The evolution of fibre networks has led to increased bandwidth, which in turn has enabled new services, but this growth in the network has not translated into a proportional growth in revenue for operators. To attempt to remedy this unsustainable situation, several European research projects have been set up to investigate new technologies for optical access networks.
‘Effectively revenues and usage of network capacity are becoming decoupled over the last decade or so. That is really demonstrated by the flat rate charging regime where you buy a bundle of capacity and services, and how much you use them does not affect how much you pay for them. That is a problem for the future because providing more bandwidth isn’t free,’ commented Professor David Payne from Trinity College Dublin.
Payne is coordinating a European-Commission-funded project called DISCUS, which is a charmingly imprecise acronym for ‘DIStributed Core for unlimited bandwidth supply for all Users and Services’. The project’s aim is ambitious – the researchers want to develop an entirely new architecture that will allow the access network to scale economically – but it is not impossible. At the European Conference on Optical Communications (ECOC) exhibition last autumn, the project team showed their first demonstrator: a software-reconfigurable, DWDM long-reach passive optical network (LR-PON) test-bed.
Does DISCUS show us the future of optical access? Certainly it may be one future, but it’s not the only possibility. To address the challenges facing optical access networks, a number of European research projects were created under the European Commission’s 7th Framework Programme, which awarded funding between 2007 and 2013. These projects are now beginning to report results. At least three major projects have presented their findings at high-profile telecoms conferences and meetings in the past year. Interestingly, these projects have taken contrasting approaches.
The direct approach
A project called CEEOALAN (Cost and Energy Efficient Optical Access and Local Area Network) was highlighted by systems vendor ADVA Optical Networking at ECOC 2015. Taking a nearer term view, this project has focused on designing next-generation access networks around direct-detection methods, with particular attention on the modulation formats that will be needed for high bit rates. The European funding allows a post-doctoral researcher from the University of Bristol to work with ADVA for two years, and devote himself to the project – which runs until August 2016.
‘Most people, and that certainly includes us at ADVA, agree that it is worth following a direct detection approaches for next-generation access, even for bit rates beyond 10Gb/s,’ asserted Klaus Grobe, director of global sustainability at ADVA Optical Networking. ‘At least, somebody has to look at the possibility of implementing higher bit rates beyond 10Gb/s cost efficiently. We expect that with the proper implementation it [direct detection] will be cheaper compared to coherent solutions and that is exactly what the CEEOALAN project set out to prove.’
Specifically the project team will carry out a theoretical analysis and comparison of key criteria – such as optical power budget, cost and power dissipation – for different modulation formats including pulse amplitude modulation (PAM), optical duobinary, carrierless amplitude and phase (CAP) modulation and discrete multitone (DMT), a form of orthogonal frequency division multiplexing (OFDM). Experimental demonstrations of one or two variants are also planned before the end of the project.
‘It is clear that the higher the bit rate the more problems that will face direct detection,’ said Grobe. This is why alternative modulation schemes are required. However, the characteristics of the access network, which tends to operate over limited distances, will help alleviate some of the technical challenges. A standard reach class of 20km is typical in commercial PON systems. Even in a site consolidation scenario, where operators bypass some of their central offices, then the upper bound for the reach of next-generation access is expected to be in the region of 50km, according to ADVA’s estimates.
‘That will of course solve many things because it eases the aspects of dispersion and also possibly sensitivity,’ he said. ‘That indeed points towards a direct-detection approach. If we are not talking hundreds of kilometres then one of the advantages of coherent detection, namely the equalisation of dispersion compensation, is partially gone,’ he added.
The project has already narrowed down the options for an alternative modulation scheme. ‘We are looking at two or maybe three contenders,’ said Grobe. First on the list is PAM4, a scheme that uses four levels of intensity modulation. ‘If I have four levels instead of two then you can immediately see that the spaces in between these levels have to become smaller and that increases noise which in turn reduces the overall reach of the signal,’ Grobe points out. However, the advantage of PAM4 is that it will enable the use of cheaper electronics. ‘The symbol rate is not much higher than 10Gb/s. This means that I may be able to re-use some of the components from 10Gb/s,’ he explained.
This concept of component re-use has become one of the recurring themes of the CEEOALAN project, Grobe adds. ‘Can we re-use products from 10Gb/s that have become sufficiently cheap? In general, the answer seems to be yes, at least to some limited extent, largely because the modulation schemes under consideration also decrease the analogue bandwidth or symbol rate.’
This principal would also hold true for DMT modulation, although in a slightly different way because the signal is split into a large number of sub-carriers. Grobe explained: ‘We are considering something in the range of 1000 sub-carriers. If you split the signal into 1000 sub-signals then each sub-signal is slower by a factor of 1000. Each sub-signal is not affected, for example, by chromatic dispersion. On the other hand you have to create the DMT multi sub-carrier signal in the digital domain. This requires fairly sophisticated and more expensive analogue-to-digital and digital-to-analogue converters that can provide a suitable resolution.’
The modulation schemes investigated have turned out to be closely matched in terms of performance, with no clear winner emerging. ‘Today there is no simple black or white answer,’ said Grobe. ‘Let’s consider for a moment a signal of 25Gb/s, then for up to 25km duo-binary is best but if it has to go up to 30km then PAM4 is best.’ Ultimately, one scheme must be selected. ‘Ideally people would like to have a single piece of optics that can cover all situations. Per bit-rate class there can be different modulation schemes, but per reach class when you are talking about a granularity of ten kilometres or even less, I don’t see that as being acceptable,’ he added.
Although the technology choice is not simple, the candidate technologies share a number of characteristics. ‘All of them use a pretty simple direct detection optical front-end, some require additional signal processing for example post processing for some dispersion compensation, and it seems all of them will be in the end much simpler compared to a coherent approach,’ he said.
The merits of direct detection versus coherent have previously been considered by the Full Service Access Network (FSAN) group, a forum for building an industry consensus around new technologies in the access network. Grobe is an active participant in the group, helping to set the direction for the latest iteration of PON technology, referred to as NG-PON2.
‘In access 100 per cent of what has been installed to date is direct detection,’ he stated. ‘There has been a proposal in FSAN for NG-PON2 on a coherent access system which is not being considered anymore. The FSAN community did not follow that contender for good reason; only one party believed that it could be made cost effective. The proposal is five or six years old now... but even from today’s perspective there are still one or two key components missing.’
However, another European research project has a contrasting opinion. The €4 million COCONUT project aims to develop a simple and therefore low-cost coherent receiver for optical access networks. COCONUT is an acronym, just about, for COst-effective COhereNt Ultra-dense-WDM-PON for lambda-To-the-user access networks.
Coordinated by Ernesto Ciaramella from the Scuola Superiore Sant’Anna in Italy, there are eight project partners in total, including the Universitat Politècnica de Catalunya in Spain and Athens Information Technology Centre in Greece on the academic side, with industrial support from operator BT, manufacturers Ericsson and III-V Lab, and two high-tech small businesses, Promax Electronica in Spain and Optronics Technologies in Greece.
‘A coherent system has two key advantages,’ explained Ciaramella. ‘First, coherent detection increases the sensitivity, which in turn increase the power budget: this may be useful especially when upgrading existing plants with high losses, for example, because of high numbers of passive splitters, or longer distances from the central office. Another key relevant feature is wavelength-selectivity – optical coherent detection allows us to select one channel in an ultra-dense grid such as 6.25GHz, which would be extremely difficult and expensive to do with optical filters.’
The ultra-dense grid would enable the ‘lambda to the user’ concept, where individual wavelengths are allocated to hundreds of user on the network – users in the broadest sense could be consumers, businesses or even base stations in a cellular network. The envisioned access network architecture represents an evolution of the almost-commercial WDM-PON system that has four or eight wavelengths today to one with an extremely high number of wavelengths. (NG-PON2 is officially regarded by FSAN as a flavour of WDM-PON, according to ADVA’s Grobe). Given these features, the COCONUT approach would enable a significant increase of the capacity in access networks.
However, while existing coherent optical systems have been designed for high-speed transmission used in the metro and core network, they would not be suitable for access networks, where the cost of the equipment drives adoption of the technology. In particular, the optical receiver that sits inside the networking unit (ONU) at the end user premises is extremely cost sensitive. However, Ciaramella stressed that the COCONUT project was targeting ‘a novel type of optical coherent detection’ that uses much simpler optical components such as directly modulated distributed feedback (DFB) lasers without the need for phase locking, and that cost savings could also be made on the electronic side by using analogue rather than digital signal processing.
The three-year project has been extended to February 2016, when participants will run the final demonstration at the Scuola Superiore Sant’Anna in Pisa in front of European Commission representatives and local authorities. The project has already put on a demonstration at the Optical Network Design and Modelling (ONDM) conference in May 2015, where they showed a prototype coherent optical receiver receiving high-definition video in real time through a Gigabit Ethernet link.
Ciaramella recently presented the results of the COCONUT project to the FSAN group. ‘As you know, the FSAN group is working towards producing a PON system roadmap and it held a dedicated workshop in Atlanta to hear from specialists about the new technologies that may come about, where I was one of the invited speakers. The technical workshop and other sources of information are taken by the group to assess what requirements, applications, etc. may be possible in the future, and as such, the COCONUT presentation has been very well received by the FSAN Group,’ he said.
The COCONUT project has already achieved real-time operation of its optical coherent system at 1.25Gb/s and has experimentally proved the feasibility of its approach at 10Gb/s. ‘These last are particularly interesting to the community of operators and manufactures working in optical access networks,’ Ciaramella said. ‘Cost estimations show that these realisations could have a competitive cost level, much lower than today’s coherent systems. Of course, they are expected to be a bit more expensive than current direct-detection schemes, which is completely acceptable since they will have much higher performance.’
Flattening the core
The DISCUS project mentioned at the start of this article shares similar goals to the other two projects, but has a more radical approach. Rather than being bound by the design of existing networks, the researchers wanted to start with a clean slate. ‘The main thing about the DISCUS project is that it is an end-to-end project; we believe that you need to use an end-to-end design if you want to get the optimal solution,’ said Trinity College Dublin’s Professor Payne. ‘If you just solve the problem in the access then you are transferring that problem to the core. And if you don’t solve the core [network] problem then the energy and cost problems simply emerge in a different way. You have got to tackle all parts of the network at the same time.’
A trio of Irish academic institutes are taking part in the project – Trinity College Dublin, Tyndall National Institute and University College Cork – as well as Aston University in the UK. Collaborative industry partners include Alcatel-Lucent, Coriant, Polatis and III-V Lab in addition to operators Telecom Italia and Telefonica. The research institutes IMEC in Belgium and KTH in Sweden, and Atesio in Germany, a spin-out from the Zuse Institute Berlin (ZIB), round the total number of project partners up to 13.
The €11.7 million DISCUS project will re-architect the fibre network, to considerably reduce the amount of equipment required and simplify the network architecture using optical switching where possible, as Payne explained: ‘We adopted a different technology called long-reach PON that, compared to the current standard fibre-to-the-home, allows us to increase the number of users per network from 32 to 1,000, and increase the distance between homes and central office from 10-20km to more than 100km. This means that over 1,000 telecom local exchanges that we have today in Ireland could be consolidated into about 20 central nodes, reducing capital and operational costs, reducing power consumption, and extending the service availability also to rural areas.’
Payne explained the reasons for selecting this design. ‘It is the electronic systems, the packet processing, all of the equipment that you have to put into the network, that drives that cost in the core,’ he says. ‘One thing that can be done with the core network is that we could do a lot more in the optical domain to reduce the number of interfaces and the number of nodes that we need. So we are looking at a network with fewer nodes and far less electronics, and we think – in fact we have proven in the DISCUS project – that the way to get that is to go with a flat optical core network.’
The resulting architecture has just two components: a flat optical core which is interwoven with a mesh of light paths to eliminate packet processing across the network, and long-reach PON for the access portion. However, what really makes the DISCUS architecture work is the synergy between removing the core nodes to reduce costs and energy while using a long-range PON system to enable a larger degree of resource sharing in the access network.
With the architecture proposed by the DISCUS project, the metro network effectively disappears, as Payne explained: ‘We bypass it and use the access network to deliver all the traffic using WDM systems from the customer all the way back to the core nodes.’ However, the trade-off here is that fibre must extend over much longer distances if it is to bypass the now redundant active sites, which are replaced by simple optical amplification.
The DISCUS project team also think that their architecture could improve the economics of connecting customers in rural areas, helping to reduce the digital divide between customers in dense urban areas and those living in remote regions. In sparsely populated rural areas, conventional PON systems struggle to achieve a sufficient number of customer connections to get adequate sharing of the physical infrastructure and achieve a low cost per customer. By extending over greater distances, a long-reach PON makes it possible to aggregate more customers.
To further optimise the architecture for rural areas, the DISCUS project proposes a chain structure for amplifier nodes, and more efficient final splitter layouts to minimise the costs of drop fibre connections. A similar design using open rings can be adopted in dense urban areas to provide protection paths. An open ring or cable chain design would allow re-use of legacy ring topologies deployed during the SDH/SONET era, and also enables resilience and optical fibre monitoring techniques to be employed economically in at least part of the network.
As well as the demonstrator, the researchers also hosted a workshop at ECOC 2015, where they presented the results of cash flow modelling, which indicate that the DISCUS LR-PON system could produce a return on investment in as a little as four or five years, similar to the expectations for fibre-to-the-cabinet (FTTC) deployments. Payne stated: ‘The old fibre-to-the-cabinet solution keeps the local exchanges and the backhaul network. As we pass a certain threshold and the bandwidths increase, the long-reach PON solution – the DISCUS solution – starts to win out.’
These research projects illustrate a variety of different ways that the access network might evolve, but they can’t all come to pass. Whether coherent or direct-detect technology comes to dominate the access network of the future, and whether or not operators decide to consolidate their central offices by employing long-reach PON, all of these projects have one thing in common: they indicate that it should be possible to achieve a significant reduction in network costs and energy consumption to help sustain the growth of broadband in the future.