5/1/2017

UNIT–V

SAMPLING

Prof K.Venkat Reddy

Reference Books

1-May-17 Prof K.Venkat Reddy 2

1
 5/1/2017

UNIT-V Syllabus
 Sampling : Sampling theorem –
Graphical and analytical proof for
Band limited signals, Impulse
(Ideal) sampling, Natural
(Chopped) Sampling and Flat top
(S&H) Sampling, Reconstruction of
signals from its samples, effect of
under sampling – Aliasing,
Introduction to Band Pass
sampling.

1-May-17 Prof K.Venkat Reddy 3

Contents
Sampling :
 Sampling theorem
 Graphical and analytical proof for
Band limited signals
 Impulse (Ideal) sampling
 Natural (Chopped) Sampling and
 Flat top (S&H) Sampling
 Reconstruction of signals from its
samples
 Effect of under sampling – Aliasing
 Introduction to Band Pass sampling.
1-May-17 Prof K.Venkat Reddy 4

2
 5/1/2017

INTRODUCTION
 Earlier we had defined a continuous –
time signal as one which is defined for
all values of time and
 A discrete-time signal as one which is
defined only over a discrete set of
points in time.
 Most of the signals that we encounter in
practice are continuous –time(analog)
signals.

1-May-17 Prof K.Venkat Reddy 5

INTRODUCTION
 Analog signal processing,
representation, transmission and
recovery fall under the category of
analog communications which have
certain drawbacks.
 In digital communications, which is
more advantages, it is required to
transform an analog signal into a
discrete-time signal.
 The process of converting a continuous-
time signal into a discrete-time signal is
called Sampling. 1-May-17 Prof K.Venkat Reddy 6

3
 5/1/2017

INTRODUCTION
 After sampling, the signal is defined at
discrete instants of time and the time
interval between two successive
sampling instants is called sampling
period or sampling interval.
 In the process of sampling, one of the
important factors that we have to
consider is—the sampling rate must be
kept sufficiently high so that the original
signal can be reconstructed from its
samples.
1-May-17 Prof K.Venkat Reddy 7

SAMPLING
 The sampling operation can be
represented by a fictitious switch shown
in Fig.8.1. the switch is closed for a very
short interval of time τ (ideally τ=0),
once every T sec during which the signal
is available at the output.

1-May-17 Prof K.Venkat Reddy 8

4
 5/1/2017

SAMPLING
 Therefore, if the input is x(t), then the
output xs(t) is x(nT), n=0,+1,+2,… and
x(nT) is called the sampling sequence of
x(t), where T is called the sampling
period or sampling interval. It is the
time interval between successive
samples and the sampling frequency is
given by fs=(1/T) Hz.
 Although a mechanical switch is shown
in Fig.8.1, in actual practice, an
electronic switch may be used.
1-May-17 Prof K.Venkat Reddy 9

SAMPLING THEOREM
 The sampling theorem is one of the most
useful theorems since it applies to digital
communication systems.
 The sampling theorem states that a
band limited signal x(t) with X(ω)=0 for
|ω|> ωm [i.e. X(f)=0 for f>fm] can be
represented into & uniquely determined
from its samples x[nT], if the sampling
frequency fs>2fm. Where fm is the
highest freqy component present in it.
 That is, for signal recovery, the sampling
frequency must be atleast twice the highest
frequency present in the signal.
1-May-17 Prof K.Venkat Reddy 10

5
 5/1/2017

SAMPLING THEOREM
 The theorem is known as uniform
sampling theorem since it pertains to
the specifications of a given signal by its
samples at uniform intervals of 1/2fm
sec.
 It is also called low pass sampling
theorem because it applies to low pass
signals, i.e. signals for which X(f)=0 for
all frequencies such that |f|>fm, where
fm is some finite frequency.

1-May-17 Prof K.Venkat Reddy 11

SAMPLING THEOREM
 Proof: The sampling operation can be
represented as shown in Fig.8.2. x(t) is
a continuous-time band limited signal to
be sampled which has no spectral
components above fm Hz.

1-May-17 Prof K.Venkat Reddy 12

6
 5/1/2017

SAMPLING THEOREM
 Proof: That means X(ω), the F.T of x(t)
is 0 for ω>ωm. δT(t) is an impulse train
which samples at a rate of fs Hz and
xs(t) is sampled signal. T is the
sampling period and fs=(1/T) is the
sampling frequency.
 xs(t) is the product of signal x(t) and
impulse train δT(t). It is a sequence of
impulses located at regular intervals of
T sec and having strength equal to the
values of x(t) at the corresponding
instants.
1-May-17 Prof K.Venkat Reddy 13

SAMPLING THEOREM
 Proof: 
 xs (t )  x(t )  T (t ) Where  T (t )    (t  nT )
n  
 The exponential form of Fourier series
of δT(t) is  
jn t
 T (t )    (t  nT )   C
n   n  
n e s

 Where Cn 
1 T2
 T
 (t )e  jnst 
1
T 2 T
 
1 1  jnst
  T (t )    (t  nT )   T e
n   n  
jns t
 e
T n

1 
xs (t )  x(t )  T (t )   x(t )e jnst
T n
1-May-17 Prof K.Venkat Reddy 14

7
 5/1/2017

SAMPLING THEOREM
 Proof: Taking F.T on both sides, we
have
1   1   
F .T {xs (t )}  F .T   x(t )e jnst   F .T   x(t )e jnst 
T n  T n 
1  1   2 
.i.e. X s ( )  
T n
X (  n s )   X  
T n  T
n


or X s ( f )  fs  X ( f  nf )
n  
s

 Where X(ω) or X(f) is the spectrum of

input signal and Xs(ω) or Xs(f) is the
spectrum of the sampled signal.

1-May-17 Prof K.Venkat Reddy 15

SAMPLING THEOREM
 Proof: Thus, the F.T of the sampled
signal is given by an infinite sum of
shifted replicas of the F.T of the Original
Signal.
 The signal x(t) is band limited to fm. The
term X[ω-(2π/T)n] is the shifting of X(ω)
from ω=0 to (2π/T)n. Hence Xs(ω) is the
sum of shifted replicas of (1/T)X(ω)
centering at (2π/T)n, n=0, +1, +2, ….
Fig 8.3 shows the plot of X(ω) and Xs(ω)
for various values of π/T.

1-May-17 Prof K.Venkat Reddy 16

8
 5/1/2017

SAMPLING
THEOREM

1-May-17 Prof K.Venkat Reddy 17

SAMPLING THEOREM
 Proof: It shows that [Fig.8.3.b & c] if
(π/T)>ωm, the replicas will not overlap
and as a result, the frequency spectrum
of TXs(ω) in the frequency range [-(π/T),
π/T] is identical to X(ω). X(ω) can be
recovered from Xs(ω) by passing it
through a low pass filter which has
sharp cutoff at ω= π/T. If (π/T)< ωm
(Fig.8.3d), the successive frequency
spectra will overlap and the original
signal cannot be recovered from the
sampled signal.
1-May-17 Prof K.Venkat Reddy 18

9
 5/1/2017

SAMPLING THEOREM
 Proof: Therefore, we can say that for
signal recovery,
s  m  m , .i.e. s  2m

 Or f s  f m  f m , .i.e. fs  2 fm


 Or  m , .i.e. f s  2f m
T

 Or 1

1
, i.e. T 
1
fs 2 fm 2 fm

1-May-17 Prof K.Venkat Reddy 19

SAMPLING THEOREM
 Proof: So we can conclude that if the
sampling interval T is small[<(1/2fm)],
X(ω) can be recovered from Xs(ω), but if
T becomes larger than 1/2fm, then there
is an overlap between successive cycles
& X(ω) cannot be recovered from Xs(ω).
This proves the sampling theorem.
 From the previous discussion, we can
observe that when the spectra overlap,
it is impossible to retrieve x(t) from
xs(t).

1-May-17 Prof K.Venkat Reddy 20

10
 5/1/2017

SAMPLING THEOREM
 Proof: Thus, we find that in general,
there are two basic conditions to be
satisfied if x(t) is to be recovered from
its samples.
1. Signal x(t) should be band limited to
some frequency ωm.
2. The sampling frequency ωs should be
at least twice the band limiting
frequency ωm.[.i.e. ωs>2ωm].

1-May-17 Prof K.Venkat Reddy 21

SAMPLING THEOREM
 From Fig.8.3, we can observe that
1. Xs(ω) is a repetitive version of X(ω)
with X(ω) repeating itself at regular
intervals of ωs, the sampling frequency.
2. When ωs>2ωm [Fig.8.3b], the spectral
replicates have large separation
between them, known as guard band,
which makes the process of filtering
much easier and effective. Even a non-
ideal filter which does not have a sharp
cutoff can also be used.
1-May-17 Prof K.Venkat Reddy 22

11
 5/1/2017

SAMPLING THEOREM
3. When ωs=2ωm [Fig.8.3c], there is no
separation between replicates, so no
guard band exists and X(ω) can be
obtained from Xs(ω) by using only an
ideal low pass filter with sharp cutoff.
4. When ωs<2ωm [Fig.8.3d], the low
frequency components in Xs(ω) overlap
on the high frequency components of
X(ω), there is distortion and X(ω)
cannot be recovered from Xs(ω) by
using any filter. This type of distortion
is called aliasing.
1-May-17 Prof K.Venkat Reddy 23

SAMPLING THEOREM
 Aliasing can be avoided if fs>2fm or
T<(1/2fm).
 Since it is impossible to build filters
having an infinite sharpness of cutoff, a
guard band between fm and fs-fm is
preferred.
 The impulse train at the sampler is
processed through an ideal LPF with gain
T and cutoff frequency greater than ωm
and less than ωs-ωm. The resulting
output signal will exactly equal x(t).
1-May-17 Prof K.Venkat Reddy 24

12
 5/1/2017

SAMPLING THEOREM
NYQUIST RATE OF SAMPLING
 Nyquist rate of sampling is the
theoretical minimum sampling rate at
which a signal can be sampled and still
be reconstructed from its samples
without any distortion.
 It is the ―theoretical minimum‖ because
when the Nyquist rate of sampling is
used, only an ideal LPF can be used to
extract X(ω) from Xs(ω), .i.e to recover
x(t) from xs(t). It is always equal to 2fm
where fm is the maximum frequency
component present in the signal.
1-May-17 Prof K.Venkat Reddy 25

SAMPLING THEOREM
NYQUIST RATE OF SAMPLING
 A signal sampled at greater than Nyquist
rate is said to be over sampled and a
signal sampled at less than its Nyquist
rate is said to be under sampled.
 Nyquist interval is the time interval
between any two adjacent samples when
sampling rate is Nyquist rate.
 Nyquist rate fN = 2fm Hz
 Nyquist interval=1/fN=1/2fm sec

1-May-17 Prof K.Venkat Reddy 26

13
 5/1/2017

SAMPLING THEOREM
EFFECTS OF UNDER SAMPLING-ALIASING
 When ωs<2ωm i.e. when the signal is
under sampled, X(ω), the spectrum of
x(t) is no longer replicated in Xs(ω), and
thus is no longer recoverable by low pass
filtering. This effect in which the
individual terms in equation
1 
X s ( )   X (  ns )
T n 

overlap is referred to as aliasing.

 This process of spectral overlap is also
called frequency folding effect.
1-May-17 Prof K.Venkat Reddy 27

SAMPLING THEOREM
EFFECTS OF UNDER SAMPLING-ALIASING
 In fact, aliasing is defined as the
phenomenon in which a high frequency
component in the frequency spectrum of
signal takes identity of a lower frequency
component in the spectrum of the
sampled signal.
 Aliasing can occur if either of the
following conditions exists:
1. The signal is not band-limited to a finite
range.
2. The sampling rate is too low.

1-May-17 Prof K.Venkat Reddy 28

14
 5/1/2017

SAMPLING THEOREM
EFFECTS OF UNDER SAMPLING-ALIASING
 Theoretically if the signal is not band-
limited, there is no way of avoiding the
aliasing problem with the basic sampling
scheme employed. However, the spectra
of most real life signals are such that
they may assumed to be band-limited.
Further, a common practice employed in
many sampled data systems is to filter
the continues-time signals before
sampling to ensure that it does meet the
band-limited criterion closely enough for
all practical purposes.
1-May-17 Prof K.Venkat Reddy 29

SAMPLING THEOREM
EFFECTS OF UNDER SAMPLING-ALIASING
 To avoid aliasing, it should be ensured
that:

1. x(t) is strictly band-limited( this can

be ensured by using anti-aliasing filter
before the sampler.)

1. fs is greater than 2fm.

1-May-17 Prof K.Venkat Reddy 30

15
 5/1/2017

SAMPLING THEOREM
ANTI-ALIASING FILTER

 Sampling theorem states that a signal

can be perfectly reconstructed from its
samples only if it is band-limited.

 In practice no signal is strictly band-

limited. i.e. in general signals have
frequency spectra consisting of low
frequency components as well as high
frequency noise components.

1-May-17 Prof K.Venkat Reddy 31

SAMPLING THEOREM
ANTI-ALIASING FILTER
 When a signal is sampled, with sampling
frequency fs, all signals with frequency
range higher than ωs/2 appear as signal
frequencies between 0 and ωs/2 creating
aliasing.
 Therefore, to avoid aliasing errors
caused by the undesired high frequency
signals, it is necessary to first band-limit
x(t) to some appropriate frequency fm
using an LPF such that most part of the
energy is retained.

1-May-17 Prof K.Venkat Reddy 32

16
 5/1/2017

SAMPLING THEOREM
ANTI-ALIASING FILTER

 This LPF used for band-limiting a signal

before sampling as shown in Fig.8.4 is
generally referred to as an anti-aliasing
filter since it is used primarily for
preventing aliasing.

1-May-17 Prof K.Venkat Reddy 33

SAMPLING TECHNIQUES
 Sampling of a signal is done in several
ways. Basically there are three types of
sampling techniques:
1. Instantaneous sampling or impulse sampling
2. Natural sampling
3. Flat top sampling
 Out of these three methods,
instantaneous or impulse sampling is
also called ideal sampling, where as the
natural sampling and flat top sampling
are called practical sampling methods.
1-May-17 Prof K.Venkat Reddy 34

17
 5/1/2017

SAMPLING TECHNIQUES
IDEAL OR IMPULSE SAMPLING
 Ideally, sampling should be done
instantaneously so that the kth element
of the sequence obtained by sampling,
represents the value of x(t) at t=kT. The
operation is shown in Fig.8.5a.
 Assume that the fictitious sampler closes
almost for zero time once in every T sec.
it is equivalent to transmitting the input
signal to the output for a very very short
time (almost zero time) once every T
sec.

1-May-17 Prof K.Venkat Reddy 35

SAMPLING TECHNIQUES
IDEAL OR IMPULSE SAMPLING
 The mechanical switch can be replaced
by an electronic switch which is basically
a Pulse Amplitude Modulator as shown in
Fig.8.5b.

1-May-17 Prof K.Venkat Reddy 36

18
 5/1/2017

SAMPLING TECHNIQUES
IDEAL OR IMPULSE SAMPLING
 Now, the operation is equivalent to
multiplying the input signal x(t) by an
impulse train δT(t) as shown in Fig.8.5c.
So the output of the sampler is a train of
impulses of height equal to the
instantaneous value of the input signal at
the sampling instant.
 The impulse train, also called the
sampling function is represented as:

 T (t )    (t  nT )
n  

1-May-17 Prof K.Venkat Reddy 37

SAMPLING TECHNIQUES
IDEAL OR IMPULSE SAMPLING
 The sampled signal is given by
 
xs (t )  x(t ) T (t )  x(t )   (t  nT )   x(nT ) (t  nT )
n   n  

1 
 X s ( )   X (  ns )
T n 

or X s ( f )  fs  X ( f  nf )
n  
s

 This equation gives the spectrum of

ideally sampled signal. it shows that the
spectrum Xs(ω) is an infinite sum of
shifted replicas of X(ω).
1-May-17 Prof K.Venkat Reddy 38

19
 5/1/2017

SAMPLING TECHNIQUES
IDEAL OR IMPULSE SAMPLING
 It shows that the spectrum Xs(ω) is an
infinite sum of shifted replicas of X(ω)
spaced nωs apart, where n=+1,+2, etc.
and scaled by a factor 1/T.
 However, it may be noted that ideal or
instantaneous sampling is possible only
in theory because it is impossible to have
a pulse with pulse width approaching
zero.
 Practically, the flat top sampling or
natural sampling is used.
1-May-17 Prof K.Venkat Reddy 39

SAMPLING TECHNIQUES
NATURAL (CHOPPED) SAMPLING
 Natural sampling, also called sampling,
using a sequence of pulses is the most
practical way of accomplishing sampling
of a band-limited signal.
 This is achieved by multiplying the signal
x(t) with a pulse train pT(t) as shown in
Fig.8.6.
 Each pulse of pT(t) is of short duration τ
and occurs at a sampling period of T sec.

1-May-17 Prof K.Venkat Reddy 40

20
 5/1/2017

SAMPLING TECHNIQUES
NATURAL (CHOPPED) SAMPLING
 The output of the sampler is same as the
input during that short duration τ. Hence
it is termed as natural sampling.

1-May-17 Prof K.Venkat Reddy 41

SAMPLING TECHNIQUES
NATURAL (CHOPPED) SAMPLING
 Fig.8.7
explains
the
process
of natural
sampling.

1-May-17 Prof K.Venkat Reddy 42

21
 5/1/2017

SAMPLING TECHNIQUES
NATURAL (CHOPPED) SAMPLING
 Fig.8.7a: The signal x(t) to be sampled &
Fig.8.7b: Its spectrum X(f).
 Fig.8.7c: The pulse train pT(t) &
Fig.8.7d: Its spectrum P(f).
 Fig.8.7e: The o/p of the sampler xs(t) &
Fig.8.7f: The o/p spectrum Xs(f).
 From Fig.8.7f it is clear that X(f) can be
recovered from Xs(f), i.e. x(t) can be
recovered from xs(t), if fs>2fm by using
an LPF whose gain is constant atleast up
to f=fm and whose cutoff frequency B is
such that fm<B<fs-f1-May-17
m. Prof K.Venkat Reddy 43

SAMPLING TECHNIQUES
NATURAL (CHOPPED) SAMPLING
 The output of the sampler is:
xs (t )  x(t ) pT (t )
Where 
pT (t )   p(t  nT )
n  

 As pT(t) is a periodic pulse train, let us

write its Fourier series expansion
 
pT (t )   p(t  nT ) 
n  
C e
n  
n
j 2nfs t

 Where 1 T2
Cn  
T 2 T
pT (t ) e  j 2nfst dt

1-May-17 Prof K.Venkat Reddy 44

22
 5/1/2017

SAMPLING TECHNIQUES
NATURAL (CHOPPED) SAMPLING
 Since τ the width of p(t), single pulse in
pT(t) is very much less than T and p(t)=0
for |t|> τ /2, we may write
1 T2 1 
Cn  
T 2 T
pT (t ) e  j 2nfst dt   p(t ) e  j 2nfst dt
T 

Cn  f s P(nf s )

 Where P(nf s )  F .T [ p(t )] f nf

s


 pT (t )  f s  P(nf )e
n  
s
j 2nfs t

1-May-17 Prof K.Venkat Reddy 45

SAMPLING TECHNIQUES
NATURAL (CHOPPED) SAMPLING
 And   
X s ( f )  F .T xs (t )  F .T  f s  P(nf s ) x(t )e j 2nfst 
 n  

X s ( f )  fs  P(nf ) X ( f )  ( f  nf )
n  
s s

F .T e    ( f  nf s )
j 2nfs t
Since


Hence X s ( f )  f s  P(nf s ) X ( f  nf s )
n  
If x(t) has a spectrum X(f), as shown in
Fig.8.7b, then Xs(f),the spectrum of the
sampled version of x(t) will appear as
shown in Fig.8.7f.

1-May-17 Prof K.Venkat Reddy 46

23
 5/1/2017

SAMPLING TECHNIQUES
FLAT TOP (S&H) SAMPLING
 This is the simplest and most popular
sampling method that uses the Sample
and Hold(S/H) circuit with flat top
samples. This is also called practical
sampling.
 Here the top of the samples remain
constant which is equal to the
instantaneous value of the base band
signal x(t) at the beginning of sampling.
The duration or width of each sample is τ
and the sampling rate, fs=1/T.

1-May-17 Prof K.Venkat Reddy 47

SAMPLING TECHNIQUES
FLAT TOP (S&H) SAMPLING
 The schematic of a ‗sample and hold‘
(S/H) circuit is shown in Fig.8.8a and a
typical output waveform from an S/H
circuit is shown in Fig.8.8b.

1-May-17 Prof K.Venkat Reddy 48

24
 5/1/2017

SAMPLING TECHNIQUES
FLAT TOP (S&H) SAMPLING
 The (S/H) circuit essentially consists of
two switches S1 , S2 and a capacitor C
connected as shown in Fig.8.8a.
 With S2 open, S1 closed for a very brief
period at each sampling instant. The
capacitor C then gets charged to a
voltage equal to the value of the input
signal x(t) at the sampling instant and
holds it for a period τ at the end of which
S2 is closed to allow the capacitor to
discharge.

1-May-17 Prof K.Venkat Reddy 49

SAMPLING TECHNIQUES
FLAT TOP (S&H) SAMPLING
 This sequence of operations is repeated
at the next and all subsequent sampling
instants.
 The switches S1 and S2 are generally FET
switches and are operated by giving
appropriate pulses to their gates.
 An actual S/H circuit uses one or two op-
amps also.
 The voltage across C appears as xs(t)
and is sketched in Fig.8.8b.

1-May-17 Prof K.Venkat Reddy 50

25
 5/1/2017

SAMPLING TECHNIQUES
FLAT TOP (S&H) SAMPLING
 From the Fig it is obvious that the
sampled version, xs(t) consists of a
sequence of rectangular pulses, the
leading edge of the kth pulse being at
t=kT and the amplitude of the pulse
being the value of x(t) at t=kT, i.e x(kT).
 The sampled signal xs(t) is the
convolution of rectangular pulses p(t)
and the ideally sampled version of x(t),
i.e. of xδ(t)
 X s ( f )  P( f ) X  ( f )

1-May-17 Prof K.Venkat Reddy 51

SAMPLING TECHNIQUES
FLAT TOP (S&H) SAMPLING
 Assume that x(t) has a spectrum as
shown in Fig.8.9.

 Since p(t) is a rectangular pulse of width

τ , its Fourier transform, P(f) which is a
sinc function will have a shape as shown
in Fig.8.10 and will have its first zero
values only at f=+(1/τ).
1-May-17 Prof K.Venkat Reddy 52

26
 5/1/2017

SAMPLING TECHNIQUES
FLAT TOP (S&H) SAMPLING
 Since τ <T, these zero values of |P(f)|
which occur at +(1/τ), will be far away
from fs and –fs. Since X s ( f )  P( f ) X  ( f ) , its
plot will be as shown in Fig.8.10b.
Observe that the mag of the high freqy
components in Xs(f) are relatively
reduced as compared to the mag of the
low freqy components because of the
multiplication of Xδ(f) by P(f). So we can
only get a distorted version of x(t), but
not exact x(t), by passing xs(t) through
an LPF.
1-May-17 Prof K.Venkat Reddy 53

SAMPLING TECHNIQUES
FLAT TOP (S&H) SAMPLING

1-May-17 Prof K.Venkat Reddy 54

27
 5/1/2017

SAMPLING TECHNIQUES
FLAT TOP (S&H) SAMPLING
 This distortion, wherein the amp of the
high freqy components are reduced
relative to the amp of the low freqy
components, in the reconstructed signal
x(t) obtained from the flat top sampled
version of the signal is referred to as the
aperture effect.
 This aperture effect can be reduced by
using an equalizer with transfer function
He(f) in cascade with the reconstruction
filter. Where 1
He ( f )  ; f  fm
P( f )
1-May-17 Prof K.Venkat Reddy 55

RECONSTRUCTION OF SIGNALS
FROM ITS SAMPLES
 The process of obtaining the analog
signal x(t) from the sampled signal xs(t)
is called data reconstruction or
interpolation.
 We know that

xs (t )  x(t ) T (t )  x(t )   (t  nT )
n  

Or
 xs (t )   x(nT ) (t  nT )
n  

 Since δ(t-nT) is zero except at the

sampling instants t=nT.
1-May-17 Prof K.Venkat Reddy 56

28
 5/1/2017

RECONSTRUCTION OF SIGNALS
FROM ITS SAMPLES
 The reconstruction filter, which is
assumed to be linear and time invariant,
has unit impulse response h(t).
 The reconstruction filter output, y(t) is
given by the convolution.

 
y (t )    x(nT )  (  nT ) h(t   ) d
 n  

1-May-17 Prof K.Venkat Reddy 57

RECONSTRUCTION OF SIGNALS
FROM ITS SAMPLES
 Or, upon changing the order of
summation and integration,
 
y (t )   x(nT )   (  nT )h(t   )d
n   

 .i.e.

y (t )   x(nT ) h(t  nT )
n  

1-May-17 Prof K.Venkat Reddy 58

29
 5/1/2017

IDEAL RECONSTRUCTION FILTER

 If x(t) is sampled at a frequency

exceeding the Nyquist rate and if the
sampled signal xs(t) is passed through
an ideal LPF, with bandwidth greater than
fm but less than fs-fm and a pass band
amplitude response of T, the filter output
is x(t).
 We choose the bandwidth of the ideal
reconstruction filter to be 0.5fs. The
transfer function of this ideal
reconstruction filter is, therefore,

1-May-17 Prof K.Venkat Reddy 59

IDEAL RECONSTRUCTION FILTER

T , | f | 0.5 f s
H( f )  
 0, otherwise
 As shown in Fig.8.11.

1-May-17 Prof K.Venkat Reddy 60

30
 5/1/2017

IDEAL RECONSTRUCTION FILTER

 The impulse response of the ideal

reconstruction filter is given by
fs
2

h(t )   Te j 2ft df
fs
 2 fs

e j 2ft
e 
2
T jf s t
 Which is h(t )  T j 2t
 j 2t
 e  jf st

fs
2

h(t ) 
1
f s t

e jf st e  jf st
2j
  sinftf t s

 Or h(t )  sin c f s t
1-May-17 Prof K.Venkat Reddy 61

IDEAL RECONSTRUCTION FILTER

 Substituting this value of h(t) in the expression
for output y(t), we get

y (t )   x(nT ) sin c f (t  nT )
n  
s

 A more convenient form for this expression,

which is often referred to as an interpolation
formula is: 
t 
y (t )   x(nT ) sin c T  n 
n  
 This shows that the original data signal can be
reconstructed by weighting each sample by a
sinc function centered at the sample time and
summing.
1-May-17 Prof K.Venkat Reddy 62

31
 5/1/2017

IDEAL RECONSTRUCTION FILTER

 The reconstruction filter discussed above

is non-causal and an impulse response is
not limited. So it cannot be used for real
time applications.
 In practice, several other methods are
used to reconstruct the signal. Some of
the important ones among them are:
◦ Xero order hold
◦ First order hold
◦ Linear interpolator

1-May-17 Prof K.Venkat Reddy 63

IDEAL RECONSTRUCTION FILTER

1-May-17 Prof K.Venkat Reddy 64

32
 5/1/2017

EXAMPLES
 Example.1: Determine the Nyquist rate
of the following signals:
a) x(t )  1  cos 2000t  sin 4000t
sin 4000t  sin 4000t 
2
b) x(t )  c) x(t )  
t  t 
 Solution:
 a) Given x(t )  1  cos 2000t  sin 4000t
 Highest freq in ‗1‘ is 0.
 Highest freq in cos2000πt is ωm1=2000π.
 Highest freq in sin4000πt is ωm2=4000π.

1-May-17 Prof K.Venkat Reddy 65

EXAMPLES
Solution:
So the maximum freq component in x(t) is
ωm=4000π.
Therefore 2πfm=4000π or fm=2000Hz
Therefore Nyquist rate fN=2fm=4000Hz
& Nyquist interval =1/fN=1/4000=0.25ms

1-May-17 Prof K.Venkat Reddy 66

33
 5/1/2017

EXAMPLES
 Solution: sin 4000t
 b) Given x(t ) 
t
Highest freq in sin4000πt is ωm=4000π.
So the maximum freq component in x(t) is
ωm=4000π.
Therefore
2πfm=4000π or fm=2000Hz
Therefore
Nyquist rate fN=2fm=4000Hz
& Nyquist interval =1/fN=1/4000=0.25ms

1-May-17 Prof K.Venkat Reddy 67

EXAMPLES
 B)

1-May-17 Prof K.Venkat Reddy 68

34
 5/1/2017

EXAMPLES
 Solution:
 c) Given

 sin 4000t   sin 4000t   sin 4000t 

2

x(t )        
 t t t
sin 4000t
 Highest freq in is ωm1=4000π.
t
 Highest freq in
 sin 4000t   sin 4000t  1  cos 8000t 
 t   t    t 2 
 
is ωm2=8000π.
1-May-17 Prof K.Venkat Reddy 69

EXAMPLES
 Solution:
 c) So the maximum freq component in
x(t) is ωm=8000π.
Therefore
2πfm=8000π or fm=4000Hz
Therefore
Nyquist rate fN=2fm=8000Hz
And
Nyquist interval
=1/fN=1/8000=0.125ms

1-May-17 Prof K.Venkat Reddy 70

35
 5/1/2017

EXAMPLES
 Solution:
 c)

1-May-17 Prof K.Venkat Reddy 71

EXAMPLES
 Example.2: A signal x(t) = 2 cos400πt
+ 6 cos640πt is ideally sampled at
fs=500Hz. If the sampled signal is
passed through an ideal low pass filter
with a cutoff frequency of 400Hz, what
frequency component will appear in the
output? Sketch the output spectrum.
Also find the output signal.
 Solution: Given
x(t )  2 cos 400t  6 cos 640t
 i.e. x(t )  2 cos[2 (200)]t  6 cos[2 (320)t ]

1-May-17 Prof K.Venkat Reddy 72

36
 5/1/2017

EXAMPLES
 Taking F.T both sides we get
X ( f )   ( f  200)   ( f  200)  3 ( f  320)   ( f  320)
 The spectrum of the given signal is
shown in Fig.8.16.

1-May-17 Prof K.Venkat Reddy 73

EXAMPLES
 The spectrum of the sampled signal is

X s ( f )  fs  X ( f  nf )
n  
s

as shown in Fig.8.17.

1-May-17 Prof K.Venkat Reddy 74

37
 5/1/2017

EXAMPLES
 It is a periodic repetition of X(f) at
regular intervals of +500n.
 If the sampled signal is passed through
an ideal LPF with a cutoff frequency of
400Hz (shown in Fig.8.18a), the
frequency spectrum of the output signal
will be as shown in Fig.8.18b.

1-May-17 Prof K.Venkat Reddy 75

EXAMPLES
 So the frequency components that will
appear in the output are:
-320Hz, -300Hz, -200Hz, -180Hz,
180Hz, 200Hz, 300Hz, 320Hz

 The spectrum of the output signal is:

Y( f ) 
1
 ( f  200)   ( f  200)   ( f  300)   ( f  300)
T

1
3 ( f  180)   ( f  180)  3 ( f  320)   ( f  320)
T

1-May-17 Prof K.Venkat Reddy 76

38
 5/1/2017

EXAMPLES
 Taking inverse Fourier transform on both
sides, the output signal is:

y (t ) 
1
2 cos[2 (200)t ]  2 cos[2 (300)t ]
T

1
6 cos[2 (180)t ]  6 cos[2 (320)t ]
T

y(t ) 
1
2 cos 400t  2 cos 600t  6 cos 360t  6 cos 640t
T

1-May-17 Prof K.Venkat Reddy 77

SAMPLING OF BAND PASS SIGNALS

 The sampling theorem, which we have
discussed earlier, is called low pass
sampling theorem because it applies
only to low pas signals, i.e. signals for
which X(f)=0 for f>fm, where fm has a
finite value. Here the signal which is
band-limited to fm Hz has to be sampled
at least at 2fm samples per second if the
analog signal is to be reconstructed.
 In the case of band pass signals, that is
signals for which X(f)=0 for all
frequencies outside the range f1<f<f2,
where f1>0, this rule does
1-May-17 not Reddy
Prof K.Venkat apply. 78

39
 5/1/2017

SAMPLING OF BAND PASS SIGNALS

 For this class of signals, while there will
be no aliasing if fs>2f2, there might be
no aliasing even if fs<2f2 provided fs
satisfies certain conditions.
 Sampling of a band pass signal with
fs>2f2 to prevent aliasing has two
disadvantages.
1. The spectrum of the sampled signal will
have spectral gaps.
2. If f2 is large, the sampling rate is also
very large, and therefore, have
practical limitations.
1-May-17 Prof K.Venkat Reddy 79

SAMPLING OF BAND PASS SIGNALS

1-May-17 Prof K.Venkat Reddy 80

40
 5/1/2017

SAMPLING OF BAND PASS SIGNALS

 To over come this, the band pass
sampling theorem is defined as follows
 Let the bandwidth of the band pass
signal shown in Fig.8.29a be
B=f2-f1=2fm.
 Then the band pass sampling theorem
states that x(t) can be recovered
without any error what so ever from its
samples x(nT) taken at regular
intervals of T, if the sampling rate fs is
such that

1-May-17 Prof K.Venkat Reddy 81

SAMPLING OF BAND PASS SIGNALS

1 2 f2
fs  
T m
 Which is smaller than the Nyquist rate
2f2. Where m is the largest integer not
exceeding f2/B.
 If we assume that the highest
frequency component present in the
band pass signal is multiple of
bandwidth, i.e. f2=KB=K2fm, then the
band pass sampling theorem states
that

1-May-17 Prof K.Venkat Reddy 82

41
 5/1/2017

SAMPLING OF BAND PASS SIGNALS

 The band pass signal x(t) whose
maximum bandwidth is 2fm can be
completely represented into and
recovered from its samples if it is
sampled at the minimum rate of twice
the bandwidth.
 Hence for band pass signal of bandwidth
2fm, the minimum sampling rate is equal
to twice that of bandwidth, i.e.
fs=2xBW=4fm samples per second, or
T=1/4fm sec.

1-May-17 Prof K.Venkat Reddy 83

SAMPLING OF BAND PASS SIGNALS

 If the spectrum of the band pass signal
is X(ω) then the spectrum of the
sampled band pass signal is:
1 
 X (  2nB)
X s ( ) 
T n 
 Where Xs(ω) is the sum of the original
Fourier transform X(ω) and Shifted
replicas of X(ω) and then scaled by 1/T.
 Fig.8.29b shows the spectrum of the
original signal and sampled signal for
K=3.
1-May-17 Prof K.Venkat Reddy 84

42
 5/1/2017

SAMPLING OF BAND PASS SIGNALS

 From this fig, we can find that the
original signal can be recovered by
passing x(t) through an ideal band pass
filter with a pass band given by f1<|f|<f2
with a gain of T.
 It may be noted that sampling
frequencies higher than what is given by
the equation fs=2f2/m may not be
always permit recovery of x(t) without
distortion i.e. they may not be able to
avoid aliasing unless fs>2f2. The required
sampling rate for a band pass filter
depends on m, i.e. 1-May-17
on f2/B.
Prof Kf .Venkat Reddy
2 85

SAMPLING OF BAND PASS SIGNALS

 If f2>2f2, there will not be any aliasing
and perfect reconstruction is possible.
Also if f2=KB where K is an integer, a
sampling rate fs=2B would suffice and
will not produce any aliasing.
 Example: the spectral range of a
function extends from 5.6MHz to
6.8MHz. Find the minimum sampling
rate and maximum sampling time.

1-May-17 Prof Kf2.Venkat Reddy 86

43
 5/1/2017

SAMPLING OF BAND PASS SIGNALS

 Solution:
 Given frequency range = 5.6MHz to
6.8MHz
 Bandwidth B=2fm=(6.8-5.6)MHz =
1.2MHz
 Minimum sampling rate fs= 2B = 4fm =
2x1.2 = 2.4MHz
 Maximum sampling time Ts = 1/fs =
1/(2.4x106) = 0.417μs.

1-May-17 Prof Kf2.Venkat Reddy 87

44