A scientific study of the problems
of digital engineering for space flight systems,
with a view to their practical solution.
Presented at the 1995 RADECS
Gary Swift
Jet Propulsion Laboratory, California Institute of Technology, Pasadena CA
Richard Katz
NASA Goddard Space Flight Center, Greenbelt MD
Abstract
Irradiations and subsequent failure analyses were performed to investigate single event dielectric rupture (SEDR) in Actel FPGAs as a function of ion LET (linear energy transfer), angle, bias, temperature, feature size, and device type. The small cross sections imply acceptably low risk for most spacecraft uses.
Table of Contents
I. Introduction
II. Actel Antifuse Structure
III. Experimental Approach
IV. Results
V. Alternate Experimental Approaches
VI. Conclusions
VII. References
List of Figures
Figure 1. A1280-family SEDR test matrix from test at Brookhaven, 8/2/94
Figure 2. Operating current during irradiation of A1280A device (s/n:350) to a fluence of 2.5·105 per cm2 of iodine (LET=60 MeV per mg/cm2 ). Other parameters: 5.5V, room temperature, 1.0 micron, 2.5 MHz, alternating pattern.
Figure 3. IDDQ technique applied to an A1280A, s/n: 004.
Figure 4. The measured dependence of SEDR susceptibility to normal-incidence LET.
Figure 5. Angle dependence of SEDR susceptibility.
Figure 6. Bias dependence of SEDR susceptibility.
List of Tables
Table 1. SEDR cross section variations for eight parameters on the Actel A1280-family.
Conclusions
Of the parameters investigated, only bias, normal LET, and angle exhibited strong enough effects to be considered significant, given the inherent statistical uncertainties. Susceptibility to SEDR increases rapidly with an ion's normally incident LET and with applied voltage. For a given ion and energy, susceptibility falls as incident angle deviates from perpendicular to the silicon surface. Other dependencies, if any, are more subtle.
Folding these dependencies into a calculation of the rate of SEDR in the 10% worst case GCR environment yields very low numbers: less than 3x10-5 per device-year for A1280- family devices and 7x10-6 for the A1020-family. Assuming operation at or below 5.5V. Note two additional assumptions: that a fielded design has no more than 5 times the average number of biased antifuses relative to the test design and that the heaviest fractions of the GCR environment are known to within a factor of three [7]. Thus, one can conclude that, although the existence of this new heavy ion induced failure mechanism is somewhat disturbing, the increment in mission risk from SEDR is small. For example, the eight month Mars Pathfinder primary mission, which incorporates seventeen flight-critical A1280As in single-string mission-critical circuits, has a 99.96% chance of not experiencing a SEDR. Even this risk is probably overstated since the calculation assumes that any SEDR is fatal. However, the residual risk could be reduced significantly if it were deemed necessary, by lowering the operating voltage only 5-10%, although this would slightly increase the susceptibility to SEUs.
Evidence of SEDR was recently obtained for the A1460A, an Act 3-family device. Thus, these results apply to all feature sizes of all families of currently available, nonradiation hardened Actel devices, consistent with only small differences in the antifuse area. Interest in using these FPGAs in radiation environments is quite high. Thus, Phillips Laboratory has contracted with Loral to produce them using their radiation-hardened process. However, SEDR susceptibility would be essentially unchanged unless special steps (such as thickening the antifuse) are taken. As a consequence of this work, Actel and Loral have adopted strategies for SEDR hardening, and the reported oxide equivalent thickness of their first product is up to about 99 Å.
Home
Last Revised January 09, 2002
Digital Engineering Institute
Web Designer: Richard Katz