Browse Prior Art Database

DATA Integrity Method

IP.com Disclosure Number: IPCOM000099399D
Original Publication Date: 1990-Jan-01
Included in the Prior Art Database: 2005-Mar-14
Document File: 2 page(s) / 71K

Publishing Venue

IBM

Related People

Aichelmann Jr, FJ: AUTHOR

Abstract

A method is proposed to extend the coverage of single-error-correction/ double-error-detection (SEC/DED) codes when used with multiple outputs from the same array chip. This disclosure provides a way to avoid the miscorrection when the code misinterprets a triple-bit error and inverts a fourth bit.

This text was extracted from an ASCII text file.
This is the abbreviated version, containing approximately 52% of the total text.

DATA Integrity Method

       A method is proposed to extend the coverage of
single-error-correction/ double-error-detection (SEC/DED) codes when
used with multiple outputs from the same array chip.  This disclosure
provides a way to avoid the miscorrection when the code misinterprets
a triple-bit error and inverts a fourth bit.

      Under conventional implementations of SEC/DED codes, multiple
out- puts from the same chip are not mixed in the same ECC word.
This re- quires that multiple chips be required to satisfy the ECC
word data width with each chip contributing a single-bit of
information.  The use of multiple bits from the same chip can produce
multiple failures in the ECC word which can exceed the SEC/DED code
capabilities.

      Table I tabulates the four triple-bit failure cases that could
arise when the detection capabilities of the code are exceeded.  A
com- plement re-complement (CRC) procedure is incorporated for all
single errors in order to prevent the miscorrection for these
multiple-bit defects.  Case 1 depicts three hard faults in data from
the same ECC word. The ECC code detects and invokes CRC.  In this
instance, the results are without defects.  Case 2 includes two hard
faults with a third hard fault which matches correct data. The CRC
procedure produces a correctable word from the hidden fault with no
degradation in coverage.  Case 3 consists of two hidden faults which
produce an uncorrectable error based upon the CRC procedure.  Case 4
produces no error because all three faults are hidden.  It should be
noted that these results are produced by hard faults and do not
include soft errors.  Cases 2 and 4 do not produce any situation that
degrades data integrity.  Case 1 is the case that, without CRC, could
produce a miscorrection, while Case 3 would have been correctly
corrected by the ECC code but now indicates an uncorrectable error
due to the CRC procedure.  This situation can be rectified by simply
monitoring the...