Chapter 4.

Sampling of Continuous-Time Signals

40I t d ti 4.0 Introduction 4.1 Periodic Sampling 4.2 Frequency Representation of Sampling 4 3 Reconstruction from Discrete 4.3 Discrete-Time Time Samples 4.4 Changing the Sampling Rate 4.5 Digital Processing of Analog Signals

DSP

4-1

4.0 Introduction
Q estion to answer: ans er how ho to approximate appro imate a contin o s Question continuous (analog) linear system by a digital system? Notations Notations:
Signals: time-domain frequency-domain Systems: time-domain frequency-domain continuous-time discrete-time

xc ( t )
X c ( j )
continuous-time

x[n ]

X ( e j )
discrete-time

hc (t ) H c ( j )

h[n ]

H (e j )
4-2

DSP

4.1 Periodic Sampling

time representation of a A typical method of obtaining a discrete discrete-time continuous-time signal is through periodic sampling.

x[n ] = xc (nT T ), ) - < n < .

T is the sampling period f s = 1 / T is the sampling frequency (samples per second) = 2 / T is the sampling frequency (radians per second)
s

An ideal continuous-to-discrete-time ( (C/D) ) converter

DSP

4-3

Mathematical Representation of Sampling

s (t )

xc ( t )

xs ( t )

Conversion from impulse train to discretetime sequence

x[n] = xc (nT )

s(t ) =

n =

(t nT )

(the periodic impulse train)

xs (t ) = xc (t ) s(t ) = xc (t ) (t nT )
n =

(modulation)

xs ( t ) =
DSP

n =

x (nT ) (t nT )
c

(sifting property)
4-4

Periodic Sampling Examples

DSP

4-5

4.2 Frequency-Domain Representation of Sampling: Time-Domain

We modulate the periodic impulse train with the original continuous-time signals, obtaining

xs (t ) = xc (t ) s (t ) = xc (t ) (t nT )
n =

n =

x (nT ) (t nT )
c
4-6

DSP

Frequency-Domain Representation
Given the Fourier transform of the impulse train as
2 s(t ) = (t nT ) S ( j) = T n =

2 ( k s ) (where s = ) T k =

Since

1 x s ( t ) = x c ( t ) s ( t ) X s ( j ) = X c ( j ) * S ( j ) 2

Then

1 2 X s ( j ) = X c ( j ) * 2 T

k =

( k )
s

1 = X c ( j ( k s )) T k =
DSP 4-7

Observations of Frequency-Domain Representation of Sampling

Thi equation ti provides id th l ti hi b t th i This the relationship between the F Fourier transform of continuous-time signal and discrete-time signal

1 X s ( j) = X c ( j ( k s )) T k =

periodically y repeated p and scaled copies p of the X s ( j) consists of p Fourier transform of The copies of X c ( j) are shifted by integer multiples of the sampling frequency s. All copies of replicated spectrums are superimposed to produce the Fourier transform of the sampled signal signal.
4-8

.) xc (t ), i.e., X c ( j

DSP

DSP

4-9

Sampling Rate and Bandwidth

limited Given the signal of band band-limited

X ( j) = 0, > N
There is no overlap between replicated spectrums, when we have the sampling p g rate as following g

s > 2 N

That means we CAN reconstruct the continuous-time signal with an ideal low-pass filter.

There will be aliasing distortion, or aliasing when s < 2 N

That means we CANNOT reconstruct the continuous-time signal from its samples.
4-10

DSP

How to Reconstruct a Signal?

Ideal low-pass Filtering

DSP

4-11

How to Reconstruct a Signal? (Cont'd)

sampling

Original Signal Ideal low-pass filtering

Discrete-Time Discrete Time Signal

Ideal low-pass p Filtering g

DSP

Reconstructed Signal g
4-12

Sampling and Reconstruction Example

Gi i l Given a signal

xc (t ) = cos 0t

What is the Fourier transform of the given signal? Use the Euler E ler equation eq ation, we e kno know that

According to continuous Fourier transform, we know

1 j0t xc (t ) = cos 0t = e + e j0t 2

x (t ) = e j0t X ( j) = 2 ( 0 )

Therefore, the Fourier transform of the g given signal g is

X c ( j) = ( ( 0 ) + ( + 0 ) )
DSP 4-13

Sampling and Reconstruction Example (No Aliasing)

Original Signal

Sampled Signal

Reconstructed Signal

1 j0t xr (t ) = e + e j0t = cos 0t 2

DSP 4-14

Sampling and Reconstruction Example (With Aliasing)

Original Signal

Sampled Signal

1 j ( s 0 ) t xr ( t ) = e + e j ( s 0 ) t = cos( ( s 0 )t 2
DSP 4-15

Reconstructed Signal

Nyquist Sampling Theorem

S th t xa (t ) X a () is i band-limited b d li it d t Suppose that to a f frequency interval [ N , N ], i.e.,

X () = 0 for N
Then x(t) can be exactly y reconstructed from equidistant q 2 samples xd [n ] = xa ( nTs ) = xa ( 2n / s ), if s = 2 N , Ts where T = 2 / is the sampling period period, s is the
s s

sampling frequency (radians per second), N is referred to as the Nyquist frequency, and 2 N is called the Nyquist rate.
DSP 4-16

How to obtain discrete-time Fourier transform (DTFT)?

Gi th sampled l d signal i l as Given the

xs (t ) =

n =

x (nT ) (t nT )
c

Since we have the following continuous-time Fourier transform (CTFT) pair

(t nT ) e

jnT

Thus we have the continuous-time Fourier transform of the sampled p signal g as

X s ( j) =
DSP

n =

x (nT )e
c

jTn

4-17

How to obtain discrete-time Fourier transform (DTFT)? (Cont'd)

Si l ti hi b t th l d Since we k know th the relationship between the sampled signal xc ( nT ) and the discrete-time sequence x[n ]

x[n ] = xc ( nT T)

We also have the DTFT of

x[n ]
n =

is defined as

X ( e j ) =
By comparing with

jn x [ n ] e

X s ( j) =
DSP

n =

jTn x ( nT ) e c

4-18

How to obtain discrete-time Fourier transform ( (DTFT)? ) (Cont'd) ( )

As we compare the following two equations

X (e ) =

n =

x[n]e

jn

X s ( j) =
= T

n =

T )e x (nT
c

jTn

X s ( j) = X ( e j )

= X ( e jT ).

1 X (e ) = X c T k =
j

2k j T T

( = T )

1 X s ( j) = X c ( j ( k s ) ) T k =

DTFT representation of sampling !

DSP

X (e j ) is simply a frequency-scaled version of X s ( j) with the frequency scaling specified by = T . This scaling is a normalization of the frequency axis so that the frequency = s in X s ( j) is normalized to = 2 for X (e j ) .
4-19

Example 4.1 (Without Aliasing)

time signal xc (t ) = cos( If we sample the continuous continuous-time (4000t ) with sampling period T=1/6000. Continuous-time Fourier transform X s ( j )

Discrete-time Fourier transform

X ( e j )

Problem Analysis

Fourier transform of the original signal 0 = 4000 .

X c ( j) = ( 4000 ) + ( + 4000 )

Sampling frequency s = 2 / T = 12000 . Fourier transforms of the sampled signal

1 X s ( j) = X c ( j ( k s )) T k =
DSP

1 2 X (e ) = X c ( j( k )) T k = T T
j

4-20

Example 4.1 (Cont'd)

( = T )

( / T ) = T ( )

DSP

4-21

Example 4.2 (With Aliasing)

time signal xc (t ) = cos( If we sample the continuous continuous-time (16000t ) with sampling period T=1/6000. Continuous-time Fourier transform X s ( j )

Discrete-time Fourier transform

X ( e j )

Problem Analysis

Fourier transform of the original signal 0 = 16000 .

X c ( j) = ( 16000 ) + ( + 16000 )

Sampling frequency s = 2 / T = 12000 . Fourier transforms of the sampled signal are exactly same as the previous i one, why? h ?

x[n ] = cos(16000n / 6000) = cos(2n + 4000n / 6000) = cos(2n / 3)

DSP 4-22

k=-2 k=0

k=1

k=-1

k=2

k=0

( = T )

( / T ) = T ( )

0 = 16000
DSP 4-23

Example 4.3 (with Aliasing)

time signal If we sample the continuous continuous-time with sampling period T=1/1500. Continuous-time Fourier transform X ( j ) s Discrete-time Fourier transform X ( e j ) Problem Analysis

xc (t ) = cos(4000t )

Fourier transform of the original signal

X c ( j) = ( 4000 ) + ( + 4000 )

Sampling frequency s = 2 / T = 3000 . The discrete-time Fourier transform is the same as previous one. Wh ? Why?

cos(4000n / 1500) = cos(2n + 1000n / 1500) = cos(2n / 3)

DSP 4-24

k=-2 k=0 0

k=1

k=-1

k=2

k=0

( = T )

( / T ) = T ( )

DSP

4-25

4.3 Reconstruction of a Band-limited Signal from Its Samples

met and if the If the conditions of the sampling theorem are met, modulated impulse train is filtered by an appropriate low-pass filter, then the Fourier transform of the filter output p will be identical to the Fourier transform of the original signal. Given a sequence of samples x[n], we form the impulse train

xs ( t ) =

n =

x[n] (t nT )

If the impulse train is the input to an ideal low-pass continuoustime filter with impulse response hr (t )
xr (t ) = xs (t ) * hr (t ) = x[n ] (t nT ) * hr (t ) = x[n ]hr (t nT ) n = n =

DSP

4-26

4.3.1 Ideal Reconstruction System

DSP

4-27

Ideal Reconstruction System (Cont'd)

Frequency Response

Impulse Response

sin(t / T ) t hr (t ) = = sinc t / T T
DSP 4-28

Ideal Reconstruction System (Cont'd)

The ideal reconstruction system is denoted by

xr ( t ) =

n =

x[n]h (t nT )
r

sin(t / T ) t = sinc hr (t ) = t / T T

sin[ (t nT ) / T ] xr (t ) = x[n ] (t nT ) / T n =

If x[n ] = xc ( n nT ) and we have

X c ( j) = 0 for / T

then

x r ( t ) = xc ( t )
DSP 4-29

Ideal Band-limited Interpolation

Th ideal id l l filt i t l t b t th i l The low-pass filter interpolates between the impulses of x[n ] to construct a continuous-time signal

sin[ (t nT ) / T ] xr (t ) = x[n ] (t nT ) / T n =

If there is no aliasing, the ideal low-pass filter interpolates correct reconstruction between the samples. However, the ideal low-pass p filter has infinite length g which is not realizable in practice. Finite length low-pass filtering will result in some reconstruction error.
DSP 4-30

Ideal Band-limited Interpolation (Cont'd)

Original continuous continuous-time time signal

Sampled signal

Reconstructed signal

DSP

4-31

Ideal D/C Converter

DSP

4-32

Ideal D/C Converter (Cont'd)

The properties of the ideal D/C converter are most easily seen in the frequency-domain. sin[ [ (t nT ) / T ] xr (t ) = x[n ] xr (t ) = x[n ]hr (t nT T) (t nT ) / T n = n =

X r ( j) =

n =

jTn x [ n ] H ( j ) e . r

Linearity of continuous-time Fourier transform Time shifting leads to an exponential factor in the Fourier transform Discrete-time Fourier Transform (DTFT) of x[n]
4-33

X r ( j) = H r ( j) X (e
DSP

jT

).

Can you get the original signal back?

lo pass filter selects the base period of the res lting The ideal low-pass resulting periodic Fourier transform X ( e jT ) and compensates for the 1/T scaling inherent in sampling sampling. If the sequence x[n] has been obtained by sampling a bandg at the Nyquist yq rate or higher, g , the reconstructed limited signal signal will be equal to the original band-limited signal. If there is aliasing g during g the sampling, p g the reconstructed signal g will be distorted, see Examples 4.2 and 4.3. In any case, the output of the ideal D/C converter is always band-limited to at most the cut-off frequency of the low-pass filter, which is taken to one-half the sampling frequency.
DSP 4-34

4.3.2 Zero-order Hold D/A Conversion

approximated digitally only if the An analog system is well well-approximated digital output is carefully transformed into analog form.

yd [n ]

d (e j )

g a (t )

a ()

ya ( t )

Comments:

j C Compensation ti with ith either ith d ( e ) or a () - not tb both th. The D/A block g a (t ) is not filtering - it is weighting.

No compensation is needed if

g a (t )

is the ideal reconstructor reconstructor.

4-35

DSP

Impulse Response of Zero-order Hold

This is what is usually done in practice practice, here
where

ya (t ) =
d

n =

x [n ]g
d

g a (t )
d

(t nTs )

1 0

It holdsy

[n ] at a constant level over each sampling period.

Ts

yd [n ]
1

ZOH Z.O.H

ya ( t )
1

-3
DSP

-2 -1

-3

-2 -1

n
4-36

Note: Z.O.H. introduces high frequencies, see sharp edges.

Frequency Response of ZOH

Frequency response of ZOH is a sinc function function.

sin(Ts / 2) jTs / 2 Ga ( j) = e ( / 2)
The high frequencies in the reconstructed signal (sharp steps) are introduced from side-lobes as follows.
Ga ( )
Ideal Z.O.H

DSP

Frequency response of Z.O.H

2 Ts

Ts

Ts

2 Ts
4-37

Compensation of ZOH
The phase response of ZOH corresponds to an advance time shit of T/2 which cannot be compensated and usually neglected. The magnitude g response can be compensated as follows.

(Ts / 2) ; a () = sin( i (Ts / 2) 0;

<

a ()
1

Ts else

2 2 Ts Ts Ts Ts Ideal compensation reconstruction filter

DSP 4-38

Physical Configuration for ZOH D/A Conversion

The D/A converter followed by an ideal compensated reconstruction filter is shown as follows.

DSP

4-39

4.4 Changing the Sampling Rate

necessar to change the sampling rate of a discrete It is often necessary discretetime signal, i.e., to obtain a new discrete-time representation of the underlying continuous-time continuous time signals signals.

x[n ] = xc ( nT ) and x ' [n ] = xc ( nT ' ) where T T '

It is of interest to consider methods of changing the sampling rate that involve discrete-time operations.

x[n ] x ' [n ]
DSP 4-40

Sampling Rate Change Examples (Down-sampling)

What happened during down sampling? down-sampling?

DSP 4-41

4.4.1 Sampling Rate Reduction by an Integer Factor (Down-sampling)

The sampling rate of a sequence can be reduced by "sampling it" by defining a new sequence

xd [n ] = x[nM ] = xc ( nMT ).

DSP

4-42

Frequency Representation of Down-Sampling [W-D]

First recall that the DTFT of x[n ] = xc ( nT T ) is

1 X (e ) = X c T k =
j

2k j T T

Similarly the DTFT of Similarly,

xd [n ] = x[nM ] = xc ( nMT ) is
Xc r =

1 X d (e ) = MT
j

2r j MT MT

Questions: what is the relationship Q p between them?

X ( e j ) X d ( e j )
DSP 4-43

Frequency Representation of Down-Sampling (Cont'd)

We can represent

1 X d (e ) = MT
j

Xc r =

2r j MT MT

r is still an interger ranging from -inf and inf

( r = i + kM )

< k < , 0 i M 1

1 j X d (e ) = M

M 1

1 2 ( kM + i ) X c j MT i =0 T k = MT
M 1

1 j X d (e ) = M
DSP

1 2k 2i X c j T MT i =0 T k = MT
4-44

Frequency Representation of Down-Sampling (Cont'd)

We then have

1 j X d (e ) = M
j j

M 1

1 Xc i =0 T k =

2i 2k j T MT

We know that
1 2 k k 1 2 = X c j X((e e )) = X (DTFT from CTFT) T kk= T T T = T

1 j ( 2 i ) / M ) = T X c X (e k =

2i 2k j T MT
j ( 2 r ) / M

Therefore, we have

1 j X d (e ) = M
DSP

M 1 r =0

X (e

)
4-45

Frequency Representation of Down-Sampling (Cont'd)

We can have the following conclusions by observing
M 1 r =0 j ( 2 r ) / M ( ) X e

1 j X d (e ) = M

There is a strong analogy between X d ( e j ) and X ( e j ). ) X d ( e j ) can be composed of M copies of the periodic Fourier transform X ( e j ) , frequency scaled by M and shifted by integer multiples lti l of f 2.

X d ( e j ) is periodic with period 2.

Aliasing can be avoided if

X ( e j ) = 0, N (band - limited) (narrow - banded) ( b d d) 2 / M 2N

DSP 4-46

2 / T = 4 N s = 4 N

N = N T = / 2

-4

DSP

4-47

Frequency Representation of Down-Sampling: Example

DSP

4-48

Down-sampling after Pre-filtering

sampling we need to reduce If aliasing occurs during down down-sampling, the band-width of signal x[n] prior to down-sampling. Signal x[n] will be pre-filtered pre filtered by an ideal low-pass low pass filter with cut-off frequency /M.

DSP

4-49

Down-sampling after Pre-filtering (Example)

hlp [n ] =

sin c n sin( (n / M ) = n n

DSP

4-50

4.4.2 Increasing the Sampling Rate by an Integer Factor

Consider a signal x[n] whose sampling rate we wish to increase by a factor of L (up-sampling).

] n = kL and k Z x[n / L], xe [ n ] = 0, otherwise

For example,

xe [n ] = (1 0 2 0 3 0 4 0 5 0) ( n = 0, 1, 2, 3, 4, 5, 6, 7, 8, 9 and L = 2)
Or equivalently,
xe [ n ] =
DSP
k =

x[n ] = (1 2 3 4 5) ( n = 0, , 1, , 2, , 3, , 4) )

x[k ] [n kL]

(not LTI convolution)

4-51

Sampling Rate Change Examples (Up-sampling) Demo

> 8 kHz) Speech (4kHz ->

DSP

4-52

Frequency Representation of Up-Sampling

Th Fourier F i transform t f f the th up-sampled l d signal i li The (DTFT) of is
jn j X e ( e ) = ( x[k ] [n kL ]) e n = k =

k =

jkL jL x [ k ] e = X ( e )

We can see the Fourier transform of the output of the expander is a frequency-scaled frequency scaled version of the Fourier transform of the input, i.e, is replaced by L.

DSP

4-53

Frequency Representation of Up-Sampling (Example)

DSP

4-54

General System for Up-sampling

To fill missing samples sampling is samples, the operation of up up-sampling therefore considered to be synonymous with interpolation.

sin c n sin n / L hlp [n ] = L = n / L n

G i Gain
DSP 4-55

Ideal Low-pass Filtering after Up-sampling

As in the case of D/C converter converter, it is possible to obtain an interpolation formula with an ideal low-pass filter as

sin[ ( n kL) / L] xi [n ] = x[k ]hlp [n kL] = x[k ] (n kL) / L k = k =

The impulse response of the low pass filter has properties

hi [0] = 1 sin(n / L) hi [n ] = hi [n ] = 0 , n = L,2 L,3L,.... n / L

Thus for the ideal low-pass interpolation filter, we have

xi [n ] = x[n / L], ] n = L,2 L,3L,...

DSP 4-56

Frequency Representation of Up-Sampling and Ideal Low-pass Filtering (Example)

DSP

4-57

Linear Interpolation after Up-sampling

Linear interpolation can be accomplished by the

1 n / L, n L hlin = otherwise 0,

DSP

4-58

Linear Interpolation after Up-sampling (Example)

1 sin(L / 2) H lin ( e j ) = L sin( / 2)

DSP

4-59

Linear Interpolation after Up-sampling (Example, Cont'd)

Please note that
hi [0] = 1 1 n / L, n L hlin li [ n ] = hi [n ] = 0 , n = L,2 L,3L,.... otherwise 0,

So that

xi [n ] =

k =

x[k ]h

lin

[n kL]

xlin [n ] = x[n / L], n = L,2 L,3L, ,...

The amount of distortion in the intervening samples can be gauged by comparing the frequency response of the linear interpolator with that of the ideal low-pass interpolator, as

1 sin(L / 2) H lin li ( e ) = L sin( / 2 )

j
DSP

4-60

4.5 Digital Processing of Analog Signals

There are only two approaches to avoiding aliasing

Sample at a faster rate - perhaps not possible (why?). Use an anti-aliasing g filter.

DSP

4-61

How to reduce aliasing?

anti aliasing filter is a low-pass low pass analog filter (LPF) that An anti-aliasing is applied to the continuous signal prior to sampling.

The idea is sample: p remove the high g frequencies q . The ideal frequency response of the anti-aliasing filter is an ideal low-pass filter as > / 2 ,
s

where the cut off

1; F = 0;

1 c c < s = Ts 2 > c

Even the LPF destroys information, it is better than the aliasing effect effect. Ideal sampler

xa (t )
DSP

Fa ()
Anti-aliasing filter (low-pass)

xd (n)
Ts
4-62

Anti-aliasing: Formulation
aliasing the sampled signal becomes With anti anti-aliasing,

xd [n ] = xa (nT Ts ) * f a ( nT Ts )

1 X d (e ) = Ts
j

2m 2m X Fa a T T m = s s

The repeated spectra X a () Fa () will not fold f or overlap. If Fa () is an ideal LPF with cutoff c , then
1 X d ( e j ) = Ts X a (); 2m where X a = Xa T m = s 0;

c > c

Usually, an ideal LPF cannot be realized and must be approximated.

DSP 4-63

Anti-aliasing: Example-1
X a ()
Analog Signal Spectrum

X a () Fa ()
Anti-aliased Spectrum

Sampled Signal Spectrum (without aliasing)

c
c < s / 2

X d (e j )

2
DSP

cTs

cTs

2
4-64

Anti-aliasing: Example-2

Original image

Down-sampling with aliasing

Down-sampling with anti-aliasing

DSP

4-65

Anti-aliasing: Digital Filter Output

R ll th ll system t fi t t Recall the overall of interest:
xd [ n ]
xa ( t )

yd [n ]
H d ( e j )

ya ( t )

H a ()

The response p g filter H d ( e j ) Yd ( e j ) of the digital

1 Yd ( e ) = Ts
j

2m j X H ( e ) a d m = Ts

without anti-aliasing

1 Yd ( e ) = Ts
j

2m 2m j X F H ( e ) a a d m = Ts Ts

with anti-aliasing

DSP

4-66