元器件 > 高级半导体博客



Is there a Killer App for Silicon Photonic ICs?

by Eric Higham | 3月 31, 2015

Well, the weather is starting (just barely) to get warmer around Boston, so that means time for travel and conferences. Travel is like riding a horse; you climb into the saddle and off you go. But just like the horse, you have to pay attention to the small details before you get where you want to go. You have to remember where you parked your car and where you put your parking ticket. In Amsterdam, you have to remember to remove every cord, piece of electronics and plug from your briefcase and in JFK airport, apparently the bus between terminals has to remember to yield to oncoming aircraft, which makes perfect sense!

Having navigated all those pitfalls, I did have the pleasure of attending two conferences this month. I presented the opening paper at CS International in Frankfurt and I participated in a panel session at OFC 2015 (Optical Fiber Communications) in Los Angeles. Both conferences were packed with all the latest developments from academia and industry researchers, coupled with exhibitors showing their latest technology and products. Since both conferences covered fiber communications, I’ll focus on that topic 

If you have heard me speak, I usually talk about the growth in data consumption 2009 and how there is no end in sight. Most of the time, I’m talking in terms of mobile data, because the growth of wireless data has driven the compound semiconductor industry to more than $7 billion of revenue. However, you’ve probably also heard me say that mobile data represents only the tip of the iceberg. The “iceberg” represents applications that use fiber networks. These fiber networks make up some or all of the CATV/broadband networks, intranet connections in enterprises, transport networks that connect cities and countries to one another and increasingly, the connections in the data centers that support the ubiquitous “cloud” that we all use.

Data center traffic is still less than 1% of total data traffic, but it is growing quickly and driving one of the trends I went to OFC to understand. The following chart, from a Finisar presentation shows the changing nature of data center communications. According to Cisco, about 75% of the data consumption in a data center occurs within the data center, as opposed to between data centers or from data center to user. This trend, coupled with growing data consumption results in the need for very short range, high-speed data links. This chart speculates that by 2018, 50% of all links in a data center will be less than 3 meters and 90% will be less than 100 meters!

Is there a Killer App for Silicon Photonic ICs?

Source: Finisar

 

This is the driving force behind silicon photonics. This technology looks to use the integration properties and existing installed base of high volume silicon CMOS processing to produce low cost, high-speed photonicsICs. One presentation mentioned that there were 50 papers dealing with silicon photonics at this year’s conference and some 1250 in total since the topic started gaining traction.

There is no question that there is a tremendous amount of development activity around silicon photonics, but it seems to be another technology with undeniable advantages and lots of work that still needs to be done. The scope of work ranges from fundamental process development to passive and active structure development to how to implement the processes in a production CMOS fab. The progress and need is undeniable, but like all new technologies, it will take time. I happened to be sitting in a session, in between two industry “insiders” when it was announced that Intel would not be presenting a paper on how they successfully integrated the process into their CMOS fab. The reason given was they could not get the proper “approval”, but the insiders were whispering that this was more a reflection of the challenges that Intel was facing with this task, rather than any approval issue.

What’s the reality? I’m not sure, but with the amount of effort aimed at this technology and the shift toward shorter and faster interconnections, silicon photonics seems more like a “when” rather than an “if”. Stay tuned as I keep tabs on what is sure to be an interesting technology story!

 

-Eric

Previous Post: The Butterfly Effect; A Falling Snowflake Leads to a Massive Semiconductor Opportunity | Next Post: Backhaul Trends Point to Less GaAs Revenue
Leave a comment