In this issue I interview Dr. Alex Fang, founder and CEO of Aurrion in Santa Barbara, CA. I do so because integrating photonic and electronic circuits gets photonics onto the same playing field as electronics in the digital age. It is not unfair to compare this development to VLSI in the 1980s that brought us the processing capability we enjoy today. I feel that this is a "big deal," enough so that readers should know more about it even though I am an investor and on the board of Aurrion.
Milton Chang: Briefly, why the big excitement about silicon photonics?
Dr. Alex Fang: Until now, lasers and other photonic components have been expensive because manufacturing processes are captive and are not scalable. Silicon photonics makes production of photonic devices compatible with silicon foundry processes. That is, photonics devices can now be manufactured and packaged on the same production lines with silicon integrated circuits. This drastically impacts photonic applications, because we can have more functionality, lower costs, and lower power consumption. At the end of the day, photonics still needs to do what photonics does best. This means we need integrated lasers so we can take advantage of WDM (wavelength-division multiplexing).
MC:Why do some people think silicon photonics is hype and others think it is a major milestone?
AF: Most silicon photonic solutions have neglected to integrate the laser in a way that is cost efficient, power efficient, and scalable for WDM architectures. This has resulted in poor market penetration. At Aurrion, we integrate lasers by integrating III-V materials with silicon and processing it within the silicon foundry process. Our WDM PICs (photonics integrated circuits) are uncooled and are co-packaged with electronics using the standard electronic OSAT supply chain. This makes all the "hype" real.
MC:What makes III-V and CMOS processes compatible?
AF: The process flow used is important. CMOS processes are typically highly controlled in the front-of-line process, but once you get into the back-end, materials acceptance opens up. The real key is to have foundry partners who understand this and the importance of what we're doing.
MC: How do you achieve lower power consumption?
AF: We can make systems-in-package that have optical ports by bringing photonics into the electronic packaging ecosystem. This eliminates numerous interfaces between the photonic chip and the electronic chip that drives power reduction. Optical modules are placed on the edge of boards today, therefore getting the bits from the ASIC out to the photonic chip involves a lot of handoff through signal re-conditioners and other interfaces. By placing the optics close to the drivers, we save power by throwing away the need to make everything with 50-ohm termination. This is only possible by co-packaging optics that can operate in the electronic environment uncooled.
For a standard 100G LR4 transceiver, power consumption can be reduced by a factor better than 4X from the current level, and cost could be reduced by at least of factor of 10 initially, and drop further as we move up the learning curve. Prices limit the subscription ratio between downlink and uplink ports in data centers, leading to bottlenecks. The reduction in cost would enable optimized network architecture.
MC: How so?
AF: The rise of dynamic content through hosted applications and user-targeted content has led to an increase in machine-to-machine traffic in data centers. The system architecture is going through a major overhaul to address the need for greater connectivity between servers and switches, new methodologies for network management via software-defined networking, and associated protocols that direct efficient traffic flow. The connectivity paradigm is driving dense connections that have greater reach and bandwidth than what is achievable now.
MC:When will silicon photonics devices be on the market?
AF: Silicon photonics variable optical attenuators and active optical cables are available now. These niche products have decent volumes, but do not follow any optical standards that enable higher value network architectures. Some customers are sampling alphas of our products to do internal demonstrations and are excited by what our products can do for their business. We recently published link results at OFC that exceed both power and performance standards. There are no physics showstoppers; we have the core technology in hand, and now it is a matter of execution. We should be sampling more parts early next year and be ready to ramp delivery.
MC: How was the company started?
AF: I joined Professor John Bowers' group at the University of California, Santa Barbara, because I found physics, and photonics in particular, a meaningful, creative, and intellectually stimulating field. John puts students in front of customers (contract program managers) to report out on the status of research so they know the importance of their problems. He and I decided to start this company when I graduated because we saw potential for the technology. John is also savvy in business, having worked with VCs to start his previous companies. He wanted to build a company that had solid technology.
I was able to learn as we went along, given the business started slowly. I have guidance from John and our directors to help me along. I recently enrolled in the Owner/President Management program at Harvard Business School.
MC:What are the alternatives for growing the business?
AF: We tried to be capital efficient from the start, by getting SBIR and research contracts from government agencies like DARPA, AFRL, DOE, and ONR, and also from for-profit companies. When the time was right, we brought in angel investors to broaden our set of mentors and later strategic investors who would be our customers and business partners.
