Examples of Syndrome Decoding So c = v – e2 = (1 1 1 1 0 1) - (0 1 0 0 0 0) = (101101).

The original message was simply the information symbols in c (the

Ex 1 Let C1 be linear binary [6,3,3] code with generator matrix first three positions), or 101.
1 0 0 0 1 1
G = 0 1 0 1 0 1 Say we receive the vector v = c + e = (1 0 0 1 0 0).
0 0 1 1 1 0
We compute v HT = (111), and e is in the coset of e1+e4. But there
and parity check matrix
are three equally likely alternatives for the error vector. We have
0 1 1 1 0 0 detected errors (probably two errors), but we cannot correct them.
H = 1 0 1 0 1 0
1 1 0 0 0 1
Note we can always correct one error in a block, and occasionally
The syndromes and coset leaders are: we can detect two errors.

Syndrome Coset Leader(s) There are C(6,2) = 15 ways in which two errors can occur.
000 0 For 3 of these ways, and can detect (but not correct) the errors.
001 e6 0 = (0,0,0,0,0,0), For the other 12, we compute e and c incorrectly (although in
010 e5 e1= (1,0,0,0,0,0), 3 of them, only the parity-check positions are affected).
011 e1 ---
e6= (0,0,0,0,0,1),
100 e4
101 e2 Ex 2 Let C2 be linear binary [7,2,4] code with generator matrix
110 e3
1 0 1 1 1 0 1
111 e1+e4, e2+e5, e3+e6 G= 0 1 1 1 1 1 0

Say we receive the vector v = (1 1 1 1 0 1). and parity check matrix

1 1 1 0 0 0 0
We know v = c + e, where c is the codeword transmitted, and e is
1 1 0 1 0 0 0
the error vector.
H = 1 1 0 0 1 0 0
e HT = (v - c) HT = v HT - c HT = v HT - 0 = v HT = (101). 0 1 0 0 0 1 0
1 0 0 0 0 0 1
The table tells us that e (and v) are in the coset with leader e2.
Under nearest-neighbor decoding, we want wt(e) to be as small as
possible, so we assume e = e2. The syndromes and coset leaders are
Syndrome Coset Leader(s) Say we receive the vector v = c + e = (1 1 1 0 1 1 0).
00000 0 We compute v HT = (1 0 1 0 1), and e is in the coset of e1+e4. Since
00001 e7 e1+e4 is the unique coset leader, we assume e = e1+e4, and compute
00010 e6 c = v – e2 = (1 1 1 0 1 1 0) - (1 0 0 1 0 0 0) = (0 1 1 1 1 1 0). We then
00011 e1+e2, e6+e7 decode to the information symbols (first two positions), obtaining
00100 e5 0 1.
00101 e5+e7
00110 e5+e6
00111 e1+e2+e5, e5+e6+e7 Code C2 can
01000 e4 i) Always correct one error in a 7-bit encoded block,
01001 e4+e7
01010 e4+e6 ii) Always detect two errors in a 7-bit encoded block, and
01011 e1+e2+e4, e4+e6+e7 usually correct them. There are C(7,2) = 21 ways in which
01100 e4+e5 two errors can occur. Of these, 15 can be corrected, and
01101 e1+e3 the other 6 only detected.
01110 e2+e3 iii) Sometimes detect (but not correct) three errors in a 7-bit
01111 e1+e3+e6, e2+e3+e7 block. There are C(7,3) = 35 ways in which 3 errors can
10000 e3 occur. Of these, 12 can be detected; the remaining 23
10001 e3+e7 cause us to determine the error vector e and codeword c
10010 e3+e6 incorrectly (although in 7 of the 23, only the parity-check
10011 e1+e2+e3, e3+e6+e7 positions are affected).
10100 e3+e5
10101 e1+e4
10110 e2+e4
10111 e1+e4+e6, e2+e4+e7
11000 e3+e4
11001 e1+e5
11010 e2+e5
11011 e1+e5+e6, e2+e5+e7
11100 e1+e7, e2+e6
11101 e1
11110 e2
11111 e1+e6, e2+e7