361 Lec4.1 ECE 361 Computer Architecture Lecture 4: MIPS Instruction Set Architecture.
Lecture Set 4
-
Upload
shamim-ali -
Category
Documents
-
view
218 -
download
0
description
Transcript of Lecture Set 4
3/17/2015
1
ECE 219 : Signal, System & Control
Chapter 4: Continuous-Time Signal Analysis: The Fourier Transform
Navneet Upadhyay Dept. of Electronics & Communication Engineering
The LNM Institute of Information Technology, Jaipur
Link Between Fourier Transform (FT) and Fourier Series (FS) (I)
Limiting process: An a-periodic signal can be expressed as a continuous sum (integral) of everlasting exponentials (or sinusoids).
xTo(t) x (t)
txtx TT 0
0
lim
i.e. pulses in periodic signal repeats after an infinite interval.
FS representing xT0(t) will also represent x(t) in the limit T0→∞
where
and
The Exponential FS of xT0(t)
Link Between Fourier Transform (FT) and Fourier Series (FS) (II)
i.e. Dn are (1/T0) times the samples of X(ω), which are uniformly spaced at intervals of ω0
0000 nDT
Doubling T0, halves the fundamental frequency ω0 and envelop 1/T0 X(ω), but not shape
Integrating xT0(t) over (-T0/2, T0/2) is the same as integrating x(t) over (-∞.∞)
To see the nature of the spectrum as T0 ↑, define X(ω) as a
continuous function of ω
Relation between Dn and X(ω)
i.e. spectrum is so dense that the spectral component are spaced at zero intervals, and amp. of each component is zero.
Link Between FT and FS (III)
From Eq. (1) and Eq. (4)
000 T replace ω0 by more appropriate notation Δω, Δω=2π/T0
xT0(t) is the sum of everlasting exponentials of frequencies 0,+Δω,+2Δω,+3Δω,
(FS). Amount of component of frequency nΔω is . In the limit
and
,....3,2,,0
n
2
nX
00 T txtxT 0
tjeX Area under the function . Thus
dtetxtxFX tj
)()(
The signal is approximated by sum of complex exponentials with Dn = X(n∆ω). [EFS] FS becomes the Fourier integral in the
limit of T0→∞
3/17/2015
2
The Fourier Transform Spectrum
The Inverse Fourier transform:
deXXFtx tj)(2
1)( 1
The Fourier transform:
The Amplitude (Magnitude) Spectrum The Phase Spectrum
i.e. if x(t) is a real function of t, then X(ω) and X(-ω) are complex conjugate.
The amplitude spectrum is an even function and the phase is an odd function.
)()()(
sincos)()(
X
tj
eXX
tdttxjtdttxdtetxX
dtetxX tj
)()(
*
*)(
XX
XX
Transform pair ( ) ( )x t X
For all ω
Summary: Definition of Fourier Transform
FS is used to represent periodic signal in term of sinusoidal or exponentials ejn0t.
Fourier Transform is used to represent a-periodic (not periodic) signal in term of exponentials ejt. The forward and inverse FT are defined for a-periodic signal as:
Existence of the Fourier Transform (I) A signal x(t) is said to have a Fourier transform in the ordinary
sense if the above integral converges
Fact: The integral does converge if
1. the signal x(t) is “well-behaved”
2. and x(t) is absolutely integrable, namely,
well behaved means that the signal has a finite number of
discontinuities, maxima, and minima within any finite time interval
then X(ω) exists for every frequency ω and is continuous.
Fact: if x(t) has finite energy, i.e.,
then X(ω) exists for “most” frequencies ω and has finite energy.
Existence of the Fourier Transform (II) Fact: if x(t) is periodic and has a Fourier series, then
is a weighted sum of impulses in frequency domain.
Consider the signal
| ( ) |x t dt dt
Clearly x(t) does not satisfy the first requirement since
Therefore, the constant signal does not have a Fourier transform in the
ordinary sense
Consider the signal
1tx
tuetx at
t
t
tjatjatja
tjat
eja
dtedte
dtetuex
0
0
0 10
If a<0, FT, X(ω) does not exist
3/17/2015
3
Fourier Transform Examples
One-sided exponential decay is defined by e−atu(t) with a > 0:
The Fourier transform of one-sided decay is:
Since x(t) has finite area, its transform is continuous.
x(t) is real but its transform is complex valued.
Fourier integral of e−atu(t) does not converge for a < 0. Hence FT of e−atu(t) does not exixt for a < 0 (growing exponential).
If a=0, x(t)=u(t), and X(ω) DOES NOT EXIST IN ORDINARY SENCE.
Fourier Transform Examples (Cont’d) We can rationalize X(ω).
We can better picture X(ω) using polar representation
FT produce continuous frequency spectra while FS produce discrete line spectra
with nonzero values at specific frequencies.
Fourier Transform Examples (Cont’d)
Fourier transform at ω = 0:
is the area under x(t), called the DC value.
For x(t) = e−atu(t), X(0) =1/a is the only real value and largest in magnitude.
For one-sided decay, X(−ω) = X∗(ω), complex conjugate of X(ω).
This is true for all real-valued signals.
We need only positive frequencies for real-valued signals.
Define Three Useful Functions
A unit rectangular window (also called a unit gate) function rect(x):
A unit triangle function Δ(x):
Interpolation function Sinc(x): The sinc function is very important. Sadly, it has two definitions.
2/|| 1
2/|| 5.0
2/|| 0
x
x
xx
rect
2/|| /21
2/|| 0
xx
xx
x
xxcor
x
xxc
sinsin
sinsin
3/17/2015
4
More about Sinc(x) function
sinc(x) is an even function of x.
02sin nc ;If n ≠ 0
sinc(x) = 0 when sin(x) = 0 except when
x=0, i.e. x = ±π,±2π, ±3π…..
sinc(0) = 1 (derived with L’Hôpital’s rule)
1sin
0sin lim0
x
xc
x
;1sin 2xdxc Thus sinc(x) has a finite energy FT
dtcxsin
sinc (x) is a band-limited pulse with no frequency content for f >1/2.
Fourier Transform of x(t) = rect(t/τ) Evaluation:
Since rect(t/τ) = 1 for -τ/2 < t <τ/2 and 0 otherwise
Fact: every finite width pulse has a transform with unbounded frequencies.
dX
2
2
tf
Amplitude spectrum
Phase spectrum
Fourier Transform of unit impulse x(t) =δ(t)
Using the sampling property of the impulse, we get:
IMPORTANT – Unit impulse contains COMPONENT AT EVERY FREQUENCY.
dtett tj)()]([F 10
t
tje
3/17/2015
5
Inverse Fourier Transform of δ(ω)
Using the sampling property of the impulse, we get:
Spectrum of a constant (i.e. d.c.) signal x(t)=1 is an impulse 2πδ(ω).
1Ft
de tj
2
1
22
tjtj ed
e
090,2
1 tt
Inverse Fourier Transform of δ(ω-ω0)
Using the sampling property of the impulse, we get:
Spectrum of an everlasting exponential ejω0t is a single impulse at ω = ω0.
Fourier Transform of everlasting Sinusoid cosω0t
Remember Euler formula:
Spectrum of cosine signal has two impulses at positive and negative frequencies.
Use result from previouss slide:
The impulse pairs at + and − frequencies correspond to two
phases.
u(t) is not absolutely integrable.
Approach this by considering u(t) to be a decaying exponential e-atu(t) in the limit a→0
tuetu at
a
0lim
da
a22
220lim
a
a
a
j
U1
when a→0, function approaches to
1 for all ω≠0 and its area π
concentrate at a single point ω =0
FT of Unit Step Function u(t)
Thus
ja
a
aj
a
aU
jaUtuF
a
a
a
1lim
lim
1lim
220
22220
0
Area under this function is π regardless the value of a
()
0
|U()|
3/17/2015
6
sgn(t) in terms of u(t)
tutut sgn
Not absolutely integrable, approximate sgn(t) to be sum of exponentials e-atu(t)–eat u(t) in the limit a→0
This looks like
FT of sgn(t)
0,2
0,2
0tansgn,
2sgn
2sgn
limsgn
1
0
jt
tuetuet atat
a
Example: FT of u(t) interms of sgn(t)
)sgn(2
1
2
1ttu
Transform of the signum function (or sign function) is
Therefore
let a→0 the exponential function resembles more and more closely the signum function.
Fourier Transform of any periodic signal
Fourier series of a periodic signal x(t) with period T0 is given by:
Take Fourier transform of both sides, we get:
n
tjn
neDtxX 0)]([)( FF
n
tjn
n eD ][ 0F
n
n nD )(2 0
n
n nD )(2 0
n
n nDX )(2 0
0
0
2)( 0
TeDtx
tjn
n
n
The FT of a periodic function consists of a sequence of equidistant impulses located at the harmonic frequencies of the function.
3/17/2015
7
Fourier Transform of a unit impulse train
Consider an impulse train:
The Fourier series of this impulse train :
Therefore using results from the previous slide, we get:
Fourier Transform Table (1)
Fourier Transform Table (2)
Fourier Transform Table (3)
3/17/2015
8
Linearity & Conjugate Properties of FT
If and
If
If is real
then
then
then
Time-Frequency Duality (Symmetry) of Fourier Transform
Near symmetry between direct and inverse Fourier transforms:
If F is the system that produces the Fourier transform, then
The forward transform results in the reversal of the inverse transform. This is called the principle of duality.
xtXFXtxF 2)]([
Duality (Symmetry) Property of FT
Proof: From definition of inverse FT (previous slide), we get
If
then
Hence
Change t to ω yield, and use definition of forward FT, we get:
Duality Property Example
Consider the FT of a rectangular function:
By duality, Fourier transform of sinc x is Π(f).
3/17/2015
9
Scaling Property of FT
That is, compression of a signal in time results in spectral expansion (and reduction in amplitude), and vice versa
If
then for any real constant a,
The sharper the pulse the wider the spectrum
Time Shifting Property of FT
Consider a sinusoidal wave, time shifted:
If
then
Obvious that phase shift increases with frequency (t0 is constant).
Frequency-Shifting (Modulation)
Property of FT
If
Then
)(Xtx
)( 00
Xetxj
dteetxetx tjtjtj
00 )(])([F
dtetxtj
)( 0)(
)( 0 X
Proof:
Frequency-Shifting Example
Find and sketch the Fourier transform of the signal x(t) cos10t where x(t) = rect(t/4).
3/17/2015
10
Convolution Properties of FT
Let H(ω) be the Fourier transform of the unit impulse response h(t), i.e.
If
Then
and
Applying the time-convolution property to y(t) = x(t)*h(t), we get:
That is: the Fourier Transform of the system impulse response is the system Frequency Response
Proof of the Time Convolution Properties
By definition
The inner integral is Fourier transform of x2(t-τ), therefore we can use time-shift property and express this as X2(ω)e-jωτ .
Frequency Convolution Example
Find the spectrum of x(t) = cos10t where x(t) = rect(t/4). Using convolution property.
Time Differentiation Property of FT
If
Then
)(Xtx
)(Xjtxdt
d
Proof: dtetxtx
dt
dF tj
)('
dtetxjetx tjtj )()(
)(Xj Generalize case:
If )(Xtx
Then )( Xjtxnn
3/17/2015
11
Time Integration Property of FT
If
Then
and
Summary of Fourier Transform Operations (1)
Summary of Fourier Transform Operations (2)
ECE 219 : Signal, System & Control
Signal Transmission & Windowing Effects
Navneet Upadhyay Dept. of Electronics & Communication Engineering
The LNM Institute of Information Technology, Jaipur
3/17/2015
12
Signal Transmission Through LTIC Systems
H
thLTI System,
Input signal Output signal
thtxty *
XHY
tx
X
X
YH )(
= Impulse response of the system
=Frequency response of the system
H
th
Example
Find the zero-state response of a stable LTI system with transfer function
and the input is x(t) = e-t u(t).
The FT of input x(t) is:
Since the system is stable, therefore H(jω) = H(ω). Hence
Therefore
Using partial fractions, we get:
Signal Distortion during Transmission
In polar form
H
thLTI System,
Input signal Output signal
thtxty *
XHY
tx
X
X
YH )(
HjXjYj eHeXeY )(
H
During transmission, input signal spectral component of frequency ω
is modified in amplitude by a factor and shifted in phase by
an angle
H
= Amplitude response of the system
= Phase response of the system
H
H
Distortion Types In applications such as signal amplification or message signal
transmission over a communication channel, we require that the output waveform be a replica of the input waveform
In such cases we need to minimize distortion
Two Types of Distortion:
1. Amplitude response is not constant over a frequency band (interval) of interest amplitude distortion
2. Phase response is not linear over a frequency band of interest phase distortion
3/17/2015
13
Distortionless Transmission
Transmission is said to be distortionless if input signal and output signal have identical wave shapes within a multiplicative constant and a Time delay constant .
H
thLTI System,
Input signal Output signal
dttGxty
Y
tx
X
dtj
d
GeHX
Y
ttGxty
FR required for distortionless
transmission
Constant G is used
for change in amplitude
Constant td accounts
for delay in transmission
GH
i.e. Amplitude response must be constant and Phase response must be a linear function of ω with slope -td and intersect at zero.
i.e.
dtH
Same shape as input, Different
magnitude with delay
linear phase means that every spectral component is
delayed by td seconds.
Therefore, a distortionless transmission needs a flat amplitude response and a linear phase response (i.e. phase is not only a linear function of ω but also
pass through origin at ω=0):
dtH
slope -td
Phase linearity measure: The phase distortion of a
linear system can be characterized using group delay. plot the slope of as a function of ω H
Hd
dtg
If tg (ω) is constant, signal is delayed by tg (assuming constant H(ω)).
If tg (ω) is not constant constant, signal is distorted.
tg (ω) is known as Group delay or Envelope delay.
Human ears are sensitive to amplitude distortion, but not phase distortion.
Human eyes are sensitive to phase distortion, but not (so much) amplitude
distortion.
Bandpass Systems & Group Delay
If one applies an input z(t) = x(t) cosωct, then the output y(t) is:
That is, the output is the delayed version of input z(t) and the output carrier acquires an extra phase φ0.
The envelope of the signal is therefore distortionless.
Consider a bandpass system with amplitude and phase characteristics as shown:
H
Example
A signal z(t) shown below is given by z(t) = x(t)cosωct where ωc=2000π.
The pulse x(t) is a lowpass pulse of duration 0.1sec and has a bandwidth of
about 10Hz. This signal is passed through a filter whose frequency response
is shown below. Find and sketch the filter output y(t).
Spectrum Z(ω) is a narrow band of bandwidth of 20Hz centered around 1kHz (=f0).
The gain at 1kHz (=f0) is 2. The group delay is:
The vertical axis intercept of phase response is
3/17/2015
14
Parseval’s Theorem: Energy Preserving
The energy of a signal x(t) in time or frequency domain:
Proof:
Energy in time domain = Energy in Frequency domain
Parseval’s Theorem
Band Energy:
B
B
w
w
x dXE
2
2
1
Energy Spectral Density of a signal
The energy over a small frequency band Δω (Δω→0) is:
Total energy is area under the curve of vs ω (divided by 2π)
Energy Spectral Density of a REAL Signal
If x(t) is a real signal, then X(ω) and X(-ω) are conjugate
This implies that X(ω) is an even function. Therefore
Consequently, the energy contributed by a real signal by spectral components between ω1 and ω2 is (i.e. Band Energy):
3/17/2015
15
Example
Find the energy E of signal x(t) = e-at (t). Determine the frequency W (rad/s) so that the energy contributed by the spectral component from 0 to W is 95% of the total signal energy E.
Take FT of x(t):
By Parseval’s theorem:
Energy in band 0 to W is 95% of this, therefore:
Note: For this signal, 95% of energy is in small frequency band from 0 to 12.706a rad/s or (2.02a Hz). All remaining bands from 12.706a rad/s to ∞ contribute only 5% of energy.
Bandlimited Signals
A signal x(t) is said to be band-limited if its Fourier transform
X(ω) is zero for all |ω|>2πB, where B is some positive number, called the bandwidth of the signal.
It turns out that any band-limited signal must have an infinite duration in time, i.e., band-limited signals cannot be time limited
If a signal x(t) is not band-limited, it is said to have infinite bandwidth or an infinite spectrum.
Time-limited signals cannot be band-limited and thus all time-limited signals have infinite bandwidth