片上多核系统
事件(粒子物理)
现场可编程门阵列
激光器
计算机科学
单事件翻转
嵌入式系统
电气工程
物理
工程类
计算机硬件
芯片上的系统
光学
量子力学
静态随机存取存储器
作者
Weitao Yang,Chenguang Zhang,Yaxin Guo,Wuqing Song,Heng An,Yonghong Li,Hang Zang,Chaohui He,Longsheng Wu
标识
DOI:10.1109/tns.2024.3384394
摘要
The article explores the susceptibility of a deep neural network implemented on a Xilinx Ultrascale+ MPSoC to single event effects under pulsed laser irradiations. The irradiation test involves scanning the MPSoC using a 1064nm wavelength single-photon laser with three different energy levels: 5nJ, 8nJ and 11nJ. During these scans, three types of soft errors were detected: misidentification number varied, system halt, and system re-initialization. The distribution of these soft errors across the MPSoC is determined. Notably, a significant burnout occurs when the current reaches approximately 1000mA, and this phenomenon does not reoccur once the threshold current is set to 800mA. Additionally, software fault injection is performed, and misidentification number variation and system halt soft errors are also detected during this process. These findings confirm a possible source of the observed soft errors in pulsed laser irradiation. It suggests that single event upsets in the configuration memory could be a potential cause of the observed soft errors during laser scanning.
科研通智能强力驱动
Strongly Powered by AbleSci AI