‫‪Fourier‬‬

‫‪Transform‬‬
‫‪Dr/ Ahmed Omara‬‬

‫‪2015048‬‬ ‫عبدالرحمن طاهر‬

‫‪20140703‬‬ ‫محمد احمد عبدالرحيم‬
‫‪20140545‬‬ ‫عثمان حسن عثمان‬
 Introduction to the Fourier
Transform

The introduction section gives an overview of

why the Fourier Transform is worth learning.
It turns out the Fourier Transform is required
to understand one of the fundamental
secrets of the universe.....

Virtually everything in the world can be

described via a waveform - a function of time,
space or some other variable. For instance, sound
waves, electromagnetic fields, the elevation of a
hill versus location, a plot of VSWR versus
frequency, the price of your favorite stock versus
time, etc. The Fourier Transform gives us a
unique and powerful way of viewing these
waveforms.
The purpose of this entire website is to explain
the following fundamental fact:
 One of the Fundamental Secrets of the
Universe
All waveforms, no matter what you scribble or
observe in the universe, are actually just the sum
of simple sinusoids of different frequencies.

The above fact, is exceedingly cool, as we will

see. The Fourier Transform decomposes a
waveform - basically any real world waveform,
into sinusoids. That is, the Fourier Transform
gives us another way to represent a waveform.
(Need a refresher on sinusoids? See Sinusoid
Properties)
As an example, lets break down the waveform in
Figure 1 into its 'building blocks' (or constituent
frequencies). This decomposition can be done
with a Fourier transform (or Fourier series for
periodic waveforms), as we will see.
The first component is a sinusoidal wave with
period T=6.28 (2*pi) and amplitude 0.3, as
shown in Figure 1.
 Figure 1. First fundamental frequency
(left) and original waveform (right)
compared.

The second frequency will have a period half as

long as the first (twice the frequency). The
second component is shown on the left in Figure
2, along with the sum of the first two frequencies
compared to the original waveform.
Figure 2. Second fundamental frequency
(left) and original waveform compared
with the first two frequency
components.

We see that the sum of the first two frequencies

is starting to look like the original waveform.
The third frequency component is 3 times the
frequency as the first. The sum of the first 3
components are shown in Figure 3.

Figure 3. Third fundamental frequency

(left) and original waveform compared
with the first three frequency
components.
Finally, adding in the fourth frequency
component, we get the original waveform,
shown in Figure 4.
Figure 4. Fourth fundamental frequency
(left) and original waveform compared
with the first four frequency components
(overlapped).
Note that if we are taking the Fourier
Transform of a spatial function (a function
that varies with position, instead of time),
then our function g(x-a) would behave the
same way, with x in place of t.

Fourier Transform Properties

On this page, we'll get to know our new friend the

Fourier Transform a little better. Some
simple properties of the Fourier Transform will
be presented with even simpler proofs. On the next
page, a more comprehensive list of the Fourier
Transform properties will be presented, with less
proofs:
Linearity of Fourier Transform
First, the Fourier Transform is a linear transform.
That is, let's say we have two functions g(t) and
h(t), with Fourier Transforms given by G(f) and
H(f), respectively. Then the Fourier Transform of
any linear combination of g and h can be easily
found:
[Equation
1]
In equation [1], c1 and c2 are any constants (real or
complex numbers). Equation [1] can be easily
shown to be true via using the definition of the
Fourier Transform:

Shifts Property of the Fourier Transform

Another simple property of the Fourier Transform
is the time shift: What is the Fourier Transform of
g(t-a), where a is a real number?
 [Equation
2]

In the second step of [2], note that a simple variable

substition u=t-a is used to evaluate the integral.
Equation [2] should make some intuitive sense. If
the original function g(t) is shifted in time by a
constant amount, it should have the same
magnitude of the spectrum, G(f). That is, a time
delay doesn't cause the frequency content of G(f) to
change at all. This should make sense. Since the
complex exponential always has a magnitude of 1,
we see the time delay alters the phase of G(f) but
not its magnitude.
Note that if we are taking the Fourier Transform of a
spatial function (a function that varies with position,
instead of time), then our function g(x-a) would behave
the same way, with x in place of t.
Let g(t) have Fourier Transform G(f). If the function g(t)
is scaled in time by a non-zero constant c, it is written
g(ct). The resultant Fourier Transform will be given by:

Scaling Property of the Fourier Transform

 [Equation 3]

The proof of Equation [3] can be found using the

definition:

Now, if c is positive, the result is very simple:

If c is negative, the integration limits flip which

introduces an extra minus sign:
Hence, you can see that for the general case of
scaling with a real number c we get Equation [3].
(To see properties 2 and 3 in action together, this
link uses the scaling and shifting property on the
Gaussian.)
Derivative Property of the Fourier Transform
(Differentiation)
The Fourier Transform of the derivative of g(t) is
given by:

[Equation
4]

Convolution Property of the Fourier Transform

The convolution of two functions in time is defined
by:

[Equation
5]

The Fourier Transform of the convolution of g(t)

and h(t) [with corresponding Fourier Transforms
G(f) and H(f)] is given by:
[Equation
6]
Modulation Property of the Fourier Transform
A function is "modulated" by another function if
they are multiplied in time. The Fourier Transform
of the product is:
[Equation
7]
Parseval's Theorem
We've discussed how the Fourier Transform gives
us a unique representation of the original
underlying signal, g(t). That is, G(f) contains all the
information about g(t), just viewed in another
manner. To further cement the equivalence, in this
section we present Parseval's Identity for Fourier
Transforms.
Let g(t) have Fourier Transform G(f). Then the
following equation is true:

[Equation
8]

The integral of the squared magnitude of a function

is known as the energy of the function. For
example, if g(t) represents the voltage across a
resistor, then the energy dissipated in the resistor
will be proportional to the integral of the square of
g(t). Equation [8] states that the energy of g(t) is
the same as the energy contained in G(f). This is a
powerful result.
Duality
Suppose g(t) has Fourier Transform G(f). Then we
automatically know the Fourier Transform of the
function G(t):
[Equation
9]
This is known as the duality property of the Fourier
Transform.
 Fourier Transform
Applications

Signal processing
Filtering :

The Fourier Transform is extensively used

in the field of Signal Processing. In fact,
the Fourier Transform is probably the most
important tool for analyzing signals in that
entire field.
So what exactly is signal processing? I'll
try to give a one paragraph high level
overview.
A signal is any waveform (function of
time). This could be anything in the real
world - an electromagnetic wave, the
voltage across a resistor versus time, the air
pressure variance due to your speech (i.e. a
sound wave), or the value of Apple Stock
versus time. Signal Processing then, is the
act of processing a signal to obtain more
useful information, or to make the signal
more useful.
How can a signal be made better? Suppose
that you are listening to a recording, and
there is a low-pitched hum in the
background. By applying a low-
frequency filter, we can eliminate the hum.
Or suppose you have a digital photograph,
and it is very noisy (that is, there are
random specs of light everywhere). We can
use signal processing and fourier
transforms to filter out this undesirable
"noise".

LTI Systems
How exactly does the signal processing or
filtering work? I'm glad you asked. We'll
start by looking at common input-output
systems, known as Linear Time-Invariant
(LTI) systems [also known as Linear Shift-
Invariant (LSI)].
Suppose we have a box that accepts an
input signal and produces an output signal
from that. Such a box can be thought of as
a system:
 Figure 1. A System which takes an input
signal and produces and output signal.
The box of Figure 1 could be anything.
Examples are given below:

Input Output
Voltage Signal to an
Sound Waveform
Acoustic Speaker
Electric Current Voltage Across the
Through a Capacitor Capacitor
A plane
A Plane
Electromagnetic
Electromagnetic
Wave at another
Wave at Position 1
Location
Arbitrary Waveform
2*x(t)
x(t)
Arbitrary Waveform
z(t - 5)
z(t)

What are the restrictions on an LTI system?

Linearity. An LTI system must be linear.
To define this, suppose an input signal
x1(t) produces an output y1(t), and another
signal x2(t) produces the output y2(t). Then
the system is linear if the sum of the two
input signals produces the sum of the two
output signals. That is, if x1(t)+x2(t) is the
input signal, the output must be
y1(t)+y2(t). This must be true for every
possible input signal.
Time (Shift) Invariance. An LTI system
must be time-invariant. Suppose a signal
x(t) applied to the system produces an
output y(t). Then the delayed version of the
signal, x(t-c) (where c is some real
number), must give an output equal to y(t-
c). Loosely speaking, this means the
system doesn't care when the input is
applied, the result is always the same.
These two requirements define LTI
systems. They are not very restrictive, so
LTI system theory is very general
(although definitely not every system is
linear and shift-invariant). On this broad
class of systems, we will now discuss
filtering.
A Bit of System Theory
Before we jump into filtering, I'll have to give
some background on LTI system analysis. The
fundamental way to describe the response of an
LTI system is via the impulse response. That is, we
use the impulse function as the input signal and
view the corresponding output signal, known as the
impulse response.
Why is the impulse used? The answer again comes
from Fourier Transforms: the Fourier Transform of
the impulse is a constant with respect to frequency.
This means that if the input is an impulse function
in time, we are essentially sending equal energy at
all frequencies. That is, in the frequency domain,
the energy density is the same at every frequency.
As a result, when we view the Fourier Transform
of the output, we now know how the system reacts
to every possible frequency. The reason for this
goes back to the linearity of the Fourier Transform:
the impulse in time can be thought of as an infinite
sum of sinusoids at every possible frequency. The
output result then is the sum of the responses to
each frequency. [Side Note: The output of an LTI
system due to a single frequency input will have
only the same single frequency in the output signal.
This is a consequence of the time-invariance
property. The signal can be scaled in magnitude
and phase delayed, but the frequency cannot
change].
Hopefully a couple figures can clear this up if it is
cloudy. Figure 1 represents the time-domain view
of the LTI system. That is, an impulse is used as
the input signal, i(t), and the output signal h(t) is
observed:

Figure 1. Time-Domain View of LTI System.

Equivalently, we can view the above signals and
system in the frequency domain. By taking the
Fourier Transforms of the input and output signals,
we see that a constant input signal [I(f)=1] gives
rise to the output H(f), which is the frequency
response, in Figure 2:
Figure 2. Frequency-Domain View of LTI System.
Note that H(f) of Figure 2 is in general a complex
number. If he impulse response of a system is
written as h(t), then the Fourier Transform of the
impulse response is H(f), and is also known as
the Transfer Function. This Transfer Function
describes what the system does to every frequency
(i.e. the amplitude and phase changes).
In the next section, we'll describe frequency filters
using the transfer function.