16-09-2025

TILL MID SEM

UNIT- I: Introduction: Review of Discrete Time Signals and Systems and z-Transforms, Solution of Difference Equations Using One-sided z-
Transform, Frequency domain Characteristics of LTI Systems, LTI Systems as Frequency-Selective Filters.

UNIT- II: Discrete Fourier Transform (DFT) and Fast Fourier Transform (FFT): Discrete Fourier Transform and its Properties, Divide and
Conquer Approach, Decimation in Time and Decimation in Frequency FFT Algorithms.

DSP
DSP

EEN 15251

| 5TH SEM
| 5TH SEM

9/16/2025
9/16/2025
AFTER MID SEM
Digital Signal Processing

| EE-MNNIT
| EE-MNNIT

UNIT- III: Digital Filter Structure: Describing Equation of digital filter, Structures for FIR Systems: Direct Form Structure, Cascade Form
Structure, Structure for IIR Systems: Direct Form Structures, Cascade Form Structure, Parallel Form Structure and Lattice Structure.

UNIT- IV: Design of Digital Filters: Causality and its Implications, Difference between analog filters and digital filters, FIR filter design
using windows, Design of IIR filters from analog filters using: Approximation of Derivatives, Impulse Invariance and Bilinear
Course Coordinator : Dr Kapil Chauhan Transformation, Frequency transformations
Assistant Professor, Department of Electrical Engineering
Motilal Nehru National Institute of Technology Allahabad, Prayagraj.
1 2

What is Signal ?
What is Signal ?

• The signals are functions of one or more independent variables, containing information
about the behavior or nature of some phenomenon.
DSP

DSP
| 5TH SEM

| 5TH SEM

1. Sound Wave ( Audio Voice) 2.Vibration 3.Voltage/ current / frequency

9/16/2025

9/16/2025
| EE-MNNIT

| EE-MNNIT

3. Motion of a vehicle 4. ECG of Cardiovascular System 5. Image

3 4
 16-09-2025

• defined at each point of time;

Type of Signal : (based on characteristics of the independent variable) • can take any only of amplitude
1. Continuous time-continuous amplitude signal (Analog Signal)
2. Discrete time- continuous amplitude (Sampled Signal)
3. Continuous time – discrete amplitude • defined at specific points of time;
DSP

DSP
• can take any only of amplitude
4. Discrete time-discrete amplitude (Digital Signal) • Sampling Rate & Sampling Frequency
| 5TH SEM

| 5TH SEM
9/16/2025

9/16/2025
| EE-MNNIT

| EE-MNNIT
• defined at each point of time;
• Continuous : can take any value between −∞ 𝑡𝑜 + ∞ • can take only specific value of
amplitude
• Discrete : Takes specific values as per sampling frequency or quantized rules

• defined at specific point of time;

• can take only specific value of
amplitude

5 6

• How to decide “Sampling Frequency” ?

Analog-to-Digital Conversion
 Maximum Sampling Frequency
1. Draw 50 Hz Sinusoidal Signal
2. Draw Sampled Signal at 400 Hz
3. Draw Sampled Signal at 800 Hz
DSP

DSP
| 5TH SEM

| 5TH SEM

• Increasing sampling frequency can improve resolution but increases computational cost and memory
9/16/2025

9/16/2025
Sampler Quantizer Coder • Maximum Sampling Frequency (Practical Constraint)
| EE-MNNIT

| EE-MNNIT

• There is no theoretical upper limit, but practically it is limited by:

• a. Hardware constraints
• ADC speed

• Memory and processing power

• Higher sampling → more data to store and process

 What will happen if a signal is sampled at a sampling frequency more than a hardware can support ?

7 8
 16-09-2025

• How to decide “Sampling Frequency” ? • Aliasing : Alias means faking someone else identity.
 Minimum Sampling Frequency
1. Draw 50 Hz Sinusoidal Signal
Sampling as per Shannon
2. Draw Sampled Signal at 400 Hz Re-construction
Nyquist Theorem
3. Draw Sampled Signal at 100 Hz
DSP

DSP
SHANNON–NYQUIST THEOREM
| 5TH SEM

| 5TH SEM
"To perfectly reconstruct a continuous-time (analog) signal from its samples, it must be sampled at a rate at

9/16/2025

9/16/2025
least twice its highest frequency component.“
| EE-MNNIT

| EE-MNNIT
𝑓 ≥ 2𝑓 Alias of original signal
Sampling as per below
Re-construction
fs/2
fs = 2fm is called as Nyquist Rate.

What will happen if a signal is sampled at a sampling frequency less than 2fm ?

9 10

• Alias Frequencies • What is Processing ?

𝑓 = 𝑓 − 𝑘𝐹
• Processing refers to the mathematical operation of the signal (of the information present in
𝑘 = 1, 2, 3, … 𝑎𝑛𝑑 𝑠𝑜 𝑜𝑛
the signal) to modify or improve it in some way.
• Select k such that 𝑓 falls in between 0 to
DSP

DSP
• A signal has multiple alias.
| 5TH SEM

| 5TH SEM

• In case of under-sampling, the reconstructed signal (time-domain) always take lowest

9/16/2025

9/16/2025
frequency. Calculate alias frequency for following cases-
| EE-MNNIT

| EE-MNNIT

1. A 200 Hz signal is sampled at 600 Hz

2. A 200 Hz signal is sampled at 300 Hz • Processing always be done by a processor.
3. A 1300 Hz signal is sampled at 600 Hz.
4. A 1400 Hz signal is sampled at 200 Hz. • Why do we do the processing ?

5. A 1400 Hz signal is sampled at 400 Hz.

11 12
 16-09-2025

Why do we do the processing ? To get desirable information from the signal

Analog Signal Processing

• Filtering: Removing unwanted parts of the signal, such as noise, or extracting useful parts of the signal. Filters can be low-
pass, high-pass, band-pass, or band-stop, depending on the frequencies they allow through or block.

• Transformation: Converting a signal from one form to another, such as from the time domain to the frequency domain
using Fourier Transform. This helps in analyzing the frequency components of the signal.
DSP

DSP
Signal Communication Processor
Receiver
| 5TH SEM

| 5TH SEM
• Modulation and Demodulation: Changing the signal to transmit it over a medium and then converting it back to its Generator Medium

9/16/2025

9/16/2025
original form. This is used in communication systems.
| EE-MNNIT

| EE-MNNIT
Compression: Reducing the amount of data required to represent a signal, making it easier to store or transmit.
Techniques like JPEG for images or MP3 for audio are examples of compression.
• Detection and Estimation: Extracting information from signals, such as detecting the presence of a signal or estimating
parameters like amplitude, phase, or frequency.

• Signal Generation: Creating signals with desired properties, such as waveforms for testing or simulation purposes.
• Interpolation and Decimation: Changing the sampling rate of a signal. Interpolation increases the sampling rate,
while decimation decreases it. 1. Classroom-wide 2. City wide announcement 3. Country-wide announcement
13
10m few kms Thousands kms 14

 Why normal human voice can’t reach far ? Analog Signal Processing: Radio Broadcasting
• Air Resistance ; can’t bend at wall corner, can’t travel in vacuum
DSP

 Radio Waves / Electromagnetic Waves DSP

| 5TH SEM

| 5TH SEM

• kHz ranges
9/16/2025

9/16/2025
• Can travel through vacuum
| EE-MNNIT

| EE-MNNIT

• Can diffract around corners and hills

15 16
 16-09-2025

Amplitude Modulation of Signal Frequency Modulation of Signal

DSP

DSP
| 5TH SEM

| 5TH SEM
9/16/2025

9/16/2025
| EE-MNNIT

| EE-MNNIT
Modulator Modulator

Q: Draw AM signal waveform

Q: Draw FM signal waveform

17 18

• Tuning Circuit
• By adjusting the capacitor (using a tuning knob), the resonant frequency fo changes, allowing the
circuit to tune into different radio stations.
DSP

DSP
| 5TH SEM

| 5TH SEM
9/16/2025

9/16/2025
| EE-MNNIT

| EE-MNNIT

19 20
 16-09-2025

Limitation of Analog Signal Processing : Limitation of Analog Signal Processing :

1. Lack of Flexibility 2. Lower Accuracy and Precision

DSP
DSP

| 5TH SEM
| 5TH SEM

9/16/2025
9/16/2025

| EE-MNNIT
| EE-MNNIT

• Once an analog circuit is designed, its difficult to modify or reconfigure.

• Tuning -- > manual hardware change

• Analog systems are sensitive to noise, temperature, and component tolerances.

• Analog components (like resistors, capacitors, op-amps) degrade over time, affecting system
performance.
21 22

Limitation of Analog Signal Processing :

Limitation of Analog Signal Processing : 3. Lack of Security Protocols
3. Data Storage Difficulties
DSP

DSP
| 5TH SEM

| 5TH SEM
9/16/2025

9/16/2025
| EE-MNNIT

| EE-MNNIT

Magnetic Tape Storage and Vinyl Record

• Stored data cannot be easily updated or modified, i.e., difficult to make updates on data
• Low storage density • Analog signals are generally less secure than digital signals because they can be intercepted and understood
more easily. While encryption techniques exist for analog communication, they are not as robust or widely
23 24
implemented as in digital systems
 16-09-2025

Limitation of Analog Signal Processing :

• 3. Lack of Security Protocols
• Noise and Interference: Analog systems are inherently susceptible to noise and interference,
• Analog Signals Are Unprotected which can degrade the quality of the signal. This noise can come from various sources, such
• The Hogwarts letters are like analog signals — plain, visible, and readable by anyone in the path.
as electrical equipment, atmospheric conditions, or cosmic interference. Unlike digital
• In analog systems, signals (e.g., voltage, sound, or electromagnetic waves) are not encrypted and can be
systems, where noise can often be filtered out or corrected, in analog systems, once noise is
intercepted or tampered with easily.
DSP

DSP
introduced, it is challenging to remove without affecting the desired signal.
2. No Authentication
| 5TH SEM

| 5TH SEM
•
• Signal Degradation: Over long distances, analog signals can degrade due to attenuation,

9/16/2025

9/16/2025
• The owl doesn’t check whether Harry actually receives the letter.
distortion, and other factors. This degradation can result in a loss of quality, making it
• Similarly, analog processors do not verify the integrity or authenticity of the data.
difficult to maintain a clear and accurate signal. For example, in AM and FM radio, signal
| EE-MNNIT

| EE-MNNIT
• Any attacker can modify or block the signal, and the system won’t know.
quality decreases with distance from the transmitter.
• 3. No Redundancy or Retransmission
• Once letters are destroyed, there’s no error detection or automatic retransmission. • Lack of Error Detection and Correction :Analog systems lack the sophisticated error
• Analog processors typically lack error correction or recovery mechanisms. detection and correction mechanisms found in digital systems. This means that any errors
introduced during transmission, such as noise or signal degradation, cannot be easily
• While encryption techniques exist for advanced analog communication, they are not as detected and corrected, leading to potential loss of information or reduced quality.
robust or widely implemented as in digital systems.
25 26

Transformation | Laplace, Z, Fourier

• Transformation
 Why Transformation? • X(t) / X[n]
• Transformation of independent variable
“A transformation is a change in perspective for looking at a signal”
SN Transformation Equation Equation (for causal)
DSP
1 Laplace 𝑋 𝑠 = ∫ 𝑥 𝑡 𝑒 𝑑𝑡 𝑋 𝑠 = ∫ 𝑥 𝑡 𝑒 𝑑𝑡
| 5TH SEM

(Pierre-Simon Laplace- 1785, French)

9/16/2025
2 Fourier
𝑋 𝜔 = 𝑥 𝑡 𝑒 𝑑𝑡 𝑋 𝜔 = 𝑥 𝑡 𝑒 𝑑𝑡
(Joseph Fourier – 1807, French)
| EE-MNNIT

3 Z
(Eliahu I. Jury – 1950, American ) 𝑋 𝑧 = 𝑥𝑛𝑧 𝑋 𝑧 = 𝑥𝑛𝑧
(Inspired from Hurewicz work)

• Why to do the (L/Z/F) transformation ?

• How to do it ?
• When to do it ?
27
28
 16-09-2025

• Why to do the (L/Z/F) transformation ?

• Task 1: We need to analyze the transient and steady-state stability of the excitation system
• To make mathematical operation easy (one of the major aim of engineering – makes life simple ! ) in this generator, especially under fault conditions. Also simulate its response to a step
disturbance.
1. Convolution in time domain  Multiplication in s-domain
Task 2: Here’s a noisy voice recording from our Alexa. Design a filter to remove the 60 Hz
DSP

DSP
•
x(t)∗h(t)⟶X(s)⋅H(s)
| 5TH SEM

| 5TH SEM
hum and identify any harmonic distortion in the audio.

9/16/2025

9/16/2025
2. Differential equations  Algebraic equation in s-domain
Task 3: Design a real-time digital PID controller for the robotic arm. The microcontroller
| EE-MNNIT

| EE-MNNIT
•

samples the angle every 1 ms. Make sure the system remains stable.
6 + + 𝑦 𝑡 = 6𝑠 𝑌 𝑠 + 10𝑠𝑌 𝑠 + 𝑌 𝑠
How to proceed using Laplace, Fourier or Z-transform ?

29 30

• Stability Analysis
• Finding responses (impulse response, ramp response,…)
• Laplace-Transform (if continuous system)
• Z-Transform (if discrete system)
DSP

DSP • How many frequency components are present in the signal ?

| 5TH SEM

| 5TH SEM

• Finding a particular frequency component’s magnitude

9/16/2025

9/16/2025
• Removing noise from the signal
| EE-MNNIT

| EE-MNNIT

• Fourier Series (for periodic continuous signal)

• Fourier Transform (for aperiodic continuous signal)
• Discrete Fourier Transform (DFT) (for finite duration discrete signal)
• Discrete Time Fourier Transform (DTFT) (for infinite duration discrete signal)

Solvability*

31 32
 16-09-2025

Fourier Series : An arbitrary function defined in a finite interval can always be represented
as sums of sinusoids.
DSP

DSP
If a signal f(t) is periodic with period T, its Fourier series is:
| 5TH SEM

| 5TH SEM
9/16/2025

9/16/2025
| EE-MNNIT

| EE-MNNIT
33 34
DSP

DSP
| 5TH SEM

| 5TH SEM
9/16/2025

9/16/2025
| EE-MNNIT

| EE-MNNIT

Square Wave

Fourier
Series 8 Level Fourier Decomposition,
Analysis 35 36
 16-09-2025

 Plot Magnitude vs Frequency and Phase vs Frequency  Mag = sqrt(a^2 + b^2) ; Phase = atan2(b,a)

DSP
DSP

| 5TH SEM
| 5TH SEM

9/16/2025
9/16/2025

| EE-MNNIT
| EE-MNNIT

𝐹𝑟𝑒𝑞𝑢𝑒𝑛𝑐𝑦 𝑣𝑒𝑐𝑡𝑜𝑟 = (1: 1: 𝑁)/𝑇 =

37 38

TUTORIAL –I

• Write MATLAB Code

• Draw Square Waveform of 0.5 Hz.
DSP

DSP Why on earth do we bother breaking any signal into sums of sinusoids?
• Calculate its Fourier Components
| 5TH SEM

| 5TH SEM

• Draw Fourier Decomposition for N = 5, 10 and 20 level

9/16/2025

9/16/2025
• Draw approximate cumulative sum signal for N = 5, 10 and 20.
| EE-MNNIT

| EE-MNNIT

39 40
 16-09-2025

DIRICHLET CONDITION
 Why on earth do we bother breaking any signal into sums of sinusoids? • Can you calculate Fourier Series of any signal ?
 Because sinusoids are fundamental building blocks

 Sines and cosines are like alphabets of signals.

 DSP

DSP
Just like any English word (say DIGITAL) can be built from letters A–Z, any signal can be built by combining

| 5TH SEM

| 5TH SEM

9/16/2025

9/16/2025
sinusoids.

| EE-MNNIT
They are simple, predictable, and mathematically well-behaved.
| EE-MNNIT

 To re-create something, its essential to know its building block.

41 42

GI BBS PHE N OMEN A

• Theoretically, Fourier series contains infinite terms (n = 0 to ∞ )
• Practically, it is impossible to do so.
DSP

DSP
| 5TH SEM

| 5TH SEM

FOURIER
9/16/2025

9/16/2025
TRANSFORM
| EE-MNNIT

| EE-MNNIT

-the jump near the discontinuity

43 44
 16-09-2025

• Why a finite length time domain signal has infinite length in Fourier domain Discrete Fourier Transform (DFT)
representation ? • What does it do ?
• How does it work ?
Magnitude vs Frequency plot
DSP

DSP
of different modes
| 5TH SEM

| 5TH SEM
9/16/2025

9/16/2025
DFT
| EE-MNNIT

| EE-MNNIT
Sampled Signal

Phase vs Frequency plot of

different modes

45 46

DFT and IDFT Formulae

DSP

DSP
| 5TH SEM

| 5TH SEM
9/16/2025

9/16/2025
| EE-MNNIT

| EE-MNNIT

• H is conjugate transpose or Hermitian

• k varies row-wise
47
• n varies column-wise 48
 16-09-2025

• Compute the DFT of the four-point sequence: • Compute the DFT of the four-point sequence:

𝑥 𝑛 = (0 1 2 3) 𝑥 𝑛 = (0 1 2 3)

1 1 1 1
1 −𝑗 −1 𝑗
DSP

DSP
𝑊=
1 −1 1 −1
| 5TH SEM

| 5TH SEM
1 𝑗 −1 −𝑗

9/16/2025

9/16/2025
𝑋 = [6 -2+2j -2 -2-2j]
| EE-MNNIT

| EE-MNNIT
|X| and /_X

49 50

TUTORIAL –II
 Fast Fourier Transform (FFT) using MATLAB
 Plot following signal - 𝒙 𝒕 = 𝑨𝒄𝒐𝒔 𝟐𝝅𝒇𝒕 using MATLAB ( A = 1 and f = 50 Hz)
DSP

DSP  Plot continuous in time and discrete in time plots

| 5TH SEM

| 5TH SEM

 Analyze change in sampling frequency (1000 Hz, 3200 Hz, 6400 Hz) on number of
9/16/2025

9/16/2025
samples or signal resolution
| EE-MNNIT

| EE-MNNIT

 Add 3rd , 5th , 7th harmonics in x(t)

 Add noise in x(t) and harmonics signal

51 52
 16-09-2025

• Fast Fourier Transform (FFT) using MATLAB • Fast Fourier Transform (FFT) using MATLAB
DSP

DSP
| 5TH SEM

| 5TH SEM
9/16/2025

9/16/2025
| EE-MNNIT

| EE-MNNIT
53 54

• Fast Fourier Transform (FFT) using MATLAB • Fourier Spectrum with different window length (integer length as 1, 2, and so on)

• Frequency resolution of Fourier Spectrum :

Window Length = 10
DSP

DSP
| 5TH SEM

| 5TH SEM
9/16/2025

9/16/2025
| EE-MNNIT

| EE-MNNIT

Window Length = 2
Window Length = 1

55 56
 16-09-2025

TI ME- FR E QU E NC Y T R ADE O F F
/ H E ISE NBERG U N CERTAINT Y IN D FT
• Statement
𝟏
𝒕𝒘𝒍 𝒇𝒓 ≥
𝟒𝝅
where, t_wl = effective time window length (time resolution) [it is not sampling rate]
TIME-FR EQUEN CY TRADE-OFF
DSP

DSP
f_r = frequency resolution (smallest frequency separation that can be distinguished in the spectrum)
| 5TH SEM

| 5TH SEM
• Short window → good time resolution, poor frequency resolution.

/ HEISENBERG UNCERTAINT Y IN DFT

9/16/2025

9/16/2025
• Long window → good frequency resolution, poor time resolution.

 5 Hz frequency resolution is better than 10 Hz

| EE-MNNIT

| EE-MNNIT
 3 cycle window length is shorter than 4 cycles

Digital Signal Processing

57 58

TI ME- FR E QU E NC Y T R ADE O F F TI ME- FR E QU E NC Y T R ADE O F F

DSP

DSP
| 5TH SEM

| 5TH SEM
9/16/2025

9/16/2025
| EE-MNNIT

| EE-MNNIT

59 60
 16-09-2025

• Why is there two peaks in the Fourier spectrum of

𝒙 𝒕 = 𝟏𝒄𝒐𝒔 𝟐𝝅𝟓𝟎𝒕 ; 𝒔𝒂𝒎𝒑𝒍𝒆𝒅 𝒂𝒕 𝟑𝟐𝟎𝟎 𝑯𝒛 𝒂𝒏𝒅 𝟏 𝒄𝒚𝒄𝒍𝒆 𝒘𝒊𝒏𝒅𝒐𝒘 𝒍𝒆𝒏𝒈𝒕𝒉 ?
• Draw Fourier Spectrum of e(j2pift),

DSP
DSP

| 5TH SEM
| 5TH SEM

9/16/2025
9/16/2025

| EE-MNNIT
| EE-MNNIT

61 62

• Spectral Leakage Issues in DFT • Properties of DFT

• Some false frequency bins appear in the FFT Spectrum due to noise or asynchronous sampling
• Periodicity
• Linearity
• Time reversal
• Circular Time Shift
DSP

DSP
• Circular Frequency Shift
| 5TH SEM

| 5TH SEM
• Complex Conjugate
9/16/2025

9/16/2025
• Circular Convolution
• Circular Correlation
| EE-MNNIT

| EE-MNNIT

• Parseval’s Theorem

• Need of these properties - To make DFT calculations easy !

• Page-480 DSP by John G Proakis

63 64
 16-09-2025

DIVIDE & CONQUER APPROACH • Standard Test Signal

• Impulse

• Step
• Decimation in Time
• Ramp
• Decimation in Frequency
• Sinusoidal
DSP

DSP
| 5TH SEM

| 5TH SEM
As discussed on black-board !

9/16/2025

9/16/2025
| EE-MNNIT

| EE-MNNIT
65 66

Standard Test Signal

• Standard Test Signal
DSP

DSP
| 5TH SEM

| 5TH SEM
9/16/2025

9/16/2025
| EE-MNNIT

| EE-MNNIT

67 68
 16-09-2025

• Standard Test Signal

PARSEVA L'S THEOREM
• Parseval’s Theorem states that the total energy of a signal in the time domain is equal to
the total energy in the frequency domain.

DSP
DSP

| 5TH SEM
| 5TH SEM

9/16/2025
9/16/2025

| EE-MNNIT
| EE-MNNIT

69 70

Z- TRANSFORM TIP FOR MID SEM EXAM

• Practice GATE QUESTIONS from EE and ECE Gate Paper relevant to our Syllabus
• UNIT- I: Introduction: Review of Discrete Time Signals and Systems and z-Transforms, Solution of
Difference Equations Using One-sided z-Transform, Frequency domain Characteristics of LTI Systems, LTI
DSP

DSP
Systems as Frequency-Selective Filters.
| 5TH SEM

| 5TH SEM
9/16/2025

9/16/2025
• UNIT- II: Discrete Fourier Transform (DFT) and Fast Fourier Transform (FFT): Discrete Fourier
Transform and its Properties, Divide and Conquer Approach, Decimation in Time and Decimation in
| EE-MNNIT

| EE-MNNIT

Frequency FFT Algorithms.

As discussed on black-board !

71 72
 16-09-2025

DIGITAL SIGNAL PROCESSING

DSP
| 5TH SEM

Thanks

9/16/2025
| EE-MNNIT

73