Chapter 3 Review Notes

ECS204

1 History
The idea of Fourier series, which is completely mathematical in nature, had
come from the study of physics phenonmena.It is the heat propagation and
diffusion.Like how the heat distribution will look over time, and found series
in sinusoids to be useful in representation of the temperature distribution in
the body, and also had helped to explain many physical phenomena unrelated
to heat like Brownian motion and Black Scoles equation.

But here we will just know how the Fouries series help in representing the
various broad class of signals in terms of linear combination of complex expo-
nential signal.Before moving directly on the Fourier series, let’s understand
eigenvalues and function which help in understanding the usefulness of LTI
system and their outputs

2 Eigenvalues and Eigenfunction

The general idea of eigenfunction is given as follows:
Consider an operator A and function f(x) and if f(x) after going through the
operator A scales the function by a certain value, let’s say λ In mathematical
terms it can be written as

Af (x) = λf (x).

Then here f(x) is called Eigenfunction and λ is called Eigenvalue.

2.1 Complex Exponential signals as the eigenfunction

Basics of LTI system :

1. Any Linear combination of time varying signals as input produces an

output with the same linear combination, with each output correspond-
ing to the individual input.
 2. Second important property is shifting of time in the input signal by δt
also produces the output signal with same shift in time as δt

Here LTI system act as the linear operator and the complex signal in the
form of complex function act as eigenfunction.

So the complex exponential signal in both discrete and continuous time case
is the eigenfunction in context of the LTI system.It is represented as follows:

• Continuous time:
est → H(s)est

• Discrete time:
z n → H(z)z n

So here we denote the eigenvalue of complex exponential signal by H(s)

2.2 Proof of complex exponential signal as eigenfunc-

tion
Let us consider the continuous time LTI signal by x(t)=est and output by
y(t). By definition of the LTI system the output y(t) can be determined by
the input convolution integral, so that x(t)=est becomes,

2
 Z ∞
y(t) = h(τ )x(t − τ ) dτ
−∞
Z ∞
y(t) = h(τ )es(t−τ ) dτ
−∞
Z ∞
= h(τ )e−sτ est dτ
−∞
Z ∞
= h(τ )e−sτ est dτ
−∞
Z ∞
=e st
h(τ )e−sτ dτ
−∞

Now we denote est by the H(s). Therefore, our new equation becomes,
y(t) = H(s)est where H(s) is related by the equation,
Z ∞
H(s) = h(τ )e−sτ dτ
−∞

2.2.1 Proof in terms of discrete complex number

In exactly the same manner we can show the discrete complex exponential
by the x[n] = zn and y[n] = H(z)zn
∞
X
y[n] = h[k]x[n − k]
k=−∞

∞
X
= h[k]z [n−k]
k=−∞

Since summation is running over k so our equation becomes

∞
X
= zn h[k]z −k
k=−∞

Assuming the convergence of the complex exponential signal the above equa-
tion takes the form of
y[n] = H(z)z n

3
where,
∞
X
H(z) = h[k]z −k
k=−∞

2.3 Example of continuous time eigenfunction

So for analysis of LTI system, usefulness of decomposing more general signal
in eigenfunction can be seen from below example, Now consider the signal,
x(t) = a1 es1 t + a2 es2 t + a3 es3 t
Now each of the function are the eigenfunction so there corresponding out
put function are denoted by
a1 es1 t → a1 H(s1 )es1 t
a2 es2 t → a2 H(s2 )es2 t
a3 es3 t → a1 H(s3 )es3 t
So by definition of the LTI system output y(t) can be written as
y(t) = a1 H(s1 )es1 t + a2 H(s2 )es2 t + a3 H(s3 )es3 t
So in general if x(t) X
x(t) = ak esk t
k
so by the definition of LTI system the corresponding output is in the form of
the X
y(t) = ak H(sk )esk t
k
In similar manner if the discrete Fourier series are written in the form of the
X
x[n] = ak zkn
k

then output is in the form of the

X
y[n] = ak H(zk )zkn
k

So from the above example we can see that each coefficient is multiplied by
its eigenvalue H(sk ) associated with esk

4
3 Fourier Series Representation of Continu-
ous Time Periodic Signals
3.1 Linear Combination of Harmonically Related Com-
plex Exponentials
Signal is periodic if, for some positive values of T

x(t) = x(t + T ) f or all t

Some periodic signals with fundamental frequency ω0 and period T = 2π/ω0

x(t) = cos ω0 t
x(t) = ejω0 t

Harmonically related exponential form

Φk (t) = ejkω0 t = ejk(2π/T )t

A linear combination of harmonically related complex exponential

+∞
X +∞
X
jkω0 t
x(t) = ak e = ak ejk(2π/T )t (1)
k=−∞ k=−∞

is also periodic with period T in the above equation k = 0 is constant, and the
components for k = +N and k = −N are referred to as the N th components.

The representation of a periodic signal in the form of (1) is referred to as the

Fourier series representation.

Suppose that x(t) is real and can be represented in the form of eq.(1) then
since x∗ = x(t) we obtain
+∞
X
x(t) = a∗k e−jkω0 t
k=−∞

5
Replacing k by −k in summation we get
+∞
X
x(t) = a∗−k ejkω0 t
k=−∞

Now it requires that a∗k = a−k and further rearranging and substituting these
in eq.(1) we get
X∞
x(t) = a0 + ak ejkω0 t + a∗k e−jkω0 t
k=1

Since two terms inside the summation are complex conjugate of each other
and if ak is expressed in polar form we get
∞
X
x(t) = a0 + 2 Re{Ak ej(kω0 t+θk ) }
k=1

That is ∞
X
x(t) = a0 + 2 Ak cos (kω0 t + θk )
k=1

3.2 Determination of the Fourier Series Representa-

tion of a Continuous-time Periodic signal
By assuming a given periodic signal can be represented with the series of,
(1) we need a procedure for determining the coefficients ak

Now on multiplying both sides of eq.(1) and integrating from 0 to T = 2π/ω0

and  T 
ZT +∞
X Z
x(t)e−jnω0 t dt = ak  ej(k−n)ω0 t dt
0 k=−∞ 0

after evaluating it using Euler’s formula, we obtain,

ZT ZT ZT
j(k−n)ω0 t
e dt = cos (k − n)ω0 t dt + j sin (k − n)ω0 t dt.
0 0 0

6
For k ̸= n , cos(k − n)ω0 t and sin(k − n)ω0 t are sinusoids with fundamental
period (T /|k − n|). So for k ̸= n both integrals on right-hand side of eq.(2)
are zero For k = n the integrand on left equals 1 and thus the integral equals
T and consequently the right-hand side reduces to, T an Therefore
ZT
1
an = x(t)e−jnω0 t dt
T
0

we will obtain the same result if integrate over any interval of length T .

To summarize, if x(t) has a Fourier series representation then the coefficients

are given by equation ,
Z
1
an = x(t)e−jnω0 t dt (2)
T
T

This pair of equations then defines the Fourier series of periodic continuous-
time signal
+∞
X +∞
X
x(t) = ak ejkω0 t = ak ejk(2π/T )t
k=−∞ k=−∞
Z Z
1 −jkω0 t 1
ak = x(t)e dt = x(t)e−jk(2π/T )t dt
T T
T T

4 Convergence of Fourier Series

In this section we shall learn about the behaviour of the infinite sum and
when is it defined. When constructing the Fourier series of a signal, the first
ting that comes to mind is whether the signal is periodic. This periodicity
is important as it is what defines the harmonic series so, it is one of the first
conditions, i.e The signal must be periodic. The question now is whether
this property is enough to claim that the Fourier series exist, to answer this
question consider the signal x(t) = tan(t), this signal looks like,

Notice that this signal is periodic but, due to it being not bounded, its in-
tegral is not always defined so (2) doesn’t always yield a finite answer. So

7
 Figure 1: Signal x(t) = tan(t)

x(t) = tan(t) is a signal whose Fourier is not defined, which implies period-
icity is not sufficient.

4.1 Dirichlet condition

These are the conditions proposed by Peter Gustav Lejeune Dirichlet, a Ger-
man mathematician, which guarantees that the Fourier series of a given pe-
riodic signal x(t) is defined and exists,

•
R
|x(t)|dt < ∞ , i.e. the absolute integral of the signal over the time
T
period must be finite

• The signal must have finitely many maxima and minima in the time
period

• The signal must have finite number of discontinuities within the time
period

Together, these condition ensures that the periodic signal has a Fourier rep-
resentation.

8
4.2 Gibbs Phenomenon

Figure 2: Square wave

Consider the Square wave, a periodic signal with a discontinuous jump, as

discussed previously we can write the signal as:
∞
a0 X
x(t) = + an cos(nπt) (3)
2 1

Let’s observe what happens when we take different frequency into consider-
ation, i.e:
N
a0 X
xN (t) = + an cos(nπt) (4)
2 1

The interesting effect that we can observe in Figure 2 is that the behaviour of
the partial sum in the vicinity of the discontinuity exhibits ripples and that
the peak amplitude of these ripples does not seem to decrease with increasing
N. Josiah Gibbs, a mathematical physicist, showed that this is in fact the
case. So, for any finite N, the peak amplitude of the ripples remains constant.
This behaviour has come to be known as the Gibbs phenomenon.

9
 N=5 N=10

N=50 N=100
Figure 3: Convergence of the Fourier series representation of a square wave: an
illustration of the Gibbs phenomenon. Here, we have depicted the finite series
approximation xN (t) = a20 + N
P
1 a n cos(nπt) for several values of N.

We can also, observer that although the ripple at the discontinuity does not
seem to decrease the function itself is converging, as the ripples are getting
closer and closer to the point of discontinuity, hence decreasing the area of
ripples. So we must take sufficiently large values of N while approximating
so that the energy difference between the approximation and the signal is
negligible.

10
5 Properties of Continuous-time Fourier Se-
ries
Fourier series representations possess a number of important properties that
are useful for developing conceptual insights into such representations, and
they can also help to reduce the complexity of the evaluation of the Fourier
series of many signals.
We define a shorthand notation to indicate the relationship between a peri-
odic signal x(t) and its Fourier series coefficients ak as:
s F
x(t) ←→ ak (5)

Let’s discuss its properties,

5.1 Linearity
Let x(t) and y(t) be two periodic signals that obey Dirichlet conditions, so
there exist axk s and ayk s such that,
s F s F
x(t) ←→ axk and y(t) ←→ ayk (6)

Let, z(t) be a linear combination of x(t) and y(t), i.e, z(t) = Ax(t) + By(t)
be a periodic signal then we have,
s F
z(t) = Ax(t) + By(t) ←→ Aaxk + Bayk (7)

This property is called Linearity.

5.2 Time Shifting

s F
Let x(t) a periodic signal that obey Dirichlet conditions, so x(t) ←→ ak . Let
y(t) = x(t − t0 ) and let bk be the coefficients of Fourier series of y(t). We
know,

Z T Z T
1 −2jnπt 1 −2jnπt
bn = y(t)e T dt = x(t − t0 )e T dt (8)
T 0 T 0

11
 1 T
Z
−2jnπ(t0) −2jnπ(t−t0) −2jnπ(t0)
=e T x(t − t0 )e T dt = e T an (9)
T 0
it follows from above that
F −2jnπ(t−t0)
s
x(t − t0 ) ←→ e T axk (10)
One consequence of this property is that, when a periodic signal is shifted in
time, the magnitudes of its Fourier series coefficients remain unaltered. That
is, |bn | = |an |·

5.3 Time Reversal

Let x(t) be a nice signal with Fourier coefficients an , and let y(t) = x(−t),
we know time period of y(t) is same as time period of x(t). The Fourier
coefficients (say bn ) of the signal y(t) will obey
∞ ∞
X 2jnπt X −2jnπt
bn e T = y(t) = x(−t) = an e T (11)
−∞ −∞

We rearrange the equation above to obtain,

∞ ∞
X 2jnπt X 2jnπt
bn e T = a−n e T (12)
−∞ −∞

by comparison we get,
bn = a−n (13)
If our function is even, i.e x(−t) = x(t), then we know bn = an from (13) we
have, an = a−n

5.4 Time Scaling

Time scaling is an operation that in general changes the period of the under-
lying signal. Specifically, if x(t) is periodic with period T and fundamental
frequency ω0 = 2πT
, then x(αt), where α is a positive real number, is periodic
T
with period α and fundamental frequency αω0 . If x(αt) has a Fourier series
expansion then,
∞
X 2j(αn)πt
x(αt) = an e T (14)
−∞

12
While the Fourier coefficients have not changed, the Fourier series represen-
tation has changed because of the change in the fundamental. frequency

5.5 Multiplication
Let x(t) and y(t) be periodic functions that have Fourier series expansion
say ak and bk respectively, we define z(t) = x(t)y(t), assuming z(t) obeys
Dirichlet conditions then we know it has a Fourier expansion given by,

1 T 1 T
Z Z
−2jnπt −2jnπt
cn = z(t)e T dt = (x(t)y(t))e T dt (15)
T 0 T 0

We use the Fourier expansion of x(t) and y(t) and get,

Z T ∞ ∞
1 X 2jkπt X 2jlπt −2jnπt
cn = ( ak e T )( bl e T )e T dt (16)
T 0 k=−∞ l=−∞

since both k and l are independent of time we can obtain,

∞ ∞ Z
1 X X T 2jkπt 2jlπt −2jnπt
cn = (ak e T )(bl e T )e T dt (17)
T k=−∞ l=−∞ 0

Notice, the integral will give non-zero values if and only if l +k = n, in fact in
such situation the value of integral will be ak bl T ,so it follows that k = n − l.
We substitute all this in the above equation to get,
∞
X
cn = al bn−l or cn = an ∗ bn (18)
l=−∞

Notice, that his operation is the same as convolution.

5.6 Conjugation
Let x(t) be a nice periodic signal with coefficients ak , consider the complex
conjugate of x(t) be represented by (x(t))∗ then the Fourier series expansion
of it will be,
∞
X 2jkπt
(x(t))∗ = ( ak e T ) ∗ (19)
k=−∞

13
Since (a + b)∗ = a∗ + b∗ and (ab)∗ = a∗ b∗ , we get,
∞ ∞
X 2jkπt X −2jkπt
∗ ∗
(x(t)) = (ak e T ) = ak ∗ e T

k=−∞ k=−∞

∞
X 2jkπt
= a−k ∗ e T (20)
k=−∞

By comparison we get,
F
x(t)∗ ←→
s
a−k ∗ (21)

If x(t) be a real signal then we get x(t)∗ = x(t), it follows from above that
a−k ∗ = ak , Moreover if the signal is even then we know a−k = ak . Combining
both yields, ak ∗ = ak .
So if x(t) is a real even periodic signal then, its coefficients are always real.

5.7 Parseval’s theorem

Consider x(t), we know the average power of a signal of time period T is
given by,
1 T
Z
Pavg = |x(t)|2 dt
T 0
Using the property of complex numbers we get,

1 T
Z
Pavg = x(t)x(t)∗ dt
T 0
Using 5.6 we can obtain,
" ∞
#
Z T
1 X −2jkπt
Pavg = x(t) ak ∗ e T dt
T 0 −∞

∞
" #
Z T
X 1 −2jkπt
= ak ∗ x(t)e T dt
−∞
T 0

Notice the integral is the same as analysis equation, so we reduce it to,

14
 ∞
X
Pavg = ak ∗ [ak ]
−∞
∞
X
= |ak |2 (22)
−∞

So average power is equal to the sum of average power of all harmonics, this
is called Parseval’s theorem.

6 Fourier Series Representation of Discrete-

Time Periodic Signals
In this section, we consider the Fourier series representation of discrete-time
periodic signals.

6.1 Linear Combinations of Harmonically Related Com-

plex Exponentials
We know that, a discrete time signal x[n] is periodic if ∃N ∈ N such that,

x[n + N ] = x[n]

The smallest such N is called the fundamental period and ω0 = 2π

N
is called
the fundamental frequency. We again do something similar and say,
∞
X 2jπkn
x[n] = ak e N (23)
k=−∞

for some ak s, we define,

2jπkn
Φk [n] = e N (24)
Consider the signal Φk+pN [n], for some p ∈ Z, we use (24) and get,
2jπ(k+pN )n
Φk+pN [n] = e N

2jπkn 2jπpN n
=e N e N

15
 2jπkn
=e N e2jπpn
2jπkn
=e N

= Φk [n] (25)
What (25) means is that if we change k by an integer multiple of N then the
identical signal Φ is generated, so while writing discrete-time Fourier series
we only need to consider different signals, using this fact in (33) we obtain
that we just need our sum over N successive integers or,
N N −1+q
X 2jπkn X 2jπkn X 2jπkn
x[n] = ak e N = ak e N = ak e N (26)
k=1 k=q k∈⟨N ⟩

For some q ∈ Z.This equation is called Synthesis Equation for discrete-

time Fourier series. Notice the difference in Fourier expansion of discrete-
time and continuous-time signal, here we have a finite sum, so the sum of
signal is always defined. Therefore, there is no Dirichlet condition in the
case of discrete-time signals, all we need is a periodic signal.

6.2 Determination of the Fourier Series Representa-

tion of a Periodic Signal
Consider the equation xN = 1, we know by fundamental theorem of
algebra that this equation has n different solution in complex plane, in fact
the roots of the equation are of form, e2jkπ/N where k ∈ ⟨N ⟩. consider the
sum, X 2jkπn
e N for k ̸= 0
n∈⟨N ⟩

This sum is equivalent to the sum of roots of xq = 1 for some q, since the
equation doesn’t have any x terms we have sum of roots is zero so,
X 2jkπn
e N = 0 for k ̸= 0
n∈⟨N ⟩

also if k = 0 then this sum becomes N so we have

X 2jkπn  N, if k = 0

e N = (27)
0, otherwise
n∈⟨N ⟩

16
Consider the sum,
1 X −2jkπn
x[n]e N
N
n∈⟨N ⟩

We use (26) in this to get,

" # " #
1 X X 2jπpn −2jkπn 1 X X 2j(p−k)πn
ap e N e N = ap e N
N N
n∈⟨N ⟩ p∈⟨N ⟩ p∈⟨N ⟩ n∈⟨N ⟩

" #
1 X X 2j(p−k)πn
= ap e N (28)
N
p∈⟨N ⟩ n∈⟨N ⟩

Using (27) we claim that when p ̸= k then,

X 2j(p−k)πn
e N =0
n∈⟨N ⟩

and when p = k then, X 2j(p−k)πn

e N =N
n∈⟨N ⟩

using this in (28) yields,

1 X −2jkπn
ak = x[n]e N (29)
N
n∈⟨N ⟩

This equation is called Analysis Equation for discrete-time Fourier se-

ries, and we use this equation to determine the coefficients of Fourier series
expansion of discrete-time. Notice the similarity between this and analysis
equation of continuous-time signals, here the integral has become a sum, oth-
erwise the formula is essentially the same.
Another interesting fact we can obtain by is ak+N = ak . This easily follows
from (25).

17
6.3 Differences Between Discrete-Time Fourier Series
(DTFS) and Continuous-Time Fourier Series(CTFS)

DTFS CTFS
x[n] is periodic in n x(t) is periodic in t
ejkω0 n is periodic in both k and n ejkω0 t is periodic in both t but not in
k
Coefficients are periodic Coefficients are not periodic
x[n] is defined by a finite sum of har- x(t) is defined by an infinite sum of
monic signals harmonic signals
No Dirichlet condition Dirichlet conditions are present

7 Properties of Discrete-Time Fourier Series

A lot of properties of CTFS directly translates to DTFS, some changes a bit.
Here we will use the same notions we did before and discuss them again.

7.1 Linearity
Let x[n] and y[n] be periodic discrete-time signals of the same fundamental
time period, and let ak and bk be coefficients of them respectively, then
s F
Ax[n] + By[n] ←→ Aak + Bbk (30)
This property is similar to what we have seen in CTFS.

7.2 Time Shifting

Let x[n] be a periodic discrete-time signal and let ak be its Fourier coefficients,
let y[n] = x[n − n0 ] for some n0 ∈ Z and bn be its coefficients, by (26) we
have, X −2jkπn0
X 2jkπ[n−n0 ] 2jkπn
y[n] = ak e N = e N ak e N
k∈⟨N ⟩ k∈⟨N ⟩

By comparison we get,
−2jkπn0
bk = e N ak (31)

18
7.3 Multiplication
Let y[n] and x[n] be periodic discrete-time signals with period N and Fourier
coefficients ak and, bk respectively. We define z[n] = x[n]y[n], let cn be its
Fourier coefficients then, X
cn = al bn−l (32)
l∈⟨N ⟩

7.4 Parseval’s theorem

Parseval’s relation for discrete-time periodic signals is given by,
1 X X
Pavg = |x[n]|2 = |an |2 (33)
N
n∈⟨N ⟩ n∈⟨N ⟩

This is similar to what we observe in CTFS.

19
7.5 Properties Summary
Let x[n] and y[n] be discrete-time periodic signals with same period N with
Fourier coefficients ak and bk respectively, then,

Properties Signal Coefficients

Linearity Ax[n] + By[n] Aak + Bbk
−2jkπn0
Time Shifting x[n − n0] ak e N

Time Reversal x[−n] a−k

P
Multiplication x[n] · y[n] l∈⟨N ⟩ al bk−l

|al |2
P
Parseval’s theorem x[n] Pavg = l∈⟨N ⟩

Real signal x[n] is real a−k ∗ = ak

Real and even x[n] is real and even ak are real and even

Real and odd x[n] is real and odd ak are purely imaginary
and odd

8 Fourier Series and LTI systems

The Fourier series representation can be used to construct any periodic signal
in discrete-time, and essentially all periodic continuous-time signals of prac-
tical importance. Specifically, in continuous time, if x(t) = est is the input
to a continuous-time LTI system, then the output is given by y(t) = H(s)est ,
where
Z +x
H(s) = h(t)eδτ dt (34)
−x

which h(t) is the impulse response of the LTI system. Similarly, if x[n] = z n
is the input to a discrete-time LTI system, then the output is given by y[n] =

20
H(z)z n , where
x
X
H(z) = h[k]z −k (35)
k=−x

in which h[n] is the impulse response of the LTI system.

Figure 4: (a) The square-wave sequence x[r] in Example; (b) the sequence x[r]
equal to x[r] for −3 ≤ r ≤ 3 and zero otherwise; (c) the sequence x[n − r]; (d)
the sequence w[n] equal to the periodic convolution of x[n] with itself and to the
aperiodic convolution of x[n] with x[n]

When s or z are general complex numbers, H(s) and H(z) are referred to
as the system functions of the corresponding systems. For continuous-time
signals and systems, we focus on the specific case in which Re{s} = 0, so
that s = jω, and consequently, est is of the form ejωt — i.e., H(jω) viewed
as a function of ω — is referred to as the frequency response of the system

21
and is given by
Z ∞
H(jω) = (h(t)e−jωt )dt (36)
−∞

Similarly, for discrete-time signals and systems, we focus on values of z for

which |z| = 1, so that z = ejω and z = ejn . Then the system function H(z)
for z restricted to the form z = ejω is referred to as the frequency response
of the system and is given by
∞
X
H(ejω ) = (h[n]e−jωn ) (37)
n=−∞

The response of an LTI system to a complex exponential signal of the form

ejt (in continuous time) or ejωn (in discrete time) is particularly simple to
express in terms of the frequency response of the system. Furthermore, as
a result of the superposition property for LTI systems, we can express the
response of an LTI system to a linear combination of complex exponentials
with equal ease.

In the remainder of this chapter, as a first look at this important set of

concepts and results, we focus on interpreting and understanding this notion
in the context of periodic signals. Consider first the continuous-time case,
and let x(t) be a periodic signal with a Fourier series representation given by
∞
X
x(t) = ak ejkω0 t (38)
k=−∞

Suppose that we apply this signal as the input to an LTI system with impulse
response h(t). Then, since each of the complex exponentials in (38) is an
eigenfunction of the system, it follows that the output is
+x
X
y(t) = ak H(jkω0 )ejkω0 t (39)
k=−x

Thus, y(t) is also periodic with the same fundamental frequency as x(t).
Furthermore, if {ak } is the set of Fourier series coefficients for the input

22
x(t), then {ak H(jkω0 )} is the set of coefficients for the output y(t). That
is, the effect of the LTI system is to modify individually each of the Fourier
coefficients of the input through multiplication by the value of the frequency
response at the corresponding frequency.

Example 1: Suppose that the periodic signal x(t) is the input signal to an
LTI system with impulse response

h(t) = e−t u(t) (40)

To calculate the Fourier series coefficients of the output y(t), we first compute
the frequency response:
Z x
−jω
H(jω) = e−τ e−e dt (41)
0

1 −jω ∞
=− e−e |0 x
1 + jω

1
=
1 + jω
2π
Therefore, using (39) and (40), together with the fact that ω0 = T
in this
example, we obtain
3
X
y(t) = bk ejk2t (42)
k=−3

with bk = ak H(jk2), so that

b0 = 1,
1 1
b1 = (1 + j2), b−1 = (1 − j2),
4 4
1 1 (43)
b2 = (1 + j4), b−2 = (1 − j4),
2 2
1 1
b3 = (1 + j6), b−3 = (1 − j6)
3 3

23
Note that y(t) must be a real-valued signal, since it is the convolution of x(t)
and h(t), which are both real. This can be verified by examining (41) and
observing that b∗k = b−k .
3
2X
y(t) = 1 + Dk cos(2kt + k) (44)
3 k=1

Or
3
2X
y(t) = 1 + (Ek cos 2kt − Fk sin 2kt)
3 k=1

Where

bk = Dk ejk = Ek + Fk , k = 1, 2, 3 (45)

These coefficients can be evaluated directly from (42) For example,

1
D1 = |b1 | = √ , θ1 = arg(b1 ) = − tan−1 (2)
4 1 + 42
1 1
E1 = Re{b1 } = (1 + 42 ), F1 = Im{b1 } = − (1 + 42 )
4 2

In discrete time, the relationship between the Fourier series coefficients of the
input and output of an LTI system exactly parallels (38) and (39). Specif-
ically, let x[n] be a periodic signal with Fourier series representation given
by
N
X −1
x[n] = ak ejk(2π/N )n (46)
k=0

If we apply this signal as the input to an LTI system with impulse response
h[n], then with zk = ejk(2π/N ) , the output is
N
X −1
y[n] = ak H(ej2π/N )n (47)
k=0

24
Thus, y[n] is also periodic with the same period as x[n], and the kth Fourier
coefficient of y[n] is the product of the kth Fourier coefficient of the input
and the value of the frequency response of the LTI system, H(ej2πk/N ), at
the corresponding frequency.

An LTI system with impulse response h(t) = et u(t) is unstable, and it is

easy to check that the integral in (36) for H(j) diverges for any value of t.
Similarly, the LTI system in another example from the book, with impulse
response h[n] = nu[n], is stable for |a| < 1 and has frequency response.
However, if |a| > 1, the system is unstable, and then the summation diverges.

25