September 10, 2009

Delivering Data At Light Speed

You may not have heard of nanophotonics, but it may be the technology that puts Intel and Broadcom chips to shame.

Greg Young serves as president and CEO of Luxtera [full disclosure: My venture firm, Lux Capital, is an equity investor]. Prior to Luxtera, he was vice president and general manager of the High Speed Ethernet Controller and High Definition Media PC Video business units at Broadcom. While there, Greg led the growth of the Ethernet Controller business unit from concept to hundreds of millions in revenue and the No. 1 market share position. Prior to joining Broadcom, Greg was with Intel, where he held several engineering marketing and leadership positions.

Josh Wolfe: What career path led you to Luxtera?

Greg Young: After trying some startups out of school I joined Intel ( INTC - news -people ) in the mid-90s, beginning as an engineer and then transitioning over to marketing and running product lines. I worked at Intel until 1999, when I joined Broadcom ( BRCM - news - people ). I spent eight years at Broadcom helping to pioneer the company's participation in the Ethernet market for the network interface controller business. Ultimately, I helped grow that business to about $350 million dollars a year in semiconductor revenue. Most of my career has been spent building businesses off of advanced transceiver technology (devices that both transmit and receive information), so when I recognized the opportunity within Luxtera, it was easy for me to see how the technology could be built into a large-scale enterprise.

What excited you about the company?

First, some market backdrop here: It's getting harder and harder to send fast signals over copper wires. The world of optics has been sitting out there for a long time as the performance leader, but it has been a very expensive way to get the performance that you need for the same kind of input/output speeds. When I recognized that Luxtera had the ability to create a complete optical transceiver in CMOS technology to take performance to 10 gigabits and well beyond 10 gigabits at a cost point that was previously unachievable, I saw the same kind of opportunity I was given at both Intel and Broadcom.

Put it in perspective--how fast is 10 gigabits?

If you use a cable modem at home, that's about a 1 megabit connection--a million bits per second. We're talking about ultimately transitioning people to the point where they can readily transmit 10 billion bits a second. That's the equivalent of downloading more than 300 songs every second.

Why do photons trump electrons when it comes to broadcasting bits?

When you send an electronic signal over copper wires, there is a relationship between speed, distance, and signal integrity. As you get faster and faster over the same distance of wire, your signal integrity gets worse, and you see distortion in the signal that starts to dominate the signal quality at higher speeds. Because of that relationship, there is a natural limit for how fast and far you can push a signal over a copper wire.

At 10 gigabit speeds, electrical interconnects over copper wires really start to break down--it's hard to transmit the signal even 10 meters. Alternatively, you can send a burst of photonic energy down a low-cost fiber optic waveguide, and you can easily send a 10 gigabit signal over 10 kilometers. You can do it with less power, less complexity, and with Luxtera's technology--lower cost.

Why is transceiver technology important in this industry?

While at Intel and Broadcom, I saw two things: first, mixed signal circuitry (combined analog and digital circuitry) would enhance the communications signals between systems, and second, I realized that the rate at which you come out with new transceiver technology is really what controlled the cadence of the innovation in the industry. I first saw this at Intel.

The company was able to utilize its own technology to build transceivers for 100 megabit Ethernet. At the time, 3Com ( COMS - news - people) was the dominant player, but by leveraging the cost and performance benefits of having an integrated transceiver technology in CMOS, we were able to transition the market from 10 megabit to 100 megabit Ethernet and move Intel's position from a minority player to the market leader within the network interface controller business. That was a really interesting learning experience for me.

When I joined Broadcom in 1999, the company was the leader in mixed signal in CMOS and was just entering the Ethernet space, building up their business as an Ethernet transceiver vendor. What I was handed when I came into the company was a complete, single-chip gigabit Ethernet transceiver. At the time, no other company in the world knew how to build a single-chip transceiver for 1 gigabit data rates, and by having that technology I was able to facilitate a very similar transition to what I had been involved with at Intel--driving the market from 100 megabit Ethernet to 1 gigabit Ethernet.

Today, you can barely buy a computer that doesn't have a gigabit Ethernet network controller in it, and it was that transceiver advantage that Broadcom had that allowed them to subsequently grab the No. 1 market position from Intel.

CMOS, photonics, optical transceivers--sounds complex! In the simplest of terms, what is it that Luxtera's technology does?

Our technology takes a high-speed signal and gets it from point A to point B. A transceiver sends out a signal at point A and receives the same signal at point B. We send that signal over a fiber optic cable, giving us performance and signal quality advantages. Our system is less expensive than other optical approaches because of nanophotonics--we've shrunk the optical elements down to the same scale as the transistors that sit inside your PC's CPU.

By being down at that scale, we've enabled the manufacturing of our systems with the same processes that makes computer chips, meaning we can precisely stamp them out in large quantities, without needing complex assembly. We've been able to move the world of photonic interconnects from an era equivalent to that of the vacuum tubes, to one of the modern integrated circuit.

Who's competing with Luxtera in this market?

If you look at the area of silicon CMOS photonics, Intel, IBM ( IBM -news - people ), Hewlett-Packard ( HPQ - news - people ) and many other big names within the industry are all doing research. But Luxtera is the leader in development in this space. The original foundation for the company came out of advanced research at Caltech, which stimulated the very early years of development.

We have pioneered a brand new space, moving nanophotonic structures into a CMOS-compatible silicon process. By doing that, we've figured out how to increase performance while reducing cost. We've blazed a new trail, and in doing so we've established the methods and techniques needed to bring this technology into production. Based upon research papers written by other companies exploring this area, we estimate that we're at least five years ahead of the nearest competitor.

What do you see as the current market opportunity for this technology?

There is a huge short-term opportunity for Luxtera within the high-performance computing segment. High-performance computing refers to supercomputers and computer clusters like data centers that are trying to achieve maximum performance to solve complex computations or process large amounts of data. They are all on the cutting-edge of technology, and typically that technology very quickly waterfalls down into the mainstream PC market.

High-performance computing centers are typically the starting point for many innovations in the industry. In each of these centers, there are many, many processors that are trying to communicate with one another at mind-boggling speed, and it's becoming nearly impossible to make that communication work with copper wires.

While there has always been a broad opportunity for photonics, the photonic approaches thus far have always been too expensive to implement. Our technology allows us to take the performance of optics and reduce the cost so that we're able to interconnect these high-performance computing centers economically.

When will we see this type of technology in our home computers?

Over time, optics will transition into every market as speeds get faster and faster. The move from copper to fiber optics is a very natural transition forecasted by just about every industry pundit. You can find this technology today within the high-performance computing space, where we have products that send signals over fiber optics used to connect high-performance computing data centers.

Some of the world's fastest computer systems use photonic interconnects, and over time you're going see that transition down into consumer electronics: Home PCs, DVD players and TVs will all ultimately pick up optics for communications between subcomponents. What's notable is that optics has already moved into the home. The transition from magnetic media--like VHS and cassette tapes--to digital optical data storage on CDs and DVDs is a great precedent where storage requirements exceeded the limits of magnetic, copper-type systems and transitioned over to optics. Communication interconnects are moving down that same path.

Is Luxtera still focused on research or is the company shipping products today?

We are in production with products today. While we continue to do research to move the edge of technology forward (with 23 PhDs on staff), we are a product company with development engineering and manufacturing operations. In fact, we recently announced that through a partnership with Freescale Semiconductor ( FSL - news - people), we've reached full-scale production status for CMOS photonics technology.

What does this collaboration with Freescale mean for the company?

It means we can now design and produce chips that use our structures on a very large scale. Freescale already has a process that they use to build transistors at very large scale, and they produce lots of chips for things like network processors and automotive sensors. We've been able to integrate our novel nanophotonic device structures into Freescale's process, so now their factories can produce CMOS photonic transceivers.

As anyone in the semiconductor industry knows, it takes about five years to develop a new CMOS process, and once you have that process in production, you build products in it for a number of years. By taking our process to maturity through our relationship with Freescale, we can now design a whole host of products and bring them very quickly from design into volume manufacturing.

How do you think big players like Intel and Broadcom perceive your company in the market today?

I think that Intel in particular, and others that work in silicon photonics, see silicon CMOS photonics as being part of their future roadmap. Having a company like Luxtera out there that's in production with CMOS photonics, on the cutting edge of technology, I think one, it comforts them that the roadmap in front of them is truly viable, and two, if I were in their shoes, I would be a little threatened by it. Our technology can be applied to anyone in the industry. Any company that wants to be able to adopt

CMOS photonics to gain performance benefits in a very large market can leverage our technology platform and get to market very quickly. On the other hand, I think a lot of companies view us as an opportunity to get their hands on a technology that could move them ahead on their own roadmap faster.

The ease by which we transport massive waves of data may leave many unaware of the physical systems that enable our virtual world. How do you give people a sense of appreciation for the importance of this technology?

Here's an analogy that may give people some sense of scale: Many people have gone through the transition from a 56k modem to a cable modem or DSL service. What photonics represents to high-performance computing is akin to the transition from dial-up to broadband.

Related Articles:



Josh Wolfe, Forbes, CMOS, photonics, Luxtera, Freescale, Greg Young, Lux Capital, Broadcom, electrons, optical transceivers, Freescale Semiconductor, nanophotonic structures, CMOS-compatible silicon process,

Posted via email from Global Business News

No comments: