Chapter 11.

03
Fourier Transform Pair: Frequency and Time
Domain

Introduction
In Chapter 11.02, Fourier approximations were expressed in the time domain. The amplitude
(vertical axis) of a given periodic function can be plotted versus time (horizontal axis), but it
can also be plotted in the frequency domain [1-6] as shown in Figure 1.

Figure 1 Periodic Function (see Example 1 in Chapter 11.02) In Frequency Domain.

The advantages of plotting the amplitude of a given periodic function in frequency domain
(instead of time domain) are due to the following reasons:
For a specific value k (say k = 2 ) of the Fourier series in the time domain, one has to plot
the entire curve to observe the amplitude of a given periodic function (recall
f 2 (t ) = a 0 + a1 cos(t ) + b1 sin(t ) + a 2 cos(2t ) + b2 sin( 2t ) , see Example 1 in Chapter 11.02).
However, in the frequency domain, the amplitude can be plotted as a single point. (see Figure
1a).
In the frequency domain, one can easily identify which frequency (or corresponding to which
value of k ) contributes the most to the amplitude [see Figure 1(a)], where such
information is not readily available if time domain is used.
11.03.1

11.03.2

Chapter 11.03

From the amplitude plot in frequency domain [see Figure 1(a)], one can easily identify that
contributions to the amplitude beyond the 8th frequency ( k > 8 ) are not significant any more.
In real-life structural dynamics problems, such as the dynamical (time-dependent) response
of a (building) structure subjected to oscillated loads (for example, the operational machines
attached to the structures), the displacement superposition method is often used to predict the
(time dependent) displacement response of the structure. This method basically transforms
the original (large, coupled) equation of motion into a reduced (much smaller size, uncoupled) equation of motion by making use of the few free vibration mode shapes and its
associated frequencies. Knowledge of which frequencies (and the corresponding mode
shapes) that have the most contribution to the predicted dynamical response (such as nodal
displacement response) plays crucial roles for the algorithms efficiencies.
Detailed explanations on how to obtain Figures 1(a), and 1(b) are now presented in the
following sections.
Explanation of Figure 1(a) and 1(b)
One starts with Equation (18) and (20) of Chapter 11.02

~
f (t ) = C k e ikw0t
k =

where
T

~ 1
C k = f (t ) e ikw0t dt
T 0

For the periodic function shown in Example 1 of Chapter 11.02 (or Figure 1 of Chapter
11.02), one has
w0 = 2f
2
=
T
2
=
2
=1

~ 1
ikt
C k = t e dt + e ikt dt
T 0

Define, and using integration by parts formula

A t e

ikt

1
dt = t e ikt + e ikt dt
ik
0 0 ik

ik 1 1 ikt
A =
e + e
ik
0 ik ik
0

=
ik

ik 1
e + 2

ik
1
e

Fourier Transform Pair: Frequency and Time Domain

i
1
1
= e ik + 2 e ik 2
k
k
k
2

B e

ikt

( )

ik 2
=
e ik
e
ik

i
= e ik 2 e ik
k

Thus,

dt = e ikt

ik

~ 1
Ck =
{A + B}
2
1 ik i 1 i 1 i ik 2
=
e + 2 2 + e

k k
2
k k
k

Using the following Euler identities

e ik = cos(k ) + i sin( k )
= cos(k ) i sin(k )
= cos(k )

e ik ( 2 ) = cos(k (2 )) i sin( k (2 ))
= cos(k (2 ))
Hence, one obtains (noting that cos(k 2 ) = 1 , for any integer k ):
~ 1
Ck =
2
or,
~ 1
Ck =
2
Also, since:

Hence:
Thus,

1 1 i
cos(k ) 2 2 + cos(k 2 )
k

k k

1 i
1
2 cos(k ) 2 +
k
k
k

1 for k = odd integer (= 1,3,5,7,...)

cos(k ) =
+ 1 for k = even integer (= 2,4,6,8,...)

cos(k ) = ( 1)

k
1 i
( 1)
2 2 +
k
k
k
~ 1
( 1)k 1 + 1 i
Ck =
2
2k
2k
From Equation (15) in Chapter 11.02, one has:
~ a ibk
Ck = k
2

~ 1
Ck =
2

11.03.3

11.03.4

Chapter 11.03

Hence upon comparing the above 2 equations, one concludes

1
ak 2 (1) k 1
k
1
bk =
k
Remarks:
For k = 1,2,3,4,...,8; the values for ak and bk (based on the above 2 formulas) are exactly
identical as the ones presented earlier in Example 1 in Chapter 11.02.

Thus

~ a ib1
C1 = 1
2
2
i (1)

=
2
1 1
=
+ i
2
a ib2
~
C2 = 2
2
1
0 i
2
=
2
1
= 0+ i
4
a

ib3
~
C3 = 3
2
2 1
i

9 3

=
2
1 1
= + i
9 6
a ib4
~
C4 = 4
2
1
0 i
4
=
2
1
= 0+ i
8
~ a 5 ib5
C5 =
2

Fourier Transform Pair: Frequency and Time Domain

2 1

i
25 5

=
2
1 1
=
+ i
25 10
a ib6
~
C6 = 6
2
1
0 i
6
=
2
1
= 0+ i
12
a 7 ib7
~
C7 =
2

2 1

i
49 7

=
2
1 1
=
+ i
49 14
~ a ib8
C8 = 8
2
1
0 i
8
=
2
1
= 0+ i
16
In general, one has
1 1
+ i for k = 1,3,5,7,.. = odd integer

~ k 2 2k
Ck =
1

i for k = 2,4,6,8,.. = even integer

2k

11.03.5

11.03.6

Chapter 11.03

Representation of a complex number in polar coordinates

~
In Cartesian (rectangular) coordinates, a complex number Ck can be expressed as:
~
Ck = Rk + (I k )i
~
where Rk and I k represents the real and imaginary components of Ck , respectively.
~
In polar coordinates, a complex number Ck can be expressed as:
~
C k = Ae i = A{cos( ) + i sin( )} = {A cos( )} + {A sin( )}i
~
where A and represents the amplitude and phase angle of Ck , respectively (see Figure 2).

Figure 2 Representation of a complex number in polar coordinates

Thus, one obtains the following relations between the cartesian and polar coordinate systems:
Rk = A cos( )
I k = A sin ( )
Hence:
Rk2 + I k2 = A 2 cos 2 ( ) + A 2 sin 2 ( ) = A 2 cos 2 ( ) + sin 2 ( )

2
k

A =R +I

2
k

A = Rk2 + I k2
cos( ) =

Rk
R
implies = cos 1 k
A
A

Fourier Transform Pair: Frequency and Time Domain

sin( ) =

11.03.7

Ik
I
implies = sin 1 k
A
A

~
Based on the above 3 formulas, the complex numbers Ck , for k = 1,2,3,...,8 can be expressed
as
~ 1 1
C1 =
+ i
2
= (0.59272353)ei ( 2.13770783)
~
Hence, the amplitude A and Phase angle for C1 are 0.59272353, and 2.13770783 radians,
respectively. The readers should refer to Figures 1(a) and 1(b) to confirm the plotted values.
Important Notes
If one uses the formula
R
= cos 1 k
A
1

= cos 1
0.59272353

= 2.13770783 radians
= 122.48
However, the other formula for gives:
I
= sin 1 k
A
0.5

= sin 1

0.59272353
= 1.0038848 radians
= 57.52
Since Rk is negative, and I k is positive, the angle must be in the 2nd (or upper left)

quadrant of a circle (or 90 180 ). Thus, the correct value for should be
2.13770783 radians (or 122.48 o ) and the other value for = 1.0038848 radians must be
discarded.
Similarly, one obtains
1
~
C2 = 0 + i
4

i
2

= (0.25)e
= (0.25)ei (1.57079633)

11.03.8

Chapter 11.03

~ 1 1
C3 = + i
9 6
= (0.17037798)ei (1.77990097 )
1
~
C4 = 0 + i
8

i
2

= (0.125)e
= (0.125)ei (1.57079633)
~ 1 1
C5 =
+ i
25 10
= (0.100807311)ei (1.69743886 )
1
~
C6 = 0 + i
12

i
2

= (0.08333333)e
= (0.08333333)ei (1.57079633)
~ 1 1
C7 =
+ i
49 14
= (0.07172336)ei (1.66149251)
1
~
C8 = 0 + i
16

i
2

= (0.0625)e
In summary, the given periodic function (shown in Example 1 of Chapter 11.02) can also be
expressed in complex number formats, in polar coordinate with the amplitudes and phase
angles given in the following table (also refer to Figures 1(a), and 1(b)).
Table 1 Amplitude and phase angle (in radians) for varying k values.
Phase Angle (radians )
k Amplitude
1 0.59272353 2.13770783

= 1.57079633
2
1.77990097

2
3

0.25
0.14037798

4
5

0.125
2
0.100807311 1.69743886

6
7

0.08333333
0.07172336

0.0625

2
1.66149251

Fourier Transform Pair: Frequency and Time Domain

11.03.9

Non-Periodic Function
Recall that a periodic function can be expressed in terms of the exponential form,
accordingly to Equations (18, 20) of Chapter 11.02 as

~
f (t ) = Ck eikw0 t
k =

~ 1
Ck = f (t ) e ikw0 t dt
T 0

Define the following function

F (ikw0 ) =

T
2

f (t )e

ikw0 t

dt

(1)

where F (ikw0 ) is a function of i, k , and w0

Then, Equation (20) of Chapter 11.02 can be written as
~
1
C k = F (ikw0 )
T
and Equation (18) of Chapter 11.02 becomes

1
f (t ) = F (ikw0 )e ikw0t
k = T
A non-periodic function f np can be considered as a periodic function, with the period
1
0 (see Figure 3)
T
From Equations (6) and (7) from Chapter 11.01, one gets
w0 = 2f
2
=
T
= 2 (f )

(2)

(3)

T , or f

(4)

11.03.10

Chapter 11.03

Figure 3 Discretization of frequency data.

From Equation (3), one obtains
f np (t ) = lim f (t )
T
orf 0

= lim

f 0

(f ) F (ikw )e
0

k =

ikw0t

(5)

In the above equation, the subscript "np" denotes non-periodic function.

f np (t ) = lim

f 0

(f ) F (ik 2f )e

ik 2ft

(6)

k =

Realizing that kf = f (See Figure 3), the above equation becomes

f (t ) = df F (i 2f )e i 2ft
np

f np

(t ) = F (i 2f )e

i 2ft

df

Multiplying and dividing the right-hand-side of the equation by 2 , one obtains

1
i 2ft
f np (t ) =
F (i 2f )e d (2f )
2

1
iw t
=
F (iw0 )e 0 d ( w0 ) ; inverse Fourier transform
2
Using the definition stated in Equation (1), one has
F (iw0 ) =

np

(t )e iw0t d (t ) ; Fourier transform

(7)

(8)

(9)

Fourier Transform Pair: Frequency and Time Domain

11.03.11

Thus, Equations (9) and (8) will transform a non-periodic function from time domain to
frequency domain, and from frequency domain to time domain, respectively.
References
[1] E.Oran Brigham, The Fast Fourier Transform, Prentice-Hall, Inc. (1974).
[2] S.C. Chapra, and R.P. Canale, Numerical Methods for Engineers, 4th Edition, Mc-Graw
Hill (2002).
[3] W.H . Press, B.P. Flannery, S.A. Tenkolsky, and W.T. Vetterling, Numerical Recipies,
Cambridge University Press (1989), Chapter 12.
[4] M.T. Heath, Scientific Computing, Mc-Graw Hill (1997).
[5] H. Joseph Weaver, Applications of Discrete and Continuous Fourier Analysis, John
Wiley & Sons, Inc. (1983).
[6] Larry N. Thibos, Fourier Analysis for Beginners, Email: thibos@indiana.edu (1993,
2000, 2003).
FAST FOURIER TRANSFORM
Topic
Fourier Transform Pair: Frequency and Time Domain
Summary Textbook notes on Fourier Transform Pair: Frequency and Time Domain
Major
General Engineering
Authors
Duc Nguyen
Date
July 25, 2010
Web Site http://numericalmethods.eng.usf.edu