FEATURE

Capacity conundrum

With providers facing the challenges of keeping up with internet traffic growth, terabit-scale systems are on the rise, Robert Roe writes

As internet traffic grows year on year, service providers have to deliver more capacity at cheaper rates. Simply putting more fibre into the ground is a very expensive solution to the problem in most cases, so network providers are under pressure to come up with innovative solutions that will increase the capacity of fibre that has already been deployed within the network infrastructure.
Existing capacity is beginning to reach the ceiling of what can be achieved, according to Uwe Fischer, chief technology officer for Coriant: ‘What you are seeing when you look into system design – today’s state-of-the-art technology – is 100Gb 4QAM coherent using a 50Ghz grid in the system. When you look from a spectral efficiency perspective, the 100Gb 4QAM modulation format is more or less using the complete passband.’

If the spectrum is already fully utilised by 4QAM modulation, one obvious way of squeezing more information through the fibre would be to go to different modulation formats with a higher number of modulation states – 8QAM or 16QAM. However, there is a trade-off: the reach – the distance over which the signal can be transmitted – falls because of increased noise when the system is being operated at its non-linear limit. This cannot be compensated for by just adding more power, because the system is already operating at that limit.

One approach, therefore, is to develop flexible, intelligent technology that includes both the higher modulation states and also super-channels, which bundle several carriers into one common signal. In this way, the technology will allow the modulation state appropriate to the range to be selected, while the capacity can be achieved by combining the carriers into a compound signal.
Ciena, Infinera, and Coriant have all demonstrated terabit scale network capabilities using super-channel technology. Infinera has been shipping 500G super-channel line cards for the past two years, while other DWDM vendors are expected to begin shipping 200G line cards this year, and to roll out super-channel products at some point in the future.

Ciena demonstrated this technology in collaboration with Comcast. Helen Xenos, director of product marketing at Ciena, said: ‘We used a lot of technologies here. The first one was the WaveLogic coherent technology to do 16QAM modulation – this allows it to carry 200Gbps on a single wavelength in less than 50 GHz of spectrum. That more than doubles the capacity of what they were able to do before in their existing, state-of-art 100G network.’

Ciena collaborated with Comcast in early October 2013, completing a live field trial of a 1Tbps optical transmission spanning nearly 1,000 km across a span of Comcast’s US network connecting Charlotte, NC and Ashburn, VA. The primary focus of the trial was to show the distance achieved using super-channel based technology; it also demonstrated the ability to increase the capacity of the network by a factor of 2.5. The network simultaneously carried customer traffic over 10G, 40G and 100G wavelengths, all coexisting on a mix of flexible and 50GHz-spaced fixed grid channels.

Coriant has achieved similar results in a trial it conducted in Australia. This super-channel transmission trial was conducted over a 1,066 km fibre optic ring in South East Queensland, Australia using Coriant’s innovative FlexiGrid technology on NBNCo’s Transit Network. Fischer said: ‘This was a super-channel experiment. Here we used 100Gb 4QAM per carrier as a base modulation format and we compiled 10 of these carriers into a one terabit compound signal.’

Infinera not only demonstrated a terabit capacity but also broke a world record – for fastest provisioning of long haul optical transmission capacity in 2013. Infinera collaborated with DANTE  (Delivery of Advanced Network Technology to Europe) to install and activate an 8Tbps long-haul super-channel across a section of the GÉANT Network, from Vancis Amsterdam, the Netherlands to GlobalConnect Hamburg, Germany. The GÉANT network is a high bandwidth pan-European research and education backbone that connects national research and education networks (NRENs) across Europe. Sixteen of Infinera’s 500Gb line cards and 32-fibre connections were deployed at each end of the link, and once the super-channels were in operation, a 100-Gigabit Ethernet (GbE) service was provisioned over the fibre.

Geoff Bennett, director of solutions and technology at Infinera said: ‘Service providers have to be able to deal with the exponential increase in demand on modern core networks, and 500G super-channels are a great way to do that. With the world record attempt, we showed how robust and scalable a super-channel solution can be.’

The total time from the insertion of the first super-channel line card to the activation of the 100Gb service was 19 minutes and one second, resulting in a provisioning rate of 26.02Tbps per hour. All of these trials demonstrated that terabit networks can be achieved using long established fibre that has been deployed for 20 years in some cases.

According to Fischer, some users might decide that the increased modulation route might be a good solution for metro/regional networks or for short-distance high-capacity interconnects between data centres – they may just accept that the range goes down. However, the solution does not work for the full network, he added.

‘Currently no one builds dedicated networks for special-use cases. The most efficient way is to have one network that can cover a number of different services. It is very important that these kinds of specific solutions can co-exist with other traditional services on the same network,’ he continued. That was one of the issues showcased in the Australian experiment.
All three of the company’s trials used super-channels to achieve terabit-per-second network capacity. However, this technology needs high modulation states; this increases noise, shortening the maximum distance that can be achieved. Currently this presents a trade-off that network providers must balance to achieve the results that customers want – by increasing the modulation state there must be more signal shaping and some kind of forward error-correction.

Older network architectures used QPSK modulation, but the move to 8 or 16 QAM presents some new challenges. Xenos said: ‘In QPSK, the four constellation points, also called symbols, are well separated from each other, but with 16QAM, there are 16 constellation points in the same space. This means there is significantly less separation between constellation points or symbols, hence much higher sensitivity to noise.’

As modulation technology becomes the standard to achieve high capacity across fibre networks, each company has had to develop solutions to combat this issue of increased noise.
Fischer pointed out that for 8QAM the spectral efficiency is improved by a factor of 1.5. Going to 16QAM does not doubling the 8QAM improvement again, but increases the spectral efficiency by another factor of 1.3. ‘However this trades against reach, because you suffer from more noise in the system,’ he added.

Bennett, of Infinera, said: ‘The optical transmission capacity from a super-channel can be compared to the torrent of water from a fire hydrant. Something has to manage that torrent so that it can be used in an effective way by lower-speed client services.’

Each network provider will have their own solution for squeezing the signals as close together as possible in technologies like 16QAM. This involves squeezing the traditional band gap and processing the signal using a DSP installed in the network.

Xenos said: ‘Different vendors’ solutions will have different types of performance, depending on the type of forward error correction that they are using and the DSP that they have. This means that not every solution at 16QAM will be the same, either.

‘Soft-decision forward error correction is a key enabler to a longer distance optical transmission, but not everybody’s forward error correction is the same. You will be able to tell, with the kind of deployments that are coming out, who has the better solution because of the types of distances that can be achieved,’ she concluded.

Infinera has been utilising photonic integrated circuit technology in its super-channel solutions, because of the space, power and reliability benefits associated with the reduced number of optical components. Bennett said: ‘Super-channels have become the favoured technology for implementing DWDM capacity at 100Gbps and beyond. As we scale the number of wavelengths in a super-channel it seems obvious to do that using a highly integrated approach like the PIC – and the service provider market obviously agrees because PIC-based super-channels are now one of the most popular ways to implement long-distance transmission capacity.’

He continued: ‘A PIC is the optical equivalent of a multi-core Pentium CPU – and shares all of the same advantages of electronic integration; namely smaller footprint, lower power use, higher reliability, and lower cost compared to a line card built from discrete optical components.’

Other providers have developed their own solutions. At Coriant, Flexigrid technology is a key component of DWDM it allows the flexible assignment of optical bandwidth to channels enabled by FlexiGrid technology. This allows higher bit rate transmissions by maximising flexibility in channel spacing. FlexiGrid also introduces the concept of virtualisation of physical layer resources. An operator can adapt the wavelength grid to the needs of high capacity, long reach transport as well as increase spectral efficiency and thus the overall capacity of the system.

Fischer said: ‘You have multiple carriers travelling on different optical wavelengths, or you could consider them as different colours. In a typical system, you have a filter that acts as a grid. It provides a filter gap between all the channels. This means you have one band, which is 50GHz apart from the other, and the transmission bandwidth of that band is roughly 37GHz. In between, there is a band gap where, if you were to apply a signal there it would just not go through, so you have filtering between the different colours.

‘Now if you separate the colours by electrical signal processing, then you can avoid the filtering in between the different colours. This removes the pass bands between the different transmission bands, allowing you to utilise the gaps in the signal by squeezing the transmissions closer together.

‘There are specific ROADM devices that allow this,’ said Fischer. ‘They allow you to program, in a flexible fashion, the transmission pass band and they also allow you to avoid any filtering in between the different pass bands.’

Signal shaping is important to reduce degradation of the signal when using higher modulation states like 16QAM. Fischer gave an overview of why this is necessary: ‘In the classical system, there is filtering in the optical signal. If you avoid that filtering, you need appropriate signal processing for the electrical signal. If you do not have this optical filtering, then you must shape the signal before you put it onto the fibre to avoid overlapping from neighbouring channels [producing noise], and this requires additional signal processing which is also part of the technology.’

Ciena’s solution to the problem of noise over higher distance transmission has been to implement a new version of Raman amplification. Xenos said: ‘Historically Raman has had negative connotations because it has been very difficult to deploy, so many people have stayed away from deploying Raman solutions.’
Raman amplification is a technology that was originally developed for the optical market in the 1970s but has generally played a secondary role to the more popular technique of EDFA amplification, even though Raman can provide improved signal to noise ratio.

The high power used by the module and the fact that Raman uses the fibre plant as the gain medium created challenges for turn-up and troubleshooting new optical connections as well as for future fibre maintenance which historically made Raman unsuitable for some applications. For example turning up a high-power Raman amplifier across a fibre span that has a dirty connector or inadequate splice quality can result in damage to the fibre plant. However these concerns have been addressed with Ciena’s new Raman based solution.

Xenos continued: ‘There is a new type of Raman that is required, because we need to deploy very simply and we expect Raman to become more widely deployed with these higher capacity channels.’

With competing solutions to reduce noise across super-channel based networks, it will be the end-performance that really sets apart one competitor from another. Once all the hardware has become commercially available, time will tell which solutions can achieve the best performance in terms of distance and capacity of the network.
Ciena has integrated Optical Time Domain Reflectometer (OTDR) capabilities directly into the Raman amplifier. The integration of the ODTR capabilities provides a controlled activation by autonomously testing the fibre plant to detect unacceptable connector and fibre conditions before the Raman amplifier is turned on. This controlled turn-up process prevents equipment and fibre damage which could cause additional deployment costs and delay. Coriant too is using Raman to overcome reach limitations and providing an integrated OTDR.

With terabit networks already in sight, is there a limit to what can be achieved? According to Fischer: ‘Strictly speaking, from a theoretical physics perspective, there is no limit: 20-30 terabit is the practical limit, because then the signal will degrade so much it is no longer suitable for practical transmission purposes. It is more of a practical limit, in that trade-off of distance and capacity.’

To get out of that ‘trade-off trap’, he said, one would have to apply more power into the fibre. But ‘with traditional fibres, this is not possible because they are working on the non-linear limit so then the only way out would be to use new fibres.

‘Now there are a number of fibres that have been proposed. The first is called hollow-core fibre; the other is called multi-mode fibre. This is exactly what we did with Telecom Austria, where we used space division multiplexing experiment that used this new type of fibre.’

Given the expense of putting new fibre into the ground, Fischer believes that this is still a solution for the future. Nonetheless, he believes that: ‘At some point in time, for very demanding connections, the industry will consider deploying these new fibres which would increase the non-linear limit, so you can just apply more power and then you can increase the reach again even while using the higher-state modulation formats.’ Coriant has already demonstrated record-breaking transmission rates of of 57.6Tb/s over solid-core multi-mode fibre and 57.6Tb/s over hollow-core photonic bandgap fibre.

The experiment with Telecom Austria showed that there are such alternate paths into the future. It is not yet a current solution because these fibres are not widely deployed and there are challenges in manufacturing them in mass volumes.

However, Fischer believes that, in some areas and for some applications, new fibre technology will be used. ‘There are some cases where people are either doing new deployments – for instance in Australia, where we supply the national broadband network. There are others like in the US for instance – important interconnect routes for electronic trading like in Chicago and New York. People are already deploying new fibres in a more straightforward fashion to avoid some microseconds of delay. When you talk to Corning or other fibre manufacturers, they will tell you there is a good and constant market for new fibre deployments worldwide. When people start using these new fibres in new deployments, we will start to get these capabilities into the ground and then we can start using it from a system perspective.’

He concluded: ‘People say that internet traffic is doubling year over year, so you can calculate when all the tricks and advances in system design will hit the ceiling of available capacity. That is why people start thinking, now, about beyond what is available today.’

Feature

CableLabs is spearheading efforts to develop a proposal that uses coherent optics to dramatically boost the capacity of hybrid fibre coaxial networks, reports Andy Extance

Feature

Systems vendors are using intelligent software to squeeze more performance from optical networks. Pauline Rigby reports on developments at OFC 2017