CHAPTER 6

CHANNEL CODING
 6.1 Waveform Coding

• Waveform coding use coding pattern for better

waveform to reduce detection error
EX :
orthogonal coding for signal design
Sequence coding:use Redundant code for error detection
and error correction
A. Antipodal signal
Two signals with out of phase Fig6.2 p307
: Maximum distance between two signal
 : Best performance (minimum error probability)

Figure 6.2 Example of an antipodal signal set.

B. Orthogonal signal

ò ( )s (t )dt = [
T 0 i¹ j
condition: si t j
0 E i= j
E is signal energy
 correlation coefficiency
1 T
Z ij = ò si (t )s j (t )dt = {
0 i¹ j

E 0 1 i= j

C. M-ary signal
MPSK ; MFSK ; MPAM→M-Level
D. Correlation detection in digital sequence
NA - ND
Z ij =
NT
NA : number of digit agreements
N D : number of digit disagreements
N T : total number of digit in the sequence

1 i= j
if orthogonal Z ij = {
0 otherwise
E. Orthogonal code
a. Code word matrix of orthogonal code

Data set orthogonal code

0 é 0 0ù
H1 = ê ú
1 ë 0 1 û

Expansion

é H k -1 H k -1 ù
HK = ê
ë H k -1 H k -1 úû
 EX: k=2 two bit

data set orthogonal codeword set

0 0 é0 0 0 0ù
0 1 ê0 1 0 1ú
H2 = ê ú
1 0 ê0 0 1 1ú
ê ú
1 1 ë0 1 1 0û
k
M= 2
M: number of bit in orthogonal code word
K: number of data bit
which is orthogonal since (use Bipolar ie: binary 1→ +1
binary → -1)
 Ex: 00 00 and 01 01 →(-1, -1, -1, -1),(-1, 1, -1, 1)
note (-1)(-1) + (-1)(1) + (-1)(-1) + (-1)(1) = 1 – 1 + 1-1 = 0
é0 0 0 0 0 0 0 0ù
ê0 1 0 1 0 1 0 1ú
ê ú
ê0 0 1 1 0 0 1 1ú
ê ú
éH 2 H 2 ù ê0 1 1 0 0 1 1 0ú
H3 = ê =
ëH 2 H 2 úû ê0 0 0 0 1 1 1 1ú
ê ú
ê0 1 0 1 1 0 1 0ú
ê0 0 1 1 1 1 0 0ú
ê ú
ë0 1 1 0 1 0 0 1û

b. Symbol/Bit error probability

PE = PE (M ) = (2 k - 1)Q ( Es N 0 ) upper bounded
PB 2 k -1 2 M
1
= k = k ­ PB ¯
PE 2 - 1 M - 1
 2k M
but (1). BW↑, =
k k 2k
(2). Data rate 同，transmission rate↑ k
F. Biorthogonal codes
use antipodal waveform
M
use 2 -bit signal to replace M-bit codeword
é H k -1 ù
Bk = ê ú
H
ë k -1 û
1 i= j
M
Z ij = - 1 i¹ j i- j = antipodal
2
0 i¹ j

PB =
1
2
[
(M - 2) Q KEb N 0 + Q 2 KEb N 0 ]
1
better performance than orthogonal BW→ 2
G. Trans orthogonal code →Simplex code
detect first digit of orthogonal set, since it always
Binary 0
there are (M-1) bit codeword
1 i= j
Z ij = { -1
i¹ j
M -1

minimum power since k→k-1 ie: KEb ® (K - 1)Eb

BW↓ than orthogonal
 6.2 type of error control

A. Error detection transmission

two way →use ARQ Automatic retransmission query
B. FEC: Forward error correction error detection
and correction one way
 6.2-1 Classification of communication
terminal by connectivity

Figure 6.6 Terminal connectivity classifications. (a) Simplex. (b) Half-duplex.

(c) Full-duplex.
 6.2-2 ARQ

A. Stop-and wait ARQ Fig 6.7 a

each data need ARQ using half duplex
yes: ACK acknowledgment
no: NAK not acknowledgment (need retransmission)
→ much time
B. Pull back full duplex ARQ Fig 6.7 b
Full duplex but with fixed data length in Ex 4 bit
C. Selective-error bit full duplex ARQ Fig 6.7 c
as (b) but retransmit the error one only
Figure 6.7 Automatic repeat request (ARQ). (a) Stop-and-wait ARQ (half-duplex). (b)
Continuous ARQ with pullback (full-duplex). (c) Continuous ARQ with selective repeat
(full-duplex).
 6.3 structure sequence →channel coding

A. DMC ( Discrete Memory less channel )

a. DMC
channel noise is defined to affect each symbol
independently of all other symbols the conditional
probability of given input sequenceU = u1 , u2 ,, um ,, u N
to the corresponding output sequence Z = z1 , z2 ,, zm ,, z N
is expressed as
N
P (Z | U ) = Õ P (z m | um )
m =1
b. Gaussian channel
signal in Gaussion distribution
1 é - (z - uk )2 ù
P (z | uk ) = exp ê ú
s 2p ë 2 s 2
û
uk : Mean of uk -bit
c. BSC Binary symmetric channel
the conditional probability are same
P (1 0) = P (0 1) = p
P (0 0) = P (1 1) = 1 - p
p:error probability
1-p: correct probability
B. Code rate η
in formation Bit sec = k
h=
transmission code Bit sec = n = k + r
r: redundant bit
k k
h= = £1
n k+r

C. Signal parity check code

even bit → coded 0
odd bit → coded 1
 Ex:
000 ® 0000
110 ® 0110
100 ® 1100
111 ® 1111
• Signal-bit error can be detected but not be corrected, even
bit error can’t be detected
the undetected error probability
n
2
æ n ö 2j n is even
Pnd = å çç ÷÷ p (1 - p )
n -2 j

j =1 è 2 j ø n is odd→(n-1)/2
D. Rectangular code

Figure 6.8 Parity checks for serial and parallel code structures. (a) Serial
structure. (b) Parallel structure.
N-bit word length → N+1
M codeword → M+1
MN
code rate h =
(M + 1)(N + 1)

can correct one bit error

the block code error probability
n
ænö j
PM = å çç ÷÷ p (1 - p )
n- j

j =t +1 è j ø
in rectangular code t=1, ie j=2
error more than 2-bit → can’t correct
 6.4 Linear Block Code

A. Linear block code

V j , Vi are block code
if linear V j Å Vi is also a code word
( n,k ) block n=k + r
B. Generator Matrix
é v1 ù é v11 v12  v1n ù
êv ú êv v22  v2 n ú
G= ê 2ú
= ê 21 ú
êú ê     ú
ê ú ê ú
ë vk û ë vk 1 vk 1  vkn û
m = [m1 m2  mk ] massage code
U = [u1 u2  un ] transmission code

U = mG
= m1v1 Å m2 v2 Å  Å mk vk
= [U1 U 2  U n ] : use in modulo 2 adder
C. Systematic linear block code
[m1 m2  mk ] ¾k¾
¾®[U1 U 2  U n ]
®n

= [C1 C2  Cn - k m1 m2  mk ]

then
G = [P | I k ]
é p11 p12  p1(n - k ) ù
êp p22  p2 ( n - k ) ú
P=ê ú
21

ê     ú
ê ú
ë pk 1 pk 2  pk ( n - k ) û

P is the parity array portion

D. Dual block code
æ n ö dual æ n ö æ n ö
çç ÷÷ ¬¾
¾®çç ÷÷or çç ÷÷
èk ø èn - k ø è r ø
[
H = In -k | PT ]
é In -k ù
H =ê
T
ú
ë P û
H is parity check matrix
E. Parity check matrix
note é p11 p12  p1(n - k ) | 1 0  0ù
êp p22  p2 (n - k ) | 0 1  0ú
G=ê ú
21

ê     |    ú
ê ú
ë pk 1 pk 2  pk (n - k ) | 0 0  1û
U = [U1 U 2  U n ]
m = [m1 m2  mk ]
U = mG
= [C1 C2  Cn - k m1 m2  mk ]
C1 = m1 p11 + m2 p21 +  + mk Pk 1
C2 = m1 p12 + m2 p22 +  + mk pk 2

Cn - k = m1 p1(n - k ) + m2 p2 (n - k ) +  + mk pk (n - k )

since orthogonal é1 0  0 ù
then UH T = 0 ê0 1  0 ú
ê ú
ê     ú
ê ú
0 0  1 ú
where [
H = In -k | PT ] é I ù
H T = ê n -k ú = ê
ë P û ê p11 p12  p1(n - k ) ú
ê ú
ê p21 p22  p2 (n - k ) ú
ê     ú
ê ú
êë pk 1 pk 2  pk (n - k ) úû
F. Syndrome testing
if R is received sequence
R=U+E E: is error pattern
S = RH T
= (U + E ) H T
= UH T + EH T
= EH T
the position of error bit can be determined by the error
pattern
G. Error pattern and error correction
for ( 6 , 3 ) block code with G and H above

Figure 6.11 Example of a standard array for a (6, 3) code.

6.5 coding strength
<A> hamming weight
W(U)à no. of being 1 in u code word
ex. U = 1010110 W(U) = 4
<B> hamming distance d(U,V) , between two code words
d(U,V) = W(U Å V)
ex. U = 1011010 V = 1101011
U Å V = 0110001 W(U ÅV) = 3
<C> minimum hamming distance
Max. likelihood ratio à minimum distance of codeword.
<D> Error correction capability
d min - 1
t =[ ]
2

Fig 6.13 P344 , dmin= 5

received r1 à U …1 bit error correction
r2 à U … 2 bit
r3 à V … 1 bit
M n j
® PM = å ( ) p (1 - p ) n - j
n =t +1 j

1 n n j
® average bit error PB =
n
å
j =t +1
j (
j
) p (1 - p ) n- j
Figure 6.13
Error correction and
detection capability.
(a) Received Vector r1.
(b) Received vector r2.
(c) Received vector r3.
6.7 cyclic code
<A> definition
if U = (U 0 , U1 ,......U n -1 ) is a codeword (code vector)
then U (1) = (U n -1 , U 0 , U1 ,......U n - 2 ) is also a code vector ® cyclic code
The polynomial
U ( x) = U 0 + U1 x + U 2 x 2 + ......U n -1 x n -1
shift 1 bit U (1) ( x) = U n -1 + U 0 x + U1 x 2 + ......U n - 2 x n -1

<B> Algebraic structure of cyclic code

U ( x) = U 0 + U1 x + U 2 x 2 + ......U n -1 x n -1
xU ( x) = U 0 x + U1 x 2 + ......U n -1 x n = U 0 x + U1 x 2 + ......U n -1 ( x n + 1) + U n -1
xU ( x) U n -1 + U 0 x + U1 x 2 + ......U n - 2 x n -1 U (1) ( x)
= U n -1 + = U n -1 + n
x +1
n
x +1
n
x +1
Þ U (1) ( x) = xU ( x)modulo( x n + 1)

and x iU ( x) = p ( x)( x n + 1) + U (i ) ( x)
or U (i ) ( x) = x iU ( x)modulo( x n + 1)

EX : U = (0,1,1) U ( x) = x + x 2
shift two positions
x 2U ( x) = x 3 + x 4 = ( x 3 + 1) + x( x 3 + 1) + 1 + x
U ( 2 ) ( x) = x 2U ( x)modulo( x 3 + 1) = 1 + x
or code ® (1,1,0)
<C> generator matrix of cyclic code

(a) k-bit information polynomial

m( x) = m0 + m1 x + m2 x 2 + ......mk -1 x k -1
or m = ( m0 , m1 , m2 ,......mk -1 )
(b) Generator matrix polynomial
g ( x) = g 0 + g1 x + g 2 x 2 + ......g n - k x n - k
U ( x) = m( x) g ( x) is cyclic code degree of r = n - k
(c) Parity check matrix
xn +1
h( x ) = Þ degree of k
g ( x)
xn +1
\U ( x) = m( x) g ( x) = m( x ) or m( x) = U ( x)h( x)
h( x ) xn +1
EX 6 - 10 P364
EX : g(x) = 1 + x + x 2 + x 4 note n = 7 r = 4 k =3
x7 +1
 g(x)h(x) = x + 1 7
\ h(x) = = 1+ x + x3
g(x)
code word U(x) = m(x)g(x)

EX: 111 m(x) = 1 + x + x 2

U ( x) = m( x) g ( x) = (1 + x + x 2 )(1 + x + x 2 + x 4 )
= 1 + x + x 2 + x 4 + x + x 2 + x3 + x5 + x 2 + x3 + x 4 + x6
= 1 + 0 x + x 2 + 0 x 3 + 0 x 4 + x 5 + x 6 ® (1010011)
U ( x)h( x) = (1 + x 2 + x 5 + x 6 )(1 + x + x 3 )
= 1 + x 2 + x5 + x 6 + x + x3 + x 6 + x 7 + x3 + x5 + x8 + x9
= 1 + x + x 2 + x 7 + x 8 + x 9 = ( x 7 + 1)(1 + x + x 2 )
U ( x ) h( x ) = 1 + x + x 2 ¬ decode ® (111) = m
( x + 1)
7
(d) Dual of cyclic code
(n, k ) ¬¾¾®(n, r )
dual
Þ g ( x ) h( x ) = x n + 1
generator matrix (n, k)
g ( x) ®
parity matrix (n, r)
parity matrix of (n, k) cyclic
h( x ) ®
generator matrix of (n, r) cyclic

<D> systematic cyclic code

(a) Fundamental :
m( x) = m0 + m1 x + m2 x 2 ...... + mk -1 x k -1
U ( x) = U 0 + U1 x + U 2 x 2 + ......U n -1 x n -1
if U ( x) = c0 + c1 x + ...... + cn - k -1 x n - k -1 + m0 x n - k + m1 x n - k +1...... + mk -1 x n -1
it is systematic eg. 6.65 p359 cyclic code
(b) Generator matrix of systematic code
à form same as block code
m( x) = m0 + m1 x + m2 x 2 ...... + mk -1 x k -1
U ( x ) = c ( x ) + x n - k m( x ) º g ( x) p( x)
\ x n - k m( x ) = g ( x ) p ( x ) + c ( x )
即以x n - k m( x)除以g ( x)得商p ( x), 餘數為c( x) ® parity check matrix
• General form for generator matrix
x n - k +i m( x) = g ( x) pi ( x) + ci ( x)
ci ( x) = ci , 0 + ci ,1 x + ......ci ,n - k -1 x n - k -1 i = 0,1,.....k - 1
é c0, 0 c0,1 .... c0,n - k -1 1  0ù
0
êc c1,1 .... c1,n - k -1 0 ú
1 ú
G=ê
1, 0

ê      0ú
ê ú
êëck -1, 0 ck -1,1 .... ck -1,n - k -1 0  0 1úû
EX : for (7,3) cyclic code , if g(x) = 1 + x + x 2 + x 4 systematic code
find G and code word for the systematic code

sol : n = 7, k = 3 r = 7 -3 = 4
x i m(x) = ? i = 0,1,2....k - 1 = 2
for 100 Þ i = 0 x n -k +i m0 = x 4 = p0 ( x) g ( x) + c0 ( x)
Þ p0 ( x ) = 1 c0 ( x) = 1 + x + x 2 ® c0 = (1110)
for 010 Þ i = 1 Þ x 5 = p1 ( x) g ( x) + c1 ( x)
= x(1 + x + x 2 + x 4 ) + x + x 2 + x 3
Þ p1 ( x) = x c1 ( x) = x + x 2 + x 3 ® c1 = (0111)
for 001 Þ i = 2 Þ x 6 = p 2 ( x ) g ( x ) + c2 ( x )
= x 2 (1 + x + x 2 + x 4 ) + (1 + x + x 2 + x 4 ) + 1 + x + x 3
Þ p2 ( x ) = 1 + x 2 c2 ( x) = 1 + x + x 3 ® c2 = (1101)
 é1 1 1 0 1 0 0ù
G = êê0 1 1 1 0 1 0úú
êë1 1 0 1 0 0 1úû

m0 m1 m2
0 0 0 0 0 0 0 0 0 0
1 0 0 1 1 1 0 1 0 0
0 1 0 0 1 1 1 0 1 0
0 0 1 1 1 0 1 0 0 1
1 1 0 1 0 0 1 1 1 0
0 1 1 1 0 1 0 0 1 1
1 0 1 0 0 1 1 1 0 1
1 1 1 0 1 0 0 1 1 1
<F> error defector
U(x) = g(x)m(x) U(x)h(x) = m(x) Þ g(x)h(x) = x n + 1
xn +1
correct Û U(x)可整除by g(x)

If received Z(x) = U(x)+e(x)

no error Û Z ( x) g ( x) divisible , no remainder
Error exist : ie Z ( x) = g ( x) p ( x) + s ( x) ¬ symdrome(nonzero)
ref ex.6.10 and Fig 6.19
6.8 Well known block code
<A> Hamming code
a block code
(n, k ) = (2 m - 1,2 m - 1 - m) with d min = 3
m = n-k no. of parity check bits
ex. m=3 (7,4)
m=4 (15,11)
m=5 (31,26)
d min - 1
Correct all single bit error t= =1
2
Bit error n
1 n i 2E C
PB = å i
n i =2 i
( ) p (1 - p ) n -i
where p = Q(
N0
)
 <B> Extended Golay code

Perfect code
à parity check bit
à t =3 (24,12)

Fig 6.22
PB versus Eb/No for coherently
demodulated BPSK over a
Gaussian channel for
several block codes.
<C> BCH code (Bose-Chadhuri-Hocquenghem)

à Large block code à generalize:table 6.4 p372

g(x)à octal number
Ex : g(x) = 23 à (010,011)
Fig 6.22 (127,36) t = 15 better
(127,64) t = 10 modified

BCH
Reed-solomon code 8.1
6.6 Usefulness of standard array
code capability/error capability
<A> error detection capability
Maximum detection error bits = (dmin – 1)

<B> error detection and error correct

α : no. of correct bit
d min = a + b + 1
β : no. of detection error bit
a£b
then
error detection ¾¾¾® correct
d min - 1
if a = b = a = t ® error correct capability
2
<C> perfect code
A code with t-error correction bit has all the error pattern of t
Fig 6.11 with t = 1 is perfect but no. for t = 2 à only one
But there are (6) = 15 possible 2-bit error
2
Good code à each codeword is as far as possible with
the added redundant bit
<D> erasure correction
Erasure not know exact (0或1不確定)
d min = r + 1
If correction a d min = 2a + g + 1
decode g ® 000 first decode
g ® 111 second decode