Soluções híbridas de hardware/software para a detecção de erros em systems-on-chip (SoC) de tempo real

• Ano de defesa: 2006
• Autor(a) principal: Piccoli, Leonardo Bisch
• Orientador(a): Vargas, Fabian Luis
• Tipo de documento: Dissertação
• Tipo de acesso: Acesso aberto
• Idioma: por
• Instituição de defesa: Pontifícia Universidade Católica do Rio Grande do Sul; Porto Alegre
• Palavras-chave em Português: ENGENHARIA ELETRÔNICA; SISTEMAS ELETRÔNICOS; PROCESSAMENTO EM TEMPO REAL; CIRCUITOS INTEGRADOS; TOLERÂNCIA A FALHAS (INFORMÁTICA); HARDWARE; SOFTWARE
• Link de acesso: http://hdl.handle.net/10923/3163

The always increasing number of critical applications requiring real time systems associated with integrated circuits, high density and the progressive system power supply reduction, has made embedded systems more sensitive to the occurrence of transient faults. Techniques that explore the robustness increase in integrated circuits (SoC) by means of increasing the clock duty-cycle generated by the PLL block, in order to accommodate eventual undesired delays through the logic [1] are possible solutions to increase electronic systems reliability. It is said that such systems use “error avoidance” techniques. Other techniques whose goal is not to avoid fault occurrence, but instead, to detect them, are said “error detection” techniques. This work is focused on the second type of techniques in order to increase electronic systems reliability. In other words, this work proposes the development new techniques to perform fault detection at system runtime. Real-time systems depend not only on the logical computation result, but also on the time at which these results are produced. In this scenario, many tasks are executed and the efficient time scheduling is a great concern. During system execution in electromagnetic interference (EMI) exposed environments, there is the large probability of transient faults occurrence. Thus, the use of fault detection techniques prevents faults from propagating through the system till primary outputs and them producing systems defect (and/or compromising the time characteristic of the system). Basically, these detection techniques are classified in two main categories: solutions based on software and solutions based on hardware. In this context, the goal of this work is to specify and to implement a solution based on software techniques (described in C language and inserted in the RTOS kernel) and/or hardware (described in VHDL language and connected on the processor bus) that is capable of performing real time detection of eventual errors in Systems-on-Chips. The faults considered in this work are these that affect the correct processor control flow. The proposed solution is innovative int the sense of having as target systems, those operating is a preemptive multitasking RTOS environment. Therefore, the proposed techniques perform fault detection based on a hybrid solution that combines software (YACCA [2,3]) with hardware (WDT [4,5], OSLC [6,7] and SEIS [8,9,10]). Several system versions have been proposed and implemented. Then, they were validated in on electromagnetic environment according to the standard IEC 62132-2 [11], witch defines rules for testing integrated circuits under radiated EMI. The obtained results demonstrate that the proposed methodology is very efficient, since it yields a high fault detection coverage higher than those proposed by other methodology on the literature. In other works, the proposed work associates the smallest system performance degradation with the smallest memory overhead and the highest fault detection coverage.