Upload
others
View
5
Download
0
Embed Size (px)
Citation preview
1
Continuous-Time Fourier Methods
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl
1
Representing a Signal • The convolution method for finding the
response of a system to an excitation takes advantage of the linearity and time-invariance of the system and represents the excitation as a linear combination of impulses and the response as a linear combination of impulse responses
• The Fourier series represents a signal as a linear combination of complex sinusoids
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 2
Linearity and Superposition If an excitation can be expressed as a sum of complex sinusoids the response of an LTI system can be expressed as the sum of responses to complex sinusoids.
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 3
Real and Complex Sinusoids
�
cos x( ) =e jx + e− jx
2
�
sin x( ) =e jx − e− jx
j2
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 4
Jean Baptiste Joseph Fourier
3/21/1768 - 5/16/1830
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 5
Conceptual Overview The Fourier series represents a signal as a sum of sinusoids. The best approximation to the dashed-line signal below using only a constant is the solid line. (A constant is a cosine of zero frequency.)
t-4 10
Constant
-0.6
0.6
t-4 10
x(t)1.6 Exact x(t)
1Approximation of x(t) by a constant
t0 t0 +T
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 6
2
Conceptual Overview The best approximation to the dashed-line signal using a constant plus one real sinusoid of the same fundamental frequency as the dashed-line signal is the solid line.
t-4 10
Sinusoid 1
-0.6
0.6
t-4 10
x(t)
1.6 Exact x(t)
1Approximation of x(t) through 1 sinusoid
t0 t0 +T
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 7
Conceptual Overview The best approximation to the dashed-line signal using a constant plus one sinusoid of the same fundamental frequency as the dashed-line signal plus another sinusoid of twice the fundamental frequency of the dashed-line signal is the solid line.
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 8
Conceptual Overview The best approximation to the dashed-line signal using a constant plus three sinusoids is the solid line. In this case (but not in general), the third sinusoid has zero amplitude. This means that no sinusoid of three times the fundamental frequency improves the approximation.
t-4 10
Sinusoid 3
-0.6
0.6
t-4 10
x(t) Exact x(t)
1
Approximation of x(t) through 3 sinusoids
t0 t0 +T
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 9
Conceptual Overview The best approximation to the dashed-line signal using a constant plus four sinusoids is the solid line. This is a good approximation which gets better with the addition of more sinusoids at higher integer multiples of the fundamental frequency.
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 10
Continuous-Time Fourier Series Definition
The Fourier series representation of a signal x(t)over a time t0 < t < t0 +T is
x t( ) = cx k[ ]e j2πkt /Tk=−∞
∞
∑where cx[k] is the harmonic function and k is the harmonicnumber. The harmonic function can be found from the signalusing the princple of orthogonality.
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 11
Orthogonality
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 12
3
Using Euler's identity
e j2πkt /T ,e j2πqt /T( ) = cos 2π k − qT
t⎛⎝⎜
⎞⎠⎟ + j sin 2π k − q
Tt⎛
⎝⎜⎞⎠⎟
⎡⎣⎢
⎤⎦⎥dt
t0
t0 +T
∫If k = q,
e j2πkt /T ,e j2πqt /T( ) = cos 0( ) + j sin 0( )⎡⎣ ⎤⎦dtt0
t0 +T
∫ = dtt0
t0 +T
∫ = T .
If k ≠ q, the integral
e j2πkt /T ,e j2πqt /T( ) = cos 2π k − qT
t⎛⎝⎜
⎞⎠⎟ + j sin 2π k − q
Tt⎛
⎝⎜⎞⎠⎟
⎡⎣⎢
⎤⎦⎥dt
t0
t0 +T
∫is over a non-zero integer number of cycles of a cosine and a sine and is therefore zero.
Orthogonality
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 13
Therefore e j2πkt /T and e j2πqt /T are orthogonal if k and q are not equal.
Now multiply the Fourier series expression x t( ) = cx k[ ]e j2πkt /Tk=−∞
∞
∑by e− j2πqt /T (q an integer)
x t( )e− j2πqt /T = cx k[ ]e j2π k−q( )t /T
k=−∞
∞
∑and integrate both sides over the interval t0 ≤ t < t0 +T
x t( )e− j2πqt /T dtt0
t0 +T
∫ = cx k[ ]e j2π k−q( )t /T
k=−∞
∞
∑⎡⎣⎢
⎤⎦⎥dt
t0
t0 +T
∫ .
Orthogonality
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 14
Orthogonality
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 15
Summarizing
x t( ) = cx k[ ]e j2πkt /Tk=−∞
∞
∑ and cx k[ ] = 1T
x t( )e− j2πkt /T dtt0
t0 +T
∫ .
The signal and its harmonic function form a Fourier seriespair x t( ) F S
T← →⎯⎯ cx k[ ] where T is the representation time and, therefore, the fundamental period of the CTFS representation of x t( ). If T is also period of x t( ), the CTFS representation of x t( ) is valid for all time. This is, by far, the most common use of the CTFS in engineering applications. If T is not a period of x t( ), the CTFS representation is generally valid only in the interval t0 ≤ t < t0 +T .
Continuous-Time Fourier Series Definition
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 16
CTFS of a Real Function
It can be shown that the continuous-time Fourier series (CTFS) harmonic function of any real-valued function x t( ) has the property that cx k[ ] = cx
* −k[ ].
One implication of this fact is that, for real-valued functions,the magnitude of the harmonic function is an even function and the phase is an odd function.
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 17
The Trigonometric CTFS The fact that, for a real-valued function x t( ) cx k[ ] = cx
* −k[ ]also leads to the definition of an alternate form of the CTFS, the so-called trigonometric form.
x t( ) = ax 0[ ]+ ax k[ ]cos 2πkt /T( ) + bx k[ ]sin 2πkt /T( ){ }k=1
∞
∑where
ax k[ ] = 2T
x t( )cos 2πkt /T( )dtt0
t0 +T
∫
bx k[ ] = 2T
x t( )sin 2πkt /T( )dtt0
t0 +T
∫
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 18
4
The Trigonometric CTFS
Since both the complex and trigonometric forms of the CTFS represent a signal, there must be relationships between the harmonic functions. Those relationships are
ax 0[ ] = cx 0[ ]bx 0[ ] = 0
ax k[ ] = cx k[ ]+ cx* k[ ]
bx k[ ] = j cx k[ ]− cx* k[ ]( )
⎧
⎨
⎪⎪
⎩
⎪⎪
⎫
⎬
⎪⎪
⎭
⎪⎪
, k = 1,2,3,…
cx 0[ ] = ax 0[ ]cx k[ ] = ax k[ ]− j bx k[ ]
2
cx −k[ ] = cx* k[ ] = ax k[ ]+ j bx k[ ]
2
⎧
⎨
⎪⎪⎪
⎩
⎪⎪⎪
⎫
⎬
⎪⎪⎪
⎭
⎪⎪⎪
, k = 1,2,3,…
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 19
CTFS Example #1
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 20
CTFS Example #1
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 21
CTFS Example #2 Let a signal be defined by x t( ) = 2cos 400πt( ) and letT = 10 ms which is 2T0 .
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 22
CTFS Example #2
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 23
CTFS Example #3 Let x t( ) = 1 / 2 − 3 / 4( )cos 20πt( ) + 1 / 2( )sin 30πt( ) and let T = 200 ms.
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 24
5
CTFS Example #3
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 25
CTFS Example #3
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 26
CTFS Example #3
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 27
Linearity of the CTFS
These relations hold only if the harmonic functions of all the component functions are based on the same representation time T.
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 28
CTFS Example #4 Let the signal be a 50% duty-cycle square wave with an amplitude of one and a fundamental period T0 = 1. x t( ) = rect 2t( )∗δ1 t( )
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 29
CTFS Example #4
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 30
6
CTFS Example #4
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 31
CTFS Example #4
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 32
CTFS Example #4 A graph of the magnitude and phase of the harmonic function as a function of harmonic number is a good way of illustrating it.
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 33
The Sinc Function
Let x t( ) = A rect t /w( )∗δT0t( ) , w < T0 . Then
x t( ) = A rect t /w( )∗δT0t( ) F S
T0← →⎯⎯ cx k[ ] = A sin πkw /T0( )
πk
The mathematical form sin π x( )π x
arises frequently enough
to be given its own name "sinc". That is sinc t( ) = sin πt( )πt
.
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 34
CTFS Example #5 Let x t( ) = 2cos 400πt( ) and let T = 7.5 ms which is1.5 fundamental periods of this signal.
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 35
CTFS Example #5
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 36
7
CTFS Example #5
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 37
CTFS Example #5 The CTFS representation of this cosine is the signal below, which is an odd function, and the discontinuities make the representation have significant higher harmonic content. This is a very inelegant representation.
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 38
CTFS of Even and Odd Functions
For an even function, the complex CTFS harmonic function cx k[ ] is purely real and the sine harmonic function ax k[ ] is zero.
For an odd function, the complex CTFS harmonic functioncx k[ ] is purely imaginary and the cosine harmonic functionbx k[ ] is zero.
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 39
Convergence of the CTFS
For continuous signals, convergence is exact at every point.
A Continuous Signal
Partial CTFS Sums xN t( ) = cx k[ ]e j2πkt /T0
k=−N
N
∑
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 40
Convergence of the CTFS
For discontinuous signals, convergence is exact at every point of continuity.
Discontinuous Signal
Partial CTFS Sums
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 41
Convergence of the CTFS
At points of discontinuity the Fourier series representation converges to the mid-point of the discontinuity.
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 42
8
Numerical Computation of the CTFS How could we find the CTFS of a signal which has no known functional description? Numerically.
cx k[ ] = 1
Tx t( )e− j2πkt /T dt
T∫Unknown
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 43
Numerical Computation of the CTFS
cx k[ ] ≅ 1T
x nTs( )e− j2πknTs /T dtnTs
n+1( )Ts
∫⎡
⎣⎢⎢
⎤
⎦⎥⎥n=0
N−1
∑Samples from x(t)
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 44
Numerical Computation of the CTFS
It can be shown (Web Appendix F) that, for harmonic numbersk << N
cx k[ ] ≅ 1 / N( )D F T x nTs( )( ) , k << Nwhere
D F T x nTs( )( ) = x nTs( )e- j2πnk /N
n=0
N -1
∑
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 45
Numerical Computation of the CTFS
The Discrete Fourier Transform
D F T x nTs( )( ) = x nTs( )e− j2πnk /N
n=0
N−1
∑is an intrinsic function in most modern high-level computerlanguages.
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 46
CTFS Properties
Linearity
α x t( ) + β y t( ) F ST← →⎯⎯α cx k[ ]+ β cy k[ ]
Let a signal x(t) have a fundamental period T0 x and let asignal y(t) have a fundamental period T0 y . Let the CTFSharmonic functions, each using a common period T as the representation time, be cx[k] and cy[k]. Then the following properties apply.
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 47
CTFS Properties Time Shifting x t − t0( ) F S
T← →⎯ e− j2πkt0 /T cx k[ ]
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 48
9
CTFS Properties
Frequency Shifting (Harmonic Number
Shifting) ej2πk0t /T x t( ) F S
T← →⎯ cx k − k0[ ]
A shift in frequency (harmonic number) corresponds to multiplication of the time function by a complex exponential.
Time Reversal x −t( ) F ST← →⎯ cx −k[ ]
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 49
CTFS Properties
Time Scaling Let z t( ) = x at( ) , a > 0Case 1. T = T0 x / a = T0z for z t( ) cz k[ ] = cx k[ ]Case 2. T = T0 x for z t( ) If a is an integer,
cz k[ ] = cx k / a[ ] , k / a an integer0 , otherwise
⎧⎨⎩
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 50
CTFS Properties Time Scaling (continued)
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 51
CTFS Properties
Change of Representation TimeWith T = T0 x , x t( ) F S
T← →⎯ cx k[ ]With T = mT0 x , x t( ) F S
T← →⎯ cx,m k[ ]
cx,m k[ ] = cx k /m[ ] , k /m an integer0 , otherwise
⎧⎨⎩
(m is any positive integer)
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 52
CTFS Properties Change of Representation Time
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 53
CTFS Properties
Time Differentiation
ddtx t( )( ) F S
T← →⎯ j2πk cx k[ ] /T
F ST← →⎯⎯
F ST← →⎯⎯
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 54
10
CTFS Properties Time Integration
Case 1. cx 0[ ] = 0
x λ( )dλ
−∞
t
∫ F ST← →⎯
cx k[ ]j2πk /T
, k ≠ 0
Case 2. cx 0[ ] ≠ 0
x λ( )dλ−∞
t
∫ is not periodic
Case 1 Case 2
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 55
CTFS Properties
Multiplication - Convolution Duality x t( )y t( ) F S
T← →⎯ cx k[ ]∗cy k[ ](The harmonic functions cx[k] and cy[k] must be basedon the same representation time T .)
x t( )y t( ) F ST← →⎯ T cx k[ ]cy k[ ]
The symbol indicates periodic convolution.Periodic convolution is defined mathematically by
x t( )y t( ) = x τ( )y t −τ( )dτT∫
x t( )y t( ) = xap t( )∗y t( ) where xap t( ) is any single period of x t( )
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 56
CTFS Properties
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 57
CTFS Properties
Conjugation x* t( ) F S
T← →⎯ cx* −k[ ]
Parseval’s Theorem
1T
x t( ) 2 dtT∫ = cx k[ ] 2
k=−∞
∞
∑The average power of a periodic signal is the sum of theaverage powers in its harmonic components.
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 58
Some Common CTFS Pairs
1 F ST← →⎯ δ k[ ] , T arbitrary
δT0t( ) F S
mT0← →⎯⎯
1 /T0( ) , k /m an integer0 , otherwise
⎧⎨⎩
e j2πqt /T0 F SmT0
← →⎯⎯ δ k − mq[ ]sin 2πqt /T0( ) F S
mT0← →⎯⎯ j / 2( ) δ k + mq[ ]−δ k − mq[ ]( )
cos 2πqt /T0( ) F SmT0
← →⎯⎯ 1 / 2( ) δ k − mq[ ]+δ k + mq[ ]( )rect t /w( )∗δT0
t( ) F SmT0
← →⎯⎯ w /T0( )sinc wk /mT0( )δm k[ ]tri t /w( )∗δT0
t( ) F SmT0
← →⎯⎯ w /T0( )sinc2 wk /mT0( )δm k[ ]
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 59
CTFS Examples
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 60
11
CTFS Examples
x t( ) = 12sin 2πt / 0.01( ) rect t / 0.01( )∗δ0.02 t( )⎡⎣ ⎤⎦x t( ) = 12sin 200πt( ) rect 100t( )∗δ0.02 t( )⎡⎣ ⎤⎦Find the CTFS harmonic function of x t( ) with T = 20 ms.
sin 2πqt /T0( ) F SmT0
← →⎯⎯ j / 2( ) δ k + mq[ ]−δ k − mq[ ]( )sin 200πt( ) F S
2×0.01← →⎯⎯ j / 2( ) δ k + 2 ×1[ ]−δ k − 2 ×1[ ]( )rect t /w( )∗δT0
t( ) F SmT0
← →⎯⎯ w /T0( )sinc wk /mT0( )δm k[ ]rect 100t( )∗δ0.02 t( ) F S
1×0.02← →⎯⎯ 1 / 2( )sinc k / 2( )Using x t( )y t( ) F S
T← →⎯ cx k[ ]∗cy k[ ],12sin 200πt( ) rect 100t( )∗δ0.02 t( )⎡⎣ ⎤⎦
F S0.02← →⎯⎯ 12 j / 2( ) δ k + 2[ ]−δ k − 2[ ]( )∗ 1 / 2( )sinc k / 2( )
12sin 200πt( ) rect 100t( )∗δ0.02 t( )⎡⎣ ⎤⎦F S0.02← →⎯⎯ j3 sinc k + 2( ) / 2( )− sinc k − 2( ) / 2( )( )
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 61
CTFS Examples Find the CTFS harmonic function of x t( ) with T = 10−8.
cx k[ ] = 1 /T( ) x t( )e− j2πkt /T dtT∫ ⇒ cx 0[ ] = 108 35 ×108 t( )dt
0
10−8
∫ = 35 / 2
cx k[ ] = 108 35 ×108 t( )e− j2π×108 kt dt0
10−8
∫ = 35 ×1016 te− j2π×108 kt dt0
10−8
∫
cx k[ ] = 35 ×1016 t e− j2π×108 kt
− j2π ×108 k⎡
⎣⎢
⎤
⎦⎥
0
10−8
− e− j2π×108 kt
− j2π ×108 k⎛
⎝⎜⎞
⎠⎟dt
0
10−8
∫⎧⎨⎪
⎩⎪
⎫⎬⎪
⎭⎪
cx k[ ] = 35 ×1016 − e− j2πk ×10−8
j2π ×108 k− e− j2π×108 kt
j2π ×108 k( )2
⎡
⎣⎢⎢
⎤
⎦⎥⎥
0
10−8⎧
⎨⎪
⎩⎪
⎫
⎬⎪
⎭⎪
cx k[ ] = 35 ×1016 −10−16e− j2πk
j2πk+ 1− e− j2πk
j2πk( )2 ×1016
⎡
⎣⎢
⎤
⎦⎥ = 351− e− j2πk − j2πke− j2πk
j2πk( )2
cx k[ ] = 351 / 2 , k = 0j
2πk , k ≠ 0
⎧⎨⎪
⎩⎪
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 62
LTI Systems with Periodic Excitation
The differential equation describing an RC lowpass filter is RC ′vout t( ) + vout t( ) = vin t( )If the excitation vin t( ) is periodic it can be expressed as aCTFS,
vin t( ) = cin k[ ]e j2πkt /Tk=−∞
∞
∑The equation for the kth harmonic alone is RC ′vout,k t( ) + vout,k t( ) = vin,k t( ) = cin k[ ]e j2πkt /T
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 63
LTI Systems with Periodic Excitation
If the excitation is periodic, the response is also, with thesame fundamental period. Therefore the response can beexpressed as a CTFS also. vout,k t( ) = cout e
j2πkt /T
Then the equation for the kth harmonic becomesj2kπRC /T( )cout k[ ]e j2πkt /T + cout k[ ]e j2πkt /T = cin k[ ]e j2πkt /T
Notice that what was once a differential equation is nowan algebraic equation.
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 64
LTI Systems with Periodic Excitation
Solving the kth-harmonic equation,
cout k[ ] = cin k[ ]j2kπRC /T +1
Then the response can be written as
vout t( ) = cout k[ ]e j2πkt /Tk=−∞
∞
∑ =cin k[ ]
j2kπRC /T +1e j2πkt /T
k=−∞
∞
∑
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 65
LTI Systems with Periodic Excitation
The ratio cout k[ ]cin k[ ] is the
harmonic response of the system.
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 66
12
Continuous-Time Fourier Methods
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl
67
Extending the CTFS
• The CTFS is a good analysis tool for systems with periodic excitation but the CTFS cannot represent an aperiodic signal for all time
• The continuous-time Fourier transform (CTFT) can represent an aperiodic signal for all time
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 68
CTFS-to-CTFT Transition
Its CTFS harmonic function is cx k[ ] = AwT0
sinc kwT0
⎛⎝⎜
⎞⎠⎟
As the period T0 is increased, holding w constant, the duty cycle is decreased. When the period becomes infinite (and the duty cycle becomes zero) x t( ) is no longer periodic.
Consider a periodic pulse-train signal x t( ) with duty cycle w /T0
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 69
CTFS-to-CTFT Transition
�
w =T02
�
w =T010
Below are graphs of the magnitude of cx k[ ] for 50% and 10% duty cycles. As the period increases the sinc function widens and its magnitude falls. As the period approaches infinity, the CTFS harmonic function becomes an infinitely-wide sinc function with zero amplitude.
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 70
CTFS-to-CTFT Transition This infinity-and-zero problem can be solved by normalizing the CTFS harmonic function. Define a new “modified” CTFS harmonic function T0 cx k[ ] = Awsinc wkf0( ) and graph it
versus kf0 instead of versus k. f0 = 1 /T0( )
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 71
CTFS-to-CTFT Transition In the limit as the period approaches infinity, the modifiedCTFS harmonic function approaches a function of continuousfrequency f (kf0 ).
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 72
13
CTFS-to-CTFT Transition
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 73
Definition of the CTFT
Forward f form Inverse
X f( ) = F x t( )( ) = x t( )e− j2π ft dt−∞
∞
∫ x t( ) = F -1 X f( )( ) = X f( )e+ j2π ft df−∞
∞
∫
Forward ω form Inverse
X jω( ) = F x t( )( ) = x t( )e− jωt dt−∞
∞
∫ x t( ) = F -1 X jω( )( ) = 12π
X jω( )e+ jωt dω−∞
∞
∫Commonly-used notation:
x t( ) F← →⎯ X f( ) or x t( ) F← →⎯ X jω( )
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 74
Some Remarkable Implications of the Fourier Transform
The CTFT expresses a finite-amplitude, real-valued, aperiodic signal which can also, in general, be time-limited, as a summation (an integral) of an infinite continuum of weighted, infinitesimal- amplitude, complex sinusoids, each of which is unlimited in time. (Time limited means “having non-zero values only for a finite time.”)
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 75
Frequency Content
Lowpass Highpass
Bandpass
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 76
δ t( ) F← →⎯ 1
e−αt u t( ) F← →⎯ 1/ jω +α( ) , α > 0 − e−αt u −t( ) F← →⎯ 1/ jω +α( ) , α < 0
te−αt u t( ) F← →⎯ 1/ jω +α( )2, α > 0 − te−αt u −t( ) F← →⎯ 1/ jω +α( )2
, α < 0
t ne−αt u t( ) F← →⎯ n!
jω +α( )n+1 , α > 0 − t ne−αt u −t( ) F← →⎯ n!
jω +α( )n+1 , α < 0
e−αt sin ω0t( )u t( ) F← →⎯ω0
jω +α( )2+ω0
2, α > 0 − e−αt sin ω0t( )u −t( ) F← →⎯
ω0
jω +α( )2+ω0
2, α < 0
e−αt cos ω0t( )u t( ) F← →⎯ jω +α
jω +α( )2+ω0
2, α > 0 − e−αt cos ω0t( )u −t( ) F← →⎯ jω +α
jω +α( )2+ω0
2, α < 0
e−α t F← →⎯ 2αω 2 +α 2 , α > 0
Some CTFT Pairs
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 77
Convergence and the Generalized Fourier Transform
Let x t( ) = A. Then from the definition of the CTFT,
X f( ) = Ae− j2π ft dt−∞
∞
∫ = A e− j2π ft dt−∞
∞
∫
This integral does not converge so, strictly speaking, the CTFT does not exist.
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 78
14
Convergence and the Generalized Fourier Transform
But consider a similar function,
xσ t( ) = Ae−σ t , σ > 0Its CTFT integral
Xσ f( ) = Ae−σ t e− j2π ft dt−∞
∞
∫does converge.
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 79
Convergence and the Generalized Fourier Transform
Carrying out the integral, Xσ f( ) = A 2σσ 2 + 2π f( )2 .
Now let σ approach zero.
If f ≠ 0 then limσ→0
A 2σσ 2 + 2π f( )2 = 0. The area under this
function is A 2σσ 2 + 2π f( )2 df
−∞
∞
∫ which is A, independent of
the value of σ . So, in the limit as σ approaches zero, the CTFT has an area of A and is zero unless f = 0. This exactly
defines an impulse of strength A. Therefore A F← →⎯ Aδ f( ).M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 80
Convergence and the Generalized Fourier Transform
By a similar process it can be shown that
cos 2π f0t( ) F← →⎯ 12
δ f − f0( ) +δ f + f0( )⎡⎣ ⎤⎦
and
sin 2π f0t( ) F← →⎯ j2
δ f + f0( )−δ f − f0( )⎡⎣ ⎤⎦
These CTFT’s which involve impulses are called generalized Fourier transforms (probably becausethe impulse is a generalized function).
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 81
Convergence and the Generalized Fourier Transform
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 82
Negative Frequency This signal is obviously a sinusoid. How is it described mathematically?
It could be described by x t( ) = Acos 2πt /T0( ) = Acos 2π f0t( )But it could also be described by x t( ) = Acos 2π − f0( )t( )
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 83
Negative Frequency
x(t) could also be described by
x t( ) = A ej2π f0t + e− j2π f0t
2or
x t( ) = A1 cos 2π f0t( ) + A2 cos 2π − f0( )t( ) , A1 + A2 = Aand probably in a few other different-looking ways. So who isto say whether the frequency is positive or negative? For thepurposes of signal analysis, it does not matter.
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 84
15
Negative Frequency Consider an experiment in which we multiply two sinusoidalsignals x1 t( ) = cos 2π f1t( ) and x2 t( ) = cos 200πt( ) to form
x t( ) = x1 t( )x2 t( ). x t( ) can be expressed using a trigonometricidentity as
x t( ) = 1 / 2( ) cos 2π f1 −100( )t( ) + cos 2π f1 +100( )t( )⎡⎣ ⎤⎦Now imagine that we continuously change f1 from a frequency above100 to a frequency below 100. f1 −100 becomesnegative.
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 85
More CTFT Pairs
The generalization of the CTFT allows us to extend the tableof CTFT pairs to some very useful functions.
δ t( ) F← →⎯ 1 1 F← →⎯ δ f( ) sgn t( ) F← →⎯ 1/ jπ f u t( ) F← →⎯ 1/ 2( )δ f( ) +1/ j2π f
rect t( ) F← →⎯ sinc f( ) sinc t( ) F← →⎯ rect f( ) tri t( ) F← →⎯ sinc2 f( ) sinc2 t( ) F← →⎯ tri f( ) δT0
t( ) F← →⎯ f0δ f0f( ) , f0 = 1/ T0 T0δT0
t( ) F← →⎯ δ f0f( ) , T0 = 1/ f0
cos 2π f0t( ) F← →⎯ 1/ 2( ) δ f − f0( ) +δ f + f0( )⎡⎣ ⎤⎦ sin 2π f0t( ) F← →⎯ j / 2( ) δ f + f0( )−δ f − f0( )⎡⎣ ⎤⎦
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 86
CTFT Properties
If F x t( )( ) = X f( ) or X jω( ) and F y t( )( ) = Y f( ) or Y jω( )then the following properties can be proven.
α x t( ) + β y t( ) F← →⎯ α X f( ) + β Y f( ) α x t( ) + β y t( ) F← →⎯ α X jω( ) + β Y jω( )Linearity
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 87
CTFT Properties
Time Shiftingx t − t0( ) F← →⎯ X f( )e− j2π ft0x t − t0( ) F← →⎯ X jω( )e− jωt0
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 88
CTFT Properties
Frequency Shiftingx t( )e+ j2π f0t F← →⎯ X f − f0( )x t( )e+ jω0t F← →⎯ X ω −ω0( )
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 89
CTFT Properties
Time Scaling x at( ) F← →⎯ 1
aX f
a⎛⎝⎜
⎞⎠⎟
x at( ) F← →⎯ 1a
X jωa
⎛⎝⎜
⎞⎠⎟
Frequency Scaling
1a
x ta
⎛⎝⎜
⎞⎠⎟
F← →⎯ X af( )
1a
x ta
⎛⎝⎜
⎞⎠⎟
F← →⎯ X jaω( )
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 90
16
The “Uncertainty” Principle The time and frequency scaling properties indicate that if a signal is expanded in one domain it is compressed in the other domain. This is called the “uncertainty principle” of Fourier analysis.
e−π t /2( )2 F← →⎯ 2e−π 2 f( )2
e−π t2 F← →⎯ e−π f2
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 91
CTFT Properties
Transform ofa Conjugate
x* t( ) F← →⎯ X* − f( )x* t( ) F← →⎯ X* − jω( )
MultiplicationConvolution
Duality
x t( )∗y t( ) F← →⎯ X f( )Y f( )x t( )∗y t( ) F← →⎯ X jω( )Y jω( )
x t( )y t( ) F← →⎯ X f( )∗Y f( )x t( )y t( ) F← →⎯ 1 / 2π( )X jω( )∗Y jω( )
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 92
CTFT Properties
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 93
CTFT Properties
In the frequency domain, the cascade connection multiplies the frequency responses instead of convolving the impulse responses.
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 94
CTFT Properties
Time Differentiation
ddt
x t( )( ) F← →⎯ j2π f X f( )ddt
x t( )( ) F← →⎯ jω X jω( )
Modulation x t( )cos 2π f0t( ) F← →⎯ 1
2X f − f0( ) + X f + f0( )⎡⎣ ⎤⎦
x t( )cos ω0t( ) F← →⎯ 12
X j ω −ω0( )( ) + X j ω +ω0( )( )⎡⎣ ⎤⎦
Transforms ofPeriodic Signals
x t( ) = X k[ ]e− j2πkfFtk=−∞
∞
∑ F← →⎯ X f( ) = X k[ ]δ f − kf0( )k=−∞
∞
∑
x t( ) = X k[ ]e− jkωFt
k=−∞
∞
∑ F← →⎯ X jω( ) = 2π X k[ ]δ ω − kω0( )k=−∞
∞
∑
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 95
CTFT Properties
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 96
17
CTFT Properties
Parseval’sTheorem
x t( ) 2 dt
−∞
∞
∫ = X f( ) 2 df−∞
∞
∫
x t( ) 2 dt−∞
∞
∫ = 12π
X jω( ) 2 df−∞
∞
∫
Integral Definitionof an Impulse
e− j2π xy−∞
∞
∫ dy = δ x( )
Duality X t( ) F← →⎯ x − f( ) and X −t( ) F← →⎯ x f( )
X jt( ) F← →⎯ 2π x −ω( ) and X − jt( ) F← →⎯ 2π x ω( )
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 97
CTFT Properties
Total -AreaIntegral
X 0( ) = x t( )e− j2π ft dt−∞
∞
∫⎡
⎣⎢
⎤
⎦⎥f→0
= x t( )dt−∞
∞
∫
x 0( ) = X f( )e+ j2π ft df−∞
∞
∫⎡
⎣⎢
⎤
⎦⎥t→0
= X f( )df−∞
∞
∫
X 0( ) = x t( )e− jωt dt−∞
∞
∫⎡
⎣⎢
⎤
⎦⎥ω→0
= x t( )dt−∞
∞
∫
x 0( ) = 12π
X jω( )e+ jωt dω−∞
∞
∫⎡
⎣⎢
⎤
⎦⎥t→0
= 12π
X jω( )dω−∞
∞
∫
Integration x λ( )dλ
−∞
t
∫ F← →⎯X f( )j2π f
+ 12
X 0( )δ f( )
x λ( )dλ−∞
t
∫ F← →⎯X jω( )jω
+π X 0( )δ ω( )
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 98
CTFT Properties
�
x 0( ) = X f( )df−∞
∞
∫
�
X 0( ) = x t( )dt−∞
∞
∫
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 99
CTFT Properties
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 100
Numerical Computation of the CTFT It can be shown (Web Appendix G) that the DFT can be used to approximate samples from the CTFT. If the signal x t( ) is a causal energy signal and N samples are taken from it over a finite time beginning at t = 0, at a rate fs then the relationship between the CTFT of x t( ) and the DFT of the samples taken from it is X kfs / N( ) ≅ Tse− jπk /N sinc k / N( )XDFT k[ ]For those harmonic numbers k for which k << N X kfs / N( ) ≅ Ts XDFT k[ ]As the sampling rate and number of samples are increased, this approximation is improved.
M. J. Roberts - All Rights Reserved. Edited by Dr. Robert Akl 101