Module 2: Lecture 5 Fourier Series Decomposition and Its Applications

Download as pdf or txt
Download as pdf or txt
You are on page 1of 4

Module 2: Lecture 5

Fourier Series Decomposition and its Applications


Prof. Vikram M. Gadre
EE210.1X
Indian Institute of Technology Bombay
February 19, 2015

IIT Bombay, 2015


This work is licensed under a Creative Commons Attribution-
ShareAlike 4.0 International License.
IITBombayX EE210.1X

1 Introduction
We have been decomposing periodic waveform into their periodic sinusoidal components. In this
lecture we look at a variation of the same decomposition using complex exponential functions.
One way of doing is to convert each sinusoidal component of sinusoidal decomposition into cor-
responding exponential functions. However, in some cases, we prefer to a priori decompose the
waveform in complex exponential components.
We shall also see how to go from the complex exponential components to sinusoidal components.
We then look at the applications of both sinusoidal and complex exponential decomposition, in
particular we will be analyzing its utility in determining output from a linear shift invariant sys-
tem on applying periodic input. We will also see some conditions under which we can not obtain
Fourier series decomposition for a periodic waveform.

2 Complex Exponential Fourier Series Decomposition


Suppose we are given a periodic signal x(t) with period T . We wish to write x(t) in terms of sum
of complex exponential functions.

X
x(t) = Ck e(j2πkt/T )
k=−∞

Basically, we are trying to write x(t) as a linear combination complex exponential functions rotat-
ing at angular frequencies 2πk/T for all integer k. Notice that complex exponentials with angular
frequency as positive integral multiple of 2πk/T represent anticlockwise rotating phasor while the
complex exponentials with negative angular frequency represent clockwise rotating phasor.

2.1 Orthogonality of Fourier Series Components


We now establish orthogonality between the complex exponential components. We take 2 integers
k and l and evaluate the inner product
1 (j2πkt/T ) (j2π(−l)t/T ) 1 (j2π(k−l)t/T )
Z T /2 Z T /2
e e dt = e dt (1)
−T /2 T −T /2 T
We have 2 cases now:
R T /2
• Case 1 k = l The integral in (1) becomes 1
−T /2 T
dt = 1

• Case 2 k 6= l The integral in (1) becomes


T /2
T /2
1 (j2π(k−l)t/T ) 1 e(j2π(k−l)t/T )
Z
e dt = =0
T T (j2π(k − l)/T )

−T /2
−T /2

2.2 Obtaining Complex exponential Decomposition from Sinusoidal De-


composition
Suppose we are given the sinusoidal components of the decomposition of a real waveform

X
x(t) = Ak cos(2πkt/T + φk )
0

Note that in above sum, for k = 0, we will have the component as A0 cos(φ0 ), we can merge
constant cos(φ0 ) into A0 and we will simply have constant A0 for k = 0. Now for non zero integer
k using cos(θ) = 21 (ejθ + (ejθ )∗ ), (Note: C ∗ denotes complex conjugate of C), we have
!
1
Ak cos(2πkt/T + φk ) = jφk j2πkt/T
Ak e e + (Ak e ) (e
jφk ∗ −j2πkt/T
)
2

2
IITBombayX EE210.1X

since Ak are real.


1
=⇒ Ck =
Ak ejφk
2
1
=⇒ C−k = (Ak ejφk )∗
2
And we also have trivially C0 = A0 Notice that for real waveform, we have Ck = (C−k )∗

2.3 Converting Complex Exponential to Sinusoidal Decomposition


Suppose we are given a real

X −1
X
x(t) = Ck e(j2πkt/T ) + Ck e(j2πkt/T ) + C0
k=1 k=−∞

Now, note that for integer k 6= 0,


T /2
T /2
1 e(j2πkt/T )
Z
e (j2πkt/T )
dt = =0
T (j2πk/T )

−T /2
−T /2

whereas, for k = 0,
Z T /2
1dt = T
−T /2

So, basically we can obtain C0 by integrating x(t) over time T . In the Fourier Decomposition only
C0 will give a non zero integral which is equal to C0 · T Hence,
1 T /2
Z
C0 = x(t)dt
T −T /2
Now, since for real x(t) we have Ck = (C−k )∗ , we get
Ck ej2πkt/T = (C−k e−j2πkt/T )∗
. So we have
Ck ej2πkt/T + C−k e−j2πkt/T = 2Re(Ck ej2πkt/T )
Using the polar form of Ck which is |Ck |eφk , we get

2Re(Ck ej2πkt/T ) = 2|Ck |Re(ej2πkt/T +φk = 2|Ck | cos(j2πkt/T + φk )


Finally, we have the sinusoidal decomposition

X
x(t) = C0 + 2|Ck | cos(j2πkt/T + φk )
k=1

3 Periodic Input to a Simple RC Circuit


Suppose, we apply a real periodic voltage waveform x(t) as input to a series RC-circuit with time
period T . Notice that an RC circuit is a linear shift invariant system. Suppose we can write
complex exponential decomposition of x(t) as

X
Ck e(j2πkt/T )
k=−∞

The advantage of doing this is that we can use the fact that a complex exponential input to a
linear shift invariant system simply gives the same complex exponential multiplied by a constant
as its ouput. By phasor analysis of the circuit, we have the transfer function
j(2πk/T )CR
1 + j(2πk/T )CR

3
IITBombayX EE210.1X

So, the output waveform will be simply



j(2πk/T )CR X
y(t) = Ck e(j2πkt/T )
1 + j(2πk/T )CR
k=−∞

This illustrates the power of having complex exponential decomposition of a waveform since we
can obtain the output waveform quite simply if it is passed through a linear shift invariant system.
Now, one catch in the above discussion is that we assumed that the decomposition of x(t) in to a
Fourier series is possible. It turns out it’s not always the case that we can write a Fourier series
decomposition for a periodic waveform! Although, for most of the practical waveforms, we can
write it. We look at the conditions under which we can write the Fourier series decomposition for a
periodic waveform. There are some waveforms in which we cannot find the Fourier decomposition.
One such example is x(t) = sin (1/f rac(t)) where f rac(t) denotes fractional part of t. So, x(t) is
periodic with period 1 but still its Fourier series decomposition fails to exist. There are certain
conditions under which Fourier analysis can be done, they are called the ‘Dirichlet’ Conditions.
We shall not discuss them here.

4 Periodic Inputs to a General Linear Shift Invariant Sys-


tem
In this section we use Fourier series decomposition to determine output from a linear shift invariant
system on applying a periodic input waveform. Of course for our analysis, will assume that Fourier
series decomposition of the input waveform exists.
Let S denote a linear shift invariant system, let h(t) be its impulse response. We apply a periodic
input waveform x(t). Now, using linearity of the system, output of the system is simply the sum
of output obtained for each component of Fourier series decomposition of x(t).
Let us derive the output for the k th component of x(t). The output for k th will be the convolution
of h(t) and the input itself Z ∞
h(τ ) · Ck ej2πk(t−τ )/T dτ
−∞

Since, the above integral runs over τ and not t, the above expression reduces to
Z ∞
Ck ej2πkt/T h(τ ) · e−j2πkτ /T dτ
−∞

The integral reduces to a constant depending on k. So, what we obtain is quite interesting as it
is simply the input itself multiplied by a constant! Let us denote the constant by H(k). So, the
output which is the sum of the outputs obtained for each component is

X
H(k)Ck ej2πkt/T
k=−∞

5 Conclusion
In this lecture we discussed about Fourier series decomposition, its properties and its application
in analysing linear shift invariant systems. In the coming lectures, we will discuss the significance
of the quantity H(k) we derived in previous section and begin with what is known as Fourier
Transform.

You might also like