Design Automation For Silicon Photonics: Pushing Research Into Production

Silicon photonics is a transformative technology that will have a major impact on system architectures in future IC design applications. Already a major solution for Datacom applications and emerging applications in sensing, design techniques in silicon photonics, with the ability to leverage CMOS technology to integrate large numbers of photonic components, are now being applied to enable optical computing. In almost all of these applications, however, an electrical interface is required, not just to provide modulation and detection signals, but to deliver real-time tuning to overcome manufacturing variability and temperature sensitivity inherent to silicon photonic components. In many cases, the supporting electrical connectivity and requisite I/O become dominant in the overall design process which includes floor-planning and physical implementation in photonic applications. Design tools, particularly tools for photonic circuit automation, need to be jointly aware of both the photonics and the electronics.

Figure 1. Ising machine design dominated by electrical interconnect. From “HPE’s New Chip Marks a Milestone in Optical Computing” in IEEE Spectrum.

I’ve had a front row seat to development of design tools for photonics over the past seven years. It started with my involvement in a silicon photonics workshop in 2011. At that time, photonic workshops consisted of academics and researchers spending a week learning all the necessary details on every aspect of silicon photonic design, ranging from component design and modeling, to photonic circuit design to physical layout, and verification to manufacturability. Since then, the community has evolved to apply design methodologies created by EDA companies that rely on interoperable tool flows and trusted PDKs. This is discussed in the March 2018 edition of Laser & Photonics Reviews where Professors Wim Bogaerts and Lukas Chrostowski give an excellent overview of the state of available design tools, with ideas on how to enable design automation in their article, “Silicon Photonics Circuit Design: Methods, Tools and Challenges.”

While there are existing capabilities available, we know reapplying EDA tools and design methodologies created for IC design alone does not fully meet the needs of the photonics community. As a relatively young community, growing during a time of enormous compute resources, photonic design engineers are apt to delegate many of their tasks to scripting automation. Increasingly, Python scripting is becoming their language of choice. Python is a powerful, open-sourced language with broad usage and support by many of the leading photonic vendors including Luceda Photonics, Lumerical, VPIphotonics, and OptiWave. We see several startups and academic institutions leveraging Python in their custom, automated circuit layout solutions and photonic designers now come with inherent understanding of Python scripting.

In the summer of 2016, we began working with a major photonics customer to help identify an existing automation solution to meet their specific photonic design and manufacturing needs. In short, there were none available. In our discussions, we were able to narrow down the requirements for photonic automation to these 5 requirements:

• Multi-domain awareness
• Scalability
• Support design reuse
• Interoperability
• Python driven

Numerous interactions with the members of the photonic community have confirmed these requirements as the key basis for an automated solution.

While considerations for automation continue, Mentor is actively developing photonic solutions using Tanner L-Edit. Tanner L-Edit, with its support of advanced curve editing and native OpenAccess support, is inherently suitable for photonics layout design. At Professor Chrostowski’s “Hands on: Silicon Photonics Component Design & Fabrication” workshop at the OFC 2018 conference, Mentor introduced three key capabilities in Tanner L-Edit that enables designers to quickly assemble photonic circuits and simulate using Lumerical’s INTERCONNECT tool:

• Support of IP interoperability
• Native waveguide routing
• Photonic netlist extraction

These features are currently being refined and improved and will be made available as “L-Edit Photonics” in our next Tanner release. L-Edit Photonics provides designers with a complete photonic layout implementation flow that comes with standard integration to Calibre for physical verification and it includes the flexibility to perform design analysis using simulation solutions from Luceda, Lumerical, or VPIphotonics.

Figure 2: Placement of GSiP PDK ring modulator with electrical and photonics routing.

As part of its photonics solution, Mentor has developed a fully interoperable revision of the Generic Silicon Photonics PDK, more commonly known as the “GSiP” PDK. The GSiP PDK was first introduced in 2013 with the help of Professor Chrostowski to provide a vehicle for tool demonstration and academic training in silicon photonic design methodologies, but it is not tied to a foundry process. In many cases, we found it being employed as a quick-start technology package to enable photonic researchers to target their designs to multiple foundries. With this end in mind, we developed GSiP as an interoperable PDK, or iPDK. This leverages parameterized building blocks based on Python which enables them to be easily configurable to meet the technology of most silicon photonics foundries.

For more information Mentor’s photonic solutions and the new interoperable GSiP PDK, please visit the Mentor’s Tanner booth at the Design Automation Conference (San Francisco, June 24-28) and at the Mentor booth at ECOC 2018 (Rome, September 23-27).