Theses Doctoral

Silicon photonic switching: from building block design to intelligent control

Huang, Yishen

The rapid growth in data communication technologies is at the heart of enriching the digital experiences for people around the world. Encoding high bandwidth data to the optical domain has drastically changed the bandwidth-distance trade-off imposed by electrical media. Silicon photonics, sharing the technological maturity of the semiconductor industry, is a platform poised to make optical interconnect components more robust, manufacturable, and ubiquitous. One of the most prominent device classes enabled by the silicon photonics platform is photonic switching, which describes the direct routing of optical signal carriers without the optical-electrical-optical conversions. While theoretical designs and prototypes of monolithic silicon photonic switch devices have been studied, realizing high-performance and feasible switch systems requires explorations of all design aspects from basic building blocks to control systems. This thesis provides a holistic collection of studies on silicon photonic switching in topics of novel switching element designs, multi-stage switch architectures, device calibration, topology scalability, smart routing strategies, and performance-aware control plane.

First, component designs for assembling a silicon photonic switch device are presented. Structures that perform 2×2 optical switching functions are introduced. To realize switching granularities in both spatial and spectral domains, a resonator-assisted Mach-Zehnder interferometer design is demonstrated with high performance and design robustness. Next, multi-stage monolithic switching devices with microring resonator-based switching elements are investigated. An 8×8 switch device with dual-microring switching elements is presented with a well-balanced set of performance metrics in extinction ratio, crosstalk suppression, and optical bandwidth. Continued scaling in the switch port count requires both an economic increase in the number of switching elements integrated in a device and the preservation of signal quality through the switch fabric. A highly scalable switch architecture based on Clos network with microring switch-and-select sub-switches is presented as a solution to reach high switch radices while addressing key factors of insertion loss, crosstalk, and optical passband to ensure end-to-end switching performance.

The thesis then explores calibration techniques to acquire and optimize system-wide control points for integrated silicon switch devices. Applicable to common rearrangeably non-blocking switch topologies, automated procedures are developed to calibrate entire switch devices without the need for built-in power monitors. Using Mach-Zehnder interferometer-based switching elements as a demonstration, calibration techniques for optimal control points are introduced to achieve balanced push-pull drive scheme and reduced crosstalk in switching operations. Furthermore, smart routing strategies are developed based on optical penalty estimations enabled by expedited lightpath characterization procedures. Leveraging configuration redundancies in the switch fabric, the routing strategies are capable of avoiding the worst penalty optical paths and effectively elevate the bottom-line performance of the switch device.

Additional works are also presented on enhancing optical system control planes with machine learning techniques to accurately characterize complex systems and identify critical control parameters. Using flexgrid networks as a case study, light-weight machine learning workflows are tailored to devise control strategies for improving spectral power stability during wavelength assignment and defragmentation. This work affirms the efficacy of intelligent control planes to predict system dynamics and drive performance optimizations for optical interconnect systems.


  • thumnail for Huang_columbia_0054D_15845.pdf Huang_columbia_0054D_15845.pdf application/pdf 8.02 MB Download File

More About This Work

Academic Units
Electrical Engineering
Thesis Advisors
Bergman, Keren
Ph.D., Columbia University
Published Here
July 6, 2020