基于DRSN的MG-Y激光器快速准连续波长调谐方法  

Fast Quasi-continuous Wavelength Tuning Method of MG-Y Laser Based on DRSN

作　　者：杜雨蒙 庄炜[1,2] 张旭 王乐 董明利 DU Yumeng;ZHUANG Wei;ZHANG Xu;WANG Le;DONG Mingli(Beijing Engineering Research Center of Optoelectronic Information and Instruments,Beijing Information Science&Technology University,Beijing 100016,China;Guangzhou Nansha Intelligent Photonic Sensing Research Institute,Guangzhou 511462,China)

机构地区：[1]北京信息科技大学光电测试技术及仪器教育部重点实验室,北京100016 [2]广州南沙光子感知技术研究院,广州511462

出　　处：《光子学报》2024年第8期160-169,共10页Acta Photonica Sinica

基　　金：国家自然科学基金(No.52375524);北京学者计划研究项目(No.BJXZ2021-012-00046)。

摘　　要：针对调制光栅Y分支(Modulated Grating Y-branch,MG-Y)可调谐半导体激光器难以在调谐范围内进行快速、准确的准连续波长调谐问题,提出了一种基于深度残差收缩网络的MG-Y激光器控制参数表快速生成方法。该方法使用光电二极管采集光功率,相较光谱仪采集波长的方法,其采集速率可大幅提高;采用深度残差收缩神经网络模型对MG-Y激光器电流调谐区域进行快速分类;结合激光器的电流调谐特性,采用拉格朗日插值方法,生成高精度控制参数表,实现MG-Y激光器快速、准确的准连续波长调谐。实验结果表明,用深度残差收缩模型对电流调谐区域分类的时间小于20 s,生成的控制参数表控制激光器输出波长精度为1 pm,稳定性为0.28 pm,法布里-珀罗标准具解调波长稳定性为1.8 pm,对光纤布拉格光栅应变传感器的应变解调精度为1με,稳定性为0.6με。该方法大幅提升了生成控制参数表的速度,满足光纤传感解调应用需求。The Modulated Grating Y-branch(MG-Y)tunable semiconductor laser stands out due to its short tuning time,wide tuning range,and high output power,making it a core device in the field of optical fiber sensing.However,achieving rapid and accurate quasi-continuous wavelength tuning within the tuning range of MG-Y lasers poses challenges,particularly in terms of the efficiency and accuracy of generating control parameter tables.Traditional wavelength tuning methods rely on spectrometers for wavelength acquisition,w hich a re t ime-consuming and costly,failing to meet the demands of high-precision optical fiber sensing applications.To address these issues,this paper proposes a rapid quasi-continuous wavelength tuning method for MG-Y lasers based on the Deep Residual Shrinkage Network(DRSN).This method aims to collect optical power through Photodiodes(PD)instead of spectrometers for wavelength acquisition,combined with the DRSN model to rapidly generate high-precision control parameter tables,thereby realizing fast and accurate tuning of MG-Y l asers t o m eet t he h igh-precision requirements of optical fiber sensing demodulation applications.The proposed method improves the process of control parameter table generation for MG-Y lasers.Instead of using wavelength meters,we employ photodiodes for collecting output optical power data,drastically reducing the data acquisition time from minutes to mere seconds.This significant speedup paves the way for more efficient subsequent processing steps.At the heart of our approach lies the DRSN model,which is specifically designed to rapidly classify the current tuning regions of the MG-Y laser.The model is trained on an extensive dataset comprising control currents,output optical power measurements,and precisely labeled tuning regions.The DRSN architecture incorporates residual modules,which alleviate the degradation problem commonly encountered in deep neural networks,ensuring that the model's performance remains stable as it grows deeper.Furthermore,the introduction of Residual

关 键 词：调制光栅Y分支激光器 波长控制 深度残差收缩网络 控制参数表 快速光功率采集 

分 类 号：TN248.4[电子电信—物理电子学]