Professional Documents
Culture Documents
L
Lecturer:
H
Hsien-I
i
I Lin
Li
Email: sofin@ntut.edu.tw
Office: Complex Eng. Bldg. 604-1
Phone: ext 4324
2010/11/29
BASIC IDEA
F=B
A/D
Fs2B
X (t)
Xa(t)
Sample
b/D
Xa(t)
2010/11/29
DISCRETE-TIME PROCESSING OF
CONTINUOUS-TIME SIGNALS
pestfilte
r
(1) Select the bandwidth of a signal.
(2) Decide the Fs sampling rate Speed B=3k transition 300-400Hz
Fs=8K
(3) Choose a prefilter (1) ensure that the signal to be sampled limited to the
desired frequency range
(4) A/D
A/D, D/A
3
(2) remove noises
(5) Digital signal processing
2010/11/29
X(F)
( )
D.T.F.T
1
2
x()e
Xa((F)) C.T.F.T
jn
X
( A) ( B)
F
f
Fs
1
Fs
x ( f )e
Fs
2
F
s
2
j 2 fn
F
F'
Fs
X ( F )e
j 2
F
n
Fs
df
X a ( F )e
x( f )e j 2fn df ( A)
1
2
1
dF
Fs
2
F
s
2
(k
Fs
2
F
s
2
1
) Fs
2
1
( k ) Fs
2
j 2 FnT
2 FnT
X a ( F )e
2 Fn
dF
j2
j 2
X a ( F kF s ) e
F
n
Fs
j 2
X a ( F kF s ) e
k
F
T
Fs
F
n
Fs
j2
2 fn
F
F
( F )e
( F )e
F
Fs
x a ( nT )
( 2f )
1
2
2 Ft
x ( n ) x a ( nT )
1
2
1
x(n)
dF
(B)
dF
dF
F kF s
n
Fs
j 2 nF
Fs
dF
dF
(D)
4
2010/11/29
( F kFs )
X [( f k )F ]
X (F ) Fs X a( F )
0
X (F )
x ( n)e
jn
Fs
x ( n )e
j 2
F
n
Fs
x ( n )e
Fs
2
otherwise
j 2
F
n
Fs
Since
a
(F )
Fs
2
F
s
2
xa (t )
X a ( F )e j 2Ft dF
2010/11/29
B Fs / 2
2 B Fs
B
Aliasing
-Fs
-B
Fs
6
2010/11/29
xa(t)
-B
x(n)
-Fs
Fs
Fs=1/2T
Fs2
B
X(F)
Fs=1/4T
7
2010/11/29
1
xa (t )
Fs
Fs
2
F
s
2
F
j 2 n
Fs
j 2Ft
dF
x ( n)e
e
Fs
2
F
s
2
x ( n)
j 2F ( t
n
)
Fs
X ( F ) F s X a( F )
0
Fs
2
otherwise
dF
sin(( )(t nT )
T
xa (nT )
n
( )(t nT )
T
Ideal
interpolation
formal
8
2010/11/29
g (t )
sin(
(
)t
9
)t
2010/11/29
CTFT
C.T.F.T
X a (F )
F
-B
aliasing
10
10
2010/11/29
X a (F )
X (F )
x(n) xa (nT )
D.T
11
11
2010/11/29
SAMPLING THEOREM
Nyquist rate
f = F/Fs
Fs
Fs
F
2
2
1
1
f
2
2
12
12
2010/11/29
Xa(F)
EXAMPLE
X(F)
Fs
xa (t ) cos 2F0t
1 j 2F0t
e
e j 2F0t
2
Fs
F
B s
2
2
-Fs -Fs/2-F0
Fs-F0Fs/2
Fs F0
xa (t ) cos 2 ( Fs F0 )t
Fs
F0 Fs
2
Fs
2
Fs F0
3Fs
2
Fs
F0Fs/2 Fs
Fs-F0 Fs+F0
xa (t ) cos 2 ( F0 Fs )t
F13
s
2
13
2010/11/29
EXAMPLE (CONTD)
-2Fs/3
-Fs
-Fs/2 -F0
Fs
F0
2
14
14
2010/11/29
EXAMPLE
xa (t ) e
A t
X a (F )
2A
A2 (2F ) 2
,A>0
A 1
A=1
X (F )
x ( n )e
j n
AT n
x (nT )e
jn
e jn
ATn jn
e
n 0
ATn jn
1 a2
,A0
2
1 2a cos 2 ( F / Fs ) a
15
15
2010/11/29
x a (t )
T=1/3 sec
x(n)
3Hz
-Fs -Fs/2 Fs/2 Fs
Fs=3Hz
X(F)
T=1 sec
Fs>2B
T=1/3 or 1 sec
F ),
) | F | Fs / 2
TX ((F
X a (F )
otherwise
0,
x a (t )
Fs=1Hz
16
16
2010/11/29
S/H
bits S/H
necessary
A/D
buffer
S H
11 5V
10
01 1.25
00 0V
2 bits
17
17
2010/11/29
x(n) xa (nT)
t nT
I k x k x ( n ) x kH
k 1,2,, L
x3 x2
Output of the
quantizer
IK
Quantizer
step size
Usually,
y, the quantization
q
is uniform in DSP,,
unless special applications such as speech
processing.
Full-scale range (FSR) for bipolar signals
Full-scale (FS) for unipolar signals
18
Dynamic range 8
18
2010/11/29
Q[x]
Q[x]
midrise
If zero is assigned a quantization
level. Quantizer is midtread type.
midtread
If zero is assigned to a decision
level, midrise type.
19
19
2010/11/29
QUANTIZATION ERRORS
1. eq((n)) is oniform distribuited
x(n)
Quantize
r
Xq(n)
eq
2
2
x(n)
Xq(n)
eq(n)
1
e
20
20
2010/11/29
TP
Pn
2
e
TP
x2 =E[x2]-(E[x])2)
A2
2
( A cos 0t ) dt
2
e p(e)de
power
2
e de
12
2A
2b
x
Px
SQNR 10 log10 2 20 log10
e
Pn
2
21
21
2010/11/29
DIGITAL TO ANALOG00CONVERTER
Sample
Sample-and-hold
and hold
01
Zero order
11
DSP
x a (t )
S/H
10
0
h(t)
1
h(t)
2
First order
1
T
2T
22
T 2T
22