首页> 外文OA文献 >Silicon Photonics for All-Optical Processing and High-Bandwidth-Density Interconnects
【2h】

Silicon Photonics for All-Optical Processing and High-Bandwidth-Density Interconnects

机译:硅光子学用于全光学处理和高带宽密度互连

摘要

Silicon photonics has emerged in recent years as one of the leading technologies poised to enable penetration of optical communications deeper and more intimately into computing systems than ever before. The integration potential of power efficient WDM links at the first level package or even deeper has been a strong driver for the rapid development this field has seen in recent years. The integration of photonic communication modules with very high bandwidth densities and virtually no bandwidth-distance limitations at the short reach regime of high performance computers and data centers has the potential to alleviate many of the bandwidth bottlenecks currently faced by board, rack, and facility levels. While networks on chip for chip multiprocessors (CMP) were initially deemed the target application of silicon photonic components, it has become evident in recent years that the initial lower hanging fruit is the CMP's I/O links to memory as well as other CMPs. The first chapter of the thesis provides more detailed motivation for the integration of silicon photonic modules into compute systems and surveys some of the recent developments in the field. The second chapter then proceeds to detail a technical case study of silicon photonic microring-based WDM links' scalability and power efficiency for these chip I/O applications which could be developed in the intermediate future. The analysis, initiated originally for a workshop on optical and electrical board and rack level interconnects, looks into a detailed model of the optical power budget for such a link capturing both single-channel aspects as well as WDM-operation-related considerations which are unique for a microring physical characteristics. The holistic analysis for the full link captures the wavelength-channel-spacing dependent characteristics, provides some methodologies for device design in the WDM-operation context, and provides performance predictions based on current best-of-class silicon photonic devices. The key results of the analysis are the determination of upper bounds on the aggregate achievable communication bandwidth per link, identifying design trade-offs for bandwidth versus power efficiency, and highlighting the need for continued technological improvements in both laser as well as photodetector technologies to allow acceptable power efficiency operation of such systems.The third chapter, while continuing on the theme silicon photonic high bandwidth density links, proceeds to detail the first experimental demonstration and characterization of an on-chip spatial division multiplexing (SDM) scheme based on microrings for the multiplexing and demultiplexing functionalities. In the context of more forward looking optical network-on-chip environments, SDM-enabled WDM photonic interconnects can potentially achieve superior bandwidth densities per waveguide compared to WDM-only photonic interconnects. The microring-based implementation allows dynamic tuning of the multiplexing and demultiplexing characteristic of the system which allows operation on WDM grid as well device tuning to combat intra-channel crosstalk. The characterization focuses on the first reported power penalty measurements for on-chip silicon photonic SDM link showing minimal penalties achievable with 3 spatial modes concurrently operating on a single waveguide with 10-Gb/s data carried by each mode. The chapter also details the first demonstration of WDM combined with SDM operation with six separate wavelength-and-spatial 10-Gb/s channels with error free operation and low power penalties. The fourth, fifth, and sixth chapters shift in topic from the application of silicon photonics to communication links to the evolving use of silicon waveguides for nonlinear all-optical processing. The unique tight mode confinement in sub-micron cross-sections combined with the high response of silicon have motivated the development of four-wave mixing (FWM)-based processing silicon devices. The key feature of the silicon platform for these nonlinear processing platforms is the ability to finely and uniformly control the dispersive properties of the optical structures in a way that enables completely offsetting the material dispersion and achieve dispersion profiles required for effective parametric interaction of waves in the optical structures. Chapter four primarily introduces and motivates nonlinear processing in communication applications and focuses on recent achievements in non-silicon and silicon FWM platforms. Chapter five describes some of the author's contributions on parametric processing of high speed data in silicon nonlinear devices, with first of a kind demonstrations of wavelength conversion of 160-Gb/s optically time division multiplexed (OTDM) data as well as the wavelength-multicasting of a 320-Gb/s OTDM stream. The chapter then details a methodical characterization and demonstration of several record wavelength conversion experiments of data in silicon with 40-Gb/s data wavelength-converted across more than 100 nm with only 1.4-dB of power penalties as well as the wavelength and format conversion of 10-Gb/s data across up to 168 nm with sensitivity gains stemming from the format conversion of about 2 dB and a residual conversion penalty of only 0.1 dB, achieved by implementing an improved experimental setup. Both experiments highlight the performance uniformity of the conversion process for a wide range of probe-idler detuning settings, showcasing the silicon platform's unique broadband phase matching properties. The sixth chapter presents a slight shift in motivation for parametric processing from traditional telecom-wavelength applications to functionalities developed targeting mid-IR operation. Parametric-processing in the silicon platform at long wavelengths holds large potential for performance improvements due to the elimination of two-photon absorption in silicon at long wavelengths as well as silicon's dispersion engineering capabilities which uniquely position the silicon platform for effective phase matching of significantly wavelength detuned waves. Four-wave mixing signal generation and reception at mid-IR wavelengths are attractive candidates for tunable flexible operation with modulation and detection speeds which are currently only available at telecom wavelengths. With this vision in mind, several contributions detailing extension of FWM functionalities in silicon to operate at wavelengths close to 2 μm with performance equivalent to much smaller detuning setting measurements. The contributions detail the experimental demonstration of the first silicon optical processing functionalities achieved at such long wavelengths including the wavelength conversion and unicast of 10-Gb/s signals with up to 700 nm of probe-idler detuning, the combined two-stage 10-Gb/s FWM-link in which both data generation and detection at 1900 nm is facilitated by parametric processing in silicon with only 2.1-dB overall penalty, the first ever 40-Gb/s receiver at 1900 nm based on a FWM stage for simultaneous temporal demultiplexing and wavelength conversion, and lastly, the demonstration of a 40-Gb/s FWM-link operation with only 3.6 dB of penalty. The chapter concludes with a short discussion on possible extensions to enable silicon parametric processing at even longer wavelengths targeting the mid-IR spectral transmission window of 3-5 μm.
机译:近年来,硅光子学已经成为领先技术之一,有望使光通信比以往任何时候都更深入,更紧密地渗透到计算系统中。高效的WDM链路在第一层甚至更深层的集成潜力,已成为近年来该领域快速发展的强大动力。在高性能计算机和数据中心的短距离传输机制中,具有非常高的带宽密度并且几乎没有带宽-距离限制的光子通信模块的集成有可能缓解当前板,机架和设施级别面临的许多带宽瓶颈。虽然最初将片上网络用于芯片多处理器(CMP)视为硅光子组件的目标应用,但近年来已明显地表明,最初的低端成果是CMP的I / O链接到存储器以及其他CMP。本文的第一章为将硅光子模块集成到计算系统中提供了更详细的动机,并概述了该领域的一些最新进展。然后,第二章将继续详细介绍基于硅光子微环的WDM链路针对这些芯片I / O应用的可扩展性和功率效率的技术案例研究,该技术可在未来的将来开发。该分析最初是针对光和电板与机架级互连的研讨会发起的,它针对光链路的光功率预算的详细模型进行了研究,该模型既捕获了单通道方面的问题,又捕获了与WDM操作相关的独特考虑因素具有微环的物理特性。全链路的整体分析捕获了波长通道间隔相关的特性,为WDM操作环境中的设备设计提供了一些方法,并基于当前同类最佳的硅光子设备提供了性能预测。分析的关键结果是确定每个链路可实现的总通信带宽的上限,确定带宽与功率效率之间的设计折衷,并强调需要对激光和光电检测器技术进行持续的技术改进以允许第三章在继续主题硅光子高带宽密度链路的同时,继续详细介绍了基于微环的片上空分复用(SDM)方案的首次实验演示和表征。复用和解复用功能。在更具前瞻性的片上光网络环境中,与仅使用WDM的光子互连相比,启用SDM的WDM光子互连可以潜在地实现每个波导更高的带宽密度。基于微环的实现方式允许对系统的复用和解复用特性进行动态调整,从而允许在WDM网格上进行操作以及进行设备调整以应对通道内串扰。该特性集中于首次报告的片上硅光子SDM链路的功率损失测量,该结果显示了在单个波导上同时运行3种空间模式可实现的最小损失,每种模式所载数据均为10 Gb / s。本章还详细介绍了WDM与SDM操作相结合的第一个演示,它具有六个独立的波长和空间10 Gb / s通道,且无误操作和低功耗。第四,第五和第六章的主题从硅光子学的应用转移到通信链路,再到硅波导用于非线性全光处理的发展。亚微米横截面中独特的紧密模式限制以及硅的高响应性推动了基于四波混合(FWM)的处理硅器件的发展。硅平台对于这些非线性处理平台的关键特征是能够以一种能够完全抵消材料色散并实现波在波中有效参量相互作用所需的色散分布的方式,精细而均匀地控制光学结构的色散特性。光学结构。第四章主要介绍和激发了通信应用中的非线性处理,并重点介绍了非硅和硅FWM平台的最新成就。第五章介绍了作者对硅非线性器件中高速数据的参数处理的一些贡献。,首先演示了160 Gb / s光学时分多路复用(OTDM)数据的波长转换以及320 Gb / s OTDM流的波长多播。然后,本章详细介绍了一些数据记录波长转换实验的系统表征,并演示了40 Gb / s数据在100纳米以上仅以1.4 dB的功率损失进行波长转换的40 Gb / s数据,以及波长和格式转换通过实施改进的实验设置,可在高达168 nm的波长范围内对10 Gb / s数据进行处理,并获得约2 dB的格式转换带来的灵敏度增益,而残留转换损失仅为0.1 dB。两项实验均突出显示了在多种探头-惰轮失谐设置下转换过程的性能均匀性,展示了硅平台独特的宽带相位匹配特性。第六章介绍了参数处理的动机从传统的电信波长应用到针对中红外操作开发的功能的微小转变。由于消除了长波长硅中的双光子吸收以及硅的色散工程能力,硅平台中的长波长参数化处理具有巨大的性能提升潜力,而硅的色散工程能力使硅平台独特地定位,可实现有效的显着波长相位匹配失谐的波浪。具有中红外波长的四波混频信号的产生和接收是具有调制和检测速度的可调谐灵活操作的诱人候选,而调制和检测速度目前仅在电信波长下可用。考虑到这一愿景,有几篇文章详细介绍了硅中FWM功能的扩展,以在接近2μm的波长下工作,其性能相当于更小的失谐设置测量值。这些文稿详细介绍了在如此长的波长下实现的首个硅光学处理功能的实验演示,包括波长转换和10 Gb / s信号的单播,以及高达700 nm的探头-惰轮失谐,组合的两级10 Gb / s FWM链路,其中通过在硅中进行参数处理而仅以2.1 dB的总体代价促进了1900 nm处的数据生成和检测,这是有史以来第一个基于FWM级在1900 nm上同时进行时间同步的40 Gb / s接收器解复用和波长转换,最后是40 Gb / s FWM链路操作的演示,其代价仅为3.6 dB。本章以简短的讨论结束,讨论了可能的扩展,以使更长的波长下的硅参数化处理成为可能,这些波长针对3-5μm的中红外光谱传输窗口。

著录项

  • 作者

    Ophir Noam;

  • 作者单位
  • 年度 2013
  • 总页数
  • 原文格式 PDF
  • 正文语种 English
  • 中图分类

相似文献

  • 外文文献
  • 中文文献
  • 专利

客服邮箱:kefu@zhangqiaokeyan.com

京公网安备:11010802029741号 ICP备案号:京ICP备15016152号-6 六维联合信息科技 (北京) 有限公司©版权所有
  • 客服微信

  • 服务号