0% found this document useful (0 votes)
75 views

Lecture 05 - Sampled Data Systems

Uploaded by

Dasun Induwara
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
0% found this document useful (0 votes)
75 views

Lecture 05 - Sampled Data Systems

Uploaded by

Dasun Induwara
Copyright
© © All Rights Reserved
We take content rights seriously. If you suspect this is your content, claim it here.
Available Formats
Download as PDF, TXT or read online on Scribd
You are on page 1/ 26

Digital Signal Processing and

Machine Vision
Lecture 05 – Sampled Data Systems
R . A . P R A B H AT H B U D D H I K A
D E PA R T M E N T O F E L E C T R I C A L , E L E C T R O N I C A N D S Y S T E M S E N G I N E E R I N G
FA C U LT Y O F E N G I N E E R I N G
NSBM GREEN UNIVERSITY
Sampling
Discrete-time signals are usually preferred for computers.
However, most of the natural signals are continuous-time signals.
Sampling continuous-time signals periodically can produce discrete-time signals.
The ,mathematical background of the sampling process is the subject to this lecture.

2022 2
Sampled Data System
Analog to Digital to
Sample and
Digital Analog
Hold 𝑥[𝑛] 𝑦𝑐 (𝑡)
𝑥𝑐 (𝑡) Conversion conversion

𝜔𝑠
𝑥𝑐 𝑡 is sampled at an angular rate of 𝜔𝑠 and held until the sample is converted to a digital number
𝑥𝑛.
The sample is obtained at time 𝑡 = 𝑛𝑇 𝑤ℎ𝑒𝑟𝑒 𝑇 = 2𝜋/𝜔𝑠
𝑥 𝑛 is again converted to a continuous signal by a DAC to produce 𝑦𝑐 𝑡 .
Sampling period is thus, T.

2022 3
DAC to reconstruct the continuous-time
signal
Given the same sampling period, the DAC will reconstruct the continuous-time signal as follows.

2022 4
An equivalent model for the sampled
data system
𝑥𝑐 (𝑡) 𝑥 s(t) 𝑧𝑐 (𝑡)
ZOH

𝑆𝑇 (𝑡)

… …
t
-T 0 T 2T

2022 5
An equivalent model for the sampled
data system - impulse train
𝑆𝑇 (𝑡) is an impulse train.

𝑆𝑇 𝑡 = ෍ 𝛿(𝑡 − 𝑛𝑇)
𝑛=−∞

2022 6
An equivalent model for the sampled
data system - ZOH
ZOH, Zero Order Hold has an impulse response as follows.
ℎ𝑍𝑂𝐻 (𝑡)

t
0

2022 7
Output of the equivalent model of the
sampled data system

2022 8
Sampling and Aliasing
𝑥𝑐 (𝑡) 𝑥 s(t) 𝑧𝑐 (𝑡) = 𝑦𝑐 (𝑡)
ZOH

𝑆𝑇 (𝑡)

… … t
-T 0 T 2T
𝑥𝑆 𝑡 = 𝑥𝑐 𝑡 𝑆𝑇 (𝑡)

2022 9
Sampling and Aliasing
𝑥𝑆 𝑡 = 𝑥𝑐 𝑡 𝑆𝑇 𝑡

∴ 𝑋𝑆 𝑗𝜔 = 𝑋𝑐 𝑗𝜔 ∗ 𝑆𝑇 𝑗𝜔


1
𝑋𝑆 𝑗𝜔 = න 𝑆𝑇 𝑗𝜉 𝑋𝐶 𝑗𝜔 − 𝑗𝜉 𝑑𝜉
2𝜋
−∞

𝑆𝑇 𝑗𝜔 = 𝜔𝑠 ෍ 𝛿(𝑗𝜔 − 𝑗𝑘𝜔𝑠 )
𝑘=−∞


𝜔𝑠
𝑋𝑆 𝑗𝜔 = ෍ 𝑋𝐶 (𝑗𝜔 − 𝑗𝑘𝜔𝑠 )
2𝜋
𝑘=−∞

2022 10
Sampling and Aliasing

2022 11
Sampling and Aliasing
If 𝑥𝑐 𝑡 is band limited to −𝜔𝑁 < 𝜔 < +𝜔𝑁 and
𝜔𝑁 < 𝜔𝑠 /2
Then 𝑥𝑐 𝑡 can be recovered from 𝑥𝑆 𝑡 by passing it through a low pass filter that has the cut
off frequency somewhere between 𝜔𝑁 and 𝜔𝑠 - 𝜔𝑁 .

2022 12
Sampling and Aliasing
𝜔𝑠
If 𝑥𝑐 𝑡 is not band limited or 𝜔𝑁 > , then we have Aliasing.
2

2022 13
Nyquist Theorem
Let 𝑥𝑐 𝑡 be a band limited signal with
𝑋𝐶 (𝑗𝜔) = 0 𝑓𝑜𝑟 𝜔 > 𝜔𝑁

𝑥𝑐 𝑡 is uniquely determined by its samples 𝑥 𝑛 = 𝑥𝑐 𝑛𝑇 , 𝑛𝜖 𝕫 if


2𝜋
𝜔𝑠 = > 2𝜔𝑁
𝑇
𝜔𝑁 is called the Nyquist Frequency
2𝜔𝑁 is called the Nyquist Rate

2022 14
Exercise 01
Let 𝑥𝑐 (𝑡) be a signal with Nyquist rate 𝜔0 . Determine the Nyquist rate for the following signals.
1. 𝑥𝑐 𝑡 + 𝑥𝑐 (𝑡 − 1)
𝑑𝑥𝑐 (𝑡)
2. 𝑑𝑡

3. 𝑥𝑐2 (𝑡)
4. 𝑥𝑐 𝑡 cos(𝜔0 𝑡)

2022 15
Exercise 02
The signal 𝑥𝑐 𝑡 = cos(𝜔0 𝑡) is sampled at a rate of 1000Hz to obtain the sequence, 𝑥 𝑛 =
𝜋𝑛
cos[ ].
4

What are the possible positive values for 𝜔0 ?

2022 16
Anti-Aliasing Filter
In most practical cases 𝑥𝑐 𝑡 is not band limited but beyond a certain
frequency 𝜔0 it has only a negligible energy.
Controlling aliasing for such cases is done with an anti-aliasing filter.
The cut-off frequency is chosen to be 𝜔𝑐 = 𝜔0
The sampling frequency is chosen to be 𝜔𝑠 > 2𝜔0

2022 17
Anti-Aliasing Filter

Even if 𝑥𝑐 𝑡 is band limited, using an anti-aliasing filter can help to eliminate high frequency
noise causing aliasing into 𝑥𝑐 𝑡 frequency band.

2022 18
Bandpass Sampling
Suppose a particular signal is bandpass such that its bandwidth 𝜔𝐵 is much smaller than its center
frequency, 𝜔𝑐 .
𝜔𝐵 << 𝜔𝐶

The maximum frequency component of this signal is 𝜔𝐶 + 𝜔𝐵 /2


This signal can be sampled at a much lower sampling rate than 2𝜔𝐶 + 𝜔𝐵 without being affected by
aliasing.

2022 19
Bandpass Sampling
suppose ⍵s = ⍵c - ⍵B/2

sampling and frequency translation can be done simultaneously.

2022 20
Bandpass Sampling
Aliasing does not occur if the following conditions are satisfied.

2𝜔𝐶 + 𝜔𝐵 2𝜔𝐶 − 𝜔𝐵
≤ 𝜔𝑠 ≤ 𝑓𝑜𝑟 𝑠𝑜𝑚𝑒 𝑚 𝜖 ℤ
𝑚+1 𝑚

2𝜔𝐵 < 𝜔𝑠

2022 21
Bandpass Sampling – Example
𝜔𝐵 = 10 rad/s 𝜔𝐶 = 40 𝑟𝑎𝑑/𝑠
m 2𝜔𝐶 + 𝜔𝐵 2𝜔𝐶 − 𝜔𝐵 2𝜔𝐶 + 𝜔𝐵 2𝜔𝐶 − 𝜔𝐵
≤ 𝜔𝑠 ≤
𝑚+1 𝑚 𝑚+1 𝑚
1 45 70 ✓

2 30 35 ✓

3 22.5 23.33 ✓

4 18 17.5 x

5 15 14 x

2022 22
Quantization Error of an ADC
𝑥𝑐 𝑥𝑑
ADC

𝜔𝑠

2022 23
Quantization Error of an ADC
Suppose an ADC has following properties:
◦ Reference voltages: ±𝑉
◦ Number of output bits: B
2𝑉 𝑉
Quantization step/resolution, Δ= =
2𝐵 2𝐵−1

Quantization Error, q = 𝑥𝑐 − 𝑥𝑑 can be modeled as a white noise process uniformly distributed between
± Δ/2.
Power(same as the variance) of Quantization Error:
+Δ/2 2 2
1 Δ 𝑉
𝜎𝑞2 = න 𝑞2 𝑑𝑞 = =
−Δ/2 Δ 12 3 ∗ 4𝐵

2022 24
Signal to quantization noise ratio
𝜎𝑥2 𝑉2
𝑆𝑁𝑅 = 10𝑙𝑜𝑔 2 = 6.0206𝐵 + 4.7712 − 10𝑙𝑜𝑔 2
𝜎𝑞 𝜎𝑥

𝜎𝑥2 𝑉
𝑆𝑁𝑅 = 10𝑙𝑜𝑔 2 = 6.0206𝐵 + 4.7712 − 20𝑙𝑜𝑔
𝜎𝑞 𝜎𝑥

2022 25
Exercise 03
Reference voltage of an ADC, V = 5V. For the signal given below, determine the number of bits
required in the ADC output to obtain an SNR better than 50dB.
𝑥𝑐 (𝑡) = 0.8 cos 𝜔1 𝑡 + sin(𝜔2 𝑡)

2022 26

You might also like