For Optical Fiber, More Bandwidth Looms

HANK HOGAN, CONTRIBUTING EDITOR, hank.hogan@photonics.com

The world’s data travels largely by fiber, with more being moved than ever before. San Jose, Calif.-based networking giant Cisco Systems forecasts that data traffic will grow 22 percent a year from 2015 to 2020.

Little wonder, then, that more fiber is going in the ground, under the sea and into data centers. This year the optical fiber installed in communication networks globally will total 421 million kilometers, said Richard Mack, principal analyst with market research firm CRU International.

Fiber innovations, along with improvements in optical transceivers and connectors, are needed to keep up with surging demands for both more data and higher bandwidth. Courtesy of Corning Inc.

“This total is 10 percent more than the amount installed in 2015, and it is almost twice the amount installed five years ago, in 2011,” he said.

In addition to more fiber being installed, industry and researchers are also working to boost capacity by engineering the fiber as well as improving sources and detectors. Some are looking at what could be fundamental material changes.

An example of these trends at work can be seen at Corning Inc. Based in Corning, N.Y., the company makes fiber for a wide variety of applications, said Joel Orban, product line operations manager for the single-mode and multimode optical fiber business. Single-mode fiber is often used in telecom and long-haul applications, with transmission wavelengths around 1550 nm and distances of thousands of kilometers. Multimode fiber, on the other hand, is used in shorter run applications such as data centers and elsewhere. Source wavelengths are 850 and 1310 nm.

Global data traffic, which largely travels over fiber, is projected to grow 20+ percent a year due to more users, devices and video. Courtesy of Cisco.

There are other differences in the cabling, driven by the differing application needs. Data centers prefer bend-insensitive cabling, the better to navigate tight confines. Telecom carriers are turning to smaller diameter optical fiber, achieved by shrinking the nonlight bearing protective coating so that the overall diameter drops from 242 to 200 µm. That smaller size enables 144 fibers to go into a standard sized cable that before held 96.

“These designs can effectively increase fiber density by 50 percent,” Orban said.

Mobile data traffic is one reason why fiber bandwidth demands are growing. Courtesy of Corning Inc.

Another change involves the size of the core, the part of the fiber through which light travels. Glass fiber that is 125 µm in diameter may have a core diameter of only nine or 10 µm. Increasing that to 11, 12 or more lowers the power density, decreases nonlinear optical effects, reduces loss per kilometer traversed and increases the maximum bandwidth. However, such fibers can be more sensitive to bend losses.

Larger core fibers can transmit high-speed data roughly twice as far as narrower core, standard fiber, said Robert Lingle Jr., director of market and technology strategy at OFS of Norcross, Ga. OFS and its parent company, Furukawa Electric of Japan, make fiber and cable, as well as signal and pump lasers.

More wavelengths = greater capacity

Another way to raise capacity is to use more wavelengths of light. For example, one trend is to extend the 1535-to 1560-nm range C-band out to 1610 nm by adding L-band amplifiers, Lingle said. Both long-haul and data centers are interested in this for the extra capacity.

“That’s a very active area and new fibers need to be designed for the C- and the L-bands,” Lingle said.

A new class of laser, based on research at the University of California, San Diego, and made by Ram Photonics, uses an optical comb to provide hundreds of stabilized carriers to cancel transmission distortions on complex channel modulation, thereby increasing fiber data-carrying capacity. Courtesy of Stojan Radic, UCSD and RAM Photonics.

He added that a large group of companies are now devising a 10-year multimode fiber road map. These guidelines will be used to develop fibers, cables and transceivers so that links can move from the current 100 Gbps commercially deployed capability to 400 and 800 Gbps. The goal of the group is to do so by employing wavelength division multiplexing over multimode fiber, thereby maintaining the cost advantage traditional for multimode fiber links, according to Lingle.

When increasing bandwidth to 400 Gbps and beyond, doing so without having to resort to more channels eliminates the need for higher density transmit/receive optical subassemblies and fiber cable connectors, said Craig Thompson, vice president of new markets at optical fiber component and module maker Finisar Corp. of Sunnyvale, Calif. The company makes transceivers, lasers and detectors for 850-, 1310- and 1550-nm wavelength bands.

Fewer channels are better because the amount of space for a subassembly or connector is largely fixed. So, if it takes eight instead of the current standard four channels to get to a given data rate, then the subassembly or connector density must be roughly doubled.

Demonstration of a pluggable 400G module, with four times the current commercially deployed connection speeds, interoperating with test equipment. Courtesy of Finisar.

The move to 400 Gbps also presents other challenges. For instance, there is a different and more complex signaling modulation. However, those faster speeds also present new opportunities for fiber optics. The alternative, sending electrons down a conductor, is also running into problems.

“Generally, as speeds go up the possible reach in copper reduces and to get that reach you need more copper, bulkier cables and sometimes even some active components at each end of the cable,” Thompson said. “We’re getting to the point where copper has some real limitations.”

He thinks that the 400-Gbps generation, made up of eight lanes running 50 Gbps each, will see a mix of optical and electronic transport even in the very short runs that connect a server rack together. The optical side might even be favored because it offers lower latency, or transmission lag.

In addition to increasing bandwidth demands, the age of the installed long-haul networks is also playing a role in new deployments. Many long-haul networks are decades old and may be maxed out in capacity, said Ron Johnson, director of product development at Cisco.

Beyond solving capacity constraints, putting in new fiber and associated components offers benefits. By using low-loss, large-core fiber, for instance, the distance a signal can travel before having to be regenerated can be significantly increased, cutting costs. More savings are possible by eliminating electronic components in favor of relatively new high port count reconfigurable optical add-drop multiplexers or ROADMs. Whereas previous devices might have six ports, the latest ones may have as many as 20. That makes it possible to have enough capacity at the optical switch to add and drop signals as well as transmit them on through without having to do so electronically.

Optical fiber spools. Courtesy of Corning Inc.

Still more savings arise from other characteristics of this new generation of optical switches. “You have a much wider (spectral) band and that reduces the transmission effects, which is translated to fewer or no regeneration required,” Johnson said.

Such savings lead to a lower cost per bit. That may create increased data demands, which will result in the need for more bandwidth. So, the new cables being installed typically have additional fiber built in. This provides extra capacity that could be lit when needed.

As for the future, extensive research efforts underway aim to boost the capacity of existing fiber. One such was demonstrated at University College London. In a 2015 Nature paper¹, the group reported on a technique that nearly doubled the length a signal could reach. They employed digital signal processing to undo the distortions introduced in the signal, using known physical parameters of the fiber to do so. One payoff is more throughput.

“You can go to higher orders of modulation to send more information over greater distances with these DSP [digital signal processing] techniques, ensuring that the ratio between your signal power and the noise remains high enough,” said Robert Maher, at the time a senior research associate. He was the paper’s lead author and now works for Infinera Corp, a Sunnyvale, Calif. maker of long-haul wavelength division multiplexing gear.

At the University of California, San Diego, a research group has taken another approach. As outlined in a 2015 Science paper², they passed an ordinary laser beam through a special mixer to generate hundreds of wavelengths. The sources thus produced are exceedingly stable, with relative carrier variability improved by orders of magnitude. That makes it possible to dramatically increase the bandwidth of existing fiber.

“You will more than quadruple capacity by encoding information using such a laser,” said Stojan Radic, professor of electrical and computer engineering and leader of the research group.

He added that a demonstrator of the concept will be built in 2017. Commercial applications will come after that. Radic is also a technical consultant with San Diego-based RAM Photonics, which has produced a laser source motivated by the technique.

Finally, while silica optical fiber dominates today, that may not always be the case. The interaction of high-intensity light with current fiber leads to nonlinear, performance-limiting effects. Incorporating significant amounts of aluminum, barium, yttrium or combinations of these and other materials leads to glasses that exhibit much smaller nonlinear effects. Eventually, as power levels climb in response to the demand for more bandwidth, fiber may be constructed out of something other than today’s silica standard.

John Ballato, a professor of materials science and engineering at Clemson University, noted that the high-power lasers now used in industrial and military applications are already at such a point. Using different, more linear core glasses makes higher operating power possible.

Such a fundamental material change is many decades off, if it happens at all. In the end, though, it may be necessary to make a clean break with the past to achieve a radical result.

As Ballato said, “The lightbulb did not arise from continuous improvements of the candle.”

References

1. R. Maher, et al. (2015) Spectrally shaped DP-16QAM super-shannel transmission with multi-channel digital back-propagation. Nature, Scientific Reports, Vol. 5, Article No. 8214. www.nature.com/articles/srep08214.

2. E. Temprana, et al. (2015) Overcoming Kerr-induced capacity limit in optical fiber transmission. Science, Vol. 348, Issue 6242, pp. 1445-1448. http://science.sciencemag.org/content/348/6242/1445.