Maximizing Capacity and Data Rate Across All Types of Submarine Cables

Geoff Bennett highlights submarine cable capacity options and discusses where the industry goes from here to squeeze out more bandwidth.

As published in the July Issue of SubTel Forum Magazine

By Geoff Bennett
July 25, 2022

According to TeleGeography, there are 486 active and planned submarine cables operating around the world today.  Traditionally, these cables were designed with an engineering life of 25 years, so it’s not surprising that some of them began their design and planning phases in the closing years of the last century.

Figure 1: Evolution of fiber capacity and cable capacity for trans-Atlantic cables

As you might expect, newer submarine cables have higher performance than previous generations, as measured by a combination of wavelength data rate and capacity.  This is shown for trans-Atlantic cables spanning two decades of cable evolution in Figure 1.  Note that all the figures are based on using the latest fifth-generation coherent transponders.

What you can see from this table is that fiber pair capacity peaked with the MAREA cable, which is an uncompensated large-effective-area fiber cable with very short repeater spacing.  As we move to space-division multiplexing (SDM) architectures, fiber pair capacity goes down somewhat, but SDM cables can support many more fiber pairs so that overall cable capacity continues to scale.  The latest example of this is referred to in Figure 1 as the “Meta Cable” (it does not yet have an official name) announced by Dr. Steve Grubb, Global Submarine Network Architect for Meta, at the PTC conference in January 2022.  In addition to scaling total cable capacity, SDM also helps to reduce costs for individual fiber pair operators because maritime maintenance costs for submarine cables are based on the cable as a whole, so by supporting more fiber pairs, SDM cables can help to spread that cost.

Given that global submarine capacity demand is increasing by about 35 percent per year, it would seem to make sense to deploy as many new cables as possible and forget about the older cables, right?

Well, maybe not.  A submarine cable can require five to seven years between the first meeting to discuss the new cable and the first services running over the completed system.  And a trans-Atlantic submarine cable will cost between 100 and 200 million dollars – plus ongoing maintenance costs for its entire life.

In some geographical areas, the demand profile may be such that existing cables can easily meet growth levels if they receive transponder upgrades during their lifetime.

So, it makes sense for all types of cable that, during their 25-year (or more) engineering lives, network operators will upgrade their submarine line-terminating equipment (SLTE) to extend their economic lives.

What Is “Performance”?

As a note on terminology, in this article I use the word performance as a synthesis of maximum wavelength data rate and maximum spectral efficiency for a given transponder.  These two goals are valuable to the network operator in their own ways, but it is not possible to optimize for both at the same time.  To achieve the highest wavelength data rate, one may need to compromise on spectral efficiency, and vice versa.  The best transponders are the ones that minimize the amount of compromise needed.

The Pitfalls of Conventional Wisdom

The transponders used in submarine networks are usually based on the same optical engines as terrestrial transponders, but often use specially selected optical and electronic components in order to minimize noise in the circuit.  But the best submarine transponders will go a step further and include features that help to optimize performance in all types of submarine cables.  Modern transponders are highly programmable, and in terrestrial networks, conventional wisdom dictates:

  • Always use the highest baud rate of the transponder
  • Always use probabilistic constellation shaping (PCS)

…to deliver maximum optical performance.

The reason this wisdom holds true for terrestrial networks is that we are dealing with large numbers of wavelength services over diversely routed optical paths, usually in a meshed topology.  In other words, the design philosophy is to keep things simple while achieving the highest practical performance.

In submarine networks, fiber capacity is incredibly valuable, while topology is relatively simple, so it becomes worthwhile to spend time optimizing the fiber channel plan and configuring different parts of the spectrum in different ways to achieve maximum performance.  Moreover, a high-quality subsea transponder solution will include a degree of automation for this design process.

Are High Baud Rates Always Good?

Modern transponders can operate at baud rates of up to 100 GBd, which delivers superior optical performance – especially if we are optimizing for high wavelength data rates.  The reason for this is that by using high baud rates we can create high data rates while using a less spectrally efficient (but longer-reach) setting for our probabilistic constellation shaping modulation.

What Is Probabilistic Constellation Shaping?

In optical transmission we load the data bits we’d like to transmit into modulation symbols.  Previous generations of optical engine used fixed modulation constellations such as PM-QPSK (4 bits per symbol), PM-16QAM (8 bits per symbol), or PM-64QAM (12 bits per symbol).  As we load more bits into each symbol, the spectral efficiency increases, but optical reach drops exponentially.  Probabilistic constellation shaping uses a high-order constellation, such as PM-64QAM, and probabilistic data assignment techniques to reduce the frequency of the most problematic constellation points.  In this way PCS allows us to optimize spectral efficiency at a given optical reach and provides far greater granularity of performance compared to fixed constellations.

Figure 2: Fixed high baud rates are not always the right approach for maximum capacity

However, maximum baud rate operation is not always the case for modern submarine cables that fit into the categories of large-effective-area/positive-dispersion cables, nor for emerging SDM cables.

In Figure 2 I show that the quality of spectrum in these uncompensated cables varies across the C-band.  Lower frequencies have a more benign optical environment.  The reasons for this are complex but include the facts that the effective area of a fiber has a frequency dependence and is larger at lower frequencies, dispersion is higher at lower frequencies, and amplifier noise figures are better at lower frequencies.

To avoid bit errors being experienced by the users of the cable, transponders must operate with a transmission margin above the forward error correction (FEC) threshold.  But the operator of a given fiber pair will add a safety margin to this limit to tolerate short- or long-term changes in the fiber pair caused by fiber repairs, long-term aging, or transient phenomena.  Thus, the commissioning limit shown in Figure 2 is above the FEC threshold and the difference between the two is usually referred to as the operating margin for the fiber pair.

When we configure the wavelengths on these types of cables, we start by using the highest possible baud rate and then adjust the probability distribution within the PCS settings to match the client services we need to carry – and these settings must be above the commissioning limit at the high-frequency end of the spectrum.

For example, on the MAREA trans-Atlantic cable, Infinera’s ICE6 transponder can operate wavelength data rates of 650 Gb/s.  Clearly this does not match any existing Ethernet service data rate.  But ICE6 is a dual-wavelength transponder with capacity that can be pooled to create a 1.3 Tb/s super-channel.  This capacity could be used to carry, for example, 13 x 100 GbE services, or two 400 GbE and five 100 GbE services, or three 400 GbE and one 100 GbE service.  In fact, any combination of 100 GbE and 400 GbE can be combined over the super-channel.

If we are operating at the highest baud rate, then the PCS probability distribution can be adjusted to carry fewer bits per symbol to close the more challenging high-frequency parts of the fiber spectrum.

If we used the same configuration in the low-frequency spectrum, we would be leaving excess margin in the fiber because this part of the spectrum has better optical characteristics.  By turning down the baud rate slightly, we can reduce the amount of spectrum that a given wavelength uses, and to ensure we carry the same data payload we can increase the bits per symbol of the PCS.  When we adjust PCS this way, we “soak up” the excess margin, but since this wavelength is narrower thanks to the lower baud rate, we free up spectrum for more wavelengths to be added on the fiber.

This Margin Monetization™ technique is the key to achieving record-breaking performance on uncompensated cables such as MAREA, Dunant, Seabras-1, and EllaLink, with up to 25 percent more fiber pair capacity compared to a fixed high-baud-rate approach.  So high baud rates are great, but the ability to vary them slightly to squeeze in more channels is even better.

Is PCS Always the Best Modulation?

What Is a Subcarrier?

The first types of coherent transponder transmit a single carrier that looks like one contiguous band of optical energy.  This carrier is modulated as one entity.  Infinera pioneered the use of Nyquist subcarriers that are digitally synthesized in the coherent transmitter and generated from a single laser.  Each subcarrier can be modulated independently, and because they operate at lower baud rates, they experience a lower impact from impairments such as chromatic dispersion.

Let’s be clear – PCS is almost always the best modulation to use to enhance capacity-reach in terrestrial networks, as well as in uncompensated and SDM submarine cables.

Figure 3: The use case for non-PCS modulations

But when we look at some examples of older cables that have a higher nonlinear penalty, we see that PCS may not always deliver the highest performance.  Figure 3 shows that PCS efficiency tends to fade away below the equivalent of QPSK – which would often be the case in older, dispersion-managed cables.

Infinera’s ICE6 optical engine supports a non-PCS hybrid modulation that can be applied on a per-subcarrier basis.  In Figure 3 we show the eight subcarriers from a single laser modulated using 3QAM on the outer four subcarriers, and QPSK on the inner four.  This results in a “3.5QAM” modulation that has better optical performance than using PCS for this particular cable.  Other hybrid fixed modulations are possible, as well as overlaying more sophisticated hashing schemes to ensure phase and power balancing across polarizations and subcarriers.  A highly balanced constellation like this will have significantly better performance in an environment with high nonlinear penalties.  Using these techniques, we have seen between a 25 percent and 50 percent increase in fiber pair capacity compared to a PCS-only implementation.

In Summary

It’s essential to track transponder evolution during the lifetime of a submarine cable.  Conventional wisdom gained from terrestrial network deployments does not always transfer well to submarine cables, no matter what type of cable.  The use of variable high-baud-rate transponders and non-PCS modulation types can help fiber pair operators enhance performance dramatically in all submarine cable types.

About the Author

Geoff Bennett is the Director of Solutions & Technology for Infinera, a leading manufacturer of Intelligent Transport Network solutions. He has over 25 years of experience in the data communications industry, including IP routing with Proteon and Wellfleet; ATM and MPLS experience with FORE Systems; and optical transmission and switching experience with Marconi, where he held the position of Distinguished Engineer in the CTO Office. Geoff is a frequent conference speaker and is the author of “Designing TCP/IP Internetworks”, published by VNR.

Click here to read more articles from the Regional Systems Issue of SubTel Forum Magazine or read this article on our archive site here.