7.
Sampling and Reconstruction
7.1. Sampling (7.1)
7.2. Reconstruction (7.2)
7.3. Sampling Theorem (7.1, 7.3)
7.1. Sampling
7.1.1. Time-Domain Expression of Sampling
A continuous-time signal is converted into a discrete-time signal
by sampling. Sampling can be periodic or not. We consider periodic
sampling only. It is expressed as
x(n)=x
c
(t)|
t=nT
=x
c
(nT). (7.1)
x
c
(t) is a continuous-time signal. x(n) is the corresponding discrete-
time signal. T is the sampling interval.
7.1.2. Frequency-Domain Expression of Sampling
Let x
c
(t) be a continuous-time signal, X
c
(O) be the continuous-
time Fourier transform of x
c
(t), x(n) be a discrete-time signal, and
X(e) be the discrete-time Fourier transform of x(n). If x(n)=x
c
(nT),
then
t
x
c
(t)
nT
x(n)
O
X
c
(O)
X(OT)
Figure 7.1. Spectral Relation in Sampling.
2t/T T
O
A
A/T
.
T
m 2
X
T
1
) ( X
m
c
=
|
.
|
\
|
t e
= e (7.2)
Letting e=OT in (7.2), we obtain
. m
T
2
X
T
1
) T ( X
m
c
=
|
.
|
\
|
t
O = O
(7.3)
X(OT) is the discrete-time Fourier transform of x(n) expressed in O.
(7.3) shows that X(OT) equals X
c
(O) extended with period 2t/T and
divided by T (figure 7.1).
(7.2) is proved as follows.
}
}
=
O
(
t O e o O =
O
(
e O O
t
=
e
(
O O O
t
=
e =
e = e
d ) m 2 T ( ) ( X
d ) n j exp( ) Tn j exp( ) ( X
2
1
) n j exp( d ) Tn j exp( ) ( X
2
1
) n j exp( ) nT ( x
) n j exp( ) n ( x ) ( X
m
c
n
c
n
c
n
c
n
.
T
m 2
X
T
1
d ) m 2 T ( ) ( X
m
c
m
c
|
.
|
\
|
t e
=
O t O e o O =
(7.4)
7.2. Reconstruction
A discrete-time signal is converted into a continuous-time signal
by reconstruction.
This section uses the following notation:
x
c
(t) a continuous-time signal,
O
0
the central frequency of x
c
(t),
x(n) the sequence obtained by sampling x
c
(t),
T the sampling interval,
X(e) the discrete-time Fourier transform of x(n),
x'
c
(t) the continuous-time signal reconstructed from x(n),
X'
c
(O) the continuous-time Fourier transform of x'
c
(t).
7.2.1. Frequency-Domain Expression of Reconstruction
The frequency-domain expression of reconstruction is
.
otherwise , 0
T / T / ), T ( TX
) ( X
0 0
c
t + O < O s t O O
= O
' (7.5)
(7.5) is obtained from the spectral relation in sampling.
Especially, when x
c
(t) is real, O
0
=0 and (7.5) becomes
.
otherwise , 0
T / T / ), T ( TX
) ( X
c
t < O s t O
= O
'
(7.6)
7.2.2. Time-Domain Expression of Reconstruction
The time-domain expression of reconstruction is
| |
| |. ) nT t ( j exp
T / ) nT t (
T / ) nT t ( sin
) n ( x ) t ( x
n
0 c
=
O
t
t
=
'
(7.7)
Especially, when x
c
(t) is real, O
0
=0 and (7.7) becomes
| |
.
T / ) nT t (
T / ) nT t ( sin
) n ( x ) t ( x
n
c
=
t
t
=
'
(7.8)
(7.7) is proved as follows.
}
}
t + O
t O
O O O
t
=
O O O
'
t
=
'
T /
T /
c c
0
0
d ) t j exp( ) T ( X
2
T
d ) t j exp( ) ( X
2
1
) t ( x
| |
| |
| |. ) nT t ( j exp
T / ) nT t (
T / ) nT t ( sin
) n ( x
d ) nT t ( j exp ) n ( x
2
T
d ) t j exp( ) Tn j exp( ) n ( x
2
T
n
0
n
T /
T /
T /
T /
n
0
0
0
0
}
}
=
t + O
t O
t + O
t O
=
O
t
t
=
O O
t
=
O O
(
O
t
=
(7.9)
7.3. Sampling Theorem
Let x
c
(t) be a continuous-time signal with bandwidth W and x(n)
be the discrete-time signal obtained by sampling x
c
(t) with sampling
interval T. If
2t/T>W, (7.10)
x
c
(t) can be reconstructed from x(n) without distortion (figure 7.2).
Otherwise, the reconstruction may have distortion owing to aliasing
(figure 7.3).
O
X
c
(O)
X(OT)
Figure 7.2. 2t/T>W.
2t/T
O
t
x
c
(t)
nT
x(n)
T W
W
A
A/T
2t/T is referred to as the sampling frequency, and W is referred to
as the Nyquist rate.
Especially, when x
c
(t) is real, the central frequency is 0, and (7.10)
becomes
2t/T>2O
u
, (7.11)
where O
u
is the upper limit of the band.
O
X
c
(O)
X(OT)
Figure 7.3. 2t/TsW.
2t/T
O
t
x
c
(t)
nT
x(n)
T W
W
A
A/T
Example. x
c
(t) is a continuous-time signal with central frequency
O
0
and bandwidth W. x(n) is the discrete-time signal obtained by
sampling x
c
(t) with sampling interval T. x'
c
(t) is the continuous-time
signal reconstructed from x(n). When 2t/TsW, x'
c
(t) may not equal
x
c
(t). x'
c
(t) does not equal x
c
(t) either when there is a wideband noise.
However, the difference between x'
c
(t) and x
c
(t) can be decreased by
filtering x
c
(t) before it is sampled. The filter has frequency response
(7.12) .
otherwise , 0
T T
, 1
) ( H
0 0
c
t
+ O < O s
t
O
= O
Explain how it works.
