IEC PAS 62396-2:2007

IEC/PAS 62396-2
Edition 1.0 2007-09
PUBLICLY AVAILABLE
SPECIFICATION
PRE-STANDARD
Process management for avionics – Atmospheric radiation effects –
Part 2: Guidelines for single event effects testing for avionics systems

IEC/PAS 62396-2:2007(E)
All rights reserved. Unless otherwise specified, no part of this publication may be reproduced or utilized in any form

or by any means, electronic or mechanical, including photocopying and microfilm, without permission in writing from
either IEC or IEC's member National Committee in the country of the requester.
If you have any questions about IEC copyright or have an enquiry about obtaining additional rights to this publication,

please contact the address below or your local IEC member National Committee for further information.

IEC Central Office
3, rue de Varembé
CH-1211 Geneva 20
Switzerland
Email: inmail@iec.ch
Web: www.iec.ch
About the IEC
The International Electrotechnical Commission (IEC) is the leading global organization that prepares and publishes
International Standards for all electrical, electronic and related technologies.

About IEC publications
The technical content of IEC publications is kept under constant review by the IEC. Please make sure that you have the
latest edition, a corrigenda or an amendment might have been published.
ƒ Catalogue of IEC publications: www.iec.ch/searchpub
The IEC on-line Catalogue enables you to search by a variety of criteria (reference number, text, technical committee,…).
It also gives information on projects, withdrawn and replaced publications.
ƒ IEC Just Published: www.iec.ch/online_news/justpub
Stay up to date on all new IEC publications. Just Published details twice a month all new publications released. Available
on-line and also by email.
ƒ Electropedia: www.electropedia.org
The world's leading online dictionary of electronic and electrical terms containing more than 20 000 terms and definitions
in English and French, with equivalent terms in additional languages. Also known as the International Electrotechnical
Vocabulary online.
ƒ Customer Service Centre: www.iec.ch/webstore/custserv
If you wish to give us your feedback on this publication or need further assistance, please visit the Customer Service
Centre FAQ or contact us:
Email: csc@iec.ch
Tel.: +41 22 919 02 11
Fax: +41 22 919 03 00
IEC/PAS 62396-2
Edition 1.0 2007-09
PUBLICLY AVAILABLE
SPECIFICATION
PRE-STANDARD
Process management for avionics – Atmospheric radiation effects –
Part 2: Guidelines for single event effects testing for avionics systems

INTERNATIONAL
ELECTROTECHNICAL
COMMISSION
PRICE CODE
T
ICS 03.100.50; 31.020; 49.060 ISBN 2-8318-9202-3

– 2 – PAS 62396-2 © IEC:2007(E)

CONTENTS
FOREWORD.3

1 General .5

1.1 Use of existing SEE data.5

1.2 Deciding to perform dedicated SEE tests.6

2 Availability of existing SEE data for avionics applications .6
2.1 Types of existing SEE data that may be used.6
2.1.1 Sources of data, proprietary versus published data .7
2.1.2 Data based on the use of different sources.8
2.1.3 Ground level versus avionics applications .11
2.2 Sources of existing data .12
3 Considerations for SEE Testing .13
3.1 Selection of hardware to be tested .14
3.2 Selection of test method.14
3.3 Selection of facility providing energetic particles .15
3.3.1 Spallation neutron source .15
3.3.2 Monoenergetic and quasi-monoenergetic beam sources.16
3.3.3 Thermal neutron sources .17
4 Converting test results to avionics SEE rates .17
4.1 Use of spallation neutron source .17
4.2 Use of SEU cross section curve over energy .18

Bibliography.21

Figure 1 – Comparison of Los Alamos and TRIUMF neutron spectra with terrestrial
neutron spectrum.9
Figure 2 – Variation of high energy neutron SEU cross section per bit as a function of
device feature size.10
Figure 3 — Comparison of mono-energetic SEU cross sections with Weibull and
Piece-Wise Linear Fits .20

Table 1 – Sources of existing data .13

PAS 62396-2 © IEC:2007(E) – 3 –

INTERNATIONAL ELECTROTECHNICAL COMMISSION

____________
PROCESS MANAGEMENT FOR AVIONICS –

ATMOSPHERIC RADIATION EFFECTS –

Part 2: Guidelines for single event effects testing

for avionics systems
FOREWORD
1) The International Electrotechnical Commission (IEC) is a worldwide organization for standardization comprising
all national electrotechnical committees (IEC National Committees). The object of IEC is to promote
international co-operation on all questions concerning standardization in the electrical and electronic fields. To
this end and in addition to other activities, IEC publishes International Standards, Technical Specifications,
Technical Reports, Publicly Available Specifications (PAS) and Guides (hereafter referred to as “IEC
Publication(s)”). Their preparation is entrusted to technical committees; any IEC National Committee interested
in the subject dealt with may participate in this preparatory work. International, governmental and non-
governmental organizations liaising with the IEC also participate in this preparation. IEC collaborates closely
with the International Organization for Standardization (ISO) in accordance with conditions determined by
agreement between the two organizations.
2) The formal decisions or agreements of IEC on technical matters express, as nearly as possible, an international
consensus of opinion on the relevant subjects since each technical committee has representation from all
interested IEC National Committees.
3) IEC Publications have the form of recommendations for international use and are accepted by IEC National
Committees in that sense. While all reasonable efforts are made to ensure that the technical content of IEC
Publications is accurate, IEC cannot be held responsible for the way in which they are used or for any
misinterpretation by any end user.
4) In order to promote international uniformity, IEC National Committees undertake to apply IEC Publications
transparently to the maximum extent possible in their national and regional publications. Any divergence
between any IEC Publication and the corresponding national or regional publication shall be clearly indicated in
the latter.
5) IEC provides no marking procedure to indicate its approval and cannot be rendered responsible for any
equipment declared to be in conformity with an IEC Publication.
6) All users should ensure that they have the latest edition of this publication.
7) No liability shall attach to IEC or its directors, employees, servants or agents including individual experts and
members of its technical committees and IEC National Committees for any personal injury, property damage or
other damage of any nature whatsoever, whether direct or indirect, or for costs (including legal fees) and
expenses arising out of the publication, use of, or reliance upon, this IEC Publication or any other IEC
Publications.
8) Attention is drawn to the Normative references cited in this publication. Use of the referenced publications is
indispensable for the correct application of this publication.
9) Attention is drawn to the possibility that some of the elements of this IEC Publication may be the subject of
patent rights. IEC shall not be held responsible for identifying any or all such patent rights.
A PAS is a technical specification not fulfilling the requirements for a standard but made

available to the public.
IEC-PAS 62396-2 has been processed by IEC technical committee 107: Process management
for avionics.
The text of this PAS is based on the This PAS was approved for publication
following document: by the P-members of the committee
concerned as indicated in the following
document:
Draft PAS Report on voting
107/57/NP 107/69/RVN
Following publication of this PAS, which is a pre-standard publication, the technical committee
or subcommittee concerned will transform it into an International Standard.

– 4 – PAS 62396-2 © IEC:2007(E)

This PAS shall remain valid for an initial maximum period of three years starting from 2007-09.

The validity may be extended for a single three-year period, following which it shall be revised

to become another type of normative document or shall be withdrawn.

IEC/PAS 62396 consists of the following parts, under the general title Process management

for avionics – Atmospheric radiation effects:

• Part 2: Guidelines for single event effects testing for avionics systems

• Part 3: Optimising system design to accommodate the Single Event Effects (SEE) of

atmospheric radiation
• Part 4: Guidelines for designing with high voltage aircraft electronics and potential
single event effects
• Part 5: Guidelines for assessing thermal neutron fluxes and effects in avionics
systems
PAS 62396-2 © IEC:2007(E) – 5 –

PROCESS MANAGEMENT FOR AVIONICS –

ATMOSPHERIC RADIATION EFFECTS –

Part 2: Guidelines for single event effects testing

for avionics systems
1 General
The purpose of this PAS is to provide guidance related to the testing of microelectronic

devices for purposes of measuring their susceptibility to single event effects (SEE) induced by
the atmospheric neutrons. Since the testing can be performed in a number of different ways,
using different kinds of radiation sources, it also shows how the test data can be used to
estimate the SEE rate of devices and boards due to the atmospheric neutrons in the
atmosphere at aircraft altitudes.
The type of SEE data available can be viewed from many different perspectives. As indicated,
the SEE testing can be performed using a variety of radiation sources, all of which can induce
single event effects in ICs. In addition, many tests are performed on individual devices, but
some tests expose an entire single board computer to radiation fields that can induce SEE
effects. However, a key discriminator is deciding on whether existing SEE data is available
that may be used, or whether there really is no existing data and therefore, a SEE test on the
device or board of interest has to be carried out.
1.1 Use of existing SEE data
The simplest solution is to find previous SEE data on a specific IC device. This is not nearly
as simple as it appears. First, the largest interest lies in SEE data that is directly usable for
purposes of estimating the SEE rate in avionics. Thus, SEE tests that have been carried out
on devices using heavy ions, data which is directly applicable for space missions, is data that
is not directly applicable for avionics purposes. This heavy ion SEE data can be used to
calculate SEE data from high energy neutrons and protons by utilizing a number of different
calculation methods, but this requires the active involvement of a radiation effects expert in
the process. Therefore, heavy ion SEE data should not be used for application to the
atmospheric neutron environment, except by scientists and engineers who have extensive
experience in using this kind of data. For that reason, unless otherwise stated explicitly, when
SEE data is discussed in the remainder of this PAS, it refers only to single event testing using
a neutron or proton source, not to the results from testing with heavy ions.
If SEE data on a device of interest is found from SEE tests using high energy neutrons or
protons, it will still require expertise regarding how the data is to be utilized in order to
calculate a SEE rate at aircraft altitudes. Data obtained by IC vendors for their standard
application to ground level systems are often expressed in totally different units, FIT units,
where one FIT is one error in 10 device hours, which is taken to apply at ground level.
IC devices are constantly changing. In some cases, devices which had been tested, become
obsolete and are replaced by new devices which have not been tested. The fact that a device
is made by the same IC vendor and is of the same type as the one it replaced does not mean
that the SEE data measured in the first device applies directly to the newer device. In some
cases, small changes in the IC design or manufacturing process can have a large effect in
altering the SEE response, but in other cases, the effect on the SEE response may be
minimal.
A continuing problem with the existing SEE data is that there is no single data base
containing all of the neutron or proton SEE data. Instead, portions of this kind of SEE data
can be found published in many diverse sources. The SEE data in the larger data bases are
mainly on much older devices, dating from the 1990s and even 1980s, and it is primarily from

– 6 – PAS 62396-2 © IEC:2007(E)

heavy ion tests that were performed for space applications and not from testing with protons

and neutrons.
1.2 Deciding to perform dedicated SEE tests

If existing SEE data is not available, for any one of the many reasons discussed above and

which will be further expanded upon below, then there is no real alternative but to carry out

your own SEE testing. The advantage of such a test is that it pertains to the specific device or

board that is of interest, but the disadvantage is that it entails making a number of important
decisions on how the testing is to be carried out. These pertain to selecting the most useful
test article (single chip or entire board), nature of the test (static or dynamic [mainly

applicable to board testing]), assembling a test team, choosing the facility that provides the

best source of neutrons or protons for testing, scheduling and performing the test, coping with
uncertainties that appear during the test and finally, using the test results to calculate the
desired SEE rate for avionics. Many of these issues will be discussed in the following sections.
2 Availability of existing SEE data for avionics applications
Because of the diverse ways that SEE testing is carried out, and the multitude of venues for
how and where such data is published, the availability of SEE data for avionics applications is
not a simple matter.
2.1 Types of existing SEE data that may be used
SEE data can be derived from a number of different kinds of tests, and all of the differences
between these tests need to be understood in order to make comparisons meaningful.
Although there are many different types of single event effects, for the purposes of this PAS,
the focus is on three of them, single event upset (SEU), single event functional interrupt (SEFI)
and single event latchup (SEL). SEU pertains to the energy deposited by an energetic particle
leading to a single bit being flipped in its logic state. The main kinds of devices that are
susceptible to SEU are random access memories (RAMs, both SRAMs and DRAMs), field
programmable gate arrays (FPGAs, especially those using SRAM-based configuration) and
microprocessors (the cache memory and register portions). A SEFI refers to a bit flip in a
complex device that results in the device itself or the board on which it is operating in not
functioning properly. A typical example is an SEU in a control register, which can affect the
device itself, but can also be propagated to another device on the board, leading to board
malfunction. SEL refers to the energy deposited in a CMOS device that leads to the turning on
of a parasitic p-n-p-n structure, which usually results in a high current in the device and a
non-functioning state. The high energy neutrons in the atmosphere can induce all of these
effects: SEU, SEFI and SEL.
One of the important simplifying assumptions to be used in this PAS is that, for single event

effects, including SEU, SEFI and SEL, the response from high energy protons, i.e., those with E >
100 MeV, is the same as that from high energy neutrons of the same energy. The SEE response
is generally measured in terms of a cross section (cm²/dev), which is the number of errors of a
given type divided by the fluence of particles to which the device was exposed. Therefore, for the
SEU, SEFI and SEL cross sections, measurements made with high energy protons can be used
as the same cross section from the atmospheric neutrons. This is far more than an assumption,
since it has been demonstrated by direct measurement in many different devices [1-6] . In these
references, SEU was measured in the same devices using monoenergetic proton beams and
using the neutron beam from the Weapons Neutron Research (WNR) facility at the Los Alamos
National Laboratory. The energy spectrum of the neutrons in the WNR is almost identical to the
spectrum of neutrons in the atmosphere. An estimate of the SEE rate at aircraft altitudes in a
device can be obtained by the simplified equation:
2 2
SEE rate per device = 6 000 [n/cm h] × avionics SEE cross section [cm per device] (1)

Numbers in square brackets refer to the bibliography.

PAS 62396-2 © IEC:2007(E) – 7 –

Here, the integral neutron flux in the atmosphere, E >10 MeV, is taken to be 6 000 n/cm h,

the approximate flux at 40 000 ft (12,2 km) and 45° latitude [2]. This shows the importance of

the SEE cross section. As indicated above, the avionics SEE cross section is taken to be the

SEE cross section obtained from SEE tests with a spallation neutron source such as the WNR,

and also with a proton or neutron beam at energies > 100 MeV. The simplified approach of

Equation (1) is used in the Technical Specification IEC/TS 62396-1 [2] and is the nominal flux

under the above conditions.
A more elaborate approach for calculating the SEE rate is to utilize a number of

measurements of the SEE cross section as a function of neutron or proton energy, and
integrate the curve of the SEE cross section over energy with the differential neutron flux.
The details for this approach are given in the standard JESD-89A [7], although the neutron
flux given in this standard is at ground level and would have to be multiplied by approximately
a factor of 300 to make it relevant to avionics applications (see 2.1.3).
Thus the data that is most valuable for estimating the SEE rate in avionics is from SEE cross
section measurements made with: a) a spallation neutron source such as the WNR, b) a
monoenergetic proton beam and c) a quasi-monoenergetic neutron beam. Other SEE data that are
also valuable are SEU cross sections made with a monoenergetic 14 MeV neutron beam.
Based on comparisons of SEU cross section measurements with a 14 MeV neutron beam and the
WNR, the WNR SEU cross section is approximately a factor of 1,5 to 2 higher than the 14 MeV
SEU cross section for relatively recent devices ([4], feature size < 0,5 μm), and a factor of 4 times
higher for older devices [5]. For some of the very latest devices, the factor is close to 1.
2.1.1 Sources of data, proprietary versus published data
As indicated above, SEE cross section measurements that are relevant to avionics SEE rates
are being made by a variety of different groups. These include: a) space organizations that
use only monoenergetic proton beams for their SEE testing, b) IC vendors who use neutron
sources to measure the upset rate at ground level [which they refer to as the soft error rate
(SER), rather than the SEU rate, although the terms have the same meaning], c) avionics
vendors who use neutron sources to measure the upset rate at aircraft levels. Generally, SEE
data taken and reported by government agencies contain most if not all of the relevant
information, including identifying the specific IC devices tested and the providing the
measured SEU cross sections in unambiguous units. This applies to most of the proton data
taken and reported by NASA in the open literature by the NASA centres at GSFC and JPL.
GSFC and JPL invariably publish almost all of the proton SEE data that they take. However,
even though they disseminate essentially all of the results from the proton SEE testing that
they carry out, this is data that is usually reported in the open literature in an inclusive
compilation that contains results from SEE testing with both heavy ions and protons, thus the
proton SEE data has to be carefully sought out. Examples of the most recent NASA-GSFC
compilations of SEE testing containing proton SEE test results are given in [8-11], and
examples of JPL reports of SEE testing containing proton SEE test results are given in [12-14].

Other governmental agencies do not necessarily publish the results from all of the proton SEE
tests that they perform.
Data from the other sources, primarily private companies, is not nearly as accessible. IC
vendors perform a large number of tests, but only a small fraction of that data is reported
upon in the open literature. Furthermore, when the SEE data from IC vendors is published,
the results are often disguised so that the identity of the devices tested, or the part number is
usually hidden by using an arbitrary designation and the results are expressed in units that
are ambiguous at best and often of little use quantitatively. Sometimes, the data is expressed
in FIT units, which means errors per 10 device hours, however, this does not incorporate
information on how many bits are included in the device. If only the FIT value is given, this
can be converted into a SEE cross section by using the FIT definition and dividing by 14
[14 n/cm ·h is the flux of high energy neutrons (E > 10 MeV) at ground level in New York City,
which is the value recommended by the JESD-89A standard and so most often used.] Thus,
–9 2
FIT × 10 /14 gives the SEE cross section in cm /device.

– 8 – PAS 62396-2 © IEC:2007(E)

Some reports give the SER rate in units of FIT/Mbit, which allows the SEE cross section per

–15
bit to be calculated by multiplying as follows (FIT/Mbit) × 10 /14 to obtain the SEE cross

section in cm /bit. Other papers report the FIT value in arbitrary units (a.u.) which allows the

authors to show how the FIT rate varies with a particular parameter (e.g., applied voltage),

but it allows no quantitative assessment to be made of the SEE cross section. Examples of

such reports using FIT rates are given in [3, 15-18].

Most of the SEE data that we have been discussing comes from the SEE testing of individual

components, placing those devices in a beam of neutrons or protons and monitoring changes

in the status of the device for errors. A typical procedure is to fill a portion of memory in a

RAM with a specified bit pattern and monitor that memory for bit flips in one or more

addresses. However, some tests are done using an entire board to monitor when an error has
occurred. In this case, the malfunction of the board is an indication that an error has occurred,
and such an error is referred to as a SEFI, but the functional interruption is in the board rather
than the actual device being irradiated. If the beam is collimated such that only one or two
devices are exposed to the particles in the beam during each test, the likely source of error is
a SEE error in those devices. However, this is a dynamic type of test and it may be that the
device in the beam experienced the initial error which was propagated to another device on
the board, and faulty performance of the latter device is what lead to the board malfunctioning.
There are some reports of such board level tests in the open literature, but they are less
common. NASA-JSC has a requirement to perform such testing on all electronic boards that
will be going on the Space Shuttle and related programs. This testing is carried out with a
beam of protons, and while it is recorded in a NASA-JSC report, these reports are not widely
available, examples are given in [19-21]. Furthermore, the main purpose of the test is to
screen all of the devices for the potential of a hard error induced by the protons, such as a
single event latchup, so recoverable errors are not analyzed in great detail in these reports.
Other government agency groups also perform such board level SEE testing, and the results
of these tests are often reported in the literature, but are not included in any organized data
base. In addition, private companies carry out such board level testing, often for the benefit of
specific programs for avionics applications (neutron tests for avionics vendors) or space
applications (proton tests for low earth orbit spacecraft contractors), and this data is rarely
reported in the open literature.
2.1.2 Data based on the use of different sources
In general, all SEE testing is carried out using an accelerated source of neutrons or protons,
meaning that the device or board to be tested will receive a larger fluence of particles over a
given period of time in the test environment compared to the fluence it would receive during
that same time period in the intended vehicle in the atmosphere or space. In the past, testing
was usually carried out with only one type of source, but in recent times, some engineering
groups have been exposing devices to more than one type of particle environment and
comparing the SEE responses. Two main types of sources have been used for this SEE

testing for avionics applications, neutrons and protons, although there are a variety of
different kinds of neutron sources that have been used, as will be discussed below.
2.1.2.1 Data obtained using neutron sources
Single event effects, in particular, single event upset, can be induced by neutrons in two
distinct energy ranges, at high energies and at very low energies, called thermal neutron
energy. The high energy neutrons cause the SEU by the nuclear reaction with the silicon in
the IC that creates a recoil, and it is the energy from this recoil that is locally deposited in
other silicon atoms that directly causes the upset. For the purposes of simplification, neutrons
with energies > 10 MeV are of greatest concern, but it is true that neutrons with lower
energies, e.g. (2 to 3) MeV, can also cause SEUs. However, since the SEU cross section for
E < 10 MeV is considerably lower than the cross section for E > 10 MeV, 10 MeV is used as
an effective cut-off. Estimates of the SEU contribution for electronics technology with
geometry greater than 0,2 μm by neutrons with E < 10 MeV to the total SEU rate from the
entire WNR neutron spectrum is < 10 %, but for lower feature sizes, this fraction is expected

PAS 62396-2 © IEC:2007(E) – 9 –

to increase. This is roughly consistent with SEU measurements made with monoenergetic

neutrons (3 and 14 MeV) on devices of the mid 1990’s (feature size greater than 0,5 μm),

showing that the SEU cross section at 3 MeV for these older devices was about a factor of

100 lower than that at 14 MeV for most of the SRAMs tested [22]. However, for more recent

devices, especially those with feature sizes less than 0,2 μm and even down to 45 nm, the

contribution of neutrons with energies below 10 MeV, is expected to be in the (8 to 10) %

range.
For high energy neutrons, there are three different types of sources: 1) a spallation neutron

source which has neutrons with energies over a wide energy spectrum similar to that of the

atmospheric neutrons, 2) a quasi-monoenergetic neutron source that has a peculiar energy

spectrum, roughly half of the neutrons are at a peak energy and the other half are evenly
distributed between close to the peak and ~1 MeV, and 3) a 14 MeV neutron generator, the
only source that is close to being truly monoenergetic.
The WNR at Los Alamos which was mentioned previously is the best example of a spallation
neutron source, although the neutron irradiation facility at TRIUMF (Tri University Meson
Facility, in Vancouver, Canada) is another such source. Since the WNR facility was upgraded
around the year 2000, it is sometimes referred to by its new name, the ICE (Irradiation of
Chips and Electronics) House [23]. Figure 1 compares the neutron spectra from Los Alamos
(the ICE House), the neutron facility at TRIUMF and the atmospheric neutron spectrum at
ground level.
KEY
FACILITY Multiplication
Factor
Ground Line plot is
Spectrum ground level
[latitude multiplied by
45º North]
3 × 10
ICE 1
House
(WNR)
Measured
Spectrum
TRIUMF
at 100 μA
Figure 1 – Comparison of Los Alamos and TRIUMF neutron spectra
with terrestrial neutron spectrum
SEU data on devices that were exposed to the WNR neutron beam have been published in a
number of papers [3-5, 24-25], however, many more devices have been tested at Los Alamos
and those results are considered to be proprietary. These results have not been published,
nor are they expected to be published. Reference [26] indicates that in the year 2001, at least
eight different groups carried out SEE testing, and of these, we estimate that maybe two of
the testing groups may publish some of their results, an American national laboratory and a
university. The six private companies, both IC manufacturers and avionics vendors, will keep
their test results proprietary.
The TRIUMF facility in Canada, called the TNF (TRIUMF Neutron Facility) also provides a
spallation neutron source. Until 2004, it had received limited use, but since that time, a

– 10 – PAS 62396-2 © IEC:2007(E)

number of papers on SEU results from the testing of IC devices at the TNF have been

published [27].
There are a number of quasi-monoenergetic neutron sources around the world, including

some in the United States of America, but until recently they had not been used for testing

microelectronics for SEE. The site with the most experience with such tests is the Theodor

Svedberg Laboratory (TSL) at Uppsala University, Uppsala, Sweden [28]. A few papers have

been published reporting on the results of microelectronics devices being exposed to the TSL

neutron beam [Refs 6, 29, 30]. Methodologies have been developed for extracting SEU cross

section data at the pseudo-peak energy [29, 30]. In addition, a similar facility has been
operating in Japan at Tohoku University [31] which also has been used to make some SEU

measurements. A different methodology from that of the Swedish researchers has been

developed for extracting SEU cross section data at the pseudo-peak energy [32, 33].
In Figure 2, we combine SEU measurements made by several different groups at these
various facilities to illustrate how the high energy SEU cross section per bit for SRAMs has
varied with feature size over the last 5 or more years. The trend that is illustrated in Figure 2
shows a consistency within an approximate plateau region of 10 to 30, however we cannot
predict how this might change in the future, as feature sizes continue to decline below 0,1 μm.

KEY
Symbol Data
Intel
microprocessors, L1
Data Cache
SRAMs – Granlund
(>2000) Various
Vendors
SRAMs – Slayman
(>2000) Various
Vendors
Figure 2 – Variation of high energy neutron SEU cross section per bit
as a function of device feature size

The third kind of high neutron facility is one that provides essentially monoenergetic neutrons,
and 14 MeV, from the D-T reaction, is the highest energy of such a monoenergetic neutron
beam. A number of facilities in the United States and abroad have such neutron generators.
Tests on SRAM devices fabricated in the mid-1990s indicated that the SEU response per bit
from a spallation neutron source was 3 to 5 times higher than from a 14 MeV neutron source
[5]. Tests on more recent devices have shown a closer agreement in the SEU response
between a spallation neutron source and a 14 MeV neutron sources [4, 6]. This indicates that
for current, low voltage devices, 14 MeV neutrons provide a fairly good simulation of the
atmospheric neutrons with respect to inducing SEUs. However, 14 MeV neutrons do not
provide a good simulation with respect to inducing single event latchup (SEL) [34].
There is a fourth type of neutron facility that should be considered for testing devices for
inducing SEUs: with thermal neutrons. Thermal neutrons cause SEUs through the neutron
reactions with the isotope Boron, which can be present in high enough concentrations to be
of concern mainly as a constituent of the glassivation layer above an IC, i.e., in BPSG
(borophosphosilicate glass). Many devices use a different type of glassivation (e.g., PSG) and

PAS 62396-2 © IEC:2007(E) – 11 –

11 10
in some cases, the boron in the BPSG is Boron, so there are no B reactions leading to

7 10
SEU from the reaction products (alpha particle and Li) of the B interaction. A limited

amount of data has been published on the SEU cross section induced by thermal neutrons [6,

17, 35].
2.1.2.2 Data obtained using proton sources

It was demonstrated that high energy protons cause SEUs in microelectronics nearly 25 years

ago [36]. It was also recognized that at high energies the protons, even though they are

charged particles, cause the upsets by the same mechanism as the high energy neutrons, by

nuclear reactions with the silicon, rather than by direct ionization in the silicon. Proton SEU

cross sections have therefore been published over the years, but the effectiveness of the low
energy protons in causing upsets has increased over time, as the applied voltage to the ICs
has decreased below 5 V. Thus, for DRAMs made during the 1980s and tested with protons,
the SEU cross section decreased by more than an order of magnitude for proton energies <
50 MeV [37, 38]. For more recent devices, the SEU cross section has generally not decreased
very much with energy, the cross section due to 50 MeV protons being only about a factor of 2
higher than the cross section due to 14 MeV neutrons [4]. A very useful compendium of SEU
cross sections in more than 120 different SRAMs and DRAMs was compiled by ESA in 1997
[39], mostly on 5 V devices, but a few at 3,3 V, however, few if any of these devices are used
today. In contrast, most other papers in the open literature contain measured proton SEU
response data for fewer devices, roughly 4 to 8 devices.
2.1.3 Ground level versus avionics applications
There are a number of important differences between the SEU considerations for devices in
avionics applications and those on the ground. First and foremost, the neutron flux in the
atmosphere is much higher than it is on the ground, so the SEU rate is going to be
proportionally higher. The nominal difference is taken to be a factor of 300 between the
neutron flux at 40 000 ft (12,2 km) and on the ground. As explained in JESD-89A and in
various technical papers [40, 41], there are two main sources of upsets in devices on the
ground, the atmospheric neutrons and alpha particles from trace amounts of radioactive
materials within the IC package. As the nature of IC packaging has evolved over the years,
the specific components responsible for most of the alpha particle emissions have changed.
Today, the major source is the lead in solder bumps, but because there is a movement to
eliminate the use of lead in ICs, this too may change, although the replacement solder
material (e.g., tin-silver-copper or tin-silver-bismuth) may also emit low levels of alpha
particles, and so the alpha particle problem will not be going away, but changing.
At the ground level, for some devices, the SEU or SER rate due to the alpha particles from
the IC package may be similar to that from the atmospheric neutrons. For other devices, the
neutrons are the main source of the upsets. However, in avionics, with the neutron flux in the
atmosphere being more than 100 times the neutron flux on the ground, the SEU rate from the

alphas emitted by the package is very small compared to the rate from the neutrons. Thus,
the alpha particles from the IC package can be neglected as a source of upsets for avionics
applications.
As discussed in 2.1.1, for most ground level applications, the upset rate is quantified in terms
device hours. The reason for this is that
of the FIT rate, number of upsets in a device in 10
the testing and analysis is being done primarily by IC vendors and not by companies that sell
ground level systems. That has been changing over the last five years, especially after the
possibility of cosmic ray neutrons causing upsets was publicized in the general press [42].
This occurred with the article in Forbes magazine of November, 2000 that reported that Sun
servers were having problems, with dozens of machines crashing due to bit flips in the SRAM
used for the L2 cache memory which were caused by cosmic rays or alpha particles. Sun
Microsystems received a great deal of adverse publicity and hundreds of thousands of people
became aware of the fact the cosmic rays can cause errors in memory chips. In this case, the
problem was amplified because Sun initially blamed the vendor of the SRAMs [43].

– 12 – PAS 62396-2 © IEC:2007(E)

Sun Microsystems and its competitors in the server market (e.g., Cisco Systems) have

become very involved in neutron-induced upsets, testing devices and systems to quantify the

rates and designing error correcting schemes to protect their systems against individual errors.

The testing they perform is generally considered proprietary and so the results from these

tests are not available; this applies to the testing of both individual devices and entire

computer boards.
For ground level applications, it is likely that the IC vendors perform more neutron testing than

the server vendors, and their testing is almost always on individual devices. Nevertheless,

their SEU or SER results invariably remain proprietary. In some cases, they do publish their
results, and in that case the upset information is expressed in FIT units, with the identity of

the individual devices that were tested hidden by means of generic designations (e.g., part A,

part B1, etc.). When the data is published by the IC vendors, it is often presented at a
particular annual meeting, the International Reliability Physics Symposium (IRPS). Examples
of recent IRPS papers that contain information related to SEUs induced by the atmospheric
neutrons, although expressed in units that may not be directly usable, are given in [16-18, 33].
There is one group of IC vendors who are more open about their SEU testing results. These
are two microelectronics manufacturers who make FPGAs (field programmable gate arrays).
These companies are Xilinx and Actel. Examples of some of the papers that they have
published containing relevant SEU information are given in [44-46].
2.2 Sources of existing data
In the previous Subclauses, we have referred to diverse references in the open literature that
contain SEU cross section information from tests carried out with neutron and proton sources.
In Table 1 below, we compile descriptions of the SEU information contained in some of these
references, in particular those with the largest amount of data.

PAS 62396-2 © IEC:2007(E) – 13 –

Table 1 – Sources of existing data

Device tested or Particle type,

Data contained Ref. Comments
listed energy
20 SRAMs and Hi E proton and SEU cross section, 1 Devices not identified; SEU X-Stns

26 DRAMs WNR neutron cm /bit mixture of neutron and proton data

9 SRAMs Hi E proton, 14 SER rate, FIT/Mbit 3 Devices not identified; SER rates from

(0,14 to 0,5 μm) MeV neutron WNR and from proton measurements

and WNR
neutron
8 SRAMs Hi E proton and SEU cross section, 4 Devices not identified; SEU X-Stns
WNR neutron cm /bit from WNR and from proton data
(0,14 to 0,5 μm)
6 SRAMs, Hi E proton, SEU cross section, 5 Devices identified; SEU X-Stns from
WNR neutron, cm /bit WNR, 14 MeV and from proton data
2 μprocessors,
14 MeV neutron
2 FPGAs
6 SRAMs Hi E proton and SEU cross section, 6 Devices identified; SEU X-Stns from hi
neutron 14 MeV cm /bit E proton and neutron 14 MeV and
and thermal thermal neutron data
neutron
SRAMs, DRAMs, High energy Asymptotic SEU cross 13 Devices identified; SEU X-Sections
other devices protons section, cm / bit or per from high Energy proton
device measurements
6 SRAMs WNR neutrons SER rate, FIT/Mbit 15 Test devices, SOI and bulk, from two
vendors.
(0,25, 0,13, 0,09 μm)
6 SRAMs 150 MeV SEU cross section, 16 Test devices, vendor not identified, SOI
protons arbitrary units and bulk
(0,18, 0,13, 0,09 μm)
5 SRAMs 3 and 14 MeV SEU cross section, 22 Devices identified; SEU X-Sections
neutrons cm /bit from neutron data
24 SRAMs, 6 feature WNR neutrons SER, error/bit•h at   25 Devices and 4 vendors not identified
sizes 40 000 ft (12,2 km)
5 SRAMs Quasi-mono- SEU cross section, 29 Devices identified; mono-energetic
energetic cm /bit SEU X-Stns derived from
neutrons measurements
10 SRAMs Quasi-mono- SEU cross section, 30 Devices identified (10 of the 24 SRAMs
energetic cm /bit of 25); mono-energetic SEU X-Stns
neutrons derived from measurements
87 SRAMs, High energy SEU cross section,
39 All devices identified; devices tested
48 DRAMs, protons (20, 30, cm /bit between 1989 to 1996
10 EEPROMs, 50, 60, 100,
8 Flash EPROMs, 200, 300 and
8 UV EPROMs 500 MeV)
FPGA, 4 sections Hi energy SEU cross section 45 Device and portions of device
tested protons (cm /bit), SEFI cross (configuration memory block memory

power-on-reset and external ports)
section (cm /dev)
identified
3 Considerations for SEE testing
Testing for single event effects for avionics purposes involves the consideration of a variety of
factors. These factors include the type of hardware to be tested (individual device or entire
board), the type of test used (static or dynamic), and the type of the facility providing the
neutron or proton b
...