Press releases about fiber optics, like baseball stories, report many types of records. The two new fiber-optics records reported last month at Conference on Optical Fiber Communications (OFC 2020) are definitely major league. A team from Japan's National Institute for Communications Technology (NICT) has sent a staggering 172 terabits per second through a single multicore fiber—more than the combined throughput of all the fibers in the world's highest-capacity submarine cable. Not to be outdone, Nokia Bell Labs reported a record single-stream data rate of 1.52 terabits per second, close to four times the 400 gigabits per second achieved by the fastest links now used in data centers.
The IEEE-cosponsored OFC 2020 could have used some of that capacity during the 9–12 March meeting in San Diego. Although many exhibitors, speakers, and would-be attendees dropped plans for the show as the COVID-19 virus spread rapidly around the world, the organizers elected to go ahead with the show. In the end, a large fraction of the talks were streamed from homes and offices to the conference, then streamed to remote viewers around the world.
Telecommunications traffic has increased relentlessly over recent decades, made possible by tremendous increases in the data rates that fiber-optic transmitters can push through the standard single-mode fibers that have been used since the 1990s. But fiber throughputs are now approaching the nonlinear Shannon limit on information transfer, so developers are exploring ways to expand the number of parallel optical paths via spatial-division multiplexing.
Image: National Institute for Communications Technology
Spatial-division multiplexing is an optical counterpart of MIMO, which uses multiple input and output antennas for high-capacity microwave transmission. The leading approaches: packing many light-guiding cores into optical fibers or redesigning fiber cores to transmit light along multiple parallel paths through the core that can be isolated at the end of the fiber.
Yet multiplying the number of cores has limits. They must be separated by at least 40 micrometers to prevent noise-inducing crosstalk between them. As a result, no more than five cores can fit into fibers based on the 125-micrometer diameter standard for long-haul and submarine networks. Adding more cores can allow higher data rates, but that leads to fiber diameters up to 300 µm, which are stiff and would require expensive new designs for cables meant for submarine applications.
In a late paper, Georg Rademacher of NTIC described a new design called close-coupled multi-core fibers. He explains that the key difference in that design is that “The cores are close to each other so that the signals intentionally couple with each other. Ideally, the light that is coupled into one core should spread to all other cores after only a few meters.” The signals resemble those from fibers in which individual cores carry multiple modes, and require MIMO processing to extract the output signal. However, because signals couple between cores over a much shorter distance in the new fibers than in earlier few-mode fibers, the processing required is much simpler.
Image: National Institute for Communications Technology
Earlier demonstrations of close-coupling were limited to narrow wavelength ranges of less than 5 nanometers. In San Diego, Rademacher reported testing an 80-km length of three-core 125-nm fiber with signals from a frequency comb light source. The team transmitted 24.5 gigabaud 16-quadrature amplitude modulated (16-QAM) signals to sample performance on 359 optical channels in the C and L fiber bands spanning a 75-nm bandwidth. Signals were looped repeatedly through the test fiber and an optical amplifier to simulate a total distance of 2040 kilometers.
The total data rate measured over that distance was 172 terabits per second, a record for 125-µm fibers. Fibers with much larger diameter and more cores have transmitted over 700 terabits over 2000 km, he says, but they remain in the laboratory. The world's largest capacity commercial system, the Pacific Light Cable Network, will require six fibers in order to send 144 terabits when and if it comes into full operation. The close-coupled fibers “are far from the required maturity for subsea use,” says Rademacher. His group also is studying other ways that multicore fibers might improve on today's single-mode fibers.
The Nokia record addresses the huge market for high-speed connections in massive Internet data centers. At last year's OFC, the industry showed commercial 400-gigabit links for data centers, and announced a push for 800 gigabit rates. This year, Nokia reported a laboratory demonstration that came close to doubling the 800-Gig rate.
Successfully achieving single-channel data rates exceeding 100 gigabits depends on transmitting signals coherently rather than via the simple off-on keying used for rates up to 10 gigabits. Coherent systems convert input digital signals to analog format at the transmitter, then the receiver converts the analog signals back to digital form for distribution. This makes the analog-digital converter a crucial choke point for achieving the highest possible data rates.
Speaking via the Internet from Stuttgart, Fred Buchali of Nokia Bell Labs explained how his group had used a new silicon-germanium chip to achieve a record single-carrier transmission rate of 1.52 terabits per second through 80 km of standard single-mode fiber. Their digital-to-analog converter generated 128 gigabaud at information rates of 6.2 bits per symbol for each of two polarizations. It broke a previous record of 1.3 terabits per second that Nokia reported in September 2019.
Micram Microelectronic GmbH of Bochum, Germany, designed and made the prototype for Nokia using high-speed bipolar transistors and 55-nanometer CMOS technology. Buchali said that looping the fiber and adding erbium amplifiers allowed them to reach 240 km at a data rate of 1.46 Tbit/s. The apparent goal is to reach 1.6 Tbit/s, four times the current-best 400 gigabits per second, at the typical data center distance of 80 km.
“If we are able to meet that target with a single carrier as is demonstrated here rather than a number of carriers—say, 4 at 400Gbps—then it is quite likely that the solution would be both more efficient in spectrum usage and lower cost,” says Theodore Sizer, Smart Optical Fabric & Devices Research Lab leader at Nokia Bell labs. That could be an important step in letting the fast-growing data center world handle the world’s insatiable demand for data.
Jeff Hecht writes about lasers, optics, fiber optics, electronics, and communications. Trained in engineering and a life senior member of IEEE, he enjoys figuring out how laser, optical, and electronic systems work and explaining their applications and challenges. At the moment, he’s exploring the challenges of integrating lidars, cameras, and other sensing systems with artificial intelligence in self-driving cars. He has chronicled the histories of laser weapons and fiber-optic communications and written tutorial books on lasers and fiber optics.