Browse Prior Art Database

An Eighteen Input, Nine Output XOR Stage for Syndrome Generation

IP.com Disclosure Number: IPCOM000122316D
Original Publication Date: 1991-Nov-01
Included in the Prior Art Database: 2005-Apr-04
Document File: 1 page(s) / 48K

Publishing Venue

IBM

Related People

Wade, WD: AUTHOR

Abstract

A single error correction, double error detection (SEC-DED) code is described which enables wiring and circuitry to be minimized. For instance, a total of 117 gates, which may be 4 or 3 input exclusive ORs (XORs), using only 445 inputs are needed to generate the syndrome for a 128 data bit, 9 check bit SEC-DED (137,128) code.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 64% of the total text.

An Eighteen Input, Nine Output XOR Stage for Syndrome Generation

      A single error correction, double error detection (SEC-DED)
code is described which enables wiring and circuitry to be minimized.
For instance, a total of 117 gates, which may be 4 or 3 input
exclusive ORs (XORs), using only 445 inputs are needed to generate
the syndrome for a 128 data bit, 9 check bit SEC-DED (137,128) code.

      The key feature of this code is the 9 x 18 parity check matrix
(PCM) XOR sums eight or four inputs to two or one outputs for each
syndrome bit generator.  The PCM is composed of two 1-weight, nine
3-weight, and seven 5-weight cyclic vectors.  These 18 cyclic vectors
are rotated eight times to generate the full code.  One of the two
1-weight vectors is used for only one rotation and then discarded.

      The modular structure of each rotation permits the eighteen
inputs to be fully XORed by only 9 gates.  The output of these gates
is then inputted to a second stage of gates.  Since the output to any
one syndrome bit generator is either one or two, two rotations
produces exactly the inputs needed for the second stage of gates.
This permits the wiring and circuitry of two rotations of the first
stage and the corresponding second stage to be simply replicated four
times (except for the second 1-weight vector).  This reduces the time
to construct the actual hardware.

      The second 1-weight vector must be added to complete the code.
The actual code can be...