光通信研究2024,Issue(2) :109-115.DOI:10.13756/j.gtxyj.2024.220087

基于深度学习的电光调制型可编程光学频率梳

Electro-optic Modulation Programmable Optical Frequency Comb based on Deep Learning

李洪达 郑子昂 马洛嘉 马云杰 李培丽
光通信研究2024,Issue(2) :109-115.DOI:10.13756/j.gtxyj.2024.220087

基于深度学习的电光调制型可编程光学频率梳

Electro-optic Modulation Programmable Optical Frequency Comb based on Deep Learning

李洪达 1郑子昂 1马洛嘉 1马云杰 1李培丽1
扫码查看

作者信息

  • 1. 南京邮电大学 电子与光学工程学院、柔性电子(未来技术)学院,南京 210023
  • 折叠

摘要

[目的]为了满足各种应用领域对高性能光学频率梳(OFC)的需求,特别是在宽带宽、高平坦度、中心波长和谱线间距等参数可以独立调节方面,文章提出了一种基于2次耦合射频(RF)信号来驱动单个双驱动马赫—曾德尔调制器(DDMZM)以产生OFC的方法.[方法]通过使用单个乘法器来生成2次耦合RF信号,不仅增加了所产生OFC的梳线数量,而且具有结构简单和成本低廉等优点.此外,为了进一步提高OFC的优化效率和性能,文章还采用了基于深度学习的逆向设计和分析方法.[结果]研究结果显示,基于构建级联网络的逆向设计能够在不到1 s的时间内找到目标OFC的对应参数.这种快速参数确定方法不仅可以实现梳线数量、OFC功率和梳线间隔的可编程性,还能生成具有1.769 dB平坦度的13线OFC.文章所提高效率设计方法为OFC的快速制备和应用提供了强大的支持.[结论]文章所提方案在OFC的生成技术中具有显著优势,特别是在性能、灵活性和优化效率方面表现卓越.文章所提基于2次耦合RF信号驱动DDMZM的OFC生成方法,不仅简化了系统结构,降低了成本,还通过深度学习的逆向设计方法大幅提高了设计效率.这些特点使得文章所提方案能够满足广泛的应用需求,尤其适用于需要快速、高效且灵活调节OFC参数的场合.

Abstract

[Objective]To meet the diverse application demands for high-performance Optical Frequency Comb(OFC),especial-ly in terms of independently adjustable parameters like bandwidth,flatness,central wavelength,and spectral line spacing,a method based on secondary coupled Radio Frequency(RF)signals to drive a single Dual Drive Mach-Zehnder Modulator(DDMZM)for OFC generation is proposed.[Methods]Utilizing a single multiplier to generate the secondary RF coupled sig-nals not only increases the number of comb lines produced by the OFC but also offers the advantages of a simple structure and low cost.Additionally,to further enhance the optimization efficiency and performance of the OFC,a deep learning-based in-verse design and analysis approach is adopted.[Results]The study shows that the inverse design based on the constructed cas-caded network can identify the corresponding parameters for the target OFC in less than one second.This rapid parameter de-termination method enables programmability of the number of comb lines,OFC power,and line spacing.It can also generate a 13-line OFC with a flatness of 1.769 dB.This efficient design method provides robust support for the rapid preparation and ap-plication of OFCs.[Conclusion]The proposed solution in this study demonstrates significant advantages in OFC generation technology,particularly in performance,flexibility,and optimization efficiency.The method of generating OFC through DDMZM driven by secondary coupled RF signals not only simplifies the system structure and reduces costs but also significant-ly improves design efficiency through the reverse design approach of deep learning.These characteristics make this solution suitable for a wide range of applications,especially in scenarios requiring quick,efficient,and flexible adjustment of OFC pa-rameters.

关键词

光学频率梳/深度学习/逆向设计/平坦度

Key words

OFC/deep learning/inverse design/flatness

引用本文复制引用

基金项目

江苏省研究生科研与实践创新计划资助项目(46006CX21277)

江苏省大学生创新创业训练计划资助项目(202110293065Y)

出版年

2024
光通信研究
武汉邮电科学研究院企管部

光通信研究

CSTPCD北大核心
影响因子:0.327
ISSN:1005-8788
参考文献量26
段落导航相关论文