (t )e jC t + m (t )e jC t
g LSB (t ) = m+ (t )e jC t + m (t )e jC t
1 1
gUSB (t ) = m(t )e jC t + jmh (t )e jC t
2 2
1 1
+ m(t )e jC t jmh (t )e jC t
2 2
= m(t ) cos(C t ) mh (t ) sin(C t )
1 1
g LSB (t ) = m(t )e jC t jmh (t )e jC t
2 2
1 1
+ m(t )e jC t + jmh (t )e jC t GUSB ( ) = M + ( C ) + M ( + C )
2 2
= m(t ) cos(C t ) + mh (t ) sin(C t ) GLSB ( ) = M + ( + C ) + M ( C )
Generation of SSB
Selective Filtering Method
Realization based on spectrum analysis
Phase-Shift Method
Realization based on time-domain expression
of the modulated signal
 M()

Selective Filtering
C 2B +2B C

GDSBSC()

USB LSB LSB USB

C2B C C+2B C2B C C+2B

M() (an example of an

HUSB()
audio signal) Guard Band
BW = 2B (B Hz)
of 600 Hz Center Freq = c+B

C2B C C+2B C2B C C+2B

GUSB()
5000 Hz 300 Hz 300 Hz 5000 Hz
USB USB

C2B C C C+2B

HLSB()
BW = 2B (B Hz)
Center Freq = c B

C2B C C+2B C2B C C+2B

GLSB()

LSB LSB

C C+2B C2B C
Phase Shifting
gUSB (t ) = m(t ) cos(C t ) mh (t ) sin(C t )
g LSB (t ) = m(t ) cos(C t ) + mh (t ) sin(C t ) m(t)cos(ct)
X
(a)

cos(ct)
(d)
m(t)
gSSB(t)
Phase Shifter
sin(ct) /2 gUSB(t) if
+ or
gLSB(t) if +

(b) (c)
Phase Shifter
/2 X
mh(t)sin(ct)

mh(t)

SSB Modulator
Phase-shifting Method:
Frequency-Domain Illustration
SSB Demodulation (Coherent)
g SSB (t ) = m(t ) cos(C t ) mh (t ) sin(C t )
1 1
g SSB (t ) cos(C t ) = m(t )[1 + cos(2C t )] mh (t ) sin( 2C t )
2 2
1
LPF Output = m(t )
2
HLPF()
gSSB(t)
(Upper or Lower
X BW = 2B
m(t)

Side bands)

cos(ct)

SSB Demodulator (receiver)

 MULTIPLEXING

Multiplexing methods of transmitting more than one signal along a

single transmission path/stream i.e many to one.
Demultiplexing separate the stream back into its component
transmission i.e one to many.
Path refers to the physical link.
Channel refers to a portion that carries a transmission between a
given pair of devices. One path can have many channels.

30
 MULTIPLEXING
Two common form of multiplexing are Frequency Division
Multiplexing(FDM) and Time Division Multiplexing(TDM).
Advantages:
Increase number of channels so that more info can be transmitted
Save cost by using one channel to send many info signals

31
 Frequency Division Multiplexing (FDM)
FDM multiple sources that originally occupied the same frequency
spectrum are each converted to different frequency band and
transmitted simultaneously.
FDM is an analog technique the information entering an FDM
system must be analog. If the source is digital, it must be
converted to analog before being frequency-division multiplexed.
Split the total channel bandwidth into several smaller channels of
different frequencies.
Different signal travel over the medium concurrently.
Guard bands keep the modulated signals from overlapping and
interfering with one another.
Modulation is used to lift the centre freq of the baseband signal up
into a preassigned freq slot.
32
FDM

In communication systems, Frequency Division Multiplexing (FDM) is a

method in which each signal (channel) is allocated a frequency slot within the
overall line/transmission bandwidth.
In other words the total available frequency bandwidth on the transmission
line is divided into frequency channels and each information signal occupies
one of these channels
The signal will have exclusive use of this frequency slot all the time (i.e. each
subscriber occupies his/her own slot).
FDM of Three Voiceband Signals
The bandwidth of a voice signal is
taken to be 4kHz, with an effective
spectrum of 300 to 3400Hz.

If this signal is used to amplitude modulate

a 64 kHz carrier, the spectrum becomes
the modulated signal which has a
bandwidth of 8 kHz, extending from 60 to
68kHz.

To make efficient use of bandwidth, we

transmit only the lower sideband.
If three voice signals are used to modulate
carriers at 64, 68, and 72 kHz, the
spectrum output is as shown.
Frequency Division Multiplexing (FDM)
A number of signal, mi(t), i=1..,n are to be
multiplexed onto the same Tx medium.

Each signal mi(t),is modulate onto a

carrier fi, refers as subcarrier.
Modulated signals are then summed to
produce a composite signal mb(t).

Figure (b) shows the result signal mi(t)

is shifted to be centered of fi.

fi must be chosen so that the BW of the

various signals do not overlap i.e channel
must be separated by unused BW (guard
band).

At the Rx end, the FDM signal is

demodulated to retrieved mb(t), which is
35
then passed through n BPF.a
With FDM, each narrowband
channels are stacked on top of
one another in the frequency
domain.

Figure shows a simple

FDM system where four
5kHz channels are
frequency-division
multiplexed into a single
20kHz combined channel.
Figure (a) shows how a group is formed with A-type channel bank.
Each voice band channel is bandlimited with an antialising filter prior to modulating
the channel carrier.
Figure (b) shows the output spectrum.
 Analog Hierarchy

Figure shows first stages of telephone mux

Group multiplexer takes 12 voice ch and puts them on subcarriers at 64,
68,108kHz using LSB.
The resulting spectrum extends 48kHz starting at 60kHz.
Five such carriers are combined by LSB on subcarriers at 420, 468,
612kHz to produce a supergroup from 312 to 552kHz an contains 60
voice channels. 39
Analog Hierarchy

If the BW of the tx medium permits, 10 supergroups can be combined

to form a mastergroup.
For a wider BW applications such as satellite links, it is possible to
form a jumbogroup (6 mastergroups)

40
 FDM in Telephone System

The original signal voice is in 300 to 3000Hz range.

The voice is used to modulate subcarrier. Each subcarrier is on different
frequency.
These subcarriers are then added together to form a single channel.

Voice fv+fc
Balance
fv Modulator
fv-fc fv-fc
0 - 4 kHz Ch 12 BPF 60 64kHz
56 64kHz

fc = 60kHz
0 - 4 kHz Ch 11 BPF Linear
Mixer
fc = 64kHz

104 108kHz
100 108kHz SSBSC
0 - 4 kHz Ch 1 BPF
DSBSC
Selects USB
fc = 104kHz
 FDM in Telephone System

Voice signal amplitude modulates 1 of 12 chs in the 60 to 108kHz range.

The carrier freq begin at 60kHz with a spacing of 4kHz.(slightly higher
than the highest typical freq of voice)
Output of the balance modulator DSBSC. The output of the filter is the
SB containing the original voice signal.
All 12 SSB signals are then summed in a linear mixer to produce a single
frequency multiplexed signal basic group.
Basic group freq spectrum for FDM telephone mux system is shown below.
4kHz
Channel
12 11 10 9 8 7 6 5 4 3 2 1 No.

f (kHz)
60 64 68 72 76 80 84 88 92 96 100 104 108
Carrier frequencies

42
If more than 12 voice channels are needed, multiple basic groups are used.
 Example For a particular telephone company, the first sub
carrier frequency is at 60 kHz and the total bandwidth is 96 kHz.
Design a FDM system, given a general rule of 12 channels per
basic group and 4 kHz per channel applies to the design.
i. How many basic groups are required?
ii. Draw the circuit diagram of your design
iii. Draw the frequency spectrum of your multiplexed system
Solution
i. Given BWtotal = 96kHz;12 channels/basic group
1 channel = 4 kHz,
then 12 channels = 12x4 = 48kHz/basic group
Thus 96/48 = 2 basic group

43
 ii. Block Diagram

Ch 12 BPF
fc = 60kHz Linear
Voice Mixer

Ch 1 BPF
Linear
fc = 104kHz Mixer
Ch 24 BPF
fc = 108kHz Linear
Mixer

Ch 13 BPF
fc = 152kHz

iii. Frequency Spectrum

48 kHz 48 kHz

44
f (kHz)
60 104 108 152
FDM in Telephony
FDM is done in stages
Reduce number of carrier frequencies
More practical realization of filters
Group: 12 voice channels 4 kHz = 48 kHz
occupy the band 60-108 kHz
Supergroup: 5 groups 48 kHz = 240 kHz
occupy the band 312-552
Mastergroup: 10 S-G 240 kHz = 2400 kHz
occupy the band 564-3084 kHz
FDM Hierarchy
108 k 552 k
12 5
11 4
10 3
9 2
8 1 312 k
7
6 Supergroup
5
4
3
2
1 60 k
4
0 Group
B. Frequency Modulation

Most widely used form of Angle modulation for

mobile radio applications
AMPS
Police/Fire/Ambulance Radios
Generally one form of "angle modulation"
Creates changes in the time varying phase (angle) of
the signal.
Many unique characteristics

47
FM Illustration

The frequency of
the carrier is varied
around c in
relation with the
message signal.
i(t)= c + kf m(t)
Instantaneous Frequency
The argument of a cosine function represents an angle.
The angle could be constant [cos(300)], or varying with
time, cos [(t)]
The instantaneous angular frequency (in rad/sec) is the
rate of change of the angle. That is:
i (t) = d (t)/dt .
For cos(c t +), i (t)= c as expected.
t

(t ) = i ( )d .

Representation of Angle Modulation in Time Domain

For an FM signal: i (t) = c + kf m(t)

t t
FM (t ) = i ( )d = c t + k f m( )d .

t

A cos c t + k f
g FM (t ) =

m ( )d

For Phase Modulation (PM), the phase of the carrier is

varied in relation to the message signal: (t) = kp m(t)
A cos c t + k p m (t )
g PM (t ) =
Relation Between FM and PM
PM (=
t ) c t + k p m (t ),
dm (t )
i (t ) =
c + k p c + k p m (t ).
= Frequency Modulator (FM)
dt
t
m(t)
FM Modulator gFM(t) t m (t )d Phase
()d

m(t) Modulator gFM(t)
(PM)
m(t)
PM Modulator gPM(t)

Phase Modulator (PM)

dm (t )
d () dt Frequency
m(t) Modulator gPM(t)
dt (FM)
Which is Which?
 FM and PM Modulation
kf = 2105 rad/sec/volt =105 Hz/Volt = 105 V-1sec-1
kp = 10 rad/Volt = 5 v-1
fc = 100 MHz
FM:
fi = fc + kf m(t)
108 -105 < fi < 108 +105
99.9 < fi < 100.1 MHZ
PM:
fi = fc + kp dm(t)/dt
108 -105 < fi < 108 -105
99.9 < fi < 100.1 MHZ
Power (FM or PM) = A2/2
Spectrum of FM/PM
Unlike Amplitude Modulation, it is not straightforward
to relate the spectrum of the FM/PM modulated signal
to that of the modulating signal m(t). We can deal with
it on a case-by-case basis.
We are, however, particularly interested in finding the
bandwidth occupied by an FM/PM signal.
For that purpose, we will make some assumptions and
work on simple modulating messages.
Because of the close relation between FM and PM, we
will do the analysis for FM and extend it to PM.
What is NOT the bandwidth of FM!
One may tend to believe that since the modulated
signal instantaneous frequency is varying between by
f around fc, then the bandwidth of the FM signal is
2f. False!
In fact, the motivation behind introducing FM was to
reduce the bandwidth compared to that of Amplitude
Modulation, which turns out to be wrong.
What was missing from the picture of bandwidth?
FM Visualization
Think of holding the frequency knob of a signal
generator, and wiggling it back and forth to
modulate the carries in response to some
message.
There are two wiggling parameters:
How far you deviate from the center frequency (f)
How fast you wiggle (related to Bm)
The rate of change of the instantaneous
frequency was missing!
Carsons Rule
BFM 2(f+Bm)
where f = frequency deviation = kf |m(t)|max
Bm = bandwidth of m(t)
Define the deviation ratio = f / Bm.
BFM 2( +1) Bm
The same rule applies to PM bandwidth,
BPM 2(f+Bm) = 2( +1) Bm
where (f )PM = kp |dm(t)/dt|max
Narrow Band and Wide Band FM
When f << Bm (or <<1), the scheme is
called Narrow Band (NBFM, NBPM).
BNBFM 2Bm (same for NBPM)
Therefore, no matter how small we make the
deviation around fc , the bandwidth of the
modulated signal does not get smaller than 2Bm.
 Estimate BFM and BPM
kf = 2105 rad/sec/volt =105
Hz/Volt = 105 V-1sec-1
kp = 5 rad/Volt = 2.5 v-1
fc = 1000 MHz
First estimate the Bm.
Cn = 8/2n2 for n odd, 0 n even
The 5th harmonic onward can
be neglected.
Bm = 15 kHz
For FM:
f = 100 kHz; BFM =230 KHz
For PM:
f = 50 kHz; BFM =130 KHz
 Repeat if m(t) is Doubled
kf = 2105 rad/sec/volt =105 40,000
2
Hz/Volt = 105 V-1sec-1
kp = 5 rad/Volt = 2.5 v-1
-2 -40,000
fc = 1000 MHz
For FM:
f = 200 kHz; BFM = 430 KHz
For PM:
f = 100 kHz; BFM = 230 KHz
Doubling the signal peak has
significant effect on both FM
and PM bandwidth
 Repeat if the period of m(t) is Doubled
kf = 2105 rad/sec/volt =105 4x10-4
10,000
Hz/Volt = 105 V-1sec-1
kp = 5 rad/Volt = 2.5 v-1
-10,000
fc = 1000 MHz
Bm = 7.5 kHz
For FM:
f = 100 kHz; BFM = 215 KHz
For PM:
f = 25 kHz; BFM = 65 KHz
Expanding the signal varies its
spectrum. This has significant
effect on PM.
 Example: AMPS
poor spectral efficiency
allocated channel BW = 30 kHz
actual standard uses threshold specifications :

62
SNR vs. BW tradeoff
in FM one can increase RF BW to improve SNR:
SNRout = SNR after FM detection
f 3SNRin: FM
f : peak frequency deviation of Tx the
frequency domain

63
 rapid non-linear, f 3 improvement in output
signal quality (SNRout) for increases in f
capture effect : FM Rx rejects the weaker of the
two FM signals (one with smaller SNRin) in the
same RF BW resistant to CCI
Increased f requires increasing the bandwidth and
spectral occupancy of the signal
must exceed the threshold of the FM detector,
which means that typically SNRin 10 dB (called
the capture threshold)
64
Lecture 8: Digital Modulation II

Chapter 5 Modulation Techniques for

Mobile Radio

1
 Recall our picture of the overall wireless transmission and
receiving system:

2
 Last lecture
Analog AM and FM
Benefits of Digital Modulation
Power and Bandwidth Efficiencies
Linear Modulation BPSK, DPSK, QPSK
Bit error rate computations.

3
Constant Envelope Modulation Methods

Constant Envelope as compared to AM

Linear: Amplitude of the signal varies according to
the message signal.
Constant Envelope: The amplitude of the carrier is
constant, regardless of the variation in the message
signal. It is the phase that changes.

4
 Benefits of Constant Envelope
Power efficient
low out-of-band radiation of the order of -60dB
to -70 dB
Simpler receiver design can be used.
High immunity against random FM noise and
Rayleigh fading.
Disadvantage of Constant Envelope
Occupies larger bandwidth than linear
modulation.

5
6
7
 In the figure above, MSK is a type constant
envelope modulation.
MSK has lower sidelobes than QPSK
23 dB vs. 10 dB
MSK has larger null-to-null BW than QPSK
1.5 Rb vs. 1.0 Rb
But 99% RF BW is much better than QPSK (1.2 Rb
vs. 8.0 Rb!!)
very low ACI

8
 Research
When responding to natural or man-made
emergencies, cellular systems are heavily congested.
And users cannot be expected to regulate their
behavior to allow emergency workers to use the
spectrum.
Example heard from a radio report: A press person
talked about how hard it was to make a phone call on
September 11, 2001, but never mentioned that maybe
their own need to communicate of a lower priority.
Press people have also been known to overload 9-1-1
call centers to try to get information for their reports.

9
 GSM has a mechanism for identifying priority
calls and queueing those calls if they are not
first accepted.
Called the Wireless Priority Service (WPS).
This gives a lower blocking probability for those
calls.
But this still does not alleviate congestion.
GSM uses a constant envelope modulation scheme
(discussed below) that is not bandwidth efficient.

10
 proposal:
Assume that after a disaster, the FCC might relax
power restrictions. This would remove some of the
expectation for the power efficiency for which
GSM was designed.
Allow users to switch to a linear modulation
scheme to be more bandwidth efficient, needing
less bandwidth to be used per channel, creating
more channels.
But linear modulation also has more out-of-band
ACI problems, so we must compensate for that.

11
 Software-defined radios can be used to change
modulation schemes on demand in software when a
disaster occurs.
A part of the spectrum is set aside for the new
modulation scheme.
And existing phones could still use standard GSM
using another part of the spectrum.
Research: Finding a good linear modulation scheme,
reducing ACI, and implementing the software
defined radio.

12
BFSK
BFSK Binary Frequency Shift Keying
Frequency of constant amplitude carrier shifted
between two possible frequencies fH = 1 and
fL = 0

f = frequency offset from fc

13
 BFSK signal

2 Eb 1
s (t ) = cos 2f c t 2 t

Tb b
2T

Can use a simple method to switch between two

oscillators
but this might cause discontinuities
if the switching between signals is done when either one is not
at a zero value
What problems do discontinuities cause?

14
 But the phase between bits can be made to be
continuous
no discontinuity constant envelope retained
if we design the circuits based on the definition of
FM from before:

Then even if the message signal m () is discontinuous,

the integral of it will not be and the signal will then be
continuous.
But this is more complicated than simply switching
between two oscillators.

15
 BFSK BW
If B = baseband BW of the message signal
RF BW = 2 f + 2 B
Assume that first null BW is used, the BW of rectangular pulses
is B=R
RF BW = 2 f + 2 R
BER for Coherent detection of BFSK
Eb
Pe, FSK = Q

N o

16
 MSK
MSK Minimum Shift Keying
Specific type of continuous phase (CP) FSK
Special condition: Peak frequency deviation is of
the bit rate, so f = 0.25 Rb Tb= f = 1 2Tb
This is a smaller frequency separation (half that of
conventional FSK) and has easier detection.
It possesses properties such as:
constant envelope
spectral efficiency
good BER performance
self-synchronizing capability.

17
18
19
 An MSK signal can
be thought of as a
special form of
OQPSK where the
baseband rectangular
pulses are replaced
with half-sinusoidal
pulses during a period
of 2T

20
 can be deduced that
MSK has a constant amplitude.
Phase continuity at the bit transition periods is
ensured by choosing the carrier frequency to be an
integral multiple of one fourth the bit rate, 1/4T.
the MSK signal is an FSK signal with binary
signaling frequencies of fc + 1/4T and fc - 1/4T.

21
 MSK RF signal BW
MSK has lower sidelobes than QPSK 23 dB vs. 10 dB
MSK has larger null-to-null BW than QPSK 1.5 Rb vs. 1.0 Rb
But 99% RF BW is much better than QPSK (1.2 Rb vs. 8.0 Rb !!)
very low ACI
Very popular modulation scheme for mobile radio
22
GMSK
GMSK Gaussian MSK
The spectral efficiency of MSK is further enhanced
by filtering the baseband signal of square pulses
with a Gaussian filter.

23
 Further reduces sidelobes.
Designed based on the product of the filter
bandwidth (Bb) and the symbol period (T)
Bb T = corresponds to MSK
GSM uses Bb T = 0.3, which defines the bandwidth
of the Gaussian filter
The smaller the value of Bb T, however, the higher
the error rates.
Sacrifices the irreducible error rate in exchange for
extremely good spectral efficiency and constant
envelop properties

24
 GMSK premodulation filter has an impulse response
given by
2 2
=
hG (t ) exp 2 t

ln 2 0.5887
exp( 2 f 2 )
HG ( f ) = ==
2B B

25
Summary: OQPSK (IS-95) and GMSK (GSM) are the two
main modulation methods for 2G systems.

26
27
Combined Linear and Constant Envelope Modulation Techniques

We can allow both the phase and the amplitude

to change at the same time this would be a
combination of linear and constant envelop
methods.
We can extend the idea of QPSK to create
symbols with M possible states (instead of just
2 or 4).
M = 2n so each symbol encompasses n bits of
data.

28
M-ary PSK
M-ary PSK - constant envelope with more phase
possibilities

29
 2 2
sM PSK (t ) = Es cos (i 1) , Es sin (i 1)
M M
i = 1,2,..., M

30
 the first null bandwidth of M-ary PSK signals decrease as
M increases while Rb is held constant.
31
 for fixed Rb , B and b as M .
At the same time, M implies that the
constellation is more densely packed, and hence
the power efficiency p (noise tolerance) .

32
QAM
Quadrature Amplitude Modulation (QAM)
Change both amplitude and phase.
The general form of an M-ary QAM signal
2 Emin 2 Emin
si (t ) ai cos(2 f c t ) + bi sin(2 f c t )
Ts Ts
0t T i=1, 2,..., M

( L + 1, L 1) ( L + 3, L 1) ( L 1, L 1)
( L + 1, L 3) ( L + 3, L 3) ( L 1, L 3)
{ai , bi } =



( L + 1, L + 1) ( L + 3, L + 1) ( L 1, L + 1)
where L = M

33
L= M
(3,3) (1,3) (1,3) (3,3)
(3,1) (1,1) (1,1) (3,1)
{ai , bi } =

(3, 1) (1, 1) (1, 1) (3, 1)

(3, 3) (1, 3) (1, 3) (3, 3)
34
 Basic tradeoff: Better bandwidth efficiency at
the expense of power efficiency
More bits per symbol time better use of
constrained bandwidth
Need much more power to keep constellation points
far enough apart for acceptable bit error rates.
need a large circle for M-ary PSK
symbols at corners (extreme points) of QAM
constellation use a lot of power.

35
36
 M-ary FSK
M-ary FSK
Frequencies are chosen in a special way so that they are
easily separated at the demodulator (orthogonality principle).
M-ary FSK transmitted signals:

2 Es
=si (t ) cos (nc + i )t =
0 t Ts i 0,1,..., M
Ts Ts

fc = nc / 2Ts for some integer nc

The M transmitted signals are of equal energy and
equal duration
The signal frequencies are separated by 1 / 2Ts Hz,
making the signals orthogonal to one another

37
 The bandwidth efficiency of an M-ary FSK
signal with M
Power efficiency with M
Since M signals are orthogonal, there is no
crowding in the signal space

38
Spread Spectrum Modulation (SSM)
Tx expands (spreads) signal BW many times with
a special code and the signal is then collapsed
(despread) in Rx with the same code
Other signals created with other codes just appear
at the Rx as random noise.
Trade BW for signal power like with FM

39
 Advantages
1) Resistant to narrowband interference interference
can only realistically affect part of the signal.
2) Allows multiple users with different codes to share
same the MRC
no frequency reuse needed
rejects interference from other users

40
3) Combats multipath fading if a multipath signal is
received with enough delay (more than one chip
duration), it also appears like noise.
4) Can even use shifted versions of codes to isolate and
receive different multipath components (RAKE
receiver which we will see later)
5) As # simultaneous users the bandwidth efficiency

41
 Signal spreading is done by multiplying the data
signal by a pseudo-noise (PN) code or sequence
the pseudo-noise signal looks like noise to all except
those who know how to recreate the sequence.

42
 PN Codes
Binary sequence with random properties noise-like
(called "pseudo-noise" because they technically are
not noise)
equal #s of 1s and 0s
Very low correlation between time-shifted versions of
same sequence

43
 Very low cross-correlation between different
codes
each user assigned unique code that is
approximately orthogonal to all other codes
the other users signals appear like random noise!

44
 Exactly 2m-1 nonzero states for an m-stage
feedback shift register
The period of a PN sequence can not exceed 2m-1
symbols (maximal length)

45
 Spreading codes

The correlation properties of PN codes are such

that this slight delay causes the multipath to appear
uncorrelated with the intended signal
Multipath contributions appear invisible the desired
Rx signal

46
Direct Sequence (DS)

Two types of SSM DS & FH

1) Direct Sequence (DS)
Multiply baseband data by PN code (same as
diagram above)
Spread the baseband spectrum over a wide range.
The Rx spread spectrum signal

2 Es
si (t ) m(t ) p (t ) cos ( 2 f c t + )
Ts

m(t) : the data sequence

p(t): The PN sequence

47
48
Frequency Hopping (FH)

2) Frequency Hopping (FH)

Randomly change fc with time
Spread the frequency values that are used over a
wide range.
In effect, this signal stays narrowband but moves
around a lot to use a wide band of frequencies over
time.

49
 Hopset : the set of possible carrier
frequencies
Hop duration: the time during between hops
Classified as fast FH or slow FH
fast FH: more than one frequency hop during
each Tx symbol
slow FH : one or more symbol are Tx in the time
interval between frequency hops.

50
51
 Bluetooth uses FH because it is an ad-hoc network.
DS would require more precise bit timing
coordination (because of the high data rate signal),
which is hard to do among an ad hoc collection of
devices.
Bluetooth uses frequency hopping with a dwell time
of 625 s (1600 frequency hops per second) over 79
different frequencies

Processing Gain = PG
SSM is resistant to narrowband interfering signals

52
 Part (a) shows how an interfering source can only affect
a small part of the spectrum of the signal.
Part (b) shows how the despreading process shrinks the
signal spectrum and spreads out the interference energy.

53
 Most of interfering energy will be outside of signal
bandwidth and will be removed with Low Pass
Filtering

The larger the PG, the greater the ability to suppress

in-band interference.

54
Performance of DS spread spectrum

3N
Pe = Q
K 1
K : multiple acess users
N : Chips

55
Performance of FH spread spectrum

Error rate due to multiple access interference

1 K 1
lim ( Pe ) =
Eb
2 M
N0

K : multiple acess users

M : Hopping channel

To combat the occasional hits

Applying Reed-Solomon or other burst error
correcting codes
Not as susceptible to the near-far problem

56
 With Spread Spectrum Modulation, users are
able to share a common band of frequencies
a multiple access technique
TDMA: Users share a band of frequencies, but use a
different time slot
FDMA: Users share a band of frequencies, but use a
different slice of frequency
SSM enables CDMA (Code Division Multiple Access):
Users share a band of frequencies, but each use a
different spreading code.

57
 Sprint PCS, Cingular, and AT&T Wirless
DS-SSM
Sprint PCS was the first nationwide deployment of
a CDMA system
Technology started by Qualcomm
The main disadvantage of DS-SSM is that very
good power control of mobiles is required
Near/far problem
Discussed in Chapter 8

58
 Performance of digital modulation in slow flat-fading channel

59
 Performance of digital modulation in frequency selective channel

60
61
 Next lectures: Using the concept of redundancy
to improve wireless signal quality.
Redundant antennas
diversity to overcome fading.
Redundant data bits
error control codes to detect and correct errors.

62
Multicarrier Modulation and
OFDM
Multicarrier Modulations - Introduction
In multicarrier (MC) modulation, a transmitted bit stream
is divided into many different sub-streams, which are sent
in parallel over many subchannels;
The parallel subchannels are typically orthogonal under
ideal propagation conditions;
The data rate on each of the subcarriers is much lower than
the total data rate;
The bandwidth of subchannels is usually much less than
the coherence bandwidth of the wireless channel, so that
the subchannels experience flat fading. Thus, the ISI on
each subchannel is small;
MC modulation can be efficiently implemented digitally
using the FFT (Fast Fourier Transform) techniques,
yielding the so-called orthogonal frequency division
multiplexing (OFDM);
 Multicarrier Modulations-Application
Examples
Digital audio and video broadcasting in Europe;
Wireless local area networks (WLAN) - IEEE802.11a, g,
n, ac, ad, etc.;
Fixed wireless broadband services;
Mobile wireless broadband communications;
Multiband OFDM for ultra-wideband (UWB)
communications;
Main modulation scheme in the 4th generation cellular
mobile communications systems (uplink SC-FDMA,
downlink OFDMA);
A candidate for many future generations (802.11ax, 5th
generation cellular) of wireless communications systems.
Multicarrier Modulations -
Transmitter
 Multicarrier Modulations - Principles
Consider a linearly modulated system with data rate R and
bandwidth B;
The coherence bandwidth of the channel is assumed to
be Bc < B, so signals transmitted over this channel
experience frequency-selective fading.
When employing the MC modulations:
the bandwidth B is broken into N subbands, each of which has
a bandwidth BN = B/N for conveying a data rate RN = R/N;
Usually, it is designed to make BN << Bc, so that the
subchannels experience (frequency non-selective) flat fading.
In the time-domain, the symbol duration TN1/BN of the
modulated signals is much larger than the delay-spread Tm
1/Bc of the channel, which hence yields small ISI.
 An example
Consider a MC system with a total passband bandwidth
of 1 MHz. Suppose the channel delay-spread is Tm = 20s.
How many subchannels are needed to obtain
approximately flat fading in each subchannel?
The channel coherence bandwidth is
Bc = 1/Tm = 1/0.00002 = 50 KHz;
To ensure flat fading on each subchannel, we take BN =
B/N = 0.1 Bc << Bc;
Hence, N = B/(0.1 Bc) = 1000000/5000 = 200 subcarriers.
Note for OFDM we need channels in power of 2n
therefore select next higher value, 256 channels will be
used
 Multicarrier Modulations
Transmitted Signals

where
si: complex data symbol (QAM, PSK, etc.) transmitted
on the ith subcarrier;
: phase offset of the ith subcarrier;
fi= f0 + i(BN): central frequency of the ith subcarrier;
g(t): waveform-shaping pulse, such as raised cosine
pulse
Multicarrier Modulated Signal
Receiver
Overlapping MC
OFDM modulation at baseband
 Overlapping MC - Detection
Without considering fading and noise, the received MC signal can be
expressed as

Assuming that the detector knows {i}, then, sj can be detected as

where C is a constant.
 Implementation of MC Modulation
Using DFT/IDFT
Let x[n]; 0 1; denote a discrete time sequence.
The N-point discrete Fourier transform (DFT) of {x[n]} is
defined as

Correspondingly, given {X[i]}, the sequence {x[n]} can be

recovered by the inverse DFT (IDFT) defined as
 Implementation of MC Modulation
Using DFT/IDFT
When an input data stream {x[n]} is sent through a linear time-
invariant discrete-time channel having the channel impulse response
(CIR) {h[n]}, the output {y[n]} is given by the discrete-time
convolution of the input and the CIR, expressed as

Circular Convolution: when {x[n]} is a N-length periodic sequence,

then the N-point circular convolution of {x[n]} and {h[n]} is defined as

which has the property

 Implementation of MC Modulation:
Cyclic Prefix

The original N-length data block is x[n] : x[0];.; x[N -

1];
The -length cyclic prefix block is x[N - ]; ; x[N
- 1], which is constituted by the last symbols of the
data block {x[n]};
The actually transmitted data block is length N + ,
which is [n] : x[N - ]; ; x[N - 1]; x[0]; x[1];...;
x[N 1]
 Implementation of MC Modulation:
Cyclic Prefix
Then, when {[n]} is input to a discrete-time
channel having the CIR h[n] : h[0]; .. ; h[L],
the channel outputs are

Therefore,

When {Y [i]} and {H[i]} are given, the

transmitted sequence can be recovered as
 OFDM Using Cyclic Prefixing - An
Example
Consider an OFDM system with total bandwidth B = 1
MHz and using N = 128 subcarriers, 16QAM
modulation, and length = 8 of cyclic prefix.
Then
The subchannel bandwidth is BN = B/128 = 7.812 kHz;
The symbol duration on each subchannel is
TN = 1/BN = 128 s;
The total transmission time of each OFDM block is
T = TN + 8/B = 136 s;
The overhead due to the cyclic prefix is 8/128 = 6.25%;
The total data rate is
128 x log2 16 x 1/(T = 136 x 10-6) = 3.76 Mbps.
OFDM - System Structure
OFDM - Transmitter
OFDM - Receiver
 OFDM - Transmitted Signal
Let the N data symbols (thought as in the frequency-
domain) to be transmitted on the N subcarriers within
a DFT period is given by
X= [X0; X1; ; XN-1]T (14)
After the IDFT on X, it generates N time-domain
coefficients expressed as

Let be a fast Fourier transform (FFT) matrix given

by the next slide. Then, we can express x = [x0;
x1;; xN-1]Tas
x = FH X (16)
 FFT/IFFT Matrices
FFT matrix:

where WN = / ;
IFFT matrix is given by FH;
Main Properties: Unitary- FHF= FFH = IN.
A square complex matrix with orthonormal columns
is called unitary
 OFDM - Transmitted Signal
After adding the cyclic-prefix (CP) of length , x is modified to of
length N+ ;
The normalized transmitted base-band OFDM signal is formed as

where
A: Constant related to the transmit power;
T(t): time-domain waveform construction function, such as the sinc(.)-
function;
T: chip-duration and T =1/B.
 OFDM - Representation of Received
Signals
When the OFDM signal of (18) is transmitted over a frequency
selective fading channel with the CIR hn; 0 as well as
Gaussian noise, the discrete-time received observation samples in
correspondence to x0; x1;; xN-1are obtained from sampling the
received signal, which can be expressed as

Let y = [y0; y1;; yN-1]T. Then, it can be shown that y can be

expressed as

= + (20)
.
Here, it is very important to represent the matrix
 Circulant matrix
A circulant matrix is a matrix where each row (column) is a cyclic
shift of the preceding row (column):

Note that a single row (or column) completely determines the

entire matrix.
The beauty of circulant matrices lies in the fact that they are all
diagonalized in the Fourier basis,

that is, every circulant matrix can
be decomposed as: =
where U is the unitary discrete Fourier transform matrix, U* is its
conjugate transpose and is the diagonal matrix of eigenvalues of
C. The eigenvalues can be found by taking the discrete Fourier
transform of the first row of C.
 Linear Convolution
Please Refer any Basic book in Signals and
Systems
Circular Convolution
 Illustration of circular convolution for
N=8

x[n]spread clockwise
y[n]spread counter clockwise
Example
 Relationship between Linear
Convolution and Circular Convolution
Why are we interested?
Linear convolution is the filtering operation
Filters are a very important application of DSP
Would like to implement filters efficiently FFT
is a fast DFT
Maybe we can use FFTs to implement filters (?)
 Relationship between Linear
Convolution and Circular Convolution
We desire that circular and linear convolution
give identical results,
then we can use FFTs for fast filtering
This can be achieved by applying zero-padding to
the signals before performing circular convolution
For a signal x1(n) of length N1 and signal x2(n) of
length N2
Zero-pad x1(n) with N2-1 zeros
Zero-pad x2(n) with N1-1 zeros
Example
OFDM - Representation of Received
Signals (Linear Convolution)
OFDM - Representation of Received
Signals (Another Way)
 OFDM - Representation of Received
Signals
From the previous slide, we can see that
OFDM - Representation of Received
Signals
 OFDM - Representation of Received
Signals
Therefore, we have
 OFDM - Representation of Received
Signals
In (22), if CP is used and set as x-i = xN-i; i =
1;;L, then, (22) can be represented as
OFDM - Representation of Received
Signals
 Example
Let we assume x = [x0; x1; x2; x3]T and L = 2.
Then, we have
 OFDM - Signal Detection
In (24), H is a circulant channel matrix, which can be
decomposed into = , where = diag{H0; H1;...;HN-1}
is a (NX N) diagonal matrix, and Hn is in fact the fading gain of
the nth subcarrier.
Using x = FHX of (16), we can re-write (24) as

Carrying out the FFT on y gives

Therefore, for n = 0; 1;; N - 1,

based on which {Xn} can be detected.

Explicitly, there is no ISI.
 Transmit signal processing

F=RH is the normalised IDFT matrix

The columns of RH are the transmit signal components
kth transmit symbol (xk-1) scales the kth column of RH
Transmit multiplex is the sum of N scaled complex exponentials
(scaled by the data)
Can be implemented in O(N log N) using the IFFT
 Receive signal processing

R is the normalised DFT matrix

Rows of R contain the receive signal components
Multiplication of R with r corresponds to computing N
correlations (one per subcarrier/subchannel)
Can be implemented in O(N log N) using the FFT
 Channel: linear convolution matrix
We describe the channel dispersion using a convolution
matrix H, where we assume L = M:

Both DMT (H real-valued) or OFDM (H complex-valued)

can be represented
The first L received samples r(n), n = -L, . . . ,-1 are
implicitly ignored
OFDM modulation at baseband using
IFFT
 SSZG520_Lecture 14

Error Detection and Correction

A code taxonomy
 Error-control coding: basics of Forward Error
Correction (FEC) channel coding
Coding is used for error detection and/or error correction
Coding is a compromise between reliability, efficiency, equipment
complexity
In coding extra bits are added for data security
Coding can be realized by two approaches
ARQ (automatic repeat request)
stop-and-wait

go-back-N

selective repeat

FEC (forward error coding)

block coding Topic today
convolutional coding

ARQ includes also FEC

Implementations, hardware structures
 What is channel coding?
Coding is mapping of binary source (usually) output sequences of length
k into binary channel input sequences n (>k)
A block code is denoted by (n,k)
Binary coding produces 2k codewords of length n. Extra bits in
codewords are used for error detection/correction
In this course we concentrate on two coding types: (1) block, and (2)
convolutional codes realized by binary numbers:
Block codes: mapping of information source into channel inputs
done independently: Encoder output depends only on the current
block of input sequence
Convolutional codes: each source bit influences n(L+1) channel
input bits. n(L+1) is the constraint length and L is the memory
depth. These codes are denoted by (n,k,L).
k-bits (n,k) n-bits
block coder
 Representing codes by vectors
Code strength is measured by Hamming distance that tells how different
code words are:
Codes are more powerful when their minimum Hamming distance
dmin (over all codes in the code family) is large
Hamming distance d(X,Y) is the number of bits that are different
between code words
(n,k) codes can be mapped into n-dimensional grid:
3-bit repetition code 3-bit parity code

valid code word

 Hamming distance: The decision sphere
interpretation
Consider two block code (n,k) words c1 and c2 at the Hamming
distance d = min d (ci , c j ) in the n-dimensional code space:
i , j

d /2
c1 c2

It can be seen that we can detect l=dmin-1 errors in the code words.
This is because the only way to not to detect the error is that the error
transforms the code into another code word. This requires change in d
code bits.
Also, we can see that we can correct t=(dmin-1)/2 errors. If more errors
occur, the received word may fall into the decoding sphere of another
code word.
 Example: repetition coding
In repetition coding bits are repeated several times
Can be used for error correction or detection
For (n,k) block codes d min n k that
+ 1 is a bound achieved by repetition
codes. Code rate is anyhow very small
Consider for instance (3,1) repetition code, yielding the code rate
=
RC k=
/ n 1/ 3
Encoded word is formed by the simple coding rule:
1 111 0 000
Code is decoded by majority voting, e.g. for instance:
001 0, 101 1
In a three bit code word
one error can be corrected always, because majority voting can detect and
correct one code word bit error always
two errors can be detected always, because all code words must be all
zeros or all ones (but now the encoded bit can not be recovered)
 Error rate for a simple repetitive code

error rate pe
Note that by increasing word length
more and more resistance to channel
introduced errors is obtained.

code length
n n
 Parity-check coding

Repetition coding can greatly improve transmission reliability

However, due to repetition transmission rate is reduced. Here the code
rate was 1/3 (that is the ratio of the bits to be coded to the encoded bits)
In parity-check coding a check bit is formed that indicates number of 1
in the word to be encoded.
Even number of 1 means that the the encoded word has even parity
Example: coding 2-bit words by even parity is realized by

00 000, 01 011
10 101, 11 110
Question: How many errors can be detected/corrected by parity-check
coding?
 Generating block codes: Systematic block codes

In (n,k) block codes each sequence of k information bits is mapped into a

sequence of n (>k) channel inputs in a fixed way regardless of previous
information bits
The formed code family should be selected such that the code minimum
distance is as large as possible -> high error correction or detection capability
A systematic block code:
the first k elements are the same as the message bits
the following q = n - k bits are the check bits
Therefore the encoded word is
X= ( b1 b2 .... br m1 m2 ... mk ), r= n k

check message
or as the partitioned representation

X = ( B | M)
 Block codes by matrix representation

Given the message vector M, the respective linear, systematic block

code X can be obtained by the matrix multiplication by
X = MG
The matrix G is the generator matrix with the general structure
G = ( P | Ik )
where Ik is kxk identity matrix and P is a kxr binary submatrix
ultimately determining the generated codes

p11 p12 p1r

p p22 p2 r
P = 21

p pk 2 pkr
k1
X = (B | M )
 Generating block codes
For u message vectors M (each consisting of k bits) the respective n-bit
block codes X are therefore determined by
p1,1 p1,2 p1,r 1 0 0

p2,1 p2,2 p2,r 0 1 0
G = (P | Ik )

pk ,1 pk ,2 pk ,r

0 0 1

m1,1 m1,2 m1,k b1,1 b1,2 b1,r m1,1 m1,2 m1,k
m m m b b b m m m
X = MG
2,1 2,2 2,k 2,1 2,2 2,r 2,1 2,2 2,k

m

u ,1 mu ,2 m b
u ,k u ,1 bu ,2 bu ,r mu ,1 mu ,2 mu ,k

M X =( B|M )

Generated check bits, from above, as for instance for k=4,

cu ,2 = mu ,1 p1,2 mu ,2 p2,2 mu ,3 p3,2 mu ,4 p4,2
 Forming the P matrix
The check vector B that is appended to the message in the encoded word
is thus determined by the multiplication
B = MP
The j:th element of B on the u:th row is therefore encoded by
bu , j =
mu ,1 p1, j mu ,2 p2, j mu ,k pk , j , j =
1...r

For the Hamming code P matrix of k rows consists of all r-bit words
with two or more "1":s arranged in any order! Hence P can be for
instance
1 0 1
1 1 1
P=
1 1 0
0 1 1

 Generating a Hamming code: An example
For the Hamming codes n=2r-1, k = n - r, dmin=3 (P is k x r matrix)
Take the systematic (n,k) Hamming code with r=3 (the number of
check bits) and n=23-1=7 and k=n - r=7-3=4. Therefore the generator
matrix is

1 0 1 1 0 0 0
1 1 1 0 1 0 0
G =
1 1 0 0 0 1 0
0 1 1 0 0 0 1

P M
Note that in Hamming code the three last columns make up the P
submatrix including all the 3-bit words that have 2 or more 1:s.
For a physical realization of the encoder we now assume that the
message contains the bits M = ( m1 m2 m3 m4 )
 Realizing a (7,4) Hamming code encoder

For these four message bits we have a four element message register
implementation
Note that here the check bits [b1,b2,b3] are obtained by substituting the
elements of P into equation B=MP or
bj =
m1 p1 j m2 p2 j .... mk pkj
 Example*

*S. Lin, E. Costello: Error Control Coding: Fundamentals and Applications

 Listing generated Hamming codes

Going through all the combinations of the input vector X yields all the
possible output vectors

Note that for the Hamming codes the minimum distance or weight w =
3 (the number of 1 on each row)
 Decoding block codes
A brute-force method for error correction of a block code includes
comparison to all possible same length code structures and choosing the one
with the minimum Hamming distance when compared to the received code.
In practical applications codes can be very long and the extensive
comparison would require much time and memory. For instance, to get the
code rate of 9/10 with a Hamming code it is required that
k k nr 9
= =
n 2 r 1 2 r 1 10
This equation fulfills if the code length is at least k=57, and now n = 63.
There are 2= 1.4 10 different block codes in this case! Decoding by
k 17

direct comparison would be quite unpractical!

This approach of comparing Hamming distance of the received code to the
possible codes, and selecting the shortest one is the maximum likelihood
detection and will be discussed more details with convolutional codes
 Syndrome decoding for error detection
In syndrome decoding a parity checking matrix H is designed such that
multiplication with a code word produces all-zero matrix:
XHT = (0 0 0)
Therefore error detection of the received signal Y can be based on
syndrome:
S = YHT
that is always zero when a (correct) code word is received. (Note that
the syndrome does not reveal errors if channel noise has produced
another code word!)
The parity checking matrix is determined by
I
H = ( I rxr | PT ) or H = rxr
T

P
Having parity checking matrix designed such that the rows of HT are all
different and contain at least one "1" a distinct syndrome for each single
error pattern can be obtained -> enables error correction!
 Syndrome decoding for error correction
Syndrome decoding can be used for error correction by checking the
one-bit error pattern for each syndrome:
Example: Consider a (7,4) Hamming code with a parity check matrix
1 0 0 1 1 1 0
= H (= I rxr | PT ) 0 1 0 0 1 1 1

0 0 1 1 1 0 1
The respective syndromes and error vectors (showing the position of
errors by "1") are

= =
S YH T
( c + e)
HT

e = YHT where Y is any valid code with

the error in the position indicated by the
respective syndrome
 Syndrome is independent of code words
This design enables that the syndrome depends entirely on error pattern
but not on particular code. Consider for instance
X = (1 0=11 0) Y (1 0 0 11) E = (0 0 1 0 1) (Y= X + E)
=S YHT
=
( X + E) H T XH
=T
+ EHT EHT
0, that follows from the defintion of H

Syndrome does not determine the error pattern uniquely because there
exists only 2q different syndromes (syndrome length is q) but there
exists 2k different codes (for each symbol that must be encoded).
After error correction decoding double errors can turn out even triple
errors
Therefore syndrome decoding is efficient when channel errors are not
too likely, e.g. probability for double errors must be small.
For difficult channels there are more elaborated schemes using for
instance extended Hamming codes or maximum likelihood methods (as
the Viterbi-decoding)
 Table lookup syndrome decoder circuit
The error vector is used for error correction by the circuit shown bellow:

Error subtraction

P
S = YH T
H =
T

Iq
 Defining cyclic codes: code polynomial
An (n,k) linear code X is called a cyclic code when every cyclic shift of
a code X, as for instance X, is also a code, e.g.
X = ( x0 x1 xn2 xn1 )
X ' = ( xn1 x0 xn3 xn2 )

Each cyclic code has the associated code vector with the polynomial
X( p ) = x0 + x1 p + + xn2 p n2 + xn1 p n1
Note that the (n,k) code vector has the polynomial of degree of n-1 or
less. Mapping between code vector and code polynomial is one-to-one,
e.g. they specify the same code uniquely
Manipulation of the associated polynomial is done in a Galois field (for
instance GF(2)) having elements {0,1}, where operations are performed
mod-2
For each cyclic code, there exist only one generator polynomial whose
degree equals the number of check bits in the encoded word
 An example of a (7,4) cyclic code,
generator polynomial G(p)=1+p+p3
M( p) X( p ) M ( p )G ( p )

(1 + X )(1 + X + X )
3

= 1+ X + X 3 + X + X 2 + X 4
 The common factor of cyclic codes
GF(2) operations (XOR and AND):

Modulo-2 Addition Modulo-2 Multiplication

+ 0 1 * 0 1
0 0 1 0 0 0
1 1 0 1 0 1

Cyclic codes have a common factor pn+1. In order to see this we

consider summing two (unity shifted) cyclic code vectors:
X( p ) = x0 + x1 p + + xn2 p n2 + xn1 p n1 Unshifted
X '( p ) = xn1 + x0 p + x1 p 2 + + xn2 p n1 Right rotated
pX( p ) = x0 p + x1 p 2 + + xn2 p n1 + xn1 p n Right shifted by multiplication

Question is how to make the cyclic code from the multiplied code?
Adding the last two equations together reveals the common factor:
pX( p ) + X '( p )= xn1 p n + xn1= xn1 ( p n + 1)
 Factoring cyclic code generator polynomial
Any factor of pn+1 with the degree q=n-k generates an (n,k) cyclic code
Example: Consider the polynomial p7+1. This can be factored as
p 7 + 1 = (1 + p )(1 + p + p 3 )(1 + p 2 + p 3 )

For instance the factors 1+p+p3 or 1+p2+p3, can be used to generate an

unique cyclic code. For a message polynomial 1+p2 the following
encoded word is generated:
(1 + p 2 )(1 + p + p 3 ) =1 + p + p 2 + p 5
and the respective code vector (of degree n-1, in this case) is
(111 0 0 1 0)
 Obtaining a cyclic code from another cyclic code
Therefore unity cyclic shift is obtained by multiplication by p where
after division by the common factor yields a cyclic code
=X '( p ) pX( p ) mod( p n + 1) 1

and by induction, any cyclic shift is obtained by

1 + p3 p + p3
1 + p3
=X( i ) ( p ) p ( i ) X( p ) mod( p n + 1)
1+ p
Example: 101 X( p ) =
1 + p2
right shift 101
pX( p )= p + p 3 not a three-bit code,
divide by the common factor
pX( p ) 1
=
1+ 110
1+ p 3
1+ p
Important point is that division by mod pn+1 and multiplication by the
generator polynomial is enabled by tapped shift register.
 Using shift registers for multiplication
Figure shows a shift register to realize multiplication by 1+p2+p3

In practice, multiplication can be realized by two equivalent topologies:

 Example: multiplication by using a shift register

out
1 1 0 0 0 0 0 0 0 0 0
0 1 1 0 0 0 0 0 0 0 0
0 0 1 1 0 0 0 0 0 0 0
0 0 0 1 1 0 0 0 0 0 1
adding dashed line would 0 0 0 0 1 1 0 0 0 0 1
enable division by 1+pn determined by 0 0 0 0 0 1 1 0 0 0 1
the tapped connections 0 0 0 0 0 0 1 1 0 0 0
word to be
0 0 0 0 0 0 0 1 1 0 1
encoded
0 0 0 0 0 0 0 0 1 1 0
(1 + p )(1 + p 2 + p 3 )
=+
1 p 2 + p3 + p + p3 + p 4
=1 + p + p 2 + p 4 11101 Encoded word
 Examples of cyclic code generator polynomials
The generator polynomial for a (n,k) cyclic code is defined by
G ( p ) =1 + g1 p + + g q1 p q1 + p q , q =n k
and G(p) is a factor of pn+1. Any factor of pn+1 that has the degree q
may serve as the generator polynomial. We noticed that a code is
generated by the multiplication
X( p ) = M ( p )G ( p )
where M(p) is a block of k message bits. Hence this gives a criterion to
select the generating polynomial, e.g. it must be a factor of pn+1.
Only few of the possible generating polynomials yield high quality
codes (in terms of their minimum Hamming distance)
G ( p) = p 3 + 0 + p + 1

Some cyclic codes:

 Systematic cyclic codes
Define the length q=n-k check vector C and the length-k message
vector M by M ( p ) = m + m p + + m p k 1
0 1 k 1

C( p ) = c0 + c1 p + + cq1 p q1
Thus the systematic n:th degree codeword polynomial is
X(=
p ) p nk (m0 + m1 p + + mk 1 p k 1 )
+ c0 + c1 p + + cq1 p q1 message bits
= p q M ( p ) + C( p ) check bits

Check bits determined by:

C( p ) = mod [ p nk M ( p ) / G ( p ) ]

Question: Why these denote the message bits still

the message bits are M(p) ???
 Determining check-bits
Prove that the check-bits C( p ) can be calculated from the message bits
M(p) by
C( p ) = mod [ p M ( p ) / G ( p ) ]
nk

p nk M ( p ) / G=
( p ) Q( p ) + C( p ) / G ( p )
p nk M ( p ) / G ( p ) + C( p ) / G ( p ) =+
Q( p ) C( p ) / G ( p ) + C( p ) / G ( p )

=0
p M ( p ) + C( p ) =
nk
G ( p )Q( p )
must be a systematic code
=X ( p )
message check based on its definition
(previous slide)
Example: (7,4) Cyclic code:

G ( p ) = p + p + 1 p nk
M ( p ) / G ( p ) = p 3
+ p 2
+1+ 1
C (p )
3 2

= +
Q( p)
nk
3
M ( p ) p p
p 74 M ( p= p M ( p ) + C ( p ) = p 3
( p 3
+ p ) + 1= p 6
+ p 4
+1
) p +p
6 4

Q( p )G ( p ) = ( p 3 + p 2 + 1)( p 3 + p 2 + 1) = p 6 + p 4 + 1


Example: Encoding of systematic cyclic codes
Circuit for encoding systematic cyclic codes
0

We noticed earlier that cyclic codes can be generated by using shift registers
whose feedback coefficients are determined directly by the generating
polynomial
For cyclic codes the generator polynomial is of the form
G ( p ) = 1 + g1 p + + g q2 p q2 + g q1 p q1 + p q
In the circuit, first the message flows to the transmitter, and feedback switch is
set to 1, where after check-bit-switch is turned on, and the feedback switch to
0, enabling the check bits to be outputted
 Decoding cyclic codes
Every valid, received code word R(p) must be a multiple of G(p), otherwise an
error has occurred. (Assume that the probability for noise to convert code
words to other code words is very small.)
Therefore dividing the R(p)/G(p) and considering the remainder as a syndrome
can reveal if the error has happed and sometimes also to reveal in which bit
(depending on code strength)
The error syndrome of n-k-1 degree is therefore
S( p ) = mod [ R ( p ) / G ( p ) ]
This can be expressed also in terms of error E(p) and the
code word X(p)
R=
( p ) X ( p ) + E( p )
hence
=S( p ) mod {[ X( p ) + E( p ) ] / G ( p )}
S( p ) = mod [ E( p ) / G ( p ) ]
Decoding cyclic codes: example

Using denotation of this example:

16.20 s ( x) = mod [ e( x) / g ( x) ]
 Table 16.6
Decoding cyclic codes (cont.)

s ( x) = mod [ r ( x) / g ( x) ]

g ( x)
 Decoding circuit for (7,4) code
syndrome computation

0
received code 1 syndrome

G ( p) = p 3 + p + 1

While first receiving the code, switch is set to 0

The shift register is stepped until all the received code bits have entered the
register
This results 3-bit syndrome (n - k = 3 ) that is then left to the register
Then the switch is turned to the direction 1 that drives the syndrome out of the
register
 Convolutional coding
Block codes are memoryless
Convolution codes have memory that utilizes previous bits to encode or
decode following bits
Convolutional codes are specified by n, k and constraint length that is
the maximum number of information symbols upon which the symbol
may depend
Thus they are denoted by (n,k,L), where L is the code memory depth
Convolutional codes are commonly used in applications that require
relatively good performance with low implementation cost
Convolutional codes are encoded by circuits based on shift registers and
decoded by several methods as
Viterbi decoding that is a maximum likelihood method

Sequential decoding (performance depends on decoder

complexity)
Feedback decoding (simplified hardware, lower performance)
 Convolutional codes
(n, k, m) convolutional codes
k input bits will produce n output bits.
The most recent m k-message-bit input blocks will be
recorded (buffered).
The n output bits will be given by a linear
combination of the buffered input bits.
Constraint length of a convolutional code
Number of k-message-bit shifts over which a single k-
message-bit block can influence the encoder output.
Based on the above definition, constraint length
=m+1.
 Convolutional codes
Effective code rate of a convolutional code
In practice, km zeros are often appended at the end to
clear the shift register contents.
Hence, kL message bits will produce n(L+m) output
bits.
The effective code rate is therefore given by:

Since L is often much larger than m,

This is named the code rate of a convolutional
code.
Example: convolutional encoder
x ' j = m j2 m j1 m j
x=
'' j m j2 m j

X out = x '1 x ''1 x '2 x ''2 x '3 x ''3 ... (n,k,L) = (2,1,2) encoder
Convolutional encoder is a finite state machine processing information bits in
a serial manner
Thus the generated code word is a function of input and the state of the
machine at that time instant
In this (n,k,L)=(2,1,2) encoder each message bit influences a span of n(L+1)=6
successive output bits that is the code constraint length
Thus (n,k,L) convolutional code is produced that is a 2n(L-1) state finite-state
machine
(3,2,1) Convolutional encoder

x ' j = m j 3 m j 2 m j
Here each message bit influences
x '' j = m j3 m j1 m j
a span of n(L+1)=3(1+1)=6
x=
''' j m j2 m j successive output bits
 Generator sequences
(n,k,L) Convolutional code can be described by the generator sequences
g1 , g 2 ,...g n that are the impulse responses for each coder output branch:

(2,1, 2) encoder

g1 = [1 0 11] Note that the generator sequence length

g 2 = [1111] exceeds register depth by 1

Generator sequences specify convolutional code completely by the

generator matrix
Encoded convolution code is produced by matrix multiplication of the
input and the generator matrix
 Encoding equations
Encoder outputs are formed by modulo-2 discrete convolutions:
v (1) = u * g (1)
v(2) = u * g(2)
Therefore the l:th bit of the j:th output branch is
vl( j ) = im=0 ul i g l( j ) =ul g 0( j ) + ul 1 g1( j ) + ... + ul m g m( j ) , j = 1, 2...m, m = L + 2
Input for extraction of generator sequences is
i ,l = i g1 = [1 0 11]
u l i
0,otherwise
Hence for this circuit the following equations result: g 2 = [111 1]

v=
l
(1)
ul + u l 2 + u l 3
vl( 2 ) =ul + ul 1 + ul 2 + ul 3

Encoder output:
v = [v0(1) v0( 2 ) v1(1) v1( 2 ) v2(1) v2( 2 ) ...]
Example of using generator matrix

g1 = [1 0 11]
g = [111 1]
2
11
00 01
11 =
01

11 10

01

Verify that you can obtain the result shown!

 Representing convolutional code: code tree

x ' j = m j2 m j1 m j
x=
'' j m j2 m j
X out = x '1 x ''1 x '2 x ''2 x '3 x ''3 ...

Tells how one input bit

is transformed into two output bits
(initially register is all zero)
Representing convolutional codes compactly:
code trellis and state diagram
Input state 1
indicated by dashed line

Shift register states

Lecture 9: Diversity

Chapter 7 Equalization, Diversity, and Coding

1
I. Introduction
MRC Impairments:
1) ACI/CCI system generated interference
2) Shadowing large-scale path loss from LOS
obstructions
3) Multipath Fading rapid small-scale signal
variations
4) Doppler Spread due to motion of mobile unit
All can lead to significant distortion or attenuation
of Rx signal
Degrade Bit Error Rate (BER) of digitally modulated
signal
2
 Three techniques are used to improve Rx signal
quality and lower BER:

1) Equalization
2) Diversity
3) Channel Coding

Used independently or together

We will consider Diversity and Channel Coding

3
 These techniques improve mobile radio link
performance
Effectiveness of each varies widely in practical
wireless systems
Cost & complexity are also important issues
Complexity in mobile vs. in base station

4
III. Diversity Techniques
Diversity : Primary goal is to reduce depth &
duration of small-scale fades

Spatial or antenna diversity most common

Use multiple Rx antennas in mobile or base station
Why would this be helpful?

Even small antenna separation ( ) changes phase of

signal constructive /destructive nature is changed
Other diversity types polarization, frequency, &
time

5
 Exploits random behavior of MRC
Goal is to make use of several independent
(uncorrelated) received signal paths
Why is this necessary?

Select path with best SNR or combine multiple

paths improve overall SNR performance

6
Microscopic Diversity Techniques
These techniques exploit small-scale fading
characterized by deep and rapid amplitude
fluctuations as mobile moves over distances of
just a few wavelength
Example:
In case of small-scale fading, if two antennas are
separated by a fraction of meter, one may receive a
weak signal while other may receive a strong signal
By selecting the best signal at all times, a receiver
can mitigate small-scale fading effects
This is called antenna diversity or space diversity

7
Macroscopic Diversity Techniques
These techniques exploit large-scale fading
caused by shadowing due to variations in both
the terrain profile and the nature of the
surroundings
Example:
By selecting a base station which is not shadowed
when others are, the mobile can improve
substantially the average SNR on the forward link
Macroscopic diversity is also useful at the base
station receive

8
 Microscopic diversity combat small-scale
fading

Most widely used

Use multiple antennas separated in space

At a mobile, signals are independent if separation > / 2

But it is not practical to have a mobile with multiple
antennas separated by / 2 (7.5 cm apart at 2 GHz)
Can have multiple receiving antennas at base stations, but
must be separated on the order of ten wavelengths (1 to 5
meters).

9
 Since reflections occur near receiver, independent
signals spread out a lot before they reach the base
station.
a typical antenna configuration for 120 degree
sectoring.
For each sector, a transmit antenna is in the center,
with two diversity receiving antennas on each side.
If one radio path undergoes a deep fade, another
independent path may have a strong signal.
By having more than one path one select from, both
the instantaneous and average SNRs at the receiver
may be improved

10
The instantaneous SNR(i) is usually defined as i = instantaneous
signal power per branch/mean noise power per branch.
For Rayleigh fading channels, has a Rayleigh distribution and so
2 and consequently i have a chi-square distribution with two
degrees of freedom.

11
 Spatial or Antenna Diversity 4 basic types
M independent branches
Variable gain & phase at each branch G
Each branch has same average SNR: SNR = = Eb 2
identically Rayleigh distributed N0
is RV used to represent amplitude fading values of channel
Inst.. SNR = i = Inst Signal power per branch/Mean Noise per
branch,
Therefore, the pdf of i (Chi-Square distribution 2 degree

freedom ) 1 i
p ( i )
= e i 0 (6.155)

pdf of single branch that SNR is less than threshold
i
1
Pr [ i ] = p ( i )d i = e
d i =
1 e

0 0

12
 The probability that all M independent diversity branches Rx signal
which are simultaneously less than some specific SNR threshold
(all of them less than )
Pr [ 1 ,... M ] =(1 e / ) M =PM ( )
The prob. that at least one of branch has SNR > (CDF of fading )
Pr [ i > ] =1 PM ( ) =1 (1 e / ) M

The pdf of : pM=( )

d
d
( )
PM=
M

(1 e ) e
M 1

Average SNR improvement offered by selection diversity

( )
M 1
M
=
x
p ( ) d =
Mx 1 e e x dx, x =
0 0

M
1
=
k =1 k

13
14
 Space diversity methods:

1) Selection diversity
2) Feedback diversity
3) Maximal radio combining
4) Equal gain diversity

15
Selection Diversity

16
1) Selection Diversity simple & cheap
Rx selects branch with highest instantaneous SNR
new selection made at a time that is the reciprocal of the
fading rate
this will cause the system to stay with the current signal
until it is likely the signal has faded
SNR improvement :
is new avg. SNR
: avg. SNR in each branch

17
18
 Example:
Average SNR is 20 dB
Acceptable SNR is 10 dB
Assume four branch diversity
Determine that the probability that one signal has
SNR less than 10 dB

19
20
2) Scanning Diversity
scan each antenna until a signal is found that is above
predetermined threshold
if signal drops below threshold rescan
only one Rx is required (since only receiving one signal
at a time), so less costly still need multiple antennas

21
3) Maximal Ratio Diversity
signal amplitudes are weighted according to each
SNR
summed in-phase
most complex of all types
a complicated mechanism, but modern DSP makes
this more practical especially in the base station
Rx where battery power to perform computations is
not an issue

22
 The resulting signal envelop applied to detector:
M
rM = Gi ri
i =1

Total noise power:

M
NT = N Gi2
i =1

SNR applied to detector:

rM2
M =
2 NT

23
 The voltage signals i from each of the M diversity
branches are co-phased to provide coherent voltage
addition and are individually weighted to provide
optimal SNR

( rM is maximized when Gi = ri / N )

The SNR out of the diversity combiner is the sum of the

SNRs in each branch.

24
 The probability that M less than some specific
SNR threshold

25
 gives optimal SNR improvement :
i: avg. SNR of each individual branch
i = if the avg. SNR is the same for each branch

M M
M = i =
=i 1 =i 1
i = M

26
27
4) Equal Gain Diversity
combine multiple signals into one
G = 1, but the phase is adjusted for each received
signal so that
The signal from each branch are co-phased
vectors add in-phase
better performance than selection diversity

28
IV. Time Diversity
Time Diversity transmit repeatedly the
information at different time spacings

Time spacing > coherence time (coherence time is

the time over which a fading signal can be
considered to have similar characteristics)
So signals can be considered independent
Main disadvantage is that BW efficiency is
significantly worsened signal is transmitted more
than once
BW must to obtain the same Rd (data rate)

29
 If data stream repeated twice then either
1) BW doubles for the same Rd or
2) Rd is reduced by for the same BW

30
 RAKE Receiver
Powerful form of time diversity available in spread
spectrum (DS) systems CDMA
Signal is only transmitted once
Propagation delays in the MRC provide multiple
copies of Tx signals delayed in time

31
 attempts to collect the time-shifted versions of the
original signal by providing a separate correlation
receiver for each of the multipath signals.
Each correlation receiver may be adjusted in time
delay, so that a microprocessor controller can cause
different correlation receivers to search in different
time windows for significant multipath.
The range of time delays that a particular correlator
can search is called a search window.

32
 If time delay between multiple signals > chip
period of spreading sequence (Tc) multipath
signals can be considered uncorrelated
(independent)
In a basic system, these delayed signals only appear
as noise, since they are delayed by more than a chip
duration. And ignored.
Multiplying by the chip code results in noise
because of the time shift.
But this can also be used to our advantage, by
shifting the chip sequence to receive that delayed
signal separately from the other signals.
33
** The RAKE Rx is a time diversity Rx that collects
time-shifted versions of the original Tx signal **

34
 M branches or fingers = # of correlation Rxs
Separately detect the M strongest signals
Weighted sum computed from M branches

faded signal low weight

strong signal high weight
overcomes fading of a signal in a single branch

35
 In outdoor environments

the delay between multipath components is usually

large, the low autocorrelation properties of a
CDMA spreading sequence can assure that
multipath components will appear nearly
uncorrelated with each other.

36
 In indoor environments

RAKE receiver in IS-95 CDMA has been found to

perform poorly
since the multipath delay spreads in indoor channels
(100 ns) are much smaller than an IS-95 chip duration
( 800 ns).
In such cases, a RAKE will not work since multipath is
unresolveable
Rayleigh flat-fading typically occurs within a single chip
period.

37
Lecture 16 Review
 Question 1

2
 GMSK premodulation filter has an impulse response
given by
2 2
=
hG (t ) exp 2 t

ln 2 0.5887
exp( 2 f 2 )
HG ( f ) = ==
2B B

3
 Question 2
A digital signaling system is required to
operate at 9.6 kbps. If a signal element
encodes a 4-bit word, what is the minimum
bandwidth of the channel?
 Solution
Channel capacity (C) = 9.6 kbps = 9600 bps. A
signal element encodes a 4-bit word. That
means log2M = 4. Bandwidth (B) = ? Then by
Nyquists formula, C = 2Blog2(M) Hence, 9600
= 2B * 4 = 8B i.e. B = 9600/8 = 1200 Hz. = 1.2
KHz. Hence the minimum bandwidth of the
channel = 1.2 KHz
 Question 3
Given a channel with intended capacity of 20
Mbps, the bandwidth of the channel is 3 MHz.
What signal-to-noise ratio is required to
achieve this capacity?
 Solution
Capacity (C) = 20 Mbps = 20*106 bps.
Bandwidth (B) = 3 MHz = 3*106 Hz. Signal-to
noise ratio (SNR) = ?
By Shannons capacity formula, C = B log2
(1+SNR) i.e. 20*106 = 3*106 * log2(1+SNR)
Therefore log2(1+SNR) = 20/3 i.e. log(1+)/
2 = 20/3
i.e. log (1+SNR) = 20/2 3 = 2.01
 Question 4
Determine the possible Bit-rate and the
number of Levels over a channel for these
cases?
a. BW = 2.4 kHz, noiseless channel with L=16.
b. BW = 2.4 kHz, SNR = 20 dB
BW = 3.0 kHz, SNR = 40 dB.
Solution
 Question 5
A mobile radio channel Multipath delay spread
= 1s, Doppler spread Bd = 0.02Hz, Channel
bandwidth B= 5Hz, Symbol duration Ts = 10s.
a. Determine the coherence bandwidth and the
coherence time.
b. Does this channel exhibit frequency selective
fading? Explain.
c. Does this channel exhibit slow or fast fading?
Explain.
Solution
 Question 6
Determine the maximum and minimum
spectral frequencies received from a
stationary GSM transmitter that has a center
frequency of exactly 1950MHz, assuming that
the receiver is traveling at speeds of: a)
1Km/hr; b) 100Km/hr
Solution
Solution contd..
 Question 7
If a total of 33MHz of bandwidth is allocated to a
particular FDD cellular telephone system which
uses two 25kHz simplex channels to provide full
duplex voice and control channels, computer the
number of channels available per cell if a system
uses: (a) 4-cell reuses, (b) 7-cell reuses, (c) 12 cell
reuse. If 1 MHz of the allocated spectrum is
dedicated to control channels, determine an
equitable distribution of control channels and
voice channels is each cell for each of the three
systems.
 Solution
Given: Total bandwidth=33MHz
Channel bandwidth=25 kHz2 simplex channels=50 kHz/duplex channel
Total available channels = 33000/50 = 660 channels.
A 1 MHz spectrum for control channels implies that there are 1000/50 =
20 control channels out of the 660 channels available.
In practice, only the 640 voice channels would be allocated since the
control channels are allocated separately as 1 per cell.
Here, the 640 voice channels must be equitable distributed to each cell
within the cluster.
For N = 4 , we can have 5 control channels and 160 voice channels per cell.
In practice, however, each cell only needs a single control channel (the
controls have a greater reuse distance than the voice channels). Thus, one
control channel and 160 voice channels would be assigned to each cell.
For N=7, each cell have one control channel, four cells would have 91
voice channels, and three cells would have 92 voice channels.
(491+392=640)
For N=12, each cell would have one control channel, 8 cells would have 53
voice channels, and, 4 cells would have 54 voice channels.
(853+454=640)
 A CDMA system delivers message to 20 users
with equal power. The message is transmitted
at a rate of 5 kbps, using a DS spread
spectrum signal operating at a chip rate of 1
MHz. (a) Find the message bit period and PN
code period (b) Calculate the processing gain.
 Message bit period=1/5k=0.2ms
PN code period=1/1M= 1 micros
PG = tm/tc = Bs/Bm = 1M/5k = 200
 A cellular system has a total RF bandwidth of 25
MHz and simplex channel spacing of 60 kHz. The
system is required to cover a total area of 7200
km2.
(a) Calculate the number of traffic channels/cell,
if the cluster size is 18.
(b) Find the reuse distance for part (a), if the area
of each cell is 16 km2.
(c) Determine the total number of cells needed to
cover entire area.
(d) How many calls can be simultaneously
processed by each cell if 16 users can share each
channel?
 Bandwidth of duplex channel = 60 x 2 = 120 k
Total No. of Duplex channels = 25 M/ 120 k = 208.
(1). No. of Traffic channels/cell = 208/18 = 11
(2). Area of cell = 2.598 x R2 = 16 therefore R =
2.48 km
Reuse distance = D = sqrt(3N) x R = 18.22 km
(3). total cell needed = 7200/16= 450
(4). No of calls = 16 x 11 = 176. (16* 9)
 A certain city has an area of 1,300 square miles and is
covered by a cellular system using a 7-cell reuse pattern.
Each cell has a radius of 4 miles and the city is allocated 40
MHz of spectrum with a full duplex channel bandwidth of
60 kHz. Assume a grade of service (GOS) of 2% for an Erlang
B system is specified. If the offered traffic per user is 0.02
Erlangs, compute:
(a) The number of cells in the service area, the number of
channels per cell, and per cluster,
(b) Traffic intensity of each cell, the maximum carried
traffic,
(c) The total number of users that can be served,
(d) The number of mobiles per channel,
(e) The theoretical maximum number of users that could be
served at one time by the system.
 (a) Given: Total coverage area = 1300 miles. Cell radius = 4 miles . The area of a cell
(hexagon) can be shown to be , thus each cell covers 2.5981 x (4) 2= 41.57 sq mi.
Hence, the total number of cells are N = 1300/41.57 = 31 cells.
The total number of channels per cell C= allocated spectrum / ( channel width
frequency reuse factor ) = 40, 000,000/ (60,000 x 7) = 95 channels/cell
Given: C = 95, and GOS = 0.02
From the Erlang B chart, we have traffic intensity per cell A = 84 Erlangs/cell
Maximum carried traffic = number of cells x traffic intensity per cell =3184=2604
Erlangs
Given traffic per user=0.02 Erlangs Total number of users = Total traffic / traffic per
users= 2604/0.02 = 130200 users.
Number of mobiles per channel = number of users/number of channels
= 130200/665 = 195 mobiles/channel.
The theoretical maximum number of served mobiles is the number of available
channels in the system (all channels occupied) = C x N C = 95 x 31 = 2945 users,
which is 3.4% of the customer base.
 Consider a wireless system where the power received
at a distance of 100 meters is equal to 0.1 W and
calculate the following
(a) The power received at a distance equal to 500
meters assuming the path loss exponent in the range
from 100 meters to 500 meters is equal to 2.
(b) The power received at a distance of 1000 meters
assuming the power exponent at the distances in range
from 500 meters to 2000 meters is equal to 3
 In the figure Q4, assume PT= 10W, GT= 10dB,
GR = 3dB at 900MHz. The received power for
the diffracted signal (PrLoS + Gd in dB) is -70
dBm. (a). Calculate the height of obstruction
above ground. (b). How many complete
Fresnel Zones does the obstruction cover?
 (1). Consider a cellular system with 7-cell cluster and a
total of 630 full duplex voice channels available in the
system. For each the following cases determine the
optimized distribution of voice channels. (a) Omni-
directional antennas are used in each cell. (b) 120
sectoring is used (c) 60 sectoring is used
(2). Consider a cellular system having 5 Macro-cells
with each cell having 60 channels assigned to it. (a)
Find the total number of channels in the system. (b)
Repeat the above if all of the Macro-cells are split into
4 microcells (c) The system capacity, if each Micro-cell
is further divided into four Femto-cells (further cell
splitting of Micro-cells)
1. The antenna transmits in all directions therefore we get 630/7 =
90 channels.
2. The antenna would transmit only in a 120 degree sector, therefore
we would get 90/3 = 30 channels from 1 antenna.
3. The antenna would transmit only in a 60 degree sector, therefore
we would get 90/6 = 15 channels from 1 antenna
4. For A, Number of channels per cell = 60
Number of cells = 5
Therefore the number of channels = 60 * 5 = 300
For B, now the number of cells becomes 4 * 5 = 20
Number of channels per cell is same
Therefore, the no of channels = 20 * 60 = 1200
Similarly for C no of channels = 1200 * 4 = 4800
 Theory Questions
Compare Constant Envelope modulation with Amplitude
modulation
What are some of the major advantages and disadvantages of
Constant Envelope modulation
Compare MSK and QPSK
What is Power Spectral Density (PSD) of modulated Bandpass
signal. How Bandwidth is dependent on PSD
Compare different types of Amplitudue modulation techniques
What is M-ary digital modulation. Briefly discuss major M-ary
modulation schemes.
Compare different types of PSK and M-ary schemes.
What is Spread spectrum modulation. What are its major
advantages and disadvantages ?
Define Processing Gain in Spread Spectrum communication?
What are different types of spread spectrum modulation? What are
its major advantages and disadvantages?
 What is Modulation? Why it is requires in communication?
Explain FSK modulation and compare it with PSK modulation?
Discuss in brief M-ary FSK. What are its major advantages and
disadvantages?
What is QAM. What are its major advantages and disadvantages?
Explain Processing Gain SS-modulation? How it helps in combating
narrowband interference? What are different types of receiver antenna
diversity techniques used in Wireless communication? Compare and
contrasts different diversity techniques.
Compare Digital and Analog modulation?
What is Offset QPSK modulation? Why it is required?
Explain pi/4 QPSK modulation scheme. Discuss its advantages and
disadvantages.
Explain time diversity? Mention its major advantage and disadvantage
 What is Muti-carrier Communication? How it can help in combating
ISI?
Explain major difference between FDM and OFDM.
What is Coherence Bandwidth and frequency selective fading?
How they are related to each other?
Draw block diagram and explain each part of OFDM TX and RX.
Briefly describe GMSK modulation technique. What are its major
advantages over MSK modulation.
Describe briefly different types of small scale fading encountered in
wireless communication?
Compare BPSK, Pi/4 QPSK, Offset QPSK and QPSK modulation in
terms of Bandwidth, Power efficiency and BER?
Discuss QAM in brief
Explain Microscopic and Macroscopic diversity techniques