PROGRAMMABLE TECHNOLOGIES WEB SITE

A scientific study of the problems of digital engineering for space flight systems,
with a view to their practical solution.

Anatomy of an In-flight Anomaly: Investigation of Proton-Induced SEE Test Results for Stacked IBM DRAMs

Presented at the 1998 IEEE Nuclear and Space Radiation Effects Conference

Kenneth A. LaBel¹, Paul W. Marshall², Janet L. Barth¹, Richard Katz¹, Robert A.  Reed¹, Henning Leidecker¹, Hak S. Kim³, Cheryl J. Marshall⁴

¹NASA/Goddard Space Flight Center (GSFC)
²Consultant
³Jackson and Tull Chartered Engineers
⁴NRL (currently at NASA/GSFC)

Abstract

We present ground test and space flight data describing a single event anomaly that affects multiple bytes in a stacked DRAM module. A 12 Gbit solid state recorder containing 1,440 DRAM die experiences the anomalous events at a rate requiring testing of a large sample set of these modules.

Table of Contents

I. Background of Anomaly

A. Description of Solid State Recorder
B. Description of Anomaly

II. Pre-Flight Single Event Effect Test Results

III. Proton Tests of the IBM DRAM

A. Proton SEE Test Plan
B. Test Facility and Test Method
C. Details of Proton SEE Test Results

IV. Correlation of In-flight Anomaly and Ground Test Results

V. Additional Ground Test Data Examples

VI. Discussion

VII. Summary

List of Figures

Figure 1. ISC 160 Mbit DRAM Module
Figure 2. An illustration of a block SEFI mapped to the logical address space of a die (IBM Luna ES Rev. C DRAM). Darkened areas indicate the occurrence of errors. In this case, whenever the column address in this page of die memory had a value of x01H, the data was in error.

Conclusion

We presented a description of an in-flight HST SSR anomaly as well as SEE ground test data for the potentially error-causing device. A probable correlation was determined between the ground and flight data including identifying a possible target on the die to cause such an event. Further data were presented on two additional device types. These data sets illustrate the need for larger sample sizes. With only three DRAM die previously proton characterized, performing a statistically significant test to simulate 1440 die becomes important.

Home
Last Revised January 09, 2002
Digital Engineering Institute
Web Designer: Richard Katz