Professional Documents
Culture Documents
Chapter 3-2
g ( t ) =
o Ts sampling period
o fs = 1/Ts sampling rate
g (nT ) (t nT )
s
n =
G ( f ) =
n =
n =
Chapter 3-3
Po-Ning Chen@ece.nctu
o Claim: G ( f ) = f s
Po-Ning Chen@ece.nctu
n =
G ( f mf )
s
m =
Chapter 3-4
m =
n =
cn =
1
fs
n
1
exp j 2
f , where cn =
f s
fs
fs / 2
fs / 2
2n
L( f ) exp j
fs
n
j 2
L
(
f
)
exp
f df
f /2
f s
fs / 2
s
f df
f /2
2n
= f / 2 G ( f mf s ) exp j
fs
m=
f df
Chapter 3-5
Po-Ning Chen@ece.nctu
cn =
m =
fs / 2
m =
m =
2n
G ( f mf s ) exp j
f df , s = f mf s
f s
f s / 2 + mf s
f s / 2 + mf s
fs / 2
f s / 2 + mf s
f s / 2 + mf s
2n
G ( s ) exp j
( s + mf s ) ds
fs
2n
G ( s ) exp j
s ds
f s
2n
= G ( s ) exp j
s ds
f
s
= g ( nTs )
n
L( f ) = g ( nTs ) exp j 2
f
f s
n =
m =
Po-Ning Chen@ece.nctu
Chapter 3-6
g (nT ) (t nT ) G ( f ) = f G( f mf )
s
n =
m =
Chapter 3-7
Po-Ning Chen@ece.nctu
G( f ) =
Po-Ning Chen@ece.nctu
1
G ( f ) for | f | W .
2W
Chapter 3-8
Chapter 3-9
Po-Ning Chen@ece.nctu
2W
fs
Po-Ning Chen@ece.nctu
Chapter 3-10
G(f )
G (f )
1
G (f ) for |f | W
fs
1
X
L(f ) = fs
G(f
1
X
1
fs
1
fs
g (nT )
n =
2W
fs
g (nT )
s
n =
m= 1
mfs )
m= 1
sin[2W (t nTs )]
2W (t nTs )
n =
Po-Ning Chen@ece.nctu
g (t ) =
g (nT ) sinc T
s
n =
Po-Ning Chen@ece.nctu
Chapter 3-12
g (nT ) sinc T
s
n =
t
t
g (t ) * sinc = g ( )sinc
Ts
Ts
n =
Ts
g (nT )
s
n =
t
Ts
( nTs )sinc
Chapter 3-13
Po-Ning Chen@ece.nctu
t
Reconstruction filter (interpolation filter) h(t ) = sinc
Ts
H ( f ) = Ts rect (Ts f )
g (t )
H( f )
g (t )
fs / 2
Po-Ning Chen@ece.nctu
fs
fs / 2
Chapter 3-14
Po-Ning Chen@ece.nctu
Chapter 3-15
Chapter 3-16
Chapter 3-17
Po-Ning Chen@ece.nctu
n =
0<t <T
1,
Po-Ning Chen@ece.nctu
Chapter 3-18
= f s M ( f kf s ) H ( f )
k =
= f s M ( f kf s )H ( f )
k =
Po-Ning Chen@ece.nctu
S ( f ) = f s M ( f kf s )H ( f )
k =
= f s M ( f ) H ( f ) + f s M ( f kf s ) H ( f )
|k |1
Reconstruction Filter
Po-Ning Chen@ece.nctu
Equalizer
M ( f )H ( f ) M ( f )
Chapter 3-20
10
1
1
=
exp( jfT ), | f | W
E ( f ) = H ( f ) Tsinc ( fT )
0,
otherwise
Po-Ning Chen@ece.nctu
~
E( f )
1 1
> = f s > 2W .
T Ts
0.8
, | f | W
~
E ( f ) = Tsinc ( fT )
0,
otherwise
0.6
E.g., T = 1, W = 1/8.
0.4
0.2
-1
-0.5
0.5
i (t )
~
E( f )
A lowpass filter
o1 (t ) (t + T / 2) or
exp( jfT )
o (t )
non-realizable! Why?
Because " o1 (t ) = 0 for t < 0" does not imply " o(t ) = 0 for t < 0"
Po-Ning Chen@ece.nctu
Chapter 3-22
11
i (t )
h (t )
o (t )
Chapter 3-23
Po-Ning Chen@ece.nctu
n Simplified Proof:
t
h(t ) = 0 for t < 0
o(t ) = h( )i (t )d = 0h( )i (t )d
i (t ) = 0 for t < 0
0
for
some
a
>
0
,
then
take
i
(
t
)
=
1, for t 0.
a
o( a ) = h( )d 0, which means that
If
(a )
= 0 for a > 0.
a
a
a
h
(
)
d
=
0
for
every
a
>
0
h( )d = h( a ) = 0 for a > 0.
Po-Ning Chen@ece.nctu
Chapter 3-24
12
Po-Ning Chen@ece.nctu
, | f | W
~
E ( f ) = Tsinc ( fT )
0,
otherwise
and
1 1
> = f s > 2W .
T Ts
1
, | f | 0.04
~
E ( f ) = sinc ( f )
for T = 1, Ts = 10,W = 0.04
0,
otherwise
~
E( f )
1.00264
1
0.8
0.6
0.4
0.2
-0.06
Po-Ning Chen@ece.nctu
-0.04
-0.02
0.02
0.04
0.06
Chapter 3-26
13
Chapter 3-27
Po-Ning Chen@ece.nctu
Chapter 3-28
14
Pulse trains
PDM
PPM
Po-Ning Chen@ece.nctu
Chapter 3-29
Chapter 3-30
15
1 BT ,Carson 1
Chapter 3-31
Po-Ning Chen@ece.nctu
Chapter 3-32
16
Chapter 3-33
Po-Ning Chen@ece.nctu
midtread
Po-Ning Chen@ece.nctu
midrise
Chapter 3-34
17
Uniform midtread
quantizer
Chapter 3-35
Po-Ning Chen@ece.nctu
2mmax
L
Chapter 3-36
18
q<
0,
2
q 1
Pr{Q q} = Pr ( M mod ) q = + , q <
2
2
2
1,
q
2
Or pdf f Q ( q) =
1
1 q <
2
2
Chapter 3-37
Po-Ning Chen@ece.nctu
P
2 1
/ 2 q dq
/2
P
P
3P
=
= 2 L2
2
1 2
mmax
1 2mmax
12
12 L
Chapter 3-38
19
A2m
2
and mmax = Am
) SN RO =
3(Am2 /2) 2
L
A2m
SNRO (dB)
32
31.8
64
37.8
128
43.8
256
49.8
Po-Ning Chen@ece.nctu
Chapter 3-39
Po-Ning Chen@ece.nctu
Chapter 3-40
20
v1
v2
vL-1
vL
Partitions
I1
I2
IL-1
IL
= [ A, A)
k =1
Po-Ning Chen@ece.nctu
{Ik }
{vk }
{Ik }
k =1 I k
Chapter 3-42
21
min d ( m, vk ) f M ( m)dm
{vk }
Since
v j
k =1 I k
d ( m, vk ) f M ( m)dm = d ( m, v j ) f M ( m)dm
k =1 I
v j I
d ( m, v j )
=
f M ( m)dm
v j
I
k
d (m, v j )
v j
Ij
f M (m)dm = 0.
Po-Ning Chen@ece.nctu
I1
Po-Ning Chen@ece.nctu
v2
vL-1
vL
I2
IL-1
IL
Chapter 3-44
22
mj+1
mj
@(m vj )2
fM (m)dM =
@vj
m j +1
v j ,optimal
mj
mf M (m)dm
m j +1
mj
f M (m)dm
mj+1
(m
vj )fM (m)dm = 0.
mj
= E[ M | m j M < m j +1 ]
Po-Ning Chen@ece.nctu
Chapter 3-45
Po-Ning Chen@ece.nctu
Chapter 3-46
23
Po-Ning Chen@ece.nctu
Chapter 3-47
3.7 Laws
o Quantization Laws
n A-law
o 13-bit uniformly quantized
o Conversion to 8-bit code
n -law
o 14-bit uniformly quantized
o Conversion to 8-bit code.
n These two are referred to as compression laws since
they uses 8-bit to (lossily) represent 13-(or 14-)bit
information.
Po-Ning Chen@ece.nctu
Chapter 3-48
24
m,
1 + log( A)
FA-law ( m) =
1 + log( A | m |)
sgn(m)
,
1 + log( A)
1
A
1
m 1
A
Linear mapping
Logarithmic mapping
Chapter 3-49
Po-Ning Chen@ece.nctu
FA-law ( m)
1
0.8
0.6
0.4
output
0.2
0
-0.2
-0.4
-0.6
-0.8
-1
-1
-0.8
Po-Ning Chen@ece.nctu
-0.6
-0.4
-0.2
0
input
0.2
0.4
0.6
0.8
Chapter 3-50
25
FA-law ( m)
64
48
output
32
0
-32
-48
-64
256
128
64
-64
-128
-256
-80
-96
-112
-128
-4096
-1024 -512
-2048
0 512 1024
input
2048
4096
Po-Ning Chen@ece.nctu
0
0
0
0
0
0
1
a
0
0
0
0
0
1
a
b
0
0
0
0
1
a
b
c
0
0
0
1
a
b
c
d
0
0
1
a
b
c
d
x
0
1
a
b
c
d
x
x
a
a
b
c
d
x
x
x
b
b
c
d
x
x
x
x
c
c
d
x
x
x
x
x
d
d
x
x
x
x
x
x
x
x
x
x
x
x
x
x
0
0
1
1
0
0
1
1
0
1
0
1
0
1
0
1
a
a
a
a
a
a
a
a
b
b
b
b
b
b
b
b
c
c
c
c
c
c
c
c
d
d
d
d
d
d
d
d
Po-Ning Chen@ece.nctu
Chapter 3-52
26
0
0
1
1
0
0
1
1
0
1
0
1
0
1
0
1
a
a
a
a
a
a
a
a
b
b
b
b
b
b
b
b
c
c
c
c
c
c
c
c
d
d
d
d
d
d
d
d
0
0
0
0
0
0
0
1
0
0
0
0
0
0
1
a
0
0
0
0
0
1
a
b
0
0
0
0
1
a
b
c
0
0
0
1
a
b
c
d
0
0
1
a
b
c
d
1
0
1
a
b
c
d
1
0
a
a
b
c
d
1
0
0
b
b
c
d
1
0
0
0
c
c
d
1
0
0
0
0
d
d
1
0
0
0
0
0
1
1
0
0
0
0
0
0
Po-Ning Chen@ece.nctu
log(1 + m )
for m 1.
1 + log( )
Po-Ning Chen@ece.nctu
Chapter 3-54
27
F -law (m)
0.8
0.6
0.4
0.2
0
-0.2
-0.4
-0.6
-0.8
-1
-1
-0.8
-0.6
-0.4
-0.2
0.2
0.4
0.6
Chapter 3-55
Po-Ning Chen@ece.nctu
0.8
128
112
96
80
64
F -law (m)
48
32
16
0
-16
-32
-48
479
223
95
31
-31
-95
-223
-479
-64
-80
-96
-112
-128
-8159
-4063
-2015 -991
991 2015
4063
8159
Chapter 3-56
28
Bits:12 11 10 9 8 7 6 5 4 3 2 1 0
0
0
0
0
0
0
0
1
0
0
0
0
0
0
1
a
0
0
0
0
0
1
a
b
0
0
0
0
1
a
b
c
0
0
0
1
a
b
c
d
0
0
1
a
b
c
d
x
0
1
a
b
c
d
x
x
1
a
b
c
d
x
x
x
a
b
c
d
x
x
x
x
b
c
d
x
x
x
x
x
c
d
x
x
x
x
x
x
d
x
x
x
x
x
x
x
x
x
x
x
x
x
x
x
0
0
0
0
1
1
1
1
0
0
1
1
0
0
1
1
0
1
0
1
0
1
0
1
a
a
a
a
a
a
a
a
b
b
b
b
b
b
b
b
c
c
c
c
c
c
c
c
d
d
d
d
d
d
d
d
Po-Ning Chen@ece.nctu
0
0
1
1
0
0
1
1
0
1
0
1
0
1
0
1
a
a
a
a
a
a
a
a
b
b
b
b
b
b
b
b
c
c
c
c
c
c
c
c
d
d
d
d
d
d
d
d
0
0
0
0
0
0
1
a
0
0
0
0
0
1
a
b
0
0
0
0
1
a
b
c
0
0
0
1
a
b
c
d
0
0
1
a
b
c
d
1
0
1
a
b
c
d
1
0
1
a
b
c
d
1
0
0
a
b
c
d
1
0
0
0
b
c
d
1
0
0
0
0
c
d
1
0
0
0
0
0
d
1
0
0
0
0
0
0
1
0
0
0
0
0
0
0
Chapter 3-58
29
A-law
mu-law
-0.6
-0.8
-1
-0.8
Po-Ning Chen@ece.nctu
-0.6
-0.4
-0.2
0.2
0.4
0.6
0.8
Chapter 3-59
3.7 Coding
o After the quantizer provides a symbol representing one of
256 possible levels (8 bits of information) at each sampled
time, the encoder will transform the symbol (or several
symbols) into a code character (or code word) that is
suitable for transmission over a noisy channel.
o Example. Binary code.
0 = change
11100100
1 1 1 0 0 1 0 0 1 = unchange
Po-Ning Chen@ece.nctu
Chapter 3-60
30
3.7 Coding
o Example. Ternary code (Pseudo-binary code).
00011011
0 0
0 1 1 0 1 1
A
B
C
00011011ACABBCBB
Through the help of coding, the receiver may be able to
detect (or even correct) the transmission errors due to noise.
For example, it is impossible to receive ABABBABB, since
this is not a legitimate code word (character).
Po-Ning Chen@ece.nctu
Chapter 3-61
3.7 Coding
o Example of error correcting code Three-times repetition
code (to protect Bluetooth packet header).
00011011 000,000,000,111,111,000,111,111
Then majority law can be applied at the receiver to
correct one-bit error.
oChannel (error correcting) codes are designed to
compensate the channel noise, while line codes are simply
used as the electrical representation of a binary data stream
over the electrical line.
Po-Ning Chen@ece.nctu
Chapter 3-62
31
Chapter 3-63
Po-Ning Chen@ece.nctu
PSD = lim
T
1
E[ S ( f ) S2*T ( f )], where s2T (t ) = s(t ) 1{| t | T }.
2T
1
X
1
X
an e
j2f nTb
n= 1
) PSD =
Po-Ning Chen@ece.nctu
n= 1
N
X1
an e
n= N
1
X
N
X1
n= 1 m= N
j2f nTb
Chapter 3-64
32
1
N 1
| G ( f ) |2 E[an am* ]e j 2f ( n m )T
N 2 NT
n = m = N
b
1 N 1
j 2f ( n m ) T
=| G ( f ) |2 lim
a ( n m)e
N 2 NT
b m = N n =
PSD = lim
1 N 1
j 2fkT
a ( k )e
N 2 NT
b m = N k =
1
=| G ( f ) |2 a (k )e j 2fkT
Tb k =
=| G ( f ) |2 lim
1
For i.i.d. {an },
Tb
1
X
k= 1
a (k)e
j2f kTb
1
2a X
+
e
Tb
Tb
2
a
j2f kTb
k= 1
1
2 X
+ a2
Tb
Tb
2
a
(f
k/Tb )
k= 1
Po-Ning Chen@ece.nctu
Chapter 3-65
Po-Ning Chen@ece.nctu
Chapter 3-66
33
1
2 X
+ a2
Tb
Tb
2
a
|G(f )|2
A
(f
k/Tb )
k= 1
Tb2 sinc2 (f Tb )
1
2 X
+ a2
Tb
Tb
2
a
(f
k/Tb )
k= 1
1
A2 T b
1 X
sinc2 (f Tb ) 1 +
(f
4
Tb
k= 1
A2 T b
1
sinc2 (f Tb ) 1 +
(f )
4
Tb
=
=
k/Tb )
Chapter 3-67
Po-Ning Chen@ece.nctu
{an }n = is 1 i.i.d.,
PSDP-NRZ
=
=
Po-Ning Chen@ece.nctu
|G(f )|
1
2 X
+ a2
Tb
Tb
2
a
A2 Tb sinc2 (f Tb )
k= 1
(f
k/Tb )
Chapter 3-68
34
Chapter 3-69
Po-Ning Chen@ece.nctu
!
1
2 X
|G(f )|2
+ a2
(f k/Tb )
Tb
Tb
k= 1
!
1
2
A2 Tb2
f Tb
2a X
a
2
sinc
+ 2
(f k/Tb )
4
2
Tb
Tb
k= 1
!
1
A2 T b
f Tb
1 X
2
sinc
1+
(f k/Tb )
16
2
Tb
k= 1
!
1
A2 T b
f Tb
1 X
2
sinc
1+
(f k/Tb )
16
2
Tb
2
a
k=1
Po-Ning Chen@ece.nctu
Chapter 3-70
35
s(t ) =
A, 0 t < Tb / 2
otherwise
a g (t nT ), where g (t ) = 0,
n
n =
Po-Ning Chen@ece.nctu
Chapter 3-71
1
1
1 1
E[an2 ] = (0) + ( 1) 2 + ( +1) 2 =
2
4
4 2
1
1
E[an an +1 ] = ( 1) =
4
4
1
1
1
1
E[an an + 2 ] = (1)(1) + (1)(1) + ( 1)(1) + ( 1)(1) = 0
16
16
16
16
!
E[an an + m ] = 0 for m > 1.
Po-Ning Chen@ece.nctu
Chapter 3-72
36
=
=
=
=
1
|G(f )|
Tb
2
1
X
k= 1
a (k)e
j2f kTb
f Tb
1
1 j2f Tb 1 1
A2 Tb2
sinc2
e
e
+
4
2
Tb
4
2 4
A2 Tb2
f Tb
1 1 1
2
sinc
cos(2f Tb )
4
2
Tb 2 2
f Tb
A2 T b
sinc2
sin2 (f Tb )
4
2
Po-Ning Chen@ece.nctu
j2f Tb
Chapter 3-73
A, 0 t < Tb / 2
n =
g (t ) = A, Tb / 2 t < Tb
0,
otherwise
Po-Ning Chen@ece.nctu
Chapter 3-74
37
=
=
!
1
2a X
|G(f )|
+ 2
(f k/Tb )
Tb
Tb
k= 1
1
2
f Tb
f Tb
2 X
2
2
2 2
A Tb sinc
sin
+ a2
(f
2
2
Tb
Tb
k= 1
f Tb
f Tb
2
2
2
A Tb sinc
sin
2
2
2
2
a
k/Tb )
Chapter 3-75
Po-Ning Chen@ece.nctu
Let Tb=1, and adjust A such that the total power of each line
code is 1. This gives a fair comparison among line codes.
power=
1
2
+ 12
PSDU-NRZ =
1
1
sinc2 (f ) + (f )
2
2
A=
A=1
PSDP-NRZ = sinc2 (f )
1
1
f
1 X
k
2
2
1
power= 2 PSDU-RZ = sinc
+
sinc
(f k)
4
2
4
2
+1
power= 1
k= 1
f
power= 1 PSDBRZ = sinc
sin2 (f )
2
f
f
2
2
power= 1 PSDManchester = sinc
sin
2
2
2
Po-Ning Chen@ece.nctu
A=2
A=2
A=1
Chapter 3-76
38
U-NRZ
P-NRZ
U-RZ
BRZ
Manchester
0.8
0.6
0.4
0.2
1/ 2
0.5
1.5
2
Chapter 3-77
Po-Ning Chen@ece.nctu
on
dn
dn = dn
Po-Ning Chen@ece.nctu
on = dn
on
Chapter 3-78
39
3.7 Regeneration
o Regenerative repeater for PCM system
n It can completely remove the distortion if the decision
making device makes the right decision (on 1 or 0).
Po-Ning Chen@ece.nctu
Chapter 3-79
Po-Ning Chen@ece.nctu
Chapter 3-80
40
Po-Ning Chen@ece.nctu
Chapter 3-81
Chapter 3-82
41
Chapter 3-83
Po-Ning Chen@ece.nctu
BER
4.3
10-2
10-3 second
8.4
10-4
10-1 second
10.6
10-6
10 seconds
12.0
10-8
20 minutes
13.0
10-10
1 day
14.0
10-12
3 months
Chapter 3-84
42
Po-Ning Chen@ece.nctu
Chapter 3-85
Chapter 3-86
43
Chapter 3-87
Po-Ning Chen@ece.nctu
Chapter 3-88
44
Po-Ning Chen@ece.nctu
Chapter 3-89
Chapter 3-90
45
193 bit/frame
1
1 sample (from each of 24 voice channels)/frame
Po-Ning Chen@ece.nctu
Chapter 3-91
Chapter 3-92
46
Po-Ning Chen@ece.nctu
Chapter 3-93
Chapter 3-94
47
Chapter 3-95
Po-Ning Chen@ece.nctu
Chapter 3-96
48
Chapter 3-97
Po-Ning Chen@ece.nctu
Chapter 3-98
49
Po-Ning Chen@ece.nctu
Chapter 3-99
Chapter 3-100
50
Po-Ning Chen@ece.nctu
L 1
M
L
= 185.88082902 s
= 186.31178707 s
= 186.52849741 s
fin
fout
fin
M
4(L 1)
=S
+ (1
fout
4fin
S)
4L
4fin
M
L 1
L
= S
+ (1 S )
f out
f in
f in
#"!
One bit is replaced
by a stuffed bit.
S = L
f in
1.544
M = 288
1176 = 0.334601
f out
6.312
Po-Ning Chen@ece.nctu
Chapter 3-102
51
L 1
L 1
L M
L
maxS
+ (1 S )
min S
+ (1 S )
S
[
0
,
1
]
S[ 0,1]
f in f out
f in
f in
f in
L
M
L 1
L
L 1
f out f in
f out
f in
f out
f in
M
M
1.5458 =
288
287
6.312 f in
6.312 = 1.54043
1176
1176
Chapter 3-103
Po-Ning Chen@ece.nctu
106 + a ppm
106
=
1.54043 1.544
1.5458
a ppm = 1164.8 and bppm = 2312.18
o This tolerance is already much larger than the
expected change in the bit rate of the incoming DS1
bit stream.
Po-Ning Chen@ece.nctu
Chapter 3-104
52
Chapter 3-105
Po-Ning Chen@ece.nctu
Chapter 3-106
53
Po-Ning Chen@ece.nctu
Chapter 3-107
Po-Ning Chen@ece.nctu
Chapter 3-108
54
mq [n ] = mq [n 1] + eq [n ] =
e [n],
q
j =
Chapter 3-109
Po-Ning Chen@ece.nctu
mq [n ] = mq [n 1] + eq [n ] =
e [n],
q
j =
3.12 Delta
Modulation
m[n] = mq [n 1] + e[n]
mq [n] = mq [n 1] + eq [n]
) mq [n]
m[n] = eq [n]
Chapter 3-110
55
Po-Ning Chen@ece.nctu
Chapter 3-111
dm(t )
max
Ts
dt
Chapter 3-112
56
Po-Ning Chen@ece.nctu
Chapter 3-113
Po-Ning Chen@ece.nctu
Chapter 3-114
57
Po-Ning Chen@ece.nctu
Chapter 3-115
A straightforward
structure
Since integration is
a linear operation,
the two integrators
before comparator
can be combined
into one after
comparator.
Po-Ning Chen@ece.nctu
Chapter 3-116
58
nTs
Chapter 3-117
Chapter 3-118
59
o Consider a finite-duration impulse response (FIR) discretetime filter, where p is the prediction order, with linear
prediction
p
x[n ] = wk x[n k ]
k =1
Po-Ning Chen@ece.nctu
Chapter 3-119
Po-Ning Chen@ece.nctu
Chapter 3-120
60
J = E x[n ] wk x[n k ]
k =1
k =1 j =1
p
p
p p
p
i 1
J = 2 RX [i ] + 2 w j RX [i j ] + 2 wk RX [k i ] + 2 wi RX [0]
wi
k =1
j =i +1
= 2 RX [i ] + 2 w j RX [i j ] = 0
j =1
Chapter 3-121
Po-Ning Chen@ece.nctu
w R
j
[i j ] = RX [i ] for 1 i p.
j =1
RX [1]
RX [0]
R [1]
RX [0]
X
!
!
RX [ p 1] RX [ p 2]
RX [ p 1] w1 RX [1]
RX [ p 2] w2 RX [2]
=
! !
#
!
"
RX [0] w p RX [ p ]
Chapter 3-122
61
a1
a0
a
a0
1
"
"
a p 1 a p 2
a p 1
a p 2
# "
! a0 p p
is said to be Toeplitz.
o A Toeplitz matrix can be uniquely determined by p
elements, [a0, a1, , ap-1].
Po-Ning Chen@ece.nctu
Chapter 3-123
Po-Ning Chen@ece.nctu
Chapter 3-124
62
gi
J
wi
1
w i [n + 1] = w i [n ] gi [n ]
2
where is a chosen constant step size, and is
Chapter 3-125
Po-Ning Chen@ece.nctu
gi [n ] J / wi = 2 RX (i ) + 2 w j RX (i j )
j =1
)w
i [n + 1]
=
Po-Ning Chen@ece.nctu
w
i [n] + x[n
w
i [n] + x[n
i] @x[n]
i]e[n]
p
X
w
j [n]x[n
j=1
j]A
Chapter 3-126
63
Chapter 3-127
Po-Ning Chen@ece.nctu
Po-Ning Chen@ece.nctu
Chapter 3-128
64
e[n ] = m[n ] m [n ],
where m[n ] is the linear prediction of m[n ].
Chapter 3-129
Po-Ning Chen@ece.nctu
3.14 DPCM
2
1 2mmax
mmax
=
12 L
3L2
Po-Ning Chen@ece.nctu
Chapter 3-130
65
3.14 DPCM
o Derive:
= m [n ] + e[n ] + q[n ]
= m[n ] + q[n ]
So we have the same
relation between mq[n] and
m[n] (as in Slide 3-110) but
with smaller q[n].
Po-Ning Chen@ece.nctu
eq [n ]
mq [n]
m [n ]
Chapter 3-131
3.14 DPCM
o Notes
n DM system can be treated as a special case of DPCM.
Po-Ning Chen@ece.nctu
Chapter 3-132
66
3.14 DPCM
oDistortions due to DPCM
nSlope overload distortion
oThe input signal changes too rapidly for the prediction
filter to track it.
nGranular noise
Po-Ning Chen@ece.nctu
Chapter 3-133
E [m 2 [n ]]
E [ q 2 [n ]]
o We can re-write SNRO as:
SNRO =
E [m 2 [n ]] E [e 2 [n ]]
= G p SNRQ
E [e 2 [n ]] E [ q 2 [n ]]
where e[n] = m[n] m [n] is the prediction error.
SNRO =
Po-Ning Chen@ece.nctu
Chapter 3-134
67
E[m 2 [n ]]
G
=
p E[e 2 [n ]] processing gain
2
SNR = E[e [n ]] signal to quantization noise ratio
Q
E[q 2 [n ]]
Po-Ning Chen@ece.nctu
Chapter 3-135
Po-Ning Chen@ece.nctu
Chapter 3-136
68
3.14 DPCM
o Final notes on DPCM
n Comparing DPCM with PCM in the case of voice
signals, the improvement is around 4-11 dB, depending
on the prediction order.
n The greatest improvement occurs in going from no
prediction to first-order prediction, with some additional
gain resulting from increasing the prediction order up to
4 or 5, after which little additional gain is obtained.
n For the same sampling rate (8KHz) and signal quality,
DPCM may provide a saving of about 8~16 Kbps
compared to standard PCM (64 Kpbs).
Chapter 3-137
Po-Ning Chen@ece.nctu
3.14 DPCM
Source: IEEE Communications Magazine, September 1997.
Excellent
ADPCM
G.723.1 G.729
Good
G.723.1
Speech Quality
IS-641
US-1
MELP 2.4
FS-1016
FS-1015
Poor
Unacceptable
Po-Ning Chen@ece.nctu
G.726
PCM
G.711
G.727
GSM
JDC2
Fair
G.728
8
16
Bit rate (kb/s)
IS = Interim Standard
FS = Federal Standard
IS96
IS54
JDC
GSM/2
32
64
Chapter 3-138
69
Po-Ning Chen@ece.nctu
Po-Ning Chen@ece.nctu
Chapter 3-140
70
Po-Ning Chen@ece.nctu
Chapter 3-141
3.15 AQF
o AQF is in principle a more accurate estimator. However it
requires
n an additional buffer to store unquantized samples for the
learning period.
n explicit transmission of level information to the receiver
(the receiver, even without noise, only has the quantized
samples).
n a processing delay (around 16 ms for speech) due to
buffering and other operations for AQF.
o The above requirements can be relaxed by using AQB.
Po-Ning Chen@ece.nctu
Chapter 3-142
71
3.15 AQB
Chapter 3-143
Chapter 3-144
72
3.15 ADPCM
Adaptive prediction
with backward
estimation (APB).
Chapter 3-145
Po-Ning Chen@ece.nctu
eq [n ]
eq [n 1]
Chapter 3-146
73
e[n]
eq [n ]
eq [n 1]
Po-Ning Chen@ece.nctu
Chapter 3-147
1 e [n 1]
, if [n 1] min
[n 1] 1 + q
2 e [n]
[n] =
q
,
if [n 1] < min
min
Chapter 3-148
74
ADM
Chapter 3-149
Chapter 3-150
75
Po-Ning Chen@ece.nctu
Chapter 3-151
Po-Ning Chen@ece.nctu
Chapter 3-152
76
Chapter 3-153
Po-Ning Chen@ece.nctu
Chapter 3-154
77
Chapter 3-155
Chapter 3-156
78