• Skip to main content
  • Skip to search
  • Skip to footer
Cadence Home
  • This search text may be transcribed, used, stored, or accessed by our third-party service providers per our Cookie Policy and Privacy Policy.

  1. Blogs
  2. Breakfast Bytes
  3. Photonics Summit: Lumerical-Cadence Flow
Paul McLellan
Paul McLellan

Community Member

Blog Activity
Options
  • Subscribe by email
  • More
  • Cancel
curvycore
Lumerical
silicon photonics
Virtuoso
photonics

Photonics Summit: Lumerical-Cadence Flow

5 Dec 2019 • 3 minute read

 breakfast bytes logoJames Pond, the CTO of Lumerical, wrapped up the Photonics Summit recently. He started by giving an overview of the Lumerical-Cadence flow but the real focus of his short presentation was creating robust designs using photonics inverse design, or PID.

Last year at the Photonics Summit, I interviewed James. See my post An Illuminating Chat with Lumerical's CTO.

Lumerical-Cadence Flow

The heart of the Lumerical-Cadence flow is the capability to combine optical and electrical geometry in the same layout in the Virtuoso environment, and then seamlessly combine Spectre simulation for the electrical part with Lumerical's System Suite and Device Suite for the optical part. The flow is shown in the above diagram.

Virtuoso has been enhanced with CurvyCore to allow it to handle the curvilinear geometry that photonics requires, since light doesn't like to go around corners, it likes to go around bends. (See my blog post Yoga Is Passé, the Future Is CurvyCore.) There are more details about this Cadence-Lumerical integration in my interview with James Pond above, and also in my post Schematic-Driven Simulation and Layout of Complex Photonic ICs (although this post pre-dates CurvyCore so the layout part is now out-of-date).

Photonics Inverse Design

I first heard about inverse design last year, and I wrote about it in my post Superhuman Photonic Design. But there have been advances since then that James talked about at the Summit. One challenge with designing photonics devices is getting something that works at all. But the second challenge is to design them to be robust in the face of manufacturing tolerances. For transistors, we have the same problem and so we do a lot of simulation at various "corners". For photonics, the biggest problems are under and over-etching, the design being slightly undersized or being slightly oversized. This is equivalent to co-optimizing two devices, the slightly small one and the slightly large one. For example, the splitter shown below:

If you have seen the superhuman design post referenced above, you will know that the inverse design exploration comes up with structures that are beyond human comprehension but that meet the most aggressive specs. James had a photomicrograph of the manufactured device, which has an impressive measured insertion loss of 0.1dB.

He went into more details of what is going on under the hood and had a couple of other examples.

CompoundTek's Grating Coupler

 How well does inverse design work? Lumerical just put out a press release about some work they did with Singapore-based CompoundTek designing a grating coupler. This is a large, complex device that forms the interface between the end of an incoming fiber and taking the light into the chip world. At the scale of the chip, of course, optical fibers are really big. One of CompountTek's couplers is in this image. You only have to glance at it to see that it is clearly a design with a lot of degrees of freedom (20 carefully shaped slots swept out along arcs of a circle). 

Using photonics inverse design, they shrank the design by 20X. That's not a misprint for 20%, it was 20X smaller. Moreover, the design just took two weeks from start to tapeout. Working in the cloud they tried eight different variants in that time.

As described in the press release:

The SiPh grating coupler is a key functional block for photonic I/O, enabling light to be coupled from fiber into and out of a photonic integrated circuit. Unlike end-fire edge couplers, grating couplers can be located anywhere on chip, enabling additional applications such as sense and test. With this flexibility comes the requirement for reduced footprint while maintaining high coupling efficiency. However, with over 100 design parameters, grating couplers are geometrically complex, rendering traditional optimization techniques impractical. Lumerical’s PID technology enables designers to automatically and reliably generate optimal grating couplers with hundreds of free parameters.Testimony to its effectiveness, CompoundTek’s SiPh new grating couplers have been reduced in size by 20X and promise improved coupling efficiency.

 

Sign up for Sunday Brunch, the weekly Breakfast Bytes email.