TN 412: Digital Signal

Processing
5: Frequency Analysis of Signals and
Systems
Introduction
n The Fourier transform is one of the several mathematical tools that is
useful in the analysis and design of LTI systems.
n Another is the Fourier series.
n These signal representations involves the decomposition of the signals in
terms of sinusoidal (complex exponentials) components.
n With such a decomposition, a signal is said to be represented in the
frequency domain.
n For class of periodic signals, such a decomposition is called a Fourier
series.
n For class of finite energy signals, such a decomposition is called a
Fourier transform.
Introduction
n These decomposition 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.
n The process of obtaining the spectrum of a given signal using the basic
mathematical tools is known as frequency or spectral analysis.
n The process of determining the spectrum of a signal in practice based on
actual measurements of a signal is called spectrum estimation.
n Frequency range for discrete-time signals is unique over the interval (−𝜋, 𝜋)
or (0,2𝜋).
n A discrete-time signal of fundamental period N can consist of frequency
components separated by 2𝜋⁄𝑁 radians of 𝑓 = 1⁄𝑁 cycles.
n The Fourier series representation of the discrete periodic signal will contain
at most N frequency components.
Introduction
n Frequency analysis of a signal involves the resolution of the signal into its
frequency (sinusoidal) components.
n The recombination of the sinusoidal components to reconstruct the
original signal is basically Fourier synthesis problem.
n If any component is missing, the reconstructed component will be
different from the original signal.
n Frequency analysis is performed for the purpose of extracting information
from the observed signal.
n Eg: from biological signal, seismic signal, electromagnetic signal
Introduction
n For the biological signals, such as an ECG signal, the analytical tools are
used to extract information relevant for diagnostic purpose.
n For processing any signal for the purpose of measuring parameters, one
must know approximately the range of frequencies contained by the
signal.
The Fourier Series of DT Periodic Signals

𝑒 !"#$%/' 𝑘 = 0,1, , , , 𝑁 − 1

'*+
and is expressed as
𝑥 𝑛 = + 𝑐$ 𝑒 !"#$%/'
$()

v Where the 𝑐$ are the coefficient in the series representation

v To derive the expression for the Fourier coefficients, we use the following formula:
 The Fourier Series of DT Periodic Signals
n The proof for the summation follows the immediately from the
application of the geometric summation formula.

n The expression for Fourier coefficients 𝑐! can be obtained by

The Fourier Series of DT Periodic Signals

n The equation (4.2.7) is often called the discrete-time Fourier series (DTFS)
n The coefficients 𝑐- provides the description of the x(n) in the frequency
domain, in the sense that 𝑐- represents the amplitude and phase associated
with frequency components
The Fourier Series of DT Periodic Signals
n The spectrum of a signal x(n), which is periodic with period N, is
periodic sequence with period N.
n Consequently, any N consecutive samples of the signal or its
spectrum provide a complete description of the signal in the time
or frequency domains.
n Fourier coefficients 𝑐! , when viewed beyond the range
k=1,2,….,N-1, also satisfy a periodicity condition.
The Fourier Series of DT Periodic Signals
The Fourier Series of DT Periodic Signals
The Fourier Series of DT Periodic Signals
 Power Density Spectrum of Periodic Signals

n The average power of a discrete-time periodic signal with

period N is defined as

(a)

n Using 4.2.7 in (a) we obtain

 Power Density Spectrum of Periodic Signals

n Interchanging the order of two summation and make use of

(4.2.8)

n The average power in the periodic signal is the sum of the

powers of individual frequency components.
 Power Density Spectrum of Periodic Signals

n The energy of the sequence over single period is given by

 Power Density Spectrum of Periodic Signals
n Determine the Fourier Series coefficients and the power
spectrum of the periodic signal shown below
Power Density Spectrum of Periodic Signals
Power Density Spectrum of Periodic Signals
 Fourier Transform of DT Aperiodic Signals
n Just as in the case of continuous time aperiodic energy
signals, the frequency analysis of discrete-time aperiodic
finite-energy signals involves a Fourier Transform of time
domain signal.
n The Fourier transform of finite-energy discrete time signal x(n)
is defined as

n 𝑋 𝜔 𝑟𝑒𝑝𝑟𝑒𝑠𝑒𝑛𝑡 𝑡ℎ𝑒 𝑓𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦 𝑐𝑜𝑛𝑡𝑒𝑛𝑡 𝑜𝑓 𝑎 𝑠𝑖𝑔𝑛𝑎𝑙 𝑥 𝑛 .

n 𝑋 𝜔 𝑖𝑠 𝑎 𝑑𝑒𝑐𝑜𝑚𝑝𝑜𝑠𝑖𝑡𝑖𝑜𝑛 𝑜𝑓 𝑥 𝑛 𝑖𝑛𝑡𝑜 𝑓𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦 𝑐𝑜𝑚𝑝𝑜𝑛𝑒𝑛𝑡𝑠
Fourier Transform of DT Aperiodic Signals
 Properties of Fourier transform for discrete time signals
n Fourier transform for aperiodic finite-energy discrete time signal
posses a number of properties that are very useful in reducing the
complexity of frequency analysis problem in many practical
applications.
n Similar properties holds for the Fourier transform of aperiodic
finite-energy continuous-time signals.
n We refer x(n) and X(w) as the Fourier Transform Pairs and are
denoted by

n X(w) is periodic with period 2𝜋.Any interval of length 2𝜋 is

sufficient for specification of spectrum.
Symmetry Properties of Discrete Time Fourier Transform
Properties of Fourier transform for discrete time signals
 Energy Density Spectrum of Aperiodic Signals
n Recall
 Energy Density Spectrum of Aperiodic Signals
n Example
 Energy Density Spectrum of Aperiodic Signals
n Example
LTI systems as frequency selective filters, inverse systems and
deconvolution
 LTI as Frequency Selective Filters

n A filter is used to describe a device that discriminates , according

to some attribute of the objects applied at its input, what passes
through it.
n A linear time-invariant system also performs a type of
discrimination or filtering among the various frequency
components at its input.
n The nature of this filtering action is determined by frequency
response characteristics 𝐻 𝑤 , which depends on the choice of
the system parameters ie coefficients 𝑎! and 𝑏! in the
difference equation.
 LTI as Frequency Selective Filters

n By proper selection of coefficients, we can design frequency-selective

filters that pass signals with frequency components in some bands while
they attenuate signals containing frequency components in other frequency
bands.
n LTI system modifies the input signal spectrum 𝑋(𝜔) according to its
frequency response H(𝜔) to yield an output signal with spectrum Y 𝜔 =
𝐻 𝑤 𝑋(𝑤) .
n H(𝜔) acts as a weighting function or a spectral shaping function to the
different frequency components in the input signal,
n LTI systems can be viewed as frequency selective filters even though it may
not necessarily completely block any or all frequency component.
 LTI as Frequency Selective Filters

n We term filter to describe an LTI system used to perform spectral shaping or

frequency-selective filtering.
n Filtering is used in DSP in the following aspects
q Removal of undesirable noise from desired signals
q Spectral shaping, such as equalization of communication channel
q Signal detection in radars
q Performing spectral analysis of signals
 LTI as Frequency Selective Filters

Ideal Filters:
Ø Low-Pass Filters: Low-Pass Filters are designed to pass low frequencies,
from zero to a certain cut off frequency and to block high frequencies.

ì1 for w Î< 0, w0 > i.e. w Î pass - band

H (e jw
) =í
î0 for w Î (w0 , p > i.e. w Î stop - band

H ( e jw ) Ideal magnitude frequency

response
1

31
0 w p w
 LTI as Frequency Selective Filters

Ø High-Pass Filters: High-pass filters are designed to pass high

frequencies, from a certain cut off frequency to p, and to block low
frequencies.

ì0 for w Î< 0, w0 ) i.e. w Î stop - band

H (e jw
) =í
î1 for w Î< w0 , p > i.e. w Î pass - band

H ( e jw )
Ideal magnitude
frequency response
1

32
0 w0 p w
 LTI as Frequency Selective Filters

Ø Band-Pass Filters: Band-pass filters are designed to pass a certain

frequency range, which does not include zero, and to block other
frequencies.

ì0 for w Î< 0, w1 ) È (w2 , p ) i.e. w Î stop - band

H (e jw
) =í
î1 for w Î< w1 , w2 > i.e. w Î pass - band

H ( e jw )
Ideal magnitude
frequency response
1

33 0 w1 w2 p w
 LTI as Frequency Selective Filters

Ø Band-Stop Filters: Band-stop filters are designed to block

a certain frequency range, which does not include zero, and to
pass other frequencies.

ì1 for w Î< 0, w1 > È< w2 , p ) i.e. w Î pass - band

H (e jw
) =í
î0 for w Î (w1 , w2 ) i.e. w Î stop - band
Ideal magnitude
H (e jw
) frequency response
1

0
34 w1 w2 p w
LTI as Frequency Selective Filters

Ø Multi-band Filters: This type of filters generalizes the previous four

types of filters in that it allows for different gains or attenuations in different
frequency bands. A piecewise –constant multi-band filter is characterized by the
following parameters:
• A division of the frequency range to a finite union of intervals. Some of these
intervals are pass bands, some are stop bands, and the remaining can be
transition bands.
H ( e jw )
Possible ideal
• A desired gain and a 1 magnitude
permitted tolerance for frequency
each pass band. response
• An attenuation threshold
for each stop band. 0
w1 w 2 w3 w 4 w5 w 6 p w
35
 LTI as Frequency Selective Filters

n Notch Filter and Comb Filters : Reading Assignment

 Inverse Systems and Deconvolution

n In many practical applications, we are given an output signal from a system

whose characteristics are unknown and we are asked to determine the input
signal.
n In digital communication, a corrective system which when cascaded with the
channel, yields a replica of the desired transmitted signal is called equalizer.
n In linear system theory, the corrective system is called an inverse system,
because the corrective system has a frequency response which is basically
the reciprocal of the frequency of the system that has caused the distortion.
n Since, the distortive system yields an output y(n) that is the convolution of the
input x(n) with the impulse response h(n), the inverse system operation that
takes y(n) and produces x(n) is called deconvolution.
 Inverse Systems and Deconvolution
n A system is invertible if there is a one-to-one correspondence between its
input and output signals. Knowing the output sequence y(n) of an invertible
system 𝜏, we can uniquely determine its input.
n The inverse system with input y(n) and output x(n) is denoted by 𝜏 ./.

n Implying that
n In z-domain
n If H(z) has a rational system function then
 Inverse Systems and Deconvolution

Inverse System:
v Given an LTI system H(z) the 1
Hi (z ) =
inverse system Hi(z) is given as H(z )
v The cascade of a
G(z) = H(z)Hi (z) = 1
system and its inverse
yields unity g[n] = h[n] * hi [n] = d[n]

v If it exists, the frequency

response of the inverse ( )
Hi e jw =
1
( )
H e jw
system is
v Not all systems have an inverse: zeros cannot be inverted

(
1 - ck z -1 ) (
1 - dk z -1 )
M N
v The inverse of rational
system functions æb ö Õ æa ö Õ
H(z) = ç 0 ÷ k =1
ça ÷ ¾® H (z) = ç 0 ÷ k =1
¾¾ i çb ÷
Õ (1 - d z ) Õ (1 - c z )
N M
39
è 0ø -1 è 0ø -1
k k
k =1 k =1
 Inverse Systems and Deconvolution
n Thus, the zeros of H(z) become the poles of the inverse system, and vice
versa.
n If H(z) is an FIR system, then 𝐻/(z) is an all-pole system or if H(z) is an all-
pole system , then 𝐻/(z) is an FIR system
Inverse Systems and Deconvolution
 Inverse Systems and Deconvolution
n Suppose that excite an unknown LTI system with an input sequence x(n)and
we observe the output sequence y(n).
n From the output sequence, we wish to determine the impulse response of
unknown system.
n The problem is system identification and is solved by deconvolution.

n Analytical solution of deconvolution problem can be obtained with z-transform

 Inverse Systems and Deconvolution
n Hence

n Example
 Inverse Systems and Deconvolution
n Example
 Inverse Systems and Deconvolution
Deconvolution
Example:

45
Questions
Questions
End!!