Cyclic Codes for Error Detection

W. W. Peterson and D. T. Brown

by
Maheshwar R Geereddy

1
Definition

A code is called cyclic if [xnx0x1...xn-1] is a

codeword whenever [x0x1...xn-1xn] is also a
codeword.

2
 Notations

k = Number of binary digits in the message before

encoding
n = Number of binary digits in the encoded
message
n – k = number of check bits

n k

3
b = length of a burst of errors.
G (X) = message polynomial
P (X) = generator polynomial
R (X) = remainder on dividing X n-k G (X) by P(X).
F (X) = encoded message polynomial
E (X) = error polynomial

H (X) = Received encoded message polynomial

H (X) = F (X) + E (X)

4
 Polynomial Representation of Binary
Information

It is convenient to think of binary digits as

coefficients of a polynomial in the dummy variable X.

Polynomial is written low-order-to-high-order.

Polynomials are treated according to the laws of
ordinary algebra with an exception addition is to be
done modulo two.

5
Algebraic Description of Cyclic Codes

A cyclic code is defined in terms of a generator

polynomial P(X) of degree n-k.

If P(X) has X as a factor then every code polynomial

has X as a factor and therefore zero-order coefficient
equal to zero.
Only codes for which P(X) is not divisible by X are
considered.

6
 Encoded Message Polynomial F(X)

• Computer X n – k G (X)
• R(X) = X n – k G (X) / P(X)
• Add the remainder to the X n – k G (X)
– F(X) = X n – k G (X) + R(X)

X n – k G (X) = Q(X) P(X) + R(X)

Also F(X) = Q(X) P(X)

7
 Principles of Error Detection and Error
Correction
• An encoded message containing errors can be represented
by
H (X) = F (X) + E (X)

H (X) = Received encoded message polynomial

F (X) = encoded message polynomial
E (X) = error polynomial

8
 Principles of Error Detection and Error
Correction Contd…

• To detect error, divide the received, possible erroneous

message H(X) by P(X) and test the remainder.
• If the remainder is nonzero an error has been detected.
• If the remainder is zero, either no error or an undetectable
error has occurred

9
 DETECTION OF SINGLE ERRORS

• Theorem 1: A cyclic code generated by an polynomial

P(X) with more than one term detects all single errors.
• Proof:
– A single error in the i’th position of an encoded message
corresponds to an error polynomial X i.
– For detection of single errors, it is necessary that P(X) does not
divide X i.
– Obviously no polynomial with more than one term divides X I.

10
 DETECTION OF SINGLE ERRORS
Contd….

• Theorem 2: Every polynomial divisible by 1 + X has an

even number of terms.
– Proof:

• Also if P(X) contains a factor 1 + X any odd numbers of

errors will be detected.

11
 Double and Triple Error Detecting Codes
(Hamming Codes)
• Theorem 3: A code generated by the polynomial P(X)
detects all single and double errors if the length n of the
code is no greater than the exponent e to which P(X)
belongs.
– Detecting double errors requires that P(X) does not divisible by
X i + X j for any i, j < n

12
 Double and Triple Error Detecting Codes
Contd…
• Theorem 4: A code generated by P (X) = (1 + X) P1(X)
detects all single, double, and triple errors if the length n of
the code is no greater than the exponent e to which P1(X)
belongs.
– Single and triple errors are detected by presence of factor 1 + X as
proved in Theorem 2.
– Double errors are detected because P1(X) belong to the exponent
e >= n as proved in Theorem 3
– Q.E.D

13
 Detection of a Burst-Error

• A burst error of length b will be defined as any pattern of

errors for which the number of symbols between first and
last errors including these errors, is b.

14
 Detection of a Burst-Error Contd…

• Theorem 5: Any cyclic code generated by a polynomial of

degree n-k detects any burst –error of length n-k or less.
– Any burst polynomial can be factored as E(X) = Xi E1(X)
– E1(X) is of degree b-1
– Burst can be detected if P(X) does not divide E(X)
– Since P(X) is assumed not to have X as a factor, it could divide
E(X) only if it could divide E1(X).
– But b < = n – k

15
 Theorem 5 Contd…

• There fore P(X) is of higher degree than E1(X) which implies that P(X)
could not divide E1(X)
• Q.E.D

16
 Detection of a Burst-Error Contd…

• Theorem 6: The fraction of bursts of length b > n-k that are

undetected is
2-(n-k) if b > n – k + 1
2-(n – k – 1) if b = n – k + 1

17
Detection of Two Bursts of Errors (Abram
son and Fire Codes)
• Theorem 7: The cyclic code generated by P (X) = (1 + X) P1 (X)
detects any combination of two burst-errors of length two or less if the
length of the code, n, is no go greater than e, the exponent to which P1
(X) belongs.
– Proof
– There are four types of error patterns
• E(X) = X i + X j
• E(X) = (X i + Xi+1) + X j
• E(X) = X j (X j + Xj+1)
• E(X) = (X i + Xi+1) + (X j + Xj+1)

18
 Other Cyclic Codes

• There are several important cyclic codes which have not

been discussed in this paper.
– BCH codes (developed by Bose, Chaudhuri, and Hocquenghem)
are a very important type of cyclic codes.
– Reed-Solomon codes are a special type of BCH codes that are
commonly used in compact disc players.

19
 Implementation

• Briefly, to encode a message, G (X), n-k zeros are annexed

(I.e. multiplication of Xn-1G (X) is performed) and then Xn-
1G (X) is divided by the polynomial P (X) of degree n-k.

The remainder is then subtracted from Xn-1G (X). (It

replaces the n-k zeroes).
• This encoded message is divisible by P (X) for checking
out errors

20
 Implementation Contd…

• It can be seen that modulo 2 arithmetic has simplified the

division considerably.
• Here we do not require the quotient, so the division to find
the remainder can be described as follows.
• 1) Align the coefficient of the highest degree terms of the
divisor and dividend and subtract (same as addition)
• 2) Align the coefficient of the highest degree terms of the
divisor and difference and subtract again
3) Repeat the process until the difference has the lower degree
than the divisor
21
 Implementation Contd…

• The hardware to implement this algorithm is a shift register

and a collection of modulo two adders.
• The number of shift register positions is equal to the
degree of the divisor, P (X), and the dividend is shifted
through high order first and left to right.

22
 Implementation Contd…

• As the first one (the coefficient of the high order term of

the dividend) shifts off the end we subtract the divisor by
the following procedure:
1. In the subtraction the high-order terms of the divisor and dividend
always cancel. As the higher order term of the dividend is shifted
off the end of the register, this part of the subtraction is done
automatically.
2. Modulo two adders are placed so that when a one shifts off the end
of the register, the divisor is subtracted from the contents of the
register. The register than contains a difference that is shifted until
another comes off the end and then the process is repeated. This
continues until the entire dividend is shifted into the register.

23
24
Input 100010001101011

0 -> 10 00 1
0 -> 11 10 1
0 -> 11 01 1
1 -> 11 00 0
1 -> 11 10 0
0 -> 11 11 0
1 -> 01 11 1
0 -> 00 01 0
1 -> 00 00 1
1 -> 00 10 1

25
26
27
 Implementation Contd…

• To minimize the hardware it is desirable to use the same

register for both encoding and error detection.
• If circuit of fig : 3 is used for error detection, the remainder
on dividing X n-k H(X) by P(X) instead of remainder on
dividing H(X) by P(X)
• This makes no difference, because if H(X) is not evenly
divisible by P(X) than obviously X n-k H(X) will not be
divisible either.

28
 Implementation Contd…

• Error Correction: It is a much more difficult task than error

detection.
• It can be shown that each different correctable error pattern
must give a different remainder after division buy P(X).
• There fore error correction can be done.

29
 Conclusion

• Cyclic codes for error detection provides high efficiency

and the ease of implementation.
• It provides standardization like CRC-8 and CRC-32

30
Reference:

1) Cyclic Codes for Error Detection by W. W. Peterson

and D. T. Brown.
• 2) http://www.cs.nmsu.edu/~pfeiffer/
• 3) http://web.syr.edu/~rrosenque/ecc/cyclic.htm
• 4) http://www.microconsultants.com/tips/crc/crc.txt
• 5) Error detecting and Error Correcting codes by R. W.
Hamming
• 6) http://web.usa.navy.mil/~wdj/teach/codes_magma_ex
cercises1/node1.html

31
32