Browse Prior Art Database

Monitor for Single-Bit Failures in Memories

IP.com Disclosure Number: IPCOM000036654D
Original Publication Date: 1989-Oct-01
Included in the Prior Art Database: 2005-Jan-29
Document File: 4 page(s) / 100K

Publishing Venue

IBM

Related People

Michels, H: AUTHOR [+4]

Abstract

The described monitor provides statistical information on the frequency of single-bit failures (SBFs) in memory. This information is used for maintenance and service tasks during design and manufacture. As the normal processing cycle must not be interfered with and the implementation cost must be kept to a minimum, such a monitor is difficult to realize.

This text was extracted from a PDF file.
At least one non-text object (such as an image or picture) has been suppressed.
This is the abbreviated version, containing approximately 48% of the total text.

Page 1 of 4

Monitor for Single-Bit Failures in Memories

The described monitor provides statistical information on the frequency of single-bit failures (SBFs) in memory. This information is used for maintenance and service tasks during design and manufacture. As the normal processing cycle must not be interfered with and the implementation cost must be kept to a minimum, such a monitor is difficult to realize.

Therefore, the described monitor is designed such that it counts only the errors detected/corrected by the SCRUB function but not SBFs occurring during normal processing. As all memory areas are system

(Image Omitted)

atically analyzed, well-defined statistical error information is obtained which is substantially independent of the respective application.

The described monitor, which is based on IBM internally used memory hardware, comprises the following elements and functions for error detection and correction. 1. ECC (error correction code) (logic means) for

on-line

1-bit error correction, 2-bit error detection;

2. a SCRUB function

preventing 2-bit soft errors;

3. a RETRY function

for 2-bit hard and soft error correction; and

4. an independent service processor (SP) for on-line

machine checks and error isolation.

The basic concept is to analyze errors detected by the SCRUB function rather than SBFs occurring during normal processing. This is possible and expedient, since the desired statistical information does not depend on whether SBFs are detected during normal processing or by SCRUBBING. It is essential that all memory areas are analyzed at regular intervals, a requirement that is readily met by the SCRUB function which provides statistical data that is more reliable and easier to obtain.

An SBF state generally indicates that there are correctable, i.e., 1-bit soft or hard, errors. Such errors are detected by the ECC logic using the SCRUB function when reading a word stored in memory. Detailed information on the respective type of error is obtained by means of an ECC evaluator logic analyzing the error, the memory address and the symptom code.

As a result, different items of error information are obtained which may be used as a basis for a more detailed analyis. It may be expedient, for example, to determine how frequently, say, word line, bit line, single-cell and/or other types of error actually occur in the field. Such data is valuable to process engineers, array designers and other experts who are interested in quality improvement. Being available on-line in the machine, this data may also be used to actively support maintenance work.

1

Page 2 of 4

Another variant of the described concept is to detect hard and soft errors which occur during normal processing rather than during SCRUBBING. For this category of errors, separate statistical information is provided.

The statistics program used for this purpose provides an initial statistical evaluation with the aim of continuously storing the result in a compact form in the machine...