DIGITAL SIGNAL PROCESSING

Lecture 26,27,28
Frequency Analysis of Signals : Review
4.1-4.2 (text book)
Chapter 3 & 5- SNS Book
YouTube lecture reference: SNS

https://www.youtube.com/watch?v=Ztts-r7XDqQ
 Frequency Analysis of Signals
2

¨ Most signals of practical interest can be decomposed into a sum of sinusoidal

signal components.
¨ Frequency Analysis of a signal involves the resolution of a signal into its sinusoidal
(frequency) components.
¨ For the class of periodic signals, such a decomposition is called a Fourier series.
¨ For the class of finite energy signals, the decomposition is called the Fourier
transform.
¨ These decompositions are extremely important in the analysis of LTI systems
because the response of an LTI system to a sinusoidal input signal is a sinusoid of
the same frequency but of different amplitude and phase. Furthermore, the
linearity property of the LTI system implies that a linear sum of sinusoidal
components at the input produces a similar linear sum of sinusoidal components at
the output, which differ only in the amplitudes and phases from the input sinusoids.
3

¨ Periodic Sequences → Fourier Series

¨ Aperiodic finite energy sequences → Fourier Transform

 Fourier Spectra
4

¨ The plot of Fourier coefficients vs frequency constitutes the Fourier Spectrum.

¨ The spectrum is in general complex valued and can be expressed in polar
form as a magnitude and phase.
̶ A plot of magnitude vs frequency is termed as the magnitude spectrum . The
magnitude spectrum is even symmetric.
̶ A plot of phase vs frequency is termed as the magnitude spectrum . The magnitude
spectrum is odd symmetric.
̶ The magnitude squared spectrum represents the distribution of power/ energy as a
function of frequency and is termed as Power Density Spectrum (PSD)/ Energy
Density Spectrum (ESD) .
̶ According to Parseval’s Relation, the integral of PSD/ESD determines signal
Power/Energy.
 Summary
5
 Observations: Fourier Spectra
6

¨ CT signals have aperiodic spectra

¨ DT signals have periodic spectra
¨ Periodic signals have discrete spectra
¨ Aperiodic signals have continuous spectra
¨ The spectrum of real signals has Hermitian symmetry i.e., magnitude
spectrum is even symmetric and phase spectrum is odd symmetric
FREQUENCY ANALYSIS OF DISCRETE-TIME SIGNALS
4.2
 Discrete Time Fourier Series (DTFS)
8

¨ As a consequence of periodicity of complex exponentials, DTFS pair is

described as

x[n] = å ak e jkw0 n = å ak e
jk ( 2p / N ) n
Synthesis Equation
k =N k =N

1 1
ak =
N
å x[n]e- jkw0n =
k =N N
å x[n]e
k =N
- jk ( 2p / N ) n
Analysis Equation

¨ For example, k could take on the values 𝑘 = 0, 1, … , 𝑁 − 1 , or 𝑘 =

3, 4, … , 𝑁 + 2. In either case, by virtue of periodicity of exponentials,
exactly the same set of complex exponential sequences appears in the
summation on the right-hand side of the synthesis equation.
¨ The synthesis equation is termed as the Discrete-Time Fourier Series (DTFS)
and the coefficients 𝑎! as the Fourier series coefficients.
 Frequency variable: CT vs. DT
9

¨ The Fourier representation of a continuous-time signal can consist of an infinite

number of frequency components, where the frequency spacing between two
successive harmonically related frequencies is 1/𝑇𝑝 with 𝑇𝑝 being the
fundamental period. Since the frequency range for continuous-time signals
extends from−∞ to∞, it is possible to have signals that contain an infinite
number of frequency components.
¨ In contrast, the frequency range for discrete-time signals is unique over the
interval (−𝜋, 𝜋) or (0, 2𝜋). A discrete-time signal of fundamental period 𝑁
can consist of frequency components separated by 2𝜋/𝑁 radians or 𝑓 =
1/𝑁 cycles. Consequently, the Fourier series representation of the discrete-
time periodic signal will contain at most 𝑁 frequency components.
¨ This is the basic difference between the Fourier spectra for continuous-time
and discrete-time periodic signals.
 Example 1
10
¨ Find the discrete-time Fourier series (DTFS) for 𝑥[𝑛] = sin(0.1𝜋𝑛). Sketch
the amplitude and phase spectra.
¨ Practice: Example 4.2.1
 Solution
11
 Spectrum

12
 Fourier Transform for DT Aperiodic Signals
Discrete Time Fourier Transform (DTFT)
13

¨ For a general sequence 𝑥[n] of finite duration such that 𝑥 𝑛 = 0 outside the
range −𝑁1 ≤ n ≤ 𝑁2 the DTFT is given as
¥
X (w ) = å x[n]e
n =-¥
- jw n
Analysis Equation

¨ The plot of Fourier Transform 𝑋(𝜔) will be referred to as the spectrum of 𝑥[𝑛]
because it provides with the information of how 𝑥[𝑛] is composed of complex
sinusoids of different frequencies.
¨ The inverse DTFT or the synthesis equation is given as
1
òp X (w )e
jw n
x[n] = dw Synthesis Equation
2p 2
 Convergence of DTFT for infinite duration
signals
14

¨ The DTFT equations are valid for an extremely broad category of signals including
infinite duration signals provided the condition of convergence is met.
¨ Specifically the analysis equation will converge if following two equations are met:
̶ The sequence converges i.e.
¥

å | x[n] |< ¥
n =-¥

̶ The sequence has finite energy i.e.

¥

å | x[n] | < ¥
n =-¥
2

¨ In contrast to the situation for the analysis equation, there are generally no
convergence issues associated with the synthesis equation since the integral in this
equation is over a finite interval of integration.
 Example 1 :
15

¨ Determine the DTFT of

16
 Spectrum
17
 Example 2: Inverse DTFT (Page 252 book)
18

¨ Determine the sequence 𝑥[𝑛] for the transform shown in figure

ì wc
ïï p n=0
x[n] = í
For n=0 , we have ï sin wc n n¹0
p ïî p n
1
x[0] =
2p òp X (w )dw
-
wc
1
=
2p òw 1dw
- c

1 wc 2w w Given 𝜔# the, sequence can be obtained by

= w -w = c = c
2p c
2p p varying values of n for 𝑥[𝑛]
 Observations from previous example
19

¨ The previous example involved

determining the time domain expression
for a discrete time low pass filter. The
following observations can be made
about 𝑥[𝑛]
̶ Sequence non-causal
̶ Sequence infinite in length
 Example 3 : (Book Example 4.2.4)
20

¨ Finite energy signal, therefore DTFT is possible

21
 ESD and PSD
22

¨ Energy spectral density represents the distribution of the energy of the signal in
the frequency domain. (Aperiodic signals)
2
x X (w ) = X (w )
¨ The power spectral density (PSD) of a signal describes the distribution of signal
power over frequency domain (Periodic signals)
2
S X (w ) = X (w )
¨ Parseval’s Theorem relates signal energy/power in time and frequency domain.
The integral of ESD/PSD over all frequencies determines signal Energy/power

Ex = ò x(t ) = ò x X (w )d w
2

1
ò x(t ) = ò S X (w )d w
2
Px = limT ®¥
2T
 4.2.6 Relationship of the Fourier
Transform to the z-transform
23
4.2.9 Frequency-domain classification of
Signals: The concept of bandwidth
24
Narrowband, Wideband & Ultra-Wideband
25

¨ In the case of a low-frequency/bandpass signal, the terms are defined by the ratio of
bandwidth at -10 dB to center frequency. The -10 dB point represents the spectral
power of a signal at 10 dB lower than its peak power.

Or
¨ The fractional bandwidth as the ratio of signal bandwidth to center frequency.

fH - fL
Bf = 2´ ´100%
fH + fL
Narrowband B f < 1%
Wideband 1% < B f < 20%
Ultra-Wideband B f > 20%
 Properties of DTFT
26
 Some Useful DTFT pairs
27
 Numerical Problems
28

¨ FT of aperiodic sequences : 4.9 (excluding h), 4.14, 4.18, 4.19

¨ FT of periodic sequences: 4.5, 4.6, 4.7 (Note: You have to determine FT
and not Fourier series. FS is the first step in this regard. The method done in
class should be followed)
¨ Properties of FT to be covered : Linearity, Time shifting, Time reversal,
frequency shifting
¨ Tips:
̶ Always decompose cosines and sines in complex exponentials and apply
properties to determine transform.
̶ For determination of values at a set of frequencies combining exponentials to
form sinusoids makes calculation easier
 Problem 4.6
29

Done in class, Solutions shared as well

 Problem 4.18
30

[ 1.0000 0.7159 -0.0000 -0.7936 -1.2361 -1.0000 0.0000 1.5575 3.2361 4.5201 5.0000
4.5201 3.2361 1.5575 0.0000 -1.0000 -1.2361 -0.7936 -0.0000 0.7159 1.0000]

Done in class, Solutions shared as well

 Practice Examples
31

¨ Fourier Series
¨ Example 4.1.1, 4.2.1, 4.2.2 (Textbook)
¨ Example 3.10, 3.11 (SNS Book)
¨ Fourier Transform
¨ Example 4.2.3, 4.2.4