Reliability-Book-2013-Ch 33

Total Page:16

File Type:pdf, Size:1020Kb

Reliability-Book-2013-Ch 33 107 37. EFFECT OF RADIATION DAMAGE – SOFT ERRORS 37.1 Review/Background: This lecture will focus on radiation-induced reliability issue, which is the fourth reliability mechanism other than NBTI, HCI, and TBBD. In the previous lecture, the importance of radiation related to a large of number of failures of memories as well as logic circuits is covered. Today, how radiation causes soft errors will be discussed. Soft errors refer to reversible errors, which is opposite to hard errors that result from punch- through and is permanent. 37.2 Source of radiation The source of the radiation determines impact on the transistors. For example, those come from cosmos are very different than those from solar wind or packaging. Thus, the understanding of radiation sources is very important to understand the potential damage. Figure. 37.1. Four different sources of radiation 107 108 From Figure. 37.1, we have three different types of sources. They are comic ray, solar wind, and packaging process. When we hold our cell phones in classroom, then the source of disturbing our electronics will be comic ray, some of which have enough energy to penetrate to the ground level. Solar wind, which is the radiation from the sun, causes problems to satellites that are out of the atmosphere primarily. Packaging itself has radiative components, such as trace amount of Thorium, which emits alpha particle. A danger situation is the interaction between comic ray and packaging materials. For example, there are two types of p-type doping material Boron: Boron-10 and Boron-11. There is a small fraction of Boron-10 in the packing materials. When Boron 10 absorbs low-energy comic ray, it breaks up as Li(7) and alpha particle – each of which creates a huge number of electron-hole pairs. Moreover, proton and α-particle are charged particles. According to Columbs law ,they will generate electron-hole pairs through electromagnetic interaction when they come through. Neutron, however, is not a charged particle. So it can penetrate a long distance without anything happening to it at most time. The longest distance it can go is 40 cm on average in Silicon based on Blackwall theory as discussed in the last lecture. So the likelihood of a neutron can hit a silicon atom in a 10-um device is very low. But when it happens, there will be a nuclear reaction between the neutron and silicon atom. The silicon atom will break apart and create three protons, a C-12, and a few α-particles. Therefore, once nuclear reaction happens, the device is guaranteed to fail. In order to calculate the probability of failure due to particles from a certain source, we must know three things – the flux of particles from the source, and the efficiency of electron/hole pair generation due to each particle, and finally, the critical charge necessary to upset operation. Let us being with calculating the flux. We will derive the Bethe formula for the efficiency of charge generation in the next chapter. Comic Ray : Let’s first talk about the cosmic ray. It consists of 92% proton, 6% α-particle (or 70% proton, 30% neutron). An important fact to know is that the cut-off of cosmic ray to reach the ground is 1 GeV. Any ray below this energy will be turned around by the earth’s magnetic field. As the plot shown in Figure. 37.2, the part we are primarily interested in is a small region on the top. The letters P, α, L, H and M stand for proton, α-particle, Lithium, Hydrogen, and Magnesium respectively. Interestingly, the 108 109 scattering time for these particles is 300 million years in the space. The reason is that the density of them in cosmos is 1.7 × 10 g/cm . So we may find only single one particle in miles and miles. Solar Wind . Next, we will take a look at solar wind. Unlike the earth, gas-like materials form the sun. So what actually happens is that the equator of the sun moves faster than the poles by around eight days. It results in tangled magnetic field that allows big holes, which is referred as the black sunspots. These are opening on the sun, through which the gas can escape. The energy density at surface is 1365 W/cm . But only a small portion of the solar wind can reach the earth and the other can deflected by the earth’s magnetic field. The process takes approximately 4 days. Figure. 37.2. Energy vs. Flux intensity (comic ray) One of the most interesting curves generated in the last century is shown in Figure. 37.3. 109 110 Figure. 37.3. Radiation Flux Density vs. Altitude The left side of the curve is the ground and the right side is the surface of atmosphere. The pressure measured by Hg presents the altitude. As shown, there is a peak of the flux near the surface of the atmosphere. The reason is that when the primary particles strike the atmosphere, it generates a creation number of electron-hole pairs with very high energy. Subsequently, a secondary generation that produces even much more pairs occurs due to these high-energy carriers. As the altitude goes down, recombination is dominant. It is the reason why the intensity decreases exponentially as approach the ground. A noteworthy fact is that a flying plane is near the height of the intensity peak, which explains why there is a guaranteed failure in one flight. Moreover, the source of the radiation damaging the satellites is solar wind due to similar reasons. We also can interpret the altitude dependency of radiation flux mathematically. @ − @ 37.1 = ( ) = 1033 − (0.0364 ) + (4.26 10 #) g/cm 37.2 In Eq. 37.1, I stands for the flux intensity. L presents the mean free time before the particles recombine. For example, = 100 g/cm for electron. A is a variable 110 111 depending on the height H of the position where we want to calculate the intensity as shown in Eq. 37.2. The exponential relation in Eq. 37.1 is due to the fact that recombination is exponential with position. Let’s do some simple calculation to demonstrate the formula. The height of Denver is about 5280 ft. So A equals to 1033 g/cm at sea level and 862 g/cm at Denver. The mean free used here is 14 g/cm . So the final number for I at Denver will be 3.4. On the other hand, a plane usually fly at 15 km above the sea level. So after calculation, the intensity will be 100 times larger than Denver. To proceed, let’s discuss about the radiation intensity distribution with respect to energy at the Earth level. On the left side of Figure. 37.4, there are a large number of particles generated by secondary generation so most of them are below 1 GeV and they are the most dangerous particles which will be discuss later. The list in Figure. 37.4 shows the composition of the flux and people are most concerned about neutrons and protons. Similar to the altitude distribution, people also developed mathematical model for energy distribution. @ − @ 37.3 = $ % 1.5 e(p *+(), = ln (/) +() = −5.2 0 2.6 + 0.6 0 0.01 + 0.00361 2 37.4 With Eq. 37.3and Eq. 37.4, it is convenient to calculate the flux intensity at given height and given energy. 111 112 Figure. 37.4. Energy distribution of different particles at the Earth level Packaging: Third, let’s move to radiation from packaging. There are several earth sources. For example, boro-phospho-silicate glass contains a certain number of thermal neutron and Lithium will inject proton as well as thermal proton. If taking a close look at Figure. 37.5, we will find the fluxes of α-particle coming from different components. Here, we don’t worry too much about the high-energy particle, which will zip through the device very fast so generate as many as electron-hold pairs as it could. Instead, the low-energy particles are most dangerous to the devices, because it will wander around and create a lot of pairs. Figure. 37.6 can explain it by showing the relation between particle energy and the number of generated charges. Figure. 37.5. Alpha particle fluxes from a variety of sources 112 113 Figure. 37.6. Charge generated vs. Alpha particle energy 37.3 Charge generation and potential fluctuation After study the sources of radiation, it is the time to see how radiation influences device. We will start with an intrinsic device. From Eq. 37.5 or Eq. 37.6, it is not difficult to calculate the absorption of the particles. Eq. 37.5 is for direct band materials and Eq. 37.6 is for indirect band. 7 (28) 9:; − /< 37.5 3456 = => : ?@ B / 4 7 8(:; − /<) 37.6 35A456 = 3 => : ?@?C:D We can first examine the case when particles strike on intrinsic structure as shown in Figure. 37.7 113 114 Figure. 37.7. Intrinsic structure with particle strike Figure. 37.8 Post-strike Net charge distribution Lrad is the radius of the strike and electron-hole pairs are created here. Afterwards, electrons and holes are diffusing toward the two ends of the device. Since they have different motilities, the speed of the diffusion of electron and hole is different then we will have net charge built up along the device as shown in Figure. 37.8. If the mobility of electron and hole is the same such as in graphene, however, the concentration of electron and hole will cancel each other out.
Recommended publications
  • Radiation-Induced Soft Errors in Advanced Semiconductor Technologies Robert C
    IEEE TRANSACTIONS ON DEVICE AND MATERIALS RELIABILITY, VOL. 5, NO. 3, SEPTEMBER 2005 305 Radiation-Induced Soft Errors in Advanced Semiconductor Technologies Robert C. Baumann, Fellow, IEEE Invited Paper Abstract—The once-ephemeral radiation-induced soft error has bit upset (SBU). While MBUs are usually a small fraction of become a key threat to advanced commercial electronic compo- the total observed SEU rate, their occurrence has implications nents and systems. Left unchallenged, soft errors have the poten- for memory architecture in systems utilizing error correction tial for inducing the highest failure rate of all other reliability mechanisms combined. This article briefly reviews the types of [3], [4]. Another type of soft error occurs when the bit that failure modes for soft errors, the three dominant radiation mech- is flipped is in a critical system control register such as that anisms responsible for creating soft errors in terrestrial applica- found in field-programmable gate arrays (FPGAs) or dynamic tions, and how these soft errors are generated by the collection of random access memory (DRAM) control circuitry, so that the radiation-induced charge. The soft error sensitivity as a function error causes the product to malfunction [5]. This type of soft of technology scaling for various memory and logic components is then presented with a consideration of which applications are most error, called a single event interrupt (SEFI), obviously impacts likely to require soft error mitigation. the product reliability since each SEFI leads to a direct product malfunction as opposed to typical memory soft errors that may Index Terms—Radiation effects, reliability, single-event effects, soft errors.
    [Show full text]
  • Chips for Discovering the Higgs Boson and Other Particles at CERN: Present and Future
    Chips for discovering the Higgs boson and other particles at CERN: present and future W. Snoeys CERN PH-ESE-ME, Geneva, Switzerland [email protected] Abstract – Integrated circuits and devices revolutionized particle physics experiments, and have been essential in the recent discovery of the Higgs boson by the ATLAS and CMS experiments at the Large Hadron Collider at CERN [1,2]. Particles are accelerated and brought into collision at specific interaction points where detectors, giant cameras of about 40 m long by 20 m in diameter, take pictures of the collision products as they fly away from the collision point. These detectors contain millions of channels, often implemented as reverse biased silicon pin diode arrays covering areas of up to 200 m2 in the center of the experiment, generating a small (~1fC) electric charge upon particle traversals. Integrated circuits provide the readout, and accept collision rates of about 40 MHz with on-line selection of potentially interesting events before data storage. Important limitations are power consumption, radiation tolerance, data rates, and system issues like robustness, redundancy, channel-to-channel uniformity, timing distribution and safety. The already predominant role of Figure 1. Aerial view of CERN indicating the location of the 27 silicon devices and integrated circuits in these detectors km LHC ring with the position of the main experimental areas the is only expected to increase in the future. ring. The city of Geneva, its airport, the lake and Mont Blanc are visible © 2015 CERN. Keywords-particle detection; silicon pin diodes; charge sensitive readout II. MAIN CONSTRAINTS AND LIMITATIONS I.
    [Show full text]
  • MARS-C: Modeling and Reduction of Soft Errors in Combinational Circuits
    MARS-C: Modeling and Reduction of Soft Errors in Combinational Circuits Natasa Miskov-Zivanov, Diana Marculescu Department of Electrical and Computer Engineering Carnegie Mellon University {nmiskov,dianam}@ece.cmu.edu ABSTRACT completely masked before it reaches the latch; Due to the shrinking of feature size and reduction in supply voltages, • latching-window masking – only if the glitch reaches the latch nanoscale circuits have become more susceptible to radiation induced and satisfies setup and hold time conditions, it will be latched. transient faults. In this paper, we present a symbolic framework based In this work, we estimate the likelihood that a transient fault will on BDDs and ADDs that enables analysis of combinational circuit lead to a soft error. Our main goal is to allow for symbolic modeling reliability from different aspects: output susceptibility to error, and efficient estimation of the susceptibility of a combinational logic influence of individual gates on individual outputs and overall circuit circuit to soft errors. We further use this framework to reduce the cost reliability, and the dependence of circuit reliability on glitch duration, of radiation hardening techniques by selectively resizing the gates that amplitude, and input patterns. This is demonstrated by the set of have the largest impact on circuit error. experimental results, which show that the mean output error The rest of this paper is organized as follows. In Section 2 we susceptibility can vary from less than 0.1%, for large circuits and outline the contribution of our work. In Section 3 we give an overview small glitches, to about 30% for very small circuits and large enough of related work.
    [Show full text]
  • Phonon Spectroscopy of the Electron-Hole-Liquid W
    PHONON SPECTROSCOPY OF THE ELECTRON-HOLE-LIQUID W. Dietsche, S. Kirch, J. Wolfe To cite this version: W. Dietsche, S. Kirch, J. Wolfe. PHONON SPECTROSCOPY OF THE ELECTRON- HOLE-LIQUID. Journal de Physique Colloques, 1981, 42 (C6), pp.C6-447-C6-449. 10.1051/jphyscol:19816129. jpa-00221192 HAL Id: jpa-00221192 https://hal.archives-ouvertes.fr/jpa-00221192 Submitted on 1 Jan 1981 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. JOURNAI, DE PHYSIQUE CoZZoque C6, suppZe'ment au nOl2, Tome 42, de'cembre 1981 page C6-447 PHONON SPECTROSCOPY OF THE ELECTRON-HOLE-LIQUID W. ~ietschejS.J. Kirch and J.P. Wolfe Physics Department and Materials Research ihboratory, University of IZZinois at Urbana-Champaign, Urbana, IL. 61 801, U. S. A. Abstract.-We have observed the 2kF cut-off in the phonon absorption of electron-hole droplets in Ge and measured the deformation potential. Photoexcited carriers in Ge at low temperatures condense into metallic drop- lets of electron-hole liquid (EHL).' These droplets provide a unique, tailorable system for studying the electron-phonon interaction in a Fermi liquid. The inter- 2 action of phonons with EHL was considered theoretically by Keldysh and has been studied experimentally using heat In contrast, we have employed mono- chromatic phonons5 to examine the frequency dependence of the absorption over the range of 150 - 500 GHz.
    [Show full text]
  • Scaling and Technology Issues for Soft Error Rates Allan
    Presented at the 4th Annual Research Conference on Reliability, Stanford University, October 2000 Scaling and Technology Issues for Soft Error Rates Allan. H. Johnston Jet Propulsion Laboratory California Institute of Technology Pasadena, California Abstract - The effects of device technology and scaling on Figure 2 shows how various contributions to the soft error rates are discussed, using information obtained from terrestrial error rate are affected by critical charge [3]. The both the device and space communities as a guide to determine work was done with SRAM cells, fabricated with a 0.35 µm the net effect on soft errors. Recent data on upset from high- CMOS process. For critical charge < 35 fC it is possible to energy protons indicates that the soft-error problem in DRAMs and microprocessors is less severe for highly scaled devices, in upset the cell with alpha particles. The largest contribution contrast to expectations. Possible improvements in soft-error from alphas comes from solder, but there is also a significant rate for future devices, manufactured with silicon-on-insulator contribution from impurities in the metallization. By technology, are also discussed. increasing the critical charge it is possible to eliminate errors from alpha particles, but terrestrial neutrons are still able to I. INTRODUCTION induce errors. The gradual decrease in neutron-induced error Soft-errors from alpha particles were first reported by rate with increasing critical charge is due to the distribution May and Woods [1], and considerable effort was spent by the of neutron energies, which extends over a very wide range. semiconductor device community during the ensuing years to Figure 2.
    [Show full text]
  • Ity in the Topological Weyl Semimetal Nbp
    Extremely large magnetoresistance and ultrahigh mobil- ity in the topological Weyl semimetal NbP Chandra Shekhar1, Ajaya K. Nayak1, Yan Sun1, Marcus Schmidt1, Michael Nicklas1, Inge Leermakers2, Uli Zeitler2, Yurii Skourski3, Jochen Wosnitza3, Zhongkai Liu4, Yulin Chen5, Walter Schnelle1, Horst Borrmann1, Yuri Grin1, Claudia Felser1, & Binghai Yan1;6 ∗ 1Max Planck Institute for Chemical Physics of Solids, 01187 Dresden, Germany 2High Field Magnet Laboratory (HFML-EMFL), Radboud University, Toernooiveld 7, 6525 ED Nijmegen, The Netherlands 3Dresden High Magnetic Field Laboratory (HLD-EMFL), Helmholtz-Zentrum Dresden- Rossendorf, 01328 Dresden, Germany 4Diamond Light Source, Harwell Science and Innovation Campus, Fermi Ave, Didcot, Oxford- shire, OX11 0QX, UK 5Physics Department, Oxford University, Oxford, OX1 3PU, UK 6Max Planck Institute for the Physics of Complex Systems, 01187 Dresden, Germany Recent experiments have revealed spectacular transport properties of conceptually simple 1 semimetals. For example, normal semimetals (e.g., WTe2) have started a new trend to realize a large magnetoresistance, which is the change of electrical resistance by an external magnetic field. Weyl semimetal (WSM) 2 is a topological semimetal with massless relativistic arXiv:1502.04361v2 [cond-mat.mtrl-sci] 22 Jun 2015 electrons as the three-dimensional analogue of graphene 3 and promises exotic transport properties and surface states 4–6, which are different from those of the famous topological 1 insulators (TIs) 7, 8. In this letter, we choose to utilize NbP in magneto-transport experiments because its band structure is on assembly of a WSM 9, 10 and a normal semimetal. Such a combination in NbP indeed leads to the observation of remarkable transport properties, an extremely large magnetoresistance of 850,000% at 1.85 K (250% at room temperature) in a magnetic field of 9 T without any signs of saturation, and ultrahigh carrier mobility of 5×106 cm2 V−1 s−1 accompanied by strong Shubnikov–de Hass (SdH) oscillations.
    [Show full text]
  • Radiation Hardening Efficiency of Gate Sizing and Transistor Stacking Based on Standard Cells
    Radiation hardening efficiency of gate sizing and transistor stacking based on standard cells Y.Q. Aguiar, Frédéric Wrobel, S. Guagliardo, J.-L. Autran, P. Leroux, F. Saigné, A.D. Touboul, V. Pouget To cite this version: Y.Q. Aguiar, Frédéric Wrobel, S. Guagliardo, J.-L. Autran, P. Leroux, et al.. Radiation hardening efficiency of gate sizing and transistor stacking based on standard cells. Microelectronics Reliability, Elsevier, 2019, 100-101, pp.113457. 10.1016/j.microrel.2019.113457. hal-02515096 HAL Id: hal-02515096 https://hal.archives-ouvertes.fr/hal-02515096 Submitted on 24 Mar 2020 HAL is a multi-disciplinary open access L’archive ouverte pluridisciplinaire HAL, est archive for the deposit and dissemination of sci- destinée au dépôt et à la diffusion de documents entific research documents, whether they are pub- scientifiques de niveau recherche, publiés ou non, lished or not. The documents may come from émanant des établissements d’enseignement et de teaching and research institutions in France or recherche français ou étrangers, des laboratoires abroad, or from public or private research centers. publics ou privés. Radiation Hardening Efficiency of Gate Sizing and Transistor Stacking based on Standard Cells Y. Q. Aguiara,*, F. Wrobela, S. Guagliardoa, J-L. Autranb, P. Lerouxc, F. Saignéa, A. D. Touboula and V. Pougeta a Institut d’Electronique et des Systèmes, University of Montpellier, Montpellier, France b Institut Materiaux Microelectronique Nanoscience de Provence, Aix-Marseille University, Marseille, France c Advanced Integrated Sensing Lab, KU Leuven University, Leuven, Belgium Abstract Soft error mitigation schemes inherently lead to penalties in terms of area usage, power consumption and/or performance metrics.
    [Show full text]
  • Design of Robust CMOS Circuits for Soft Error Tolerance
    Design of Robust CMOS Circuits for Soft Error Tolerance Debopriyo Chowdhury, Mohammad Amin Arbabian Department of EECS, Univ. of California, Berkeley, CA 94720 Abstract - With the continuous downscaling of technology, take care of the problems and analyze the merits and lowering of supply voltage and increase of operating demerits. Finally, we want to design robust latches and frequency, integrated circuits become increasingly susceptible combinational blocks that have good soft-error tolerance. to single event effects (SEE) caused by high energy particles The report is organized into four sections; section II covers like alpha particles, neutrons from cosmic rays etc. A SEU the background, origin and effect of soft error on may cause a bit flip in some latch or memory element, thereby altering the state of the system, leading to a ‘soft error’. Soft nanometer circuits. Section III is a literature review with an errors in memory have traditionally been a much greater analytical flavor, while section IV outlines the proposed concern than soft errors in logic circuits. However, as process work for the rest of the semester as well as shows some technology scales below 100 nanometers, voltage levels go initial simulation results. down and noise margins reduce, soft errors in logic circuits seem to be a potential threat too. In this work, we propose to analyze the effect of various circuit parameters on soft error susceptibility of logic circuits. Also, we plan to design a robust II. SOFT ERRORS: ORIGIN AND EFFECT ON latch that has simultaneous SET and SEU tolerance. INTEGRATED CIRCUITS Index Terms - Soft Errors, SET, SEU Hardened Latch A soft error occurs when a radiation event causes enough of a charge distribution to reverse or flip the data state of a memory cell, latch, flip-flop or even a node in a I.
    [Show full text]
  • Evaluation of Soft Errors Rate in a Commercial Memory Eeprom
    2011 International Nuclear Atlantic Conference - INAC 2011 Belo Horizonte,MG, Brazil, October 24-28, 2011 ASSOCIAÇÃO BRASILEIRA DE ENERGIA NUCLEAR - ABEN ISBN: 978-85-99141-04-5 EVALUATION OF SOFT ERRORS RATE IN A COMMERCIAL MEMORY EEPROM Luiz H. Claro1, A. A. Silva1, José A. Santos1, Suzy F. L. Nogueira2, Ary G. Barrios Jr2. 1 Divisão de Energia Nuclear Instituto de Estudos Avançados , IEAv Caixa Postal 6044 12231-970 São José dos Campos, SP [email protected] 2 Faculdade de Tecnologia São Francisco, FATESF Av. Siqueira Campos, 1174 12307-000 Jacareí, SP [email protected] ABSTRACT Soft errors are transient circuit errors caused by external radiation. When an ion intercepts a p-n region in an electronic component, the ionization produces excess charges along the track. These charges when collected can flip internal values, especially in memory cells. The problem affects not only space application but also terrestrial ones. Neutrons induced by cosmic rays and alpha particles, emitted from traces of radioactive contaminants contained in packaging and chip materials, are the predominant sources of radiation. The soft error susceptibility is different for different memory technology hence the experimental study are very important for Soft Error Rate (SER) evaluation. In this work, the methodology for accelerated tests is presented with the results for SER in a commercial electrically erasable and programmable read-only memory (EEPROM). 1. INTRODUCTION From all possible kinds of radiation damages to electronic components, the soft errors are included in the class of transitory errors induced by a single particle. In the other class, the hard errors, the ionizing radiation permanently changes the structural array in the electronic components.
    [Show full text]
  • Soft Error Modeling and Analysis of the Neutron Intercepting Silicon Chip (NISC) C
    Soft Error Modeling and Analysis of the Neutron Intercepting Silicon Chip (NISC) C. Çelik,1,2 K. Ünlü,1,2 N. Vijaykrishnan,3 M. J. Irwin3 Service Provided: Neutron Beam Laboratory Sponsors: National Science Foundation, U. S. Department of Energy, INIE Mini Grant, the Penn State Radiation Science and Engineering Center, and the Penn State Department of Computer Science and Engineering detailed records for particles, including the mother Introduction particle of the particle that causes the soft error. Advances in microelectronic technologies result in semiconductor memories with sub-micrometer NISC Simulation Model transistor dimensions. While the decrease in the dimensions satisfy both the producers’ and consumers’ The semiconductor device node represents the basic requirements, it also leads to a higher susceptibility of data storage unit in a semiconductor memory, and for the NISC design it is chosen to be as simple as possible the integrated circuit designs to temperature, magnetic 10 7 interference, power supply and environmental noise, in order to focus on the B(n,α) Li reaction. A cross and radiation. sectional view of the memory node model is illustrated in Figure 1. The BPSG layer is designed to produce Soft errors are transient circuit errors caused due to energetic α and 7Li particles, hence it acts as a source excess charge carriers induced primarily by external for producing soft errors. In a semiconductor memory, radiations. The Neutron Intercepting Silicon Chip (NISC) depending on the architecture and vendors, there are promises an unconventional, portable, power efficient different layers to produce depletion regions, gates, neutron monitoring and detection system by enhancing and isolation layers.
    [Show full text]
  • ECE 571 – Advanced Microprocessor-Based Design Lecture 17
    ECE 571 { Advanced Microprocessor-Based Design Lecture 17 Vince Weaver http://web.eece.maine.edu/~vweaver [email protected] 3 April 2018 Announcements • HW8 is readings 1 More DRAM 2 ECC Memory • There's debate about how many errors can happen, anywhere from 10−10 error/bit*h (roughly one bit error per hour per gigabyte of memory) to 10−17 error/bit*h (roughly one bit error per millennium per gigabyte of memory • Google did a study and they found more toward the high end • Would you notice if you had a bit flipped? • Scrubbing { only notice a flip once you read out a value 3 Registered Memory • Registered vs Unregistered • Registered has a buffer on board. More expensive but can have more DIMMs on a channel • Registered may be slower (if it buffers for a cycle) • RDIMM/UDIMM 4 Bandwidth/Latency Issues • Truly random access? No, burst speed fast, random speed not. • Is that a problem? Mostly filling cache lines? 5 Memory Controller • Can we have full random access to memory? Why not just pass on CPU mem requests unchanged? • What might have higher priority? • Why might re-ordering the accesses help performance (back and forth between two pages) 6 Reducing Refresh • DRAM Refresh Mechanisms, Penalties, and Trade-Offs by Bhati et al. • Refresh hurts performance: ◦ Memory controller stalls access to memory being refreshed ◦ Refresh takes energy (read/write) On 32Gb device, up to 20% of energy consumption and 30% of performance 7 Async vs Sync Refresh • Traditional refresh rates ◦ Async Standard (15.6us) ◦ Async Extended (125us) ◦ SDRAM -
    [Show full text]
  • Properties of Liquid Argon Scintillation Light Emission
    Properties of Liquid Argon Scintillation Light Emission Ettore Segreto∗ Instituto de F´ısica \Gleb Wataghin" Universidade Estadual de Campinas - UNICAMP Rua S´ergio Buarque de Holanda, No 777, CEP 13083-859 Campinas, S~aoPaulo, Brazil (Dated: December 14, 2020) Liquid argon is used as active medium in a variety of neutrino and Dark Matter experiments thanks to its excellent properties of charge yield and transport and as a scintillator. Liquid argon scintillation photons are emitted in a narrow band of 10 nm centered around 127 nm and with a characteristic time profile made by two components originated by the decay of the lowest lying 1 + 3 + ∗ singlet, Σu , and triplet states, Σu , of the excimer Ar2 to the dissociative ground state. A model is proposed which takes into account the quenching of the long lived triplet states through the + self-interaction with other triplet states or through the interaction with molecular Ar2 ions. The model predicts the time profile of the scintillation signals and its dependence on the intensity of an external electric field and on the density of deposited energy, if the relative abundance of the unquenched fast and slow components is know. The model successfully explains the experimentally observed dependence of the characteristic time of the slow component on the intensity of the applied electric field and the increase of photon yield of liquid argon when doped with small quantities of xenon (at the ppm level). The model also predicts the dependence of the pulse shape parameter, Fprompt, for electron and nuclear recoils on the recoil energy and the behavior of the relative light yield of nuclear recoils in liquid argon, Leff .
    [Show full text]