Escolar Documentos
Profissional Documentos
Cultura Documentos
f (t ) =
In other words, any periodic function can be resolved as a summation of constant value and cosine and sine functions:
f (t ) = a0 + (an cos nt + bn sin nt ) 2 n =1
The computation and study of Fourier series is known as harmonic analysis and is extremely useful as a way to break up an arbitrary periodic function into a set of simple terms that can be plugged in, solved individually, and then recombined to obtain the solution to the original problem or an approximation to it to whatever accuracy is desired or practical.
f(t)
Periodic Function
a0 2
f (t ) =
where
=
t
a1 cos t
b1 sin t
+
a2 cos 2t
+
T
a0 =
b2 sin 2t
2 T
f (t )dt
0
an =
2 f (t ) cos ntdt T 0
bn =
2 f (t ) sin ntdt T 0
T /2
T / 2
10/24/12
Example 1
Determine the Fourier series representation of the following waveform.
Solution
First, determine the period & describe the one period of the function:
T=2
f (t + 2) = f (t )
a0 =
an =
2 f (t ) cos ntdt T 0
1 2 1
Or, since f (t )dt is the total area below graph a y = f(t) over the interval [a,b], hence
sin n ,= 0
a0 =
an = .0
bn =
2 T
1
f (t ) sin ntdt
0 2 1
Finally,
f (t ) = = =
cos n = (1) n
1 (1) n 2 / n = n 0 , n odd , n even
Therefore, bn =
10/24/12
[Supplementary]
The sum of the Fourier series terms can evolve (progress) into the original waveform
For n integers,
sin n = 0 sin 2n = 0
cos n = (1) n
cos 2n = 1
It can be demonstrated that the sum will lead to the square wave:
(a)
2
(b)
2 2 sin 3t 3
(e)
sin t
sin t +
sin t +
(c)
(d)
(f)
sin t +
2 2 sin 3t + sin 5t 3 5
sin t +
Example 2
Given
f (t ) = t ,
1 t 1
f (t + 2) = f (t )
Sketch the graph of f (t) such that
3 t 3.
10/24/12
Solution
The function is described by the following graph:
a0 = =
2 T
f (t )dt
1
1 t 2 2 11 tdt = = =0 2 1 2 2 1
T=2
We find that
2 = = T
an =
bn =
2 T
cos( x) = cos x
Example 3
Finally,
f (t ) =
Given
2 t v(t ) = 0
v(t + 4) = v(t )
0 t 12.
10/24/12
Solution
The function is described by the following graph:
v (t) 2 0 2 T=4 4 6 8 10 12 t
a0 = = =
2 v(t ) dt T 0
2 4 2 ( 2 t ) dt + 0dt 4 0 2 2 1 1 t 2 ( 2 t ) dt = 2t = 1 20 2 2 0 2
We find that
2 = T 2
an = =
bn = = = =
1 cos nt = 0 + 2 2 2 n 0 =
1 sin 2n 1 2 = = n 2n 2 2 n n
sin 2n = sin n = 0
since
Symmetry Considerations
Finally,
v(t ) = =
10/24/12
Even symmetry
Any function f (t) is even if its plot is symmetrical about the vertical axis, i.e.
f (t ) = f (t )
f (t ) = cos t
Odd symmetry
Any function f (t) is odd if its plot is antisymmetrical about the vertical axis, i.e.
f (t ) = f (t )
t
A +A
f (t ) = t
f (t ) = sin t
+A
odd
(t ) dt = 0
10/24/12
Symmetry consideration
From the properties of even and odd functions, we can show that: for even periodic function;
an = 4 T
T /2
f (t ) cos ntdt
0
bn = 0
4 T
T /2
f (t ) sin ntdt
0
T /2 T /2
T 2
T 2
T /2
T 2
T 2
T /2
an =
2 4 / 2f (t ) cos ntdt = T T T
(even) (even)
||
f (t ) cos ntdt bn =
2 / 2f (t ) sin ntdt = 0 T T
(even) (odd)
||
a0 =
2 / 2f (t )dt = 0 T T
(odd)
T /2
bn =
2 4 / 2f (t ) sin ntdt = T T T
(odd) (odd)
||
T /2
f (t ) sin ntdt
0
an =
(even)
(odd)
2 / 2f (t ) cos ntdt = 0 T T
(odd) (even)
||
T /2
(even)
(odd)
Example 4
Given
1 , 2 < t < 1 f (t ) = t , 1 < t < 1 1 , 1< t < 2 f (t + 4) = f (t )
Solution
The function is described by the following graph:
f (t) 1
6 t 6.
T=4
2 = T 2
10/24/12
bn = =
2 T
2 a0 = T
and
f (t )dt = 0
cos nt t cos nt cos nt = + dt + n n n 0 0 1 cos n sin nt cos 2n cos n + 2 2 n n n 0 cos 2n sin n 2 cos n = + 2 2 = n n n since sin 2n = sin n = 0 =
1
an =
2 T
f (t ) cos ntdt = 0
Example 5
Finally,
a f (t ) = 0 + (an cos nt + bn sin nt ) 2 n =1 nt 2 cos n = sin n 2 n =1 n +1 (1) nt = 2 sin n 2 n =1
Solution
The function is described by
1 , 0 < t < 1 f (t ) = 2 , 1 < t < 2 1 , 2 < t < 3
T=3
f (t + 3) = f (t )
2 4 f (t )dt = T T 0
3/ 2
3/ 2
and
2 2 = = T 3
Or, simply
a0 =
3 2 2 Total area below graph 2 8 f (t )dt = = 3 4 = 3 in a period T 0 T
10/24/12
an = =
2 T
0 1
f (t ) cos ntdt =
3/ 2
4 T
3/ 2
f (t ) cos ntdt
0
and bn = 0 Finally,
f (t ) = =
4 1 cos ntdt + 3 0
1
2 cos ntdt
1 3/ 2
4 sin nt 4 2 sin nt = + 3 n 0 3 n 1 = = = 4 3n
4 3n sin n 2 sin 3n 2 2 n
2 3
Therefore, any non-periodic function must be extended to a periodic function first, before computing its Fourier series representation Normally, we prefer symmetry (even or odd) periodic extension instead of normal periodic extension, since symmetry function will provide zero coefficient of either an or bn This can provide a simpler Fourier series expansion
y=
t2
y=
Non-periodic function y (t )
Periodic extension
f (t )
f even (t )
t
3l 2l l 0
T
2l
3l
The Fourier series of the even or odd periodic extension of a non-periodic function is called as the half-range Fourier series This is due to the non-periodic function is considered as the half-range before it is extended as an even or an odd function
f odd (t )
t
3l 2l l
0
T
2l
3l
10/24/12
If the function is extended as an even function, then the coefficient bn= 0, hence
f (t ) = a0 + an cos nt 2 n =1
If the function is extended as an odd function, then the coefficient an= 0, hence
f (t ) = bn sin nt
n =1
which only contains the cosine harmonics. Therefore, this approach is called as the halfrange Fourier cosine series
which only contains the sine harmonics. Therefore, this approach is called as the halfrange Fourier sine series
Example 6
Compute the half-range Fourier sine series expansion of f (t), where
Solution
Since we want to seek the half-range sine series, the function to is extended to be an odd function:
f (t) 1 0 t 2 f (t) 1 0 1 T = 2 2 t
f (t ) = 1 , 0 < t <
2 =1 T
Example 7
Determine the half-range cosine series expansion of the function
a0 = an = 0
and
bn =
=
4 T
T /2
f (t ) = 2t 1 , 0 < t < 1
4 / n 2 cos nt 2 = (1 cos n ) = n 0 n 0
n odd
, n even
Sketch the graphs of both f (t) and the periodic function represented by the series expansion for 3 < t < 3.
Therefore,
f (t ) =
2 (1 cos n ) sin nt = n =1 n
n =1 n odd
n sin nt
10
10/24/12
Solution
Since we want to seek the half-range cosine series, the function to is extended to be an even function:
f (t)
1 1
1 2
f (t)
1
1
an =
t
4 T
T /2
f (t ) cos ntdt =
1
2
1 1
1
T=2
2 = T
4(cos n 1) 8 / n 2 2 = n 2 2 0
n odd
, n even
bn = 0
Therefore,
Parseval s Theorem
f (t ) = a0 + an cos nt
n =1
= 0+
n =1 n odd
n
2
8 8 cos nt = 2 2
n =1 n odd
1
2
Parserval s theorem states that the average power in a periodic signal is equal to the sum of the average power in its DC component and the average powers in its harmonics
cos nt
f(t)
Pavg
Pdc
a0 2
=
a1 cos t
Pa1
b1 sin t
Pb1
+
a2 cos 2t
+
Pa2
b2 2t b2 sin
11
10/24/12
Pavg =
1 2 1 2 2 a0 + (an + bn ) 4 2 n =1
yields
cos x = e jx + e jx 2
and
sin x =
e jx e jx j2
e jnt + e jnt e jnt e jnt a0 + bn + an 2 n =1 2 j2 e jnt + e jnt e jnt e jnt a0 = + an jbn 2 n =1 2 2 an jbn jnt an + jbn jnt a0 = + + e e 2 n =1 2 2
a jb
a + jb
c0
n =1
cn
n =1
cn
n = 1
c e
n n =1
jnt
n =
c e
n
jnt
+ c0 + cn e jnt =
n =
c e
n
jnt
f (t ) sin ntdt
0
In fact, in many cases, the complex Fourier series is easier to obtain rather than the trigonometrical Fourier series In summary, the relationship between the complex and trigonometrical Fourier series are:
c0 = a0 1 = 2 T
T
= =
1 T 1 T
f (t )dt
0
cn =
1 T
f (t )e
0
jnt
dt
cn = c n
f (t )e
0
jnt
dt
an jbn 2 an + jbn = 2
or
cn = cn
12
10/24/12
Example 8
Obtain the complex Fourier series of the following function f (t )
e 2
t
Solution
Since T = 2 , . Hence =1
c0 = =
1 T
f (t )dt
0 2
1
4 2 0 2 4
1 2
f (t )=
e dt
0
1 t e 2
[ ]
e 2 1 2
cn = =
1 T
f (t )e
0 2
jnt
dt 1 2
2
cn
n =0
1 2
e 2 1 e 2 1 = = c0 2 (1 jn) n =0 2
e e
0
jnt
dt =
2
e
0
(1 jn ) t
dt
1 = 2 =
e 1 jn 0
(1 jn ) t
f (t ) =
n =
cne jnt =
e 2 1 jnt e n = 2 (1 jn)
*Notes: Even though c0 can be found by substituting cn with n = 0, sometimes it doesn t works (as shown in the next example). Therefore, it is always better to calculate c0 alone.
Example 9
Obtain the complex Fourier series of the function in Example 1.
cn =
e 2 1 2 1 + n 2
In other words, we have transformed the function f (t) in the time domain (t), to the function cn in the frequency domain (n).
13
10/24/12
Solution
= T 1 1 1 1 c0 = f (t )dt = 1dt = T 0 20 2
1 cn = T
T
But Thus,
j (e jn 1) 2n
j / n j [(1) n 1] = 2n 0 , n odd , n even
f (t )e
0
jnt
1 dt = 1e jnt dt + 0 20 1
1 e jnt j = (e jn 1) = 2 jn 0 2n
cn
n =0
.c0
n jnt
f (t ) =
n =
c e
1 j jnt e 2 n = n n0 n odd
c0 =
1 2
1 , n odd cn = n n even 0,
0.5
Introduction
Fourier transform is another method to transform a signal from time domain to frequency domain The basic idea of Fourier transform comes from the complex Fourier series Practically, many signals are non-periodic Fourier transform use the principal of the Fourier series, with assumption that the period of the nonperiodic signal is infinity (T ).
c e
n
jn0t
(2.1)
dt
where
cn = 1 T
T /2
T / 2
f (t )e
jn0t
(2.2)
n =
T f (t )e
T / 2
T /2 jn0t
dt e jn0t
(2.3)
14
10/24/12
n =
T f (t )e
T / 2
T /2 jn0t
dt e jn0t dt e jn0t
n =
2 f (t )e
T / 2
T /2 jn0t
= (n + 1)0 n0 = 0 =
2 T
or
T=
T / 2 jn t 1 jn t = f (t )e 0 dt e 0 2 n = T / 2
Now if we let T , the summation becomes integration, becomes , and the discrete harmonic frequency n0 becomes a continuous frequency , i.e.:
n =
Thus, as T ,
f (t ) = 1 2
n =
f (t )e
T / 2
T / 2
jn0t
dt e jn0t
n 0
jt 1 j t f (t )e dt e d 2
F ( )
Therefore
f (t ) = 1 2
Or in mathematical expression,
F ( )e
jt
j t
F ( ) = F { f (t )} =
f (t )e
j t
d
jt
where
F ( ) =
f (t )e
and
f (t ) = F 1{F ( )} =
1 2
F ( )e
We say that F() is the Fourier transform of f (t), and f (t) is the inverse Fourier transform of F()
where F and F 1 is the Fourier transform and inverse Fourier transform operator respectively:
f (t )
F
F 1
F ( )
15
10/24/12
Generally, the Fourier transform F() exists when the Fourier integral converges A condition for a function f(t) to have a Fourier transform is, f(t) can be completely integrable, i.e.
f (t ) dt <
Principle of duality
The definition of Fourier transform is
F ( ) = F { f (t )} =
From (2.5),
f (t ) = 1 2
F ( )e
j t
f (t )e
1 2
j t
(2.4)
If we interchange t and
f ( ) = 1 2
F (t )e
jt
j t
dt
F ( )e
jt
(2.5)
Note that the function f (t) and its transform F () can be derived from each other
F (t )e
dt =
1 F {F (t )} 2
or
F {F (t )} = 2f ( )
This is an important property to find the Fourier transform of certain functions which their Fourier integral diverges For example, using integration, the Fourier transform of 1 is
F {1} = 1e
jt
*Note that
= cos j sin = ??
where cos and sin do not converge and | cos | 1 and |sin | 1. But
e jt e j e + j dt = =? = j j
At first, we may conclude that 1 has no Fourier transform, but in fact, it can be found using the principle of duality!
16
10/24/12
Example 1
Using the definition, find the Fourier transform of (t). Then deduce the Fourier transform of 1.
Solution
(t a) f (t )dt =
f (a)
F ( ) = F { (t )} = (t )e jt dt = e j 0 = e 0 = 1
F {1} = 2f ( ) = 2 ( ) = 2 ( )
since ( ) = ( .)
Example 2
Find the Fourier transform of the following function: (a)
(t a )
(a)
Solution
Using the definition,
F { (t a)} = (t a)e jt dt = e ja
(b)
e j t
(c)
cos t
(b)
From the result in (a), by replacing = a, and using the duality principle,
F {e jt } = 2f ( ) = 2 (( )) = 2 ( )
(c)
Example 3
Using the definition, find the Fourier transform of the following: (a)
e t u (t )
e jt + e jt F {cos t} = F 2 1 = F {e jt } + F {e jt } 2 1 = [2 ( ) + 2 ( + )] 2 = [ ( ) + ( + )]
(b)
e |t |
17
10/24/12
Solution
(a)
(b)
F {e |t | } = =
e e
t jt
dt + e t e jt dt
0
F {e
( j ) t
u (t )} = e
0
j t
dt = e
0
( + j ) t
dt
dt + e ( + j ) t dt
0 0
e 0 1 1 = = = ( + j ) 0 ( + j ) + j
( + j ) t
e ( j ) t e ( + j ) t = + j ( + j ) 0 1 1 2 = + = j + j 2 + 2
If F() = F{f (t)}, G() = F{g (t)}, a and b are constants, then: 1. Linearity F {af (t ) bg (t )} = aF ( ) bG ( ) 2. Time scaling
F { f (at )} =
1 F | a | a
3. Time shifting
5. Time differentiation
F { f (t t0 )} = e jt0 F ( )
that is, a delay in the time domain corresponds to a phase shift in the frequency domain. *Phase shift: Say
d n f (t ) n F = ( j ) F ( ) n dt
6. Frequency differentiation
F ( ) =| F | . Then
F ( ) = (1 ) (| F | ) =| F | ( . )
7. Duality
F {t n f (t )} = j n
d n F ( ) d n
4. Frequency shifting
F { f (t )e
j 0 t
} = F ( 0 )
F {F (t )} = 2f ( )
18
10/24/12
8. Convolution
Example 4
Derive the Fourier transform of a single rectangular pulse of width and height A, as shown below:
F { f (t ) g (t )} = F ( )G( )
F { f (t ) g (t )} = 1 F ( ) G ( ) 2
Solution
Using the definition of the Fourier transform,
F ( ) = F { f (t )} =
/2
Example 5
From the result in Example 4, using the time shifting theorem, obtain the Fourier transform of the function g(t):
f (t )e
j t
dt
/2
= =
e jt Ae jt dt = A j / 2 / 2
Solution
Set A = 1 and = 1 of the function f(t) to get the following waveform:
F ( ) = sinc
where
1 g1 (t ) = f (t + 1 2 ) and g 2 (t ) = f (t . 2 ) Hence,
g (t ) = g1 (t ) + g 2 (t )
1 = f (t + 1 2 ) f (t 2 )
Then write
g (t ) = g1 (t ) + g2 (t ) as shown below:
= +
= e j / 2 F ( ) e j / 2 F ( ) = (e j / 2 e j / 2 ) F ( ) = j 2 sin
sinc
j 4 sin 2 ( / 2)
19
10/24/12
Example 6
Using time differentiation, find the Fourier transform of the following function
and
F { f (t t0 )} = e jt0 F ( )
Solution
f=t+1 f=1t
The idea is we differentiate f(t) until its derivative is expressed in the terms of delta functions. This required twice differentiation as shown below:
df dt
d2 f dt 2
Hence,
d 2 f (t ) = (t + 1) 2 (t ) + (t 1) dt 2
d 2 f (t ) F = F { (t + 1) 2 (t ) + (t 1)} 2 dt ( j ) 2 F ( ) = e j 2 + e j F ( ) = (e
2 j
Example 7
Given the Fourier transform of f(t) is
F ( ) = 3 5 + j
+e
)2
Thus,
F ( ) =
2(1 cos )
(b) (d)
f (2t 1)
(c) (cos t ) f (t )
e t f (t )
20
10/24/12
Solution
(a) Using time scaling property,
t 1 F f = 1 F 1 2 2 2 3 6 = 2 F (2 ) = 2 5 + j (2 ) = 5 + j 2
(b)
* Note that the time shift is , not 1. The time shift must be determined when the coefficient of t is 1.
(c)
(d)
e jt + e jt (cos t ) f (t ) = 2
The frequency shifting property cannot be apply directly, since there are no term e jat. However, note that j 2 = 1, that is
et f (t ) = e j t f (t ) = e jjt f (t )
2
[ {
}]
3 3 = 5 + j ( j ) 6 + j
Convolution :
f (t ) g (t ) =
f ( ) g (t )d = g ( ) f (t )d
21
10/24/12
The steps of convolution: 1. Create a dummy variable (i.e. ) and make each waveform a function of . f (t ) f ( ) g (t ) g ( ) 2. Time-invert one of the waveforms (doesn t matter which). g ( ) g ( ) 3. Shift the inverted waveform by t. this allow the function to travel along -axis.
g ( ) g (( t )) = g (t )
5. For a given time t, integrate the product for the intersection range a < < b.
b
f ( ) g (t )d
a
6. Combine all the integrals after the travelling waveform moved for < t < .
Example 8
Find the convolution of the two signals:
f()
Solution
1. Write f and g as functions of .
g()
g( ) =
2. Choose one of the functions as the travelling wave. Suppose that we choose f. Reflect f() about the vertical axis to get f().
f()
g( ) =
g()
3. Write f() as f(t) . Since t varies from to +, it means we bring f to and shift it to + direction.
f(t) g()
t1
4&5. Along the journey of f(t ), observe the intersections between f and g. Then integrate along the intersection ranges:
t 2 t2 f g = (1)( )d = = 2 0 2 0
22
10/24/12
t<0
0
1
2 t f g = (1)( )d = = (2 t ) 2 t 1 2 t 1
t2
1
1 2
t (2 t )
t
Example 9
Find x(t ) h(t.) Given
1 , 1 < t < 0 x(t ) = 1 , 0 < t < 1 0 , otherwise
Solution
Sketch the graph x and h :
x(t) h(t) 1
and
1 t
0 1
Suppose that we choose h to be the travelling wave. After express x and h as functions of , then flip h and bring it to :
x() h(t ) 1
t1
t+1
0 1
t1
1 t + 1 1
t +1
xh =
(1)(1)d = (t + 1) 1 = t 2
23
10/24/12
x() h(t ) 1
t1 1
0 1
t+1 1
t1 0 1
1 t+1
t +1
x h = (1)(1)d + (1)(1)d = (0 1) + (t + 1 0) = t
1 0
xh =
t 1
(1)(1)d + (1)(1)d = (0 + t 1) + (1 0) = t
0
x() h(t ) 1
0 1
1
t1 1
t+1
xh =
t 1
(1)(1)d = 1 (t 1) = t + 2
2 1 0
x *h (t) 1 t
1 1
144
24
10/24/12
Introduction
Jean Baptiste Joseph Fourier (*1768-1830) French Mathematician La Thorie Analitique de la Chaleur (1822)
Fourier Series
Any periodic function can be expressed as a sum of sines and/or cosines
Fourier Series
145
146
Fourier Transform
Even functions that
are not periodic have a finite area under curve
Contents
Complex numbers etc. Impulses Fourier Transform (+examples) Convolution theorem Fourier Transform of sampled functions Sampling theorem Aliasing Discrete Fourier Transform Application Examples
can be expressed as an integral of sines and cosines multiplied by a weighing function Both the Fourier Series and the Fourier Transform have an inverse operation: Original Domain Fourier Domain
147
148
Complex numbers
Complex number
149
150
25
10/24/12
Euler s formula
Im
Sin ()
?
Re
Cos () ?
151
152
Complex math
Complex (vector) addition
Contents
Complex number etc. Impulses Fourier Transform (+examples) Convolution theorem Fourier Transform of sampled functions Sampling theorem Aliasing Discrete Fourier Transform Application Examples
Multiplication with i
is rotation by 90 degrees
153
154
26
10/24/12
Impulse train
Contents
Complex number etc. Impulses Fourier Transform (+examples) Convolution theorem Fourier Transform of sampled functions Sampling theorem Aliasing Discrete Fourier Transform Application Examples
T = 1
157
158
Fourier Series
Periodic with period T Series of sines and cosines, see Euler s formula
Fourier Transform
Continuous Fourier Transform (1D)
with
159
160
Symmetry: The only difference between the Fourier Transform and its inverse is the sign of the exponential.
161
162
27
10/24/12
Examples Block 1
If f(t) is real, then F() is complex F() is expansion of f(t) multiplied by sinusoidal terms t is integrated over, disappears F() is a function of only , which determines the frequency of sinusoidals Fourier transform frequency domain
163
-W/2
W/2
164
Examples Block 2
Examples Block 3
?
165 166
Examples Impulse
Euler
constant
167 168
28
10/24/12
Real part
Imaginary part
169
170
171
172
Contents
Complex number etc. Impulses Fourier Transform (+examples) Convolution theorem Fourier Transform of sampled functions Sampling theorem Aliasing Discrete Fourier Transform Application Examples
173
174
29
10/24/12
Fourier + Convolution
What is the Fourier domain equivalent of convolution?
What is
175
176
Intermezzo 1
What is Let , so ?
Intermezzo 2
177
178
Recapitulation 1
Convolution in one domain is multiplication in the other domain
179
180
30
10/24/12
Recapitulation 2
Shift in one domain is multiplication with complex exponential in the other domain
Contents
Complex number etc. Impulses Fourier Transform (+examples) Convolution theorem Fourier Transform of sampled functions Sampling theorem Aliasing Discrete Fourier Transform Application Examples
181
182
Sampling
Sampled function can be written as
Sampling - 2
183
184
Sampling - 3
Sampling - 4
185
186
31
10/24/12
FT of sampled functions
Fourier transform of sampled function
FT of sampled functions
Contents
Complex number etc. Impulses Fourier Transform (+examples) Convolution theorem Fourier Transform of sampled functions Sampling theorem Aliasing Discrete Fourier Transform Application Examples
Sifting property
189
190
Sampling theorem
Band-limited function
Sampling theorem 2
Sampling theorem:
If copy of can be isolated from the periodic sequence of copies contained in , can be completely recovered from the sampled version.
Since is a continuous, periodic function with period 1/T, one complete period from is enough to reconstruct the signal. This can be done via the Inverse FT.
192
32
10/24/12
Sampling theorem 3
Extracting a single period from equal to is possible if
Nyquist frequency
Contents
that is Complex number etc. Impulses Fourier Transform (+examples) Convolution theorem Fourier Transform of sampled functions Sampling theorem Aliasing Discrete Fourier Transform Application Examples
193
194
Aliasing
If , aliasing can occur
Contents
Complex number etc. Impulses Fourier Transform (+examples) Convolution theorem Fourier Transform of sampled functions Sampling theorem Aliasing Discrete Fourier Transform Application Examples
195
196
197
198
33
10/24/12
We need only one period to characterise If we want to take M equally spaced samples from in the period = 0 to = 1/, this can be done thus
200
Contents
Substituting Complex number etc. Impulses Fourier Transform (+examples) Convolution theorem Fourier Transform of sampled functions Sampling theorem Aliasing Discrete Fourier Transform Application Examples
Into yields
201
202
Examples
Cardiac tagging analysis Dither removal
203
204
34
10/24/12
206
Euler s formula
Recall
207
208
Recall
Cos(t)
1/2
1/2i
Sin(t)
209
210
35
10/24/12
Contents
Fourier Transform of sine and cosine 2D Fourier Transform Properties of the Discrete Fourier Transform
Inverse
213
214
Examples
215
216
36
10/24/12
Fourier Series
Periodic with period T Series of sines and cosines, see Euler s formula
Examples
with
217
218
Periodicity
2D Fourier Transform is periodic in both directions
Periodicity
2D Inverse Fourier Transform is periodic in both directions
219
220
Fourier Domain
221
222
37
10/24/12
Contents
Fourier Transform of sine and cosine 2D Fourier Transform Properties of the Discrete Fourier Transform
223
224
Real Real
Even Real
38
10/24/12
Properties
Odd function
229
230
Properties - 2
231
232
Im
Re
233
234
39
10/24/12
Re
Im
235
236
1 f (t ) = 2
F () exp (i t) d
F ( ) =
f (t) exp(it) dt
40
10/24/12
Time
This light wave has many frequencies. And the frequency increases in time (from red to blue).
It will be nice if our measure also tells us when each frequency occurs.
Joseph Fourier
The resulting wave is periodic, but not harmonic. Essentially all waves are anharmonic.
Any function can be written as the sum of an even and an odd function.
sin(t)
E(-x) = E(x)
sin(3t)
O(x)
O(-x) = -O(x)
sin(5t)
f(x)
41
10/24/12
1 if 0 & % $ # " = ! mn
, mn
f ( t)
1 !
m =0
F m cos( mt )
where the set {Fm; m = 0, 1, } is a set of coefficients that define the series. And where we ll only worry about the function f(t) over the interval (,).
f (t )
F
m=0
cos(mt )
To find Fm, multiply each side by cos(m t), where m is another integer, and integrate:
f (t ) cos(m ' t ) dt
F
m=0
But:
$
m= 0
So:
f (t ) cos(m ' t) dt
from the m:
F
m=0
m,m '
Dropping the
Fm
f (t) cos(mt) dt
where we ll only worry about the function f(t) over the interval (,).
m,
Fourier Series
So if f(t) is a general function, neither even nor odd, it can be written:
f (t )
F sin(mt )
m=0
To find Fm, multiply each side by sin(m t), where m is another integer, and integrate:
But:
f (t ) sin(m ' t ) dt
f (t )
m =0
Fm cos(mt )
m =0
sin(mt ) Fm
even component
odd component
So:
f (t ) sin(m ' t ) dt
from the m:
Fm
F only the m
m =0
= m term contributes
F m =
f (t ) sin(mt ) dt yields the coefficients
for any f(t)!
Dropping the
F != m
"
f (t ) sin(mt ) dt
42
10/24/12
Fm vs. m
Let the integer m become a real number and let the coefficients, Fm, become a function F(m).
m
We really need two such plots, one for the cosine series and another for the sine series.
m
Again, we really need two such plots, one for the cosine series and another for the sine series.
f (t ) cos(mt ) dt i
f (t ) sin(mt ) dt
Let s now allow f(t) to range from to , so we ll have to integrate from to , and let s redefine m to be the frequency, which we ll now call :
F ( ) = f (t ) exp ( i t ) dt The Fourier
Transform
F() is called the Fourier Transform of f(t). It contains equivalent information to that in f(t). We say that f(t) lives in the time domain, and F () lives in the frequency domain. F() is just another way of looking at a function or wave.
f (t )
F cos(mt )
m m=0
F sin(mt )
' m m=0
Now transform the sums to integrals from to , and again replace Fm with F(). Remembering the fact that we introduced a factor of i (and including a factor of 2 that just crops up), we have:
f (t ) =
1 2
F ( ) exp(i t ) d
F ( )
=
1 2
f (t ) exp(it ) dt
FourierTransform
f(t) F()
If the function is already labeled by an upper-case letter, such as E, we can write: or:
f (t )
F ( ) exp(i t ) d
So we can transform to the frequency domain and back. Interestingly, these transformations are very similar. There are different definitions of these transforms. The 2 can occur in several places, but the idea is generally the same.
E (t ) F {E (t )}
% E (t ) E ( )
43
10/24/12
The Spectrum
We define the spectrum, S(), of a wave E(t) to be:
F ( ) = =
1/ 2
exp(it )dt =
1 2 [exp(it )]1/ 1/ 2 i
S ( ) F {E (t )}
F()
F ( ) = sinc(/2)
This is the measure of the frequencies present in a light wave.
Imaginary Component = 0
(t )
1
sinc2 ( / 2)
1
1/2
F ( ) = exp( at ) exp( it ) dt
0
F {exp(at 2 )} =
exp( at
) exp(it ) dt
exp( 2 / 4a)
exp( at 2 )
exp( 2 / 4a)
F ( 1 ) ! ia " = i"
A complex Lorentzian!
44
10/24/12
(t )
if t = 0 0 if t 0
It s best to think of the delta function as the limit of a series of peaked continuous functions.
fm(t) = m exp[-(mt)2]/
(t)
(t)
f3(t) f2(t) t f1(t) t
(t)
(t ) dt = 1
(t a) f (t ) dt = (t a) f (a) dt =
f (a)
exp(it ) dt = 2 ( ) exp[i( )t ]
dt = 2 ( )
1 exp(it ) dt = 2 ( )
2()
1
t
{exp(i0 t )}
exp(i t ) exp(i t ) dt
0
0
{cos(0t )}
=
= 1 2
exp(i [ ] t ) dt
t 0 t 0
= 2 ( 0 )
1 2
cos( t ) exp(i t ) dt
0
exp(i0t)
Im Re
F {exp(i0t)}
exp(i [ 0 ] t ) dt
1 2
exp(i [ + ] t ) dt
0
= ( 0 ) + ( + 0 )
cos(0t)
F {cos(0t )}
t 0
The function exp(i0t) is the essential component of Fourier analysis. It is a pure frequency.
-0
0
+0
45
10/24/12
F ( ) =
O(t ) dt = 0
= 0 if Re{f(t)} is odd
= 0 if Im{f(t)} is even
Re{F()}
F ( ) =
Functions that do not asymptote to zero in both the + and directions generally do not have Fourier transforms. So we ll assume that all functions of interest go to zero at .
f (t ) dt <
+ i
Re{ f (t )} sin(t) dt
Odd functions of
= 0 if Re{f(t)} is even
Im{F()}
46