Вы находитесь на странице: 1из 17

3/7/2011

1
Communication Systems
Instructor:
Mr. Salman M. Khan
Lecture # 4
What if we make a non periodic signal to repeat itself after an
infinite time period?
Then the exponential Fourier Series for the signal that repeats
itself after infinite period is given by
where,
Fourier series of Aperiodic Signals

0
T
) ( ) ( lim
0
0
t g t g
T
T
=

=
=
n
t jnw
n T
e D t g
0
0
) (
0
0
2
,
T
w

=

=
2 /
2 /
0
0
0
0
0
) (
1
T
T
t jnw
T n
dt e t g
T
D

0
T
3/7/2011
2
The Concept
Bridge Between Fourier Series and Transform
(contd)
As the period T, the fundamental frequency
0
0. So, the
distance between the two consecutive a
k
s becomes zero, and the
sketch of a
k
becomes continuous, what is called as Fourier
Transform.
At the other side, as T , the signal x(t) becomes aperiodic and
takes the form:
This means the Fourier Transform can represent an aperiodic signal
on the frequency-domain.
x(t)
t
-T
1
0 T
1
3/7/2011
3
Now, let us define G(w) a continuous function of w
Hence, where, is the envelop of the
coefficients
And,
As, , (infinitesimally small). Hence we replace w by more
appropriate term
where,

=
=
n
t jnw
T
e
T
nw G
t g
0
0
0
0
) (
) (

= dt e t g
T
D
t jnw
n
0
) (
1
0

0
T

= dt e t g w G
jwt
) ( ) (
) (
1
0
0
nw G
T
D
n
=
0
0
w
) (
1
0
w G
T
|
|

\
|
n
D


=
n
t w n j
T
e
w w n G
t g
) (
2
) (
) (
0

w
T

=
2
0
w
We have represented a non-periodic signal by a Fourier Integral rather
than a Fourier series
We call G(w) the direct Fourier transform of g(t), and g(t) the Inverse
Fourier transform of G(w)
g(t) and G(w) are a Fourier transform pair
The Fourier Transform G(w) is generally complex. Typically represented
by its amplitude and phase
Conjugate Symmetry Property
For real signals,
and
) ( ) ( w G t g


= dw e w G t g
jwt
) (
2
1
) (

) ( ) ( w G w G

=
) (w G
) (w G
) ( ) ( w G w G = ) ( ) ( w G w G =
3/7/2011
4
Rectangular pulses are a window in the time domain and a filter in the frequency
domain
Fourier Transform is a sinc function with infinite frequency content
Stretching a pulse in a time domain causes its Fourier Transform to shrink in the
frequency domain
A signal cannot be both time-limited and band-limited
Fourier Transform of a Rectangular Pulse
.5T -.5T
A
t
w
Infinite Frequency Content
( ) T A w G T t Arect t g w Tsinc ) ( ) / ( ) ( = =
Linearity
Time Shift
Frequency Shift (modulation)
Multiplying in time by any exponential leads to a frequency shift
DC Calculation
Duality & Conjugation (homework exercise)
Time Scaling (homework exercise)
Convolution & Multiplication (homework exercise)
Multiplication in time leads to Convolution in frequency and vice versa
and
Useful properties of Fourier Transform
) ( ) ( ) ( ) (
2 1 2 1
w bG w aG t bg t ag + +
|

\
|

a
w
G
a
at g
1
) (
) ( ) ( w G e t g
jw


= dt t g G ) ( ) 0 (
) ( ) ( ) ( * ) (
2 1 2 1
w G w G t g t g ) ( * ) ( ) ( ) (
2 1 2 1
w G w G t g t g
) ( ) (
c
t jw
w w G e t g
c

3/7/2011
5
Delta Function
Exponentials
Sinusoids (modulation)
Delta Function Train
Special Functions
d(t)
0
w
c
Ad(w-w
c
)
Ae
jwct

Acos(w
c
t)
w
c
.5Ad(w-w
c
)
-w
c
.5Ad(w+w
c
)
T
s

n
d(t-nT
s
)
0

n
d(t-n/T
s
)
0
T
s 2T
s
3T
s
-T
s
-2T
s
-3T
s
1/T
s
-1/T
s
Center for Advanced Studies in
Engineering
Output equals input except for amplitude scaling and/or delay
Simple equalizers invert channel distortion
Can enhance noise power
Distortion-less Transmission
x(t)
Kx(t-t)
h(t)=Kd(t-t)
X(w)
H(w)=Ke
-jwt
Ke
-jwt
X(w)
X(w) X(w)+(w)/H(w) H(w) 1/H(w)
(w)
Channel Equalizer
+
3/7/2011
6
Filter response to is an impulse response
For any input x(t), filter output is x(t)*h(t)
Filtering
d(t) h(t)
LTI
Filter
x(t)
x(t)*h(t)
h(t)
X(w) H(w)X(w)
Much easier to study filtering in the frequency domain
) (t
The Fourier Transform represents the spectral components
of the signal
For band-limited signals, bandwidth B defined as range of
positive frequencies for which |G(w)| > 0
In practice, all signals are time-limited
Need alternative definitions of bandwidth that indicate how much spectrum
a signal occupies
Common definition includes Null-to-Null and 3dB bandwidth definitions
When a real baseband signal is up-converted to a carrier frequency, its
Bandwidth doubles
Signal Bandwidth
|G(w)|
2B
0
Band-limited
|G(w)|
2B
0
ull-to-ull
|G(w)|
2B
0
3dB
-3dB
3/7/2011
7
Fourier transform is the spectral components of a signal
Rectangle in time is a sinc in frequency
Time-limited signals are not band-limited and vice versa
Delta function, sinusoids and exponentials key in Fourier analysis
Delta function train in time is a delta function train in freq.
Communication channels introduce distortion
Equalizers compensate for channel distortion but enhance noise
Signal bandwidth definition depends on its use
Main Points
Examples
) ( ) ( t t x =
) ( ) (
0
t t t x =
3/7/2011
8
Fourier Transform of a Periodic
Signal

+

+

=
=
) ( 2 ) ( ,
) ( ,
0
0

k a X Then
e a t x if
k
t jk
k
t j IFFT
e
0
) ( 2 ,
0

Q
TRAI IMPULSE t x = ) (
FT of an Impulse Train
T
s

n
d(t-nT
s
)
0

n
d(t-n/T
s
)
0
T
s 2T
s
3T
s
-T
s
-2T
s
-3T
s
1/T
s
-1/T
s
3/7/2011
9
Examples Cont.
(Unit Step Function)


j
t u t x
FFT
1
) ( ) ( ) ( + =
FT of a sinusoidal function
3/7/2011
10
Output equals input except for amplitude scaling and/or delay
Simple equalizers invert channel distortion
Can enhance noise power
Distortion-less Transmission
x(t)
Kx(t-t)
h(t)=Kd(t-t)
X(w)
H(w)=Ke
-jwt
Ke
-jwt
X(w)
X(w) X(w)+(w)/H(w) H(w) 1/H(w)
(w)
Channel Equalizer
+
We can determine Signal Energy from its Fourier
Transform using Parsevals Theorem
Generalized form of Signal Energy is
g*(t) can be related to its spectrum as
Hence,
Parsevals Theorem
dt t g E
g
+

=
2
| ) ( |
dt t g t g E
g
+

= ) ( ) (
*

= dw e w G t g
jwt
) (
2
1
) (
* *

dw w G dw w G w G E
dw dt e t g w G
dt dw e w G t g E
g
jwt
jwt
g



+

+

+

+

+

+

= =

=
2 *
*
*
| ) ( |
2
1
) ( ) (
2
1
) ( ) (
2
1
) (
2
1
) (

3/7/2011
11
Signal transmission through a linear
system
Signal transmission through a linear
system
3/7/2011
12
Signal transmission through a linear
system
Distortion caused by multipath effects
3/7/2011
13
Distortion caused by multipath effects
For band-limited signals, bandwidth B defined as range of
positive frequencies for which |G(w)| > 0
The range of frequencies that the signals FT has a power above a
certain threshold
The range of frequencies which contains a specific percentage of the
total power
Need alternative definitions of bandwidth indicating the spectrum
occupied by the signal
Eg: Null-to-Null bandwidth & 3dB Bandwidth definitions
Signal Bandwidth in Analog Systems
|G(w)|
2B
0
ull-to-ull
|G(w)|
3dB
-3dB
3/7/2011
14
In Digital Systems:
It is a synonym for the BAUD Rate (the rate at which
symbols may be transmitted)
E.g.: Transmitting a 56kb/s information over a phone line
using a bandwidth of only 8 kHz.
In Optical Fiber Comm. Systems:
The width of the optical frequency range that can be
transmitted by the fiber
Depends on the dispersion characteristics of the fiber
and many other factors
The bandwidth of the Source (Laser/LED) or Receiver
(Photo diodes) depends on the rise & fall time of the
light source
Signal Bandwidth
Measures Distribution of Signal Energy over Frequency
Energy Spectral Density and autocorrelation are FT pairs
Auto-correlation ESD
Energy Spectral Density (ESD)
dw w dt t g E
g
) (
2
1
| ) ( |
2

+

+

= =


g
(w)=|G(w)|
2
w
) ( ) ( ) ( ) ( ) ( ) ( ) ( ) ( ) (
2
*
w w G w G w G g g dt t g t g
g g
= = = + =



) (
g
t
0
) ( ) ( w
g g

3/7/2011
15
Example: Estimate the essential bandwidth W (rad/sec) of the
signal if the essential band is required to contain
95% of the signal energy.
95% of the energy is
Energy of the signal defined in terms of its spectral contents is
Energy Spectral Density (ESD) cont
) 0 ( ) (

a t u e
at

d E
g g
) (
2
1

+

=

d
a
G
a j
G t u e
g
F at
) (
1
) (
1
) ( ) (
2 2
2
=
+
=
+
=

( )
a
dt t u e E
at
g
2
1
) (
2
= =

a
E
g
2
95 . 0
% 95
=
Example: Estimate the essential bandwidth W (rad/sec) of the
signal if the essential band is required to contain
95% of the signal energy.
The 95% BW is contained in (W,+W)
Energy Spectral Density (ESD) cont
Hz a OR s rad a W
a a a
d
a a
d E
W
W
W
W
W
W
g g
) ( 02 . 2 / ) ( 706 . 12
tan
2
1
2
95 . 0
1
2
1
2
95 . 0
) (
2
1
1
2 2
% 95
=
|

\
|
=
+
=
=
+

\
|
=
+

a
x
a
dx
a x
1
2 2
tan
1 1
Q
) 0 ( ) (

a t u e
at
3/7/2011
16
Similar to ESD but for power signals (periodic or random)
Distribution of signal power over frequency
Power Spectral Density (PSD)



+



=

=
=
dw
T
w G
dw w G
T
T
E
P
T
T
T T
gT
T g
2
2
| ) ( |
lim
2
1
| ) ( |
2
1 1
lim
lim

2
) (
1
lim ) ( w G
T
w S
T
T
g

=
g(t) g
T
(t)
-T T 0

T=

Density Spectral Power
Just like ESD, The autocorrelation of power signals
and the PSD are FT pairs
Power Spectral Density (PSD) (cont)
) (
1
lim
) (
1
lim ) ( ) (
1
lim ) ( ) (
1
lim
) ( ) (
1
lim ) (
2 *
w S
T
w G
T
w G w G
T
g g
T
dt t g t g
T
R
g T
T T T T T T T T
T T T g



=
= =
+ =



R
g
(t)
t
0
) ( ) ( w S R
g g

3/7/2011
17
Signal bandwidth definition depends on its use
Energy spectral density measures signal energy distribution
across frequency domain
Autocorrelation function of energy signal measures signal self
similarity versus delay
A energy signals autocorrelation and ESD are FT pairs
Power signals are characterized with PSD and autocorrelation,
which are FT pairs.
Main Points
Table

Вам также может понравиться