EEE6432

Convolutional code
Prof. Jie Zhang & Dr. Jiliang Zhang
{jie.zhang, jiliang.zhang}@sheffield.ac.uk
Block code and convolutional codes

• Convolutional codes differ from block codes in that the

encoder contains memory. For convolution codes, the
encoder outputs at any given time unit depend not only on
the inputs at that time unit but also on some number of
previous inputs.
• Reed Solomon, turbo codes
• In block codes, the information bits are followed by the
parity bits. In convolution codes the information bits are
spread along the sequence. That means that the
convolution codes map information to code bits not block
wise, but sequentially convolve the sequence of
information bits according to some rule.
• Block codes have fixed length, convolution codes do not.
Convolution
Convolution is a mathematical operation on two sequences
(d and g) to produce a third sequence that expresses how
the shape of one is modified by the other.
∞
𝑥 ( 𝑛 ) =𝑑 ( 𝑛 ) ∗ 𝑔(𝑛)= ∑ 𝑔 (𝑚 ) 𝑑(𝑛− 𝑚)
𝑚=0

d9 d8 d7 d6 d5 d4 d3 d2 d1 d0 × 𝑔 ( 0)
+¿
d9 d8 d7 d6 d5 d4 d3 d2 d1 d0 0 × 𝑔 ( 1)
+¿
d9 d8 d7 d6 d5 d4 d3 d2 d1 d0 0 0 × 𝑔 ( 2)
+¿
d9 d8 d7 d6 d5 d4 d3 d2 d1 d0 0 0 0 × 𝑔 ( 3)

x12 x11 x10 x9 x8 x7 x6

¿
x5 x4 x3 x2 x1 x0
Realization of convolution by shift
register
∞
𝑥 ( 𝑛 ) =𝑑 ( 𝑛 ) ∗ 𝑔(𝑛)= ∑ 𝑔 (𝑚 ) 𝑑(𝑛− 𝑚)
𝑚=0

Example 1: Example 2: Example 3:

𝑔1 ( 𝑚) = {
0 else 2
0 {
1 𝑚=0 , 1, 2 𝑔 ( 𝑚 )= 1 𝑚=0 , 1
else
𝑔 3 (𝑚 )= {
1 𝑚=0 , 2
0 else

m0 m1 m2 m0 m1 m2 m0 m1 m2

x1 x2 x3
 Convolutional code-encoder
m0 m1 m2 m1 m2 State
Example: Assuming bits 011011 is
0 0 A
transmitted, what is the output of the
encoder? 0 1 B
Transmitted 1 0 C
x1 x2 x3 Answer: 000 111 001 011 010 001
bits 1 1 D

m0 m1 m2 Current Next x1 x2 x3
state state
Solid lines denote inputting 0
0 0 0 A A 0 0 0 Dash lines denote inputting 1
1 0 0 A C 1 1 1
0 0 1 B A 1 0 1 A B
1 0 1 B C 0 1 0 000 101

0 1 0 C B 1 1 0 111 110
011
1 1 0 C D 0 0 1 C 010 D
0 1 1 D B 0 1 1 001 100

1 1 1 D D 1 0 0
Trellis diagram m0 m1 m2

m0 m1 m2 Current Next x1 x2 x3
state state
0 0 0 A A 0 0 0 x1 x2 x3
1 0 0 A C 1 1 1
0 0 1 B A 1 0 1
1 0 1 B C 0 1 0 A 000
0 1 0 C B 1 1 0 111 101
1 1 0 C D 0 0 1
B
0 1 1 D B 0 1 1 010
1 1 1 D D 1 0 0
110
C

Solid lines denote inputting 0 011 001

Dash lines denote inputting 1 D 100
Encoding using trellis diagram
Example: Assuming bits 011011 are transmitted.

0 1 1 0 1 1
A 000 000 000 000 000 000
111 101 111 101 111 101 111 101 111 101 111 101

B
010 010 010 010 010 010
110 110 110 110 110 110
C
011 001 011 001 011 001 011 001 011 001 011 001
D 100 100 100 100 100 100

A A C D B C D
000 111 001 011 010 001
Exercise:

m0 m1 m2

x1 x2

(a) Plot the trellis diagram of the convolutional code in the figure.
(b) Assuming the shift register is initialised with 00, find the
output sequence for an input bit sequence of 110110.
 A B
Viterbi Algorithm 000 101
111 110
Definition: The Viterbi algorithm is a dynamic 011
C 010 D
programming algorithm for finding the most likely
sequence of hidden states that results in a sequence 001 100
of observed events.
0 1 1 0 1 1
A 000 000 000 000 000 000
111 101 111 101 111 101 111 101 111 101 111 101

B
010 010 010 010 010 010
110 110 110 110 110 110
C
011 001 011 001 011 001 011 001 011 001 011 001
D 100 100 100 100 100 100

A A C D B C D
000 111 001 011 010 001
After received bit sequence with errors at the receiver, as long as we can find the red
route, the information bits are obtained.
 000
111 101
Decoding using the Viterbi algorithm
010
110
Example: Assuming bits 001 111 011 011 000 001
011 001
are transmitted.
100
Input
bits 001 111 011 011 000 001
A 1 1 4 4 6 5 7 5 5 4 5 5
5 6 8 6
1 5 5 4 7
B 3 3 2 6 6
4 4 3 8
2
3
3 6 6 6
C 2 1 4 4 3 6
2 5 5 3
4
4 2 7 6
D 4 7 2 5 5 6 5 7 3

Decoded
bits 000 111 001 011 010 001
Information 0 1 1 0 1 1
bits
Exercise:

m0 m1 m2

x1 x2

Decode 01 01 11 11 01 11 using the Viterbi algorithm.

Thank you!