Digital Signal Processing

Soma Biswas
Department of Electrical Engineering
IISc, Bangalore
Office: C 320 (EE)
E-mail: soma.biswas@ee.iisc.ernet.in

Class Hours: Tuesday and Thursday 3:45 to 5.15 pm

Class Room: EE 302 (old no)

Signals
Signal: A physical quantity that varies with time, space, or any other independent
variable or variables.
Mathematically, we describe a signal as a function of one or more independent
variables.

()=

()=

(,)=++,
,
However, there are cases where such functional relationships are unknown or too
complicated to be of any practical use.

Digital Signal Processing - Lecture 2

Speech Signals
Cannot be described functionally using such expressions.
Generated by forcing air through the vocal cord.

Digital Signal Processing - Lecture 2

ECG (Electrocardiogram)
ECG provides doctor with information about the condition of the patients heart.

Digital Signal Processing - Lecture 2

EEG (Electroencephalogram)
ECG provides doctor with information about the activity of the brain.

Digital Signal Processing - Lecture 2

1-D, 2-D Signal

2D

1D
-Speech, ECG, EEG all are 1D
Signals
- Functions of a single
independent variable - time

Lenna or Lena (1973). originally cropped from

Nov 1972 issue of Playboy magazine. It is a picture
of Lena Sderberg, a Swedish model, shot by
photographer Dwight Hooker.
- Image function of 2 independent variables
- Obtained by exposing a photographic film to a
scene or an object.

Digital Signal Processing - Lecture 2

Basic Definitions
Amplitude: The value of the signal at a specified value of the independent variable.
Waveform: The variation of the amplitude as a function of the independent variable.
If the independent variable is continuous, it is called continuous-time signal. Typically
for 1-D signal the independent variable is labelled as time. If the independent
variable is discrete, it is known as discrete-time signal.

Analog and Digital Signals

A continuous-time signal with a continuous amplitude is
called an analog signal. E.g., speech
A continuous-time signal with discrete-valued amplitude
is referred to as quantized boxcar signal.
Digital Signal Processing - Lecture 2

Analog and digital signal

A discrete-time signal with continuous-valued amplitude is known as sampled-data
signal.
A discrete-time signal with discrete-valued amplitude represented by finite number
of digits is called digital signal.

Digital Signal Processing - Lecture 2

Digital Processing Systems

Analog signal scan be processed directly by analog systems for changing their characteristics
or extracting some desired information.
Alternate method for processing analog systems: digital signal processing
O/P required is analog: (speech communication) D/A converter required
Radar signals: information extracted like aircraft position and speed can be given in digital
form no need for D/A converter

Analog signal processing

Digital signal processing system

Digital Signal Processing - Lecture 2

Advantages of Digital Signal Processing

Flexibility in reconfiguring the digital signal processing operations simply by
changing the program eg. Programmable digital computer
Reconfiguration of analog systems usually requires redesign of hardware followed
by testing and verification to see that it operates properly.
Better control in accuracy requirements tolerances in analog circuit components
make it difficult to control the accuracy of an analog signal processing system
Signals can be easily stored in media (like tape or disk) so signals become
transportable and can be processed offline.
Complicated mathematical operations are easy to do on digital signal compared to
analog.
Some cases digital is cheaper digital hardware is cheaper and flexibility of
modifications
So DSP is widely used eg in speech processing, signal transmission on telephone
channels, image processing, etc.
Digital Signal Processing - Lecture 2

10

Limitations

Speed of operation of A/D converters and digital signal processors.

Extremely high bandwidth signal requires fast-sampling-rate A/D converters and
fast digital signal processors.
There are analog signals with large bandwidths for which a digital processing
approach is beyond the state of the art of digital hardware.

Digital Signal Processing - Lecture 2

11

Concept of Frequency

()=(+),<<
Subscript a denotes x(t) is an analog signal.
This signal is completely characterised by three parameters. is
the amplitude of the sinusoid, is the frequency in radian per
second (rad/s), and is the phase in radians.
Instead of , we often use the frequency in cycles per second or
hertz (Hz), where =

()=(+),
<<

Digital Signal Processing - Lecture 2

12

Properties of sinusoid

Complex Exponential Signals:

()=(+)

()=(+)=/(+)+/(+)
Digital Signal Processing - Lecture 2

13

Phasor

()=(+)=/(+)+/(+)
Sinusoidal signal obtained by adding 2 equal
magnitude complex conjugate exponential signals
- called phasors.
With time, Phasors rotate in opposite directions
with angular frequency +- Omega rad/sec.
Positive frequency counterclockwise uniform
angular motion
Negative frequency - clockwise uniform angular
motion
-inf < F < inf

Digital Signal Processing - Lecture 2

14

Discrete-time Sinusoid Signals

()=(+),<<
is an integer variable, called sample number. is the
amplitudeof the sinusoid, is the frequencyin radian per sample,
and is the phasein radians.
Instead of , we use the frequency in cycles per sample, where
=
()=(+),
<<

=/6
(=1/12
)
=/3
Digital Signal Processing - Lecture 2

15

Properties of discrete-time sinusoid

A discrete-time sinusoid is periodic only if its frequency is a
rational number. (+)=(), . The smallest value of
(>) for which it is true is called fundamental period.
Discrete-time sinusoids whose frequencies are separated by an
integer multiple of are identical.
Therefore, all sinusoidal sequences, ()=(+), =0,1,2,
where, =0+2, 0 are indistinguishable(i.e., identical).
Any sequence resulting from a sinusoid with a frequency ||>, or ||>1/2, is
identical to a sequence obtained from a sinusoidal signal with frequency ||<.
We call the sinusoid having the frequency ||> an aliasof a corresponding
sinusoid with frequency ||<.

, or 1/21/2, unique and all frequencies ||>, or ||>

1/2, an alias
Digital Signal Processing - Lecture 2

16

Properties cont
The highest rate of oscillation in a discrete-time sinusoid is
attained when = or equivalently =1/2.
Discrete-time sinusoids whose frequencies are separated by an
integer multiple of are identical.

Digital Signal Processing - Lecture 2

17

Properties cont
Negative frequencies for discrete time signals:

()=(+)=/(+)+/(+)

Fundamental range or 02 (01 or 1/2

1/2)

Discrete time sinusoids with frequencies separated by integer multiple of 2 pi are

identical.
Frequencies in any interval
constitute all the existing discrete
time sinusoids.
So frequency range for discrete time sinusoids is finite with duration 2 pi.

Digital Signal Processing - Lecture 2

18

Analog-to-Digital and Digital-to-Analog Conversion

Most signals of practical interest, such as speech, biological signals, seismic signals,
radar signals, sonar signals, and various communication signals such as audio and
video signals are analog.

Digital signal processing system

To convert analog signals into a sequence of numbers having finite precision, we
use analog-to-digital (A/D) conversion, and the corresponding devices are called A/D
Converters (ADCs).

Digital Signal Processing - Lecture 2

19

A/D converter

Sampling: Takes samples at discrete time interval. ()=

(), .
Quantization: converts discrete-time continuous-valued signal into
discrete-time discrete-valued signal. Difference between () and
() is called quantization error.
Coding: Each discrete value of ()is represented by b-bit
binary sequence.
The operations of sampling and quantization can be performed in either order
but, in practice, sampling Digital
is always
performed before quantization.
Signal Processing - Lecture 2

20

D/A converter
In many practical cases (e.g., speech processing) it is desirable to convert the
processed digital signals into analog form.
The process of converting digital signal into an analog signal is called digital-toanalog (D/A) conversion.
D/A converters do some form of interpolation
Zero-order hold/staircase approximation
- Linear interpolation- linearly connect pair of
successive samples
- Quadratic interpolation, etc.

Sampling does not result in a loss of information, nor does it introduces distortion
in the signal if the signal bandwidth is finite.
On the other hand quantization is a noninvertible or irreversible process that
results in signal distortion.
Digital Signal Processing - Lecture 2

21

Uniform or Periodic Sampling

Uniform sampling: ()=(),<<. The time interval
between successive samples is called the sampling period or
sample interval and its reciprocal /= is called the
sampling rate (samples per second) or the sampling frequency
(hertz).

==/
Digital Signal Processing - Lecture 2

22

Sampling
There exists a relationship between the frequency variable (or )
for analog signals and the frequency variable (or ) for discretetime signals.

()=(+)is sampled periodically at a rate =

/
()=()=(+)=(/+)
=/ =
or 1/21/2
/==/or1/2=/2
/2=1/2
Digital Signal Processing - Lecture 2

23

Sampling

Digital Signal Processing - Lecture 2

24

Sampling
Sampling introduces an ambiguity. The highest frequency in a continuous-time
signal that can be uniquely distinguished when such a signal is sampled at a rate
=1
/ is =/2, =.
Higher frequencies will be aliased. The frequency /2(=) is called the
folding frequency.

Digital Signal Processing - Lecture 2

25

Sampling Theorem
Q ) How do we select sampling rate Fs?
Required: information about maximum frequency content of the signal Fmax
Sometimes signal passed through a filter that severely attenuates frequency components
above Fmax
- Highest frequency that can be unambiguously reconstructed is Fs/2
- Any frequency >Fs/2 or <-Fs/2 results in aliases
- To avoid aliasing, Fs is selected so that

Digital Signal Processing - Lecture 2

26

Quantization
The process of converting a discrete-time continuous-amplitude
signal into a digital signal by expressing each sample value as a
finite number of digits is called quantization.
The error introduced in this process is called quantization error or
quantization noise.

()=[()]
()=()()

Digital Signal Processing - Lecture 2

27

Levels increase ->Step size decrease

->quantization error decrease

Digital Signal Processing - Lecture 2

28

Digital Signal Processing - Lecture 2

29

Quantization of sinusoid

Digital Signal Processing - Lecture 2

30

Coding of quantized samples

The coding process in an A/D converter assigns a unique binary
number to each quantization level.
With a word length of bits one can create different binary
numbers.
The number of bits required in the coder is the smallest integer
greater than or equal to where we have levels.
Generally, the higher the sampling speed and the finer the quantization,
the more expensive the device becomes.

Digital Signal Processing - Lecture 2

31

D/A conversion
The task of a D/A converter is to interpolate between samples.
The sampling theorem specifies an optimum interpolation,
however, it is too complicated.
Zero-order hold is the simplest D/A converter.
Additional improvements can be done using linear interpolation or
higher order interpolation.

Digital Signal Processing - Lecture 2

32