Por favor, use este identificador para citar o enlazar este ítem:
http://inaoe.repositorioinstitucional.mx/jspui/handle/1009/836| A method to estimate the reliability of FPGA-based architectures for space systems | |
| JUAN ANDRES PEREZ CELIS | |
| RENE ARMANDO CUMPLIDO PARRA | |
| Acceso Abierto | |
| Atribución-NoComercial-SinDerivadas | |
| Space systems FPGAs Reliability Markov chains Fault mitigation technique Voters placement | |
| The use of Static Random-Access Memory (SRAM)-based Field Programmable Gate Arrays (FPGAs) in space systems is gaining a growing interest. The high-computational power, the ability to reconfigure and the flexibility of FPGAs at low power make them suitable devices for space applications. Nevertheless, FPGAs are susceptible to Single Event Upsets which may cause an operational failure, leading to a decrease of the reliability. Fault Mitigation Techniques (FMT) are used to increase the reliability, making the architecture suitable for operating in the space. However, estimating the reliability of a FPGA design can be challenging. The preferred method consists of placing the FPGA under a high-energy particle beam and performing the test iteratively to estimate the reliability. The iterative process makes radiation testing prohibitive for many space applications, as the monetary cost and time spent for radiation testing is extremely high. This thesis proposes a method for determining the reliability of an FPGA architecture with FMTs applied based on the failure rate of an unmitigated design. The method uses Markov chains to model the architecture and estimate the reliability. An unmitigated design and a mitigated design were developed and placed under a neutron beam to estimate the reliability. The mitigated design decreases the sensible area, i.e. increases the reliability, by a factor of 20:18. The mitigated design is a triplication of the mitigated design with additional voters placed throughout the architecture. Two models resembling the behavior of the mitigated design with a Medium Grain Triple Modular Redundancy (MGTMR) are proposed. The difference of the models lies in the addition of shared-resources modules. The results of the models show that the estimation of reliability is close to that obtained through neutron testing. | |
| Instituto Nacional de Astrofísica, Óptica y Electrónica | |
| 2016-08 | |
| Tesis de maestría | |
| Inglés | |
| Estudiantes Investigadores Público en general | |
| Perez-Celis J.A. | |
| CIENCIA DE LOS ORDENADORES | |
| Versión aceptada | |
| acceptedVersion - Versión aceptada | |
| Aparece en las colecciones: | Maestría en Ciencias Computacionales |
Cargar archivos:
| Fichero | Tamaño | Formato | |
|---|---|---|---|
| PerezCJA.pdf | 12.48 MB | Adobe PDF | Visualizar/Abrir |