Chapter 4: Continuous-Time and Discrete-Time Fourier Transforms

4.1 Fourier Transform

Purpose
Non-periodic signals can be represented using the Fourier Transform (FT).
The Fourier transform provides an effective and reversible link between the time domain and frequency domain representations of a
signal.
2π
For non-periodic signals, the period T0 ​ → ∞, so the fundamental frequency ω0 = ​

T0 ​
​ → 0.
This means the spacing between spectral components becomes infinitesimally small, resulting in a continuous spectrum.

4.1.1 Definition of Fourier Transform

The Fourier Transform of a continuous-time signal x(t) is defined as:
∞ ∞
Fourier Transform: X(ω) = ∫ ​ x(t)e −jωt
dt or X(f ) = ∫ ​ x(t)e−j2πf t dt (4.1.1)
−∞ −∞

Where:

x(t) is the time-domain representation of the signal,

X(ω) or X(f ) is the frequency-domain representation,
ω is the angular frequency in radians per second (ω = 2πf ),
f is the frequency in Hertz (Hz).

Sometimes X(ω) is also written as X(jω), especially in system analysis.

Inverse Fourier Transform

The original signal x(t) can be recovered from X(ω) using the Inverse Fourier Transform:
∞ ∞
1
Inverse Fourier Transform: x(t) = ∫ X(ω)ejωt dω or x(t) = ∫ X(f )ej2πf t df (4.1.2)
2π −∞
​ ​ ​

−∞

This shows that the Fourier transform pair is reversible.

Fourier Transform Pair

A Fourier transform pair is denoted as:
FT FT
x(t) ​
X(ω) or x(t) ​
X(f )

Note:

X(ω) is often used in theoretical analysis.

X(f ) is common in engineering and communication systems.

4.1.2 Existence of Fourier Transform – Dirichlet Conditions

For the Fourier transform of a function x(t) to exist, the following Dirichlet conditions should be satisfied:

i) Single-valued property

x(t) must have only one value at any time instant over a finite time interval T .

ii) Finite discontinuities

x(t) should have at most a finite number of discontinuities over any finite time interval T .
iii) Finite peaks

The signal x(t) should have a finite number of maxima and minima over any finite time interval T .

iv) Absolute integrability

x(t) must be absolutely integrable, i.e.,

∞
∫ ​ ∣x(t)∣ dt < ∞
−∞

This condition ensures that the integral defining the Fourier transform converges.

Note: These conditions are sufficient but not necessary for the existence of the Fourier transform. Some signals (e.g., sinusoids,
impulse functions) do not satisfy these conditions but still have Fourier transforms in the distributional sense.

4.1.3 Properties of Fourier Transform

[Link] Linearity
If

FT FT
x(t) ​ X(ω) and y(t) ​ Y (ω)

then,

FT
z(t) = ax(t) + by(t) ​ Z(ω) = aX(ω) + bY (ω) (4.1.3)

Meaning:
The Fourier transform of a linear combination of signals is equal to the same linear combination of their individual Fourier transforms. This
is also known as the principle of superposition.

Proof:
By definition of Fourier transform:
∞ ∞
Z(ω) = ∫ ​
z(t)e−jωt dt = ∫ [ax(t) + by(t)]e−jωt dt
​

−∞ −∞

Distribute the integral:

∞ ∞
= a∫ ​ x(t)e−jωt dt + b ∫ ​ y(t)e−jωt dt = aX(ω) + bY (ω)
−∞ −∞

Thus,

Z(ω) = aX(ω) + bY (ω) ​

This shows that the Fourier transform is a linear operator.

[Link] Time Shift

If

FT
x(t) ​
X(ω)

then,
 FT
y(t) = x(t − t0 ) ​ ​ Y (ω) = e−jωt0 X(ω) ​

(4.1.4)
Meaning:

A time shift of t0 in the time domain corresponds to a phase shift of −ωt0 in the frequency domain.
​ ​

The magnitude of the spectrum remains unchanged.

The phase is altered by a linear term −ωt0 . ​

Proof:

By definition of Fourier transform:

∞ ∞
Y (ω) = ∫ ​
y(t)e−jωt dt = ∫ ​ x(t − t0 )e−jωt dt​

−∞ −∞

Let τ = t − t0 . Then t = τ + t0 , and dt = dτ . Since the limits are from −∞ to ∞, they remain unchanged.
​ ​

So:
∞ ∞
Y (ω) = ∫ ​ x(τ )e−jω(τ +t0 ) dτ = ∫ ​

​ x(τ )e−jωτ ⋅ e−jωt0 dτ ​

−∞ −∞

Factor out the constant:

∞
= e−jωt0 ∫ ​

​
x(τ )e−jωτ dτ = e−jωt0 X(ω) ​

−∞

Thus,

Y (ω) = e−jωt0 X(ω) ​

This shows that time delay introduces a linear phase shift in the frequency domain.

[Link] Frequency Shift

If

FT
x(t) ​
X(ω)

then,

FT
y(t) = ejβt x(t) ​ Y (ω) = X(ω − β) (4.1.5)
Meaning:

Multiplying a signal by a complex exponential ejβt in the time domain results in a shift of the frequency spectrum by β radians per
second.
This is equivalent to shifting the frequency axis by β .

Proof:
∞ ∞ ∞
Y (ω) = ∫ ​ y(t)e−jωt dt = ∫ ​ [ejβt x(t)] e−jωt dt = ∫ ​ x(t)e−j(ω−β)t dt
−∞ −∞ −∞

But this is exactly the definition of the Fourier transform of x(t) evaluated at ω − β:

= X(ω − β)

Thus,
 Y (ω) = X(ω − β) ​

This is known as the modulation property: frequency shifting in the frequency domain corresponds to multiplication by a complex
exponential in the time domain.

[Link] Time Scaling

If

FT
x(t) ​
X(ω)

then,

1 ω
X( )
FT
y(t) = x(at) Y (ω) = (4.1.6)
∣a∣
​ ​

a
Meaning:

Compression of a signal in the time domain (when a > 1) results in expansion of its spectrum in the frequency domain.
Expansion in time (when 0 < a < 1) leads to compression in frequency.
1
The magnitude of the spectrum is scaled by ∣a∣ to preserve energy.
​

This property reflects the inverse relationship between time and frequency domains.

Proof:

By definition of Fourier transform:

∞ ∞
Y (ω) = ∫ ​ y(t)e−jωt dt = ∫ ​ x(at)e−jωt dt
−∞ −∞

Let τ = at. Then t = τa , and dt = a1 dτ

​ ​

Since a can be positive or negative, we must account for the absolute value when changing limits.

But since the integration is over (−∞, ∞), and τ runs from −∞ to ∞ regardless of the sign of a, we have:
∞
1 dτ dτ
Y (ω) = ∫ x(τ )e−jω(τ /a) ⋅ dτ (Note: dt = , but ∣dt∣ = )
∣a∣ ∣a∣
​ ​ ​ ​

−∞ a

So:
∞
1 1 ω
Y (ω) = ∫ x(τ )e−j(ω/a)τ dτ = X( )
∣a∣ −∞ ∣a∣
​ ​ ​ ​

Thus,

1 ω
Y (ω) = X( )
∣a∣
​ ​ ​

This shows that time scaling affects both magnitude and frequency scaling of the spectrum.

[Link] Frequency Differentiation

If

FT
x(t) ​ X(ω)

then,
 FT d
−jt x(t) ​ ​
X(ω) (4.1.7)
dω
Meaning:

Differentiating the frequency-domain representation X(ω) corresponds to multiplying the time-domain signal x(t) by −jt.

This is useful in analyzing systems where derivatives of spectra are involved.

Proof:

We start with the definition of Fourier transform:

∞
X(ω) = ∫ ​ x(t)e−jωt dt
−∞

Now differentiate both sides with respect to ω :

∞ ∞
d d −jωt
​X(ω) = ∫ x(t) [e
​ ] dt = ∫ x(t) (−jt e−jωt ) dt
​ ​

dω −∞ dω −∞
∞
= −jt ∫ ​ x(t)e−jωt dt = −jt X(ω)
−∞

Therefore,

d FT d
​X(ω) = −jt X(ω) ⇒ −jt x(t) X(ω)
​ ​

dω dω

This property links time-domain multiplication by −jt to frequency-domain differentiation.

[Link] Time Differentiation(Reading Assignment)

[Link] Convolution
If

FT FT
x(t) ​ X(ω) and y(t) ​ Y (ω)

then,

FT
z(t) = x(t) ∗ y(t) ​ Z(ω) = X(ω)Y (ω) (4.1.9)
Meaning:

Convolution in the time domain corresponds to multiplication in the frequency domain.

This is one of the most important properties of the Fourier transform and is fundamental in system analysis.

Proof:

By definition of Fourier transform:

∞ ∞
Z(ω) = ∫ ​
z(t)e−jωt dt = ∫ ​
[x(t) ∗ y(t)] e−jωt dt
−∞ −∞

Recall convolution:
∞
x(t) ∗ y(t) = ∫ ​ x(τ )y(t − τ ) dτ
−∞

Substitute into Z(ω):

 ∞ ∞
Z(ω) = ∫ ​ (∫ ​ x(τ )y(t − τ ) dτ ) e−jωt dt
−∞ −∞

Interchange order of integration:

∞ ∞
=∫ ​ x(τ ) (∫ ​
y(t − τ )e−jωt dt) dτ
−∞ −∞

Let α = t − τ . Then t = α + τ , dt = dα. Limits remain (−∞, ∞).

So:
∞ ∞ ∞ ∞
=∫ ​ x(τ ) (∫ ​
y(α)e−jω(α+τ ) dα) dτ = ∫ ​ x(τ )e−jωτ (∫ ​
y(α)e−jωα dα) dτ
−∞ −∞ −∞ −∞
∞ ∞
= (∫ ​ x(τ )e−jωτ dτ ) ⋅ (∫ ​ y(α)e−jωα dα) = X(ω) ⋅ Y (ω)
−∞ −∞

Thus,

Z(ω) = X(ω)Y (ω) ​

This confirms that convolution in time domain becomes multiplication in frequency domain.

[Link] Integration (Reading Assignment)

[Link] Modulation
If

FT FT
x(t) ​ X(ω) and y(t) ​ Y (ω)

then,

FT 1
z(t) = x(t)y(t) Z(ω) = [X(ω) ∗ Y (ω)] (4.1.11)
2π
​ ​

Meaning:
1
Multiplication (modulation) in the time domain corresponds to convolution in the frequency domain, scaled by 2π . ​

This is the dual of the convolution property:

Time convolution → Frequency multiplication

Time multiplication → Frequency convolution

Proof:

By definition of Fourier transform:

∞ ∞
Z(ω) = ∫ ​ z(t)e−jωt dt = ∫ ​
x(t)y(t)e−jωt dt (4.1.12)
−∞ −∞

Now use the inverse Fourier transform of x(t):

∞
1
x(t) = ∫ X(λ)ejλt dλ
2π −∞
​ ​

Substitute into Equation (4.1.12):

∞ ∞
1
Z(ω) = ∫ ( ∫ X(λ)ejλt dλ) y(t)e−jωt dt
2π −∞
​ ​ ​

−∞
Interchange integrals:
∞ ∞
1
= ∫ X(λ) (∫ y(t)e−j(ω−λ)t dt) dλ
2π −∞
​ ​ ​

−∞

The inner integral is the Fourier transform of y(t) evaluated at ω − λ:

∞
1 1
= ∫ X(λ)Y (ω − λ) dλ = [X(ω) ∗ Y (ω)]
2π −∞ 2π
​ ​ ​

Thus,

1
Z(ω) = [X(ω) ∗ Y (ω)]
2π
​ ​

1
This confirms that time-domain multiplication results in frequency-domain convolution, scaled by 2π . ​

[Link] Duality(Reading Assignment)

[Link] Symmetry(Reading Assignment)

[Link] Parseval’s Theorem or Rayleigh’s Theorem(Reading Assignment)

Example 4.1.1
Obtain the Fourier transform of the signal x(t) = e−at u(t) and plot its magnitude and phase spectrum.

Solution:
Given:

x(t) = e−at u(t)

where a > 0, and u(t) is the unit step function.

We compute the Fourier transform:

∞ ∞
X(ω) = ∫ ​ x(t)e−jωt dt = ∫ ​ e−at u(t)e−jωt dt
−∞ −∞

Since u(t) = 1 for t ≥ 0, the limits become from 0 to ∞:

∞ ∞
X(ω) = ∫ ​ e−at e−jωt dt = ∫ ​ e−(a+jω)t dt
0 0

Integrate:
∞
e−(a+jω)t 1
=[ ] = (since e−∞ = 0)
−(a + jω) 0 a + jω
​ ​ ​

Thus,

1
X(ω) = (4.1.17)
a + jω
​ ​

So,

FT 1
e−at u(t)
a + jω
​ ​
To Obtain Magnitude and Phase Spectrum

We express X(ω) in terms of real and imaginary parts:

1 a − jω a − jω a ω
X(ω) = = = 2 2
= 2 2
−j 2
a + jω (a + jω)(a − jω) a +ω a +ω a + ω2
​ ​ ​ ​ ​

Magnitude Spectrum:

2 2
a2 + ω 2 1
( 2 ) ( )
a ω
∣X(ω)∣ = + = =
​

a + ω2 a2 + ω 2 a2 + ω 2 a2 + ω2
​ ​ ​ ​ ​

So,

1
∣X(ω)∣ =
a2 + ω2
​ ​

Phase Spectrum:

−ω/(a2 + ω 2 )
∠X(ω) = tan−1 ( ) = tan−1 (− ) = − tan−1 ( )
ω ω
a/(a2 + ω 2 )
​ ​ ​

a a

So,

ω
∠X(ω) = − tan−1 ( ) ​ ​

a
Plots

For a = 1, we have:

Magnitude Spectrum:

1
∣X(ω)∣ =
1 + ω2
​

→ Decreases as ∣ω∣ increases, peaks at ω =0

Phase Spectrum:

∠X(ω) = − tan−1 (ω)

→ Ranges from 0 to − π2 as ω → ∞, and from 0 to π

2
​as ω → −∞

Fig. 5.1.1 shows these plots:

 Magnitude Plot: Symmetric about ω = 0, maximum at ω = 0
Phase Plot: Odd function, continuous, goes from π2 to − π2
​ ​

Summary:

The signal e−at u(t) has a low-pass characteristic.

1
Its spectrum decays as ∣ω∣ at high frequencies.
​

It is widely used in system analysis and filtering.

Fourier Transform of a Periodic Signal (Reading Assignment)

4.1.5 Inverse Fourier Transform

The inverse Fourier transform can be obtained by the basic definition.
It can also be obtained with the help of partial fraction expansions.
Following Fourier transform pairs are used for partial fraction expansions:

FT 1
e−at u(t)
a + jω
​ ​

FT 1
te−at u(t)
(a + jω)2
​ ​

Example 4.1.11
Determine inverse Fourier transform of:

2jω+1
(i) X(jω) = (jω+2) 2 by partial fraction expansions.
​

1
(ii) X(jω) = (a+jω)2 by convolution property.
​
Solution:
(i)

The given function is:

2jω + 1
X(jω) =
(jω + 2)2
​

It can be expanded in partial fractions as:

A1 A2
X(jω) = +
​ ​

jω + 2 (jω + 2)2
​ ​

Values of A1 and A2 are calculated as follows:

​ ​

2jω + 1
A2 = (jω + 2)2 ⋅ X(jω) = (jω + 2)2 ⋅
(jω + 2)2
​ ​ ​ ​ ​ ​

jω=−2 jω=−2

= (2jω + 1) jω=−2
​ ​ = 2(−2) + 1 = −3

d d
A1 =
​ ​[(jω + 2)2 X(jω)]jω=−2 = [2jω + 1] ​ ​ ​ ​ =2
d(jω) d(jω) jω=−2

Thus,

2 3
X(jω) = −
jω + 2 (jω + 2)2
​ ​

Using standard Fourier transform pairs, we obtain:

x(t) = 2e−2t u(t) − 3te−2t u(t)

This is the required time-domain signal.

(ii)

The given function is:

1 1
X(jω) = =
(a + jω) 2 (a + jω)(a + jω)
​ ​

= X1 (jω) ⋅ X2 (jω)
​ ​

where

1 1
X1 (jω) = , X2 (jω) =
a + jω a + jω
​ ​ ​ ​

⇒ x1 (t) = e−at u(t),

​ x2 (t) = e−at u(t)
​

Using convolution property:

x(t) = x1 (t) ∗ x2 (t) ​ ​

∞
=∫ ​ e−aτ u(τ ) ⋅ e−a(t−τ ) u(t − τ ) dτ
−∞

Here, u(τ ) = 0 for τ < 0.

And u(t − τ ) = 1 for τ ≤ t.

So,
 t t
x(t) = ∫ e−aτ e−a(t−τ ) dτ = e−at ∫ dτ = e−at [τ ]t0 = te−at u(t)
​ ​ ​

0 0

Thus, the inverse Fourier transform of

1
(a + jω)2
​

is

x(t) = te−at u(t)

4.2 Discrete Time Fourier Transform (DTFT)

4.2.1 Definition of DTFT

The DTFT is used for the analysis of nonperiodic discrete-time signals.
Consider the discrete-time signal x(n). Its DTFT is denoted as X(Ω) and is given by:
∞
X(Ω) = ∑ x(n)e−jΩn ​ (Analysis Equation) (4.2.1)
n=−∞

Here, Ω is the frequency variable of the discrete-time signals.

The range of Ω is from −π to π , or equivalently 0 to 2π .
Thus, Ω is continuous over this range.
X(Ω) is also called the spectrum of the discrete-time signal.
Even though x(n) is discrete, its spectrum is continuous.
This means x(n) can have all frequencies from −π to π .
The DTFT of periodic signals can also be calculated.
In this case, the spectrum X(Ω) will be discrete, because periodic signals have discrete spectra.

The Inverse DTFT is given by:

π
1
x(n) = ∫ X(Ω)ejΩn dΩ (Synthesis Equation) (4.2.2)
2π −π
​ ​

Key Notes

The discrete-time sequence x(n) can be obtained from X(Ω) using integration.
Since X(Ω) is a continuous function of Ω, the integration range is −π to π .
2π is the period of integration.
x(n) and X(Ω) form a DTFT pair, written as:
DTFT
x(n) ​
X(Ω)

4.2.2 Existence of DTFT

In the definition of the Discrete-Time Fourier Transform (DTFT) (Equation 5.2.1), there is a summation over an infinite range of n.
Therefore, for the DTFT to exist, convergence of this infinite series is required.

The DTFT is defined as:

∞
X(Ω) = ∑ x(n)e−jΩn ​
(4.2.1)
n=−∞

For this sum to converge, one of the following conditions must be satisfied:
i) Absolute Summability
The DTFT exists if x(n) is absolutely summable, i.e.,
∞
∑ ∣x(n)∣ < ∞
​
(4.2.3)
n=−∞

This ensures that the infinite series converges absolutely.

ii) Finite Energy (Square-Summable)

If x(n) is not absolutely summable, it may still have a DTFT if it has finite energy, i.e.,
∞
∑ ∣x(n)∣2 < ∞
​ (4.2.4)
n=−∞

In such cases, the DTFT may exist in the mean-square sense.

Note: The inverse DTFT involves integration over a finite interval (−π, π), so it does not suffer from convergence issues — unlike the
forward transform.

Most physical signals satisfy at least one of these conditions.

4.2.3 Properties of DTFT

Now we discuss the properties of the Discrete-Time Fourier Transform. These are analogous to those of the Continuous-Time Fourier
Transform.

[Link] Periodicity
The DTFT X(Ω) is periodic with period 2π , i.e.,

X(Ω + 2πk) = X(Ω) for any integer k (4.2.5)

Proof:
By definition of DTFT:
∞
X(Ω) = ∑ x(n)e−jΩn ​

n=−∞

Now replace Ω by Ω + 2πk :

∞ ∞
X(Ω + 2πk) = ∑ x(n)e−j(Ω+2πk)n = ∑ x(n)e−jΩn ⋅ e−j(2πk)n
​ ​ (4.2.6)
n=−∞ n=−∞

Now,

e−j(2πk)n = cos(2πkn) − j sin(2πkn)

Since k and n are integers:

cos(2πkn) = 1
sin(2πkn) = 0

Hence,

e−j(2πk)n = 1

So,
 ∞
X(Ω + 2πk) = ∑ x(n)e−jΩn = X(Ω)
​

n=−∞

Thus,

X(Ω + 2πk) = X(Ω) ​

This periodicity arises because e−jΩn is periodic in Ω with period 2π .

[Link] Linearity
This property states that the DTFT is a linear transform.

If
DTFT DTFT
x(n) X(Ω) and y(n) ​
Y (Ω)

then,

DTFT
z(n) = ax(n) + by(n) ​ Z(Ω) = aX(Ω) + bY (Ω) (4.2.7)

Proof:
By definition of DTFT:
∞
Z(Ω) = ∑ z(n)e−jΩn ​

n=−∞

Substitute z(n) = ax(n) + by(n):

∞ ∞ ∞
Z(Ω) = ∑ [ax(n) + by(n)]e −jΩn
= a ∑ x(n)e ​
−jΩn
+ b ∑ y(n)e−jΩn
​

n=−∞ n=−∞ n=−∞

= aX(Ω) + bY (Ω)

Thus,

Z(Ω) = aX(Ω) + bY (Ω) ​

This shows that the DTFT satisfies the superposition principle: linear combinations in time domain map to linear combinations in
frequency domain.

[Link] Time Shift

This property states that a time shift in the sequence results in a phase shift in the frequency domain.

If

DTFT
x(n) ​ X(Ω)

then,

DTFT
y(n) = x(n − n0 ) ​ ​ Y (Ω) = e−jΩn0 X(Ω) ​

(4.2.8)

Proof:
By definition of DTFT:
 ∞ ∞
Y (Ω) = ∑ y(n)e−jΩn = ∑ x(n − n0 )e−jΩn
​ ​ ​

n=−∞ n=−∞

Let m = n − n0 . Then n = m + n0 , and as n ranges from −∞ to ∞, so does m.

​ ​

So:
∞ ∞
Y (Ω) = ∑ x(m)e ​
−jΩ(m+n0 ) ​

= ∑ x(m)e−jΩm ⋅ e−jΩn0 ​

m=−∞ m=−∞

Factor out the constant:

∞
=e −jΩn0 ​

∑ x(m)e−jΩm = e−jΩn0 X(Ω)

​
​

m=−∞

Thus,

Y (Ω) = e−jΩn0 X(Ω) ​

A time delay of n0 samples introduces a linear phase shift −Ωn0 in the frequency domain.
​ ​

[Link] Frequency Shift

This property states that multiplying a sequence by a complex exponential in the time domain results in a shift in the frequency domain.

If

DTFT
x(n) ​
X(Ω)

then,

DTFT
y(n) = ejΩ0 n x(n)​

​ Y (Ω) = X(Ω − Ω0 ) ​ (4.2.9)

Proof:
By definition of DTFT:
∞ ∞ ∞
Y (Ω) = ∑ y(n)e ​
−jΩn
= ∑ [e ​
jΩ0 n ​

x(n)] e −jΩn
= ∑ x(n)e−j(Ω−Ω0 )n
​
​

n=−∞ n=−∞ n=−∞

But this is exactly the DTFT of x(n) evaluated at Ω − Ω0 : ​

= X(Ω − Ω0 ) ​

Thus,

Y (Ω) = X(Ω − Ω0 ) ​ ​

This shows that frequency shifting in the frequency domain corresponds to modulation by ejΩ0 n in the time domain. ​

[Link] Scaling
Let the discrete-time sequence be scaled as:

y(n) = x(pn), where p is an integer

In this case, information in x(n) is discarded because only values at multiples of p are retained. For example, if p = 2, then y(n) =
x(2n) uses only even-indexed samples.
Hence, the scaling property is meaningful only when:
n
x(n) = 0 for ​/Z
∈
p

i.e., x(n) is zero at non-multiple indices of p. Then x(pn)  0 for all n, and no data is lost.
=

The scaling property is given as:

If

DTFT
x(n) ​ X(Ω)

then,

Ω
Y (Ω) = X ( )
DTFT
y(n) = x(pn) ​ ​ (4.2.10)
p

Proof:
By definition of DTFT:
∞ ∞
Y (Ω) = ∑ y(n)e−jΩn = ∑ x(pn)e−jΩn
​ ​

n=−∞ n=−∞

m
Let m = pn. Then n = p , and since
​
p is an integer, m runs over all integers as n does.

So:
∞ ∞
Ω
Y (Ω) = ∑ x(m)e−jΩ(m/p) = ∑ x(m)e−j(Ω/p)m = X ( )
​ ​ ​

m=−∞ m=−∞
p

Thus,

Ω
Y (Ω) = X ( ) ​ ​

This means that expanding in time domain (i.e., inserting zeros between samples) corresponds to compressing the spectrum in
frequency domain.

Note: This is valid only if x(n) has support such that x(pn) retains all necessary information.

[Link] Differentiation in Frequency Domain

This property states that differentiating the DTFT with respect to Ω corresponds to multiplying the time-domain sequence by −jn.

If

DTFT
x(n) ​ X(Ω)

then,

DTFT d
−jnx(n) ​ X(Ω) ​
(4.2.11)
dΩ
Proof:
By definition of DTFT:
∞
X(Ω) = ∑ x(n)e−jΩn ​

n=−∞

Differentiate both sides with respect to Ω:

∞ ∞
( ∑ x(n)e−jΩn ) = ∑ x(n) ⋅
d d d
X(Ω) =
​ ​ (e−jΩn )
​ ​ ​

dΩ dΩ n=−∞ n=−∞
dΩ
∞ ∞
= ∑ x(n) ⋅ (−jn)e
​
−jΩn
= ∑ [−jnx(n)]e−jΩn ​

n=−∞ n=−∞

This is the DTFT of −jnx(n). Therefore,

DTFT d
−jnx(n) ​ ​X(Ω) ​

dΩ

This property links frequency-domain differentiation to time-domain multiplication by −jn.

[Link] Time Reversal

This property states that reversing a sequence in time results in a reversal of its spectrum.

If

DTFT
x(n) ​
X(Ω)

then,

DTFT
y(n) = x(−n) Y (Ω) = X(−Ω) (4.2.12)

Proof:
By definition of DTFT:
∞ ∞
Y (Ω) = ∑ y(n)e−jΩn = ∑ x(−n)e−jΩn
​ ​

n=−∞ n=−∞

Let m = −n. Then n = −m, and as n runs from −∞ to ∞, so does m.

So:
∞ ∞
Y (Ω) = ∑ x(m)e−jΩ(−m) = ∑ x(m)ejΩm
​

m=−∞ m=−∞

But this is the complex conjugate of the DTFT of x(m) evaluated at −Ω, or directly:
∞
= ∑ x(m)e−j(−Ω)m = X(−Ω)
​

m=−∞

Thus,

Y (Ω) = X(−Ω) ​
 This shows that time reversal in the time domain corresponds to frequency reversal in the frequency domain.

[Link] Convolution
This property states that convolution in the time domain corresponds to multiplication in the frequency domain.

If

DTFT DTFT
x(n) X(Ω) and y(n) ​ Y (Ω)

then,

DTFT
z(n) = x(n) ∗ y(n) ​ Z(Ω) = X(Ω)Y (Ω) (4.2.13)

Proof:
By definition of DTFT:
∞ ∞
Z(Ω) = ∑ z(n)e−jΩn = ∑ [x(n) ∗ y(n)] e−jΩn
​ ​

n=−∞ n=−∞

Recall the convolution sum:

∞
x(n) ∗ y(n) = ∑ x(k )y (n − k) ​

k=−∞

Substitute into Z(Ω):

∞ ∞
Z(Ω) = ∑ ( ∑ x(k)y(n − k)) e−jΩn
​ ​

n=−∞ k=−∞

Interchange the order of summations:

∞ ∞
= ∑ x(k) ∑ y(n − k)e−jΩn
​ ​

k=−∞ n=−∞

Let m = n − k . Then n = m + k , so:

∞ ∞ ∞ ∞
= ∑ x(k) ∑ y(m)e−jΩ(m+k) = ∑ x(k)e−jΩk ∑ y(m)e−jΩm
​ ​ ​ ​

k=−∞ m=−∞ k=−∞ m=−∞

∞ ∞
= ( ∑ x(k)e−jΩk ) ( ∑ y(m)e−jΩm ) = X(Ω)Y (Ω)
​ ​

k=−∞ m=−∞

Thus,

Z(Ω) = X(Ω)Y (Ω) ​

This is one of the most important properties: time-domain convolution becomes frequency-domain multiplication.

[Link] Multiplication in Time Domain (Modulation)

1
This property states that multiplication in the time domain corresponds to convolution in the frequency domain, scaled by 2π . ​

If

DTFT DTFT
x(n) X(Ω) and y(n) ​ Y (Ω)
then,

DTFT 1
z(n) = x(n)y(n) Z(Ω) = [X(Ω) ∗ Y (Ω)] (4.2.14)
2π
​ ​

Proof:
By definition of DTFT:
∞ ∞
Z(Ω) = ∑ z(n)e ​
−jΩn
= ∑ x(n)y(n)e−jΩn ​
(4.2.15)
n=−∞ n=−∞

Now use the inverse DTFT of x(n):

π
1
x(n) = ∫ X(λ)ejλn dλ
2π −π
​ ​

Substitute into Equation (4.2.15):

∞ π
1
Z(Ω) = ∑ ( ∫ X(λ)ejλn dλ) y(n)e−jΩn
2π −π
​ ​ ​

n=−∞

Interchange summation and integration:

∞
∫ X(λ) ( ∑ y(n)e−j(Ω−λ)n ) dλ
π
1
=
2π −π
​ ​ ​

n=−∞

The term in brackets is the DTFT of y(n) evaluated at Ω − λ:

π
1 1
= ∫ X(λ)Y (Ω − λ) dλ = [X(Ω) ∗ Y (Ω)]
2π −π 2π
​ ​ ​

Thus,

1
Z(Ω) = [X(Ω) ∗ Y (Ω)]
2π
​ ​

1
This shows that time-domain multiplication results in frequency-domain convolution, scaled by 2π . ​

[Link] Parseval’s Theorem

Parseval’s theorem states that the total energy of a discrete-time signal can be computed either in the time domain or in the frequency
domain.

If
DTFT
x(n) ​ X(Ω)

then, the energy of the signal is given by:

π
1
E= ∫ ∣X(Ω)∣2 dΩ (4.2.16)
2π −π
​ ​

Proof:
We know that the energy of a discrete-time signal is defined as:
 ∞
E = ∑ ∣x(n)∣2 ​

n=−∞

Since ∣x(n)∣2 = x(n)x∗ (n), we can write:

∞
E = ∑ x(n)x∗ (n) ​

n=−∞

From the inverse DTFT (Equation 5.2.2):

π
1
x(n) = ∫ X(Ω)ejΩn dΩ
2π −π
​ ​

Similarly,
π
1
x∗ (n) = ∫ X ∗ (Ω)e−jΩn dΩ
2π −π
​ ​

Substitute into the energy expression:

∞ ∞ π π
1 1
E = ∑ x(n)x (n) = ∑ (∗
∫ X(Ω)ejΩn dΩ) ( ∫ X ∗ (λ)e−jλn dλ)
2π −π 2π −π
​ ​ ​ ​ ​

n=−∞ n=−∞

Interchange summation and integration:

∞
X(Ω)X ∗ (λ) ( ∑ ej(Ω−λ)n ) dΩ dλ
π π
1
= ∫ ∫
(2π)2 −π −π
​ ​ ​ ​

n=−∞

∞
The sum ∑n=−∞ ej(Ω−λ)n is a Dirac comb, which equals 2πδ(Ω − λ). So:
​

π π π
1 1
E= ∫ ∫ X(Ω)X ∗
(λ) ⋅ 2πδ(Ω − λ) dΩ dλ = ∫ X(Ω)X ∗ (Ω) dΩ
(2π)2 −π −π 2π −π
​ ​ ​ ​

π
1
= ∫ ∣X(Ω)∣2 dΩ
2π −π
​ ​

Thus,

∞ π
1
E = ∑ ∣x(n)∣2 = ∫ ∣X(Ω)∣2 dΩ
2π −π
​ ​ ​ ​

n=−∞

This confirms that energy is preserved under the DTFT transformation.

Example 4.2.1
Determine the Fourier transform of x(n) = an u(n) for −1 < a < 1.

Solution:
Given:

x(n) = an u(n), where − 1 < a < 1

We need to compute the Discrete-Time Fourier Transform (DTFT):

∞
X(Ω) = ∑ x(n)e−jΩn ​

n=−∞
Since u(n) = 1 for n ≥ 0, we have:
∞ ∞
X(Ω) = ∑ a e ​
n −jΩn
= ∑(ae−jΩ )n
​

n=0 n=0

This is an infinite geometric series with common ratio r = ae−jΩ .

For convergence, we require ∣r∣ < 1. Since ∣a∣ < 1 and ∣e−jΩ ∣ = 1, we have:
∞
∣ae−jΩ ∣ = ∣a∣ < 1 ⇒ ∑(ae−jΩ )n converges ​

n=0

The sum of a geometric series is:

∞
1
∑ rn = , for ∣r∣ < 1
1−r
​ ​

n=0

So,

1
X(Ω) = (4.2.19)
1 − ae−jΩ
​

Thus,

DTFT 1
x(n) = an u(n) X(Ω) =
1 − ae−jΩ
​ ​ ​

To Obtain Magnitude and Phase Spectrum

We now compute ∣X(Ω)∣ and ∠X(Ω).

Start with:

1 1 1
X(Ω) = = =
1 − ae −jΩ 1 − a(cos Ω − j sin Ω) (1 − a cos Ω) + ja sin Ω
​ ​ ​

Multiply numerator and denominator by the conjugate of the denominator:

(1 − a cos Ω) − ja sin Ω
=
(1 − a cos Ω)2 + (a sin Ω)2
​

Denominator:

(1 − a cos Ω)2 + a2 sin2 Ω = 1 − 2a cos Ω + a2 cos2 Ω + a2 sin2 Ω = 1 − 2a cos Ω + a2 (cos2 Ω + sin2 Ω) = 1 − 2a cos Ω + a2

So,

1 − a cos Ω − ja sin Ω 1 − a cos Ω a sin Ω

X(Ω) = = −j
1 − 2a cos Ω + a 2 1 − 2a cos Ω + a 2 1 − 2a cos Ω + a2
​ ​ ​

Magnitude Spectrum:

1 − a cos Ω
2
a sin Ω
2 (1 − a cos Ω)2 + a2 sin2 Ω
( ) +( ) =
​

∣X(Ω)∣ =
1 − 2a cos Ω + a2 1 − 2a cos Ω + a2 1 − 2a cos Ω + a2
​ ​ ​ ​

Simplify numerator:

(1 − a cos Ω)2 + a2 sin2 Ω = 1 − 2a cos Ω + a2 cos2 Ω + a2 sin2 Ω = 1 − 2a cos Ω + a2

So,
 1 − 2a cos Ω + a2 1
∣X(Ω)∣ = =
​

1 − 2a cos Ω + a2 1 − 2a cos Ω + a2
​ ​

1
∣X(Ω)∣ =
1 − 2a cos Ω + a2
​ ​

Phase Spectrum:
−a sin Ω a sin Ω
∠X(Ω) = tan−1 ( ) = − tan−1 ( )
1 − a cos Ω 1 − a cos Ω
​ ​

a sin Ω
∠X(Ω) = − tan−1 ( )
1 − a cos Ω
​ ​

This result shows that the DTFT of a right-sided exponential sequence has a magnitude that depends on cos Ω, and a phase that
varies with Ω. For different values of a, the spectrum changes — useful in filter design.

Frequency Response of Discrete-Time and Continuous-Time LTI Systems

4.3 Introduction
There are mainly two applications of Fourier representations:

i) Analysis of the interaction between signals and systems

ii) Numerical evaluation of signal properties or system behavior

The Continuous-Time Fourier Transform (CTFT) and Discrete-Time Fourier Transform (DTFT) are most commonly used for
analysis applications.
The Discrete-Time Fourier Series (DTFS) is mainly used for computational applications.
The DTFS is the only Fourier representation that can be evaluated on a computer. Therefore, DTFS is used extensively in numerical
algorithms.

4.3 Frequency Response of LTI Systems

4.3.1 Frequency Response of CT Systems

The output y(t) of a Linear Time-Invariant (LTI) system is given by convolution:
∞
y(t) = ∫ ​ h(τ )x(t − τ ) dτ
−∞

Let the input be a complex exponential: x(t) = ejωt

Substitute into the convolution equation:

∞ ∞
y(t) = ∫ ​
h(τ )ejω(t−τ ) dτ = ejωt ∫ ​
h(τ )e−jωτ dτ
−∞ −∞

The integral is the Fourier transform of h(τ ):

y(t) = ejωt H(ω) (4.2.1)

Where:
∞
H(ω) = ∫ ​ h(τ )e−jωτ dτ = F{h(t)}
−∞

This shows that when a sinusoidal input ejωt is applied to an LTI system, the output is the same sinusoid multiplied by H(ω), which is
the frequency response of the system.
Thus, H(ω) is called the frequency response of the system.

We can also write:

y(t) = x(t) ∗ h(t) ⇒ For x(t) = ejωt , y(t) = H(ω)ejωt

This means:

The system does not change the frequency of the input

It modifies the amplitude and phase via H(ω)
H(ω) characterizes how the system responds to different frequencies

By the convolution property of the Fourier transform, we can write the output in the frequency domain as:

Y (ω) = X(ω)H(ω) (4.2.2)

where:

X(ω) is the Fourier transform of the input x(t)

H(ω) is the Fourier transform of the impulse response h(t)
Y (ω) is the Fourier transform of the output y(t)

From this, we can solve for the frequency response:

Y (ω)
H(ω) = ​
(4.2.3)
X(ω)

Here, H(ω) represents the frequency response of the LTI continuous-time (CT) system. It is also known as the system transfer
function.

4.3.2 Frequency Response of DT Systems

For discrete-time (DT) systems, the input and output are related by convolution:

y(n) = x(n) ∗ h(n) (4.2.4)

By the convolution property of the Discrete-Time Fourier Transform (DTFT), we can write the above equation in the frequency domain
as:

Y (Ω) = X(Ω)H(Ω) (4.2.5)

where:

X(Ω) is the DTFT of x(n)

H(Ω) is the DTFT of h(n)
Y (Ω) is the DTFT of y(n)

Alternatively, solving for H(Ω):

Y (Ω)
H(Ω) = ​
(4.2.6)
X(Ω)
This is the frequency response of the discrete-time LTI system.

Key Points:
The frequency response H(ω) or H(Ω) characterizes how a system responds to sinusoidal inputs.
For CT systems: H(ω) = F{h(t)}
For DT systems: H(Ω) = DTFT{h(n)}
The frequency response is a complex function that provides both magnitude (gain) and phase (delay) information at each frequency.
Example 4.2.1
The impulse response of a continuous-time system is given as:

1 −t/RC
h(t) = e ​
u(t)
RC
Determine the frequency response and plot its magnitude and phase plots.

Solution:
Take the Fourier transform of the given impulse response:
∞ ∞ ∞
1 −t/RC −jωt 1
H(ω) = ∫ ​ h(t)e −jωt
dt = ∫ ​ e e​ dt = ∫ e−(1/RC+jω)t dt ​

−∞ 0 RC RC 0

Integrate:
∞
1 e−(1/RC+jω)t 1 1 1
= [ ] = ⋅ =
RC −(1/RC + jω) 0 RC 1/RC + jω 1 + jωRC
​ ​ ​ ​ ​ ​

Thus,

1
H(ω) =
1 + jωRC
​ ​

To Determine Magnitude and Phase Response

We now compute ∣H(ω)∣ and ∠H(ω).

Rewrite H(ω):

1 1 − jωRC 1 − jωRC
H(ω) = = =
1 + jωRC (1 + jωRC)(1 − jωRC) 1 + (ωRC)2
​ ​

So,

1
Real part: 1+(ωRC) 2
​

−ωRC
Imaginary part: 1+(ωRC)2 ​

Magnitude Response:
2 2
1 −ωRC 1 + (ωRC)2 1
∣H(ω)∣ = ( ) + ( ) = =
​

1 + (ωRC)2 1 + (ωRC)2 1 + (ωRC)2 1 + (ωRC)2

​ ​ ​ ​ ​

1
∣H(ω)∣ =
1 + (ωRC)2
​ ​

This is the magnitude response — it decreases with increasing ω , indicating a low-pass filter behavior.

Phase Response:
−ωRC
∠H(ω) = tan−1 ( ) = − tan−1 (ωRC)
1
​

∠H(ω) = − tan−1 (ωRC) ​

π
This shows that the phase shifts from 0 to − 2 as ω ​
→ ∞.
Special Case: Let RC =1
Then:

1
∣H(ω)∣ = 1+ω 2
​

∠H(ω) = − tan−1 (ω)

This represents a first-order low-pass filter with cutoff frequency at ωc ​

= 1 rad/sec.

Plots (Fig. 4.2.1)

Magnitude Plot: Starts at 1 at ω = 0, rolls off at −20 dB/decade.
Phase Plot: Starts at 0°, approaches −90∘ as ω → ∞.

Fig. 4.2.1 (a) Magnitude Response (b) Phase Response

(a) Magnitude Response ∣H(ω)∣

The magnitude response is symmetric about ω =0

It peaks at ω= 0 with value ∣H(0)∣ = 1
Decreases monotonically as ∣ω∣ increases
Asymptotically approaches zero as ω → ±∞

This shape indicates a low-pass filter behavior: it passes low frequencies and attenuates high frequencies.

(b) Phase Response ∠H(ω)

The phase response is antisymmetric about ω =0

At ω = 0, ∠H(0) = 0
 For ω > 0, phase decreases from 0 to − π2 ​

π
For ω < 0, phase increases from 0 to + 2 ​

1
The cutoff frequency is at ω = RC , where the phase crosses − π4
​ ​

Observations:
Magnitude symmetry: ∣H(−ω)∣ = ∣H(ω)∣ → even function
Phase antisymmetry: ∠H(−ω) = −∠H(ω) → odd function
The magnitude response is monotonically decreasing for ω > 0
The system exhibits linear phase only if the phase is a linear function of ω , which is not the case here — it's nonlinear but smooth

Conclusion:
This system is a first-order low-pass filter with:

1
Cutoff frequency: ωc ​ = RC
​

Gain: ∣H(0)∣ =1
Phase lag increasing with frequency