Frequency Domain Representation of Signals and LTI

Systems: Continuous Time Case

Dr. J. Aravinth
Frequency Response of LTI Systems
Frequency Response of Continuous-Time Systems

Consider a continuous-time LTI system with impulse response ℎ(𝑡).

Suppose a complex sinusoid is applied as input. i.e.,

𝑥(𝑡) = 𝑒
𝑗𝜔𝑡
.
Then, the otput 𝑦(𝑡) is given by
∞ ∞
𝑗𝜔(𝑡−𝜏)
𝑦(𝑡) = ℎ(𝜏)𝑥(𝑡 − 𝜏)𝑑𝜏 = ℎ(𝜏)𝑒 𝑑𝜏
∫ ∫
−∞ −∞

∞
𝑗𝜔𝑡 −𝑗𝜔𝜏 𝑗𝜔𝑡
= 𝑒 ℎ(𝜏)𝑒 𝑑𝜏 = 𝐻(𝑗𝜔)𝑒 , where
∫
−∞

∞
−𝑗𝜔𝜏
𝐻(𝑗𝜔) = ℎ(𝜏)𝑒 𝑑𝜏 = Frequency Response.
∫
−∞
Frequency Response of Continuous-Time Systems

In summary, the output of a continuous-time LTI system for a

complex sinusoid input is a complex sinusoid of the same
frequency, multiplied by a complex number 𝐻 (𝑗𝜔) , called the
frequency response of the system, which is a function of only the
frequency 𝜔 and not the time 𝑡 .

Therefore, we say that the complex sinusoid 𝜓(𝑡) = 𝑒𝑗𝜔𝑡 is an

eigenfunction of the LTI system with an associated eigenvalue
𝜆 = 𝐻(𝑗𝜔).
Frequency Response of Continuous-Time Systems

Substituting 𝐻(𝑗𝜔) = |𝐻(𝑗𝜔)|𝑒

𝑗 arg{𝐻(𝑗𝜔)}
, we obtain the
output corresponding to a complex sinusoid 𝑒𝑗𝜔𝑡 as

𝑗𝜔𝑡 𝑗(𝜔𝑡+arg{𝐻(𝑗𝜔)})
𝑦(𝑡) = 𝐻(𝑗𝜔)𝑒 = |𝐻(𝑗𝜔)|𝑒 ,

which implies that in steady state the LTI system modi es the
magnitude of the input by a factor |𝐻(𝑗𝜔)| and modi es the
phase by a shift of arg{𝐻(𝑗𝜔)}.

Therefore, |𝐻(𝑗𝜔)| is called the magnitude response and

arg{𝐻(𝑗𝜔)} is called the phase response of the LTI system.
Example: RC Circuit

1 𝑡

The impulse response is ℎ(𝑡) = 𝑒

−
𝑅𝐶 𝑢(𝑡) . Then, we have
𝑅𝐶

𝑅𝐶
𝐻(𝑗𝜔) = .
1
𝑗𝜔 +
𝑅𝐶

1/𝑅𝐶 −1
⇒ |𝐻(𝑗𝜔)| = and arg{𝐻(𝑗𝜔)} = − tan (𝜔𝑅𝐶).
2
√‾‾‾‾‾‾‾‾‾‾‾‾
+ (1/𝑅𝐶)‾
2
𝜔

Taking 𝑥(𝑡) = 𝑒
𝑗𝜔𝑡
𝑢(𝑡) , it can be shown that

𝑡
𝑗𝜔𝑡 − 𝑗𝜔𝑡
𝑦(𝑡) = 𝐻(𝑗𝜔) 𝑒 − 𝑒 𝑅𝐶 ⟶ 𝐻(𝑗𝜔)𝑒 in steady state.
( )
Frequency Response of Discrete-Time Systems

Consider a discrete-time LTI system with impulse response ℎ[𝑛] .

Suppose a complex sinusoid is applied as input. i.e.,

𝑥[𝑛] = 𝑒
𝑗Ω𝑛
.
Then, the otput 𝑦[𝑛] is given by

∞ ∞

𝑗Ω(𝑛−𝑘)
𝑦[𝑛] = ℎ[𝑘]𝑥[𝑛 − 𝑘] = ℎ[𝑘]𝑒
∑ ∑
𝑘=−∞ 𝑘=−∞

𝑗Ω𝑛 −𝑗Ω𝑘 𝑗Ω 𝑗Ω𝑛

= 𝑒 ℎ[𝑘]𝑒 = 𝐻(𝑒 )𝑒 , where
∑
𝑘=−∞

𝑗Ω −𝑗Ω𝑘
𝐻(𝑒 ) = ℎ[𝑘]𝑒 = Frequency Response.
∑
𝑘=−∞
Frequency Response of Discrete-Time Systems

In summary, the output of a discrete-time LTI system for a complex

sinusoid input is a complex sinusoid of the same frequency,
multiplied by a complex number 𝐻 (𝑒𝑗Ω ), called the frequency
response of the system, which is a function of only the frequency
Ω and not the time 𝑛.

𝑗Ω

Similarly, we can write 𝐻(𝑒

𝑗Ω
) = |𝐻(𝑒
𝑗Ω
)|𝑒
𝑗 arg{𝐻(𝑒 )}
,
where

|𝐻(𝑒
𝑗Ω
is called the magnitude response and
)|

)} is called the phase response of the LTI

𝑗Ω
arg{𝐻(𝑒

system.
Fourier Representation of Signals
Motivation for Fourier Representations

Suppose we can decompose a general input signal 𝑥(𝑡) as a

weighted sum of complex sinusoids as
𝑗𝜔 𝑘 𝑡
𝑥(𝑡) = 𝑎𝑘 𝑒 .
∑
𝑘

Then, the output 𝑦(𝑡) is also a weighted sum of exponentials, given

by

𝑗𝜔 𝑘 𝑡
𝑦(𝑡) = 𝑎𝑘 𝐻(𝑗𝜔 𝑘 )𝑒 .
∑
𝑘

Note that the operation of convolution, ℎ(𝑡) ∗ 𝑥(𝑡) , has been

replaced by multiplication, 𝑎𝑘 𝐻 (𝑗𝜔𝑘 ), because 𝑥(𝑡) has been
expressed as a sum of eigenfunctions; analogous relationships hold
in the discrete time as well.
Four Classes of Fourier Representations
Fourier Series Representation of Continuous-Time
Periodic Signals
Fundamental (Angular) Frequency and Harmonics

A sinusoid whose (angular) frequency 𝜔 is an integer multiple of a

fundamental (angular) frequency 𝜔0 is said to be a harmonic of the
sinusoid of the fundamental frequency.

The sinusoids 𝑒±𝑗𝑘𝜔 are the 𝑘-th harmonics of 𝑒𝑗𝜔 .

0 0

Consider representing a continuous-time periodic signal 𝑥(𝑡) with

fundamental period 𝑇0 as a weighted sum of complex sinusoids as

𝑗𝜔 𝑘 𝑡
𝑥(𝑡) = 𝑎𝑘 𝑒 .
∑
𝑘

How do we choose the frequencies 𝜔𝑘 and coef cients 𝑎𝑘 ?

Fundamental (Angular) Frequency and Harmonics

The weighted sum must have the same fundamental period 𝑇0 .

Therefore, each complex sinusoid in the weighted sum must

have a fundamental period 𝑇0 /𝑛 , 𝑛 = 1, 2, 3, …, etc., so
that the LCM of the fundamenatal periods of the
constituent complex sinusoids is 𝑇0 .
Equivalently, the frequency of each constituent complex
sinusoid must be an integer multiple of the signal 𝑥(𝑡)'s
2𝜋
fundamental frequency 𝜔0 = .
𝑇0

This implies that a continuous-time periodic signal 𝑥(𝑡) with

frequency 𝜔0 can be written as a weighted sum of the harmonics of
𝑒
𝑗𝜔
0
as
∞
𝑗𝜔 𝑘 𝑡 𝑗𝑘𝜔 0 𝑡
𝑥(𝑡) = 𝑎𝑘 𝑒 = 𝑎𝑘 𝑒 .
∑ ∑
𝑘 𝑘=− ∞
Fourier Series Coe cients

The weighted sum representation of a continuous-time periodic

signal 𝑥(𝑡) with fundamental period 𝑇0 can be re-written as

𝑗𝑘𝜔 0 𝑡
𝑥(𝑡) = 𝑋[𝑘]𝑒 ,
∑
𝑘=−∞

where any coef cient 𝑋[ℓ] is given by (why?)

𝑇0
1 −𝑗ℓ𝜔 0 𝑡
𝑋[ℓ] = 𝑥(𝑡)𝑒 𝑑𝑡 .
𝑇0 ∫
0

The FS coef cients 𝑋[𝑘] are known as a frequency-domain

representation of 𝑥(𝑡) because each FS coef cient is associated
with a complex sinusoid of a different frequency.
Validity of Fourier Series Expansion

We say that 𝑥(𝑡) and 𝑋[𝑘] are a FS pair and denote this
relationship as

CTFS
𝑥(𝑡) ⟷ 𝑋[𝑘].

The foregoing equations for weighted sum decomposition of 𝑥(𝑡)

and nding the corresponding Fourier Series coef cients 𝑋[𝑘] are
valid only if the Dirichlet conditions are satis ed:

𝑥(𝑡) is bounded or absolutely integrable over one time

period, i.e., if
|𝑥(𝑡)|𝑑𝑡 ≤ 𝑀𝑥 < ∞.
∫
𝑇0

𝑥(𝑡) has a nite number of maxima and minima in one

period.
𝑥(𝑡) has a nite number of discontinuities in one period.
Validity of Fourier Series Expansion

Here, 𝑥(𝑡) is unbounded.

Validity of Fourier Series Expansion

Here, 𝑥(𝑡) has an in nite number of maxima and minima in one

time period.
Validity of Fourier Series Expansion

Here, 𝑥(𝑡) has an in nite number of jump discontinuities in one

time period.
Simple Examples

Find the Fourier Series (FS) expansion of the following signals: (i)
𝑥(𝑡) = cos(𝜔0 𝑡), (ii) 𝑥(𝑡) = sin(𝜔0 𝑡) , (iii) 𝑥(𝑡) = cos(2𝑡 + ),
𝜋

(iv) 𝑥(𝑡) = cos(4𝑡) + sin(6𝑡) , (v) 𝑥(𝑡) = sin (𝑡)

2
.

Solution: (i) We have cos(𝜔0 𝑡) =

1

2 (𝑒
𝑗𝜔 0 𝑡
+ 𝑒
−𝑗𝜔 0 𝑡
) , which
implies that 𝑋[1] = 𝑋[−1] =
1

2
and 𝑋[𝑘] = 0 for all
𝑘 ≠ ±1 .

(ii) We have sin(𝜔0 𝑡) =

1

2𝑗 (𝑒
𝑗𝜔 0 𝑡
− 𝑒
−𝑗𝜔 0 𝑡
) , which implies that
𝑋[1] =
1

2𝑗
, 𝑋[−1] = −
1

2𝑗
and 𝑋[𝑘] = 0 for all 𝑘 ≠ ±1 .
Simple Examples

Find the Fourier Series (FS) expansion of the following signals: (i)
𝑥(𝑡) = cos(𝜔0 𝑡), (ii) 𝑥(𝑡) = sin(𝜔0 𝑡) , (iii) 𝑥(𝑡) = cos(2𝑡 + ),
𝜋

(iv) 𝑥(𝑡) = cos(4𝑡) + sin(6𝑡) , (v) 𝑥(𝑡) = sin (𝑡)

2
.

Solution: (iii) We have 𝜔0 = 2 , and

𝜋 1 𝑗(2𝑡+
𝜋
−𝑗(2𝑡+
𝜋
) )
cos(2𝑡 + ) = 𝑒 4 + 𝑒 4
( )
4 2

1 𝑗
𝜋
𝑗2𝑡
1 −𝑗
𝜋
−𝑗2𝑡
= 𝑒 4 𝑒 + 𝑒 4 𝑒 ,
2 2

𝜋 𝜋

which implies that 𝑋[1] =

1

2
𝑒
𝑗
4 , 𝑋[−1] =
1

2
𝑒
−𝑗
4 and
𝑋[𝑘] = 0 for all 𝑘 ≠ .
±1
Simple Examples

Find the Fourier Series (FS) expansion of the following signals: (i)
𝑥(𝑡) = cos(𝜔0 𝑡), (ii) 𝑥(𝑡) = sin(𝜔0 𝑡) , (iii) 𝑥(𝑡) = cos(2𝑡 + ),
𝜋

(iv) 𝑥(𝑡) = cos(4𝑡) + sin(6𝑡) , (v) 𝑥(𝑡) = sin (𝑡)

2
.

Solution: (iv) We have

1 2
2𝜋 2𝜋
𝑇0 = 𝐿𝐶𝑀(𝑇 , 𝑇 ) = 𝐿𝐶𝑀 ,
( 𝜔1 𝜔
2 )

2𝜋 2𝜋
= 𝐿𝐶𝑀 , = 𝜋,
( 4 6 )

which implies that 𝜔0 =

2𝜋

𝑇0
= 2 , and

1 𝑗4𝑡
1 −𝑗4𝑡
1 𝑗6𝑡
1 −𝑗6𝑡
cos(4𝑡) + sin(6𝑡) = 𝑒 + 𝑒 + 𝑒 − 𝑒
2 2 2𝑗 2𝑗

1 𝑗2𝜔 0 𝑡
1 −𝑗2𝜔 0 𝑡
1 𝑗3𝜔 0 𝑡
1 −𝑗3𝜔 0 𝑡
= 𝑒 + 𝑒 + 𝑒 − 𝑒 ,
2 2 2𝑗 2𝑗

which implies that 𝑋[2] = 𝑋[−2] =

1

2
, 𝑋[3] =
1

2𝑗
,
𝑋[−3] = −
2𝑗
1
and 𝑋[𝑘] = 0 for all 𝑘 ≠ ±2, ±3 .
Simple Examples

Find the Fourier Series (FS) expansion of the following signals: (i)
𝑥(𝑡) = cos(𝜔0 𝑡), (ii) 𝑥(𝑡) = sin(𝜔0 𝑡) , (iii) 𝑥(𝑡) = cos(2𝑡 + ),
𝜋

(iv) 𝑥(𝑡) = cos(4𝑡) + sin(6𝑡) , (v) 𝑥(𝑡) 2

= sin (𝑡) .

Solution: (v) We have sin2 (𝑡) =

1

2
(1 − cos(2𝑡)) , which implies
that 𝜔0 = 2 , and

1 1 1 𝑗2𝑡 −𝑗2𝑡
(1 − cos(2𝑡)) = − (𝑒 + 𝑒 )
2 2 4

which implies that 𝑋[0] =

1

2
, 𝑋[1] = 𝑋[−1] = −
1

4
, and
𝑋[𝑘] = 0 for all 𝑘 ≠ 0, ±1 .
Exponential and Trigonometric Forms

The decomposition of a continuous-time periodic signal 𝑥(𝑡) with

period 𝑇0 as a weighted sum of complex exponentials, as discussed
earlier, and given by

𝑗𝑘𝜔 0 𝑡
𝑥(𝑡) = 𝑋[𝑘]𝑒 ,
∑
𝑘=−∞

is called the exponential form of FS representation.

Applying Euler's identity, we get the trigonometric form of FS

representation as

𝑗𝑘𝜔 0 𝑡 −𝑗𝑘𝜔 0 𝑡
𝑥(𝑡) = 𝑋[0] + (𝑋[𝑘]𝑒 + 𝑋[−𝑘]𝑒 )
∑
𝑘=1
,
∞

= 𝑋[0] + 𝐵[𝑘] cos(𝑘𝜔 0 𝑡) + 𝐴[𝑘] sin(𝑘𝜔 0 𝑡)

∑
𝑘=1
Exponential and Trigonometric Forms

where for any 𝑥(𝑡) (real- or complex-valued) we have

𝑇0
1
𝑋[0] = 𝑥(𝑡)𝑑𝑡 = time-averaged value of 𝑥(𝑡)
𝑇0 ∫0

𝑇0
2
𝐵[𝑘] = (𝑋[𝑘] + 𝑋[−𝑘]) = 𝑥(𝑡) cos(𝑘𝜔 0 𝑡)𝑑𝑡
𝑇0 ∫0

𝑇0
2
𝐴[𝑘] = 𝑗 (𝑋[𝑘] − 𝑋[−𝑘]) = 𝑥(𝑡) sin(𝑘𝜔 0 𝑡)𝑑𝑡
𝑇0 ∫
0

Recall that any FS coef cient 𝑋[ℓ] is given by

𝑇0
1 −𝑗ℓ𝜔 0 𝑡
𝑋[ℓ] = 𝑥(𝑡)𝑒 𝑑𝑡 .
𝑇0 ∫
0
Exponential and Trigonometric Forms

From the previous equation, it is easy to see that if 𝑥(𝑡) is real-

∗
valued, then 𝑋[−ℓ] = 𝑋 [ℓ] , where 𝑋 ∗ [ℓ] denotes the
complex conjugate of 𝑋[ℓ].

∗
Substituting 𝑋[−ℓ] = 𝑋 [ℓ] , we get the coef cients of the
trigonometric form of FS representation for a real-valued signal
𝑥(𝑡) as

∗ ∗
𝐵[𝑘] = (𝑋[𝑘] + 𝑋 [𝑘]) and 𝐴[𝑘] = 𝑗 (𝑋[𝑘] − 𝑋 [𝑘])

If 𝑥(𝑡) is an even (resp. odd) signal, then the trigonometric form of

FS expansion will consist of only cosine (resp. sine) terms.
Square Wave Example 1
𝑋[0] = 𝐴/2, 𝑋[2𝑚] = 0, 𝑋[2𝑚 + 1] = 𝐴/ [𝑗 (2𝑚 + 1) 𝜋]

𝐴[𝑘] =?𝐵[𝑘] =?
Square Wave Example 2 𝑚
𝑋[0] = 𝐴/2, 𝑋[2𝑚] = 0, 𝑋[2𝑚 + 1] = (−1) 𝐴/ [(2𝑚 + 1) 𝜋]

𝐴[𝑘] =?𝐵[𝑘] =?

Can we get the coef cients directly from Example 1?

Square Wave Example 3
𝑋[0] = 0, 𝑋[2𝑚] = 0, 𝑋[2𝑚 + 1] = 2𝐴/ [𝑗 (2𝑚 + 1) 𝜋]

𝐴[𝑘] =?𝐵[𝑘] =?

Can we get the coef cients directly from Example 1?

Amplitude and Phase Spectra of a CT Periodic Signal

The FS coef cients can be written in polar form as

𝑗𝜃𝑘
𝑋[𝑘] = |𝑋[𝑘]|𝑒 ,

where |𝑋[𝑘]| and 𝜃𝑘 as functions of discrete frequency 𝑘𝜔0 are

called the magnitude spectrum and phase spectrum, respectively.

Recall that for a real-valued signal 𝑥(𝑡) we have

∗
𝑋[−𝑘] = 𝑋 [𝑘] , which implies that
Amplitude and Phase Spectra of a CT Periodic Signal
∗
(i) |𝑋[−𝑘]| = |𝑋 [𝑘]| = |𝑋[𝑘]| , i.e., the magnitude
spectrum is an even function of the discrete index 𝑘.
(ii) 𝜃−𝑘 = arg{𝑋 ∗ [𝑘]} = − arg{𝑋[𝑘]} = −𝜃𝑘 , i.e.,
the phase spectrum is an odd function of the discrete index
𝑘.

For the above reason of symmetry, when dealing with real-valued

signals 𝑥(𝑡), one can ignore the negative frequencies altogether.

If the signal 𝑥(𝑡) is complex-valued, then the above

symmetry will not hold and one has to consider both
positive and negative frequencies separately.
Example: Periodic Rectangular Pulses

2𝜋
For 𝑥(𝑡) shown below, we have 𝜔0 = .
𝑇
Example: Periodic Rectangular Pulses

2𝑇0
It can be shown that 𝑋[0] = and for all 𝑘 ≠ 0 , we have
𝑇

2 sin(𝑘𝜔 0 𝑇0 ) 2𝑇0 sin(𝑘𝜔 0 𝑇0 )

𝑋[𝑘] = =
𝑘𝜔 0 𝑇 𝑇 𝑘𝜔 0 𝑇0

2𝑇0 sin(𝑘2𝜋𝑇0 /𝑇 ) 2𝑇0 2𝑇0

= = sinc 𝑘 ,
𝑇 𝑘2𝜋𝑇0 /𝑇 𝑇 ( 𝑇 )

where
sin(𝜋𝑥)
sinc(𝑥) = .
𝜋𝑥
Example: Periodic Rectangular Pulses

Can you sketch the amplitude and phase spectra?

Output Computation of an LTI System using FS

Suppose a continuous-time periodic signal 𝑥(𝑡) is applied as input

to an LTI system with impulse response ℎ(𝑡).

From time-domain analysis, we know that the output 𝑦(𝑡) can be

obtained by taking a convolution of the input 𝑥(𝑡) and impulse
response ℎ(𝑡).

An alternate method that does not require convolution:

CTFS
Obtain the FS spectrum of 𝑥(𝑡). Let 𝑥(𝑡) ⟷ 𝑋[𝑘] .
Obtain the frequency response of the LTI system 𝐻(𝑗𝜔).
Obtain the FS spectrum of 𝑦(𝑡) from that of 𝑥(𝑡) as
𝑌 [𝑘] = 𝐻(𝑗𝑘𝜔 0 )𝑋[𝑘] .
CTFS
Obtain the output 𝑌 [𝑘] ⟷ 𝑦(𝑡) .
Example: Square Wave Through an RC Circuit

Recall that for the RC circuit, the frequency response is given by

𝑅𝐶
𝐻(𝑗𝜔) = .
1
𝑗𝜔 +
𝑅𝐶

The magnitude and phase responses are given by (plotted on the

next slide)

1/𝑅𝐶 −1
|𝐻(𝑗𝜔)| = and arg{𝐻(𝑗𝜔)} = − tan (𝜔𝑅𝐶).
2
√𝜔 ‾‾‾‾‾‾‾‾‾‾‾‾
2
+ (1/𝑅𝐶)‾
Example: Square Wave Through an RC Circuit
Example: Square Wave Through an RC Circuit

Suppose a square wave as shown below is applied as input 𝑥(𝑡).

We had obtained the FS coef cients 𝑋[𝑘] earlier.

Example: Square Wave Through an RC Circuit

The output is then given by

∞ ∞

𝑗𝜔 0 𝑡 𝑗𝜔 0 𝑡
𝑦(𝑡) = 𝑌 [𝑘]𝑒 = 𝐻(𝑗𝑘𝜔 0 )𝑋[𝑘]𝑒
∑ ∑
𝑘=−∞ 𝑘=−∞

𝑗𝜔 0 𝑡
≈ 𝐻(𝑗𝑘𝜔 0 )𝑋[𝑘]𝑒
∑
𝑘=−𝐾

Can you sketch the output (at least approximately)?

Ex: Square Wave Through An Ideal Lowpass Filter
Ex: Square Wave Through An Ideal Lowpass Filter
Ex: Square Wave Through An Ideal Lowpass Filter
Ex: Square Wave Through An Ideal Lowpass Filter
CTFS Properties
Linearity
Time Shifting (Translation)
Frequency Shifting (Modulation)
Time Reversal (Re ection)
Conjugation
Periodic Convolution
Multiplication
Di erentiation in Time

We have

𝑑 CTFS
𝑥(𝑡) ⟷ 𝑗𝑘𝜔 0 𝑋[𝑘].
𝑑𝑡
Scaling in Time

We have

CTFS
𝑥(𝑎𝑡) ⟷ 𝑋[𝑘],

but the spacing between the Fourier coef cients changes from 𝜔0
to 𝑎𝜔0 .
Even and Odd Symmetry
Real-Valued Functions
Real-Valued Functions
Parseval's Relation
Other Properties
Table of CTFS Properties
Table of CTFS Properties 2
Continuous Time Fourier Transform (CTFT)
Motivation for CTFT
CTFT from CTFS

Recall the example of periodic rectangular pulses 𝑥(𝑡) discussed

earlier.
CTFT from CTFS

Recall the corresponding CTFS coef cients 𝑋[𝑘] obtained earlier.

CTFT from CTFS

If we make the period 𝑇 grow very large, then 𝑥(𝑡) becomes a non-
2𝜋
periodic signal and the fundamental frequency 𝜔0 = reduces
𝑇
to Δ𝜔 ≈ 0.

→
Then, it is intuitively clear that in the limit 𝑇 ∞ the
discrete-index function 𝑋[𝑘] becomes a function 𝑋(𝜔) of
continuous frequency 𝜔.
CTFT from CTFS

The FS coef cient 𝑋[𝑘] becomes

2𝑇0 sin(𝑘2𝜋𝑇0 /𝑇 ) 2𝑇0 sin(𝑘Δ𝜔𝑇0 ) 1 2 sin(𝑘Δ𝜔𝑇0 )

𝑋[𝑘] = = = Δ𝜔,
𝑇 𝑘2𝜋𝑇0 /𝑇 𝑇 𝑘Δ𝜔𝑇0 2𝜋 𝑘Δ𝜔

→
and in the limit 𝑇 ∞, we have Δ𝜔 → 𝑑𝜔 and 𝑘Δ𝜔 → 𝜔 , and
the FS representation of 𝑥(𝑡) becomes

∞ ∞
𝑗𝑘𝜔 0 𝑡
1 2 sin(𝜔𝑇0 ) 𝑗𝜔𝑡
𝑥(𝑡) = 𝑋[𝑘]𝑒 −−−−→ 𝑒 𝑑𝜔
∑ 2𝜋 ∫−∞ 𝜔
T→∞ 
𝑘=−∞

𝑋(𝜔)
CTFT from CTFS

In general, for any periodic signal 𝑥(𝑡) with period 𝑇 , we have

∞ ∞ 𝑇 /2
𝑗𝑘𝜔 0 𝑡
1 −𝑗𝑘(2𝜋/𝑇 )𝑡 𝑗𝑘𝜔 0 𝑡
𝑥(𝑡) = 𝑋[𝑘]𝑒 = 𝑥(𝑡)𝑒 𝑑𝑡 𝑒
∑ ∑ (𝑇 ∫ )
−𝑇 /2
𝑘=−∞ 𝑘=−∞

∞ ∞ ∞
1 −𝑗𝜔𝑡 𝑗𝜔𝑡
1 𝑗𝜔𝑡
−−−−→ 𝑥(𝑡)𝑒 𝑑𝑡 𝑒 𝑑𝜔 = 𝑋(𝜔)𝑒 𝑑𝜔,
T→∞
2𝜋 ∫−∞ (∫−∞ ) 2𝜋 ∫−∞

∞
where 𝑋(𝜔) = 𝑥(𝑡)𝑒
−𝑗𝜔𝑡
𝑑𝑡 is called the Fourier
∫
− ∞
Transform of the non-periodic signal 𝑥(𝑡).
Dirichlet Conditions for the Existence of CTFT

The foregoing equation representing 𝑥(𝑡) as an integration

(weighted sum) of a continuum of frequency components and the
equation representing the corresponding Fourier Transform 𝑋(𝜔)
are valid at all points, except at discontinuities, only if the Dirichlet
conditions are satis ed:

𝑥(𝑡) is absolutely integrable, i.e., if

∞

|𝑥(𝑡)|𝑑𝑡 ≤ 𝑀𝑥 < ∞.
∫
−∞

𝑥(𝑡) has a nite number of maxima, minima and

discontinuities in any nite interval, and the size of each
discontinuity is nite.
Dirichlet Conditions for the Existence of CTFT

If the FT exists, then we say that 𝑥(𝑡) and 𝑋(𝜔) form a FT pair and
denote this relationship as

CTFT
𝑥(𝑡) ⟷ 𝑋(𝜔).

Example: Consider the signal 𝑥(𝑡) = 𝑒−𝑎𝑡 𝑢(𝑡). Does this signal
have a Fourier Transform? If yes, nd the FT.

Solution: FT exists only if 𝑎 > 0 . If FT exists, then we have

1
𝑋(𝜔) =

⇒
𝑎 + 𝑗𝜔
−1/2
2 2
|𝑋(𝜔)| = (𝑎 + 𝜔 ) and arg{𝑋(𝜔)} =

−1
− tan (𝜔/𝑎).
CTFT Properties
Linearity
Time Shifting (Translation)
Frequency Shifting (Modulation)
Time and Frequency Scaling
Conjugation
Duality
Convolution
Multiplication
Di erentiation in Time Domain
Di erentiation in Frequency Domain
Integration in Time Domain
Even and Odd Symmetry
Real-Valued Functions
Parseval's Relation
Other Properties
Time-Bandwidth Product
Table of CTFT Properties
Table of CTFT Properties
References:
[1] Simon Haykin and Barry Van Veen, Signals and Systems, Second
Edition, John Wiley and Sons, 2003.

[2] Lecture Notes by Michael D. Adams.

https://www.ece.uvic.ca/~frodo/sigsysbook/downloads/lecture_slides_for_signals_and_systems
2.0.pdf
(https://www.ece.uvic.ca/~frodo/sigsysbook/downloads/lecture_slides_for_signals_and_system
2.0.pdf)

[3] Lecture Notes by Richard Baraniuk.

https://www.di.univr.it/documenti/OccorrenzaIns/matdid/matdid018094.pdf
(https://www.di.univr.it/documenti/OccorrenzaIns/matdid/matdid018094.pdf)