系统级电路辐照效应的复杂性对建模方法提出了高精度与高速度的双重要求。CMOS(互补金属氧化物半导体)工艺收发器作为系统级电路的核心常用器件,其总剂量效应的精准建模仿真至关重要。为此,本文提出一种适用于CMOS收发器的总剂量效应行...系统级电路辐照效应的复杂性对建模方法提出了高精度与高速度的双重要求。CMOS(互补金属氧化物半导体)工艺收发器作为系统级电路的核心常用器件,其总剂量效应的精准建模仿真至关重要。为此,本文提出一种适用于CMOS收发器的总剂量效应行为级仿真方法:采用输入输出缓冲区信息规范(input/output buffer information specification,IBIS)模型表征Hi-1573器件的缓冲区特性,通过VHDL-AMS语言完成器件功能区的精细化建模。为验证方法有效性,开展了^(60)Co伽马射线辐照实验,基于实验数据优化总剂量效应模块参数,将其与IBIS总剂量效应模型融合进行仿真。结果显示,仿真结果与实验数据的性能退化趋势高度吻合,充分证明了该行为级仿真方法在CMOS收发器总剂量效应建模中的可行性与可靠性。展开更多
To generate a neutron beam exhibiting a Maxwellian energy distribution with narrow emission angles for measuring the neutron capture reaction rates of the s-process nuclides,a monoenergetic 3.4 MeV proton beam produce...To generate a neutron beam exhibiting a Maxwellian energy distribution with narrow emission angles for measuring the neutron capture reaction rates of the s-process nuclides,a monoenergetic 3.4 MeV proton beam produced by the tandem-accelerator in the China Institute of Atomic Energy was utilized.The proton beam was first transmitted through a 60.5μm aluminum foil and then impinged on a natural LiF target to produce neutron beam via^(7)Li(p,n)7Be reaction.The quasi-Gaussian energy distribution of protons in the LiF target resulted in neutron energy spectra that agreed with a Maxwellian energy distribution at kT=(22±2)keV,which was achieved by integrating neutrons detected within an emission angle of 65.0°±2.6°using a ^(6)Li glass detector positioned at 65°relative to the proton beam direction.The narrow angular spread of the Maxwelliandistributed neutron beam enables direct measurement of neutron capture cross-sections for most s-process nuclides,overcoming previous experimental limitations associated with broad angular distributions.展开更多
近期提出的单体相移深度神经网络(single phase-shift deep neural network,SPDNN),因其网络规模小、学习精度高,成为首个复杂中子共振截面拟合与评价的实用深度学习工具。在SPDNN学习共振截面的过程中,诸多因素显著影响网络的训练效果...近期提出的单体相移深度神经网络(single phase-shift deep neural network,SPDNN),因其网络规模小、学习精度高,成为首个复杂中子共振截面拟合与评价的实用深度学习工具。在SPDNN学习共振截面的过程中,诸多因素显著影响网络的训练效果、训练效率以及训练模型的泛化性。这些因素包括:决定网络相移层大小的共振截面频谱范围与频段宽度、隐藏层的数目、每层神经元的数目、激活函数、损失函数、训练步数和训练数据的预处理等。为了进一步提升SPDNN在共振截面研究中的实用性,详细考察了这些因素对网络拟合性能的影响。通过考察,确定了SPDNN在共振截面研究中适宜的网络构建和训练方法,助力推动SPDNN的广泛应用。展开更多
文摘系统级电路辐照效应的复杂性对建模方法提出了高精度与高速度的双重要求。CMOS(互补金属氧化物半导体)工艺收发器作为系统级电路的核心常用器件,其总剂量效应的精准建模仿真至关重要。为此,本文提出一种适用于CMOS收发器的总剂量效应行为级仿真方法:采用输入输出缓冲区信息规范(input/output buffer information specification,IBIS)模型表征Hi-1573器件的缓冲区特性,通过VHDL-AMS语言完成器件功能区的精细化建模。为验证方法有效性,开展了^(60)Co伽马射线辐照实验,基于实验数据优化总剂量效应模块参数,将其与IBIS总剂量效应模型融合进行仿真。结果显示,仿真结果与实验数据的性能退化趋势高度吻合,充分证明了该行为级仿真方法在CMOS收发器总剂量效应建模中的可行性与可靠性。
基金National Natural Science Foundation of China(12125509,11961141003,12275361,U2267205,12175152,12175121)National Key Research and Development Project(2022YFA1602301)Continuous-support Basic Scientific Research Project。
文摘To generate a neutron beam exhibiting a Maxwellian energy distribution with narrow emission angles for measuring the neutron capture reaction rates of the s-process nuclides,a monoenergetic 3.4 MeV proton beam produced by the tandem-accelerator in the China Institute of Atomic Energy was utilized.The proton beam was first transmitted through a 60.5μm aluminum foil and then impinged on a natural LiF target to produce neutron beam via^(7)Li(p,n)7Be reaction.The quasi-Gaussian energy distribution of protons in the LiF target resulted in neutron energy spectra that agreed with a Maxwellian energy distribution at kT=(22±2)keV,which was achieved by integrating neutrons detected within an emission angle of 65.0°±2.6°using a ^(6)Li glass detector positioned at 65°relative to the proton beam direction.The narrow angular spread of the Maxwelliandistributed neutron beam enables direct measurement of neutron capture cross-sections for most s-process nuclides,overcoming previous experimental limitations associated with broad angular distributions.
文摘近期提出的单体相移深度神经网络(single phase-shift deep neural network,SPDNN),因其网络规模小、学习精度高,成为首个复杂中子共振截面拟合与评价的实用深度学习工具。在SPDNN学习共振截面的过程中,诸多因素显著影响网络的训练效果、训练效率以及训练模型的泛化性。这些因素包括:决定网络相移层大小的共振截面频谱范围与频段宽度、隐藏层的数目、每层神经元的数目、激活函数、损失函数、训练步数和训练数据的预处理等。为了进一步提升SPDNN在共振截面研究中的实用性,详细考察了这些因素对网络拟合性能的影响。通过考察,确定了SPDNN在共振截面研究中适宜的网络构建和训练方法,助力推动SPDNN的广泛应用。
