Professor Daniel Limbrick of North Carolina A&T presented at the Vanderbilt Seminar Series.
Fault injection studies on microprocessors examine how radiation‑induced soft errors and other physical faults affect the functional correctness of modern computational devices. As autonomous systems, AI accelerators, and safety‑critical platforms become ubiquitous, characterizing reliability at the processor level is essential for ensuring trustworthy operation in space, military, and emerging edge applications.
The resource details a hierarchical methodology for modeling and simulating soft errors across microcontroller‑ to microprocessor‑scale designs. It describes the use of an open‑source RTL‑to‑GDSII flow (OpenLane/LibraLane) to extract design‑space metrics, perform transistor‑level fault injection, and evaluate error propagation through critical circuit nodes identified via large‑scale ALU campaigns. Mitigation strategies such as selective transistor sizing, exploitation of inherent design slack, and targeted hardening of high‑vulnerability nodes are presented, together with quantitative assessments of performance, area, and power overheads.
With this knowledge, learners can articulate the motivation behind fault injection research, compare the effectiveness and cost of various mitigation techniques, and explain how soft‑error models are constructed and simulated in processor designs. They also acquire the ability to pinpoint the circuit nodes that most strongly influence reliability and to devise fault‑injection experiments that balance reliability gains against architectural trade‑offs.