CHAPTER 3:

FREQUENCY-DOMAIN REPRESENTATION
OF SIGNALS

Lecture 1: Fourier representation of signals

Lecture 2: Discrete Fourier Transform
Lecture 3: Spectral analysis of signals using DFT
 Why frequency-domain approach?

! Let us know other characteristics of signals, which are not

exhibited on time-domain representation
" E.g.: some voiced sounds are hardly distinguishable if only time-
domain features are used

! Let us know other characteristics of systems, which are not

exhibited on time-domain representation
" Response of a LTI system to a sinusoid is a sinusoid with same
frequency but different amplitude and phase
" Details are presented in next chapter

2
 Basic ideas

! Any signal can be described as a sum or integral of

sinusoidal signals
! Exact form of the representation depends on whether
" signal is continuous-time or discrete-time
" signal is periodic or aperiodic

# Clear understanding of sinusoidal signals are essential

3
 Lecture 1
Fourier representation of signals

$ Outline:
1. Sinusoidal signals and their properties
2. Summary of Fourier series and Fourier transforms
3. Properties of discrete-time Fourier transform (DTFT)

4
Continuous-time sinusoids
Definition

A: amplitude
angular frequency (rad/s)
θ: phase (radians)
F0: frequency (Hz)

5
 Continuous-time sinusoids
Relation with complex exponentials

" Every sinusoidal signal is a sum of two complex exponentials with

the same frequency
" Negative frequencies for mathematical convenience

6
 Continuous-time sinusoids
Properties

" Two sinusoids with different frequencies are different

" Rate of oscillation increases indefinitely w/ increasing frequency

7
 Continuous-time sinusoids
Harmonically related complex exponentials

" Fundamental frequency:

" sk(t) is the kth harmonic (“hài bậc k”)
" Example of a periodic signal composed of three sinusoids with
harmonically related frequencies

8
 Discrete-time sinusoids
Definition
Often obtained by sampling the continuous- time sinusoid

or

" Normalized frequency:

" Normalized angular frequency:

9
 Discrete-time sinusoids
Properties

" x[n] = A cos(2𝜋f0n + θ) is periodic if and only if f0 = k/N (f0 is a

rational number)
! If k and N are a pair of prime numbers, then N is the
fundamental period of x[n]
" Sinusoidal sequences with angular frequencies separated by
integer multiples of 2𝜋 are identical (Periodicity in frequency)
" All distinct sinusoidal sequences have frequencies within an
interval of 2𝜋 radians
# interested frequency ranges:

10
 DT sinusoids
Properties

" Low frequencies (slow

oscillations) at ~ ω0 = k2𝜋
" High frequencies (rapid
oscillations) at ~ ω0 = 𝜋 + k2𝜋

11
 Discrete-time sinusoids
Harmonically related complex exponentials

" For sk[n] to be periodic with fundamental period N, the frequency ωk

should be a rational multiple of 2𝜋, that is, ωk = 2𝜋k/N
" Fundamental frequency: f0 = 1/N
" There are only N distinct harmonics at frequencies fk = k/N (0≤k≤N-1)

12
Frequency variables and units

13
 Lecture 1
Fourier representation of signals

$ Outline:
1. Sinusoidal signals and their properties
2. Summary of Fourier series and Fourier transforms
3. Properties of discrete-time Fourier transform (DTFT)

14
 Basic ideas revisited

! Any signal can be described as a sum or integral of

sinusoidal signals
! Exact form of the representation depends on whether
" signal is continuous-time (CT) or discrete-time (DT)
" signal is periodic or aperiodic

15
 Fourier series for CT periodic signals

! Continuous-Time Fourier Series (CTFS) pair

" Analysis equation: analyzes (“breaks-up”) a periodic signal x(t) into

a set of harmonic components {ckexp(jkΩ0t)}
" Synthesis equation: synthesizes the signal x(t) from its harmonic
components

16
 Fourier series for CT periodic signals

! Spectrum of CT periodic signals

" {ck}: Fourier series coefficients (k = 0,±1,±2,...)
" Spectrum: the plot of ck as a function of frequency F = kF0, which
constitutes a description of the signal in the frequency-domain

|ck|: magnitude spectrum of x(t) ∠ck: phase spectrum of x(t)

17
 Fourier series for CT periodic signals

! Spectrum of CT periodic signals (cont.)

" {ck}: Fourier series coefficients (k = 0,±1,±2,...)
" Parseval’s relation:

# |ck|2: the portion of the average power of signal x(t) that is

contributed by the kth harmonic component
" Power spectrum: the plot of |ck|2 as a function of frequency F = kF0
# the power is distributed at a set of discrete frequencies
# discrete or line spectra with uniform spacing F0 = 1/T0
(F0: fundamental frequency of x(t))

Here is the foundation for F0 estimation on frequency domain

18
 Fourier series for CT periodic signals

! Example: Spectrum of rectangular pulse train

(c0 = Aτ/T0 ) 19
 Fourier series for CT periodic signals
! Example: Spectrum of rectangular pulse train (cont.)

=5(s) 20
 Fourier Transform for CT aperiodic signals

! Continuous-Time Fourier Transform (CTFT) pair

or

" X(j2𝜋F): spectrum of the aperiodic signal x(t) (called Fourier Transform)
" CTFT is of the same nature as CTFS with fund. frequency F0 = 1/T0 → 0
21
 Fourier Transform for CT aperiodic signals
! Transition from the CTFS to CTFT

CTFS of periodic signal x(t) when T0 = 10τ

CTFT of aperiodic signal x(t) when T0 → ∞

22
 Fourier Transform for CT aperiodic signals

! Parseval’s relation for aperiodic signals with finite energy

# total energy of x(t) may be obtained either from the signal itself
or from its spectrum
" |X(j2πF)|2∆F, for a small ∆F, provides the energy of the signal in a
narrow frequency band of width ∆F
" Energy-density spectrum: the plot of |X(j2πF)|2

23
 Fourier Transform for CT aperiodic signals

! Example 1: Spectrum of causal exponential signal

This signal is absolutely integrable if a > 0.

Magnitude and phase spectra:

24
 Fourier Transform for CT aperiodic signals

! Example 1: Spectrum of causal exponential signal (cont.)

x(t) is a real function of t # |X(j2πF)| has even symmetry, ∠X(j2πF) has odd symmetry

25
 Fourier Transform for CT aperiodic signals

! Example 2: Spectrum of rectangular pulse signal

This signal is absolutely integrable for any finite τ. Its spectrum is computed as

Magnitude spectrum (A=2, τ =1) Phase spectrum

26
 Fourier series for DT periodic signals

! Discrete-Time Fourier Series (DTFS) pair

" Analysis equation: analyzes (“breaks-up”) a periodic signal x[n] into

a set of N harmonic components {ckexp(jkω0n)}, where ω0=2𝜋/N
" Synthesis equation: synthesizes the signal x[n] from its N harmonic
components
" N: fundamental period of periodic sequence x[n]

27
 Fourier series for DT periodic signals

! Parseval’s relation
The average power in one period of x[n] can be expressed

" |ck|2: portion of the average power of x[n] that is contributed by its kth
harmonic component
" Power spectrum: The graph of |ck|2 as a function of f = k/N, ω = 2πk/N, or
simply k

28