Intel Silicon Photonics Wafer
You know a technology is going to be big when major players start getting involved, and there aren’t many players more major than Intel. So, when the company says its photonics technology is a game changer, it’s probably worth listening. To find out more, Optical Connections sat down with Robert Blum , Intel’s Senior Director of Marketing and New Business, and Darron Young , Intel’s Senior Director of Product Management, both in the company’s Silicon Photonics Product Division, for a deep dive into the company’s photonics strategy going forward. INTEL PHOTONICS: INTEGRATION AT THE WAFER LEVEL
How did Intel first get into photonics?
photonics has significant benefits for integrating passive components, but for us the ability to make active devices on chip is equally, if not even more, relevant.
RB The initial research dates back to about 20 years ago, but our first qualified product was released in 2016. It’s been quite an incredible journey since launching this first 100G PSM4 transceiver for data centre applications, ramping photonics to high volume production and getting all the benefits of a mature silicon photonics process. We’re now shipping two million transceivers a year, with excellent reliability, and we can say that not just based on our internal qualification data but from years of deployment in the field. Manufacturing optics as wafer scale and the resulting products have really disrupted the industry.
What sectors of optical communications is Intel targeting with its photonic products?
in a wafer level process,
DY As Robert said, the first transceiver was a data centre product, and this continues to be our focus. All of our data centre revenue today is from pluggable transceivers, and we have shipped more than 5 million 100G transceivers and generated over $1 billion in revenue to date. We started with 100G, 4 to 5 years ago, but we also see this exponential traffic growth in the data centre, with data doubling every two to three years. So it’s really a cadence going from 100G to 200G to 400G, and both 200G FR4 and 400G DR4 are ramping this year. We have also started sampling 800G DR8 transceivers. One of the reasons why silicon photonics is unique and why we believe it’s an enabling technology, and we’ve been saying this from day one, is that pluggable transceivers with discrete optics will not be able to keep increasing in data rate for much longer, and we’re running into these limits now. 800G is really, I think, the last generation where using standard pluggable transceivers make sense - the next step is
including writing the gratings in the silicon. This allows us to make highly complex chips that integrate multiple lasers with different wavelengths together with many other passive and active components. The ability to integrate optical gain on the photonic chip is a significant benefit. Our DFB (Distributed Feedback) lasers, for example, are fundamentally different in design from discrete lasers, with inherent reliability advantages, including not having any exposed facets. We are also able to burn them in and fully test them at the wafer level. All this allows us to make highly complex structures. For some of the LiDAR applications, we can put multiple SOAs (semiconductor optical amplifiers) on the chip as well, to really boost the optical output power. That really changes the game as opposed to, for instance, using an external fibre and an EDFA (Erbium- Doped Fibre Amplifier) outside of the chip. I think many people are aware that silicon
What makes Intel’s photonics so different?
The game changer is being able to integrate and make optics at the wafer scale with all the associated
benefits such as tighter process control, better performance, cost, and scalability. And on our platform we manufacture both passive and active devices at the wafer level – we basically put the Indium Phosphide epitaxial material down on the wafer, pattern it, and then make the lasers
Made with FlippingBook Ebook Creator