Alpha-particle-induced soft errors in dynamic memories

May, T.C.; Woods, M. H.

doi:10.1109/t-ed.1979.19370

Cited by 730 publications

(215 citation statements)

References 3 publications

Supporting

Mentioning

205

Contrasting

Unclassified

Order By: Relevance

“…5. Today's spacecraft systems operate in high-radiation environments with possible memory corruptions caused by cosmic rays or alpha particles [103,144]. For the layers of the memory hierarchy close to the CPU (e. g., registers, L1-cache, L2-cache, L3-cache, .…”

Section: Discussionmentioning

confidence: 99%

See 1 more Smart Citation

From Supervised to Unsupervised Support Vector Machines and Applications in Astronomy

Gieseke

2013

Künstl Intell

View full text Add to dashboard Cite

A common task in the field of machine learning is the classification of objects. The basis for such a task is usually a training set consisting of patterns and associated class labels. A typical example is, for instance, the automatic classification of stars and galaxies in the field of astronomy. Here, the training set could consist of images and associated labels, which indicate whether a particular image shows a star or a galaxy. For such a learning scenario, one aims at generating models that can automatically classify new, unseen images. In the field of machine learning, various classification schemes have been proposed. One of the most popular ones is the concept of support vector machines, which often yields excellent classification results given sufficient labeled data.However, for a variety of real-world tasks, the acquisition of sufficient labeled data can be quite time-consuming. In contrast to labeled training data, unlabeled one can often be obtained easily in huge quantities. Semi-and unsupervised techniques aim at taking these unlabeled patterns into account to generate appropriate models. In the literature, various ways of extending support vector machines to these scenarios have been proposed. One of these ways leads to combinatorial optimization tasks that are difficult to address.In this thesis, several optimization strategies will be developed for these tasks that (1) aim at solving them exactly or (2) aim at obtaining (possibly suboptimal) candidate solutions in an efficient kind of way. More specifically, we will derive a polynomial-time approach that can compute exact solutions for special cases of both tasks. This approach is among the first ones that provide upper runtime bounds for the tasks at hand and, thus, yield theoretical insights into their computational complexity. In addition to this exact scheme, two heuristics tackling both problems will be provided. The first one is based on least-squares variants of the original tasks whereas the second one relies on differentiable surrogates for the corresponding objective functions. While direct implementations of both heuristics are still computationally expensive, we will show how to make use of matrix operations to speed up their execution. This will result in two optimization schemes that exhibit an excellent classification and runtime performance.Despite these theoretical derivations, we will also depict possible application domains of machine learning methods in astronomy. Here, the massive amount of data given for today's and future projects renders a manual analysis impossible and necessitates the use of sophisticated techniques. In this context, we will derive an efficient way to preprocess spectroscopic data, which is based on an adaptation of support vector machines, and the benefits of semi-supervised learning schemes for appropriate learning tasks will be sketched. As a further contribution to this field, we will propose the use of so-called resilient algorithms for the automatic data analysis taking place aboard toda...

show abstract

Section: Discussionmentioning

confidence: 99%

“…Further, these systems operate in high-radiation environments with interference caused by cosmic rays and alpha particles [103,144]. 1 These disturbing influences can cause memory corruptions (e. g., bit flips) in the systems' memory hierarchies which, in turn, can severely affect the computations aboard such systems.…”

Section: Motivationmentioning

confidence: 99%

From Supervised to Unsupervised Support Vector Machines and Applications in Astronomy

Gieseke

2013

Künstl Intell

View full text Add to dashboard Cite

show abstract

“…[12] Because of their high rate of energy loss, low energy alpha particle production is a likely direct means to induce errors, since it has long been known that 5 MeV alpha particles from natural radioactivity are dangerous sources of SEU. [1] Reaction model calculations for protons and pions on Si provide spectra for emerging alpha particles that peak near this energy of 5 MeV [13], but measurements have not gone to such low energies.…”

Section: Reaction Mechanismsmentioning

confidence: 99%

“…[1] These temporary or soft errors did not damage the chip, but an erroneous 'zero' had replaced a 'one', or vice versa, spoiling the contents of the memory. This problem was traced to small amounts of alpha-emitting impurities in the materials.…”

Section: Introductionmentioning

confidence: 99%

Radiation-induced errors in memory chips

Peterson¹

2003

Braz. J. Phys.

View full text Add to dashboard Cite

We have measured probabilities for proton, neutron and pion beams from accelerators to induce temporary or soft errors in a wide range of modern 16 Mb and 64 Mb DRAM memory chips, typical of those used in aircraft electronics. Relations among the cross sections for these particles are deduced. Measurement of alpha particle yields from pions on aluminum, as a surrogate for silicon, indicate that these reaction products are the proximate cause of the charge deposition resulting in errors. I IntroductionIn 1978 a serious industrial problem was causing errors in the memory contents of Intel 2107 16Kb DRAM (Dynamic Random Access Memory) chips.[1] These temporary or soft errors did not damage the chip, but an erroneous 'zero' had replaced a 'one', or vice versa, spoiling the contents of the memory. This problem was traced to small amounts of alpha-emitting impurities in the materials. These 5 MeV alpha particles caused ionization in the Si of the chip, and the collected charge was enough to redefine the memory state.Later, in repair records from 1984, IBM found a correlation between these errors, here called 'soft error upsets' or SEU, and altitude. These errors were then shown to be closely correlated with cosmic ray intensities. [2,3] In Denver, at 1700m, the SEU rate was ten times the national average.It is easy to understand how heavily-ionizing low energy alpha particles, with a range of 25 microns in Si, could deposit disruptive charge, but how would high energy neutrons, protons and pions from cosmic ray reactions induce errors? At sea level, most cosmic rays are muons, with only small cross sections for any reactions, but evidently energetic hadrons can cause reactions within microcircuits to produce short range heavily ionizing reaction products.A series of experiments with intense accelerator beams has been used to measure the probabilities or cross sections for hadrons to induce SEU in a wide range of modern memory chips. These measurements are particularly important for commercial and military aircraft, with their high altitude operations.Here we define σ SEU = induced errors number of sample bits × particles/cm 2 . II Accelerator testingCommercially available memory chips were placed directly in our beams. Proton beams were mostly from the Harvard cyclotron, at energies from 50 to 150 MeV. III ComparisonsSEU cross sections for 16 Mb samples showed a very wide range, strongly dependent upon the chip construction. Cross sections are shown in Fig. 2. For some manufacturers, several different chips of the same model were studied, with little scatter. The large differences can be explained by the smaller volume of Si available to deposit stray charge in the newer generations.

show abstract

“…This definition is unambiguous in the case of dynamic RAMs [92], where the node "sensitive" to charge collected from ions and the node storing information are collocated. For digital circuits in which information storage is spatially "distributed" and involves regions not vulnerable to charge collection directly from ion tracks, the term Q., is still used to represent the total integrated charge perturbation associated with upset (such as SRAM cells, latches and combinational logic), although in these circuits, not only the total charge perturbation, Q, t, but also the time profile of charge deposition and transport to information-storage areas actually determines vulnerability to single events.…”

Section: Upset Threshold Measuresmentioning

confidence: 99%