Channel Coding

To reduce the effect of noise

Redundancy
Error detection
Data word Codeword
0 00
1 11
Invalid code-words : 01 or 10
 Hamming Distance

Words of same length

Number of different digits
00 and 11
Distance : 2
000 and 111
Distance : 3
Hamming distance for Error
detection

t errors to be detected.
Hamming distance: t + 1
Words : 00 and 11
Hamming distance : 2
1 error can be detected.
Error correction Coding

Data word Codeword

0 000
1 111
Majority rule
1 bit error correction
 Rate reduction in coding
accumulated digits
Rate in bits per sec : 
Accumulation over time : T seconds
T bits coded by T bits
T bits sent in t1 seconds
T sent in t = ( /) t1 seconds
t/t1 = /
 Error Probability

Fraction = 2 T/ 2 T
= 2 –(-)T
Pe reduces as T is increased
Hamming Distance for error
correction

Number of errors : t
Distance between codewords : 2t + 1
0 000
1 111
Distance is 3 bits
 Hamming Bound

k : Number digits in the

dataword
n : Number of digits in the
codeword
2 m >=  ncj ; j =0 to j = t
m = n – k : Parity or check digits
 Table for n,k pairs
t n k

1 3 1
4 1
7 4
2 10 4
15 8
3 10 2
15 5
23 12
 Linear Block Codes
c = (c1,c2,c3 …….cn)
d = (d1,d2,d3 ….dk)
c1 = d1 ; c2 = d2 ; ….. ck = dk
c k+1 = h11d1 + h12d2 + …+h1kdk
c k+2 = h21d1 + h22d2 + …h2kdk
.
cn = hm1d1 + hm2d2 + …. + hmkdk
 Generator Matrix

c=dG
1 0 .. 0 h11 h21 .. hm1
0 1 .. 0 h12 h22 ..hm2
G= . .
0 0 .. 1 h1k h2k ..hmk
Ik(k by k) P( k by m)
 For a (6,3) Code the generator matrix is
1 0 0 1 0 1
G  0 1 0 0 1 1
0 0 1 1 1 0

Data word d Code word c

111 111000
110 110110
101 101011
100 100101
011 011101
010 010011
001 001110
000 000000
 The Example

1 0 0 1 0 1

c   d1 d 2 d 3 0 1 0 0 1 1 
0 0 1 1 1 0
c1  d1 ; c 2  d 2 ; c3  d 3
c 4  d 1  d 3 ; c5  d 2  d 3 ; c 6  d 1  d 2
Code word property

C = dG
= d [ Ik , P]
= [ d, dP ]
P C HT = [dP+dP]
HT= = 0
Im HT : n by m
 Error Vector
d : 100
r : received word 101101
c : Code word 100101
e : Error vector 001000
r=c+e
c=r+e
 Syndrome Vector

S=rHT
= ( c + e) H T
= eHT
1 by n n by m
S : 1 by m
Decoding table
s  eH T

1 0 1 
0 1 1 
 
 1 1 0 
H 
T

1 0 0
0 1 0 
 
0 0 1
Decoding Table Continued
e s
000000
100000
010000
001000
000100
000010
000001
100010
 Decoding Table
e s
000000 000
100000 101
010000 011
001000 110
000100 100
000010 010
000001 001
100010 111
 Decoding

(1)Calculate syndrome by
S = r HT
(2) Obtain the error vector by
using the look up table
(3) c = r + e
 Algebraic Codes

c   c1, c2,     cn 
n 1 n2
c( x)  c1 x  c2 x     cn
d  (d1 , d 2      d k )
k 1 k 2
d ( x)  d1 x  d2 x     dk
 Examples
n7 , k  4
d  (1 0 1 0 )
d ( x)  x  x
3

c  (1 0 1 0 1 0 0)
c( x)  x  x  x
6 4 2
Modulo’2 addition

x x x
6 4

x x x
5 4

x x
6 5
Generation of Cyclic Codes

(1) Find g ( x) Generator polynomial

(n  k ) factor of x  1
th n

(2) For each d calculate d ( x)

c( x)  d ( x) g ( x)
Hence get c
 An Example
n7, k 4 , nk  3
Third order factors
x  1 , x  x  1, etc.
3 3

We first try x  1
3
 Division

x  1) x  1( x  x
3 7 4

x7  x4

x 1
4

x4  x

x 1
Trial Division

x 1
7


x  x 1
3 2
 The Result

x 1
7

  x  x  x 1
4 3 2

x  x 1
3 2
 Code generation
d  (1 0 1 0)
d ( x)  x  x
3

c( x)  d ( x).g ( x)
c( x)  ( x  x)( x  x  1)
3 3 2

x x x x
6 5 4

c  (1 1 1 0 0 1 0)
 Systematic cyclic code
• The first k digits of c
are same as that of k

c( x)  x d ( x)   ( x)
nk

 ( x) is remainder of
nk
x d ( x)
g ( x)
Systematic code continued

x nk
d ( x)  ( x)
 q( x) 
g ( x) g ( x)
x nk
d ( x)  ( x)
  q( x)
g ( x) g ( x)
x nk
d ( x)   ( x)  q( x).g ( x)
 Decoding
• Every valid code
polynomial is a
multiple of g(x)
• If the received word
polynomial r(x) is not r ( x)
a multiple of g(x) an
s ( x )  Re mainder of
g ( x)
error is indicated and
remainder polynomial
is used as syndrome
 Error Polynomial

r ( x )  c ( x )  e( x )
r ( x)
s ( x)  Re m.of
g ( x)
c ( x )  e( x )
 Re m. of
g ( x)
e( x )
 Re m. of
g ( x)
 Syndrome-Error Table
• For various errors syndrome is calculated
and stored in the form of table.
• Decoding steps
• Get s(x) as remainder in dividing r(x) by
g(x)
• Get the error from the table
• Get c = r + e
 Cyclic Code Generation
nk n  k 1 nk 2
g ( x)  x  g1 x  g2 x    g n  k 1 x  1

gn-k-1
g2 g1

D D D D
a
c
s2
s1
n-k number of flip-flops d b

and modulo’2 adders Output

 An Example
(7, 4 ) Code
g ( x)  x 3  x 2  1
g ( x)  x 3  1. x 2  0.x  1
g1  1
g 2  0 ; n  k  3 Thus 3 flip  flops

D1 D2 D3
 Verification

d  (1 0 0 1)
d ( x)  x  1 3

nk
x d ( x)  x ( x  1)
3 3

 x6  x3
nk
x d ( x)
To find  ( x) we do the division
g ( x)
Remainder Calculation
x  x  1) x  x ( x  x  x  1
3 2 6 3 3 2

x x x
6 5 3

5
x
x x x
5 4 2

x x
4 2

x x x
4 3

x3  x 2  x
x3  x 2  1
x 1
Remainder polynomial

 ( x)  x  1
  011
 D1 D2 D3

Clock D1 D2 D3
0 0 0 D = 1001
1 1 0 0
2 0 1 0
3 0 0 1
4 0 0 1
5 1 0 1
6 1 1 1
7 1 1 0
 Interlaced Coding for Burst errors

x1 x2 x3 ------- x14 x15

y1 y2 y3 ------- y14 y15

z1 z2 z3 ------- z14 z15

15 , 8 code t = 2
 Convolution Coder
• N stage shift register
• V modulo 2 adders
 An Example
11010
s1 s2 s3 v1 v2
S1 S2 S3 0 0 0
Data 1 0 0 1 1
in 1 1 0 0 1
0 1 1 0 1
1 0 1 0 0
0 1 0 1 0
v1 v2 0 0 1 1 1
0 0 0 0 0
 Number of output digits
• n = (k+N-1).v
• = (5+3-1).2 = 14
 Channel coding in GSM

50 132 78

Parity bits Trailing bits

50 + 3 132 + 4
189 bits
Convolution Coder

378bits
456 bits
Code Tree 00 a
00
00 a 11 b

11 10 c

0 01 d
b
11 a
10
1
c 00 b
11
01 c
01
10 d
d
 a 00
S1 S2 S3
Data b 01
in
c 10
d 11
v1 v2 From to
10

d a a &b
01 01
10 b c &d
b c
00
11 c a & b
11
a
d c & d
 00 00 00
a
11 11 11
11
10 b
00
Trellis 10
01 c
Diagram 01
01
d
10
Received Digits : 01 00 01 00
Paths into node a 00 00 00 HD : 2 Survivor path
11 10 11 HD : 3
Paths into node b : 00 00 11 HD : 2 Survivor path
11 10 10 HD : 3
Error Probability in Un-coded
system
PEU  k . Q ( 2 )
Si
 : f 0   Bit rate
 f0
Si : Signal Power
 : PSD of noise
k : Number of bits in data word
Q( x) : Pr obability ( X  x)
For Gaussian Function
Error Probability in coded system

t 1
n  2k 
PEC    Q 
 t  1  n 
n : Number of bits in coded word
t : Number of correctable errors
An Example

n  15
k  11
  9.12
t 1
4
PEU  1.1 10
6
PEC  1.96 10