DT Processing of Conti - Signals

Digital Processing of Digital Processing of
Continuous-Time Signals Continuous-Time Signals

• Digital processing of a continuous-time
• Conversion of a continuous-time signal into
signal involves the following basic steps:
digital form is carried out by an analog-to-
(1) Conversion of the continuous-time digital (A/D) converter
signal into a discrete-time signal,
• The reverse operation of converting a
(2) Processing of the discrete-time signal, digital signal into a continuous-time signal
(3) Conversion of the processed discrete- is performed by a digital-to-analog (D/A)
time signal back into a continuous-time converter
signal
1 2
Copyright © 2005, S. K. Mitra Copyright © 2005, S. K. Mitra
Digital Processing of Digital Processing of

Continuous-Time Signals Continuos-Time Signals
• Since the A/D conversion takes a finite • To prevent aliasing, an analog anti-aliasing
amount of time, a sample-and-hold (S/H) filter is employed before the S/H circuit
circuit is used to ensure that the analog • To smooth the output signal of the D/A
signal at the input of the A/D converter converter, which has a staircase-like
remains constant in amplitude until the waveform, an analog reconstruction filter
conversion is complete to minimize the is used
error in its representation
3 4
Digital Processing of
Sampling of Continuous-Time
Continuous-Time Signals
Complete block-diagram
Signals
Anti- Digital Reconstruction
• As indicated earlier, discrete-time signals in
aliasing S/H A/D processor D/A
filter filter many applications are generated by
sampling continuous-time signals
• Since both the anti-aliasing filter and the
reconstruction filter are analog lowpass • We have seen earlier that identical discrete-
filters, we review first the theory behind the time signals may result from the sampling
design of such filters of more than one distinct continuous-time
• Also, the most widely used IIR digital filter function
design method is based on the conversion of
5
an analog lowpass prototype 6
1
Sampling of Continuous-Time Sampling of Continuous-Time
Signals Signals
• In fact, there exists an infinite number of • If these conditions hold, then it is possible
continuous-time signals, which when to recover the original continuous-time
sampled lead to the same discrete-time signal from its sampled values
signal • We next develop this correspondence and
• However, under certain conditions, it is the associated conditions
possible to relate a unique continuous-time
signal to a given discrete-time signals
7 8
Effect of Sampling in the Effect of Sampling in the

Frequency Domain Frequency Domain
• Now, the frequency-domain representation of
• Let g a (t ) be a continuous-time signal that is
g a (t ) is given by its continuos-time Fourier
sampled uniformly at t = nT, generating the
transform (CTFT):
sequence g[n] where ∞
g [n] = g a (nT ), − ∞ < n < ∞ Ga ( jΩ) = ∫−∞ g a (t )e − jΩt dt
• The frequency-domain representation of g[n]
with T being the sampling period
is given by its discrete-time Fourier transform
• The reciprocal of T is called the sampling (DTFT):
frequency FT , i.e.,
G ( e jω ) = ∑ ∞
n = −∞ g[ n] e
− jω n
FT = 1
9 T 10

• p(t) consists of a train of ideal impulses
• To establish the relation between Ga ( jΩ)
with a period T as shown below
and G (e jω ) , we treat the sampling operation
mathematically as a multiplication of g a (t )
by a periodic impulse train p(t):
∞
p (t ) = ∑ δ(t − nT )
n = −∞ • The multiplication operation yields an
g a (t ) × g p(t ) impulse train: ∞
p (t ) g p (t ) = g a (t ) p(t ) = ∑ g a ( nT )δ(t − nT )
11 12
Copyright © 2005, S. K. Mitra n = −∞ Copyright © 2005, S. K. Mitra
2
Frequency Domain
Frequency Domain
• There are two different forms of G p ( jΩ):
• g p (t ) is a continuous-time signal consisting
of a train of uniformly spaced impulses with • One form is given by the weighted sum of
the impulse at t = nT weighted by the the CTFTs of δ(t − nT ) :
sampled value g a (nT ) of g a (t ) at that instant G p ( jΩ ) = ∑ ∞
n = −∞ g a ( nT ) e
− jΩnT
• To derive the second form, we note that p(t)

can be expressed as a Fourier series:
p(t ) = 1 ∑∞ e j ( 2π / T )kT = 1 ∑∞ e jΩT kt
T k = −∞ T k = −∞
where ΩT = 2π / T
13 14

• The impulse train g p (t ) therefore can be • Hence, an alternative form of the CTFT of
expressed as g p (t ) is given by
∞
⎛ ∞ ⎞
g p (t ) = ⎜⎜ 1 ∑ e jΩT kt ⎟⎟ ⋅ g a (t )
T
G p ( jΩ ) = 1
T ∑ Ga ( j (Ω − kΩT ) )
⎝ k = −∞ ⎠ k = −∞
• From the frequency-shifting property of the • Therefore, G p ( jΩ) is a periodic function of

CTFT, the CTFT of e jΩT kt g a (t ) is given by Ω consisting of a sum of shifted and scaled
Ga ( j (Ω − kΩT ) ) replicas of Ga ( jΩ) , shifted by integer
multiples of ΩT and scaled by 1
15 16 T
Effect of Sampling in the

Frequency Domain
Frequency Domain
• Assume g a (t ) is a band-limited signal with a
• The term on the RHS of the previous CTFT Ga ( jΩ) as shown below
equation for k = 0 is the baseband portion
of G p ( jΩ) , and each of the remaining terms
are the frequency translated portions of
G p ( jΩ ) • The spectrum P ( jΩ) of p(t) having a
sampling period T = Ω2 π is indicated below
• The frequency range T
Ω Ω
− T ≤Ω≤ T
2 2
• is called the baseband or Nyquist band
17 18
3
Frequency Domain Effect of Sampling in the
• Two possible spectra of G p ( jΩ) are shown
Frequency Domain
below • It is evident from the top figure on the
previous slide that if ΩT > 2Ω m , there is no
overlap between the shifted replicas of Ga ( jΩ)
generating G p ( jΩ)
• On the other hand, as indicated by the figure
on the bottom, if ΩT < 2Ω m , there is an
overlap of the spectra of the shifted replicas
of Ga ( jΩ) generating G p ( jΩ)
19 20

Frequency Domain
Frequency Domain • The spectra of the filter and pertinent
If ΩT > 2Ω m , g a (t ) can be signals are shown below
recovered exactly from g a (t ) by passing it
through an ideal lowpass filter H r ( jΩ) with
a gain T and a cutoff frequency Ωc greater
than Ω m and less than ΩT − Ω m as shown
below
21 22

• On the other hand, if ΩT < 2Ω m , due to the Sampling theorem - Let g a (t ) be a band-
overlap of the shifted replicas of Ga ( jΩ) , limited signal with CTFT Ga ( jΩ) = 0 for
the spectrum Ga ( jΩ) cannot be separated by Ω > Ωm
filtering to recover Ga ( jΩ) because of the • Then g a (t ) is uniquely determined by its
distortion caused by a part of the replicas samples g a (nT ) , − ∞ ≤ n ≤ ∞ if
immediately outside the baseband folded
back or aliased into the baseband ΩT ≥ 2Ω m
where ΩT = 2π / T
23 24
4
• The condition ΩT ≥ 2Ω m is often referred to • Given {g a (nT )}, we can recover exactly g a (t )
as the Nyquist condition by generating an impulse train
• The frequency ΩT is usually referred to as g p (t ) = ∑∞
n = −∞ g a ( nT )δ(t − nT )
2
the folding frequency and then passing it through an ideal lowpass
filter H r ( jΩ) with a gain T and a cutoff
frequency Ωc satisfying
Ω m < Ωc < (ΩT − Ω m )
25 26

• Oversampling - The sampling frequency is
• The highest frequency Ω m contained in g a (t ) higher than the Nyquist rate
is usually called the Nyquist frequency
• Undersampling - The sampling frequency is
since it determines the minimum sampling
lower than the Nyquist rate
frequency ΩT = 2Ω m that must be used to
fully recover g a (t ) from its sampled version • Critical sampling - The sampling frequency
is equal to the Nyquist rate
• The frequency 2Ω m is called the Nyquist
rate • Note: A pure sinusoid may not be
recoverable from its critically sampled
27 28
version

• In digital telephony, a 3.4 kHz signal • In high-quality analog music signal
bandwidth is acceptable for telephone processing, a bandwidth of 20 kHz has been
conversation determined to preserve the fidelity
• Here, a sampling rate of 8 kHz, which is • Hence, in compact disc (CD) music
greater than twice the signal bandwidth, is systems, a sampling rate of 44.1 kHz, which
used is slightly higher than twice the signal
bandwidth, is used
29 30
5
Frequency Domain
Frequency Domain
• These three transforms are plotted below
• Example - Consider the three continuous-
time sinusoidal signals:
g1(t ) = cos(6πt )
g 2 (t ) = cos(14πt )
g3 (t ) = cos( 26πt )
• Their corresponding CTFTs are:
G1( jΩ) = π[δ(Ω − 6π) + δ(Ω + 6π)]
G2 ( jΩ) = π[δ(Ω − 14π) + δ(Ω + 14π)]
31
G3 ( jΩ) = π[δ(Ω − 26π) + δ(Ω + 26π)] 32

Frequency Domain
Frequency Domain
• Plots of the 3 CTFTs are shown below
• These continuous-time signals sampled at a
rate of T = 0.1 sec, i.e., with a sampling
frequency ΩT = 20π rad/sec
• The sampling process generates the
continuous-time impulse trains, g1 p (t ) ,
g 2 p (t ) , and g3 p (t )
• Their corresponding CTFTs are given by
Glp ( jΩ) = 10∑∞
k = −∞ Gl ( j (Ω − kΩT ) ), 1 ≤ l ≤ 3
33 34

Frequency Domain
Frequency Domain
• These figures also indicate by dotted lines
the frequency response of an ideal lowpass • Moreover, the reconstructed output is
filter with a cutoff at Ωc = ΩT / 2 = 10π and precisely the original continuous-time
a gain T = 0.1 signal
• The CTFTs of the lowpass filter output are • In the other two cases, the sampling rate
also shown in these three figures does not satisfy the Nyquist condition,
resulting in aliasing and the filter outputs
• In the case of g1(t ), the sampling rate
are all equal to cos(6πt)
satisfies the Nyquist condition, hence no
aliasing
35 36
6
Frequency Domain
Frequency Domain
• Note: In the figure below, the impulse
appearing at Ω = 6π in the positive • We now derive the relation between the
frequency passband of the filter results from DTFT of g[n] and the CTFT of g p (t )
the aliasing of the impulse in G2 ( jΩ) at • To this end we compare
Ω = −14π G ( e jω ) = ∑ ∞ − jω n
n = −∞ g[ n] e
• Likewise, the impulse appearing at Ω = 6π with
in the positive frequency passband of the G p ( jΩ ) = ∑ ∞n = −∞ g a ( nT ) e
− jΩnT
filter results from the aliasing of the impulse and make use of g[n] = g a (nT ), − ∞ < n < ∞
in G3 ( jΩ) at Ω = 26π
37 38

• Observation: We have we arrive at the desired result given by
G ( e jω ) = G p ( jΩ ) ∞
G (e jω ) = T1 ∑ Ga ( jΩ − jkΩT )
Ω =ω / T
or, equivalently, k = −∞ Ω =ω / T
∞
G p ( jΩ ) = G ( e jω ) = 1 ∑ Ga ( j ω − jkΩT )
ω=ΩT T T
• From the above observation and k = −∞
∞ ∞
G p ( jΩ ) = 1
T ∑ Ga ( j (Ω − kΩT ) ) = 1 ∑ Ga ( j ω − j 2 π k )
T T T
k = −∞ k = −∞
39 40

Frequency Domain
Frequency Domain
• The relation derived on the previous slide
can be alternately expressed as • Now, the CTFT Gp ( jΩ) is a periodic
function of Ω with a period ΩT = 2π / T
G (e jΩT ) = T1 ∑∞ k = −∞ Ga ( jΩ − jkΩT )
• From • Because of the mapping, the DTFT G (e jω )
G ( e jω ) = G p ( jΩ ) is a periodic function of ω with a period 2π
Ω =ω / T
or from
G p ( jΩ ) = G ( e jω )
ω=ΩT
it follows that G (e jω ) is obtained from Gp ( jΩ)
41
by applying the mapping Ω = ω 42
T
7
Recovery of the Analog Signal Recovery of the Analog Signal
• We now derive the expression for the output
g^ a (t ) of the ideal lowpass reconstruction • Thus, the impulse response is given by
filter H r ( jΩ) as a function of the samples hr (t ) = 1 ∞ H ( jΩ) e jΩt dΩ = T Ω c e jΩt dΩ
∫
2 π −∞ r 2 π −Ωc ∫
g[n]
sin(Ωct )
• The impulse response hr (t ) of the lowpass = , −∞ ≤t ≤ ∞
ΩT t / 2
reconstruction filter is obtained by taking • The input to the lowpass filter is the
the inverse DTFT of H r ( jΩ): impulse train gp(t ) :
T , Ω ≤ Ωc
H r ( jΩ) = ⎧⎨ g p (t ) = ∑∞
⎩ 0, Ω > Ωc n = −∞ g[ n] δ(t − nT )
43 44

• Therefore, the output g^ a (t ) of the ideal • The ideal bandlimited interpolation process
lowpass filter is given by: is illustrated below
∞
g^ a (t ) = hr (t ) * g p (t ) = ∑ g[n]hr (t − nT )
n = −∞
• Substituting hr (t ) = sin(Ωct ) /(ΩT t / 2) in the

above and assuming for simplicity
Ωc = ΩT / 2 = π / T , we get
∞ sin[ π(t − nT ) / T ]
45
g^ (t ) =
a ∑ g[ n]
π(t − nT ) / T Copyright © 2005, S. K. Mitra 46
n = −∞ Copyright © 2005, S. K. Mitra

• It can be shown that when Ωc = ΩT / 2 in
sin( Ω ct )
hr (t ) = • The relation
ΩT t / 2
h r(0) = 1 and h r(nT ) = 0 for n ≠ 0 g^ a (rT ) = g[r ] = g a (rT )
• As a result, from holds whether or not the condition of the
sin[ π(t − nT ) / T ]
g^ a (t ) = ∑∞n = −∞ g[ n] π(t − nT ) / T
sampling theorem is satisfied
we observe • However, g^a (rT ) = g a (rT ) for all values of
g a (rT ) = g[r ] = g a (rT ) t only if the sampling frequency ΩT satisfies
the condition of the sampling theorem
for all integer values of r in the range
47 −∞ < r <∞ 48
8
Implication of the Sampling Implication of the Sampling
Process Process
• Consider again the three continuous-time • From the plot, it is apparent that we can
signals: g1(t ) = cos(6πt ) , g 2 (t ) = cos(14πt ) , recover any of its frequency-translated
and g3 (t ) = cos(26πt ) versions cos[(20k ± 6)π t] outside the
• The plot of the CTFT G1p ( jΩ) of the baseband by passing g1p (t ) through an ideal
sampled version g1p (t ) of g1(t ) is shown analog bandpass filter with a passband
below centered at Ω = ( 20k ± 6) π
49 50
Implication of the Sampling Implication of the Sampling

Process Process
• For example, to recover the signal cos(34πt), • Likewise, we can recover the aliased
it will be necessary to employ a bandpass baseband component cos(6πt) from the
filter with a frequency response sampled version of either g 2 p (t ) or g3 p (t )
0.1, (34 − ∆ )π ≤ Ω ≤ (34 + ∆ )π by passing it through an ideal lowpass filter
H r ( jΩ) = ⎧⎨ with a frequency response:
⎩ 0, otherwise
0.1, (6 − ∆ )π ≤ Ω ≤ (6 + ∆ )π
where ∆ is a small number H r ( jΩ) = ⎨⎧
⎩ 0, otherwise
51 52
Implication of the Sampling Sampling of Bandpass Signals

Process
• The conditions developed earlier for the
• There is no aliasing distortion unless the unique representation of a continuous-time
original continuous-time signal also signal by the discrete-time signal obtained
contains the component cos(6πt) by uniform sampling assumed that the
• Similarly, from either g 2 p (t ) or g3 p (t ) we continuous-time signal is bandlimited in the
can recover any one of the frequency- frequency range from dc to some frequency
translated versions, including the parent Ωm
continuous-time signal g 2(t ) or g3(t ) as the • Such a continuous-time signal is commonly
case may be, by employing suitable filters referred to as a lowpass signal
53 54
9
Sampling of Bandpass Signals Sampling of Bandpass Signals
• There are applications where the continuous-
time signal is bandlimited to a higher • However, due to the bandpass spectrum of
frequency range Ω L ≤ Ω ≤ Ω H with Ω L > 0 the continuous-time signal, the spectrum of
the discrete-time signal obtained by sampling
• Such a signal is usually referred to as the
will have spectral gaps with no signal
bandpass signal
components present in these gaps
• To prevent aliasing a bandpass signal can of
• Moreover, if Ω H is very large, the sampling
course be sampled at a rate greater than
rate also has to be very large which may not
twice the highest frequency, i.e. by ensuring
be practical in some situations
ΩT ≥ 2Ω H
55 56

• A more practical approach is to use under- • We choose the sampling frequency ΩT to
sampling satisfy the condition
2Ω
• Let ∆Ω = Ω H − Ω L define the bandwidth of ΩT = 2(∆Ω) = H
M
the bandpass signal which is smaller than 2Ω H , the Nyquist
• Assume first that the highest frequency Ω H rate
contained in the signal is an integer multiple • Substitute the above expression for ΩT in
of the bandwidth, i.e., ∞
Ω H = M (∆Ω) G p ( jΩ ) = 1
T ∑ Ga ( j (Ω − k ΩT ))
57 58
k = −∞

• This leads to • Figure below illustrate the idea behind
G p ( jΩ ) = 1 ∑ ∞ (
k = −∞ Ga jΩ − j 2k (∆Ω)
) Ga ( jΩ)
T
• As before, G p( jΩ) consists of a sum of Ga ( jΩ)
and replicas of Ga ( jΩ) shifted by integer
Ω
multiples of twice the bandwidth ∆Ω and − ΩH − ΩL 0 ΩL ΩH
scaled by 1/T G p ( jΩ )
• The amount of shift for each value of k

ensures that there will be no overlap
Ω
− ΩH − ΩL ΩL ΩH
between all shifted replicas no aliasing 0
59 60
10
Sampling of Bandpass Signals
• As can be seen, g a (t ) can be recovered from
g p (t ) by passing it through an ideal
bandpass filter with a passband given by
Ω L ≤ Ω ≤ Ω H and a gain of T
• Note: Any of the replicas in the lower
frequency bands can be retained by passing
g p (t ) through bandpass filters with
passbands Ω L − k ( ∆Ω) ≤ Ω ≤ Ω H − k (∆Ω) ,
1 ≤ k ≤ M − 1 providing a translation to
61 lower frequency ranges
Copyright © 2005, S. K. Mitra
11

DT Processing of Conti - Signals

Uploaded by

Document Information

Original Title

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

DT Processing of Conti - Signals

Uploaded by

Copyright:

Available Formats

Digital Processing of Digital Processing of

Continuous-Time Signals Continuous-Time Signals

Digital Processing of Digital Processing of

Effect of Sampling in the Effect of Sampling in the

Effect of Sampling in the Effect of Sampling in the

• To derive the second form, we note that p(t)

Effect of Sampling in the Effect of Sampling in the

• From the frequency-shifting property of the • Therefore, G p ( jΩ) is a periodic function of

Effect of Sampling in the

Effect of Sampling in the

Effect of Sampling in the Effect of Sampling in the

Effect of Sampling in the Effect of Sampling in the

Effect of Sampling in the Effect of Sampling in the

Effect of Sampling in the Effect of Sampling in the

Effect of Sampling in the Effect of Sampling in the

Effect of Sampling in the Effect of Sampling in the

Effect of Sampling in the Effect of Sampling in the

Recovery of the Analog Signal Recovery of the Analog Signal

• Substituting hr (t ) = sin(Ωct ) /(ΩT t / 2) in the

Recovery of the Analog Signal Recovery of the Analog Signal

Implication of the Sampling Implication of the Sampling

Implication of the Sampling Sampling of Bandpass Signals

Sampling of Bandpass Signals Sampling of Bandpass Signals

Sampling of Bandpass Signals Sampling of Bandpass Signals

• The amount of shift for each value of k

You might also like