The 190-Terabit Optical Fiber Jigsaw Puzzle

Multiplexing optical signals has become one of the keys to increasing throughput across fiber-optic connections. Now, splicing multiple “core” fiber optic cables together may be the answer toward pushing optical bandwidth to a mind-blowing 190 terabits per second.

The problem is actually aligning the end of an optical cable with the beginning of another, so that signals remain in sync. With a single cable, that’s not a problem. When you wrap so-called multicore fiber together (including nineteen 10-terabit core fibers) the problem’s complexity increases tremendously.

The bandwidth involved is even a hair faster than the maximum potential of the 400 Gbits/s network that connects Paris and Lyon, which is expected to reach 17.6 terabits per second if 44 available optical wavelengths are successfully multiplexed.

That’s the problem Wenxin Zheng, manager of splice engineering at cable manufacturer AFL, suggests he’s solved. Zheng is scheduled to speak at the Optical Fiber Communication Conference and Exposition/National Fiber Optic Engineers Conference (OFC/NFOEC) next week in Anaheim, where academia meets industry in trying to solve problems related to high-speed data transmission. The difference is that Zheng and AFL are trying to develop production techniques, rather than solve the problem in a lab. (Another multicore fiber from OFS Labs has successfully transmitted seven upstream 1310 nm and seven downstream 1490 nm signals at 2.5 Gbit/s, each over distances of 11.3 km.)

AT&T is also expected to release a paper on how, through improved signal modulation, next-generation bandwidths of 400-Gbit/s can be achieved on existing gigahertz networks. Both the AFL and AT&T papers are significant because they take existing technologies and, with modifications, extend them into the future.

Zheng’s 190-terabit threshold work assumes that both time multiplexing and wavelength multiplexing (of 200 different wavelengths) can be enabled on each fiber core. But the real challenge is then splicing them together, again and again and again. “To align the multiple cores simultaneously is a big challenge,” Zheng said, according to the conference, which released some of the work being presented before it begins. “If two fibers to be spliced have random core locations, there is no way to align the entire core.”

Under normal circumstances a single core can be laser-spliced. Zheng’s multicore technique involves a Fujikura FSM-100P+ fusion splicer, where the fibers to be spliced are stripped and loaded into the splicer, then rotated and imaged with two video cameras so that their cores can be roughly aligned using a pattern-matching algorithm. Next, using a power-feedback method and image processing, a pair of corresponding cores in each fiber are finely aligned, as is the cladding around the cores. Finally, the cores are heat-spliced.

If (and only if) the component cores are made identically, and then distributed symmetrically within the outer cladding, then aligning one core will align the others automatically, according to Zheng.

AT&T’s 400-Gbit/s Optics

AT&T is likewise expected to show that it has been able to tune the modulation spectral efficiency, the measure of the data rate that can be sent over a given bandwidth. That has implications, AT&T noted, because next-generation Ethernet technologies ditched terabit speeds due to cost concerns.

Xiang Zhou of AT&T Labs-Research in Middletown, N.J. used Nyquist-shaped 400Gb/s signals with tunable spectral efficiency that were generated using modulated subcarriers. Eight 100 GHz-spaced, 400 Gb/s wavelength-division-multiplexed signals were combined and then transmitted over a re-circulating transmission test platform consisting of 100-km fiber spans, AT&T said.

Using the new modulation technique and a new low-loss, large-effective area fiber from OFS Labs, the team transmitted the signals over a record-breaking 12,000 kilometers (7,500 miles)—surpassing their own previous distance record (using the 50 gigahertz-grid) by more than 9,000 km.

Other papers include one from IBM, which said has it has cut the power needed to operate an optical transmission link by half (to 24 mW) while increasing throughput by 66 percent, to 25 Gbits/s. All told, that’s one picojoule per bit.

It’s hard to believe that just a few years ago, networks with hundreds of gigabits of bandwidth per second was considered jaw-dropping. There may be no guarantee that the work being done by either AFL or AT&T will pan out, but it is exciting.


Image: FooTToo/

Post a Comment

Your email address will not be published.