Demonstrations of high-speed fiber-optic transmission reached a major milestone at the March 10 postdeadline session of this year’s Optical Fiber Communications Conference (OFC; Mar. 6–10, 2011, Los Angeles, CA). Two groups reported sending more than 100 Tbit/s through single optical fibers using different techniques. That caps three years of rapid increases in fiber-optic speed records following a hiatus in hero experiments during the post-telecom-bubble years.
Those transmission rates are far beyond the installed capacities of current backbone or submarine cables, says Tim Strong of Telegeography (Washington, DC). But with traffic growing at about 50% each year, carriers need to upgrade their networks. Submarine cable operators have begun moving to 40 Gbit optical channels, and a little over a year ago Verizon switched on the first operational 100 Gbit optical channel between Paris and Frankfurt. At the end of March, Verizon announced it would install the first three 100 Gbit optical channels in US long-haul systems.
How to further upgrade that capacity has been an issue. At the OFC postdeadline sessions, a team from Verizon (Richardson, TX) and NEC Labs (Princeton, NJ) described mixing 100 Gbit/s optical channels with “superchannels” transmitting 450 Gbit/s and 1.15 Tbit/s across wider spectral regions through 3560 km of fiber in the Verizon network.1 The superchannels combined orthogonal frequency-division multiplexing (OFDM) with dual-polarization quadrature phase-shift key (DP-QPSK) modulation of multiple subcarriers within the transmission band. Tiejun Xia of Verizon and colleagues reported that the superchannels “provide higher channel capacity and higher spectral efficiency than the existing 100 Gbit/s system” in installed fiber. That shows the feasibility of upgrading existing networks to terabit transmission, says Ting Wang, manager of NEC Labs’ optical networking department.
NEC Labs stretched the superchannel approach to demonstrate 101.7 Tbit/s transmission over three 55 km lengths of standard single-mode fiber.2 They started with 370 laser transmitters spaced at 25 GHz intervals in the C-band at 1527.4 to 1565.7 nm and the L-band at 1570 to 1607 nm. They modulated odd and even channels independently, and partitioned each 25 GHz band into four subbands, each carrying a 6-GHz-wide OFDM signal, with 1 GHz left as a guard band. They modulated and combined those signals, with polarization-division-multiplexed 128QAM applied to each subcarrier. This built up a 294 Gbit/s signal for each of the 370 transmitting lasers, a raw data rate of 108.8 Tbit/s, which corresponds to 101.7 Tbit/s after error correction is done. Dispersion was compensated electronically.
Dayou Qian of NEC Labs reported the system met forward-error-control limits when transmitting over the whole system, with Raman amplification in each 55 km length of fiber. “We want to explore high capacity to see where the limitations are,” says Wang. Key tradeoffs are capacity, transmission distance, and spectral efficiency. In addition to setting a record for total data rate through a single core, NEC Labs set a record for spectral efficiency in wavelength-division multiplexing (WDM), 11 bits per second per hertz.
Transmission distance fell far short of the 3560 km achieved at a lower speed with Verizon. However, Wang says that very high speeds will be needed over short distances in data centers operated by companies such as Facebook and Google long before they are required for long-haul transmission.
Immediately after Qian’s postdeadline talk, Jun Sakaguchi of Japan’s National Institute of Information and Communications Technology (NIICT; Tokyo, Japan) described a different approach to 100 Tbit transmission: spatial-division multiplexing by splitting the signals among seven separate cores in the same fiber (see figure).3 Multicore fibers have become a hot topic in the past couple of years as developers look for new ways to increase fiber capacity. Sakaguchi and colleagues from Sumitomo Electric’s R&D Lab (Yokohama, Japan) and Optoquest (Saitama, Japan) described WDM of 10 Gbit/s signals in seven-core fibers during the regular sessions.4
In the postdeadline sessions, Sakaguchi reported a big step forward: transmitting 97 WDM channels on a 100 GHz grid, each carrying DP-QPSK signals at 172 Gbit/s, through all seven cores. The raw data rate adds up to 15.6 Tbit per core or 109 Tbit/s through the entire fiber, excluding the overhead for forward-error correction. The experiments divided output from a single WDM transmitter among all seven cores, with different time delays added between the transmitter and the optics that focused the light into each of the singlemode cores, which were spaced on 45 µm centers in a 150 µm cladding. The signals then went through 16.8 km of the multicore fiber without amplification before being demultiplexed.
Crosstalk among cores is a potential issue, but Sakaguchi says his group reduced it by increasing core spacing and by bending the fibers to reduce phase matching between the cores. The multicore fiber had chromatic dispersion of about 22.2 ps/nm/km—larger than the 16 to 17 ps/nm/km of standard singlemode fiber—which helps suppress nonlinear effects. They used dispersion-compensating fiber to control the chromatic dispersion, but Sakaguchi says electronic compensation should be possible. He is optimistic that combining the NIICT-Sumitomo technology with NEC’s could push fiber transmission closer to petabit rates.
However, Wang has his doubts about multicore fibers, citing issues that include multiplexing seven data streams into the closely spaced cores, crosstalk among cores, and making a seven-core optical amplifier. He thinks a better way to add capacity would be using few-mode fibers, citing demonstration of spatial-mode separation in a three-mode fiber, also reported at the OFC postdeadline session by Roland Ryf of Bell Labs (Holmdel, NJ).5
OFC also recorded a terabit milestone on a different scale: the fastest-ever passive optical network (PON), transmitting 1.2 Tbit/s through 90 km of fiber. Neda Cvijetic of NEC Labs says the goal is building a PON that could provide a thousand homes with the gigabit service that Google has proposed. At OFC, she reported transmitting 48 Gbit/s on 25 wavelengths through standard singlemode fiber with passive branching to 32 delivery fibers.6 She says the design could be a next-generation standard to follow the recently approved 10 Gbit PON standard.
1. T. Xia et al., OFC 2011, postdeadline paper PDPA3.
2. D. Qian et al., OFC 2011, postdeadline paper PDPB5.
3. J. Sakaguchi et al., OFC 2011, postdeadline paper PDPB6.
4. J. Sakaguchi et al., OFC 2011, regular session, paper OWJ2.
5. R. Ryf et al., OFC 2011, postdeadline paper PDPB10.
6. N. Cvijetic et al., OFC 2011, postdeadline paper PDPD7.