Lecture-2

Digital Signal Processing
Topic -02
Dr Kazi Rafiqul Islam

Overview
• Digital Signal Processing System
• Analog to Digital Conversion
• Nyquist–Shannon Sampling Theorem
• Aliasing
• Sampling Effect in Time Domain
• Sampling Effect in Frequency Domain
• Anti Aliasing Filter
• Under-sampling
• Sampling of Band Limited Signals
• Over-sampling
• Digital to Analog Conversion
Analog vs. Digital Signal Processing
Analog input Signal x(t) Analog output Signal y(t)

Analog
Signal Processor
Analog Signal Processing
Analog input Analog output

Signal x(t) Signal y(t)
A/D Digital D/A
converter Signal Processor converter
Digital Signal Processing

3
Typical Digital Signal Processing System
It consists of
• an analog filter called (anti-imaging) filter,
• an analog-to-digital conversion (ADC) unit,
• a digital signal (DS) processor,
• a digital-to-analog conversion (DAC) unit,
• and an analog filter called reconstruction (anti-image) filter.
4
Typical Digital Signal Processing System
5
A/D & D/A Conversion
6
Analog to Digital (A/D) Conversion
• Most signals of practical interest are analog in nature

Examples: Voice, Video, RADAR signals, Transducer/Sensor
output, Biological signals etc
• So in order to utilize those benefits, we need to convert our

analog signals into digital
• This process is called A/D conversion
7
Analog to Digital Conversion
A/D conversion can be viewed as a three step process
8
A/D conversion can be viewed as a three step process
9
Sample & Hold (Sampler)
• Analog signal is continuous in time and continuous in

amplitude.
• It means that it carries infinite information of time and infinite

information of amplitude.
• Analog (continuous-time) signal has some value defined at

every time instant, so it has infinite number of sample points.
10
• It is impossible to digitize an infinite number of points.
• The infinite points cannot be processed by the digital signal

(DS) processor or computer, since they require an infinite
amount of memory and infinite amount of processing power
for computations.
• Sampling is the process to reduce the time information or

sample points.
11
• The first essential step in analog-to-digital (A/D) conversion is

to sample an analog signal.
• This step is performed by a sample and hold circuit, which

samples at regular intervals called sampling intervals.
• Sampling can take samples at a fixed time interval.
• The length of the sampling interval is the same as the

sampling period, and the reciprocal of the sampling period is
the sampling frequency fs.
12
• After a brief acquisition time, during which a sample is

acquired, the sample and hold circuit holds the sample steady
for the remainder of the sampling interval.
• The hold time is needed to allow time for an A/D converter to
generate a digital code that best corresponds to the analog
sample.
• If x(t) is the input to the sampler, the output is x(nT), where T
is called the sampling interval or sampling period.
• After the sampling, the signal is called “discrete time
continuous signal” which is discrete in time and continuous in
amplitude.
13
14
Figure below shows an analog (continuous-time) signal (solid
line) defined at every point over the time axis (horizontal line)
and amplitude axis (vertical line).
Hence, the analog signal contains an infinite number of points.
15
• Each sample maintains its voltage level during the sampling
interval 𝑻 to give the ADC enough time to convert it.
• This process is called sample and hold.
16
Nyquist–Shannon Sampling Theorem
The sampling theorem guarantees that an analogue signal can be

perfectly recovered as long as the sampling rate is at least twice
as large as the highest-frequency component of the analogue
signal to be sampled.
17
18
Examples
19
Example: For the following analog signal, find the Nyquist sampling
rate, also determine the digital signal frequency and the digital
signal
20
Example: Find the sampling frequency of the following signal.
So sampling frequency should be
21
Exercise
Determine the Nyquist sampling rate of a signal

x(t) = 3sin(5000t + 17o)
• In this case, the signal x(t) = 3sin(5000 t + 17o) has a frequency of
5000 Hz. Therefore, according to the Nyquist-Shannon sampling
theorem, the signal should be sampled at a rate of at least 10000
samples per second (or 10 kHz) to avoid aliasing and ensure
accurate representation of the signal.
• Therefore, the Nyquist sampling rate of the given signal is 10 kHz.
22
Aliasing
23
Aliasing
How many hertz can the human eye see?
• Most don't notice unless it is under 50 or 60 Hz.
• Generally, people notice when the frame-rate is less than the

refresh rate of the display.
• Depending on the type of CRT, you couldn't see flicker at 30

Hz or you could still see it at 120 Hz.
24
Aliasing
• When the minimum sampling rate is not respected, distortion
called aliasing occurs.
• Aliasing causes high frequency signals to appear as lower

frequency signals.
• To be sure aliasing will not occur, sampling is always preceded

by low pass filtering.
• The low pass filter, called the anti-aliasing filter, removes all
frequencies above half the selected sampling rate.
25
Aliasing
• Figure illustrates sampling a 40 Hz sinusoid
• The sampling interval between sample points is T = 0.01 second,
and the sampling rate is thus fs = 100 Hz.
• The sampling theorem condition is satisfied
26
Aliasing
• Figure illustrates sampling a 90 Hz sinusoid
• The sampling interval between sample points is T = 0.01 second,
and the sampling rate is thus fs = 100 Hz.
• The sampling theorem condition is not satisfied
27
Aliasing
28
Sampling Effect in Time Domain
Example of Aliasing in the time domain

of various sinusoidal signals ranging
from 10 kHz to 80 kHz with a sampling
frequency Fs = 40 kHz.
29
Time & Frequency Domains
• There are two complementary signal descriptions.
• Signals seen as projected onto time or frequency domains.
30
Time & Frequency Domains
31
Signal & Spectrum
32
Frequency Range of Analog & Digital Signals
• For analog signals, the frequency range is from -∞ Hz to ∞ Hz
• For digital signals, the frequency range is from 0 Hz to Fs/2 Hz
33
Sampling Effect in Frequency Domain
• Sampling causes images of a signal’s spectrum to appear at

every multiple of the sampling frequency fs.
• For a signal with frequency f, the sampled spectrum has

frequency components at kfs ± f
34
35
36
Anti Aliasing Filter
• A signal with no frequency component above a certain
maximum frequency is known as a band-limited signal.
• In our case we want to have a signal band-limited to ½ Fs.
• Some times higher frequency components (both harmonics

and noise) are added to the analog signal (practical signals are
not band-limited).
• In order to keep analog signal band-limited, we need a filter,

usually a low pass that stops all frequencies above ½ Fs.
• This is called an “Anti-Aliasing” filter. 37

Anti Aliasing Filter
• Anti-aliasing filters are analog filters.
• They process the signal before it is sampled.
• In most cases, they are also low-pass filters unless band-pass

sampling techniques are used.
38
Under Sampling
• If the sampling rate is lower than the required Nyquist rate, that
is fS < 2W, it is called under sampling.
• In under sampling images of high frequency signals erroneously

appear in the baseband (or Nyquist range) due to aliasing.
39
Sampling of Band Limited Signals
Signals whose frequencies are restricted to a narrow band of

high frequencies can be sampled at a rate similar to twice the
Bandwidth (BW) instead of twice the maximum frequency.
Fs ≥ BW
40
• While this under-sampling is normally avoided, it can be
exploited.
• For example, in the case of band limited signals all of the
important signal characteristics can be deduced from the copy
of the spectrum that appears in the baseband through
sampling.
• Depending on the relationship between the signal frequencies
and the sampling rate, spectral inversion may cause the shape
of the spectrum in the baseband to be inverted from the true
spectrum of the signal.
41
Figure: Signal recovered

From Nyquist range are
Base band versions of the
Original signal. Sampling rate is
Important to make sure no aliasing
and spectral inversion occurs.
(a) Fs = 80 kHz, signal spectrum

is Inverted in the baseband.
(b) Fs = 100 kHz, the lowest

Frequencies In the signal alias
to the highest frequencies.
(c) Fs = 120 kHz, No spectral

42
Inversion occurs.
Over Sampling
• Oversampling is defined as sampling above the minimum
Nyquist rate, that is, fS > 2fmax.
• Oversampling is useful because it creates space in the

spectrum that can reduce the demands on the analog anti-
aliasing filter.
43
Over Sampling
• In the example below, 2x oversampling means that a low order analog filter is
adequate to keep important signal information intact after sampling.
• After sampling, higher order digital filter can be used to extract the information.
44
Over Sampling
• The ideal filter has a flat pass-band and the cut-off is very sharp,
since the cut-off frequency of this filter is half of that of the
sampling frequency, the resulting replicated spectrum of the
sampled signal do not overlap each other. Thus no aliasing
occurs.
• Practical low-pass filters cannot achieve the ideal
characteristics.
• Firstly, this would mean that we have to sample the filtered
signals at a rate that is higher than the Nyquist rate to
compensate for the transition band of the filter
45
Spectra of Sampled signals
Figure: Signal ‘s Spectra

(i) Over sampled
(ii) Nyquest Rate
(iii) Under Sampled
46
Sampling Low Pass Signals
47
Exercise
Exercise-1: If the 20 kHz signal is under-sampled at 30 kHz, find the aliased
frequency of the signal.
Exercise-2: A voice signal is sampled at 8000 samples per second.

i. What is the time between samples?
ii. What is the maximum frequency that will be recovered from the signal?
Exercise-3: An analog Electromyogram (EMG) signal contains useful

frequencies up to 3000 Hz.
i. Determine the minimum required sampling rate to avoid aliasing.
ii. Suppose that we sample this signal at a rate of 6500 samples/s. what is
the highest frequency that can be represented uniquely at this sampling
rate?
48
Exercise
Exercise-4: Humans can hear sounds at frequencies between 0 and 20 kHz.
What minimum sampling rate should be chosen to permit perfect recovery
from samples?
Exercise-5: An ECG signal is sampled at 250 samples per second.

Exercise-6: An ultrasound signal ranging in frequency from 900 kHz to 900.5

kHz is under-sampled at 200 kHz. If a 200 Hz target appears in the baseband,
what is the actual frequency of the target?
49
Quantizer
• After the sampling, the discrete time continuous signal still

carry infinite information (can take any value) in terms of
amplitude.
• Quantization is the process to reduce infinite information of

the amplitude.
• Quantizer do the conversion of discrete time continuous

valued signal into a discrete-time discrete-value signal.
• The value of each signal sample is represented by a value

selected from a finite set of possible values.
50
Quantizer
• The A/D converter chooses a quantization level for each

analog sample.
• Number of levels of quantizer is equal to L = 2N
• An N-bit converter chooses among 2N possible quantization

levels.
• So 3 bit converter has 8 quantization levels, and 4 bit

converter has 8 quantization levels.
51
The quantization step size or resolution is calculated as:
Δ = Q = R/2N
where
R is the full scale range of the analog signal (i.e. Ymax - Ymin)
N is the number of bits used by the converter
• Resolution of a quantizer is the distance between two

successive quantization levels
• More quantization levels, a better resolution!
• What's the downside of more quantization levels?
The strength of the signal compared to that of the quantization

errors is measured by dynamic range and signal-to-noise ratio.
52
4-bit Quantizer
53
4-bit Quantizer
54
Quantization Error
• The error caused by representing a continuous-valued signal
(infinite set) by a finite set of discrete-valued levels.
• The larger the number of quantization levels, the smaller the

quantization errors.
• The quantization error is calculated as the difference between

the quantized level and the true sample level.
• Most quantization errors are limited in size to half a

quantization step Q or Δ .
55
Quantization Error
• Suppose a quantizer operation given by Q(.) is performed on
continuous-valued samples x[n] is given by Q(x[n]), then the
quantization error is given by
56
• Lets consider the signal which is to be quantized.
In the figure, we can see that there is a difference between the

original signal (Blue Line) and the quantized signal (Red Lines).
This is the error produced while quantization
57
Quantization error can be reduced, however, if the number of
quantization levels is increased as illustrated in the figure
58
Quantization of unipolar data (maximum error = full step)
59
Quantization of unipolar data (maximum error = half step)
60
Example: Analog pressures are recorded using a pressure transducer as
voltages between 0 and 3 V. The signal must be quantized using a 3-bit
digital code. Indicate how the analog voltages will be covered to digital
values.
The quantization step size is

Q = 3 V/23 = 0.375 V
The half of quantization step is

0.1875 V
61
Quantization of bipolar data (maximum error = half step)
62
Three-bit A/D Conversion
63
Dynamic Range
• Quantization errors can be determined by the quantization
step.
• Quantization errors can be reduced by increasing the number
of bits used to represent each sample.
• Unfortunately these errors can not be entirely eliminated and
their combined effect is called quantization noise.
• The dynamic range of the quantizer is the number of levels it

can distinguish in noise.
• It is a function of the range of signal values and the range of
error values, and is expressed in decibels, dB.
𝑅
• 𝐷𝑦𝑛𝑎𝑚𝑖𝑐 𝑅𝑎𝑛𝑔𝑒 = 20𝑙𝑜𝑔2
𝑄
64
Signal-to-Quantization-Noise Ratio
• Provides the ratio of the signal power to the quantization

noise (or quantization error)
• Mathematically,
where
• Px= Power of the signal ‘x’ (before quantization)
• Pq= Power of the error signal ‘xq’
65
66
2 bit Flash ADC
67
Digital-to-Analog (D/A) Conversion
Block Diagram of D/A Conversion
68
• Once digital signal processing is complete, digital-to-analog

(D/A) conversion must occur.
• This process begins by converting each digital code into an
analog voltage that is proportional in size to the number
represented by the code.
• This voltage is held steady through zero order hold until the next
code is available, one sampling interval later.
• This creates a staircase-like signal that contains frequencies
above W Hz.
• These signals are removed with a smoothing analog low pass
filter, the last step in D/A conversion.
69
• In the frequency domain, the high frequency elements

present in the zero order hold signal appear as images, copies
of the original signal spectrum situated around integer
multiples of the sampling frequency.
• The smoothing analog filter removes these images and so is

given the name of Anti-Imaging Filter.
• Only the frequencies in the baseband, between 0 and fS/2 Hz,

remain.
70
Three bit D/A Conversion
71
Comparing Signals in the A/D & D/A Chain
72
Comparing Signals in the A/D & D/A Chain
73
Summary
74
Summary
75
Problem - solution
Exercise-1: If the 20 kHz signal is under-sampled at 30 kHz, find the aliased
frequency of the signal.
If a signal with a frequency of 20 kHz is under-sampled at a rate of 30 kHz, then it is being

sampled at a rate that is less than twice the frequency of the signal. As a result, the signal
will be aliased, and the aliased frequency can be calculated using the formula:
aliased frequency = sampled frequency - (n x signal frequency)
where n is an integer.
In this case, the signal frequency is 20 kHz, and the sampled frequency is 30 kHz. To
determine the aliased frequency, we need to find the value of n that makes the aliased
frequency less than the Nyquist frequency (which is half the sampling frequency, or 15 kHz).
We can do this by rearranging the formula:
aliased frequency = sampled frequency - (n x signal frequency) n = (sampled frequency -
aliased frequency) / signal frequency
If we set the Nyquist frequency as the maximum allowable aliased frequency (15 kHz), then
we can calculate the maximum value of n that will give an aliased frequency below 15 kHz:
n = (30 kHz - 15 kHz) / 20 kHz = 0.75
Since n must be an integer, the actual value of n that will give the largest aliased frequency
below 15 kHz is 0 (because rounding up to 1 would result in an aliased frequency above 15
kHz). Therefore, the aliased frequency is:
aliased frequency = 30 kHz - (0 x 20 kHz) = 30 kHz
So the aliased frequency of the 20 kHz signal under-sampled at 30 kHz is 30 kHz. 76
Problem - solution
Exercise-2: A voice signal is sampled at 8000 samples per second.
The time between samples is the inverse of the sampling rate.
Therefore, in this case, the time between samples is:
time between samples = 1 / sampling rate = 1 / 8000 = 0.000125
seconds
So the time between samples is 0.000125 seconds, or 125
microseconds.
The maximum frequency that can be recovered from a sampled signal
is half the sampling rate, according to the Nyquist-Shannon sampling
theorem. In this case, the sampling rate is 8000 samples per second, so
the maximum frequency that can be recovered from the signal is:
maximum frequency = sampling rate / 2 = 8000 / 2 = 4000 Hz
Therefore, the maximum frequency that can be recovered from the
voice signal sampled at 8000 samples per second is 4000 Hz. Any
frequency component above this will be aliased, resulting in distortion
of the original signal. 77
Problem - solution
Exercise-3: An analog Electromyogram (EMG) signal contains useful frequencies up to
3000 Hz.
i. Determine the minimum required sampling rate to avoid aliasing.
ii. Suppose that we sample this signal at a rate of 6500 samples/s. what is the highest
frequency that can be represented uniquely at this sampling rate?
To avoid aliasing, the signal should be sampled at a rate that is at least
twice the maximum frequency of the signal. In this case, the analog EMG
signal contains useful frequencies up to 3000 Hz, so the minimum required
sampling rate to avoid aliasing is:
minimum sampling rate = 2 x maximum frequency = 2 x 3000 Hz = 6000
samples/s
Therefore, the minimum required sampling rate to avoid aliasing is 6000
samples/s.
If the signal is sampled at a rate of 6500 samples/s, then the highest
frequency that can be represented uniquely is half the sampling rate,
according to the Nyquist-Shannon sampling theorem. Therefore:
highest frequency = sampling rate / 2 = 6500 / 2 = 3250 Hz
So the highest frequency that can be represented uniquely at a sampling
rate of 6500 samples/s is 3250 Hz. Any frequency component above this 78
may be aliased, resulting in distortion of the original signal.
Problem - solution
Exercise-4: Humans can hear sounds at frequencies between 0 and 20 kHz. What
minimum sampling rate should be chosen to permit perfect recovery from samples?
According to the Nyquist-Shannon sampling theorem, in order to

perfectly recover a signal without any loss of information, the sampling
rate should be at least twice the maximum frequency present in the
signal. In this case, the maximum frequency of human hearing is 20
kHz. Therefore, the minimum sampling rate that should be chosen to
permit perfect recovery from samples is:
minimum sampling rate = 2 x maximum frequency = 2 x 20 kHz = 40
kHz
So the minimum sampling rate that should be chosen to permit perfect
recovery from samples of sounds within the human hearing range is 40
kHz. This is why the standard sampling rate for digital audio is typically
44.1 kHz or 48 kHz, which provides a small margin of safety above the
minimum requirement.
79
Problem - solution
Exercise-5: An ECG signal is sampled at 250 samples per second.
The time between samples is the inverse of the sampling rate.

Therefore, in this case, the time between samples is:
time between samples = 1 / sampling rate = 1 / 250 = 0.004 seconds
So the time between samples is 0.004 seconds, or 4 milliseconds.
The maximum frequency that can be recovered from a sampled signal
is half the sampling rate, according to the Nyquist-Shannon sampling
theorem. In this case, the sampling rate is 250 samples per second, so
the maximum frequency that can be recovered from the signal is:
maximum frequency = sampling rate / 2 = 250 / 2 = 125 Hz
Therefore, the maximum frequency that can be recovered from the
ECG signal sampled at 250 samples per second is 125 Hz. Any
frequency component above this will be aliased, resulting in distortion 80
Problem - solution
Exercise-6: An ultrasound signal ranging in frequency from 900 kHz to 900.5 kHz
is under-sampled at 200 kHz. If a 200 Hz target appears in the baseband, what is
the actual frequency of the target?
When a signal is under-sampled, frequencies above the Nyquist frequency fold back and alias to lower
frequencies. In this case, the Nyquist frequency is half of the sampling rate, which is 100 kHz. Since the
ultrasound signal ranges from 900 kHz to 900.5 kHz, which is above the Nyquist frequency, it will be
aliased to frequencies between 100 kHz and 200 kHz.
The aliasing can be determined using the following formula:
aliased frequency = original frequency - n x sampling rate
where n is an integer that determines how many times the sampling rate has been folded into the
original frequency.
The 200 Hz target in the baseband corresponds to a frequency of 900.2 kHz in the original signal, which
is above the Nyquist frequency. Therefore, it will be aliased to a lower frequency. We can calculate the
actual frequency of the target using the above formula:
aliased frequency = original frequency - n x sampling rate 900.2 kHz = 900.5 kHz - n x 200 kHz
Solving for n, we get:
n = (900.5 kHz - 900.2 kHz) / 200 kHz = 0.0015
Since n must be an integer, we can round it to the nearest integer to get n = 0. The aliased frequency is
therefore:
aliased frequency = original frequency - n x sampling rate aliased frequency = 900.2 kHz - 0 x 200 kHz =
900.2 kHz
Therefore, the actual frequency of the 200 Hz target in the under-sampled ultrasound signal is 900.2
kHz.
81

Lecture-2

Uploaded by

Copyright:

Available Formats

Lecture-2

Uploaded by

Document Information

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Lecture-2

Uploaded by

Copyright:

Available Formats

Digital Signal Processing

Dr Kazi Rafiqul Islam

Analog input Signal x(t) Analog output Signal y(t)

Analog Signal Processing

Analog input Analog output

Digital Signal Processing

• Most signals of practical interest are analog in nature

• So in order to utilize those benefits, we need to convert our

• This process is called A/D conversion

• Analog signal is continuous in time and continuous in

• It means that it carries infinite information of time and infinite

• Analog (continuous-time) signal has some value defined at

• It is impossible to digitize an infinite number of points.

• The infinite points cannot be processed by the digital signal

• Sampling is the process to reduce the time information or

• The first essential step in analog-to-digital (A/D) conversion is

• This step is performed by a sample and hold circuit, which

• Sampling can take samples at a fixed time interval.

• The length of the sampling interval is the same as the

• After a brief acquisition time, during which a sample is

The sampling theorem guarantees that an analogue signal can be

Example: Find the sampling frequency of the following signal.

So sampling frequency should be

Determine the Nyquist sampling rate of a signal

• Most don't notice unless it is under 50 or 60 Hz.

• Generally, people notice when the frame-rate is less than the

• Depending on the type of CRT, you couldn't see flicker at 30

• Aliasing causes high frequency signals to appear as lower

• To be sure aliasing will not occur, sampling is always preceded

Example of Aliasing in the time domain

• For analog signals, the frequency range is from -∞ Hz to ∞ Hz

• For digital signals, the frequency range is from 0 Hz to Fs/2 Hz

• Sampling causes images of a signal’s spectrum to appear at

• For a signal with frequency f, the sampled spectrum has

• In our case we want to have a signal band-limited to ½ Fs.

• Some times higher frequency components (both harmonics

• In order to keep analog signal band-limited, we need a filter,

• This is called an “Anti-Aliasing” filter. 37

• They process the signal before it is sampled.

• In most cases, they are also low-pass filters unless band-pass

• In under sampling images of high frequency signals erroneously

Signals whose frequencies are restricted to a narrow band of

Figure: Signal recovered

(a) Fs = 80 kHz, signal spectrum

(b) Fs = 100 kHz, the lowest

(c) Fs = 120 kHz, No spectral

• Oversampling is useful because it creates space in the

Figure: Signal ‘s Spectra

Exercise-2: A voice signal is sampled at 8000 samples per second.

Exercise-3: An analog Electromyogram (EMG) signal contains useful

Exercise-5: An ECG signal is sampled at 250 samples per second.

Exercise-6: An ultrasound signal ranging in frequency from 900 kHz to 900.5

• After the sampling, the discrete time continuous signal still

• Quantization is the process to reduce infinite information of

• Quantizer do the conversion of discrete time continuous

• The value of each signal sample is represented by a value

• The A/D converter chooses a quantization level for each

• Number of levels of quantizer is equal to L = 2N

• An N-bit converter chooses among 2N possible quantization

• So 3 bit converter has 8 quantization levels, and 4 bit