We report a low-power silicon photonic ringresonator switch integrated with a CMOS driver that achieves sub 3.5-ns transition times, below -20-dB crosstalk, and ~1-mW combined switch and driver power consumption. Results are compared to nonresonant Mach-Zehnder based switches.
IntroductionDiscrete silicon photonic switches have previously demonstrated hundreds of Gb/s of throughput bandwidth and nanosecond-scale switching speeds [1][2][3][4][5]. Monolithic integration of these optical switches with digital complementary metal-oxide-semiconductor (CMOS) driver circuits is ultimately required to minimize power and area. Here, we compare two photonic switches based on ring resonator and Mach-Zehnder devices, each integrated with digital CMOS drivers, to explore fundamental tradeoffs in optical bandwidth, power dissipation and footprint.
Passive photonic components consisting of sub-micrometer waveguides, grating couplers, and ring resonators are demonstrated in an unmodified commercial complementary metal-oxide-semiconductor (CMOS) process. Waveguides demonstrate propagation losses of approximately 1 dB/mm at 1.3-µm wavelengths. Grating couplers achieve better than −5.8-dB coupling efficiencies. Ring resonators exhibit Qs greater than 5000 with extinction ratios of more than 20 dB. Furthermore, due to the utilization of front-end-of-the-line design layers for creating the waveguide cores coupled with a relatively thick buried-oxide layer standard within the process, no post-processing is required at the chip or wafer level to achieve the reported device performance. Hence, for the first time, designers without processing capabilities may design custom CMOS photonic circuits for a broad range of applications. One such application which may appropriately leverage the low-cost platform could be disposable bio-photonic sensor arrays.
We propose an HPC network architecture with co-packaged optics enabling 128-port 51.2-Tb/s switches. Simulations for a >34,000-GPU system show up to 11.2x throughput improvement over a Summit-like supercomputer, opening the way to direct-network-attached GPUs.
scite is a Brooklyn-based organization that helps researchers better discover and understand research articles through Smart Citations–citations that display the context of the citation and describe whether the article provides supporting or contrasting evidence. scite is used by students and researchers from around the world and is funded in part by the National Science Foundation and the National Institute on Drug Abuse of the National Institutes of Health.