1.6T OSFP-XD 2×DR4光收发模块系统性能分析
DOI:
CSTR:
作者:
作者单位:

1.华中科技大学,光学与电子信息学院;2.东莞立讯技术有限公司,光电产品处

作者简介:

通讯作者:

中图分类号:

基金项目:

湖北省自然科学基金创新群体项目(项目编号:2024AFA016)


1.6T OSFP-XD 2×DR4 optical transceiver module system performance analysis
Author:
Affiliation:

Photoelectric department, DongGuan Luxshare Technology Co., Ltd

Fund Project:

  • 摘要
  • |
  • 图/表
  • |
  • 访问统计
  • |
  • 参考文献
  • |
  • 相似文献
  • |
  • 引证文献
  • |
  • 资源附件
  • |
  • 文章评论
    摘要:

    在生成式AI、大规模分布式计算推动数据中心流量爆炸式增长的背景下,为满足数据中心向1.6T网络升级需求,1.6T OSFP-XD 2×DR4光收发模块的设计与研究至关重要。本论文提出双DSP协同调度架构的创新方案,采用OSFP-XD封装标准与PAM4技术,通过8个差分通道实现1.6Tbit/s总传输速率,内置前向纠错(FEC)算法与自适应LMS均衡器,保障高速信号处理性能。测试结果显示,模块在0°C、25°C、70°C下,发射端眼图性能稳定,平均光功率、消光比、TDECQ等指标均符合 IEEE802.3dj 协议;常温光谱宽度<0.12nm,边模抑制比(SMSR)>30dB,接收端各通道灵敏度均低于- 4.7dBm,满足协议要求。该模块可广泛应用于AI训练集群、超大规模云平台等场景,为下一代智能网络构建提供重要技术支撑。

    Abstract:

    Against the backdrop of explosive growth in data center traffic driven by generative AI and large-scale distributed computing, the design and research of the 1.6T OSFP-XD 2×DR4 optical transceiver module are essential to meet data centers’ requirements for upgrading to 1.6T networks. This paper proposes an innovative solution integrating a dual-DSP collaborative scheduling architecture . It adopts the OSFP-XD packaging standard and PAM4 technology, achieving a total transmission rate of 1.6 Tbit/s through 8 differential channels, and features built-in forward error correction (FEC) algorithms and adaptive LMS equalizers to ensure high-speed signal processing performance.Test results indicate the module maintains stable transmitter eye diagram characteristics at 0°C, 25°C, and 70°C, with key metrics including average optical power, extinction ratio, and TDECQ all complying with the IEEE 802.3dj standard. At room temperature, the spectral width is less than 0.12 nm, the side-mode suppression ratio (SMSR) exceeds 30 dB, and each receiver channel’s sensitivity is better than -4.7 dBm—fully meeting protocol specifications. The module can be widely deployed in AI training clusters, hyperscale cloud platforms, and other scenarios, providing critical technical support for building next-generation intelligent networks.

    参考文献
    相似文献
    引证文献
引用本文
分享
相关视频

文章指标
  • 点击次数:
  • 下载次数:
  • HTML阅读次数:
  • 引用次数:
历史
  • 收稿日期:2025-11-20
  • 最后修改日期:2025-12-05
  • 录用日期:2025-12-17
  • 在线发布日期:
  • 出版日期:
文章二维码