Professional Documents
Culture Documents
CSE311 1stPartMerged
CSE311 1stPartMerged
Data Communication
Instructor:
Dr. Md. Monirul Islam
Course Outline
• Introduction to digital data • Analog Modulation
communication; • Pulse modulation
• Introduction to signals • Inter symbol interference
• Review of Fourier Transform • Pulse shaping
• Frequency Response of • Digital modulation
Linear Systems; • Multiple access techniques:
• Sampling theorem; TDM, FDM; Random processes;
Quantization; • Additive White Gaussian Noise
• Line coding; (AWGN);
• Error rate due to noise
• Introduction to information
theory; Concept of channel
coding and capacity.
Course Objectives
• Learn continuous and discrete-time signals, linear systems and their
properties
• Learn and apply Fourier Transform for analyzing signals and linear
systems,
• Investigate the concepts of modulation, multiplexing and multiple access
• have in-depth knowledge and understanding of Data Comm. Techniques
• identify and compare pros and cons of different CoE techniques
Course Outcomes
CO Statement Domains and Delivery
CO Correspondi
After undergoing this course, students Taxonomy Method(s) and Assessment Tool(s)
No. ng PO(s)*
should be able to: level(s)** Activity(-ies)
Understand and analyze continuous
Class Tests or
and discrete-time signals, linear
Assignments or
CO1 systems, their properties, and the PO1, PO2 C2, C4 Lecture
Projects, and Final
sampling process in time and frequency
Exam
domains
Class Tests or
Apply mathematical concepts for Assignments or
CO2 PO1 C3 Lecture
analysis of signals and linear systems Projects, and Final
Exam
Class Tests or
Understand and analyze underlying
Assignments or
CO3 notions of analog and digital - C2, P7 Lecture
Projects, and Final
communication systems.
Exam
Evaluate and analyze different Class Tests or
modulation, multiplexing and multiple PO2, PO4, Assignments or
CO4 C6, A5 Lecture
access principals, practices and PO5 Projects, and Final
measures. Exam
Text Books
1. Modern Digital and Analog Communication Systems,
International 4th edition, B P Lathi and Zhi Ding
2. An Introduction to Analog and Digital Communications,
2nd edition, S. Haykin and M. Moher
Assessment
• Slow
• Unreliable
Communication systems
• Continued from ancient times
– Runners
– Carrier pigeons
– Light, mirror, lens
– Smoke and fire
• Slow
• Unreliable
• Email
• E-commerce
• Intercontinental business meeting, zoom, team, and many more…
Communication system
breakthroughs
• Telegraph (1844, Samuel Morse)
– Transmitted “What hath God wrought” between
Washington, D.C. and Baltimore, Maryland.
Communication system
breakthroughs
• Telephone (1875, Alexander Graham Bell)
– Made real-time transmission of speech by
electrical
encoding and
replication
of sound
Communication system
breakthroughs
• Radio (1901,Marconi ) and TV (1925, Jenkins)
broadcasting
• 1906:
– AM radio
• 1935:
– FM radio
• 1953
– Color TV
• Recent innovation: internet radio/TV
Communication system
breakthroughs
• Satellite Communication
– 1955: John R. Pierce proposed
– 1957: Sputnik I launched by Soviet U transmitted
telemetry signals for 21 days.
– 1962: Telstar I launched by Bell Lab.
• Capable of relaying
TV programs
across
the Atlantic.
Communication system
breakthroughs
• Cellular Communication
– 1947: cellular concept first proposed at Bell labs.
– 1978: first cellular trial by AT&T.
– 1991:first GSM
cellular service
launched in Finland.
– 1996: 1st commercial
CDMA cellular
service launched
in Finland.
Communication system
breakthroughs
• Wireless LAN
– 1971: 1st wireless
computer network
1997- to date: defined
a no. of different IEEE
Wireless LAN
standards.
Communication system
breakthroughs
• Internet: Network of Networks
– 1971: ARPANET put into service.
• Later renamed the Internet, in 1985.
– 1990: Tim Berners-Lee
proposed a Hypermedia
software interface
to
Internet
WWW
Components of
Communication systems
Distortion
and Noise
Components of
Communication systems
Distortion
and Noise
Main challenges
Challenges of
Communication systems
1. Distortion
–systematic undesirable changes in signals
–Linear or non-linear
2. Noise
– Unwanted signal that interfere with the transmitted
signal
– Random signals from internal or external sources
Components of
Communication systems
g(t)
amplitude
t
time
CSE 311:
Data Communication
Instructor:
Dr. Md. Monirul Islam
Messages/Signals: Definition
- A signal is a set of information or data.
- A signal is a function of independent variables
that carry some information.
- A signal is a physical quantity that varies with
time, space or any other independent variable by
which information can be conveyed.
Example of Signals
• Voice signal
• Telephone or television signal
• Monthly sales figure
• Opening or closing stock prices
• Charge density over a surface
- In this course we deal with signals that are functions
of time.
Signal representation: Time Domain
g(t)
amplitude
t
time
Classification of Signals
• Based on continuity in time axis
– Continuous time
– Discrete time
• Based on continuity in amplitude axis
– Continuous amplitude
– Discrete amplitude
Classification of Signals
Continuous amplitude Discrete amplitude
Continuous time
Discrete time
Analog and Digital Signal
Analog Signal
- Continuous amplitude, i.e., takes any value in a
continuous range.
- May be both continuous and discrete time.
Digital Signal
- Discrete amplitude, i.e., amplitude can take only
a finite number of values.
- Values need not be always integer.
- Not necessarily always binary, rather M-ary.
- May be both continuous and discrete time.
Analog and Digital Signal: Examples
Analog Digital
Thermometer
Clock
Blood Pressure
Monitor
Components of
Communication systems
Distortion
and Noise
Main challenges
Challenges of
Communication systems
1. Distortion
–systematic undesirable changes in signals
–Linear or non-linear
2. Noise
– Unwanted signal that interfere with the transmitted
signal
– Random signals from internal or external sources
Benefits of Digital
Message/Signal
1. Quality, e.g., enhanced noise immunity
2. Economics
Benefits of Digital
Message/Signal
1. noise immunity
• Represented by binary or M-ary pulses
Sent
Benefits of Digital
Message/Signal
1. noise immunity
• Represented by binary or M-ary pulses
Sent
Received
w/o noise
Benefits of Digital
Message/Signal
1. noise immunity
• Represented by binary or M-ary pulses
Sent
Received
w/o noise
Received
with noise
Benefits of Digital
Message/Signal
1. noise immunity
• Recovered despite small distortion and noises
Sent
Received
w/o noise
Received
with noise
Recovered,
though delayed
Repeater’s Role in Digital
Message/Signal
• Distortion and noise are unavoidable in channel
• Repeaters and nodes regenerates digital pulses
Repeater’s Role in Digital
Message/Signal
• Distortion and noise are unavoidable in channel
• Repeaters and nodes regenerates digital pulses
Channel
Channel repeater
Channel amplifier
Channel amplifier
Remember
Nyquist’s theorem!
Analog to Digital Conversion
of Message/Signal
• Increasing Quantization level L
– increases accuracy
– more noise immunity
– but requires higher channel bandwidth
Analog to Digital Conversion
of Message/Signal
Sent
Received
w/o noise
Received
with noise
Recovered,
though delayed
Analog to Digital Conversion
of Message/Signal
• Detection is easy when A >> noise
Recall this figure • Usually A >> 5-10 times of noise
Sent
Received
w/o noise
Received
with noise
Recovered,
though delayed
Analog to Digital Conversion
of Message/Signal
• Problem: quantization error is unavoidable
Analog to Digital Conversion
of Message/Signal
• Problem: quantization error is unavoidable
Analog to Digital Conversion
of Message/Signal
• Problem: quantization error is unavoidable
Analog to Digital Conversion
of Message/Signal
• Problem: quantization error is unavoidable
Analog to Digital Conversion
of Message/Signal
• Problem: quantization error is unavoidable
Analog to Digital Conversion
of Message/Signal
• Problem: quantization error is unavoidable
Quantization
error can be
minimized
increasing L
Representation of Digital
Signal: Pulse Coded Modulation
2m p
L
-mp
Representation of Digital
Signal: Pulse Coded Modulation
• Assume, L =16
mp
m(t) Quantized samples m(t)
Allowed quantization levels
2m p
L
-mp
Representation of Digital
Signal: Pulse Coded Modulation
• Assume, L =16
– Each quantized value is represented
by a sequence of FOUR binary
pulses.
CSE 311:
Data Communication
Instructor:
Dr. Md. Monirul Islam
Channel & signal
Characteristics
– Signal power, Ps
Channel & signal
Characteristics
• Bandwidth
– Channel Bandwidth
– Signal bandwidth
Channel & signal
Characteristics
• Bandwidth
– Channel Bandwidth
• Range of frequencies that a channel can transmit
• Ex: if transmit 0-5 kHz frequencies, Channel bandwidth B = 5 kHz
– Signal bandwidth
Channel & signal
Characteristics
• Bandwidth
– Channel Bandwidth
• Range of frequencies that a channel can transmit
• Ex: if transmit 0-5 kHz frequencies, Channel bandwidth B = 5 kHz
– Signal bandwidth
• Maximum frequency that is available in a signal
Channel & signal
Characteristics
• Bandwidth
– Channel Bandwidth
• Range of frequencies that a channel can transmit
• Ex: if transmit 0-5 kHz frequencies, Channel bandwidth B = 5 kHz
– Signal bandwidth
• Maximum frequency that is available in a signal
Fourier Transform
f
Frequency
Time domain
domain
Channel & signal
Characteristics
• Bandwidth
– Channel Bandwidth
• Range of frequencies that a channel can transmit
• Ex: if transmit 0-5 kHz frequencies, Channel bandwidth B = 5 kHz
– Signal bandwidth
• Maximum frequency that is available in a signal
maximum
frequency
f
Channel & signal
Characteristics
• Bandwidth
– Channel Bandwidth
• Range of frequencies that a channel can transmit
• Ex: if transmit 0-5 kHz frequencies, Channel bandwidth B = 5 kHz
– Signal bandwidth B
• Maximum frequency that is available in a signal
maximum
frequency
f
Channel & signal
Characteristics
Compressed Expanded in
in time frequency
Channel & signal
Characteristics
• Frequency
– Compressing in time, increases frequency, means higher channel
Bandwidth
N pulses/sec
Channel with B
Channel & signal
Characteristics
• More transmission speed requires channel with higher
bandwidth
N pulses/sec
Channel with B
N pulses/sec
Channel with B
NK pulses/sec
Channel & signal
Characteristics
• More transmission speed requires channel with higher
bandwidth
N pulses/sec
Channel with B
NK pulses/sec
New requirement:
Channel with KB
Channel & signal
Characteristics
• Signal Power, Ps
– Dual role
• Higher Quality
• Less channel bandwidth
Analog to Digital Conversion
of Message/Signal
• Detection is easy when A >> noise
• Usually A >> 5-10 times of noise
Sent
Received
w/o noise
Received
with noise
Recovered,
though delayed
Channel & signal
Characteristics
Baseband signal
Channel & signal
Characteristics
Baseband signal 1
Baseband signal 2
Channel & signal
Characteristics
• Assume, these 2 signals need
to be transmitted
Baseband signal 1
Baseband signal 2
Channel & signal
Characteristics
• Assume, these 2 signals need
to be transmitted
– 2 different channels?
Baseband signal 1
Baseband signal 2
Channel & signal
Characteristics
• Assume, these 2 signals need
to be transmitted
– Even, channel bandwidth and
signal bandwidth may NOT
match!!
Baseband signal 1
Baseband signal 2
Channel & signal
Characteristics
• Assume, these 2 signals need
to be transmitted
– 2 different channels?
Baseband signal 1
Solution is
modulation
Baseband signal 2
Channel & signal
Characteristics: modulation
• Use carrier signal to shift these
2 signals to different frequency
positions
Baseband signal 1
Baseband signal 2
Channel & signal
Characteristics: modulation
• Use carrier signal to shift these
2 signals to different frequency
positions
Baseband Baseband
After signal 1 signal 2
modulation:
Frequency
axis
0 f1 f2
Channel & signal
Characteristics: modulation
• other examples of modulation
(shown in time domain)
Baseband
signal
Carrier signal
Modulated
signal
Channel & signal
Characteristics: modulation
• other examples of modulation
(shown in time domain)
Baseband
signal
Carrier signal
Frequency
Modulated modulation
signal (FM)
Channel & signal
Characteristics: modulation
• other examples of modulation
(shown in time domain)
Baseband
signal
Carrier signal
Amplitude
Modulated
signal
modulation
(AM)
Channel & signal
Characteristics: modulation
• Phase modulation (PM)
• Changes phase angle of the signal
Channel & signal
Characteristics: modulation
Other reasons for modulation
Higher Generating
Low frequency
wavelength antenna requires
higher dimension
Channel & signal
Characteristics: modulation
Other reasons for modulation
Higher Generating
Low frequency
wavelength antenna requires
higher dimension
Lower Generating
Modulating a High
wavelength antenna requires
frequency carrier
lower dimension
Channel & signal
Characteristics: modulation
Other reasons for modulation
Lower Generating
Modulating a High
wavelength antenna requires
frequency carrier
lower dimension
Baseband Baseband
After signal 1 signal 2
modulation:
Frequency
axis
0 f1 f2
Channel & signal
Characteristics: modulation
• Use carrier signal to shift these
2 signals in different frequency
positions
also called
Frequency division
multiplexing (FDM) Baseband
Baseband
After signal 1 signal 2
modulation:
Frequency
axis
0 f1 f2
Channel & signal
Characteristics: modulation
• Time division multiplexing
(TDM)
– Interleave pulses from different
signals in time domain signal
Channel & signal
Characteristics: modulation
• Time division multiplexing
(TDM)
– Interleave pulses from different
signals in time domain signal
Channel & signal
Characteristics: DeModulation
• Done at the receiving end
– Bandpass filter separates appropriate signal
– Makes necessary corrections for amplitude, frequency
and phase changes
CSE 311:
Data Communication
Instructor:
Dr. Md. Monirul Islam
Channel & signal
Characteristics: modulation
• Use carrier signal to shift these
2 signals in different frequency
positions
also called
Frequency division
multiplexing (FDM) Baseband
Baseband
After signal 1 signal 2
modulation:
Frequency
f1 f2 axis
0
Channel & signal
Characteristics: modulation
• Time division multiplexing
(TDM)
– Interleave pulses from different
signals in time domain signal
Channel & signal
Characteristics: DeModulation
• Done at the receiving end
– Bandpass filter separates appropriate signal
– Makes necessary corrections for amplitude, frequency
and phase changes
Signal Characteristics: digital
source coding and error
correction coding
• 2 opposite procedures
No NO news Nothing to
randomness at all transmit
No NO news Nothing to
randomness at all transmit
Topics
Periodic
signal
Period, T0 =2
Aperiodic
signal
Periodic and Aperiodic signal
• Periodic signal starts at t = -α and continues forever
Periodic
signal
Period, T0 =2
Periodic and Aperiodic signal
• Periodic signal starts at t = -α and continues forever
• cannot start at an finite time, say, t = 0, otherwise
g(t) = g(t + T0) cannot be satisfied
Periodic and Aperiodic signal
• Periodic signal starts at t = -α and continues forever
• cannot start at an finite time, say, t = 0, otherwise
g(t) = g(t + T0) cannot be satisfied
Zero or Undefined at
this region
Started here
Periodic and Aperiodic signal
• Periodic signal starts at t = -α and continues forever
• cannot start at an finite time, say, t = 0, otherwise
g(t) = g(t + T0) cannot be satisfied
Zero or Undefined at
this region
1 T 2
2
0 Pg lim g (t )dt
T T T 2
Energy and Power signals
2
E g g (t )dt
1 T2 2
0 Pg lim g (t )dt
T T T 2
• Finite Eg signal has zero Pg
• Finite Pg Signal has infinite Eg
• A signal CANNOT be both energy and power signal
• Real life signals are energy signals
• Power signals have infinite duration; impractical to
generate
• Periodic signals are power signals
Deterministic and Random
signals
• Deterministic
– has complete physical description, mathematically or
graphically
• Random
– has only probabilistic description, e.g., mean value, rms,
distribution
• All message signals are random
Signal Properties
Time shifting property
Or,
Signal Properties
Time shifting property
Beginning T
seconds later
Beginning T
seconds earlier
Signal Properties
Time scaling property
• Compression or expansion
• Compression:
– Whatever happens in g(t) at t second also happens in (t)
at t/a
,a>1
Signal Properties
Time scaling property
• Compression or expansion
• Compression:
– Whatever happens in g(t) at t second also happens in (t)
at t/a
,a>1
• Expansion:
– Whatever happens in g(t) at t second also happens in (t)
at at
,a>1
Signal Properties
Time scaling property
Compression
Expansion
Signal Properties
Time inversion property
if t 0
(t )
0 if t 0
(t )dt 1
Unit Impulse Signal
A function with unit area under curve
Unit Impulse Signal
Compress the function leaving
the area unchanged
Unit Impulse Signal
Keep compressing. . . . .
Unit Impulse Signal
Keep compressing. . . . .
Ultimately we got Dirac Delta function
(t)
t
Dirac Delta
Function
Unit Impulse Signal
if t 0
(t )
0 if t 0 (t)
(t )dt 1
t
Unit Impulse Signal
if t 0
(t )
0 if t 0 (t)
(t )dt 1
t
• Impulse location is at t =0
Multiplication of a Function
by Impulse
(t ) (t ) (0) (t )
(t)
t
(t )
Multiplication of a Function
by Impulse
(t ) (t T ) (T ) (t T ) (t-T)
t
T
(t )
Sampling Property of the Unit
Impulse Function
(0)
t
(t )
Sampling Property of the Unit
Impulse Function
(t-T)
(t T ) (t )dt
t
(T ) (t T )dt T
(t )
(T )
Sampling Property of the Unit
Impulse Function
b (t-T)
(t T ) (t )dt
a
b t
T
(T ) (t T )dt (t )
a
Sampling Property of the Unit
Impulse Function
b
(t-T)
(t T ) (t )dt
a
b a b
t
(T ) (t T )dt T
(t )
a
(T )
Sampling Property of the Unit
Impulse Function
b
(t T ) (t )dt
a
(t-T)
b
(T ) (t T )dt a b
t
a T
(t )
(T ) a T b
0 otherwise
CSE 311:
Data Communication
Instructor:
Dr. Md. Monirul Islam
Signals and systems
Topics
1 if t 0
u (t )
0 if t 0
Unit step signal/function, u(t)
• Causal function, e,g., starts at t = 0
1 if t 0
u (t )
0 if t 0
• g(t) is causal if
Use of u(t) to start an
everlasting function from t = 0
e-at
a>0
t
Use of u(t) to start an
everlasting function from t = 0
e-at
×
t
Use of u(t) to start an
everlasting function from t = 0
Alternate representation of u(t)
in terms of (t)
Alternate representation of u(t)
in terms of (t)
()
Alternate representation of u(t)
in terms of (t)
=1 ()
t
Alternate representation of u(t)
in terms of (t)
=0 ()
t
Alternate representation of u(t)
in terms of (t)
()
t
Alternate representation of u(t)
in terms of (t)
()
t
0 if t 0
u (t )
1 if t 0
Alternate representation of u(t)
in terms of (t)
()
t
0 if t 0
u (t )
1 if t 0
Alternate representation of u(t)
in terms of (t)
()
t
Alternate representation of u(t)
in terms of (t)
(t)
t
Fourier Series of a periodic signal
‘Any function that periodically repeats itself can be
expressed as the sum of sines and/or cosines of
different frequencies’
- Joseph Fourier
Fourier Series of a periodic signal
‘Any function that periodically repeats itself can be
expressed as the sum of sines and/or cosines of
different frequencies’
- Joseph Fourier
n
jn0t
g (t ) n
D e
n with, T0 = period of g(t)
n f0 = frequency of g(t)
jn 2f 0 t
D e
n
n
1 jn 2f 0 t
where, Dn g (t )e
T0 T0
Fourier Series
Fourier says:
Simpler
sine/cosine
functions
Fourier Series and Light Spectrum
A complicated Simpler
function/signal functions/signals
Representing a Periodic signal using
Fourier Series
Representing a Periodic signal using
Fourier Series
n n
jn0t j 2 nt
(t ) n
D e
n
n
D e
n
n
j 2 nt
(t ) n
D e
n
Representing a Periodic signal using
Fourier Series
• Dn are exponential Fourier spectra
• Dn are complex though (t) is a real periodic
• Dn and D-n are complex conjugates
Representing a Periodic signal using
Fourier Series
• Dn are complex though (t) is a real periodic
–
Representing a Periodic signal using
Fourier Series
• Dn are complex though (t) is a real periodic
–
–
Representing a Periodic signal using
Fourier Series
• Dn are complex though (t) is a real periodic
–
–
• Therefore,
Representing a Periodic signal using
Fourier Series
Im Im
e j 0 t e j 0 t
Re Re
Fourier series of Some Useful Signals
Periodic
square wave
Period, T0 = 2
Frequency, f0 = 1/T0 = 1/ 2
Fourier series of Some Useful Signals
Periodic
square wave
Period, T0 = 2
Frequency, f0 = 1/T0 = 1/ 2
where, ω0 =2f0
Fourier series of Some Useful Signals
Periodic
square wave
Period, T0 = 2
Frequency, f0 = 1/T0 = 1/2
Fourier series of Some Useful Signals
Periodic
square wave
Period, T0 = 2
Frequency, f0 = 1/T0 = 1/2
Fourier series of Some Useful Signals
Periodic
square wave
Period, T0 = 2
Frequency, f0 = 1/T0 = 1/2
Fourier series of Some Useful Signals
Periodic
square wave
Period, T0 = 2
Frequency, f0 = 1/T0 = 1/2
Fourier series of Some Useful Signals
Periodic
square wave
Period, T0 = 2
Frequency, f0 = 1/T0 = 1/2
D0 is undefined
Fourier series of Some Useful Signals
Periodic
square wave
Period, T0 = 2
Frequency, f0 = 1/T0 = 1/2
1 1
D0 w(t )dt
T0 T0 2
Fourier series of Some Useful Signals
1
D0
2
Exponential
spectra of a
Periodic
square wave
f
Fourier series of Some Useful Signals
1
D0
2
Exponential
spectra of a
Periodic
square wave
CSE 311:
Data Communication
Instructor:
Dr. Md. Monirul Islam
Signals and systems
Topics
t
Impulse train
g (t ) T0 (t )
t
Impulse train
g (t ) T0 (t )
where,
Fourier Series of Unit
Impulse Train
g (t ) T0 (t )
(t)
t
Fourier Series of Unit
Impulse Train
g (t ) T0 (t )
(t)
t
Fourier Series of Unit
Impulse Train
g (t ) T0 (t )
(t)
t
Fourier Series of Unit
Impulse Train
g (t ) T0 (t )
g(t)
t
Fourier’s theorem for
aperiodic Signal
• aperiodic signal but with finite area under curve can be
represented as integral of sines and cosines
j t
g (t ) G ( f )e df
jt
where, G ( f ) g (t ) e dt
and ω = 2f
Fourier’s theorem for
aperiodic Signal
• These two are called Fourier transform pair
j t j 2ft
G( f ) g (t ) e dt g (t ) e dt
jt j 2ft
g (t ) G ( f )e df G ( f )e df
Fourier’s theorem for
aperiodic Signal
• These two are called Fourier transform pair
j t j 2ft
G( f ) g (t ) e dt g (t ) e dt
jt j 2ft
g (t ) G ( f )e df G ( f )e df
j g ( f )
G( f ) G( f ) e
Fourier Integral for
aperiodic signal
j t j 2ft
G( f ) g (t )e dt g (t )e dt
• Replacing f by -f
j 2ft
G ( f ) g (t ) e dt
Fourier Integral for
aperiodic signal
j t j 2ft
G( f ) g (t )e dt g (t )e dt
• Replacing f by -f
j 2ft
G ( f ) g (t ) e dt
Complex
conjugate
* j 2ft
G (f) g (t )e dt provided, g(t) is real
Fourier Integral for
aperiodic signal
j t j 2ft
G( f ) g (t )e dt g (t )e dt
• Replacing f by -f
j 2ft
G ( f ) g (t ) e dt
* j 2ft
G (f) g (t )e dt G ( f )
Fourier Integral for
aperiodic signal
• The following is complex conjugate symmetry
G * ( f ) G ( f )
Fourier Integral for
aperiodic signal
• The following is complex conjugate symmetry
G * ( f ) G ( f )
Therefore,
*
G ( f ) G ( f ) G ( f )
Fourier Integral for
aperiodic signal
• The following is complex conjugate symmetry
G * ( f ) G ( f )
Therefore,
*
G ( f ) G ( f ) G ( f )
but
g ( f ) g ( f )
Conjugate Symmetry Property in
Fourier Series of Periodic signals
• Dn are complex though (t) is a real periodic
–
–
• Therefore,
Fourier Transform Example
Find Fourier transform of e-atu(t)
g(t)
e-atu(t)
0 t
Fourier Transform Example
Find Fourier transform of e-atu(t)
j 2ft at j 2ft at j 2ft
G( f ) g (t ) e dt e u (t ) e dt e dt
e
0
g(t)
e-atu(t)
0 t
Fourier Transform Example
Find Fourier transform of e-atu(t)
j 2ft at j 2ft at j 2ft
G( f ) g (t ) e dt e u (t ) e dt e dt
e
0
g(t) e ( a j 2f ) t dt
0
e-atu(t)
0 t
Fourier Transform Example
Find Fourier transform of e-atu(t)
j 2ft at j 2ft at j 2ft
G( f ) g (t ) e dt e u (t ) e dt e dt
e
0
1
g(t) e ( a j 2f ) t
dt
a j 2f
e ( a j 2f )t
0
0
e-atu(t)
0 t
Fourier Transform Example
Find Fourier transform of e-atu(t)
j 2ft at j 2ft at j 2ft
G( f ) g (t ) e dt e u (t ) e dt e dt
e
0
1
e ( a j 2f ) t
dt
a j 2f
e ( a j 2f )t
0
0
1 1
G( f )
a j a j 2f
Fourier Transform Example
Find Fourier transform of e-atu(t)
1 1
G( f )
a j a j 2f
Therefore,
and
Fourier Transform Example
Find Fourier transform of e-atu(t)
Fourier Transform Example
Remember, we got this result for a > 0
g(t)
1
e-atu(t) G( f )
a j 2f
e ( a j 2f )t
0
1
0 t a j 2f
Now as t->α, e-(a+j2f )t= e-at e-j2f)t = 0 provided a > 0
Fourier Transform Example
What happens if result for a < 0?
g(t)
1
e-atu(t) G( f )
a j 2f
e ( a j 2f )t
0
1
0 t a j 2f
Now as t->α, e-(a+j2f )t= e-at e-j2f)t = 0 provided a > 0
Existence of Fourier Transform
Conditions
• Area under curve must be finite
• The function must be integrable
Dirichlet condition
Existence of Fourier Transform
Conditions
• Area under curve must be finite
• The function must be integrable
Then,
a1 g1 (t ) a2 g 2 (t ) a1G1 ( f ) a2G2 ( f )
Linearity of Fourier Transform
Assume, g1 (t ) G1 ( f ) and g 2 (t ) G2 ( f )
and a1 and a2 are arbitrary constants
Then,
a1 g1 (t ) a2 g 2 (t ) a1G1 ( f ) a2G2 ( f )
This means,
a g
k
k k (t ) ak Gk ( f )
k
Linearity of Fourier Transform
More general case
a g
k
k k (t ) ak Gk ( f )
k
Superposition Theorem
Area Under Curve of Fourier
Transform G(f)
j 2ft
g (t ) G ( f )e df
Area Under Curve of Fourier
Transform G(f)
j 2ft
g (t ) G ( f )e df
Assume t = 0
g (0) G ( f )e j 2f .0 df G ( f )df
Area Under Curve of Time domain
Function g(t)
j 2ft
G( f ) g (t )e dt
Area Under Curve of Time domain
Function g(t)
j 2ft
G( f ) g (t )e dt
Assume f = 0
j 2t .0
G ( 0) g (t ) e dt g (t )dt
Fourier Transform of Some Useful
Functions
Unit Rectangular Function
Fourier Transform of Some Useful
Functions
Unit Rectangular Function
Other names
Box Function
Gate Function
Fourier Transform of Some Useful
Functions
Unit Rectangular Function
Expanded
function
Fourier Transform of Some Useful
Functions
Unit Triangular Function
, ,
, ,
Fourier Transform of Some Useful
Functions
Sinc Function, sinc (x)
sin x
x
Fourier Transform of Some Useful
Functions
Sinc Function, sinc (x)
sin x
x
Fourier Transform of Some Useful
Functions
Sinc Function, sinc (x)
= g(t)
g(t) =
Fourier Transform of Some Useful
Functions
Unit scaled rectangular function
= g(t)
g(t) =
Fourier Transform of Some Useful
Functions
Unit scaled rectangular function
= g(t)
g(t) =
Fourier Transform of Some Useful
Functions
Unit scaled rectangular function
= g(t)
g(t) =
Fourier Transform of Some Useful
Functions
Unit scaled rectangular function
= g(t)
Instructor:
Dr. Md. Monirul Islam
Signals and systems
Topics
= g(t)
g(t) = (t)
t
Fourier Transform of Some Useful
Functions
Unit Impulse function (t)
g(t) = (t)
t
Fourier Transform of Some Useful
Functions
Unit Impulse function (t)
G(f) = 1
g(t) = (t) 1
f
t 0
G(f) = (f)
f
Fourier Transform of Some Useful
Functions
Inverse Fourier Transform of (f)
G(f) = (f)
g (t ) 1[ ( f )]
G ( f )e j 2ft df ( f )e j 2ft df
j 2t .0
e 1
Fourier Transform of Some Useful
Functions
Inverse Fourier Transform of (f)
g(t) = 1
1 G(f) = (f)
f
0 t
g (t ) 1[ ( f )]
G ( f )e j 2ft df ( f )e j 2ft df Fourier Transform Pair
j 2t .0
e 1
Fourier Transform of Some Useful
Functions
Inverse Fourier Transform of (f-f0)
G(f) = (f-f0)
f
f0
Fourier Transform of Some Useful
Functions
Inverse Fourier Transform of (f-f0)
G(f) = (f-f0)
g (t ) 1[ ( f f 0 )]
f
j 2ft j 2ft f0
G ( f )e df ( f f 0 )e df
e j 2tf 0 e j 2f 0t
Fourier Transform of Some Useful
Functions
Inverse Fourier Transform of (f-f0)
G(f) = (f-f0)
g (t ) 1[ ( f f 0 )]
f
j 2ft j 2ft f0
G ( f )e df ( f f 0 )e df
e j 2tf 0 e j 2f 0t
Fourier Transform Pair
j 2f 0t
e ( f f0 )
Fourier Transform of Some Useful
Functions
Inverse Fourier Transform of (f-f0)
G(f) = (f-f0)
j 2f 0t
e ( f f0 )
f
f0
Fourier Transform of Some Useful
Functions
Inverse Fourier Transform of (f-f0) and (f+f0)
G(f) = (f-f0)
j 2f 0t
e ( f f0 )
f
f0
G(f) = (f+f0)
j 2f 0 t
e? ( f f0 )
f
-f0
Fourier Transform of Some Useful
Functions
Inverse Fourier Transform of (f-f0) and (f+f0)
G(f) = (f-f0)
j 2f 0t
e ( f f0 )
f
f0
G(f) = (f+f0)
j 2f 0 t
e ( f f0 )
f
-f0
t
1
cos 2f 0t [ ( f f 0 ) ( f f 0 )]
2
Fourier Transform of Some Useful
Functions
Fourier Transform of sine function
g(t)
1
e-atu(t) G( f )
a j 2f
e ( a j 2f )t
0
1
0 t a j 2f
Now as t->α, e-(a+j2f )t= e-at e-j2f)t = 0 provided a > 0
Fourier Transform of Some Useful
Functions
Fourier Transform of exponential functions
u (t )e at u (t )e at
1
a j 2f
Fourier Transform Pair
Exponential rising fnc
exp(at )u (t )
Fourier Transform of Some Useful
Functions
Fourier Transform of sgn(t) function
sgn(t)
+1
0 t
-1
Fourier Transform of Some Useful
Functions
Fourier Transform of sgn(t) function
g(t)
+1
t
0
-1
Can be approximated by
Fourier Transform of Some Useful
Functions
Fourier Transform of sgn(t) function
g(t)
+1
t
0
-1
Can be approximated by
Fourier Transform of Some Useful
Functions
Fourier Transform of sgn(t) function
g(t)
+1
t
0
Limiting a to 0
-1
Fourier Transform of Some Useful
Functions
Fourier Transform of sgn(t) function
g(t)
+1
t
0
Limiting a to 0
-1
-T0/2 +T0/2
Fourier Transform of Some Useful
Functions
Fourier Transform of a periodic signal gT0 (t )
gT0 (t )
-T0/2 +T0/2
g (t )
-T0/2 +T0/2
Fourier Transform of Some Useful
Functions
Fourier Transform of a periodic signal gT0 (t )
gT0 (t )
-T0/2 +T0/2
g (t )
Assume that
g (t ) G ( f )
-T0/2 +T0/2
Fourier Transform of Some Useful
Functions
Fourier Transform of a periodic signal gT0 (t )
gT0 (t )
-T0/2 +T0/2
g (t )
Assume that
g (t ) G ( f )
-T0/2 +T0/2
Fourier Transform of Some Useful
Functions
Fourier Transform of a periodic signal gT0 (t )
gT0 (t )
-T0/2 +T0/2
g (t )
Assume that
g (t ) G ( f )
-T0/2 +T0/2
Fourier Transform of Some Useful
Functions
Fourier Transform of a periodic signal gT0 (t )
gT0 (t )
-T0/2 +T0/2
n
As it is periodic,
gT0 (t ) c
n
n exp( j 2nf 0t )
T0 2
1
where, cn g T0 (t ) exp( j 2nf 0t )dt and
T0 T0 2
Fourier Transform of Some Useful
Functions
Fourier Transform of a periodic signal gT0 (t )
gT0 (t )
-T0/2 +T0/2
T0 2
1
cn g T0 (t ) exp( j 2nf 0t )dt
T0 T0 2
Fourier Transform of Some Useful
Functions
Fourier Transform of a periodic signal gT0 (t )
gT0 (t )
-T0/2 +T0/2
T0 2
1 g (t )
cn
T0 gT0 (t ) exp( j 2nf 0t )dt
T0 2
-T0/2 +T0/2
Fourier Transform of Some Useful
Functions
Fourier Transform of a periodic signal gT0 (t )
gT0 (t )
-T0/2 +T0/2
T0 2 T0 2
1 1
cn g (t ) exp( j 2nf t ) dt
T0 T0 2
T0 0
T0 g (t ) exp( j 2nf t )dt
T0 2
0
Fourier Transform of Some Useful
Functions
Fourier Transform of a periodic signal gT0 (t )
g (t )
-T0/2 +T0/2
T0 2 T0 2
1 1
cn g (t ) exp( j 2nf t ) dt
T0 T0 2
T0 0
T0 g (t ) exp( j 2nf t )dt
T0 2
0
Fourier Transform of Some Useful
Functions
Fourier Transform of a periodic signal gT0 (t )
g (t )
-T0/2 +T0/2
T0 2 T0 2
1 1
cn g (t ) exp( j 2nf t ) dt
T0 T0 2
T0 0
T0 g (t ) exp( j 2nf t )dt
T0 2
0
Fourier Transform of Some Useful
Functions
Fourier Transform of a periodic signal gT0 (t )
gT0 (t )
-T0/2 +T0/2
T0 2
1
cn
T0
g
T0 2
T0 (t ) exp( j 2nf 0t )dt
-T0/2 +T0/2
n
gT0 (t ) c
n
n exp( j 2nf 0t ) and we found, cn = f0G(nf0)
Fourier Transform of Some Useful
Functions
Fourier Transform of a periodic signal gT0 (t )
gT0 (t )
-T0/2 +T0/2
n
gT0 (t ) c
n
n exp( j 2nf 0t ) and we found, cn = f0G(nf0)
Fourier Transform of Some Useful
Functions
Fourier Transform of a periodic signal gT0 (t )
gT0 (t )
-T0/2 +T0/2
n n
gT0 (t ) f 0 G (nf 0 ) exp( j 2nf 0t ) f 0 G (nf 0 )exp( j 2nf 0t )
n n
n
f0 G(nf
n
0 ) ( f nf 0 )
Fourier Transform of Some Useful
Functions
Fourier Transform of a periodic signal gT0 (t )
n
gT0 (t ) f 0 G(nf 0 ) ( f nf 0 ) g (t ) G ( f )
n
Instructor:
Dr. Md. Monirul Islam
Signals and systems
Topics
(t)
• T0 (t ) is a periodic signal with period T0
t
Impulse train T (t ) Impulse signal
0
t
-2T0 -T0 0 T0 2T0 3T0
n
FT of Impulse train
T0 (t ) f 0 ( f nf 0 )
n
f
-4f0 -3f0 -2f0 -f0 0 f0 2f0 3f0 4f0 5f0 6f0
Fourier Transform of Some Useful
Functions
a>0
a>0
a>0
a>0
a>0
Properties of Fourier Transform
Time-Frequency duality
Properties of Fourier Transform
Time-Frequency duality
• Difference is minor
• Remarkably similar
Properties of Fourier Transform
Time-Frequency duality
• Difference is minor
• Remarkably similar
Let, g (t ) G ( f )
Let we have
Properties of Fourier Transform
Time-Frequency duality
Let, g (t ) G ( f )
Let we have
This means
Properties of Fourier Transform
Duality property
If g (t ) G ( f )
then G (t ) g ( f )
Properties of Fourier Transform
Duality property Proof
then G (t ) g ( f )
Properties of Fourier Transform
Duality property Proof
then G (t ) g ( f )
Properties of Fourier Transform
Duality property Proof
G (t )
Properties of Fourier Transform
Duality property Proof
This means, G (t ) g ( f )
Properties of Fourier Transform
Use of Duality property G (t ) g ( f )
t
g (t ) ( ) G(f) = sinc (f)
Properties of Fourier Transform
Use of Duality property G (t ) g ( f )
t
g (t ) ( ) G(f) = sinc (f)
f f
g ( f ) ( ) ( )
Properties of Fourier Transform
Use of Duality property G (t ) g ( f )
t
g (t ) ( ) G(f) = sinc (f)
f f
g ( f ) ( ) ( )
Properties of Fourier Transform
Use of Duality property G (t ) g ( f )
t
g (t ) ( ) G(f) = sinc (f)
f
g ( f ) ( )
G(t) = sinc (t)
f
t
Properties of Fourier Transform
1 f
Time scaling or dilation g ( at ) G( )
property a a
Properties of Fourier Transform
1 f
Time scaling or dilation g ( at ) G( )
property a a
Properties of Fourier Transform
1 f
Time scaling or dilation g ( at ) G( )
property a a
1 f
F g (at ) G( ) assuming a < 0
a a
Properties of Fourier Transform
1 f
Time scaling or dilation g ( at ) G( )
property a a
1 f
F g (at ) G( ) assuming a < 0
a a
1 f
Therefore, g (at ) G ( )
a a
Properties of Fourier Transform
1 f
Significance of Time g (at ) G ( )
scaling property a a
t
g (t ) ( ) G(f) = 2 sinc (2f)
2
Properties of Fourier Transform
Reflection property g (t ) G ( f )
1 f
g (at ) G ( )
a a
Properties of Fourier Transform
Time shifting property g (t t0 ) G ( f ) exp( j 2ft0 )
Properties of Fourier Transform
Time shifting property g (t t0 ) G ( f ) exp( j 2ft0 )
F g (t t0 ) g (t t 0 ) e j 2ft
dt
Properties of Fourier Transform
Time shifting property g (t t0 ) G ( f ) exp( j 2ft0 )
F g (t t0 ) g (t t 0 ) e j 2ft
dt
j 2f ( t 0 ) assuming t =t0 +
g ( ) e d
Properties of Fourier Transform
Time shifting property g (t t0 ) G ( f ) exp( j 2ft0 )
F g (t t0 ) g (t t 0 ) e j 2ft
dt
j 2f ( t 0 ) assuming t =t0 +
g ( ) e d
e j 2ft0 g ( ) e j 2f
d
e j 2ft0 G ( f )
Properties of Fourier Transform
Time shifting property g (t t0 ) G ( f ) exp( j 2ft0 )
F g (t t0 ) g (t t 0 ) e j 2ft
dt
j 2f ( t 0 ) assuming t =t0 +
g ( ) e d
e j 2ft0 g ( ) e j 2f
d
e j 2ft0 G ( f )
Properties of Fourier Transform
Time shifting property g (t t0 ) G ( f ) exp( j 2ft0 )
Lower frequency
signal
Higher frequency
signal
Properties of Fourier Transform
Time shifting property g (t t0 ) G ( f ) exp( j 2ft0 )
Lower frequency
signal
Lower frequency
signal, phase shift =/2
F g (t )e j 2f 0t
g (t )e j 2f 0t j 2ft
e dt
j 2 ( f f 0 ) t
g (t )e
dt
G( f f0 )
Properties of Fourier Transform
Frequency shifting g (t ) exp( j 2f 0t ) G ( f f 0 )
property
F g (t )e j 2f 0t
g (t )e j 2f 0t j 2ft
e dt
j 2 ( f f 0 ) t
g (t )e
dt
G( f f0 )
Properties of Fourier Transform
Frequency shifting g (t ) exp( j 2f 0t ) G ( f f 0 )
property
F g (t )e j 2f 0t
g (t )e j 2f 0t j 2ft
e dt
j 2 ( f f 0 ) t
g (t )e
dt
G( f f0 )
Properties of Fourier Transform
Frequency shifting g (t ) exp( j 2f 0t ) G ( f f 0 )
property
F g (t )e j 2f 0t
g (t )e j 2f 0t j 2ft
e dt
j 2 ( f f 0 ) t
g (t )e
dt
because, j 2ft
G( f f0 ) G( f ) g (t ) e dt
Properties of Fourier Transform
Frequency shifting g (t ) exp( j 2f 0t ) G ( f f 0 )
property
Replacing f0 by –f0
g (t ) exp( j 2f 0t ) G ( f f 0 )
Properties of Fourier Transform
Frequency shifting g (t ) exp( j 2f 0t ) G ( f f 0 )
property
Significance:
Multiplying g(t) by exp(j2f0t) shifts the spectrum by f0 to a new
location
Properties of Fourier Transform
Frequency shifting g (t ) exp( j 2f 0t ) G ( f f 0 )
property
Significance:
Multiplying g(t) by exp(j2f0t) shifts the spectrum by f0 to a new
location
Recall FDM!
Properties of Fourier Transform
Frequency shifting g (t ) exp( j 2f 0t ) G ( f f 0 )
property
Significance:
Multiplying g(t) by exp(j2f0t) shifts the spectrum by f0 to a new
location
Significance:
As exp(j2f0t) is complex, g(t) is multiplied by a real sinusoid
cos(2f0t)
Properties of Fourier Transform
Frequency shifting g (t ) exp( j 2f 0t ) G ( f f 0 )
property
Significance:
Multiplying g(t) by exp(j2f0t) shifts the spectrum by f0 to a new
location
Significance:
As exp(j2f0t) is complex, g(t) is multiplied by a real sinusoid
cos(2f0t)
1 j 2f 0t
But, cos( 2f 0t ) (e e j 2f 0t )
2
Properties of Fourier Transform
Frequency shifting g (t ) exp( j 2f 0t ) G ( f f 0 )
property
1 j 2f 0t j 2f 0t
Therefore, g (t ) cos(2f 0t ) ( g (t )e g (t )e )
2
Properties of Fourier Transform
Frequency shifting g (t ) exp( j 2f 0t ) G ( f f 0 )
property
1 j 2f 0t j 2f 0t
Therefore, g (t ) cos(2f 0t ) ( g (t )e g (t )e )
2
1
This results, g (t ) cos(2f 0t ) [G ( f f 0 ) G ( f f 0 )]
2
Properties of Fourier Transform
Frequency shifting g (t ) exp( j 2f 0t ) G ( f f 0 )
property
1
g (t ) cos(2f 0t ) [G ( f f 0 ) G ( f f 0 )]
2
Significance:
g(t)cos(2πf0t)
CSE 311:
Data Communication
Instructor:
Dr. Md. Monirul Islam
Signals and systems
Topics
g(t)cos(2πf0t)
Properties of Fourier Transform
Frequency shifting g (t ) exp( j 2f 0t ) G ( f f 0 )
property
1
g (t ) cos(2f 0t ) [G ( f f 0 ) G ( f f 0 )]
2
Further phase change of modulated signal by 0:
1
g (t ) cos( j 2f 0t ) [G ( f f 0 ) G ( f f 0 )]
2
Significance:
Time Convolution
g1 (t ) g 2 (t ) G1 ( f )G2 ( f )
g1 (t ) g 2 (t ) G1 ( f ) G2 ( f )
Properties of Fourier Transform
What is Convolution?
g1 (t ) g 2 (t ) g ( ) g
1 2 (t )d
Properties of Fourier Transform
What is Convolution?
g1 (t ) g 2 (t ) g ( ) g
1 2 (t )d
g1 (t ) g 2 (t ) g 2 (t ) g1 (t )
Properties of Fourier Transform
What is Convolution?
g1 (t ) g 2 (t ) g ( ) g
1 2 (t )d
g1 (t ) g 2 (t ) g 2 (t ) g1 (t )
g1 (t ) g 2 (t ) g ( ) g
1 2 (t )d g
2 ( ) g1 (t )d
Properties of Fourier Transform
Time Convolution g1 (t ) g 2 (t ) G1 ( f )G2 ( f )
Property
j 2ft
g1 (t ) g 2 (t ) g1 ( ) g 2 (t )d e dt
j 2ft
g1 ( ) g 2 (t )e dt d
Properties of Fourier Transform
Time Convolution g1 (t ) g 2 (t ) G1 ( f )G2 ( f )
Property
j 2ft
g1 (t ) g 2 (t ) g1 ( ) g 2 (t )d e dt
j 2ft
g1 ( ) g 2 (t )e dt d
Properties of Fourier Transform
Time Convolution g1 (t ) g 2 (t ) G1 ( f )G2 ( f )
Property
j 2ft
g1 (t ) g 2 (t ) g1 ( ) g 2 (t )d e dt
j 2ft
g1 ( ) g 2 (t )e dt d
j 2ft j 2f
g
2 ( t ) e dt is FT of - time shifted g 2 (t ), e. g ., G 2 ( f ) e
Properties of Fourier Transform
Time Convolution g1 (t ) g 2 (t ) G1 ( f )G2 ( f )
Property
j 2ft
g1 (t ) g 2 (t ) g1 ( ) g 2 (t )d e dt
j 2ft
g1 ( ) g 2 (t )e dt d
j 2f j 2f
g (
1 2 )G ( f ) e d G 2 ( f ) g
1 ( ) e d
G2 ( f ) g1 (t )e j 2ft dt G2 ( f )G1 ( f )
Properties of Fourier Transform
Time Convolution g1 (t ) g 2 (t ) G1 ( f )G2 ( f )
Property
j 2ft
g1 (t ) g 2 (t ) g1 ( ) g 2 (t )d e dt
j 2ft
g1 ( ) g 2 (t )e dt d
j 2f j 2f
g (
1 2 )G ( f ) e d G 2 ( f ) g
1 ( ) e d
G2 ( f ) g1 (t )e j 2ft dt G2 ( f )G1 ( f )
Properties of Fourier Transform
Time Convolution g1 (t ) g 2 (t ) G1 ( f )G2 ( f )
Property
j 2ft
g1 (t ) g 2 (t ) g1 ( ) g 2 (t )d e dt
j 2ft
g1 ( ) g 2 (t )e dt d
j 2f j 2f
g (
1 2 )G ( f ) e d G 2 ( f ) g
1 ( ) e d
G2 ( f ) g1 (t )e j 2ft dt G2 ( f )G1 ( f )
Properties of Fourier Transform
Modulation or frequency g1 (t ) g 2 (t ) G1 ( f ) G2 ( f )
Convolution Property
Properties of Fourier Transform
Modulation or frequency g1 (t ) g 2 (t ) G1 ( f ) G2 ( f )
Convolution Property
j 2ft
G1 ( f ) G2 ( f ) G1 ( )G2 ( f )d e df
1
j 2ft
G1 ( ) G2 ( f )e df d
G1 ( ) g 2 (t )e j 2t d g 2 (t ) G1 ( )e j 2t d
g 2 (t ) G1 ( f )e j 2ft df g 2 (t ) g1 (t )
Properties of Fourier Transform
Modulation or frequency g1 (t ) g 2 (t ) G1 ( f ) G2 ( f )
Convolution Property
j 2ft
G1 ( f ) G2 ( f ) G1 ( )G2 ( f )d e df
1
j 2ft
G1 ( ) G2 ( f )e df d
G1 ( ) g 2 (t )e j 2t d g 2 (t ) G1 ( )e j 2t d
g 2 (t ) G1 ( f )e j 2ft df g 2 (t ) g1 (t )
Properties of Fourier Transform
Modulation or frequency g1 (t ) g 2 (t ) G1 ( f ) G2 ( f )
Convolution Property
j 2ft
G1 ( f ) G2 ( f ) G1 ( )G2 ( f )d e df
1
j 2ft
G1 ( ) G2 ( f )e df d
G1 ( ) g 2 (t )e j 2t d g 2 (t ) G1 ( )e j 2t d
g 2 (t ) G1 ( f )e j 2ft df g 2 (t ) g1 (t )
Properties of Fourier Transform
Modulation or frequency g1 (t ) g 2 (t ) G1 ( f ) G2 ( f )
Convolution Property
j 2ft
G1 ( f ) G2 ( f ) G1 ( )G2 ( f )d e df
1
j 2ft
G1 ( ) G2 ( f )e df d
G1 ( ) g 2 (t )e j 2t d g 2 (t ) G1 ( )e j 2t d
g 2 (t ) G1 ( f )e j 2ft df g 2 (t ) g1 (t )
Properties of Fourier Transform
t
Time Integration G( f ) 1
Property g ( )d j 2f 2 G(0) ( f )
Properties of Fourier Transform
t
Time Integration G( f ) 1
Property g ( )d j 2f 2 G(0) ( f )
Recall 1 if t 0
u (t )
0 if t 0
Properties of Fourier Transform
t
Time Integration G( f ) 1
Property g ( )d j 2f 2 G(0) ( f )
Recall 1 if t 0
u (t )
0 if t 0
Therefore,
1 if t 0 or t
u (t )
0 if t 0 or t
Properties of Fourier Transform
t
Time Integration G( f ) 1
Property g ( )d j 2f 2 G(0) ( f )
u(t-) u(t-)
1 1
0 t
- t 0
1 if t 0
u (t )
1 if t 0 if t 0
u (t )
0 if t
Properties of Fourier Transform
t
Time Integration G( f ) 1
Property g ( )d j 2f 2 G(0) ( f )
u(t-)
g (t ) u (t ) g ( )u (t )d
1
t
g ( )u (t )d g ( )u (t )d
t
t t - t 0
g ( ).1d g ( ).0d g ( )d
t
1 if t 0
u (t )
0 if t 0
Properties of Fourier Transform
t
Time Integration G( f ) 1
Property g ( )d j 2f 2 G(0) ( f )
u(t-)
g (t ) u (t ) g ( )u (t )d
1
t
g ( )u (t )d g ( )u (t )d
t
t t - t 0
g ( ).1d g ( ).0d g ( )d
t
1 if t 0
u (t )
0 if t 0
Properties of Fourier Transform
t
Time Integration G( f ) 1
Property g ( )d j 2f 2 G(0) ( f )
u(t-)
g (t ) u (t ) g ( )u (t )d
1
t
g ( )u (t )d g ( )u (t )d
t
t t - t 0
g ( ).1d g ( ).0d g ( )d
t
1 if t 0
u (t )
0 if t 0
Properties of Fourier Transform
t
Time Integration G( f ) 1
Property g ( )d j 2f 2 G(0) ( f )
t
g (t ) u (t ) g ( )d
Properties of Fourier Transform
t
Time Integration G( f ) 1
Property g ( )d j 2f 2 G(0) ( f )
t
g (t ) u (t ) g ( )d
t
g ( )d g (t ) u (t ) G( f )U ( f )
Properties of Fourier Transform
u(t) and sgn(t) function
sgn(t)
u(t)
+1
1
0 t
0 t
-1
1 if t 0
u (t )
0 if t 0
Properties of Fourier Transform
u(t) and sgn(t) function
sgn(t)
u(t)
+1
1
0 t
0
t
-1
1 if t 0
u (t )
0 if t 0
1
Therefore, u (t ) sgn(t ) 1
2
Properties of Fourier Transform
u(t) and sgn(t) function
1
u (t ) sgn(t ) 1
2
1 1 1 1
Therefore, U ( f )
( f )
(f )
2 jf j 2f 2
Properties of Fourier Transform
t
Time Integration G( f ) 1
Property g ( )d j 2f 2 G(0) ( f )
t
g (t ) u (t ) g ( )d
t
g ( )d g (t ) u (t ) G( f )U ( f )
1 1 1 1
U ( f ) ( f ) (f )
2 jf j 2f 2
Properties of Fourier Transform
t
Time Integration G( f ) 1
Property g ( )d j 2f 2 G(0) ( f )
t
g (t ) u (t ) g ( )d
t
g ( )d g (t ) u (t ) G( f )U ( f )
G( f ) 1
G ( f ) ( f )
j 2f 2
G( f ) 1
G (0) ( f )
j 2f 2
Properties of Fourier Transform
Time Differentiation dg (t )
Property j 2fG ( f )
dt
Properties of Fourier Transform
Time Differentiation dg (t )
Property j 2fG ( f )
dt
g (t ) G ( f )e j 2ft df
Properties of Fourier Transform
Time Differentiation dg (t )
Property j 2fG ( f )
dt
g (t ) G ( f )e j 2ft df
dg (t ) d
G ( f )e j 2ft df j 2fG ( f )e j 2ft df
dt dt
Properties of Fourier Transform
Time Differentiation dg (t )
Property j 2fG ( f )
dt
g (t ) G ( f )e j 2ft df
dg (t ) d
G ( f )e j 2ft df j 2fG ( f )e j 2ft df
dt dt
dg (t )
j 2fG ( f )
dt
Properties of Fourier Transform
Time Differentiation dg (t )
Property j 2fG ( f )
dt
d n g (t ) n
n
j 2f G( f )
dt
Properties of Fourier Transform
Time Differentiation
d n g (t ) n
Property to find FT of n
j 2f G( f )
complex signal dt
Properties of Fourier Transform
Time Differentiation
d n g (t ) n
Property to find FT of n
j 2f G( f )
complex signal dt
Properties of Fourier Transform
Time Differentiation
d n g (t ) n
Property to find FT of n
j 2f G( f )
complex signal dt
Properties of Fourier Transform
Time Differentiation
d n g (t ) n
Property to find FT of n
j 2f G( f )
complex signal dt
Let, (g(t)) G ( f )
Properties of Fourier Transform
Time Differentiation
d n g (t ) n
Property to find FT of n
j 2f G( f )
complex signal dt
Properties of Fourier Transform
Time Differentiation
d n g (t ) n
Property to find FT of n
j 2f G( f )
complex signal dt
Properties of Fourier Transform
Time Differentiation
d n g (t ) n
Property to find FT of n
j 2f G( f )
complex signal dt
Properties of Fourier Transform
Time Differentiation
d n g (t ) n
Property to find FT of n
j 2f G( f )
complex signal dt
Time Domain and Frequency Domain
Representation
f
t
Indefinite spread in time Limited in frequency
Time Domain and Frequency Domain
Representation
• A signal cannot be BOTH band limited and time limited
• Relation is inverse
• Relation changes in inverse manner
• Any change in time specification changes spectrum
speciation inversely
• We can specify EITHER of them, NOT BOTH of them
CSE 311:
Data Communication
Instructor:
Dr. Md. Monirul Islam
Signals and systems
Topics
f
t
Indefinite spread in time Limited in frequency
Time Domain and Frequency Domain
Representation
• A signal cannot be BOTH band limited and time limited
• Relation is inverse
• Relation changes in inverse manner
• Any change in time specification changes spectrum
speciation inversely
• We can specify EITHER of them, NOT BOTH of them
Bandwidth of a Signal
Lowpass signal: B
Lowpass signal: B
B
Bandpass signal:
Frequency spectrum is centered
around ±f0
Bandwidth of a Signal
Lowpass signal: B
Lowpass signal: B
B
Shifting a low pass signal to f0
location DOUBLES its
bandwidth
3-dB Bandwidth of a Signal
B
Lowpass signal: 1
G (0f )
Separation betn 2
Bandpass signal: 1
B G ( f0)
Separation betn two positions 2
where amplitudes are 1 of its
2
peak
rms Bandwidth of a Signal
Root mean square bandwidth:
the square root of the second moment of a properly
normalized form of the squared amplitude spectrum of the
signal about a suitably chosen point
rms Bandwidth of a Signal
Root mean square bandwidth:
the square root of the second moment of a properly
normalized form of the squared amplitude spectrum of the
signal about a suitably chosen point
k 2
kth-moment
origin
of squared amplitude about f
G ( f ) df
rms Bandwidth of a Signal
Root mean square bandwidth:
the square root of the second moment of a properly
normalized form of the squared amplitude spectrum of the
signal about a suitably chosen point
2 2
2nd moment of squared amplitude about m2 f G ( f ) df
origin
2
0-th moment of squared amplitude about m0 G( f ) df
origin
rms Bandwidth of a Signal
Root mean square bandwidth:
the square root of the second moment of a properly
normalized form of the squared amplitude spectrum of the
signal about a suitably chosen point
2 2
t B
g (t ) ( ) sinc (f)=
2 2 2 2
t
g (t ) dt f
G ( f ) df
Trms
Wrms
2 2
g (t )
dt G( f )
df
Time-Bandwidth Product
1
Trms wrms
4
Time-Bandwidth Product
1
Trms wrms
4
y(t) = S(x(t))
Signal Transmission through
Linear Time Invariant System
System
y(t) = S(x(t))
y1(t) = S(x1(t))
y2(t) = S(x2(t))
Signal Transmission through
Linear Time Invariant System
Linear System
y(t) = LS(x(t))
y1(t) = LS(x1(t)) y2(t) = LS(x2(t))
Signal Transmission through
Linear Time Invariant System
Linear System
y(t) = LS(x(t))
y1(t) = LS(x1(t)) y2(t) = LS(x2(t))
Principle of Superposition
the response of a linear system to a number of excitations applied
simultaneously is equal to the sum of the responses of the system when
each excitation is applied individually.
Signal Transmission through
Linear Time Invariant System
Linear System
Principle of Superposition
the response of a linear system to a number of excitations applied
simultaneously is equal to the sum of the responses of the system when
each excitation is applied individually.
Examples
Filters and Communication Channels
Signal Transmission through
Linear Time Invariant System
Time Invariant System
Time Invariant
x(t) System, TIS y(t)
y(t) = TIS(x(t))
Signal Transmission through
Linear Time Invariant System
Time Invariant System
Time Invariant
x(t) System, TIS y(t)
y(t) = TIS(x(t))
y1(t) = TIS(x1(t))
y2(t) = TIS(x2(t))
Signal Transmission through
Linear Time Invariant System
Time Invariant System
Time Invariant
x1(t-ta) System, TIS y1(t-ta)
y(t) = TIS(x(t))
y1(t-ta) = TIS(x1(t-ta)) y2(t-tb) = TIS(x2(t-tb))
Signal Transmission through
Linear Time Invariant System
Linear Time Invariant System
Linear Time
x(t) Invariant System, y(t)
LTI
y(t) = LTI(x(t))
Signal Transmission through
Linear Time Invariant System
Linear Time Invariant System
Linear Time
x(t) Invariant System, y(t)
LTI
y(t) = LTI(x(t))
y1(t) = LTIS(x1(t))
y2(t) = LTIS(x2(t))
Signal Transmission through
Linear Time Invariant System
Linear Time Invariant System
Linear Time
a1x1(t-ta) + a2x2(t-tb) Invariant System, a1y1(t-ta) + a2y2(t-tb)
LTI
y(t) = LTI(x(t))
a1y1(t-ta) + a2y2(t-tb) = LTI(a1x1(t-ta) + a2x2(t-tb))
Signal Transmission through
Linear Time Invariant System
Linear Time Invariant System
Linear Time
Invariant System,
x(t) y(t)
h(t)
Instructor:
Dr. Md. Monirul Islam
Signals and systems
Topics
y(t) = LS(x(t))
y1(t) = LS(x1(t)) y2(t) = LS(x2(t))
Time Invariant
x1(t-ta) System, TIS y1(t-ta)
y(t) = TIS(x(t))
y1(t-ta) = TIS(x1(t-ta)) y2(t-tb) = TIS(x2(t-tb))
Signal Transmission through
Linear Time Invariant System
Linear Time Invariant System
Linear Time
a1x1(t-ta) + a2x2(t-tb) Invariant System, a1y1(t-ta) + a2y2(t-tb)
LTI
y(t) = LTI(x(t))
a1y1(t-ta) + a2y2(t-tb) = LTI(a1x1(t-ta) + a2x2(t-tb))
Signal Transmission through
Linear Time Invariant System
Linear Time Invariant System
Linear Time
Invariant System,
x(t) y(t)
h(t)
Linear Time
Invariant System,
x(t) = (t) y(t) = h(t)
h(t)
Linear Time
Invariant System,
x(t-t0) y(t-t0)
h(t)
Linear Time
Invariant System,
(t-t0) h(t-t0)
h(t)
Linear Time
Invariant System,
a(t-t0) ah(t-t0)
h(t)
Linear Time
Invariant System,
a1(t-t1) + a2(t-t2) a1h(t-t1) + a2h(t-t2)
h(t)
Linear Time
Invariant System,
a (t t
k k ) a h(t t
k k )
k h(t) k
Linear Time
Invariant System,
x(t) y(t)
h(t)
t
t
x()
t
x()
0 1 2 . . . n
t
x()
0 1 2 . . . n
t
x()
t
x()
t
Linear Time
Invariant System,
x(t) δ(t nτ)x(nτ)τ
n h(t)
Signal Transmission through
Linear Time Invariant System
Linear Time Invariant System
Linear Time
Invariant System,
x(t) δ(t nτ)x(nτ)τ y(t) h(t nτ)x(nτ)τ
n h(t) n
Signal Transmission through
Linear Time Invariant System
Linear Time Invariant System
Linear Time
Invariant System,
x(t) δ(t nτ)x(nτ)τ y(t) h(t nτ)x(nτ)τ
n h(t) n
As approaches zero,
y (t ) h(t nτ)x(nτ)τ h(t τ ) x( τ )τ h(t τ ) x( τ )dτ
n τ
Signal Transmission through
Linear Time Invariant System
Linear Time Invariant System
Linear Time
Invariant System,
x(t) δ(t nτ)x(nτ)τ y(t) h(t nτ)x(nτ)τ
n h(t) n
As approaches zero,
y (t ) h(t nτ)x(nτ)τ h(t τ ) x( τ )τ h(t τ ) x( τ )dτ
n τ
Signal Transmission through
Linear Time Invariant System
Linear Time Invariant System
Linear Time
Invariant System,
x(t) δ(t nτ)x(nτ)τ y(t) h(t τ)x(nτ)τ
n h(t) n
y (t ) h(t τ ) x( τ )dτ is convolution integral and can be written as
y (t ) h(t ) x(t )
Signal Transmission through
Linear Time Invariant System
Linear Time Invariant System
Linear Time
Invariant System,
x(t ) y (t ) h(t ) x(t )
h(t)
Signal Transmission through
Linear Time Invariant System
Linear Time Invariant System
Linear Time
Invariant System,
x(t ) y (t ) h(t ) x(t )
h(t)
Now Recall
g1 (t ) g 2 (t ) G1 ( f )G2 ( f )
Signal Transmission through
Linear Time Invariant System
Linear Time Invariant System
Linear Time
x(t ) Invariant System, y (t ) h(t ) x(t )
X(f)
h(t)
Y(f) = H(f)X(f)
H(f)
H(f) is
• Fourier transform of h(t)
• the frequency response of LTI system
• a transfer function, because it transfers or passes desired frequencies of X(f)
• a complex function
Signal Transmission through
Linear Time Invariant System
Linear Time Invariant System
Linear Time
x(t ) Invariant System, y (t ) h(t ) x(t )
X(f)
h(t)
Y(f) = H(f)X(f)
H(f)
H ( f ) H ( f ) e j h ( f )
• |H(f)| is amplitude response of LTI system
• h(f) is phase response of LTI system, measured in radians
Signal Transmission through
Linear Time Invariant System
Linear Time Invariant System
Linear Time
x(t ) Invariant System, y (t ) h(t ) x(t )
X(f)
h(t)
Y(f) = H(f)X(f)
H(f)
H ( f ) H ( f ) e j h ( f )
ln H ( f ) ln H ( f ) j h ( f ) ( f ) j h ( f )
Signal Transmission through
Linear Time Invariant System
Linear Time Invariant System
Linear Time
x(t ) Invariant System, y (t ) h(t ) x(t )
X(f)
h(t)
Y(f) = H(f)X(f)
H(f)
H ( f ) H ( f ) e j h ( f )
ln H ( f ) ln H ( f ) j h ( f ) ( f ) j h ( f )
Linear Time
x(t ) Invariant System, y (t ) h(t ) x(t )
X(f)
h(t)
Y(f) = H(f)X(f)
H(f)
H ( f ) H ( f ) e j h ( f )
( f ) 20 log10 ln H ( f ) dB
h(t)=0, t<0
Signal Transmission through
Linear Time Invariant System
Causal LTI System
Necessary and sufficient condition for causal system
where ( f ) ln H ( f )
Signal Transmission through
Linear Time Invariant System
Causal LTI System
Necessary and sufficient condition for causal system
where ( f ) ln H ( f )
Paley-Wiener Criterion
Signal Transmission through
Linear Time Invariant System
Causal LTI System
Necessary and sufficient condition for causal system
where ( f ) ln H ( f )
Paley-Wiener Criterion
Significance
• If α(f) satisfy P-W criterion, α(f) can be combined with h(f) to find h(t) for a
causal system
• Infinite attenuation (α(f) = - or |H(f)| =0 ) for a band of frequencies DOES NOT
satisfy P-W criterion
• However, α(f) = - allowed for a discrete set of frequencies
Signal Transmission through
Linear Time Invariant System
Stable LTI System
• output signal is bounded for all bounded input signals.
• The criteria is also known as bounded input–bounded output (BIBO)
stability criterion
Signal Transmission through
Linear Time Invariant System
Stable LTI System
For bounded input signal, x(t)
h(τ ) dτ
Signal Transmission through
Linear Time Invariant System
Stable LTI System
Necessary and sufficient condition for BIBO stability
h(τ ) dτ
Signal Transmission through
Linear Time Invariant System
System Response in System Response in
Frequency domain Time domain
|H(f)|
1 h(t)
f
-B B
td
Signal Transmission through
Linear Time Invariant System
System Response in System Response in
Frequency domain Time domain
|H(f)|
1 h(t)
f
-B B
td
Instructor:
Dr. Md. Monirul Islam
CT2
Tomorrow, Wednesday
Signals and systems
Topics
Linear Time
x(t ) Invariant System, y (t ) h(t ) x(t )
X(f)
h(t)
Y(f) = H(f)X(f)
H(f)
H ( f ) H ( f ) e j h ( f )
Signal Transmission through
Linear Time Invariant System
Linear Time Invariant System
Linear Time
x(t ) Invariant System, y (t ) h(t ) x(t )
X(f)
h(t)
Y(f) = H(f)X(f)
H(f)
H ( f ) H ( f ) e j h ( f )
Signal Distortion during
Transmission
Linear Time
x(t ) Invariant System, y (t ) h(t ) x(t )
X(f)
h(t)
Y(f) = H(f)X(f)
H(f)
Y ( f ) e jy ( f ) H ( f ) X ( f ) e j h ( f ) x ( f )
H ( f ) H ( f ) e j h ( f )
Signal Distortion during
Transmission
Linear Time
x(t ) Invariant System, y (t ) h(t ) x(t )
X(f)
h(t)
Y(f) = H(f)X(f)
H(f)
Y ( f ) e jy ( f ) H ( f ) X ( f ) e j h ( f ) x ( f )
• System distorts the input signal
• Input signal amplitude |X(f)| is changed to |X(f)| |H(f)|
• Input signal phase is shifted by h(f)
• Plot of |H(f)| and h(f) vs. f shows how the system changes diffrerent
frequency components differently (attenuation/boosting)
Distortionless Transmission
Linear Time
x(t ) Invariant System, y (t ) x(t )
X(f)
h(t)
H(f) Y( f ) X( f )
Linear Time
x(t ) Invariant System, y (t ) kx(t )
X(f)
h(t)
H(f) Y ( f ) kX ( f )
Linear Time
x(t ) Invariant System, y (t ) kx(t t d )
X(f)
h(t)
H(f) Y ( f ) kX ( f )e j 2ft d
Linear Time
x(t ) Invariant System, y (t ) kx(t t d )
X(f)
h(t)
H(f) Y ( f ) kX ( f )e j 2ft d
Y ( f ) kX ( f )e j 2ftd ke j 2ftd X ( f ) H ( f ) X ( f )
j 2ft d
H ( f ) ke
Distortionless Transmission
Linear Time
x(t ) Invariant System, y (t ) kx(t t d )
X(f)
h(t)
H(f) Y ( f ) kX ( f )e j 2ft d
j 2ft d
H ( f ) ke
H ( f ) k and h ( f ) 2ft d
• |H(f)| is constant
• h(f) is a linear function of f
Distortionless Transmission
|H(f)|
h(f)
H( f ) k
Y( f ) H( f ) X ( f )
• |H(f)| = k means that every frequency amplitude of |X(f)| is
passed to |Y(f)| with same gain
• Everything is present |Y(f)|, nothing is missing!
All Pass Vs. Distortionless
System
|H(f)|
H( f ) k
Y( f ) H( f ) X ( f )
• |H(f)| = k means that every frequency amplitude of |X(f)| is
passed to |Y(f)| with same gain
• Everything is present |Y(f)|, nothing is missing!
• However, this does NOT guarantee distortionless system
All Pass Vs. Distortionless
System
All pass system:
H( f ) k
Consider a real composite signal consisting of multiple sinusoids
of different frequencies
• |H(f)| = k guarantees only presence of all frequency components
H ( f ) k and h ( f ) 2ft d
1 d h ( f )
td ( f )
2 df
All Pass Vs. Distortionless
System
How to Check Distortionless transmission?
1 d h ( f )
td ( f )
2 df
1 d h ( f )
td ( f )
2 df
• If the slope of h(f) is constant, it guarantees a constant time delay
td for every sinusoid
Human ear is
• sensitive to amplitude distortion which results if |H(f)| ≠ k
Human eye is
• insensitive to amplitude distortion
• sensitive to phase distortion
Nature of Distortion in Video
Signals
|H(f)|
1
f
-B B
h(f) = -2ftd
Ideal Vs. Practical Filters
This Ideal Highpass filter
• allows transmission for frequencies f |B|
• suppresses all other frequencies f |B|
|H(f)|
f
-B B
h(f) = -2ftd
Ideal Vs. Practical Filters
This Ideal Bandpass filter
• allows transmission for the band centered at ±f0
• suppresses all other frequencies
|H(f)|
f
-f0 f0
h(f) = -2ftd
Ideal Vs. Practical Filters
This Ideal Lowpass filter
• allows transmission for frequencies f |B|
• suppresses all other frequencies f |B|
|H(f)|
1
f
-B B
h(f) = -2ftd
Ideal Vs. Practical Filters
This Ideal Lowpass filter
• allows transmission for frequencies f |B|
• suppresses all other frequencies f |B|
• Slope of h(f) is constant –td, means time delay is td
|H(f)|
1
f
-B B
h(f) = -2ftd
Ideal Vs. Practical Filters
This Ideal Lowpass filter
• allows transmission for frequencies f |B|
• suppresses all other frequencies f |B|
• Slope of h(f) is constant –td, means time delay is td
f
-B B
h(f) = -2ftd
Ideal Vs. Practical Filters
This Ideal Lowpass filter
• allows transmission for frequencies f |B|
• suppresses all other frequencies f |B|
• Slope of h(f) is constant –td, means time delay is td
f
y(t) = g(t -td) H ( f ) ( ) and
2B
|H(f)| h ( f ) 2ft d
1
f
-B B
h(f) = -2ftd
Ideal Vs. Practical Filters
This Ideal Lowpass filter
• allows transmission for frequencies f |B|
• suppresses all other frequencies f |B|
• Slope of h(f) is constant –td, means time delay is td
f
y(t) = g(t -td) H ( f ) ( ) and
2B
|H(f)| h ( f ) 2ft d
1 f j 2ftd
H ( f ) ( )e
2B
f
-B B
h(f) = -2ftd
Ideal Vs. Practical Filters
Impulse response of Ideal Lowpass filter
Frequency Transform of
Impulse response
f j 2ft d
H ( f ) ( )e
2B
|H(f)|
1
f
-B B
h(f) = -2ftd
Ideal Vs. Practical Filters
Impulse response of Ideal Lowpass filter
Frequency Transform of Impulse response
1 f j 2ftd
Impulse response h(t ) H ( f ) ( )e
1
f j 2ft d
2B
H ( f ) ( )e 2 Bsinc(2B (t t d ))
2B
|H(f)|
1
f
-B B
h(f) = -2ftd
Ideal Vs. Practical Filters
Impulse response of Ideal Lowpass filter
Frequency Transform of Impulse response
1 f j 2ftd
Impulse response h(t ) H ( f ) ( )e
1
f j 2ft d
2B
H ( f ) ( )e 2 Bsinc(2B (t t d ))
2B
|H(f)|
1 h(t)
f
-B B
td
h(f) = -2ftd
CSE 311:
Data Communication
Instructor:
Dr. Md. Monirul Islam
Signals and systems
Topics
f
-B B
h(f) = -2ftd
Ideal Vs. Practical Filters
f
-B B
h(f) = -2ftd
Ideal Vs. Practical Filters
f
-B B
td
h(f) = -2ftd
Ideal Vs. Practical Filters
Impulse response of Ideal Lowpass filter
ILP Filter in Frequency Impulse Response
1 f j 2ftd
domain h(t ) H ( f ) ( )e
1
f j 2ft d
2B
H ( f ) ( )e 2 Bsinc(2B (t t d ))
2B
|H(f)|
1 h(t)
f
-B B
td
h(f) = -2ftd
Ideal Vs. Practical Filters
Impulse response of Ideal Lowpass filter
Frequency Transform of Impulse Response
1 f j 2ftd
Impulse response h(t ) H ( f ) ( )e
1
f j 2ft d
2B
H ( f ) ( )e 2 Bsinc(2B (t t d ))
2B
|H(f)|
1 h(t)
f
-B B
td
h(f) = -2ftd
Ideal Vs. Practical Filters
Impulse response of Ideal Lowpass filter
Impulse response
h(t) h(t ) 2 Bsinc(2B(t t d ))
td
Response begins
before t = 0
Ideal Vs. Practical Filters
Impulse response of Ideal Lowpass filter
Impulse response
h(t) h(t ) 2 Bsinc(2B(t t d ))
td
Response begins
before t = 0
Noncausal and
unrealizable filter
Ideal Vs. Practical Filters
Impulse response of Ideal Lowpass filter
Impulse response
h(t) h(t ) 2 Bsinc(2B(t t d ))
Condition to be causal
td h(t) = 0 for all t < 0
Response begins
before t = 0
Noncausal and
unrealizable filter
Ideal Vs. Practical Filters
Causal and realizable Lowpass filter
td
Response begins
on or after t = 0
Ideal Vs. Practical Filters
Causal and realizable Lowpass filter
Condition in frequency
td
domain
Response begins
on or after t = 0
where ( f ) ln H ( f )
Paley-Wiener Criterion
Ideal Vs. Practical Filters
Causal and realizable Lowpass filter??
Condition in frequency
|H(f)|
1
domain
f
-B B
where ( f ) ln H ( f )
h(f) = -2ftd Paley-Wiener Criterion
Ideal Vs. Practical Filters
Causal and realizable Lowpass filter??
Condition in frequency
|H(f)|
1
domain
f
-B B
where ( f ) ln H ( f )
h(f) = -2ftd Paley-Wiener Criterion
f
-B B
where ( f ) ln H ( f )
h(f) = -2ftd Paley-Wiener Criterion
where ( f ) ln H ( f )
td
Paley-Wiener Criterion
hˆ(t ) h(t )u (t )
Ideal Vs. Practical Filters
Other Causal and realizable Lowpass filters from ILP filter
Condition in frequency
|H(f)|
1
domain
f
-B B
where ( f ) ln H ( f )
h(f) = -2ftd Paley-Wiener Criterion
f
-B B
h(f) = -2ftd
f
B
h(f) = -2ftd
f
B
j 2ft
g (t ) G ( f ) e df dt
Parseval’s Theorem: Signal
Energy in Frequency Domain
2
E g g (t ) dt
g (t ) g (t ) dt
j 2ft
G ( f ) g (t )e dt df
Parseval’s Theorem: Signal
Energy in Frequency Domain
2
E g g (t ) dt
g (t ) g (t ) dt
j 2ft
G ( f ) g (t )e dt df
G ( f ) G ( f )df
Parseval’s Theorem: Signal
Energy in Frequency Domain
2
E g g (t ) dt
g (t ) g (t ) dt
j 2ft
G ( f ) g (t )e dt df
G ( f ) G ( f )df
2
G ( f ) df
Parseval’s Theorem: Signal
Energy in Frequency Domain
2 2
E g g (t ) dt G ( f ) df
Application of Parseval’s
Theorem
2 2
E g (t ) dt G ( f ) df
g
Find
|G(f)|2
Energy
G ( f ) Tsinc(fT )
Application of Parseval’s
Theorem
2 2
E g (t ) dt G ( f ) df
g
Find
|G(f)|2
Energy
f
0
G ( f ) Tsinc(fT )
Sampling
and
Analog-to-Digital Conversion
Analog to Digital Conversion of
Message Signal
• 2 major steps
– Sampling
– Quantization
mp
m(t) Quantized samples m(t)
Allowed quantization levels
2m p
L
-mp
Analog to Digital Conversion of
Message Signal
• Sampling
4
Sampling Theorem
g(t)
g(t)
g(t)
t
Sampling Theorem
Ts
Sampling Theorem
The result is g (t )
T (t ) (t nTs )
s
T (t ) n
s
Ts
Sampling Theorem
T (t ) (t nTs )
s
n
Ts
Sampling Theorem
T (t ) (t nTs )
s
n
Ts
Sampling Theorem
T (t ) (t nTs )
s
n Alternate Representation
1 jn 2f s t
Ts (t )
Ts
n
e
Ts
Sampling Theorem
1 j 2nf s t
Ts
n
g (t ) e
1 j 2nf s t
Ts
n
g (t ) e
FT of g (t ) is defined by
1
G ( f ) g (t ) G ( f nf )s
Ts n
Sampling Theorem
1 j 2nf s t
Ts
n
g (t ) e
FT of g (t ) is defined by G(f)
G ( f ) consists of G(f)
scaled by fs repeated
1 periodically with
G ( f ) g (t ) G ( f nf )s
Ts n period fs
Sampling Theorem
G( f )
g(t)
g (t ) G( f )
Can we reconstruct
g(t) from g (t ) ?
g (t ) G( f )
fs> 2B
Low pass filter of
Alternately, bandwidth B G( f )
1
Sampling interval of g(t), Ts
2B
Use an Ideal lowpass filter of bandwidth
B to isolate one cycle.
fs = 2B
Instructor:
Dr. Md. Monirul Islam
Sampling
and
Analog-to-Digital Conversion
Sampling Theorem
G( f )
g(t)
fs = 2B
fs > 2B
fs > 2B
fs 2B
fs > 2B
fs 2B
sin 2Bt jG ( f )
-B B
Sampling Theorem
sin 2Bt jG ( f )
-B B
f
H ( f ) Ts ( )
2B -2fs -fs fs 2fs
Signal Reconstruction from
Uniform Samples
f
• To recover g(t), the transfer function will be H ( f ) Ts ( )
2B
H(f)
Ts
f
-B B
f
H ( f ) Ts ( )
2B
Signal Reconstruction from
Uniform Samples
f
• To recover g(t), the transfer function will be H ( f ) Ts ( )
2B
H(f) h(t)
Ts 2BTs
f
-B B
f
-B B
h(t ) sinc(2Bt )
Impulse response
Signal Reconstruction from
Uniform Samples
f
• To recover g(t), the transfer function will be H ( f ) Ts ( )
2B
If we pass g (t ) through an h(t)
LTI with this impulse
response, the output will be
g(t).
h(t ) sinc(2Bt )
Impulse response
Signal Reconstruction from
Uniform Samples
f
• To recover g(t), the transfer function will be H ( f ) Ts ( )
2B
If we pass g (t ) through an LTI h(t)
with this impulse response, the
output will be g(t).
g (t ) g (nTs ) (t nTs )
n
h(t ) sinc(2Bt )
Signal Reconstruction from
Uniform Samples
f
• To recover g(t), the transfer function will be H ( f ) Ts ( )
2B
If we pass g (t ) through an LTI h(t)
with this impulse response, the
output will be g(t).
g (t ) g (nTs ) (t nTs )
n
h(t ) sinc(2Bt )
Linear Time
Invariant System,
a (t t
k k ) a h(t t
k k )
k h(t) k
Signal Reconstruction from
Uniform Samples
f
• To recover g(t), the transfer function will be H ( f ) Ts ( )
2B
If we pass g (t ) through an LTI h(t)
with this impulse response, the
output will be g(t).
g (t ) g (nTs ) (t nTs )
n
h(t ) sinc(2Bt )
Linear Time
Invariant System,
x(t) δ(t nτ)x(nτ)τ y(t) h(t τ)x(nτ)τ
n h(t) n
Signal Reconstruction from
Uniform Samples
f
• To recover g(t), the transfer function will be H ( f ) Ts ( )
2B
If we pass g (t ) through an LTI h(t)
with this impulse response, the
output will be g(t).
g (t ) g (nTs ) (t nTs )
n
h(t ) sinc(2Bt )
Linear Time
Invariant System,
g (t ) g (nTs ) (t nTs ) g (t ) g (nTs )h(t nTs )
n h(t) n
Signal Reconstruction from
Uniform Samples
Linear Time
Invariant System,
g (t ) g (nTs ) (t nTs ) g (t ) g (nTs )h(t nTs )
n h(t) n
sinc(2Bt )
Signal Reconstruction from
Uniform Samples
Linear Time
Invariant System,
g (t ) g (nTs ) (t nTs ) g (t ) g (nTs )h(t nTs )
n h(t) n
sinc(2Bt )
g (t ) h(t )
Signal Reconstruction from
Uniform Samples
Linear Time
Invariant System,
g (t ) g (nTs ) (t nTs ) g (t ) g (nTs )h(t nTs )
n h(t) n
sinc(2Bt )
g (t ) h(t )
g (t ) g (nTs )h(t nTs )
n
g (nTs )sinc(2Bt n )
n
Signal Reconstruction from
Uniform Samples
Linear Time
Invariant System,
g (t ) g (nTs ) (t nTs ) g (t ) g (nTs )h(t nTs )
n h(t) n
sinc(2Bt )
g (t ) h(t )
g (t ) g (nTs )sinc(2Bt n )
n
Interpolation formula to find the value g(t)
at arbitrary time t.
Signal Reconstruction from
Uniform Samples
Significance of Interpolation
formula
g (t ) g (nTs )sinc(2Bt n )
n
ILPF
Practical Signal Reconstruction
(Interpolation)
H(f)
Ts
generates g (t ) g (nTs )sinc(2Bt n )
n
f
-B B Interpolation formula
ILPF:
Noncausal
and
Unrealizable
Practical Signal Reconstruction
(Interpolation)
H(f)
Ts
generates g (t ) g (nTs )sinc(2Bt n )
n
f
-B B Interpolation formula
ILPF: h(t)
Noncausal
and
Unrealizable
h(t)
Sharp transition of
ILPF at ±B causes
this problem
p(t)
t
Practical Signal Reconstruction
(Interpolation)
Original signal
Practical Signal Reconstruction
(Interpolation)
Original signal
g~ (t ) g (nTs ) p (t nTs )
n
Practical Signal Reconstruction
(Interpolation)
Original signal
g~ (t ) g (nTs ) p (t nTs ) g (t ) p (t )
n
Practical Signal Reconstruction
(Interpolation)
g~ (t ) g (nTs ) p (t nTs ) g (t ) p (t )
n
Practical Signal Reconstruction
(Interpolation)
g~ (t ) g (nTs ) p (t nTs ) g (t ) p (t )
n
g~ (t ) g (nTs ) p (t nTs ) g (t ) p (t )
n
1
Remember, G ( f )
Ts
G ( f nf )
n
s
Practical Signal Reconstruction
(Interpolation)
g~ (t ) g (nTs ) p (t nTs ) g (t ) p (t )
n
g~ (t ) g (nTs ) p (t nTs ) g (t ) p (t )
n
~
G ( f ) E ( f )G ( f )
E(f)P(f) must
1
E ( f ) P( f )
Ts
G ( f nf )
s REMOVE this
n portion
~
G( f )
~
G ( f ) E ( f )G ( f )
E(f)P(f) must
1
E ( f ) P( f )
Ts
G ( f nf )
s REMOVE this
n portion
~
G( f )
~
G ( f ) E ( f )G ( f )
E(f)P(f) must
1
E ( f ) P( f )
Ts
G ( f nf )
s PRESERE this
n portion
~
G( f )
~
G ( f ) E ( f )G ( f )
E(f)P(f) must
1
E ( f ) P( f )
Ts
G ( f nf )s PRESERE this
n portion
~
G( f )
E(f)P(f) is flexible
~ in this portion
G ( f ) E ( f )G ( f )
1 B < |f| < fs- B
E ( f ) P( f )
Ts
G ( f nf )
n
s
~
G( f )
~
G ( f ) E ( f )G ( f )
1
E ( f ) P( f )
Ts
G ( f nf )
n
s
~
G( f )
3 conditions together
Ts P ( f ) f B
E ( f ) Flexible B f f s B
0 f fs B
-2fs -fs fs 2fs
Practical Signal Reconstruction
(Interpolation)
Samples
Sampling at g(nTs) g~ (t ) Equalizer
g(t)
frequency fs × E(f)
g(t)
Pulse
generator
Practical Signal Reconstruction
(Interpolation)
Samples
Sampling at g(nTs) g~ (t ) Equalizer
g(t)
frequency fs × E(f)
g(t)
Pulse
generator
t
0 Tp
Practical Signal Reconstruction
(Interpolation)
Samples
Sampling at g(nTs) g~ (t ) Equalizer
g(t)
frequency fs × E(f)
g(t)
Pulse
generator
Pulse
generator
0
Practical Signal Reconstruction
(Interpolation)
Samples
Sampling at g(nTs) g~ (t ) Equalizer
g(t)
frequency fs × E(f)
g(t)
Pulse
generator
0
Practical Signal Reconstruction
(Interpolation)
Samples
Sampling at g(nTs) g~ (t ) Equalizer
g(t)
frequency fs × E(f)
g(t)
Pulse
generator
0
Practical Signal Reconstruction
(Interpolation)
Samples
Sampling at g(nTs) g~ (t ) Equalizer
g(t)
frequency fs × E(f)
g(t)
Pulse
generator
0
Ts
Practical Signal Reconstruction
(Interpolation)
Samples
Sampling at g(nTs) g~ (t ) Equalizer
g(t)
frequency fs × E(f)
g(t)
Pulse
generator Generates
this pulse
p(t)
1
0 t
Ts 0 Tp
Practical Signal Reconstruction
(Interpolation)
Samples
Sampling at g(nTs) g~ (t ) Equalizer
g(t)
frequency fs × E(f)
g(t)
Pulse
generator
p(t)
1
t
Ts 0 Tp
Tp
Practical Signal Reconstruction
(Interpolation)
Samples
Sampling at g(nTs) g~ (t ) Equalizer
g(t)
frequency fs × E(f)
g(t)
Pulse
generator What happens
if Tp→0?
p(t)
1
t
Ts 0 Tp
Tp
Practical Signal Reconstruction
(Interpolation)
Samples
Sampling at g(nTs) g~ (t ) Equalizer
g(t)
frequency fs × E(f)
g(t)
Pulse
After
generator
equalizing
p(t)
1
t
Ts 0 Tp
Tp
Practical Signal Reconstruction
(Interpolation)
Samples
Sampling at g(nTs) g~ (t ) Equalizer
g(t)
frequency fs × E(f)
g(t)
t
Ts 0 Tp
Tp
Practical Signal Reconstruction
(Interpolation)
Let’s find the FT
response of the
rectangular
reconstruction p(t) t 0.5T p
1 p (t )
pulse T
p
t
0 Tp
Practical Signal Reconstruction
(Interpolation)
Let’s find the FT
response of the
rectangular
reconstruction p(t) t 0.5T p
1 p (t )
pulse T
p
t
0 Tp
jfT p
P ( f ) T p sinc(fT p )e
Practical Signal Reconstruction
(Interpolation)
Let’s find the FT
response of the
rectangular
reconstruction p(t) t 0.5T p
1 p (t )
pulse T
p
t
0 Tp
jfT p
P ( f ) T p sinc(fT p )e
t
sinc (f )
g (t t0 ) G ( f ) exp( j 2ft0 )
Practical Signal Reconstruction
(Interpolation)
rectangular |P(f)|
reconstruction pulse
p(t)
1
t
0 Tp
3 2 1 0 1 2 3
t 0.5T p Tp Tp Tp Tp Tp Tp
p (t )
T
p P ( f ) T p sinc(fT p )e
jfT p
Practical Signal Reconstruction
(Interpolation)
rectangular |P(f)|
reconstruction pulse
FT of Ideal
p(t) Reconstruction
pulse
1
t
0 Tp
3 2 1 0 1 2 3
t 0.5T p Tp Tp Tp Tp Tp Tp
p (t )
T
p P ( f ) T p sinc(fT p )e
jfT p
Practical Signal Reconstruction
(Interpolation)
Samples
Sampling at g(nTs) g~ (t ) Equalizer
g(t)
frequency fs × E(f)
g(t)
Pulse
After
generator
equalizing
p(t)
1
t
Ts 0 Tp
Tp
Practical Signal Reconstruction
(Interpolation)
~ 1
G ( f ) P( f )
Ts
G ( f nf )
n
s
~
G ( f ) E ( f )G ( f )
1
E ( f ) P( f )
Ts
G ( f nf )
n
s
~
G( f )
Equalizer
Ts P ( f ) f B
E ( f ) Flexible B f f s B
0 f fs B
-2fs -fs fs 2fs
Practical Signal Reconstruction
(Interpolation)
~ 1
G ( f ) P( f )
Ts
G ( f nf ) s p(t)
n 1
~
G ( f ) E ( f )G ( f ) t
0 Tp
1
E ( f ) P( f )
Ts
G ( f nf )
n
s
~
G( f )
Ts P ( f ) f B
E ( f ) Flexible B f f s B
0 f fs B
-2fs -fs fs 2fs
Practical Signal Reconstruction
(Interpolation)
~ 1
G ( f ) P( f )
Ts
G ( f nf ) s p(t) t 0.5T p
p (t )
n 1 T
~ p
G ( f ) E ( f )G ( f ) t
0 Tp
1
E ( f ) P( f )
Ts
G ( f nf )
n
s
~
G( f )
Ts P ( f ) f B
E ( f ) Flexible B f f s B
0 f fs B
-2fs -fs fs 2fs
Practical Signal Reconstruction
(Interpolation)
~ 1
G ( f ) P( f )
Ts
G ( f nf ) s p(t) t 0.5T p
p (t )
n 1 T
~ p
G ( f ) E ( f )G ( f ) t
0 Tp
1
E ( f ) P( f )
Ts
G ( f nf )
n
s
jfT p
P( f ) T p sinc(fT p )e
Ts P ( f ) f B
E ( f ) Flexible B f f s B
0 f fs B
Practical Signal Reconstruction
(Interpolation)
~ 1
G ( f ) P( f )
Ts
G ( f nf ) s p(t) t 0.5T p
p (t )
n 1 T
~ p
G ( f ) E ( f )G ( f ) t
0 Tp
1
E ( f ) P( f )
Ts
G ( f nf )
n
s
jfT p
P( f ) T p sinc(fT p )e
Ts P ( f ) f B
f
E ( f ) Flexible B f f s B
E ( f ) Ts f B
0 f fs B sin(fT p )
Practical Signal Reconstruction
(Interpolation)
p(t)
1
t
Ts 0 Tp
Tp
t
Ts 0 Tp
Tp
t
Ts 0 Tp
Tp
Instructor:
Dr. Md. Monirul Islam
Sampling
and
Analog-to-Digital Conversion
Sampling Theorem
G( f )
g(t)
g (t ) G( f )
G( f )
f
B fs
Realizability of Reconstruction
Filters
Spectrum G ( f ) when signal g(t) is
sampled at the Nyquist rate fs = 2B
G( f )
f
B fs
Repetitions of G(f) without any gap
between successive cycles
Realizability of Reconstruction
Filters
Spectrum G ( f ) when signal g(t) is
sampled at the Nyquist rate fs = 2B
G( f )
f
B fs
Ideal reconstruction
filter (Box filter)
Realizability of Reconstruction
Filters
Spectrum G ( f ) when signal g(t) is
sampled at the Nyquist rate fs = 2B
G( f )
f
B fs
Ideal reconstruction
filter (Box filter)
But box filter is unrealizable; can be
approximated only with infinite time delay
Realizability of Reconstruction
Filters
Alternate: Sample at a rate higher than
Nyquist rate fs > 2B
G( f )
f
B fs
Realizability of Reconstruction
Filters
Alternate: Sample at a rate higher than
Nyquist rate fs > 2B
G( f )
f
B fs
f
B fs
f
B fs
A signal of 60 Hz frequency
Wrong Assumption on Bandwidth:
Treachery of Aliasing
Definition of aliasing
Sampled Signal
Wrong Assumption on Bandwidth:
Treachery of Aliasing
Definition of aliasing
Spectrum of
0 f
unsampled g(t) -B B
G( f )
Spectrum of over
f
sampled g(t) -2fs -fs -B 0 B fs 2fs
G( f )
Spectrum of critically
f
sampled g(t) -2fs -fs -B 0 B fs 2fs
G( f )
Spectrum of under
f
sampled g(t) -2fs -fs 0 fs 2fs
G(f)
Spectrum of
0 f
unsampled g(t) -B B
G( f )
Spectrum of over
f
sampled g(t) -2fs -fs -B 0 B fs 2fs
G( f )
Spectrum of critically
f
sampled g(t) -2fs -fs -B 0 B fs 2fs
G( f )
Overlapping cycles
f
of G(f) -2fs -fs 0 fs 2fs
G(f)
Spectrum of
0 f
unsampled g(t) -B B
G( f )
Spectrum of over
f
sampled g(t) -2fs -fs -B 0 B fs 2fs
G( f )
Spectrum of critically
f
sampled g(t) -2fs -fs -B 0 B fs 2fs
G( f )
Spectrum of
0 f
unsampled g(t) -B B
G( f )
Spectrum of over
f
sampled g(t) -2fs -fs -B 0 B fs 2fs
G( f )
Spectrum of critically
f
sampled g(t) -2fs -fs -B 0 B fs 2fs
G( f )
Spectrum of
0 f
unsampled g(t) -B B
G( f )
Spectrum of over
f
sampled g(t) -2fs -fs -B 0 B fs 2fs
G( f )
Spectrum of critically
f
sampled g(t) -2fs -fs -B 0 B fs 2fs
G( f )
However,
• all practical signals are time-limited
Wrong Assumption on Bandwidth:
Treachery of Aliasing
While sampling and reconstructing,
• Signal g(t) is assumed to be band-limited to B
However,
• all practical signals are time-limited
• time limited signal CANNOT be band-limited
Wrong Assumption on Bandwidth:
Treachery of Aliasing
While sampling and reconstructing,
• Signal g(t) is assumed to be band-limited to B
However,
• all practical signals are time-limited
• time limited signal CANNOT be band-limited
• practical signals are therefore NON-band-limited
Wrong Assumption on Bandwidth:
Treachery of Aliasing
G(f)
f
0
A practical signal with
infinite bandwidth (B)
Wrong Assumption on Bandwidth:
Treachery of Aliasing
G(f)
f
0
A practical signal with
infinite bandwidth (B)
-fs 0 fs
-fs 0 fs
Overlapping
tails
-fs 0 fs
-fs 0 fs
-fs 0 fs
-fs 0 fs
-fs/2 fs/2
-fs 0 fs
-fs/2 fs/2
Reconstructed
spectrum Ga(f)
Wrong Assumption on Bandwidth:
Treachery of Aliasing
H(f) Spectrum of
G( f ) Sampled signal
-fs 0 fs
-fs/2 fs/2
Lost tail is
folded back Lost tail
Reconstructed
spectrum Ga(f)
Wrong Assumption on Bandwidth:
Treachery of Aliasing
H(f) Spectrum of
G( f ) Sampled signal
-fs 0 fs
-fs/2 fs/2
Lost tail is
folded back Lost tail
Folding
frequency
Wrong Assumption on Bandwidth:
Treachery of Aliasing
H(f) Spectrum of
G( f ) Sampled signal
-fs 0 fs
-fs/2 fs/2
Lost tail is
folded back Lost tail
-fs 0 fs
-fs/2 fs/2
-fs 0 fs
-fs/2 fs/2
-fs 0 fs
-fs/2 fs/2
2. Folded back tail
distorts lower
Reconstructed frequencies
spectrum Ga(f)
1. Lost tail results in
loss of higher
TWO frequencies
Problems! -fs/2 fs/2
Combatting Aliasing
H(f) Spectrum of
G( f ) Sampled signal
-fs 0 fs
-fs/2 fs/2
this portion is the
main betrayer
Combatting Aliasing
H(f) Spectrum of
G( f ) Sampled signal
-fs 0 fs
-fs/2 fs/2
REMOVE this
portion before
sampling using an
Antialiasing filter
(ILPF)
Combatting Aliasing
H(f) Spectrum of
G( f ) Sampled signal
-fs 0 fs
-fs/2 fs/2
Antialiasing filter
gaa(t) g aa (t )
g(t) Haa(f) Sampler
Gaa ( f )
T(t)
Combatting Aliasing
H(f) Spectrum of
G( f ) Sampled signal
-fs 0 fs
-fs/2 fs/2
H(f)
Gaa ( f )
Sample signal
spectrum
Reconstructed spectrum
no distortion of low frequency
-fs/2 fs/2
Lost tail results in loss of high
frequencies
Combatting Aliasing
H(f)
Gaa ( f )
Sample signal
spectrum
Reconstructed spectrum
no distortion of low frequency
-fs/2 fs/2
Lost tail results in loss of high
frequencies
Advantage: suppressed
component cannot reappear
Gaa ( f ) G ( f ) | f | f s / 2 in low frequency
components
Combatting Aliasing
H(f)
Gaa ( f )
Sample signal
spectrum
Reconstructed spectrum
no distortion of low frequency
-fs/2 fs/2
Lost tail results in loss of high
frequencies
Other characteristics:
• Antialiasing filter band-limits g(t) to fs/2
• Reduces noise too, otherwise noise might be folded back to low
frequency components
• Noises beyond fs/2 are suppressed
Sampling Effect: Non-band-limited
signal to band-limited signal
G(f)
f
0
G ( f ) = Spectrum of
G( f ) Sampled signal, g (t )
G(f)
-fs 0 fs
-fs/2 fs/2
Bandwidth of G(f) is practically infinite
Sampling Effect: Non-band-limited
signal to band-limited signal
G(f)
f
0
G ( f ) = Spectrum of
G( f ) Sampled signal, g (t )
-fs 0 fs
-fs/2 fs/2
Bandwidth of G(f) is practically infinite
fs << Nyquist rate
Sampling Effect: Non-band-limited
signal to band-limited signal
G(f)
f
0
G ( f ) = Spectrum of
G( f ) Sampled signal, g (t )
-fs 0 fs
-fs/2 fs/2
Bandwidth of G(f) is practically infinite
G ( f ) , that means, spectrum of g (t ) consists of overlapping cycles of G(f)
Sampling Effect: Non-band-limited
signal to band-limited signal
G(f)
f
0
G ( f ) = Spectrum of
G( f ) Sampled signal, g (t )
-fs 0 fs
-fs/2 fs/2
Bandwidth of G(f) is practically infinite
g (t ) are sub-Nyquist samples of g(t)!
Sampling Effect: Non-band-limited
signal to band-limited signal
G ( f ) = Spectrum of
G( f ) Sampled signal, g (t )
-fs 0 fs
-fs/2 fs/2
The above bold line spectrum can be viewed as repetition of the following Ga(f)
band-limited to fs/2
Ga(f)
-fs/2 0 fs/2
Sampling Effect: Non-band-limited
signal to band-limited signal
G ( f ) = Spectrum of
G( f ) Sampled signal, g (t )
-fs 0 fs
-fs/2 fs/2
The above bold line spectrum can be viewed as repetition of the following Ga(f)
band-limited to fs/2 Ga(f) = spectrum of ga(t)
-fs/2 0 fs/2
Sampling Effect: Non-band-limited
signal to band-limited signal
G ( f ) = Spectrum of
G( f ) Sampled signal, g (t )
-fs 0 fs
-fs/2 fs/2
-fs/2 0 fs/2
Sampling Effect: Non-band-limited
signal to band-limited signal
G ( f ) = Spectrum of
G( f ) Sampled signal, g (t )
-fs 0 fs
-fs/2 fs/2
Instructor:
Dr. Md. Monirul Islam
Sampling
and
Analog-to-Digital Conversion
Sampling Effect: Non-band-limited
signal to band-limited signal
G(f) = spectrum of g(t)
f
0
Sampling Effect: Non-band-limited
signal to band-limited signal
G ( f ) = Spectrum of
G( f ) Sampled signal, g (t )
-fs 0 fs
-fs/2 fs/2
G(f)
-fs/2 0 fs/2
Sampling Effect: Non-band-limited
signal to band-limited signal
G ( f ) = Spectrum of
G( f ) Sampled signal, g (t )
-fs 0 fs
-fs/2 fs/2
G(f)
f
0
Let
sub-Nyquist Sampling of g(t) at fs generates samples g(0), g(Ts), g(2Ts), g(3Ts),, . . .
Sampling Effect: Non-band-limited
signal to band-limited signal
Ga(f) = spectrum of ga(t)
-fs/2 0 fs/2
Let
Nyquist Sampling of ga(t) at fs generates samples ga(0), ga(Ts), ga (2Ts), ga (3Ts),, . . .
Sampling Effect: Non-band-limited
signal to band-limited signal
G(f) = spectrum of g(t) Ga(f) = spectrum of ga(t)
f
0 -fs/2 0 fs/2
sub-Nyquist Samples g(0), Nyquist Samples ga(0), ga(Ts), ga(2Ts),
g(Ts), g(2Ts), g(3Ts),, . . . ga(3Ts),, . . .
Sampling Effect: Non-band-limited
signal to band-limited signal
G(f) = spectrum of g(t) Ga(f) = spectrum of ga(t)
f
0 -fs/2 0 fs/2
sub-Nyquist Samples g(0), Nyquist Samples ga(0), ga(Ts), ga(2Ts),
g(Ts), g(2Ts), g(3Ts),, . . . ga(3Ts),, . . .
f
0 -fs/2 0 fs/2
sub-Nyquist Samples g(0), Nyquist Samples ga(0), ga(Ts), ga(2Ts),
g(Ts), g(2Ts), g(3Ts),, . . . ga(3Ts),, . . .
Therefore,
g(nTs) = ga(nTs) = gn
Sampling Effect: Non-band-limited
signal to band-limited signal
G(f) = spectrum of g(t) Ga(f) = spectrum of ga(t)
f
0 -fs/2 0 fs/2
sub-Nyquist Samples g(0), Nyquist Samples ga(0), ga(Ts), ga(2Ts),
g(Ts), g(2Ts), g(3Ts),, . . . ga(3Ts),, . . .
f
0 -fs/2 0 fs/2
sub-Nyquist Samples g(0), Nyquist Samples ga(0), ga(Ts), ga(2Ts),
g(Ts), g(2Ts), g(3Ts),, . . . ga(3Ts),, . . .
We will prove
• Maximum 2B pieces of information can be sent per second
Maximum Information Rate of a
Channel with BW = B Hz
Assume
• Error free, noise less channel
• Channel bandwidth is B
Previous Knowledge
• Channel can send a low pass signal of B Hz
• This signal can be recovered from samples uniformly taken at 2B samples per
second
Maximum Information Rate of a
Channel with BW = B Hz
Assume
• Error free, noise less channel
• Channel bandwidth is B
Previous Knowledge
• Channel can send a low pass signal of B Hz
• This signal can be recovered from samples uniformly taken at 2B samples per
second
• This means, 2B samples/second can be sent through the channel
Maximum Information Rate of a
Channel with BW = B Hz
Assume
• Error free, noise less channel
• Channel bandwidth is B
Previous Knowledge
• Channel can send a low pass signal of B Hz
• This signal can be recovered from samples uniformly taken at 2B samples per
second
• This means, 2B samples/second can be sent through the channel
Assume a sequence of samples g0, g1, g2, g3, . . . denoted as {gn} at the rate of 2B s/s
Maximum Information Rate of a
Channel with BW = B Hz
Assume a sequence of samples g0, g1, g2, g3, . . . denoted as {gn} at the rate of 2B s/s
We will always find a signal g(t) whose samples g(0), g(Ts), g(2Ts), g(3Ts), . .
matches with {gn}.
f
0
Maximum Information Rate of a
Channel with BW = B Hz
Assume a sequence of samples g0, g1, g2, g3, . . . denoted as {gn} at the rate of 2B s/s
We will always find a signal g(t) whose samples g(0), g(Ts), g(2Ts), g(3Ts), . .
matches with {gn}.
f
0
Sampling Effect: Non-band-limited
signal to band-limited signal
G(f) = spectrum of g(t) Ga(f) = spectrum of ga(t)
f
0 -fs/2 0 fs/2
sub-Nyquist Samples g(0), Nyquist Samples ga(0), ga(Ts), ga(2Ts),
g(Ts), g(2Ts), g(3Ts),, . . . ga(3Ts),, . . .
f
0 -fs/2 0 fs/2
sub-Nyquist Samples g(0), Nyquist Samples ga(0), ga(Ts), ga(2Ts),
g(Ts), g(2Ts), g(3Ts),, . . . ga(3Ts),, . . .
g (t ) g (nTs ) (t nTs )
n
Sampling Effect: Non-band-limited
signal to band-limited signal
G(f) = spectrum of g(t) Ga(f) = spectrum of ga(t)
f
0 -fs/2 0 fs/2
sub-Nyquist Samples g(0), Nyquist Samples ga(0), ga(Ts), ga(2Ts),
g(Ts), g(2Ts), g(3Ts),, . . . ga(3Ts),, . . .
g (t ) g (nTs ) (t nTs )
n
g a (nTs ) (t nTs )
n
Sampling Effect: Non-band-limited
signal to band-limited signal
G(f) = spectrum of g(t) Ga(f) = spectrum of ga(t)
f
0 -fs/2 0 fs/2
sub-Nyquist Samples g(0), Nyquist Samples ga(0), ga(Ts), ga(2Ts),
g(Ts), g(2Ts), g(3Ts),, . . . ga(3Ts),, . . .
g (t ) g (nTs ) (t nTs )
n
g a (nTs ) (t nTs )
n
g n (t nTs )
n
Sampling Effect: Non-band-limited
signal to band-limited signal
G(f) = spectrum of g(t) Ga(f) = spectrum of ga(t)
f
0 -fs/2 0 fs/2
sub-Nyquist Samples g(0), Nyquist Samples ga(0), ga(Ts), ga(2Ts),
g(Ts), g(2Ts), g(3Ts),, . . . ga(3Ts),, . . .
sampling
TDM Example
Application of Sampling Theorem
TDM: dual of FDM where different signals share channel bandwidth
TDM Example
G1(f) G2(f)
FDM Example
f
0 f1 f2
Pulse Code Modulation (PCM)
LPF Bit
Sampler Quantizer
Encoder 1 0 1 1 0 1
2m p
v
L
2m p
L
-mp
Analog to Digital Conversion of
Message Signal
• 2 major steps
– Sampling
The range (–mp, mp) is
– Quantizing
divided into L sub-
mp intervals, each of
m(t) Quantized samples m(t) magnitude v
Allowed quantization levels
2m p
v
L
L is known as
2m p quantization level
L
-mp
Analog to Digital Conversion of
Message Signal
• 2 major steps
– Sampling
A sampled value is
– Quantizing
placed into one of
mp these L sub-intervals,
m(t) Quantized samples m(t) thus gets ONE of the
Allowed quantization levels
L values
2m p
L
-mp
Analog to Digital Conversion of
Message Signal
• 2 major steps
– Sampling
A sampled value is
– Quantizing
placed into one of
mp these L sub-intervals,
m(t) Quantized samples m(t) thus gets ONE of the
Allowed quantization levels
L values
Signal is known as
L-ary digital signal
2m p
L
-mp
Analog to Digital Conversion of
Message Signal
Fidelity is compromised!
Analog to Digital Conversion of
Message Signal
Analog signal bandwidth to digital data rate
B = 3400 Hz
fs = 8000 > 2B
Analog to Digital Conversion of
Message Signal
Analog signal bandwidth to digital data rate
B = 3400 Hz
fs = 8000 > 2B
B = 3400 Hz
fs = 8000 > 2B
B = 20000 Hz
fs = 441000 Hz > 2B
Self Study
Quantization
2m p
v
L
L is known as
2m p quantization level
L
-mp
Quantization
mp is NOT the signal PEAK, rather is it’s the LIMIT of the quantizer
2m p
v
L
L is known as
2m p quantization level
L
-mp
Quantization
k-th sample value m(kTs) is replaced by the midpoint of an interval where it lies
2m p
v
L
L is known as
2m p quantization level
L
-mp
Quantization
k-th sample value m(kTs) is replaced by the midpoint of an interval where it lies
Replaced by
m(kTs ) mˆ (kTs )
The range (–mp, mp) is
divided into L sub-
mp intervals, each of
m(t) Quantized samples m(t) magnitude v
Allowed quantization levels
2m p
v
L
L is known as
2m p quantization level
L
-mp
Quantization
k-th sample value m(kTs) is replaced by the midpoint of an interval where it lies
Replaced by
m(kTs ) mˆ (kTs )
quantization error is unavoidable which is lies in (-v/2, v/2)
mp
m(t) Quantized samples m(t)
Allowed quantization levels
2m p
v
L
L is known as
2m p quantization level
L
-mp
Quantization
k-th sample value m(kTs) is replaced by the midpoint of an interval where it lies
Replaced by
m(kTs ) mˆ (kTs )
quantization error is unavoidable which is lies in (-v/2, v/2)
mp
m(t) Quantized samples m(t)
Allowed quantization levels
mˆ (kTs ) v / 2
mˆ (kTs )
mˆ (kTs ) v / 2
2m p
L
2m p
-mp v
L
Quantization
k-th sample value m(kTs) is replaced by the midpoint of an interval where it lies
Replaced by
m(kTs ) mˆ (kTs )
quantization error is unavoidable which is lies in (-v/2, v/2)
mp
m(t) Quantized samples m(t)
Allowed quantization levels
mˆ (kTs ) v / 2
m(kTs)
mˆ (kTs )
mˆ (kTs ) v / 2
2m p
L
2m p
-mp v
L
Quantization
k-th sample value m(kTs) is replaced by the midpoint of an interval where it lies
Replaced by
m(kTs ) mˆ (kTs )
quantization error is unavoidable which is lies in (-v/2, v/2)
mp
m(t) Quantized samples m(t)
Allowed quantization levels
mˆ (kTs ) v / 2
m(kTs)
mˆ (kTs )
mˆ (kTs ) v / 2
2m p
L
2m p
-mp v
L
Quantization
k-th sample value m(kTs) is replaced by the midpoint of an interval where it lies
Replaced by
m(kTs ) mˆ (kTs )
quantization error is unavoidable which is lies in (-v/2, v/2)
mp
error
m(t) Quantized samples m(t)
Allowed quantization levels
v/2
q(kTs) = q
0
-v/2
2m p
L
2m p
-mp v
L
Quantization
k-th sample value m(kTs) is replaced by the midpoint of an interval where it lies
Replaced by
m(kTs ) mˆ (kTs )
quantization error is unavoidable which is lies in (-v/2, v/2)
mp
m(t) Quantized samples m(t)
Allowed quantization levels
mˆ (kTs ) v / 2
mˆ (kTs )
m(kTs)
mˆ (kTs ) v / 2
2m p
L
2m p
-mp v
L
Quantization
k-th sample value m(kTs) is replaced by the midpoint of an interval where it lies
Replaced by
m(kTs ) mˆ (kTs )
quantization error is unavoidable which is lies in (-v/2, v/2)
mp
error
m(t) Quantized samples m(t)
Allowed quantization levels
v/2
0
q(kTs) = q
-v/2
2m p
L
2m p
-mp v
L
Quantization
k-th sample value m(kTs) is replaced by the midpoint of an interval where it lies
Binary
quantization coding
m(kTs ) mˆ (kTs ) 101101 . . . Transmission
pulse
Pulse detection
at receiving end
Quantization
k-th sample value m(kTs) is replaced by the midpoint of an interval where it lies
Binary
quantization coding
m(kTs ) mˆ (kTs ) 101101 . . . Transmission
pulse
Pulse
detection
error
Quantization
Quantization noise,
q(kTs) = Quantization
error for kth sample
Quantization
lim 1 T / 2
a1 a2 a3 dt
2
T T T / 2
Quantization
lim 1 T / 2
a1 a2 a3 dt
2
T T T / 2
lim 1 T / 2 2
T T T / 2
a1
a 2
2 a 2
3 2 a1a 2 2 a1a3 2 a1a 4 dt
Quantization
lim 1 T / 2 2
T T T / 2
a1 a 2
2 a 2
3 2 a1a 2 2 a1a3 2 a1a 4 dt
lim 1 T / 2 2
T T T / 2 k
ak 2 am an dt
m n
Quantization
2
~~~~~~
2 lim 1 T / 2
q (t )
T T T / 2 k
q (kTs )sinc(2Bt k ) dt
lim 1 T / 2 2
T T T / 2 k
ak 2 am an dt
mn
lim 1 T / 2 2 2
T T T / 2 k
q (kTs )sinc (2Bt k ) dt
lim 1 T / 2
2
T T T / 2 m n
q ( mTs ) q ( nTs )sinc( 2Bt m )sinc( 2Bt n )
dt
Quantization
~~~~~~ lim 1 T / 2
2 2 2
q (t )
T T T / 2 k
q ( kTs )sinc ( 2Bt k )
dt
lim 1 T / 2
2
T T T / 2 m n
q ( mTs ) q ( nTs ) sinc( 2Bt m ) sinc( 2Bt n )
dt
T /2
lim 1
q 2 (kTs ) sinc 2 (2Bt k )dt
T T k T / 2
T /2
lim 1
T T mn
q(mTs )q (nTs ) sinc( 2Bt m )sinc( 2Bt n )dt
T / 2
Quantization
~~~~~~ T /2
2 lim 1 2 2
q (t )
T T k
q ( kTs ) sinc (2Bt k )dt
T / 2
T /2
lim 1
T T mn
q(mTs )q (nTs ) sinc( 2Bt m )sinc( 2Bt n )dt
T / 2
Quantization
~~~~~~ T /2
2 lim 1 2 2
q (t )
T T k
q ( kTs ) sinc (2Bt k )dt
T / 2
T /2
lim 1
T T mn
q(mTs )q (nTs ) sinc( 2Bt m )sinc( 2Bt n )dt
T / 2
0 mn
sinc(2Bt m )sinc(2Bt n )dt 1 mn
2B
Quantization
~~~~~~
2 lim 1 2
q (t )
T 2BT k
q (kTs )
Quantization
~~~~~~
2 lim 1 2
q (t )
T 2BT k
q (kTs )
~~~~~~
2 lim 1 2
q (t )
T 2BT k
q (kTs )
mp
error
m(t) Quantized samples m(t)
Allowed quantization levels
v/2
0
q(kTs) = q
-v/2
2m p
L
2m p
-mp v
L
Quantization
v/2
0
q(kTs) = q
-v/2
2m p
L
2m p
-mp v
L
Quantization
~~~~~~
2
~~~
2
m 2p
N 0 q (t ) q
3L2
Quantization
power of quantization noise (N0) = mean square quantization error
~~~~~~
2
~~~
2
m 2p
N 0 q (t ) q
3L2
~~~~~~
2
Assume, power of message signal (S0) is given by S 0 m (t )
Quantization
power of quantization noise (N0) = mean square quantization error
~~~~~~
2
~~~
2
m 2p
N 0 q (t ) q
3L2
~~~~~~
2
Assume, power of message signal (S0) is given by S 0 m (t )
S0 varies
• from speaker to speaker
• due to different length of connecting circuits
Quantization
~~~~~~
2
S0 2 m (t )
SNR 3L
N0 m 2p
~~~~~~
2
• SNR is linear function of signal power, S 0 m (t )
S0 varies
• from speaker to speaker
• due to different length of connecting circuits
S0 varies
• from speaker to speaker
• due to different length of connecting circuits
However statistically,
• Small amplitudes (soft speakers) predominate in speech
• Larger amplitudes (loud speakers) are less frequent
CSE 311:
Data Communication
Instructor:
Dr. Md. Monirul Islam
Sampling
and
Analog-to-Digital Conversion
Quantization
k-th sample value m(kTs) is replaced by the midpoint of an interval where it lies
Replaced by
m(kTs ) mˆ (kTs )
quantization error is unavoidable which is lies in (-v/2, v/2)
mp
error
m(t) Quantized samples m(t)
Allowed quantization levels
v/2
0
q(kTs) = q
-v/2
2m p
L
2m p
-mp v
L
Quantization
power of quantization noise (N0) = mean square quantization error
~~~~~~
2
~~~
2
m 2p
N 0 q (t ) q
3L2
~~~~~~
2
Assume, power of message signal (S0) is given by S 0 m (t )
S0 varies
• from speaker to speaker
• due to different length of connecting circuits
However statistically,
• Small amplitudes (soft speakers) predominate in speech
• Larger amplitudes (loud speakers) are less frequent
Quantization
~~~~~~
2
S0 2 m (t )
SNR 3L
N0 m 2p
Causes of the problem
2m p
• All amplitudes are uniformly quantized with equal step size v
L
• Variation in amplitudes is NOT considered
Quantization
~~~~~~
2
S0 2 m (t )
SNR 3L
N0 m 2p
Causes of the problem
2m p
• All amplitudes are uniformly quantized with equal step size v
L
• Variation in amplitudes is NOT considered
2
N 0 v
Quantization
Solution to the problem: Non-uniform quantization
• Smaller step size for smaller amplitudes
• Lager step size for larger amplitudes
mp
2
1
-1
-2
-mp
Non-uniform quantization
Quantization
Solution to the problem: Non-uniform quantization
• Smaller step size for smaller amplitudes
• Lager step size for larger amplitudes
mp
Same result is obtained
by compressing signal
samples non-linearly and
then using a uniform 2
quantizaton 1
-1
-2
Non-linear Uniform
Compression Quantization
-mp
Non-uniform quantization Non-uniform quantization
Quantization
Solution to the problem: Non-uniform quantization
• Smaller step size for smaller amplitudes
• Lager step size for larger amplitudes
mp
output
2
1
Uniform
-1
-2
input
-mp
Non-uniform Non-uniform quantization
Quantization
Solution to the problem: Non-uniform quantization
• Smaller step size for smaller amplitudes
• Lager step size for larger amplitudes
output
input
Non-uniform
Quantization
Solution to the problem: Non-uniform quantization
• Smaller step size for smaller amplitudes
• Lager step size for larger amplitudes
output
106
6
105
5
Uniform
2
input 102 1
10
output
Non-uniform input
Quantization
Solution to the problem: Non-uniform quantization
• Smaller step size for smaller amplitudes
• Lager step size for larger amplitudes
output
106
6
105
5
Uniform
2
input 102 1
10
output
Non-uniform input
Quantization
Solution to the problem: Non-uniform quantization
• Smaller step size for smaller amplitudes
• Lager step size for larger amplitudes
output
106
6
105
5
Logarithmic
Uniform
transform 2
input 102 1
10
output
Non-uniform input
Quantization
Solution to the problem: Non-uniform quantization
• Smaller step size for smaller amplitudes
• Lager step size for larger amplitudes
output
input
Non-uniform
Quantization
Solution to the problem: Non-uniform quantization
• Smaller step size for smaller amplitudes
• Lager step size for larger amplitudes
output
2
input
N 0 v
Non-uniform
Quantization
Solution to the problem: Non-uniform quantization
• Smaller step size for smaller amplitudes
• Lager step size for larger amplitudes
output
• Logarithmic compression
makes quantization noise
proportional to signal power
Uniform
~~~~~~
2
N 0 m (t ) S 0
input
Non-uniform
Quantization
Solution to the problem: Non-uniform quantization
• Smaller step size for smaller amplitudes
• Lager step size for larger amplitudes
output
• Logarithmic compression
makes quantization noise
proportional to signal power
Uniform
~~~~~~
2
N 0 m (t ) S 0
input
S0
SNR constant
Non-uniform N0
Quantization: standard
Compression Characteristic
μ-law characteristic A-law characteristic
y y
m
m
mp
mp
A m m 1
0
1 ln A m p mp A
1 m m y
y ln1
0 1 1 1 ln Am 1 m 1
ln(1 ) mp mp 1 ln A A m
m p p
Restoring the Compressed Signal
Non-uniform quantization
Restoring the Compressed Signal
SNR improvement in μ-law compandor
S0 3L2
S0 N 0 1 ln 2
dB
N0
~~~~~~
2
Relative Signal power m (t ), dB
Restoring the Compressed Signal
SNR improvement in μ-law compandor
S0 3L2
S0 N 0 1 ln 2
dB
N0
~~~~~~
2
Relative Signal power m (t ), dB
Pulse Code Modulation (PCM)
Encoder
LPF Bit
Sampler Quantizer
Encoder 1 0 1 1 0 1
LPF Bit
Sampler Quantizer
Encoder 1 0 1 1 0 1
LPF Bit
Sampler Quantizer
Encoder 1 0 1 1 0 1
Analog-to-digital conversion:
taught at Electronics course
Self Study
Transmission Bandwidth and
Output SNR
Sampling at Encoding in
Nyquist rate Quantization n = log2L bits
m(t ) m(kTs ) mˆ (kTs ) 1 0 1 1 0 1 ...
Level: L
B Hz 2B Hz 2nB bits/s
Transmission Bandwidth and
Output SNR
Sampling at Encoding in
Nyquist rate Quantization n = log2L bits
m(t ) m(kTs ) mˆ (kTs ) 1 0 1 1 0 1 ...
Level: L
B Hz 2B Hz 2nB bits/s
S0 3L2
SNR With non-linear compression
N 0 1 ln 2
SNR vs Channel Bandwidth, BT
~~~~~~
2
S0 2 m (t )
SNR 3L Without compression
N0 m 2p
S0 3L2
SNR With non-linear compression
N 0 1 ln 2
S0
can be combined and written as, SNR c (2) 2 n = c.L2
N0
SNR vs Channel Bandwidth, BT
~~~~~~
2
S0 2 m (t )
SNR 3L
N0 m 2p
S0 3L2
SNR
N 0 1 ln 2
S0
can be combined and written as, SNR c (2) 2 n = c.L2
N0
~~~~~~2
3 m (t )
m2 uncompressed case
where, c p
and, L = 2n
3
1 ln 2 compressed case
SNR vs Channel Bandwidth, BT
S0
SNR c ( 2) 2 n
N0
Replace n by BT/B where,
S0
c (2) 2 BT / B
N0
S0
SNR c ( 2) 2 n
N0
S0 S0
10 log10
10 log10 c (2) 2 n
N 0 dB N0
SNR vs Channel Bandwidth, BT
S0
SNR c ( 2) 2 n
N0
S0 S0
10 log10
10 log10 c (2) 2 n
N 0 dB N0
S0
SNR c ( 2) 2 n
N0
S0 S0
10 log10
10 log10 c (2) 2 n
N 0 dB N0
TDM Example
PCM in TDM: T1 Carrier System
- A Case Study
T1 Carrier
• replaces earlier technology that transmitted a single analog audio signal at a time
24 output
channels
PCM in TDM: T1 Carrier System
- A Case Study
24 input
channels
8K ×24 samples/sec
PCM in TDM: T1 Carrier System
- A Case Study
24 input
channels
8K ×24 s/s
8K ×24 s/s
8K ×24 s/s
8 bit representation of
each sample
PCM in TDM: T1 Carrier System
- A Case Study
24 input
channels
8K ×24 s/s
8K ×24 s/s
Framing bit
8K ×24 s/s
8K ×24 s/s
8K ×24 s/s
8K ×24 s/s
8K ×24 s/s
1.544 Mbps
Digital Signal Level 1 (DS1)
PCM in TDM: T1 Carrier System
- A Case Study
24 input
channels
8K ×24 s/s
1.544 Mbps
Digital Signal Level 1 (DS1)
Decommutators
Synchronizing and Signaling in
T1 Carrier
Frame No.s
Framing bit
Framing bit
Framing bit
Framing bit
Framing bit
Framing bit
T1 requires to also transmit signaling bits All 8 slots are occupied for
• Dialing pulses information bits
• On-hook/off-hook signals
Synchronizing and Signaling in
T1 Carrier
signaling bit
Frame No.s
Framing bit
signaling bit
Frame No.s
ONLY every 6th sample contains signaling bit in its LSB position
Therefore, Frame No.s 1, 7, 13 … contains signaling bits
Signaling bit rate = 8K/6 bps = 1333 bps
Synchronizing and Signaling in
T1 Carrier
signaling bit
Frame No.s
signaling bit
Frame No.s
Framing bit
signaling bit
Frame No.s
Non-Signaling
frame
Frame No.s
Frame No.s
Frame No.s
Structure of ESF
ESF
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Frame No.s
Structure of ESF
ESF
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Frame No.s
Structure of ESF
ESF
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Frame No.s
Instructor:
Dr. Md. Monirul Islam
Sampling
and
Analog-to-Digital Conversion
T1 System Review
24 input
channels
8K ×24 s/s
4 KHz
analog 1.544 Mbps
signals Digital Signal Level 1 (DS1)
8K ×24 s/s
4 KHz
analog 1.544 Mbps
signals Digital Signal Level 1 (DS1)
8K ×24 s/s
4 KHz
analog 1.544 Mbps
signals Digital Signal Level 1 (DS1)
8K ×24 s/s
Word-by-word
interleaving in T1
system
Digital Multiplexing: Word
Interleaving
Channels
Identical bit rates for
all channels
Structure of ESF
ESF
1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17 18 19 20 21 22 23 24
Frame Numbers
All [48]’s : Data bits, interleaved CA, CB, CC, CD, : Stuffing bits
from 4 channels
Bit wise
interleaved from
4 channels
DM1/2 Multiplexer Frame Format
M0 [48] CA [48] F0 [48] CA [48] CA [48] F1 [48]
Bit wise
interleaved from
4 channels
bitA bitB bitC bitD bitA bitB bitC bitD bitA bitB bitC bitD . . . . so on
DM1/2 Multiplexer Frame Format
M0 [48] CA [48] F0 [48] CA [48] CA [48] F1 [48]
Complexity
• NO guaranty to hold: No. of active channels No. of available slots
Complexity
• NO guaranty to hold: No. of active channels No. of available slots
Management
• Use previous statistics to minimize the overload
Asynchronous Channels & Bit
Stuffing
• Synchronization is lost due to different reasons
Propagation time for a pulse through the cable = 1000 km / (2×108 m/s)
=1/200 second
Asynchronous Channels & Bit
Stuffing
Cable length = 1000 km
Propagation time for a pulse through the cable = 1000 km / (2×108 m/s)
=1/200 second
No. of Pulses in transit within this time = 1/200 second * (2×108 pulses/s)
=106 pulses
Asynchronous Channels & Bit
Stuffing
Cable length = 1000 km
Propagation time for a pulse through the cable = 1000 km / (2×108 m/s)
=1/200 second
No. of Pulses in transit within this time = 1/200 second * (2×108 pulses/s)
=106 pulses
Propagation time for a pulse through the cable = 1000 km / (2×108 m/s)
=1/200 second --
++
No. of Pulses in transit within this time = 1/200 second * (2×108 pulses/s) ++
=106 pulses ++
Propagation time for a pulse through the cable = 1000 km / (2×108 m/s)
=1/200 second --
No. of Pulses in transit within this time = 1/200 second * (2×108 pulses/s) ++
=106 pulses ++
Propagation time for a pulse through the cable = 1000 km / (2×108 m/s)
=1/200 second
No. of Pulses in transit within this time = 1/200 second * (2×108 pulses/s)
=106 pulses
drop in cable temperature => rate of received pulses drops => vacant slots in
multiplexer => vacant slots to be stuffed with dummy bits
Asynchronous Channels & Bit
Stuffing
Cable length = 1000 km
Findings:
• Even synchronous multiplexer system may receive bits asynchronously
• We need
• storage for / removal of extra bits
• Stuffing for vacant slots
Asynchronous Channels & Bit
Stuffing
Stuffing classification
• Positive
• Negative
• Positive/negative
Asynchronous Channels & Bit
Stuffing
Positive Stuffing
Un-stuffed signal
• The stuffed bit is the first bit following F1 in the corresponding sub-frame
Plesiochronous (almost
Synchronous) Digital Hierarchy
1 T1 multiplexer/
64 Kbps 2 Channel Bank/
DS0 Signal
24 digital switch
1
1.544 Mbps
2
DS1 Signal 3 DM 1/2
4
1
6.312 Mbps
2
DS2 Signal DM 2/3
7
• Hierarchy of multiplexers
to produce digital signals 1
44.736 Mbps
2 DM 3/4 NA
of progressively higher bit DS3 Signal
rates 3
1
1.544 Mbps
2
DS1 Signal 3 DM 1/2
4
1
6.312 Mbps
2
DS2 Signal DM 2/3
7
• Two major types of
multiplexers 1
44.736 Mbps
– One combines low–data- 2 DM 3/4 NA
DS3 Signal
rate channels 3
– Other combines high-data-
rate channels 139.264 Mbps: DS4NA
Plesiochronous (almost
Synchronous) Digital Hierarchy
1 T1 multiplexer/
64 Kbps 2 Channel Bank/
DS0 Signal
24 digital switch
1
1.544 Mbps
2
DS1 Signal 3 DM 1/2
4
1
6.312 Mbps
2
DS2 Signal DM 2/3
• Low–data-rate 7
multiplexer, Digital Data 1
44.736 Mbps
system (DDS) 2 DM 3/4 NA
DS3 Signal
– Generates 64 Kbps signal 3
– Combines channels of
rates from 2.4 Kbps up to
9.6 Kbps 139.264 Mbps: DS4NA
Plesiochronous (almost
Synchronous) Digital Hierarchy
1 T1 multiplexer/
64 Kbps 2 Channel Bank/
DS0 Signal
24 digital switch
1
1.544 Mbps
2
DS1 Signal 3 DM 1/2
4
1
6.312 Mbps
2
DS2 Signal DM 2/3
7
• high–data-rate
multiplexers, all others 1
44.736 Mbps
2 DM 3/4 NA
DS3 Signal
3
1
1.544 Mbps
2
DS1 Signal 3 DM 1/2
4
1
6.312 Mbps
2
DS2 Signal DM 2/3
7
• 24 channels in T1: not
necessarily all are voice 1
44.736 Mbps
2 DM 3/4 NA
channels DS3 Signal
• The same thing is true for 3
other higher levels
139.264 Mbps: DS4NA
Other Plesiochronous Digital
Hierarchies
PDH T-Carrier
worldwide US and Canada Japan
×4 ×3 ×4
Instructor:
Dr. Md. Monirul Islam
Differential Pulse Code
Modulation (DPCM)
PCM and Transmission Channel
Bandwidth: Example-1
Analog signal bandwidth => digital data rate => Channel Bandwidth
requirement
B = 20000 Hz
fs = 44100 Hz > 2B
Findings
• PCM generates too many bits
• Requires high bandwidth to transmit them
PCM and Transmission Channel
Bandwidth
Causes:
Every sample is represented by n = log2L bits
mp
No. of quantization levels = L
2m p
L
-mp
PCM and Transmission Channel
Bandwidth
Causes:
Every sample is represented by n = log2L bits
mp
No. of quantization levels = L
increasing v
2m p
v
L
increasing NOISE
-mp
PCM and Transmission Channel
Bandwidth
Causes:
Every sample is represented by n = log2L bits
mp
No. of quantization levels = L
increasing v
2m p
v
L
increasing NOISE
-mp
mp
No. of quantization levels = L
2m p
v
L
-mp
PCM and Transmission Channel
Bandwidth
WAY OUT
• Successive samples are NOT independent, can be predicted
• Redundancy (CORRELATION) exists among neighboring samples
• Difference can be transmitted instead of actual sample
mp
No. of quantization levels = L
Due t o redundancy,
mp
No. of quantization levels = L
decreases peak dp
2m p decreases vd
v
L
-mp decreases NOISE
PCM and Transmission Channel
Bandwidth
Alternatively,
For same NOISE or SNR level, samples can be transmitted by fewer n = log2L bits
mp
No. of quantization levels = L
decreases peak dp
2m p decreases vd
v
L
-mp decreases NOISE
PCM and Transmission Channel
Bandwidth
d[k] effect
• For same NOISE or SNR, fewer n = log2L bits
• For same n, reduced noise or higher SNR
mp
No. of quantization levels = L
decreases peak dp
2m p decreases vd
v
L
-mp decreases NOISE
PCM and Transmission Channel
Bandwidth
At the receiver
• m[k] is reconstructed from d[k] and previous samples m[k-1]
• Smaller is d[k], so is n
PCM and Transmission Channel
Bandwidth
Findings
• d[k] reduces the no. of transmitted bits
• Smaller is d[k], so is n
• Smaller is d[k], so is n
Smaller is d[k]
Smaller is n
PCM and Transmission Channel
Bandwidth
Steps
Predict mˆ [k ]
In DPCM: d [k ] m[k ] mˆ [k ]
Ts2 Ts3
m(t Ts ) m(t ) Ts m (t ) (t )
m (t )
m
2! 3!
Prediction in DPCM
Ts2 Ts3
m(t Ts ) m(t ) Ts m (t ) (t )
m (t )
m
2! 3!
Ts2 Ts3
m(t Ts ) m(t ) Ts m (t ) m (t ) (t )
m
2! 3!
m(t ) Ts m (t ) for small Ts
Prediction in DPCM
Ts2 Ts3
m(t Ts ) m(t ) Ts m (t ) m (t ) (t )
m
2! 3!
m(t ) Ts m (t ) for small Ts
Ts2 Ts3
m(t Ts ) m(t ) Ts m (t ) m (t ) (t )
m
2! 3!
m(t ) Ts m (t ) for small Ts
m[k ] m[k 1]
m (t )
Ts
Prediction in DPCM
Ts2 Ts3
m(t Ts ) m(t ) Ts m (t ) m (t ) (t )
m
2! 3!
m(t ) Ts m (t ) for small Ts
Finally,
m[k ] m[k 1]
m[k 1] m[k ] Ts
Ts
2m[k ] m[k 1]
Prediction in DPCM
Generalization,
ˆ [k ]
The RHS is the predicted value of m[k], that is, m
Prediction in DPCM
Generalized prediction,
m[k ] a1m[k 1] a2 m[k 2] a N m[k N ]
ˆ [k ]
The RHS is the predicted value of m[k], that is, m
Therefore,
mˆ [k ] m[k 1]
The previous sample is the current sample
Prediction in DPCM
Predicted output,
a1 a2 aN-1 aN
a1 a2 aN-1 aN
Predictor
ˆ [k ]
Output, m
Prediction in DPCM
Predicted output,
d [k ] m[k ] mˆ [k ]
• At receiver, we recalculate m[k] from d[k] and mˆ [k ]
m[k ] d [k ] mˆ [k ]
Prediction in DPCM
Transmitter m[k] + d[k]
To channel
-
mˆ [k ]
Receiver
d[k] + m[k]
Input Output
+
mˆ [k ] Predictor
Prediction in DPCM
Transmitter m[k] + d[k]
To channel
-
mˆ [k ]
At Receiver:
m[k ] mˆ [k ] d [k ]
a1m[k 1] a2 m[k 2] a N m[k N ] d [k ]
d[k] + m[k]
Input Output
+
mˆ [k ] Predictor
Prediction in DPCM
Transmitter m[k] + d[k]
To channel
-
mˆ [k ]
Difficulty at Transmitter:
We DO NOT transmit d[k], rather its quantized version dq[k]
d[k] + m[k]
Input Output
+
mˆ [k ] Predictor
Prediction in DPCM
Transmitter m[k] + d[k] dq[k]
Quantizer To channel
-
Difficulty at Transmitter:
We DO NOT transmit d[k], rather its quantized version dq[k]
d[k] + m[k]
Input Output
+
mˆ [k ] Predictor
Prediction in DPCM
Transmitter m[k] + d[k] dq[k]
Quantizer To channel
-
Receiver
dq[k] + mq[k]
Input Output
+
Predictor
Prediction in DPCM
Transmitter m[k] + d[k] dq[k]
Quantizer To channel
-
Receiver
dq[k] + mq[k]
Input Output
+
Predictor
Prediction in DPCM
Transmitter m[k] + d[k] dq[k]
Quantizer To channel
-
mˆ q [k ]
Predictor will generate
mˆ q [k ] a1mq [k 1] a2 mq [k 2] a N mq [k N ]
Receiver
dq[k] + mq[k]
Input Output
+
mˆ q [k ] Predictor
Prediction in DPCM
Transmitter m[k] + d[k] dq[k]
Quantizer To channel
-
mˆ q [k ]
Receiver
dq[k] + mq[k]
Input Output
+
mˆ q [k ] Predictor
Prediction in DPCM
Transmitter m[k] + d[k] dq[k]
Quantizer To channel
-
+
mˆ q [k ] +
Predictor mq[k]
Receiver
dq[k] + mq[k]
Input Output
+
mˆ q [k ] Predictor
Prediction in DPCM
Transmitter m[k] + d[k] dq[k]
Quantizer To channel
-
+
mˆ q [k ] +
Predictor mq[k]
Predictor mq[k]
Predictor mq[k]
dp
Therefore, quantization step reduction by a factor of
mp
2
dp
As quantization noise N0 = (v) /12, it reduces by a factor of
2
m p
Improvement of DPCM: Delta
Modulation
Improvement of DPCM: Delta
Modulation
DPCM Motivation:
• Successive samples are NOT independent, can be predicted
• Redundancy (CORRELATION) exists among neighboring samples
• Difference can be transmitted instead of actual sample
mp
No. of quantization levels = L
2m p
v
L
-mp
Improvement of DPCM: Delta
Modulation
DPCM Motivation:
• Successive samples are NOT independent, can be predicted
• Redundancy (CORRELATION) exists among neighboring samples
• Difference can be transmitted instead of actual sample
mp
No. of quantization levels = L
d[k] --
2m p
v
L
-mp 1-bit coding
Improvement of DPCM: Delta
Modulation
Delta Modulation:
• 1- bit DPCM
mˆ q [k ] mq [k 1]
Transmitters: DPCM Vs. DM
DPCM m[k] + d[k] dq[k]
Quantizer To channel
-
+
mˆ q [k ] +
Predictor mq[k]
Delay Ts mq[k]
Receivers: DPCM Vs. DM
DPCM
dq[k] + mq[k]
Input Output
+
mˆ q [k ] Predictor
DM
dq[k] + mq[k]
Input Output
+
mq[k-1] Delay Ts
Delta Modulation
Transmitter m[k] + d[k] dq[k]
Quantizer To channel
-
+
mq[k-1]
+
Delay Ts mq[k]
Receiver
dq[k] + mq[k]
Input Output
+
mq[k-1] Delay Ts
Delta Modulation
Transmitter m[k] + d[k] dq[k]
Quantizer To channel
-
+
mq[k-1]
+
Delay Ts mq[k]
+
mq[k-1] Delay Ts
Delta Modulation
+
mq[k-1] Delay Ts
Delta Modulation
+
mq[k-1] Delay Ts
Delta Modulation
Finally, we get k
mq [k ] d q [m]
m 0
Receiver
dq[k] + mq[k]
Input Output
+
mq[k-1] Delay Ts
Delta Modulation
Finally, we get k
mq [k ] d q [m]
m 0
Receiver
dq[k] + mq[k]
Input Output
+
mq[k-1] Delay Ts
Delta Modulation
Alternate Receiver
mˆ q (t ) ~ (t )
√ Input dq[k] Low Pass m
filter
Integrator-
amplifier
+
mq[k-1] Delay Ts
Delta Modulation
Sampler
Alternate Transmitter Comparator frequency fs
E dq[k]
m(t) dq(t)
+ d(t)
mˆ q (t ) - -E
Integrator-
amplifier
Alternate Receiver
dq[k] mˆ q (t ) ~ (t )
m
Input Low Pass
filter
Integrator-
amplifier
Delta Modulation
Sampler
Alternate Transmitter Comparator frequency fs
E dq[k]
m(t) dq(t)
+ d(t)
mˆ q (t ) - -E
E: Threshold
of coding
Integrator-
amplifier
mˆ q (t ) - -E
Integrator-
amplifier
mˆ q (t ) - -E
Integrator-
amplifier
mˆ q (t )
m(t)
dq[k]
Modulated signal to
be transmitted
Delta Modulation
Sampler
Alternate Transmitter Comparator frequency fs
E dq[k]
m(t) dq(t)
+ d(t)
mˆ q (t ) - -E
Integrator-
amplifier
mˆ q (t )
• dq[k] gives rise to a step
function in mˆ q (t ) m(t)
• dq[k] equalizes mˆ q (t )to m(t)
in small step, E
dq[k]
Delta Modulation
Alternate Receiver
dq[k] mˆ q (t ) ~ (t )
m
Input Low Pass
filter
Integrator-
amplifier
mˆ q (t )
• Coarseness of mˆ q (t ) is
removed when passed m(t)
through low pass filter at
receiver
dq[k]
Delta Modulation
Sampler
Alternate Transmitter Comparator frequency fs
E dq[k]
m(t) dq(t)
+ d(t)
mˆ q (t ) - -E
Integrator-
amplifier
mˆ q (t )
• DM encodes and transmits
NOT the actual signal but m(t)
the differences (derivatives)
mˆ q (t ) - -E
Integrator-
amplifier
mˆ q (t )
• DM encodes and transmits
NOT the actual signal but m(t)
the differences (derivatives)
mˆ q (t ) - -E
Integrator-
amplifier slope-overload
mˆ q (t )
• If m(t) changes too fast, mˆ q (t )
cannot follow m(t): slope- m(t)
overload occurs
• Reason: E is not high
enough dq[k]
Delta Modulation
Sampler
Alternate Transmitter Comparator frequency fs
E dq[k]
m(t) dq(t)
+ d(t)
mˆ q (t ) - -E
Integrator-
amplifier slope-overload
Granular noise
mˆ q (t )
• If E is too high, granular
noise occurs : cannot follow m(t)
small changes of m(t)
dq[k]
Digital-to-Digital Conversion:
Line Coding, Block Coding and
Scrambling
From
Data Communications and Networking, 5th Edition
By Behrouz A. Forouzan
Line Coding and Decoding
Line Coding
• Example:
1 → +V
0 → -V
Data Element and Signal
Element
• data element
– smallest entity to represent a piece of information
– 0, 1, etc,
• signal element
– shortest unit of a digital signal
– Each represents one or more data element
Data Element and Signal
Element
• data element
– smallest entity to represent a piece of information
– 0, 1, etc,
• signal element
– shortest unit of a digital signal
– Each represents one or more data element
• signal element
– shortest unit of a digital signal
– Each represents one or more data element
• signal rate, S
– No. of signal elements sent in 1 second
– Other names: baud rate, pulse rate, modulation rate
S ave c N (1 r ) baud
Line Coding Guidelines
• r should be higher, S = N/r should be lower
• Minimize DC components:
• Self synchronization
• Complexity
– Simpler is better
– A coding with FOUR signal levels is more difficult to interpret than one
that uses TWO signal levels
Line Coding Schemes
Unipolar
• All signal levels are on one side of the time axis - either above
or below
– The signal level does not return to zero at middle of the bit
Unipolar: NRZ
• Pros
– Simple
• Cons
r=1
Polar
• Polar NRZ
• Polar NRZ-I
• Pros
– Simple
• Cons
r=1
Polar: Return to Zero (RZ)
• 3 signal levels: +V, -V, 0
– bit 0: -V to 0 bit 1: +V to 0
Polar: Return to Zero (RZ)
• Pros
– NO Baseline wandering, NO DC components
• Cons
r = ⅟2
Polar: Biphase-Manchester and
Differential Manchester
• 2 signal levels: +V, -V
• Manchester
– Combines NRZ-L and RZ
• Differential Manchester
– Combines NRZ-I and RZ
Polar: Biphase-Manchester and
Differential Manchester
• Manchester
– Level determines bits
• AMI
• Pseudoternary
• Pseudoternary
• Pros
– r=1
• Cons
– 3 signal elements
– NO synchronization
– NO error detection
Multilevel Schemes
• Target: to increase r or bit rate
• 2m <Ln :
– Redundant signal patterns
– Very flexible mapping is possible
– Better noise immunity and error detection
Multilevel Schemes
• Coding symbol: short representation is mBnL
• L is replace by character code
– L = 2 : B meaning Binary
– L = 3 : T meaning Ternary
– L = 4 : Q meaning Quaternary
11 00 01 11 10
2B1Q Scheme +3
Rules +1
00: -3 -1
01: -1
10: +3
-3
11: +1
Multilevel Schemes
• Pros
– r=2
– Used in DSL to provide high speed connection to the Internet
• Cons
– 4 signal elements,
– Baseline wandering, DC components are possible
– NO redundancy, NO error detection
11 00 01 11 10
+3
2B1Q Scheme
Rules +1
00: -3 -1
01: -1
10: +3 -3
11: +1
Multilevel Schemes: 8B6T
• 28 = 256 different data patterns
Polar
Block Coding: mB/nB
• replaces m bits by n bits, where m < n
– division
– substitution
– combination
Block Coding: 4B/5B
• Done in combination with NRZ-I which has sync problem for long
sequence of 0’s
• 16 data sequences
• 32 available encoding sequences
• redundant sequences are for error
detection, overhead control
• Maximum 1 leading 0 and 2
trailing 0’s
Block Coding: 8B/10B
• Biphase scheme
– High bandwidth (r < 1)
• Bipolar AMI
– NO sync for long sequence of 0’s
– Other characteristics are OK
Search for best line coding:
Scrambling
• Bipolar AMI
– NO sync for long sequence of 0’s
– Other characteristics are OK
decoding
Scrambling: B8ZS
• Bipolar 8 Zeros (B8ZS)
– Replaces 8 zeros by 0 0 0 V B 0 V B
– Replaces 8 zeros by 0 0 0 V B 0 V B
– Replaces 8 zeros by 0 0 0 V B 0 V B
– Replaces 8 zeros by 0 0 0 V B 0 V B
• Replaces 4 zeros by
– 0 0 0 V if No. of nonzero pulses after last substitution is ODD
– B 0 0 V if No. of nonzero pulses after last substitution is EVEN
Scrambling: HDB3
• High Definition Bipolar 3-Zero (HDB3)
• Replaces 4 zeros by
– 0 0 0 V if No. of nonzero pulses after last substitution is ODD
– B 0 0 V if No. of nonzero pulses after last substitution is EVEN
Scrambling: HDB3
• High Definition Bipolar 3-Zero (HDB3)
• Replaces 4 zeros by
– 0 0 0 V if No. of nonzero pulses after last substitution is ODD
– B 0 0 V if No. of nonzero pulses after last substitution is EVEN
Scrambling: HDB3
• High Definition Bipolar 3-Zero (HDB3)
• Replaces 4 zeros by
– 0 0 0 V if No. of nonzero pulses after last substitution is ODD
– B 0 0 V if No. of nonzero pulses after last substitution is EVEN
Scrambling: HDB3
• High Definition Bipolar 3-Zero (HDB3)
• Replaces 4 zeros by
– 0 0 0 V if No. of nonzero pulses after last substitution is ODD
– B 0 0 V if No. of nonzero pulses after last substitution is EVEN
Odd
Scrambling: HDB3
• High Definition Bipolar 3-Zero (HDB3)
• Replaces 4 zeros by
– 0 0 0 V if No. of nonzero pulses after last substitution is ODD
– B 0 0 V if No. of nonzero pulses after last substitution is EVEN
Even
Scrambling: HDB3
• High Definition Bipolar 3-Zero (HDB3)
• Replaces 4 zeros by
– 0 0 0 V if No. of nonzero pulses after last substitution is ODD
– B 0 0 V if No. of nonzero pulses after last substitution is EVEN
Even