Integrated Photonics (Part 3)

Experts at the Table, part 3: What is the role of EDA in building photonics systems, is a viable IP market developing and how are photonics systems tested?


Semiconductor Engineering sat down to discuss the status of integrated photonics with Twan Korthorst, CEO for PhoeniX Software; Gilles Lamant, distinguished engineer for Cadence; Bill De Vries, director of marketing for Lumerical Solutions; and Brett Attaway, director of EPDA solutions at AIM Photonics, SUNY Polytechnic Institute. What follows are excerpts of that conversation. Part one can be found here and part two here.

Fig. 1: Photonics panel. Photo by Brian Bailey

SE: What role does EDA play in integrated photonics?

Lamant: We were called to the rescue. When it comes to doing photonics in the lab they were good, but when it came to making a chip with a predictable schedule that had photonics on it, pulled together with the electronics, and having the whole system working together, that was a disaster. They could not predict or tell you they would get the chip out on time. It was often taking them several tapeouts before they got something that worked. System simulation, even on the photonics side, was not done. A lot of chips were designed but had never been simulated. Some chips had not used before being sent to manufacturing. Some were not LVSed before manufacturing.

De Vries: Not that there ever was an S. They just drew it by hand.

Lamant: Exactly, and this is the role that EDA plays.

Korthorst: We start with a single component, then a few components connected together as a simple circuit, but the complexity is going up.

Lamant: It is like 25 years ago when people doing analog were drawing it by hand and saying, ‘I can do it all by hand.’ This works for maybe 10 transistors, but if you want complex modulation schemes, you can’t do it like that.

Attaway: System enablement. EDA or should we say PDA holds the key to enabling anyone to get a system designed with photonics in the middle of it. Another challenge is skillsets. Right now, most photonic design groups are made up from PhDs and they are hard to come by.

Korthorst: And finding someone who understands both the physics and the process is a problem.

Attaway: Yes, the role of EDA is to allow photonics to be an added piece of the repertoire of today’s analog or RF IC designer. Being able to add this in as a skillset, where you have design teams that are skilled at getting chips taped out and in the fab — that is lacking in the photonics world. EDA can enable it. Higher levels of abstraction are critically needed and it is possible to move in that direction.

De Vries: We sit one level lower with our tools. I hesitate to use the term, but it is kind of like the photonics TCAD world where what we are doing is component design. Half of our toolset is directed toward the Ph.D.’s who take our tools and they understand how to design one thing. Their application may be one particular photo detector or modulator, and they want to determine if it is functioning the way it is supposed to. Stitching those things together – they do it because they can. But what the AIM organization and EDA tools do is enable you to look at this foundry process, which essentially gives me pure confidence that if I take the pieces they give me within the PDK, I implement them, I pass the design rules, and whatever it is that I ask them to fabricate will be fabricated. We are not yet 100% where that is the case. The amount of vertical integration, understanding of components process, how I stitch them together for the system, is still very specialized.

SE: Can you buy IP components for photonics?

Attaway: We are moving there.

De Vries: One of the things that will really explode the market is when I can go to one of my customers and say, ‘Go talk to this fab. You want to build a 200GHz modulator, then just take their pieces and put them together the way you want. Here is the simulator that enables you to model that, and the six different ways that you want it to work and you have that and feel confident that you can fabricate it the same way that you can for high-speed CMOS SerDes at 25GHz, the same way you can buy processor cores.’ It will evolve to be the same infrastructure as that.

Lamant: There is another process in component design IP, which is taking hold within organizations. If we are able to create a platform where you can publish and use IP for photonics, then you can grow the industry and the community that can access the technology.

Attaway: One of the most important things that we are trying to do, or we ought to be trying to do, is to enable the concepts of design reuse or hierarchical design. Those concepts have to be applied to photonics. Most photonic design groups don’t grasp the importance of that yet, but they are starting to. EDA concepts, which have been beaten into submission over the past 15 years along that line, are extremely important to enabling growth and higher levels of abstraction—the ability to buy blocks of IP and put them together and know they will work. All of those pieces are critical to enabling system design with photonics in it.

SE: What does an EDA flow for photonics look like?

Lamant: We constructed it to look like a traditional EDA flow. You have schematic capture at the front end with electrical and photonic components and you have an environment for driving simulation. On the back end, instead of having a SPICE simulator, you can plug in an optical simulator. Then you have the layout-driven side, where you use the same schematic and drive your layout, just as you would for electrical. Behind that layout, or when you are creating the waveguide, you have tools that are routing things and passing back parameters. That is an interesting difference between the electrical and photonics world in that we don’t have Rs and Cs. Instead we have waveguides, and they have a fairly complex model that is not as easy as R and C to back-annotate. But basically the flow, when put in front of the designer, has the same concepts, the same flow as you would do for electronic design. We are reaching that goal mostly. This minimizes the learning curve and you don’t have to learn new tools. You have to learn what it means to make an optical network, analysis and interconnect. You do have to learn about the characteristics of waveguides, such that if you make a 90 degree angle in your waveguide, unlike in electronics, light will not be very happy. But the tool is exactly the same.

Korthorst: Our core competence is layout and verification of photonics circuits, and we combine that with some simulators. Today, for layout implementation of circuits to a certain size, our tool is the most widely used. We do get requests to collaborate with EDA vendors. Our environment is script-driven, so you describe things in a high-level language—your components and the connections—and then we synthesize and compile the layout. But that is not interactive. To combine that with an interactive flow and making it schematic-driven, you can really bring the ease of design into the hands of a larger audience.

Attaway: Enabling – this aligns with my earlier comments – is being able to let the world’s electronic designers add photonics into their repertoire. The reason why that is a viable path to grow the ecosystem is that they are doing it. Cadence is doing this today and their competitors are likely to follow.

SE: You provided a clear message that the biggest limiter today is cost, and that is limited by packaging. What about test? Is that a big issue or just a thorny issue?

Korthorst: It is a big issue. Last week there was an international roadmap workgroup for test and packaging, and their first bullet was the need to educate the design community. There is technology in place to impose constraints on your die, or even be aware of your packaging environment, and design for packaging, or design for test. But there are still designers doing real products, not only in academia but also in real companies, who do not design for test.

De Vries: That is not just for photonics.

Korthorst: Education, training, providing the tools to help them lower the cost of being able to use industry available test equipment. If there is high volume, you can make the investment to design a customized package. But if you are ramping, you need to get over that hurdle and use the tools and technology to design for test and packaging.

De Vries: For a variety of different reasons, from capital costs to time-to-market, it is not acceptable to go through what academic photonics community typically accepts – build it, measure it, tweak it, build it, tweak it until you have something good. In a commercial application that makes no sense. You cannot do that. You cannot do that in high volume CMOS IC design. You can’t do it in high-performance RF. So test and reliability are important to make sure you get it right the first time. This is the only way to make the industry successful. That is not just making sure I can build the silicon. It is what I am going to assemble and ship, and testing is a critical part of that. If it requires hand testing or the yield rate diminishes after I put it in a package, these things matter fundamentally to make it commercially viable. Development itself is one piece of the puzzle. It is not a segment of the industry that we spend a lot of time in. We are more on the component side, but I see this as a piece that will enable the commercial side to be successful.

Attaway: That is another area we are investing in. Trying to tackle assembly and test issues, which are high cost items – I am not sure we have all of the right paths figured out that will get us to good solutions. But we are attempting to do that.

Related Stories
Integrated Photonics (Part 1)
Are we there yet? Where is the demand coming from and which types of product will drive innovation?
Integrated Photonics (Part 2)
What can be done to reduce costs and improve packaging options, and what makes sense in terms of fabrication technologies.
Photonics Moves Closer To Chip
Government, private funding ramps up as semiconductor industry looks for faster low-power solutions.
Focus Shifting To Photonics
Using light to move data will save power and improve performance; laser built into process technology overcomes huge hurdle.

Leave a Reply

(Note: This name will be displayed publicly)