2. Policies and procedures
3 hours of lecture per week(MWF)
2 hours of lab per week(Wed. in CEER 118)
Homework assigned and graded weekly
Lab is due the same day
2 tests and one final
Grade break down
• 15% each test
• 25% final
• 25% labs
• 20% homework
Copyright1999 by BG Mobasseri 2
3. Lab work
Hands-on lab work is an integrated part of
the course
There will be about 10 experiments done
using MATLAB with signal processing and
COMM toolboxes
Experiments, to the extent possible,
parallel theoretical material
Professional MATALB code is expected
Copyright1999 by BG Mobasseri 3
4. Going online
Send a blank message to
ece4790_f99-subscribe@egroups.com
You will then have access to all class
notes, labs etc.
Notes/labs are in MS Office format
You can also participate in the online
discussion group and, if you wish, chat
room
Copyright1999 by BG Mobasseri 4
5. Ethical standards
This course will be online and make full
use of internet. This convenience brings
with it many responsibilities
• Keep electronic class notes/material private
• Keep all passwords/accounts to yourself
• Do not exchange MATLAB code
Copyright1999 by BG Mobasseri 5
6. Necessary Background
This course requires, at a minimum, the
following body of knowledge
• Signal processing
• Probability
• MATLAB
Copyright1999 by BG Mobasseri 6
7. Course Outlook
Introduction Pulse Shaping
• signals, channels • “best” pulse shape
• bandwidth • interference
• signal represent. • equalization
Analog Modulation Digital Modulations
• AM and FM Modem standards
Source coding Spread Spectrum
• Sampling, PAM Wireless
• PCM, DM, DPCM
Copyright1999 by BG Mobasseri 7
16. Periodic vs. Nonperiodic
A periodic signal satisfies the condition
The smallest value of To for which this
condition is met is called a period of g(t)
g(t ) = g(t + T0 )
period
Copyright1999 by BG Mobasseri 16
17. Deterministic vs. random
A deterministic signal is a signal about
which there is no uncertainty with respect
to its value at any given time
• exp(-t)
• cos(100t)
Copyright1999 by BG Mobasseri 17
18. Energy and Power
Consider the following
+ i(t)
V(t) R
Instantaneous power is given by
-
2
v (t ) 2
p( t ) = = Ri(t )
R
Copyright1999 by BG Mobasseri 18
19. Energy
Working with normalized load, R=1Ω
2 2 2
p(t ) = v(t ) = i(t ) = g(t )
Energy is then defined as
T
2
E = lim ∫ g(t ) dt
−T
T→∞
Copyright1999 by BG Mobasseri 19
20. Average Power
The instantaneous power is a function of
time. An overall measure of signal power is
its average power
T 2
1
T→∞ 2T ∫
P = lim g( t ) dt
−T
Copyright1999 by BG Mobasseri 20
21. Energy and Power of a
Sinusoid
Take m (t ) = A cos (2πfc t )
• Find the energy
T T
A2 cos2 (2πfc t )dt ⇒ ∞
2
E = lim
T→∞ ∫
−T
T →∞ ∫ 1 44 2 4 4
g(t ) dt = lim
−T
3
always>0
Copyright1999 by BG Mobasseri 21
22. Instantaneous Power
Instantaneous power
AMPLITUDE= 1, FREQ.=2
1
2 2
p(t ) = A cos ( 2πfc t ) 0.8
0.6
A2 A2 0.4
= + cos( 4πfc t ) 0.2
2 2 0
-0.2 INSTANT. POWER
ORIGINAL SIGNAL
-0.4
-0.6
-0.8
-1
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
TIME(SEC)
Copyright1999 by BG Mobasseri 22
23. Average Power
The average power of a sinusoid is
T 2 T
1 1
T →∞ 2T ∫ T→∞ 2T ∫
P = lim g(t ) dt = lim A2 cos 2 (2πfc t )dt
−T −T
from 2 cos 2 x = 1 + cos(2x ),then
T
1 A2
P = lim
T →∞ 2T ∫ 2 (1 + cos(4πfct ))dt
−T
T T
1 A2 1 A2 A2
T→ ∞ 2T ∫ 2 T →∞ 2T ∫ 2
= lim + lim cos( 4πfc t )dt =
−T 4 4 4 2 4 4 4 3
2
−T 1
averages to zero
Copyright1999 by BG Mobasseri 23
24. Average Transmitted Power
What is the peak signal amplitude in order
to transmit 50KW? Assume antenna
impedance of 75Ω.
• Note the change in Pavg for non-unit
ohm load
⎛ A2 ⎞
Pavg = ⎜ ⎟ R ⇒ A = 2RPavg = 2 × 75 × 50,000
⎝2⎠
⇒ A = 2,738volts
Copyright1999 by BG Mobasseri 24
25. Energy Signals vs.
Power Signals
Do all signals have valid energy and power
levels?
• What is the energy of a sinusoid?
• What is the power of a square pulse?
In the first case, the answer is inf.
In the second case the answer is 0.
Copyright1999 by BG Mobasseri 25
26. Energy Signals
A signal is classified as an energy signal if
it meets the following
0<E<infinity
Time-limited signals, such as a square
pulse, are examples of energy signals
Copyright1999 by BG Mobasseri 26
27. Power Signals
A power signal must satisfy
0<P<infinity
Examples of power signals are sinusoidal
functions
Copyright1999 by BG Mobasseri 27
28. Example:energy signal
Square pulse has finite energy but zero
average power
A E = A2 T
T
1 T 2 1 T 2
P = lim ∫ g(t ) dt = lim ∫ A dt ⇒ 0
T→∞ 2T −T T→∞ 2T −T
123
fixed
Copyright1999 by BG Mobasseri 28
29. Example:power signal
A sinusoid has infinite energy but finite
power
A
∞
E= ∫ A2 cos 2 (2πfc t ) dt ⇒ ∞
−∞
but
T
1 ∞
cos (2πfc t )dt ⇒ ⇒ finite
T→∞ 2T ∫
2
Pavg = lim
−T
∞
Copyright1999 by BG Mobasseri 29
30. SUMUP
Energy and power signals are mutually
exclusive:
• Energy signals have zero avg. power
• Power signals have infinite energy
• There are signals that are neither energy
or power?. Can you think of one?
Copyright1999 by BG Mobasseri 30
32. WHAT IS BANDWIDTH?
In a nutshell, bandwidth is the “highest”
frequency contained in a signal.
We can identify at least 5 definitions for
bandwidth
• absolute
• 3-dB
• zero crossing
• equivalent noise
• RMS
Copyright1999 by BG Mobasseri 32
33. ABSOLUTE BANDWIDTH
The highest frequency
Spectrum
-W W f
Copyright1999 by BG Mobasseri 33
34. 3-dB BANDWIDTH
The frequency where frequency response
drops to .707 of its peak
Spectrum
W f
Copyright1999 by BG Mobasseri 34
35. FIRST ZERO CROSSING
BANDWIDTH
The frequency where spectrum first goes to zero is
called zero crossing bandwidth.
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
-4 -3 -2 -1 0 1 2 3 4
Copyright1999 by BG Mobasseri 35
36. EQUIVALENT NOISE
BANDWIDTH
Bandwidth which contains the same power
as an equivalent bandlimited white noise
1
0.9
0.8
0.7
0.6
0.5
0.4
0.3
0.2
0.1
0
Copyright1999 by BG Mobasseri 36
-4 -3 -2 -1 0 1 2 3 4
37. RMS BANDWIDTH
RMS bandwidth is related to the second
moment of the amplitude spectrum
1
∞ 2
⎡∫−∞ f G( f ) ⎤
2 2
Wrms = ⎢ ∞ 2 ⎥
⎣ ∫−∞ G( f ) ⎦
This measures the tightness of the
spectrum around its mean
Copyright1999 by BG Mobasseri 37
38. RMS BANDWIDTH OF A
SQUARE PULSE
Take a square pulse of duration 0.01 sec.
Its spectrum is a sinc
SQUARE PULSE SPECTRUM (WIDTH=0.01 SEC.)
4.5
4
3.5
3
2.5
2
1.5
1
0.5
0
Copyright1999 by BG Mobasseri 38
0 100 200 300 400 500 600 700 800
FREQUENCY(Hz)
39. RMS BANDWIDTH
The RMS bandwidth can be numerically
computed using the following MATLAB
code
W rms= 35.34 Hz
Copyright1999 by BG Mobasseri 39
40. Bandwidth of Real Signals
This is the spectrum a 3 sec. clip sampled
at 8KHz
0
-10
-20
-30
-40
-50
-60
-70
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Frequency
Copyright1999 by BG Mobasseri 40
41. Gate Function
Gate function is one of the most versatile
pulse shapes in comm.
It is pulse of amplitude A and width T
A
-T/2 T/2
Copyright1999 by BG Mobasseri 41
42. rect function
Gate function is defined based on a
function called rect
⎧ 1 1
1 − <t<
rect(t) = ⎨ 2 2
⎪
⎩0 otherwise
Copyright1999 by BG Mobasseri 42
43. Expression for gate
Based on rect we can write
t
g(t ) = Arect ⎛ ⎞
⎝T ⎠
Note that for -T/2<t<T/2, the argument of
rect is inside {-1/2,1/2} therefore rect is 1
and g(t)=A
Copyright1999 by BG Mobasseri 43
44. Generalizing gate
Let’ say we want a pulse with amplitude A
centered at t=to and width T
T
A
t=to
Copyright1999 by BG Mobasseri 44
45. Arriving at an Expression
What we have a is a rect function shifted
to the right by to
Shift to the right of f(t) by to is written by
f(t-to)
Therefore
⎛t − to ⎞
g(t ) = Arect ⎝
T ⎠
Copyright1999 by BG Mobasseri 45
46. gate function in the Fourier
Domain
The Fourier transform of a gate function is
a sinc as follows
sinc(t)
1
0.8
g(t ) = Arect ⎛t ⎞
⎝T ⎠
0.6
0.4
G( f ) = ATsinc( fT ) 0.2
Zero crossing
0
-0.2
-0.4
-3 -2 -1 0 1 2 3
TIME(t)
Copyright1999 by BG Mobasseri 46
47. Zero Crossing
Zero crossings of a sinc is very significant.
ZC occurs at integer values of sinc
argument
⎧ x =0
1
sin c(x) = ⎨
⎩ x = ±1,±2,L
0
Copyright1999 by BG Mobasseri 47
48. Some Numbers
What is the frequency content of a 1 msec.
square pulse of amplitude .5v?
• We have A=0.5 and T=1ms
g(t ) = 0.5rect(1000t)
G( f ) = 5 × 10 −4 sinc(10 −3 f )
• First zero crossing at f=1000Hz obtained
by setting 10^-3f=1
Copyright1999 by BG Mobasseri 48
49. RF Pulse
RF(radio frequency) pulse is at the heart of
all digital communication systems.
RF pulse is a short burst of energy,
expressed by a sinusoidal function
Copyright1999 by BG Mobasseri 49
50. Modeling RF Pulse
An RF pulse is a cosine wave that is
truncated on both sides
This effect can be modeled by “gating”the
cosine wave
Copyright1999 by BG Mobasseri 50
51. Mathematically Speaking
Call the RF pulse g(t), then
t
g(t ) = Arect ⎛ ⎞ cos(2πfc t )
⎝T ⎠
This is in effect the modulated version of
the original gate function
Copyright1999 by BG Mobasseri 51
52. Spectrum of the RF
Pulse:basic rule
We resort to the following
g1 (t )g2 (t ) ⇔ G1 ( f ) * G2 ( f )
Meaning, the Fourier transform of the
product is the convolution of individual
transforms
Copyright1999 by BG Mobasseri 52
53. RF Pulse Spectrum: Result
We now have to identify each term
AT
gate → g1 (t ) ⇔ sin c(Tf )
2
cosine → g 2 (t ) ⇔ [δ ( f − fc ) + δ ( f + fc )]
Then, the RF pulse spectrum, G(f)
AT
G( f ) = ⎛ sin c(Tf )⎞ * [δ ( f − fc ) + δ ( f + fc )]
⎝ 2 ⎠
AT AT
= sin c(T ( f − fc )) + sin c(T ( f + fc ))
2 2
Copyright1999 by BG Mobasseri 53
54. Interpretation
The spectrum of the RF pulse are two
sincs, one at f=- fc and the other at f=+fc
RF PULSE SPECTRUM:two sincs at pulse freq
-4 0 4
FREQUENCY (HZ)
Copyright1999 by BG Mobasseri 54
57. Definitions:Baseband
The raw message signal is referred to as
baseband, or low freq. signal
Spectrum of a baseband audio signal
0.045
0.04
0.035
0.03
0.025
0.02
0.015
0.01
0.005
0
0 500 1000 1500 2000 2500 3000 3500 4000
FREQUENCY(HZ)
Copyright1999 by BG Mobasseri 57
58. Definitions:Bandpass
When a baseband signal m(t) is modulated,
we get a bandpass signal
The bandpass signal is formed by the
following operation(modulation)
m (t )cos( 2πfc t )
Copyright1999 by BG Mobasseri 58
60. Digital Bandpass
Baseband and bandpass concepts apply
equally well to digital signals
1 1 1 1
0
RF pulses
Copyright1999 by BG Mobasseri 60
61. Baseband vs. Bandpass
Spectrum
Creating a bandpass signal is the same as
modulation process. We have the
following
1
m (t )cos( 2πfc t ) ⇔ [ M( f − fc ) + M ( f + fc )]
2
Interpretation
Copyright1999 by BG Mobasseri 61
62. Showing the Contrast
baseband
frequency
Bandwidth doubles bandpass bandwidth
Copyright1999 by BG Mobasseri 62
64. Introduction
We need a formalism to follow a signal as
it propagates through a channel.
To this end, we have to learn a few
concepts including Hilbert Transform,
analytic signals and complex envelope
Copyright1999 by BG Mobasseri 64
65. Hilbert Transform
Hilbert transform is an operation that
affects the phase of a signal
+90
H(f) f
-90
|H(f)|=1
Phase response
Copyright1999 by BG Mobasseri 65
66. More precisely
⎧ − jπ
⎪1e 2 f ≥ 0
H( f ) = ⎨
π
⎪ 2j
⎩1e f < 0
Equivalently
H ( f ) = − jsgn( f )
sgn( f ), or signum function, extracts the sign of its argument
⎧f >0
⎪
1
sgn( f ) = ⎨ f = 0
0
⎪ f <0
⎩−1
Copyright1999 by BG Mobasseri 66
67. HT notation
The HT of g(t) is denoted by
HT ( g(t )) = g(t )
ˆ
In the frequency domain,
ˆ
G( f ) = − j sgn( f )G( f )
Copyright1999 by BG Mobasseri 67
68. Find HT of a Sinusoid
Q: what is the HT of cosine?Ans:sine
g(t ) = cos(2πfc t )
we know
⎡ ⎛ ⎞⎤
Gˆ ( f ) = − j sgn( f )G( f ) = − j sgn( f )⎢1 ⎜δ ( f − fc ) + δ ( f + fc )⎟⎥
⎢2 ⎜14 2 43 14 2 43 ⎟⎥
⎣ ⎝ pos. freq neg. freq ⎠⎦
j j 1
= − δ ( f − fc ) + δ ( f + fc ) = [δ ( f − fc ) − δ ( f + fc )]
2 2 2j
1 4 4 4 4 2 4 4 4 43
Fourier transform of sine
Copyright1999 by BG Mobasseri 68
69. HT Properties
Property 1
• g and HT(g) have the same amplitude
spectrum
Property 2
Property 3 HT ( g(t )) = −g(t )
ˆ
• g and HT(g) are orthogonal, i.e.
∫ g(t )g(t )dt = 0
ˆ
Copyright1999 by BG Mobasseri 69
70. Using HT: Pre-envelope
From a real-valued signal, we can extract a
complex-valued signal by adding its HT as
follows
g+ (t ) = g(t ) + jg(t )
ˆ
g+(t) is called the pre-envelope of g(t)
Copyright1999 by BG Mobasseri 70
71. Question is Why?
It turns out that it is easier to work with g+
(t) than g(t) in many comm. situations
We can always go back to g(t)
g(t ) = Re { g+ (t )}
where Re stands for " real part of"
Copyright1999 by BG Mobasseri 71
72. Pre-envelope Example
Find the pre-envelope of the RF pulse
g(t ) = m (t )cos( 2πfc t + θ )
We can re-write g(t) as follows
j (2πfc t +θ )
g(t ) = Re{ m (t )e }
because
j (2πfc t +θ )
e = cos( 2πfct + θ ) + j sin( 2πfc t + θ )
Copyright1999 by BG Mobasseri 72
73. Pre-envelope is...
Compare the following two
j (2πfc t +θ )
g(t ) = Re{ m (t )e }
g(t ) = Re{ g+ ( t )} g+ ( t ) = m(t )e
j ( 2πfc t +θ )
Pre-envelope of
is g(t ) = m (t )cos( 2πfc t + θ )
j ( 2πfc t +θ )
g+ ( t ) = m(t )e
Copyright1999 by BG Mobasseri 73
74. Pre-envelope in the Frequency
Domain
How does pre-envelope look in the
frequency domain?
We know g+ (t ) = g(t ) + jg( t ). Fourier transform is
ˆ
F{ g+ (t )} = G+ ( f ) = G( f ) + j[ − j sgn( f )]G( f )
Copyright1999 by BG Mobasseri 74
75. Pre-envelope in positive and
negative frequencies
Let’s evaluate G+(f) for f>0 G(f)
f >0
G+ ( f ) = G( f ) + G( f ) = 2G( f ) f
f <0
G+(f)
G+ ( f ) = G( f ) − G( f ) = 0
f
Copyright1999 by BG Mobasseri 75
76. Interpretation
Fourier transform of Pre-envelope exists
only for positive frequencies
As such per-envelope is not a real signal. It
is complex as shown by its definition
g+ (t ) = g(t ) + jg(t )
ˆ
Copyright1999 by BG Mobasseri 76
77. Corollary
To find the pre-envelope in the frequency
domain, take the original spectrum and
chop off the negative part
Copyright1999 by BG Mobasseri 77
78. Example
Find the pre-envelope of a modulated
message
g(t ) = m (t )cos( 2πfc t + θ )
G(f) G+(f)
AM signal
Copyright1999 by BG Mobasseri 78
79. Another Definition
for Pre-envelope
Pre-envelope is such a quantity that if you
take its real part, it will give you back your
original signal
g(t ) = m (t )cos( 2πfc t + θ ) original signal
j ( 2πfc t +θ )
g(t ) = Re{ m (t )e j (2πft +θ ) g+ ( t ) = m(t )e
}
g(t ) = Re{ g+ ( t )}
Copyright1999 by BG Mobasseri 79
80. Bringing Signals Down to
Earth
Communication signals of interest are
mostly high in frequency
Simulation and handling of such signals
are very difficult and expensive
Solution: Work with their low-pass
equivalent
Copyright1999 by BG Mobasseri 80
81. Tale of Two Pulses
Consider the following two pulses
Which one carries more “information”?
Copyright1999 by BG Mobasseri 81
82. Lowpass Equivalent Concept
The RF pulse has no more information
content than the square pulse. They are
both sending one bit of information.
Which one is easier to work with?
Copyright1999 by BG Mobasseri 82
83. Implementation issues
It takes far more samples to simulate a
bandpass signal
0.01 sec. Sampling rate=200Hz
4cycles/0.01 sec Sampling rate=800Hz
-->fc=400Hz
Copyright1999 by BG Mobasseri 83
84. Complex Envelope
Every bandpass signal has a lowpass
equivalent or complex envelope
Take and re write as
g(t ) = m (t )cos( 2πfc t + θ )
⎧ ⎫
j (2πfc t +θ )
⎪ jθ j 2πfc t
⎪
g(t ) = Re{ m (t )e } = Re ⎨ m (2)e3
1 t e ⎬
⎪ lowpass ⎪
⎩ complex envelop
or ⎭
Copyright1999 by BG Mobasseri 84
85. Complex Envelope:
The Quick Way
Rewrite the signal per following model
j (2πfc t +θ )
g(t ) = Re { m (t )e
} = Re { m(t )e jθ e j 2πfc t }
The term in front of j2πf t is the complex
envelope shown by e c
jθ
g( t ) = m(t )e
˜ m and theta contain all the
information
Copyright1999 by BG Mobasseri 85
86. Signal Representation
Summary
Take a real-valued, baseband signal
G(f)
g(t)
Copyright1999 by BG Mobasseri 86
88. Pre-envelope Summary:
bandpass
Baseband signal: g(t ) = m (t )cos( 2πfc t + θ )
G(f)
G+(f)
Copyright1999 by BG Mobasseri 88
89. Complex Envelope Summary
Complex/pre envelope are related
− j 2πfc t G+(f)
g( t ) = g+ (t)e
˜
or
ˆ
G( f )
j 2πfc t
g+ (t) = g(t )e
˜
Copyright1999 by BG Mobasseri 89
90. RF Pulse: Complex Envelope
Find the complex envelope of a T second
long RF pulse at frequency fc
t
g(t ) = Arect ⎛ ⎞ cos(2πfc t )
⎝T ⎠
Copyright1999 by BG Mobasseri 90
91. Writing as Re{ }
Rewrite g(t) as follows
⎧Arect ⎛ t ⎞e j 2πfct ⎫
g(t ) = Re ⎨
⎝T ⎠ ⎬
⎩ ⎭
compare
g(t ) = Re{ g (t)e j2πfct }
˜ Comp.Env=just a square
pulse
then
t
g (t) = complex _ envelope = Arect ⎛ ⎞
˜ ⎝T ⎠
Copyright1999 by BG Mobasseri 91
92. RF Pulse Pre-envelope
Recall
− j 2πfc t
Then
g( t ) = g+ (t)e
˜
g+ (t ) = Arect ⎛ t ⎞e j 2πf ct
⎝T ⎠
Copyright1999 by BG Mobasseri 92
93. Story in the Freq. Domain
RF PULSE SPECTRUM:two sincs at pulse freq RF PULSE SPECTRUM:two sincs at pulse freq
-4 0 4 -4 0 4
FREQUENCY (HZ) FREQUENCY (HZ)
Original RF pulse spectrum Pre-env. Spectrum
(only f>0 portion)
Copyright1999 by BG Mobasseri 93
96. Signal Transmission Modeling
One of the most common tasks in
communications is transmission of RF
pulses through bandpass channels
Instead of working at high RF frequencies
at great computational cost, it is best to
work with complex envelope
representations
Copyright1999 by BG Mobasseri 96
97. Channel I/O
To determine channel output, we can work
with complex envelopes
˜ 1 ˜ ˜
Y ( f ) = H ( f ) X( f )
2
where
˜
X( f ) : C.E. of input(transmitted) signal
˜
H ( f ) : C.E. of channel transfer function
˜
Y ( f ) : C.E. of output(received) signal
Copyright1999 by BG Mobasseri 97
98. Passing an RF Pulse through a
Bandpass Channel
Here is the problem: what is the output of
an ideal bandpass channel in response to
an RF pulse?
H(f)
use
˜( f ) = 1 X ( f ) H( f )
Y ˜ ˜
2
Copyright1999 by BG Mobasseri 98
99. What is the Complex envelope
of H(f)?
It is the lowpass equivalent of H(f)
2
˜ ( f ) = 2rect ⎛ f ⎞
H
1
⎝2B⎠
2B B
H(f)
Copyright1999 by BG Mobasseri 99
100. What is the Complex Envelope
of the RF Pulse?
We found this before
⎧Arect ⎛ t ⎞e j 2πfct ⎫
g(t ) = Re ⎨
⎝T ⎠ ⎬
⎩ ⎭
compare
g(t ) = Re{ g (t)e j2πfct }
˜
then
t
g (t) = complex _ envelope = Arect ⎛ ⎞
˜ ⎝T ⎠
Copyright1999 by BG Mobasseri 100
101. Channel Output
Here is what we have
f
H ( f ) = 2rect ⎛ ⎞
˜
• Channel complex envelope ⎝2B⎠
• Input complex envelope
B=bandwidth
t
x (t ) = Arect ⎛ ⎞ ⇔ X ( f ) = AT sin c( fT )
˜ ˜
• Output ⎝T ⎠
1 f
Y ( f ) = [ AT sin c( fT)] ⎡2rect( )⎤
˜
2 ⎣ 2B ⎦
Copyright1999 by BG Mobasseri 101
102. Interpretation
˜ f 1
Y ( f ) = AT sin c( fT)rect( )
2B
0.8
0.6
For the pulse to get through 0.4
unscathed, channel bandwidth 0.2
must be larger than pulse bw 0
-0.2
B>=1/T=bit rate -0.4
-10 -8 -6 -4 -2 0 2 4 6 8 10
1/T
B
Copyright1999 by BG Mobasseri 102
103. What Does Distortion Do?
Channel Distortion creates pulse
“dispersion”
Channel
interference
Copyright1999 by BG Mobasseri 103
104. Case of No Distortion
There are two “distortions” we can live
with
• Scaling
• Delay
To
Copyright1999 by BG Mobasseri 104
105. Modeling Distortion-free
Channels
The input-output relationship for a
distortion-free channel is
y(t)=Ax(t-Td)
• x(t):input
• y(t)=output
• A:scale factor
• Td: delay
Copyright1999 by BG Mobasseri 105
106. Response of a Distortion-free
channel
What is channel’s frequency response?
Take FT of the I/O expression
Y ( f ) = AX ( f )e − j2πfTd
Then
Y( f )
H( f ) = = Ae − j2πfTd
X( f )
Copyright1999 by BG Mobasseri 106
107. Amplitude and Phase
Response
|H(f)|
Const amplitude response
f
/_H(f)
Linear phase response
f
−(2πTd ) f
Copyright1999 by BG Mobasseri 107
108. Complete Model
The complete transfer function is
H ( f ) = Ae − j 2πfTd
Since this is a lowpass function, its
complex envelope is the same as H(f)
H ( f ) = Ae − j 2πfTd
˜
Copyright1999 by BG Mobasseri 108
109. Lowpass Channel
Is a first order filter an appropriate model
for a distortion-free channel?
To answer this question we have to test
the definition of the ideal channel
R
C
Copyright1999 by BG Mobasseri 109
110. Amplitude and Phase
Response
amplitude _ response =
a
H( f ) = 2
a + (2πf )
3-dB bandwidth=
phase _ response =
a/2pi=1/(2piRC)
−1 ⎛ 2πf ⎞
θh ( f ) = −tan
⎝ a ⎠
1
a=
RC
Copyright1999 by BG Mobasseri 110
112. An “ideal” Channel?
We must have constant amplitude
response and linear phase response.
Do we?. Deviation of H(f) from the ideal is
tolerated up to .707form the peak.
The frequency at which this occurs is the
3dB bandwidth
No signal distortion if input frequencies are kept
below 3dB bandwidth or 159 Hz here
Copyright1999 by BG Mobasseri 112
113. Linear Distortion
If any of the ideal channel conditions are
violated but we are still dealing with a
linear channel, we have linear distortion
amplitude
H ( f ) = { (1 + k cos (2πfT ))e− j2πftd
f
phase
Copyright1999 by BG Mobasseri 113
114. Pulse Dispersion
Putting a pulse g(t) through this filter
produces 3 overlapping copies
channel with
T distortion
>T
Copyright1999 by BG Mobasseri 114
115. Why?
Let g(t) and r(t) be the transmitted and
received signals. Then
R( f ) = G( f )H( f ) = G( f )[1 + kG( f )cos(2πfT )]e− j2πftd
= G( f )e− j2πftd + kG( f )cos(2πfT)e− j2πftd
Taking the inverse FT
k
r(t) = g(t − td ) + [ g(t − td − T ) + g(t − td + T )]
2
Copyright1999 by BG Mobasseri 115
116. Nonlinear Distortion
This is the most serious kind where input
and output are related by a nonlinear
equation
Nonlinear
g channel r
r r=g^2
g
Copyright1999 by BG Mobasseri 116
117. Impact of Nonlinear Dist.
Nonlinear channels generate new
frequencies at the output that did not exist
in the input signal. Why?
G(f)
if f
W
r(t ) = g 2 (t ) R(f)
then
R( f ) = G( f ) * G( f )
2W f
Copyright1999 by BG Mobasseri 117
118. Practice Problems
For pre-envelope: 2.23
For filtering using complex envelope: 2.32
Copyright1999 by BG Mobasseri 118