1 Key Laboratory of the Ministry of Education for Optoelectronic Measurement Technology and Instrument, Beijing Information Science & Technology University, Beijing0092, China
2 Beijing ZX Intelligent Chip Technology Co., Ltd., Beijing100876,China
3 The 11th Research Institute of China Electronic Science & Technology Group Inc., Beijing100015,China
光子人工智能芯片以光速执行运算,且具有低功耗、延迟低、抗电磁干扰的优势。小型化与集成化是实现这一技术革新的关键步骤。本文将光刻技术运用于衍射光栅的制作,提出一种基于10.6微米激光的全光衍射深度学习神经网络光栅设计及实现方法。由于光源波长由毫米波向微米波进化,神经元的特征尺度缩小至20微米,与现有光衍射神经网络相比,深度学习神经网络特征尺寸缩小了80倍,为进一步实现光子计算芯片大规模集成奠定了基础。
光子芯片 衍射光栅 深度学习 神经网络 Photonic chip diffraction grating deep learning neural network
Author Affiliations
Abstract
1 Beijing Key Laboratory for Optoelectronics Measurement Technology, Beijing Engineering Research Center of Optoelectronic Information and Instruments, Beijing Information Science and Technology University, Beijing 100016, China
2 Beijing Laboratory for Biomedical Detection Technology and Instrument, Beijing Information Science and Technology University, Beijing 100192, China
We propose a novel non-contact rotational sensor based on a fiber Bragg grating (FBG) packaged in a core of a magnetic head, which converts the introduced strain from the circular magnetic railings ruler into the rotational information. A mathematical model is built for processing the data obtained by an interrogator, and the accuracy and resolution of the measurements are analyzed by altering the radius and period of the circular magnetic railings ruler, as well as the dimension of the sensor. The experimental results show that it is in good accordance with the theoretical analysis on rotational angle, and the fitting results indicate that the results obtained from the rotational sensor match very well with the real rotational velocity with a linearity of 0.998 and a standard error of about 0.01.
光电子快报(英文版)
2016, 12(6): 421