NANOG 85 Tutorial: Optical Fiber Capacity Limits – Where Do We Go Next? (Part 3)

With such a vast installed base of optical fiber, it’s obvious that optical network operators will explore every possible avenue of capacity increase before resorting to the deployment of new fiber. Some operators are lucky enough to have dark fiber available, while others may have spare duct capacity through which new fiber can be blown. Where these options are not available, the average cost of laying fiber is highly dependent on location. As an example, in 2013, the U.K. regulator Ofcom estimated the costs per meter of duct in the U.K. at £100 on roads, £60 on footpaths, and £40 on grass verges, with an additional £2.75 per meter for the 48-fiber cable itself. This assumes, of course, that the operator needing the fiber has the appropriate licenses and rights of way to deploy this fiber.

Are Existing Ducts Available?

For operators in this position, the problem of long-haul terrestrial capacity scaling could be solved in the near term by simply deploying more fiber. Most operators are using fiber ducts with 8-14 mm inside diameter, and newer dense cable designs can fit 192 to 432 fibers in those ducts using 200-micron fiber technology. In long-haul networks, this needs to be coupled with improved fiber bend performance and low fiber attenuation to ensure that ultra-low loss is attained in deployed cables.

For example, at OFC 2021, Corning announced the industry’s first ultra-low-loss fiber with 200-micron outer diameter that also meets the fiber macrobend loss requirements of the ITU-T G.657.A1 standard. This represented a major innovation milestone that enables the industry to use advanced fibers in dense cable designs with up to 432 fibers per cable.

New routes and new fiber deployments are certainly happening, with almost half a billion kilometers of new optical fiber manufactured each year – almost all of which is conventional optical fiber.

In certain scenarios, of course, this may not be an option, and at some point in the future even installing more of today’s fiber could reach a scalability limit. When those limits are reached and network operators are looking at options for new fiber types in the future, what types should they consider?

In this blog I look at a number of emerging fiber types, with a brief description of the pros and cons of each one. They are:

• ZBLAN fiber
• Few-mode fiber
• Multi-core fiber
• Hollow-core fiber

ZBLAN Fiber

The term ZBLAN actually covers a family of fluoride glasses (fluorides of metals such as aluminum or zirconium) that are doped with an array of elements, including zirconium, barium, lanthanum, aluminum, and sodium. ZBLAN components are being evaluated for several optical components, including Bragg gratings. But ZBLAN has also been made into optical fiber in order to investigate its attenuation and nonlinear aspects compared to silica glass. Unfortunately, while some impressive wideband results have been published, for the specific application of low-loss wideband fibers, ZBLAN has to be fabricated in zero gravity in order to avoid the formation of crystallites in the glass that would result in high scattering losses. So, while the properties of ZBLAN fibers look interesting, manufacturing in bulk will be a monumental challenge!

Few-mode Fiber (FMF)

Think of a mode as a potential pathway for light to propagate along the fiber waveguide. If we have a wide core, then there may be room for more than one mode. In this case, each mode may experience a slightly different path length and, if these modes are part of the same modulation symbol, the result will be “modal dispersion,” which historically was regarded as so undesirable that the entire industry focused on creating fiber with a very narrow core – single-mode fiber.

Single-mode fiber (SMF) is historically the solution to eliminate modal dispersion, and SMFs of various types and generations constitute the vast majority of the installed fiber base worldwide. But by carefully increasing the fiber core diameter it is possible to allow a small number of transmission modes, hence the reference to “few modes” (examples in research range from two to more than 20 guided modes), with each of these acting as a capacity multiplier. It’s not easy to interpret results for FMF capacity achievements because they are expressed in absolute capacity numbers – for example, 159 Tb/s over 1,000 km, which is only around double the capacity achievable today from a fifth-generation transponder using C+L band transmission. In FMF experiments, modal dispersion had to be compensated for using offline processing. In a commercial implementation, FMF will require the inclusion of multiple input multiple output (MIMO) circuits, which are already used for tasks such as PMD compensation in today’s DSPs, along with new algorithms in the coherent detector to compensate for modal dispersion. Coherent chipsets cost many tens of millions of dollars to develop and so the industry would need to be very confident that few-mode fiber is the way forward before taking this development step – which might not be justified for only double the capacity using a totally new fiber and media ecosystem.

Multi-core Fiber

Conventional optical fiber has a single core, the region in the center of the fiber that guides the light, and it’s this core that experiences the capacity limits I’ve described in Part 1 of this series. By making optical fiber with multiple cores, it has been shown that we can increase total capacity in the same diameter size of fiber we have today.

But do the signals traveling in each core remain separate, or is there crosstalk between them? A general assumption is that, if the number of cores is four or fewer, then there is minimal crosstalk between the cores, and we refer to this type of MCF as uncoupled.

These uncoupled core fibers are a promising way to increase fiber capacity with no change needed to the transponders. In fact, Infinera has already demonstrated increased capacity over this type of fiber. But uncoupled cores need core fan-outs and fan-ins at the terminals, ROADMs, and amplifier sites so that each core can be switched or amplified independently.

Coupled core fibers do have crosstalk between cores. The good news here is that they may not need fan-outs in the amplifier sites, although they do need them in terminals and ROADMs. But the cross-core coupling has to be unraveled in the receiving devices using a MIMO approach, similar to few-mode fibers. While the technology behind this compensation is well understood, once again it raises the question of whether there is a financial justification to develop new coherent DSP designs.

So, if multicore fiber becomes economically compelling, it will most likely be uncoupled core fiber in the short term simply because it works “as is” with coherent transponders without the need for special DSP development. One step in this process will be to ensure that MCF can be manufactured with high yields, which is not easy because a fault in any one of the cores means that entire length of fiber is unusable.

Hollow-core Fiber

The capacity limits of today’s fiber are a result of the fact that the signal propagates through silica – specifically the low attenuation bandwidth limited to the C and L-bands, and the fact that optical signal power is limited by the nonlinear nature of glass. Hollow-core fiber changes this by enabling the signal to propagate through a vacuum, air, or inert gas – different researchers are pursuing different approaches depending on their objective.

The combination of a wider low-attenuation bandwidth and a far lower nonlinear penalty means that HCF could offer far higher capacity per fiber pair. In fact Infinera (Coriant at the time) still holds a Guinness World Record for the highest data transmission rate over hollow-core fiber – 73.7 Tb/s.

The U.K. company Lumenisity announced at the OFC conference in March of this year a breakthrough in low attenuation – their dual-nested antiresonant nodeless fiber (DNANF) has achieved losses that are now less than conventional G.652 (i.e., Corning SMF-28) fiber – both in the C-band and as low as 1310 nm in the O-band. This results in a far wider low-latency transmission window, and much lower nonlinear effects. So, both these properties are a major contributor to increasing capacity. Note that the latest DNANF still has a slightly higher attenuation in the C-band than an ultra-low-loss fiber like SMF-28 ULL, but HCF still has a long way to go in terms of reaching its theoretical attenuation minimum, whereas silica-core fiber is within a fraction of a dB of its theoretical minimum. In other words, HCF still has the opportunity to reduce its attenuation even more.

As a separate market driver to capacity, light travels significantly faster in a vacuum or air than in glass – 300,000 km/s vs. 200,000 km/s, respectively. This is extremely attractive in low-latency applications. In fact, low latency over relatively short distances is the most promising application of HCF today, as opposed to its potential for increased fiber capacity in long-haul networks.

Network operators such as euNetworks are already offering low-latency financial services over HCF. And BT has tested HCF in 5G fronthaul networks, where the 100 microsecond fronthaul protocol timeout limits the distance between the hub site and antenna to approximately 15-20 km. Using HCF, this distance can increase by about 50%, which has a major positive impact on 5G network economics because so many more radio masts can be served by the same CU/DU location.

HCF does not require changes in the coherent transponder DSP other than needing transponders that can operate in other wavebands (like the O, E, and S-bands). There is no technical barrier to building wideband transponders – we just need a business case to start making them.

Amplification in HCF is more challenging than conventional fiber. The Raman effect does not work in HCF, and in order to use EDFAs, the HCF is spliced into a section of regular fiber that acts as the gain medium.

While HCF is an exciting development, there are still many questions to answer. Is it possible to manufacture these intricate structures at the scale of millions of kilometers per year? How robust are these structures? How difficult is it to fix connectors and to repair cut fibers? How do we recreate the vacuums or re-gas the fiber during a fiber repair?

Conclusion

It’s a challenge to write a meaningful summary after covering so much ground in this blog series. But there are some things we know for sure:

• Existing fiber has enormous capacity, and we will continue to unlock this using advances in transponder technology and by lighting C+L line systems.
• Deploying new fiber could be an option for some operators, and it is highly likely they will deploy more conventional SMF because it is highly cost-effective, with a fully mature ecosystem.
• In the short term there is a separate, compelling business case for HCF deployment because of its lower latency in applications such as 5G fronthaul. But manufacturability of HCF remains a challenge in the short term.
• For applications where spatial compactness is a driving factor, there is an argument for FMF, MCF, and HCF fibers. For both uncoupled MCF and HCF we could see future capacity increases without the need to develop specific transponder functions such as MIMO compensators. But specialized compensators are required for FMF and for coupled MCF.

I hope this blog series has been a useful overview of the available technologies. In my final part next time I will explain the special case of how the industry is scaling capacity in submarine networks.