FPGAs are frequently used in network systems to provide the performance and flexibility that is required of modern computer networks while allowing network vendors to bring products to market quickly. Like all electronic devices, FPGAs are vulnerable to ionizing radiation which can cause applications operating on an FPGA to fail. These low-level failures can have a wide range of negative effects on the performance of a network system. As computer networks play a larger role in modern society, it becomes increasingly important that these soft errors are addressed in the design of network systems.This work presents a framework for testing the soft-error reliability of FPGA-based networking systems. The framework consists of the NetFPGA development board, a custom traffic generator, and a custom high-speed JTAG configuration device. The NetFPGA development board is versatile and can be used to implement a wide range of network applications. The traffic generator is used to exercise the network system on the NetFPGA and to determine the health of that system. The JTAG configuration device is used to manage reliability experiments, to perform fault injection into the FPGA, and to monitor the NetFPGA during radiation tests.This thesis includes soft-error reliability tests that were performed on an Ethernet switch network system. Using both fault injection and accelerate radiation testing, the soft error sensitivity of the Ethernet switch was measured. The Ethernet switch design was then mitigated using triple module redundancy and duplication with compare. These mitigated designs were also tested and compared against the baseline design. Radiation testing shows that TMR provides a 5.05x improvement in reliability over the baseline design. DWC provides a 5.22x improvement in detectability over the baseline design without reducing the reliability of the system.

Artificial neural networks (ANNs) are used in many types of computing applications. Traditionally, ANNs have been implemented in software, executing on CPUs and even GPUs, which capitalize on the parallelizable nature of ANNs. More recently, FPGAs have become a target platform for ANN implementations due to their relatively low cost, low power, and flexibility. Some safety-critical applications could benefit from ANNs, but these applications require a certain level of reliability. SRAM-based FPGAs are sensitive to single-event upsets (SEUs), which can lead to faults and errors in execution. However there are techniques that can mask such SEUs and thereby improve the overall design reliability. This thesis evaluates the SEU reliability of neural networks implemented in SRAM-based FPGAs and investigates mitigation techniques against upsets for two case studies. The first was based on the LeNet-5 convolutional neural network and was used to test an implementation with both fault injection and neutron radiation experiments, demonstrating that our fault injection experiments could accurately evaluate SEU reliability of the networks. SEU reliability was improved by selectively applying TMR to the most critical layers of the design, achieving a 35% improvement reliability at an increase in 6.6% resources. The second was an existing neural network called BNN-PYNQ. While the base design was more sensitive to upsets than the CNN previous tested, the TMR technique improved the reliability by approximately 7× in fault injection experiments.

A fault injection tool has been created to emulate single event upset (SEU) behavior within the configuration memory of an FPGA. This tool is able to rapidly and accurately determine the dynamic sensitive cross section of the configuration memory for a given FPGA design. This tool enables the reliability of FPGA designs and fault tolerance schemes to be quickly and accurately tested. The validity of testing performed with this fault injection tool has been confirmed through radiation testing. A radiation test was conducted at Crocker Nuclear Laboratory using a proton accelerator in order to determine the actual dynamic sensitive cross section for specific FPGA designs. The results of this radiation testing were then analyzed and compared with similar fault injection tests, with results suggesting that the fault injection tool behavior is indeed accurate and valid. The fault injection tool can be used to determine the sensitivity of an FPGA design to configuration memory upsets. Additionally, fault mitigation techniques designed to increase the reliability of an FPGA design in spite of upsets within the configuration memory, can be thoroughly tested through fault injection. Fault injection testing should help to increase the feasibility of reconfigurable computing in space. FPGAs are well suited to the computational demands of space based signal processing applications; however, without appropriate mitigation or redundancy techniques, FPGAs are unreliable in a radiation environment. Because the fault injection tool has been shown to reliably model the effects of single event upsets within the configuration memory, it can be used to accurately evaluate the effectiveness of fault tolerance techniques in FPGAs.

This dissertation studies the effects of radiation-induced single-event upsets (SEUs) on field-programmable gate array(FPGA)-based circuits. It analyzes and quantifies a special case in data communication, that is, the synchronization issue of signals when they are sent across clock domains in triple modular redundancy (TMR) circuits with the presence of SEUs. After demonstrating that synchronizing errors cannot be eliminated in such case, this dissertation continues to present novel synchronizer designs that can guarantee reliable synchronization of triplicated signals. Fault injection tests then show that the proposed synchronizers provide between 6 and 10 orders of magnitude longer mean time to failure (MTTF) than unmitigated synchronizers. This dissertation also studies the reliability of block random access memory (BRAM) on FPGAs. By investigating several previous reliability models for single-error correction/double-error detection (SEC/DED) memory with scrubbing, this dissertation proposes two novel MTTF models that are suitable for FPGA applications. The first one considers non-uniform write rates for probabilistic write scrubbing, and the second one combines deterministic scrubbing and probabilistic scrubbing into a single model. The proposed models reveal the impact of memory access patterns on the reliability of BRAMs. Monte Carlo simulations then demonstrate the correctness of the proposed models. At last, the memory access patterns of a type of FPGA application, digital signal processing (DSP) is studied, and mitigation mechanisms for DSP applications are discussed.

Deep learning is replacing many traditional data processing methods in computer vision, speech recognition, natural language processing and many more diverse end applications. Until only a few years ago, using deep learning networks for inference required large amount of computational resources such as memory, processing power and energy. It was not trivial to deploy the computationally-expensive deep neural networks on embedded devices with limited capabilities. In recent years however, deep learning is finding its way through the world of embedded devices. Embedded systems such as Internet of Things (IoT) devices, phones and even components in cars are being equipped with deep neural networks. This raises interesting challenges for both embedded designers and deep learning scientists to close the gap between these two domains. In this thesis work, some challenges involved in deploying deep learning for embedded systems were discussed, as well as some of the available solutions and frameworks. Moreover, focusing on the safety and fault-tolerance aspects of embedded systems, tolerance of deep neural networks against faults was investigated using an experiment-based research strategy. A fault injection framework was designed and implemented that targeted deep neural networks defined using PyTorch. The framework developed was used to perform fault injection experiments on a small deep learning network. It was found how faults have various impacts on the accuracy of the neural network depending on the type of layer targeted by the faults. Worst-case faults were identified and several architectural modifications on the deep neural network were examined to improve the fault tolerance of the neural network under study. / Djupinlärning ersätter många traditionella databehandlingsmetoder inom datorseende, röstigenkänning, språkteknologi och flera olika slutanvändningar. Fram tills några år sedan har användningen av djupinlärningsnätverk för inferens krävt stora mängder beräkningsresurser såsom minne, processorkraft och energi. Det är inte trivialt att distribuera det beräknings-dyra djupa neurala nätverket på inbyggda enheter med begränsad kapacitet. Under de senaste åren har dock djupinlärning nått världens inbyggda enheter. Inbyggda system som Internet of Things (IoT) -enheter, telefoner och till och med komponenter i bilar utrustas med djupa neurala nätverk. Detta medför intressanta utmaningar för både designers av inbyggda system och djupinlärningsforskare att minska klyftan mellan dessa två domäner. I detta examensarbete diskuteras några av utmaningarna med att distribuera djupinlärning för inbyggda system, samt tillgängliga lösningar och ramverk. Djupa neurala nätverks feltolerans undersöks. Ett ramverk för felinjicering skapades som riktade sig mot djupa neurala nätverk som använder sig av Pytorch. Ramverket användes sedan för att utföra felinjektionsexperiment på ett mindre djupinlärningsnätverk. Man kunde se hur felinjiceringar påverkade nätverkets noggrannhet på olika sätt beroende på vilket lager av nätverket felinjiceringen gjordes på. Den värsta sortens felinjicering identifierades och olika arkitektoniska modifieringar av det djupa neurala nätverket undersöktes för att förbättra feltoleransen av nätverket.

En raison de leur impact sur la fiabilité des systèmes, les effets du rayonnement cosmique sur l’électronique ont été étudiés dès le début de l’exploration spatiale. Néanmoins, de récentes évolutions industrielles bouleversent les pratiques dans le domaine, les technologies standard devenant de plus en plus attrayantes pour réaliser des circuits durcis aux radiations. Du fait de leurs fréquences élevées, des nouvelles architectures de transistor et des temps de durcissement réduits, les puces fabriquées suivant les derniers procédés CMOS posent de nombreux défis. Ce travail s’attelle donc à la simulation des aléas logiques permanents (SEU) et transitoires (SET), en technologies FD-SOI et bulk Si avancées. La réponse radiative des transistors FD-SOI 28 nm est tout d’abord étudiée par le biais de simulations TCAD, amenant au développement de deux modèles innovants pour décrire les courants induits par particules ionisantes en FD-SOI. Le premier est principalement comportemental, tandis que le second capture des phénomènes complexes tels que l’amplification bipolaire parasite et la rétroaction du circuit, à partir des premiers principes de semi-conducteurs et en accord avec les simulations TCAD poussées.Ces modèles compacts sont alors couplés à une plateforme de simulation Monte Carlo du taux d’erreurs radiatives (SER) conduisant à une large validation sur des données expérimentales recueillies sous faisceau de particules. Enfin, des études par simulation prédictive sont présentées sur des cellules mémoire et portes logiques en FD-SOI 28 nm et bulk Si 65 nm, permettant d’approfondir la compréhension des mécanismes contribuant au SER en orbite des circuits intégrés modernes / The effects of cosmic radiation on electronics have been studied since the early days of space exploration, given the severe reliability constraints arising from harsh space environments. However, recent evolutions in the space industry landscape are changing radiation effects practices and methodologies, with mainstream technologies becoming increasingly attractive for radiation-hardened integrated circuits. Due to their high operating frequencies, new transistor architectures, and short rad-hard development times, chips manufactured in latest CMOS processes pose a variety of challenges, both from an experimental standpoint and for modeling perspectives. This work thus focuses on simulating single-event upsets and transients in advanced FD-SOI and bulk silicon processes.The soft-error response of 28 nm FD-SOI transistors is first investigated through TCAD simulations, allowing to develop two innovative models for radiation-induced currents in FD-SOI. One of them is mainly behavioral, while the other captures complex phenomena, such as parasitic bipolar amplification and circuit feedback effects, from first semiconductor principles and in agreement with detailed TCAD simulations.These compact models are then interfaced to a complete Monte Carlo Soft-Error Rate (SER) simulation platform, leading to extensive validation against experimental data collected on several test vehicles under accelerated particle beams. Finally, predictive simulation studies are presented on bit-cells, sequential and combinational logic gates in 28 nm FD-SOI and 65 nm bulk Si, providing insights into the mechanisms that contribute to the SER of modern integrated circuits in orbit

CubeSat missions needs components that are tolerant against the radiation in space. The hardware components must be reliable, and it must not compromise the functionality on-board during the mission. At the same time, the cost of hardware and its development should not be high. Hence, this thesis discusses the design and development of a CubeSat architecture using a Commercial Off-The- Shelf (COTS) Multi-Processor System on Chip (MPSoC). The architecture employs an affordable Rad-Hard Micro-Controller Unit as a Supervisor for the MPSoC. Also, it uses several radiation mitigation techniques such as the Latch-up protection circuit to protect it against Single-Event Latch-ups (SELs), Readback scrubbing for Non- Volatile Memories (NVMs) such as NOR Flash and Configuration scrubbing for the FPGA present in the MPSoC to protect it against Single-Event Upset (SEU)s, reliable communication using Cyclic Redundancy Check (CRC) and Space packet protocol. Apart from such functionalities, the Supervisor executes tasks such as Watchdog that monitors the liveliness of the applications running in the MPSoC, data logging, performing Over-The-Air Software/Firmware update. The thesis work implements functionalities such as Communication, Readback memory scrubbing, Configuration scrubbing using SEM-IP, Watchdog, and Software/Firmware update. The execution times of the functionalities are presented for the application done in the Supervisor. As for the Configuration scrubbing that was implemented in Programmable Logic (PL)/FPGA, results of area and latency are reported. / CubeSat-uppdrag behöver komponenter som är toleranta mot strålningen i rymden. Maskinvarukomponenterna måste vara pålitliga och funktionaliteten ombord får inte äventyras under uppdraget. Samtidigt bör kostnaden för hårdvara och dess utveckling inte vara hög. Därför diskuterar denna avhandling design och utveckling av en CubeSatarkitektur med hjälp av COTS (eng. Custom-off-The-Shelf) MPSoC (eng. Multi Processor System-on-Chip). Arkitekturen använder en prisvärd strålningshärdad (eng. Rad-Hard) Micro-Controller Unit(MCU) som Övervakare för MPSoC:en och använder också flera tekniker för att begränsa strålningens effekter såsom kretser för att skydda kretsen från s.k. Single Event Latch-Ups (SELs), återläsningsskrubbning för icke-volatila minnen (eng. Non-Volatile Memories) NVMs som NOR Flash och skrubbning av konfigurationsminnet skrubbning för FPGA:er i MPSoC:en för att skydda dem mot Single-Event Upsets (SEUs), och tillhandahålla pålitlig kommunikation mha CRC och Space Packet Protocol. Bortsett från sådana funktioner utför Övervakaren uppgifter som Watchdog för att övervaka att applikationerna som körs i MPSoC:en fortfarande är vid liv, dataloggning, och Over- the-Air-uppdateringar av programvaran/Firmware. Examensarbetet implementerar funktioner såsom kommunikation, återläsningsskrubbning av minnet, konfigurationsminnesskrubbning mha SEM- IP, Watchdog och uppdatering av programvara/firmware. Exekveringstiderna för utförandet av funktionerna presenteras för den applikationen som körs i Övervakaren. När det gäller konfigurationsminnesskrubbningen som implementerats i den programmerbara logiken i FPGA:n, rapporteras area och latens.

This thesis evaluates the suitability of silicon-germanium technology for electronic systems intended for extreme environments, such as ambient temperatures outside of military specification (-55 degC to 125 degC) range and intense exposures to ionizing radiation. Silicon-germanium devices and circuits were characterized at cryogenic and high-temperatures (up to 300 degC) and exposed to ionizing radiation, providing empirical evidence that silicon-germanium is an excellent platform for terrestrial and space-based electronic applications.

Hydrocarbon exploration, global navigation satellite systems, computed tomography, and aircraft avionics are just a few examples of applications that require system operation at an ambient temperature, pressure, or radiation level outside the range covered by military specifications. The electronics employed in these applications are known as "extreme environment electronics." On account of the increased cost resulting from both process modifications and the use of exotic substrate materials, only a handful of semiconductor foundries have specialized in the production of extreme environment electronics. Protection of these electronic systems in an extreme environment may be attained by encapsulating sensitive circuits in a controlled environment, which provides isolation from the hostile ambient, often at a significant cost and performance penalty. In a significant departure from this traditional approach, system designers have begun to use commercial off-the-shelf technology platforms with built in mitigation techniques for extreme environment applications. Such an approach simultaneously leverages the state of the art in technology performance with significant savings in project cost. Silicon-germanium is one such commercial technology platform that demonstrates potential for deployment into extreme environment applications as a result of its excellent performance at cryogenic temperatures, remarkable tolerance to radiation-induced degradation, and monolithic integration with silicon-based manufacturing. In this dissertation the radiation response of silicon-germanium technology is investigated, and novel transistor-level layout-based techniques are implemented to improve the radiation tolerance of HBT digital logic.

Este trabalho aborda os soft errors em conversores de dados analógico-digitais e a mitigação usando redundância e diversidade. Nas tecnologias CMOS recentes, os efeitos singulares (SEEs, Single Event Effects) são um grupo de efeitos da radiação espacial que afetam a confiabilidade e disponibilidade dos sistemas. Os soft errors são SEEs que não danificam diretamente o sistema e podem ser posteriormente corrigidos. Seus principais subgrupos são o Single Event Upset (SEU), o Single Event Transient (SET) e o Single Event Functional Interrupt (SEFI). Uma das técnicas em nível de sistema amplamente usadas para proteger os circuitos eletrônicos desses efeitos é a Redundância Modular Tripla (TMR, Triple Modular Redundancy), que pode ainda ser melhorada com a adição da técnica de diversidade. Nesse contexto, esse trabalho adota um esquema baseado nessas duas técnicas para a implementação de um sistema de aquisição de dados (SAD) analógico-digital. Seus objetivos são observar o comportamento dos conversores de dados frente aos soft errors e avaliar a eficácia de um sistema baseado em TMR e diversidade espacial-temporal contra esses efeitos da radiação. A implementação desse SAD em um SoC (System-on-Chip) da Cypress Semiconductor, chamado PSoC 5LP e fabricado em tecnologia CMOS de 130 nm, propiciou a realização de dois estudos: no primeiro, é realizada a irradiação com nêutrons, caso de particular interesse para os equipamentos eletrônicos embarcados em aviões; e no segundo, são realizadas injeções de falhas por software e em tempo de execução nos registradores de controle dos periféricos e na SRAM do PSoC 5LP. O resultado da irradiação do primeiro estudo foi a não observância de erros, o que impediu cumprir os objetivos propostos para esse teste. Essa situação permitiu duas observações principais: primeiro, o fluxo de nêutrons do experimento é uma característica fundamental que impacta na capacidade de se observar os efeitos da radiação, principalmente quando a seção de choque do circuito em análise é baixa; e segundo, de que a probabilidade de ocorrerem mascaramentos de SETs nos circuitos combinacionais e analógicos é elevada, o que contribui significativamente para reduzir a sensibilidade desses circuitos. Para avaliar a eficácia do sistema baseado em TMR e diversidade espacial-temporal foi então realizada uma investigação teórica baseada em análise combinatória, e os resultados mostraram que a adição de diversidade temporal gera, em comparação ao TMR clássico, um ganho significativo na tolerância de falhas duplas e múltiplas, ao preço de um aumento do atraso do circuito. Os resultados das injeções de falhas por software e em tempo de execução nos registradores de controle dos periféricos e na SRAM mostraram que apenas um baixo percentual das falhas injetadas é detectado na forma de erros, convergindo para a justificativa de que os mascaramentos foram determinantes para a não observância de erros no primeiro estudo, de injeção de falhas por radiação. Também verificou-se que os registradores de controle dos periféricos são mais importantes no nível de aplicação do que os dados da memória SRAM. Considerações sobre a auto injeção de falhas e auto monitoramento sugerem que a utilização desses conceitos pode trazer diversas limitações e complicadores aos testes. / The present thesis addresses the soft errors in analog-to-digital data converters and mitigation of such errors using redundancy and diversity. In modern CMOS technologies, the Single Event Effects (SEEs) comprises an important group of space radiation effects that influence the reliability and availability of the systems. Soft errors are SEEs that do not directly damage the system and that can be further corrected. Their main subgroups are the Single Event Upset (SEU), the Single Event Transient (SET) and the Single Event Functional Interrupt (SEFI). One of the system level techniques broadly used to protect the electronic circuits against these effects is the Triple Modular Redundancy (TMR), which may be improved with the addition of the diversity technique. In this context, this work proposes a scheme based on these two techniques to implement a tolerant analog-to-digital data acquisition system (DAS). The main objectives are to observe the behavior of the data converters under soft errors, and evaluate the effectiveness of a system based on TMR and spatial-temporal diversity on mitigating these radiation effects. The implementation of this DAS in a Programmable SoC (System-on-Chip) from Cypress Semiconductor (PSoC 5LP) manufactured in 130 nm CMOS, allowed the development of two studies. In the first one, an irradiation with neutrons is performed, case of particular interest to electronic equipment embedded on planes. In the second study, runtime software fault injections are performed at the peripheral control registers and SRAM of the studied device. As a result from irradiation on the first study no errors were found, what does not allowed meet the objectives of this test. This situation allow two main observations: first, the neutron flux of the experiment is a key feature that influences the ability to observe the radiation effects, mainly when the cross section of the circuit in analysis is low; and second, the probability of occurring SETs masking in combinational and analog circuits is high, which contributes significantly to reduce the sensibility of these circuits. To evaluate the effectiveness of a system based on TMR and spatial-temporal diversity then was performed a theoretical investigation based on combinatorial analysis, and the results show that the addition of temporal diversity generates a significant gain in tolerating double and multiple faults, if compared to the classical TMR, at the price of an increase in the circuit delay. The results of the second study, performed by runtime software fault injections at the peripheral control registers and SRAM, showed that only a low percentage of injected faults is detected as errors, according to the justification that no errors were found on irradiation of neutrons due to masking. Also was verified that at the application level the peripheral control registers are more important than the data stored in the SRAM memory. Considerations for faults self-injection and self-monitoring were done, suggesting that the use of these concepts may bring numerous limitations to the test.