【Abstract】

        Continued with the scaling of technology nodes, many technical benefits such as the faster clock frequency and the less area overhead are realized. However, such advance also leads to more vulnerable elec-tronic devices, in which soft errors may occur induced by radiation particles. Therefore, this project aims to develop a highly reliable system by radiation-hardened designs with an accurate analysis model for soft errors.
Soft-error analysis often requires tremendous Technology Computer Aided Design (TCAD) simulation to acquire a precise timing information of a single-event transient (SET). In the 1st year, we aim at building a high-precision machine-learning (ML) prediction model through the deep neural network and the particle swarm optimization. With such models, we can further harden the system by the Delay-Adjustable D-Flip-Flop (termed DAD-FF) to optimize its soft-error rate (SER) in the 2nd year. In the 3rd year, a monitoring mechanism of performance and vulnerability to radiation is developed and embeds a built-in self-test (BIST) in the system for measuring its online aging effect. In summary, we propose an aging- and radiation-aware framework which hardens designs by DAD-FF, and meanwhile a BIST circuit monitors its reliability and the performance simultaneously to provide robustness.
The overview of this project can be shown as figure below where an aging- and radiation-aware robust system is developed to supervise its functional correctness during operation. Once a warning pops up, we will early substitute such nearly-dying device before the system failure to prevent further serious problems.