Code Design for Dependable Systems

Chapter 9.3 - Single-Bit Error Correcting And Single-Block Error Locating Codes

9.3   SINGLE-BIT ERROR CORRECTING AND SINGLE-BLOCK ERROR
LOCATING (SEC-Sb/p×bEL) CODES


This section deals with the SEC-Sb/p×bEL codes, where B = p × b that correct single-bit
errors and locate single b-bit byte errors within a B-bit block.


9.3.1   Code Conditions and Bounds

Necessary and Sufficient Conditions   Let Es be the error set consisting of all
single-bit errors, and let Ei(Ej) be the error set consisting of all single-byte errors in the
i(j)-th block excluding single-bit errors. Thus Ei Ç Es = Ej Ç Es = φ for all i j, where
φ is the empty set. The following theorem describes necessary and sufficient conditions
that characterize SEC-Sb/p×bEL codes.

Theorem 9.4   A linear code, described by the parity-check matrix H, corrects all errors
in Es and locates all errors in Ei, or Ej, where i
j, if and only if

  1. E × HT ≠ 0   for all E Î {Es È Ei È Ej},
  2. E1 × HT ≠ E2 × HT   for all E1, E2 Î Es, E1 ≠ E2,
  3. E3 × HT ≠ E4 × HT   for all E3 Î Ei, and for all E4 Î Ej,
  4. E1 × HT ≠ E3 × HT   for all E1 Î Es, and for all E3 Î Ei.

Proof   It is apparent that conditions 1 and 2, and conditions 1 and 3 are necessary
conditions for correcting all single-bit errors and for indicating the location of an
erroneous block containing single-byte errors, respectively. Conditions 1 and 4 are
also necessary conditions for distinguishing single-bit errors from a single-byte error
excluding single-bit errors. So the SEC-Sb/p×bEL codes must satisfy all the conditions
1 to 4.

Conversely, if a code satisfies conditions 1 to 4, then we can distinguish single-bit errors
from single-byte errors. We can also correct all single-bit errors and locate all single-byte
errors. Therefore this code is an SEC-Sb/p×bEL code.                                         Q.E.D.

Bounds


Theorem 9.5   Linear (N, NR) SEC-Sb/p×bEL codes satisfy

  R ≥ 2b

Proof   For two bytes each belonging to different blocks, there exist 2b column vectors in
the parity-check matrix on GF(2). These vectors should be linearly independent by condition
3 of Theorem 9.4, and therefore the number of rows of the parity-check matrix
should be 2b or more.                                                                                           Q.E.D


Theorem 9.6   Linear (N, NR) SEC Sb/p×bEL codes satisfy

 

where B = p × b.

Proof   In the SEC-Sb/p×bEL codes, in general, the syndromes caused by single-bit errors
should be different from each other, and those caused by single-byte errors excluding single-
bit errors should be different from the ones caused by single-bit errors. Therefore the
errors in one block have at least B + 2bb 1 different syndromes, and hence there are
at least (B + 2bb 1) different syndromes in the received word. Hence the following
inequality is satisfied

 

From this, the inequality (9.2) can be deduced.                                                            Q.E.D

 

UNLIMITED FREE
ACCESS
TO THE WORLD'S BEST IDEAS

SUBMIT
Already a GlobalSpec user? Log in.

This is embarrasing...

An error occurred while processing the form. Please try again in a few minutes.

Customize Your GlobalSpec Experience

Category: ISO Containers
Finish!
Privacy Policy

This is embarrasing...

An error occurred while processing the form. Please try again in a few minutes.