Publication: Herramienta gráfica para guiar a investigadores a realizar experimentos bajo radiación memorias
Loading...
Official URL
Full text at PDC
Publication Date
2021
Authors
Advisors (or tutors)
Editors
Journal Title
Journal ISSN
Volume Title
Publisher
Citations
Exportar
Abstract
Según avanza la tecnología de fabricación de memorias, éstas se vuelven cada vez más sensibles a los efectos de la radiación natural del entorno que las rodea. Diversas fuentes de radiación (protones, neutrones, partículas alfa…) son capaces de inducir cambios en la lógica combinacional de estos dispositivos o incluso afectar al contenido de las celdas de memoria, corrompiendo la información que almacenan. Esto es debido a que las partículas incidentes, al impactar con las celdas de memoria, depositan la suficiente energía como para inducir cambios en su contenido (de 0 a 1 o viceversa). Aunque hay técnicas a la hora de fabricación para reducir la aparición de estos efectos, hasta ahora no es posible evitarlos por completo. Por ello, los equipos de investigación deben realizar pruebas simulando estos efectos, y una de las técnicas más populares y realistas de conseguirlo es exponer los dispositivos a la radiación en un acelerador de partículas. Cuando una partícula impacta con el dispositivo, ésta puede causar un evento simple, que consiste en el cambio de un único bit de información; o un evento múltiple que consiste en el cambio de varios bits de información físicamente cercanos. Cuando se analizan los resultados obtenidos tras la exposición de un dispositivo a la radiación, una problemática que surge es la probabilidad de aparición de “falsos eventos múltiples”, que se manifiestan como eventos múltiples, pero realmente se produjeron por acumulación de eventos simples que casualmente afectaron a celdas vecinas o cercanas. La probabilidad de aparición de dichos falsos eventos múltiples es mayor cuanto mayor sea el tiempo de exposición del dispositivo frente a la radiación y el flujo de partículas incidente. De este modo, el objetivo de este trabajo es crear una herramienta que ayude a los investigadores a visualizar el tiempo óptimo de exposición donde se obtendrán el máximo de resultados minimizando, o al menos controlando, la probabilidad de aparición de falsos eventos múltiples en el experimento.
As memory fabrication technology scales down, memories have become more and more sensitive to the effects of natural radiation of the surrounding environment. Different sources of radiation (protons, neutrons, alpha particles…) are able to induce changes in the combinational logic of these devices or even capable of affecting the contents of the memory cells, corrupting the information that they store. This phenomenon occurs when charged particles impact with memory cells and they deposit enough energy to flip their contents (from 0 to 1 or vice versa). Even though there exist techniques that mitigate the appearance of these effects, this is still a not fully resolved problem that needs to be addressed. Thus, research teams must simulate these effects, and one of the most popular and realistic techniques of attaining this objective is to expose devices against the radiation by using a particle accelerator. When a particle hits the device, it may cause a simple event, which consists in the alteration of one unique bit of information, or a multiple event, which consists in the alteration of several bits of information that are allocated physically close. When the results of exposing the device to the radiation are analyzed, an issue that arises is the appearance of “false multiple events” that manifest as multiple events but were, in fact, produced by accumulation of various simple events that, coincidentally, affected neighbor cells. The greater the time of exposure to the radiation and to incidental flux of particles, the greater is the probability of the aforementioned events to appear. Thus, the goal of this project is to create a tool that would help the researcher visualize the optimal time in which he/she will get the maximum number of results while minimizing, or at least, controlling the probability of appearance of false multiple events in the experiment.
As memory fabrication technology scales down, memories have become more and more sensitive to the effects of natural radiation of the surrounding environment. Different sources of radiation (protons, neutrons, alpha particles…) are able to induce changes in the combinational logic of these devices or even capable of affecting the contents of the memory cells, corrupting the information that they store. This phenomenon occurs when charged particles impact with memory cells and they deposit enough energy to flip their contents (from 0 to 1 or vice versa). Even though there exist techniques that mitigate the appearance of these effects, this is still a not fully resolved problem that needs to be addressed. Thus, research teams must simulate these effects, and one of the most popular and realistic techniques of attaining this objective is to expose devices against the radiation by using a particle accelerator. When a particle hits the device, it may cause a simple event, which consists in the alteration of one unique bit of information, or a multiple event, which consists in the alteration of several bits of information that are allocated physically close. When the results of exposing the device to the radiation are analyzed, an issue that arises is the appearance of “false multiple events” that manifest as multiple events but were, in fact, produced by accumulation of various simple events that, coincidentally, affected neighbor cells. The greater the time of exposure to the radiation and to incidental flux of particles, the greater is the probability of the aforementioned events to appear. Thus, the goal of this project is to create a tool that would help the researcher visualize the optimal time in which he/she will get the maximum number of results while minimizing, or at least, controlling the probability of appearance of false multiple events in the experiment.
Description
Trabajo de Fin de Grado en Ingeniería Informática, Facultad de Informática UCM, Departamento de Arquitectura de Computadores y Automática, Curso 2020/2021