NANOG 85 Tutorial: Optical Fiber Capacity Limits – Where Do We Go Next?

In Part 1 of this blog series I looked at “where do we go next” when we look for increased fiber capacity, and I focused on technologies that are commercially available.

So, what do you do if you’ve upgraded to the latest transponders, already moved to flexi-grid, implemented C+L on your line system, need to keep your existing fiber in the ground, and don’t have the option to light more fiber pairs?  In this blog, I’ll look at the vast amount of spectrum that lies in existing optical fiber beyond the C- and L-bands.

Multiband Transmission

One potential option using the fiber you already have in the ground is to expand beyond C+L, to “multiband” transmission.  This term usually refers to transmission beyond (and including) the C- and L-band.  In an blog from August 2021, I explained how these bands got their names and why they have traditionally accepted boundary limits.

Figure 1: Optical fiber attenuation and transmission bands

Figure 1 shows the traditional wavebands in single-mode optical fiber, and the solid black line shows the attenuation of the fiber, which tells us how much of the optical signal is absorbed per kilometer of fiber.  You can see that the band of wavelengths marked as the C- and L-bands is at the attenuation minimum for silica core fibers (i.e., the type of fiber you have in your optical network today).  The attenuation level is a major factor in determining the total cost of ownership of a long-haul network because we need to amplify the signal at regular intervals along the path.  This results in three cost factors that need to be optimized: the cost of the amplifier technology (e.g., EDFA vs Raman), the costs associated with the amplifier location itself (both CapEx and OpEx), and the cost of regeneration.  Of these three, regen costs tend to dominate the calculation and so, for example, most long-haul networks will have amplifier locations at distances of 80-100 km.  Longer spans are possible using higher-power (and more costly) amplification, but these result in fewer amplifier locations, which saves money.  However, they also have a negative impact on optical signal-to-noise ratio (OSNR), which may result in the need to regen the signal.  Overall, we tend to try to minimize the number of longer spans on any given route for this reason.

The lower the attenuation at the operating wavelengths, the more favorable these economics will be.  This is why the whole industry has tended to focus on the C- and L-bands, because this is where the attenuation is lowest in conventional optical fiber.

If the C- and L-bands are already lit, dark fiber is not available, and further capacity is needed, then we need to look to other wavebands for more spectrum to use.  Generally speaking, using wavelengths longer than the L-band (referred to as the U-band) may be a practical challenge because signals sent at these wavelengths may experience much greater bending losses.  Future fiber deployments might account for the need for U-band transmission by enforcing more stringent guidelines for fiber bending during installation and operation.  But in general, this is against current trends, where more flexibility (reduced bending sensitivity) is desired to allow more fibers to be pulled into installations such as data centers.

This implies we should be looking at shorter wavelengths – below the C-band.

Fiber Capacity Considerations at Shorter Wavelengths

In terms of fiber capacity at these wavelengths, we need to consider the following important aspects:

1. Is there a practical amplification technology for this waveband that could be commercialized?
2. Even with efficient amplification, does the higher intrinsic attenuation in silica core fiber outside of the C- and L-bands make it impractical to operate multiband transmission?
3. What are the nonlinear penalties at shorter wavelengths?
4. Is it possible to manufacture transponders at shorter wavelengths?

Multiband Amplification

Figure 2: Dopant elements for multiband fiber amplifiers

The answer to the first of these questions is a qualified yes – amplification technologies based on alternative rare earth elements or extended Raman amplification have been demonstrated experimentally – as shown in Figure 2.

Lab demonstrations are a long way from the existence of the practical and economical optical amplifiers we already have for the C- and L-bands, but in theory this is simply an engineering problem that could be solved.

In addition, it may be possible to use ultra-wideband Raman amplification, which is not trivial – and many network operators prefer not to deploy Raman at all because it requires specialized safety training, higher-quality splice repairs, and more sophisticated operational procedures.  But these are all engineering or operational issues that could be solved if there was a market for multiband Raman amplification.

Optical Performance at Shorter Wavelengths

At wavelengths lower than the C-band, three factors influence optical performance:

• Higher attenuation
• Lower chromatic dispersion
• Smaller effective area

It’s clear that the attenuation in conventional fibers rises significantly as wavelengths decrease.  In Figure 1 you can see a large peak in the E-band, as well as other smaller peaks, that result from absorption by water molecules within the fiber.  It’s possible to exclude water when fibers are manufactured, but this adds cost to the fiber and historically the position was taken that these wavelengths would not be used anyway, so why pay more for the fiber?  Moreover, there are other effects, such as Rayleigh scattering and UV absorption, that also cause increased attenuation at shorter wavelengths, and these are independent of water impurities.

While amplification technologies may be available to deal with attenuation, amplifier spontaneous emission (ASE) is a major contributor to the noise that is added to an optical signal on its long-haul journey.  Higher attenuation implies more amplifiers, which means more ASE noise.

Figure 3: Chromatic dispersion for different fiber types

For a given optical power level, the nonlinear penalty depends mainly on the value of chromatic dispersion and the effective area of the fiber, and for both parameters it’s a case of the higher the better.  But both chromatic dispersion and effective area are lower at shorter wavelengths, which means that nonlinear penalties will be higher.  Figure 3 shows the chromatic dispersion vs. wavelength for three common fiber types: G.652, G.653, and G.655.  Notice the direction of the slope and the zero-dispersion wavelength in each case.

Optical Performance: Network Design Consequences

From the point of view of network design, lower optical performance implies two options.

The first is if network operators absolutely require their transponders to be configured with the highest data rate for a given optical reach in order to optimize economics.  The higher attenuation outside the C- and L-bands means that they will need to move their amp huts closer together.  I have italicized this text because it’s highly impractical for most network operators.  The amp hut locations are already in place based on C- and L-band attenuation and amplification performance.  It may be possible to add amp huts, use a more expensive amplifier technology, or regenerate the signal, but these will all increase the TCO for these new wavelengths.

Second is that the operator chooses to accept lower transponder data rates, or perhaps use wider channel spacing in order to close links with the current amp hut spacing.  Once again, the result is that the wavelengths deployed outside the C- and L-bands will have a higher cost per bit – the exact opposite of what we require from evolving optical transmission.

For operators who do not have access to dark fiber on this route, the additional cost may be acceptable.

The C-band Is Annexing the S-band

Note that the long wavelength section of the S-band may be an exception to the challenges of multiband transmission.  At these wavelengths, attenuation is only slightly higher than the C-band, dispersion in G.652 fiber is still relatively high, and the wavelength dependency of lower effective area may not be enough to significantly affect optical performance.  In fact, many operators already use parts of the S-band in the form of Extended C-band implementations.  In these cases, the development of better gain flattening on C-band EDFAs means that the “classic” 4 THz C-band has expanded to 4.8 THz, spreading into the S-band at one end and the L-band at the other.  Further expansion to Super C-band, 6 THz EDFAs, pushes the “C-band” approximately 10 nm into the S-band, and such systems will become more widely deployed in the next few years.

Transponders for Shorter Wavelengths

Transponders can be manufactured to operate across any of the wavelengths in the O-, E-, S-, C-, and L-bands. This is one area where there are no technical challenges.  However, C- and L-band transponders are already manufactured in high volumes, resulting in lower unit costs.  This will not be true for multiband transponders initially, and it’s likely that volumes for these transponders will always lag behind C- and L-band.

Summary of Multiband Transmission

Network operators are already deploying and operating multiband systems – both because C+L transmission is now normal practice, and because the C-band has expanded over time into both the L-band and the S-band.  S-band “annexation” will continue with the move from Extended C-band to Super C-band in the next few years.

While optical amplification in the O- and E-bands has been demonstrated for CWDM and in the emerging “coherent lite” initiative, long-haul and subsea transmission is a challenge in these bands because of the lower optical performance of silica core fiber at these wavelengths.

In short-reach applications in which amplification is not required, multiband communication is commonplace thanks to the use of coarse wavelength-division multiplexing (CWDM).  But CWDM is about low cost, not high fiber capacity.

In the next blog in this series, I’ll look at the options for deployments using new or emerging fiber types.