KNR2113

Signals & Systems

Topic 4:
Sampling and Reconstruction (Part 1/2)

Assoc. Prof. Ir. Ts. Dr Kismet Anak Hong Ping

Department of Electrical & Electronic Engineering
Faculty of Engineering
Introduction
• The signals we use in • To process these signals
the real world, such as in computers, we need
our voices, are called to convert the signals to
"analog" signals. "digital" form.

2
• An analog signal is • A digital signal is
continuous in both time discrete in both time
and amplitude. and amplitude.

3
4
To convert a signal from continuous time to discrete time, a
process called sampling is used.

The value of the signal is measured at certain intervals in time.

Each measurement is referred to as a sample.

5
6
Sampling
When the continuous analog signal is sampled at a frequency F,
the resulting discrete signal has more frequency components
than did the analog signal.

To be precise, the frequency components of the analog signal

are repeated at the sample rate.

That is, in the discrete frequency response they are seen at

their original position.

7
Example:

8
Frequency Sampling Method

Fig. Periodic sampling of an analog signal

9
The time interval T between successive samples is called the
sampling period.

Sampling rate / Sampling frequency

1/ T = Fs

10
Sampling

How many samples are

necessary to ensure we are
preserving the information
contained in the signal?

11
If the signal contains high frequency components, we
will need to sample at a higher rate to avoid losing
information that is in the signal.

In general, to preserve the full information in the

signal, it is necessary to sample at twice the maximum
frequency of the signal.

This is known as the Nyquist rate.

12
The Sampling Theorem states that a signal can be
exactly reproduced if it is sampled at a frequency F,
where F is greater than twice the maximum
frequency in the signal.

The condition Fs > 2 Fmax ensures that all the

sinusoidal components in the analog signal are
mapped into corresponding discrete-time frequency
components with frequencies in the fundamental
interval.

13
14
15
Nyquist rate FN = 2 Fmax , where this signal is said to be
critically sampled.

What happens if we sample the signal at a frequency that is

lower than the Nyquist rate?
FN < 2Fmax

16
Aliasing
Aliasing is potentially a very serious problem in any
system that processes sampled signals.

Example :
If aliasing happens in digital telephone systems, for
example, speech would sound strange and probably
unintelligible. Digital audio recordings which failed to
sample the sound signal at a sufficient rate would
result in extremely poor reproduction.

17
Avoid aliasing
(image overlap/noise
etc)

Fs > 2 Fmax

18
19
Preventing Aliasing
What can be done to avoid or prevent aliasing? Briefly,
there are two approaches:

1. Make sure the sampling rate is high enough to

satisfy the sampling theorem, i.e. at least twice
the highest frequency component present in the
signal to be sampled.

20
21
Preventing Aliasing
2. Use an anti-aliasing filter prior to the sampling stage. This is
a low-pass filter designed to pass all frequency
components that can be handled correctly by the sampling
process, but to block out higher-frequency components
that would give rise to aliasing.

22
23
24
Sampling
In real-world applications, sampling at higher
frequencies results in better reconstructed signals.

However, higher sampling frequencies require

i. faster converters, and
ii. more storage.

25
Example
Consider the analog signal

xa (t ) = 3 cos100πt

Determine the minimum sampling rate required to

avoid aliasing

26
Solution
The frequency of the analog signal is .

F = 50 Hz

The minimum sampling rate required to avoid aliasing

is
Fs = 100 Hz

27
Periodic Sampling
Discrete time signal x[n] often arises from periodic
sampling of continuous time signal x(t):
x[n] = x(nT ) −∞ < n < ∞

This system is called an ideal continuous-to-

discrete-time (C/D) converter or sampler:

C/D
x(t ) x[n] = x(nT )

T
28
29
C/D converter is described by the following:
• sampling period: T seconds
• sampling frequency:

fs = 1/ T samples per second

Ω s = 2π / T radians per second

In practice, sampling is usually approximately
implemented using analog-to-digital (A/D) converter.

30
The sampling process is not generally invertible, one
cannot always reconstruct x(t) unambiguously from
x[n]. However, ambiguity can be removed by
restricting input signals to sampler.

31
32
Frequency domain representation
of sampling
What is the frequency domain relation between input
and output of C/D converter?

Consider converting xc (t ) to xs (t ) by modulating it

with the periodic impulse train:
∞
s (t ) = ∑ δ (t − nT )
n = −∞

33
which has frequency representation:
2π ∞
S ( jΩ ) =
T
∑ δ (Ω − kΩ )
k = −∞
s

34
Through the sifting property of the impulse function:
xs (t ) = xc (t ) s (t )
∞
= xc (t ) ∑ δ (t − nT )
n = −∞
∞
= ∑ x (nT )δ (t − nT )
n = −∞
c xc (t )

xs (t )
t 35
The Fourier Transform X s ( jΩ) of xs (t ) = xc (t ) s (t )
is the continuous-time convolution of Fourier
Transforms X c ( jΩ) and S ( jΩ) so
1
X s ( jΩ ) = X c ( jΩ ) ∗ S ( jΩ )
2π
1 ∞
= ∑ X c ( j (Ω − kΩ s ))
T k = −∞

36
Therefore, the Fourier Transform of xs (t ) consists of
copies of X c ( jΩ) shifted by integer multiples of
sampling frequency Ω s and then superimposed:

X c ( jΩ )

0 ΩN Ω
37
 S ( jΩ )

− 2Ω s − Ω s 0 Ωs 2Ω s Ω
X s ( jΩ )

− 2Ω s − Ω s 0 ΩN Ω 38
If xc (t ) is bandlimited, with highest nonzero
frequency at Ω N then the replicas do not overlap
when
Ω s > 2Ω N

Then, we can recover xc (t ) using an ideal low-pass

filter. Otherwise, X c ( jΩ) cannot be recovered using
low-pass filtering and aliasing results:

39
 X s ( jΩ )

− 2Ω s − Ω s 0 Ωs 2Ω s Ω

“ aliasing “

40
The frequency Ω N is referred to as the Nyquist
frequency, and the frequency 2Ω N that must be
exceeded in the sampling is the Nyquist rate.

The objective now is to express the Fourier Transform

jω
X (e ) of x[n] in terms of X c ( jΩ) and X s ( jΩ) .
Taking the Fourier Transform of the relationship:
∞
xs (t ) = ∑ x (nT )δ (t − nT )
n = −∞
c

41
yields the following:
∞
X s ( jΩ ) = ∑ c
x (
n = −∞
nT ) e − jΩTn

Now, since x[n] = xc (nT ) and

∞
X ( e jω ) = ∑ x
n = −∞
[ n ]e − j ωn

42
It follows that:
X s ( j Ω ) = X ( e jω ) = X (e jΩT )
ω = ΩT

Consequently
1 ∞
jΩT
X (e ) =
T
∑X
k = −∞
c ( j (Ω − kΩ s ))

1 ∞
  ω 2πk  
jω
X (e ) =
T
∑ X c  j  −  
k = −∞  T T 

43
Thus, X (e jω ) is just a frequency scaled version of
X s ( jΩ) with the scaling specified by ω = ΩT .

Alternatively, the effect of sampling may be thought of

as a normalization of the frequency axis, so that the
frequency Ω = Ω s of X s ( jΩ) is normalized to
ω = 2π for X (e jω ) .

44
Reconstruction of bandlimited
signal from samples
If samples of a bandlimited continuous time signal are taken
frequently enough, then they are sufficient to represent the
signal exactly.

The continuous time signal can then be recovered from the

samples.

This task is ideally performed by a discrete-to-continuous-time

(D/C) converter.

The form and behaviour of such a converter is discussed in this

section.
45
Reconstruction of bandlimited
signal from samples
Given sequence of samples x[n], we can form impulse
train ∞
xs (t ) = ∑ x[n]δ (t − nT )
n = −∞

The nth sample corresponds to the impulse at time

t = nT . If appropriate sampling conditions are met,
namely the signal is bandlimited and the Fourier
Transform replicas do not overlap, then x(t) can be
reconstructed from xs (t ) by ideal continuous time
low-pass filtering:
46
Reconstruction of bandlimited
signal from samples
∞
xr (t ) = ∑ x[n]h (t − nT )
n = −∞
r

Here, hr (t ) is impulse response of an ideal LPF with

cutoff frequency at Ω c :
H r ( jΩ ) X s ( jΩ )

− 2Ω s − Ω s 0 Ωs 2Ω s Ω
Ωc 47
Reconstruction of bandlimited
signal from samples
A convenient choice for the cutoff frequency is
Ω c = Ω s / 2 = π / T corresponding to the ideal
reconstruction filter
T Ω ≤ π /T
H r ( jΩ ) = 
0 Ω > π /T

48
Reconstruction of bandlimited
signal from samples
And reconstructed signal:
X r ( jΩ) = H r ( jΩ) X (e jΩT )

TX (e jΩT ) Ω ≤ π /T
=
 0 Ω > π /T

49
Reconstruction of bandlimited
signal from samples
In the time domain, the ideal reconstruction filter has
impulse response:
sin(π t / T )
hr (t ) =
π t /T
So, the reconstructed signal is:
∞
sin[π (t − nT ) / T ]
xr (t ) = ∑ x[n]
n = −∞ π ( t − nT ) / T

50
Reconstruction of bandlimited
signal from samples
From the previous frequency domain argument, if
x[n] = xc (nT ) with X c ( jΩ) = 0 for Ω ≥ π / T then
xr (t ) = xc (t ) .

Note that the filter hr (t ) is not realizable since it has

infinite duration.

51
Reconstruction of bandlimited
signal from samples
An ideal discrete-to-continuous (D/C) reconstruction
system, therefore, has the form:

Sequence to Reconstruction
impulse train filter
x[n] xs (t ) H r ( jΩ ) xr (t )
T

D/C
x[n] xr (t )
T
52
Discrete time processing of
continuous time signals
Discrete time systems are often used to process
continuous time signals. This is accomplished by:

Discrete
C/D D/C
time
xc (t ) x[n] system
y[n] yr (t )
T T

H eff ( jΩ) = H c ( jΩ)

53
Discrete time processing of
continuous time signals
For now, it is assumed that the C/D and D/C converters
have the same sampling rate. The C/D converter
produces the discrete time signal x[n] = xc (nT ) with
Fourier Transform:
1 ∞
  ω 2πk  
X (e ) = ∑ X c  j  −
jω
 
T k = −∞   T T 

54
Discrete time processing of
continuous time signals
The D/C converter creates a continuous time output of
the form: ∞
sin[π (t − nT ) / T ]
yr (t ) = ∑ y[n]
n = −∞ π (t − nT ) / T
The continuous time Fourier Transform of yr (t )
namely Yr ( jΩ) and the discrete time Fourier
Transform of y[n] namely Y (e jΩ ) are related by:
jΩT
Yr ( jΩ) = H r ( jΩ)Y (e )
55
Discrete time processing of
continuous time signals
TY (e jΩT ) Ω < π / T
=
 0 otherwise

If the discrete time system is LTI, then

Y ( e jω ) = H ( e jω ) X ( e jω )

where H (e jω ) is the frequency response of the

system.

56
Discrete time processing of
continuous time signals
Therefore,
Yr ( jΩ) = H r ( jΩ) H (e jΩT ) X (e jΩT )
1 ∞
  2πk  
= H r ( jΩ) H (e ) ∑ X c  j  Ω −
jΩT
 
T k = −∞   T 

If X c ( jΩ) = 0 for Ω ≥ π / T then the ideal LPF H r ( jΩ)

selects only the term for k = 0 in the sum, and scales
the result:
 H (e jΩT ) X c ( jΩ) Ω < π / T
Yr ( jΩ) = 
 0 Ω ≥ π /T
57
Discrete time processing of
continuous time signals
Thus, if X c ( jΩ) is bandlimited and sampled above
the Nyquist rate, then the output is related to the
input by:
Yr ( jΩ) = H eff ( jΩ) X c ( jΩ)

where  H (e jΩT ) Ω < π /T

H eff ( jΩ) = 
 0 Ω ≥ π /T
is the effective frequency response of the system

58
Continuous time processing of
discrete time signals
It is conceptually useful to consider continuous time
processing of discrete time signal. This system is:

D/C hc (t ) C/D
x[n] xc (t ) H c ( jΩ ) yr (t ) y[n]
T T

h[n], H (e jω )

59
Continuous time processing of
discrete time signals
Since the D/C converter includes an ideal LPF, X c ( jΩ)
and therefore also Yc ( jΩ) will be zero for Ω ≥ π / T .
Thus, the C/D converter samples yc (t ) without
aliasing and we have:
∞
sin[π (t − nT ) / T ]
xc (t ) = ∑ x[n]
n = −∞ π (t − nT ) / T
∞
sin[π (t − nT ) / T ]
yc (t ) = ∑ y[n]
n = −∞ π (t − nT ) / T
where x[n] = xc (nT ) and y[n] = yc (nT ).
60
Continuous time processing of
discrete time signals
In the frequency domain,
X c ( jΩ) = TX (e jΩT ) Ω < π /T

Yc ( jΩ) = H c ( jΩ) X c ( jΩ) Ω < π /T

1 ω
jω
Y (e ) = Yc ( j ) ω <π
T T

61
Continuous time processing of
discrete time signals
The overall system, therefore, behaves like a discrete-
time system with frequency response:
ω
jω
H (e ) = H c ( j ) ω <π
T
Equivalently, the overall frequency response of the
system will be equal to a given H (e jω ) if the
frequency of the continuous-time system is
H c ( jΩ) = H (e jΩT ) Ω < π /T
Since X c ( jΩ) = 0 for Ω < π / T , H c ( jΩ) may be
chosen above π / T . 62
63