|Issue:||Latin America II 1996|
|Topic:||Passive Optical Networks|
|Author:||John Barber & Ray Guyon|
|Organisation:||British Telecom Access Engineering|
Evaluating the use of Passive Optical Networks and comparing them to other options for Fibre in the Loop.
Like most telcos around the world, British Telecom is in a regulated environment. The present broadband market of broadcast TV is not directly available, and the market for more sophisticated services is taking time to develop. Nevertheless BT has realised that in the not too distant future it will need a broadband capable network if it is to continue to prosper. However, of more immediate importance to continued competitiveness is the need to reduce operating costs and increase quality for existing services, and it is the potential in these areas which has led BT toward an integrated full service fibre network. In pilot deployments throughout the country Passive Optical Network (PON) architecture is now taking the benefits of fibre to small/medium business sites, and to street cabinet locations providing residential service. In this article John Barber and Ray Guyon of BT Access Engineering assess the development of the use of PON in the United Kingdom where the main objectives are a reduction in operating costs, an increase of quality in the short term, and also the provision of a fibre infrastructure which will allow BT to support broadband services when the demand and opportunity arises. “The technical challenges of PON” When considered against other options for fibre in the loop, PON presents a number of technical challenges. The principle of PON rests on the sharing of opto-electronics and fibre among a number of customers, in order to make fibre access more economic for smaller customers than is possible with dedicated point to point systems. A single fibre from exchange equipment carries all the traffic for all the customers served by the PON. This single fibre is fanned out into many fibres by passive optical splitters in the external network so that the entire signal from the exchange end is received at every customer unit albeit at very much reduced power. The transmission protocol used tells each customer unit where to look in the data stream to recover the data intended for that customer. In the ‘upstream’ direction the protocol tells each customer unit when to transmit in order to ensure that data interleaves as the fibres converge to give a proper datastream returning to the exchange end. The key challenges to a PON system designer are therefore: achieving sufficient optical power budget to allow a high splitting ratio, and implementing a transmission protocol to manage the point to multi-point architecture. Additionally, the cost of the customer-end opto-electronics must be very tightly controlled if the PON is to serve small customers economically, and careful consideration must be given to the power consumption (and mains fail backup arrangements) for the customer unit, since this power is supplied by the customer. The effects of these challenges are addressed in more depth later, but an early conclusion significant to the scope of this paper is that it is clear that PON systems will not be practical or economic for Fibre To The Home in the short term or on the strength of current services available to BT. It is not yet clear whether FTTH PON is a preferred technology for delivering residential broadband services of the future, though BT is exploring the issues with its Interactive Multimedia Services market trial which uses an FTTH PON alongside other technical options. However, it is certain that economic PON systems are available now for applications where lower split and larger customer units are appropriate, and where remote end powering is less difficult. BT is confident that such systems can bring considerable benefits in terms of reduced operating costs and increased quality of service to the small/medium business customer market. These benefits can also be extended to the residential market via customer units in street cabinets. TECHNICAL AND ECONOMIC CONSIDERATIONS AND COMPARISONS Multi-service capability Today’s copper Access Network is an effective multi-service bearer. It supports all narrowband services be they switched, leased, analogue or digital. Each service being characterised by the terminal equipment installed at the customer’s premises and by the features provided by the switching platform. The copper network’s ability to support frequencies down to and including DC has given rise to a plethora of signalling systems that has resulted in almost as many narrowband service interfaces as there are narrowband services. With narrowband interfaces distributed from the serving site over the copper network, that network’s flexibility ensures that the serving site based equipment is well filled, thereby controlling capital costs. With the introduction of fibre based systems and the resultant migration of the narrowband line cards out towards the customer, the ability to connect any line card to any customer’s circuit is reduced to a very restricted number of customer connections. This limitation greatly increases the risk that the line card cannot be used. This problem is most prevalent with the minority services that, typically, have low and patchy penetration, and is compounded by the service specific nature of the primary multiplexers for minority services, causing a consequential waste of unused capacity when they are moved nearer the customer. These risks illustrate that in order to be an economic alternative to copper (a prerequisite for deployment in advance of demand for broadband), it is necessary for any prospective fibre access system to have a truly integrated multi-service capability which contains the costs, and power and space requirements, of the multiplexing common equipment. This integrated multi-service requirement means that not only does the delivery mechanism have to support a variety of service line cards at the customer end, but also that it should interface to both the switch and leased service platforms at the serving site. There are three key areas in which PONs have a significant economic attraction over other options for multi-service fibre access. They are the areas of: resource sharing, which spreads the costs of system elements across many customers; planning flexibility, which de-risks the future cost implications of a planner’s decisions; and grooming and consolidation of the various service types to allow good fill, and therefore low per-channel cost of ports into the switch network. Resource Sharing PONs are characterised by their use of passive optical splitters to achieve a point to multi-point architecture. This allows a single transmission system at the serving exchange site to support many customers (or many street cabinets) without the need for any active equipment in between. This resource sharing brings obvious benefits of the sharing of costs of exchange end equipment and fibre between many customers, but more significantly and less well understood, it brings much better ulitisation of system bandwidth than is achieved with a non-sharing (point to point) architecture. The ability to dynamically allocate the system bandwidth to the points of the network that actually require capacity enables the overall system bandwidth and equipment dimensions (and hence equipment costs) to be minimised. Resource sharing is of reduced significance when serving very large sites, where there is sufficient demand to justify dedicated equipment, and so for such applications simpler point to point architectures can be equally economic. ‘Active star’ architectures can also achieve resource sharing for small customer sites, by using a single transmission system from the exchange to a remote active node nearer the customer. However, such architectures introduce significant extra transmission and switching components, the cost and maintenance liability of which will outweigh the benefits of the degree of sharing provided. Flexibility When planning point to point systems to serve a geographical area of customer demand, the planner must decide which of a range of system capacities (typically 2, 8 or 34 Mbit/s) to dedicate to a particular business customer or street cabinet location. Given the high cost of failure if the planner underestimates the demand forecast for the location in question the practical policy is always to allow significant spare capacity. Thus, particularly against smaller customers, low average system fill can be expected with high per-channel costs. When planning a PON system onto the same area, the planner can aggregate the demand over the total number of business customers and street cabinets to be served, and need only apply a safety margin to this total demand. Since as many sites will turn out below forecast as above, this safety margin can be quite small thanks to the averaging effect across the population of sites served. As demand develops, capacity can be allocated to the sites that need it, and not to those that do not. Thus high system fill can be expected, with low per-channel cost. This planning flexibility is very attractive in taking the risk out of forecasting normal demand growth, but becomes invaluable to an operator in a highly competitive environment where he must be ready to meet new customer requirements instantly, but might equally lose the customer altogether to a competitor. It is of particular competitive advantage in the area of rapid and low cost provision of ‘wideband’ (2Mbit/s) services. Without the cost of providing speculative capacity to all customers who might need it, a network operator using PON is able to satisfy demand for additional wideband service simply by remote configuration of bandwidth to that customer and the insertion of a 2Mbit/s line card in the customer’s ONU. It should be noted that this flexibility benefit is of maximum relevance when serving smaller sites. When serving customer sites which are certain to demand very high capacities, especially for 2Mbit/s services, the risk of deploying a high band-width point to point system, probably SOH, is small and so the PON approach is less attractive at such sites. Grooming and Consolidation Any fibre access technology which is a serious candidate for high volume deployment will use 2 Mbit/s interfaces into the operator’s switch network. Channel bank interfaces to the switch analogue line cards will be too costly, require too much accommodation, and do not remove the maintenance and operating cost liability of the copper Main Distribution Frame. As we have seen above, lowest costs for any technology solution are achieved when the utilisation, or the fill, of the system is high i.e. when most or all of its capacity is being used to carry traffic. This also applies to the ports into the switch network, where the cost per connection is simply the cost of the port divided by the number of connections on that port. Though future V5 interfaces and associated service platform integration promise the ability to support all service types on the same port, the practical solutin for realistic near term deployment is that each service type must be delivered to a different port type or different platform. For BT, this means that POTS, ISDN BRA and ISDN PRA, must be delivered to different port types on the PSTN switch, narrow-band digital and analogue leased lines to ports into a managed leased services platform, and wideband (2Mbit/s) leased lines to ports into a core transmission platform. A point to point system dedicated to an individual customer who required a range of the above service types would need to be provided with a dedicated port of the correct type into each platform. Since the customer is likely to want only small numbers of channels of some of the service types, these ports can be expected to be badly filled and hence to contribute significant per-channel costs. A PON system naturally consolidates all the demand for any particular service type across the whole geographic area it serves onto a single transmission element at the serving exchange site. With the timeslot flexibility resident in that element it can be expected to well fill its ports into the various service platforms. Further, if the exchange end equipment is designed to support many PONs, with a full switching capability between any channel on any PON and any timeslot in any port, then the effect is greatly amplified, yielding near 100% fill for most port types. The benefits of the wide geographic coverage of PONs become very much less significant when serving very large sites because it is possible to achieve economic port fill by exchange end grooming and consolidation features alone. In particular, SDH networks (if designed with 64kbit/s switching capacity) are able to switch any customer channel to any switch port timeslot. Thus, at very large sites, simpler point to point architectures can often be a preferred solution. Technical risks with PON There are some areas where the architecture of PON systems brings inherent technical risks which must be overcome by careful system design. Of most concern to an operator considering use of PON are: security, which potentially, could be an issue because of the downstream broadcast nature of the transmission system; and delay, introduced through the buffering required to interleave the upstream signals from the many remote ends. Security Despite the downstream broadcast nature of PON based systems, the level of security is very high given the use of optics and the extremely complex proprietary digital coding techniques employed. The high levels of integration between the decoder elements and the identification algorithms of the ONU have resulted in equipment implementations that will, in practice, prevent eavesdropping and masquerading by other ONUs. Features built into the Network Management can prevent the commissioning of rogue ONUs through the use of specific time windows and a related unique ONU identifier. Unauthorised network activity such as tampering with a ONU or the connection to the PON of an unexpected ONU will result in the creation of an event report to the operations centre. Delay Round trip delay requirements are met on the early systems as the buffering delays associated with TDMA protocols (the optimum choice for predominately narrow-band applications) are relatively short. Future use of cell based protocols will need to address the longer delays inherent in such protocols. An approach might be to partially fill the cells on voice services thereby maintaining acceptable buffering delays. An alternative approach might be to use a mixture of Time Division Multiplexing for narrowband services and cell based protocols for packet based services and other broadband services. Rationalisation of serving sites and the resultant longer reach Access systems required may force a redistribution of the delay budget in favour of Access. Summary The common theme of these considerations is that PON offers clear economic benefits over other options at smaller customer sites. Because of this PON is able to take fibre access to a greatly increased number of customers than are other architectures and thus benefits of operating cost reductions and quality of service improvement, which may be common to all fibre access, can be more extensively enjoyed from a deployment of PON than from deployment of the other options. However, PON is not the most efficient technical solution at very large sites, where high capacity dedicated fibre links make more economic sense. The size break for the changeover is not clear and needs real deployment experience, but is probably around the point at which a PON would use all of its capacity serving only one or two customers. Dimensioning In order to achieve the full benefit of the PON, careful thought must be given to system dimensioning. The first parameter that can be used to characterise a PON based system is the split ratio of the PON. This split ratio is determined principally by the optical budget available at the system bit rate. Each doubling of the split ratio nominally takes 3dB (in practice nearer 4dB) off the optical budget, the remainder being used to accommodate splice and optical fibre losses, ageing and maintenance margins. The optical budget available principally being a function of the optical safety limits employed (this dictates the maximum power that can be launched into the fibre) and the sensitivity of the receiver at the chosen bit rate. 32 way split, for initial PON deployment represents a sensible balance between technical difficulty, cost and maximum geographic coverage. The second parameter used to characterise a PON based system is the PON bit rate. The bit rate selected needs to be able to accommodate at least the traffic (plus an allowance for system marshalling and maintenance information) that is expected to result from the geographic area addressed by the PON. The level of traffic generated will be influenced by the target market segments and the services to be delivered. The bit rate selected for the PON transmission is also limited by the cost of the optical devices. Where the initial deployment strategy is based on existing narrowband services a bit rate of 20 to 50 Mbit/s provides sufficient capacity for the PON catchment area. This assumes that the few large business sites in the area are served by direct fibre. Two other parameters that need to be considered when specifying the system dimensions are the number of supportable ONUs on a given PON (this will seldom need to be the same as the number of optical ends, given the practical limitations governing the siting of splitters and fibre joints) and the capacity supportable by each ONU. Deployment modelling indicates that, on average, only 12 to 14 optical ends of a 32 way PON will be correctly sited to meet the target market of appropriate business sites. Allowing for some level of street cabinet penetration for smaller customers leads to a conclusion that a PON supporting at least 20 ONUs will meet most needs. The distribution of customer site sizes shows that the vast majority of sites require less than 30 lines. In the interests of limiting the number of ONU variants and containing the size and potential power consumption of the ONU, 30 line ONUs represent a practical choice. The ‘spare’ optical ends will allow for two or more ONUs to be sited at larger customers if required. The final dimension considered here is the size of the OLT. Very large OLTs will maximise grooming and consolidation benefits, but will incur significant cost in their switching capability and will be under-utilised at exchange sites serving smaller numbers of customers. Deployment modelling and consideration of technology costs indicates that an OLT able to support around 4 PONs is an appropriate compromise. Planning models based on a system with the key dimensioning parameters set as discussed above indicate that in most cases all capacity limits are reached approximately simultaneously. That is, neither planning ‘bottlenecks’ nor unusable capacity are encountered to any significant degree. This shows that overall the system dimensions are proportionate and accommodate the variability of customer densities and service penetrations without suffering any major penalties on the cost per line provided. Operations and Maintenance In principle any fibre/electronics system offers inherent O&M opportunities. The majority of the network can be alarmed and tested remotely, giving extremely good fault location accuracy and a high probability of reacting to a fault before it becomes a customer problem. Remote configuration can reduce most service provision operations to card fit only, with many requiring only keystrokes at a terminal. These abilities can offer substantial benefits over the largely unmanaged copper network, but to make full use of the inherent capability of fibre/electronics requires very careful attention to network management architecture, in particular to integration with existing Operational Support Systems (OSS). Because the cost of software development to achieve integration with existing OSS is high, a generic (i.e. not technology specific) approach is important. This allows the investment in network management to pay back against all the technologies deployed in the network. The TMN (Telecommunications Managed Network) architecture is the preferred approach largely because it is based on standards which reduce the development specific to a particular network operator. However, even with these generic approaches the cost of development of network management features on any fibre system is high requiring the potential for significant volume deployment in justification. The economic advantages of PON at smaller sites discussed above give a PON a broader applicability than systems more optimised for large sites only, and therefore the potential volume base to support development of an Element Manager to cover the main TMN functions (configuration, event, test, etc..), and development of TMN ‘Q’ interfaces to the network operator’s OSS. Therefore, given that an operator is moving ahead with evolution of network management towards a TMN architecture, the potential exists for all PON management functions to be supported via generic OSS which interface to existing customer facing units and systems. However, realistic progress expectations with network management developments suggest that prioritisation is necessary to maximise early payback from investments. The appropriate areas to tackle first are ‘test’ and ‘event’, since it is in these areas that the most benefit is accrued from integration with the existing and very ‘real-time’ activities of maintaining customer service. Event reports can be generated in the PON Element Manager so that the alarms raised by faulty equipment identify not only the equipment affected and its location, but also the reference numbers (e.g. directory number) of the customer lines affected. These reports can be passed over a TMN a interface from the element manager to an OSS handling event report, which correlates with reports from other element managers and alerts both repair organisations and existing customer facing units via their existing support systems in a familiar and recognisable way. In this way a high proportion of faults will be known and under repair before the customer reports them, and a significant proportion can be expected to be repaired before the customer is inconvenienced. For those faults where the customer complains before the reception unit has been made aware, test facilities can be made available which are driven via the reception unit’s standard support system in a standard way. Thus the receptionist need not be aware that the line in question is supported by PON, and follows unchanged procedures. This great operational efficiency benefit can be achieved by providing an OSS for test management which takes requests for ‘line tests’ from the existing reception unit’s support system, and converts them into interactions over the TMN Q interface to the PON element manager, which operates test features in the PON equipment. Since service provision activities can generally be planned to a greater degree than can reactions to alarms or customer fault reports, integration of the element manager with an OSS handling configuration is less critical to early efficiency. This is fortunate, since the technical complexity associated with configuration over the TMN Q interface is much higher than for test and event. Nevertheless, it is highly important to operating cost reduction, and increasing quality and responsiveness that equipment and services can be configured remotely. A practical approach is to use a terminal interface directly on the PON element manager for configuration functions and this can be designed with a screen style familiar to staff in the existing operational units which carry out service provision activities. The PON element manager can be provided with many sophisticated features to support remote configuration. Equipment and services can be ‘pre-built’ in the element manager database when a customer order is taken, to be brought into service automatically when the element manager detects that the expected equipment has been fitted. The element manager can return to the user a ‘shopping list’ of the equipment items which need to be installed in order to carry out the ‘pre-build’ entered. Various capacity monitoring features can alert planners to the need to augment parts of the PON network. This combination of maintenance and configuration features contribute to an opportunity for substantial de-skilling and reduction of workforce, bringing significant operating cost reductions. In summary, although any fibre/electronics system can In principle enjoy the operations and maintenance features discussed above, the economics of PON down to smaller customer sites give it the volume potential to support the software development necessary to integrate with existing OSS. Such integration is highly important to translating the theoretical O&M advantages of fibre into real operating cost reductions, and real quality of service improvements, when compared to an operator’s already finely tuned processes for the copper network. STATUS OF BT TRIALS The Bishop’s Stortford Trials BT’s first trials of PON networks were conducted during the Local Loop Optical Fibre Trials (LLOFT) in Bishop’s Stortford, a medium sized town about 30 miles from London. These trials explored narrowband PON for telephony services (Fibre To The Home and Fibre To The Kerb) alongside broadband PON and broadband active star architectures for TV services. BT was granted a special licence to allow it to explore TV services for these trials only. Let’s look at the key points of the trial. The trials began in 1988 and ran for five years with the project closing successfully in March 1993. The objectives were related to confirmation technical feasibility of fibre to small business and residential customers, by comparing the various architectures and learning about operational issues in a real environment. In 1990 there was a re-focusing to concentrate on understanding key specification issues for PON. This shift of emphasis was required as it was becoming clear that BT could not enter the TV market in the near term, but would need a broadband capable network which would be efficient initially for existing telecommunications services. It was clear that the narrowband PON was the only one of the architectures under study which had potential against these requirements, and BT was beginning to move toward serious procurement and needed well researched specifications. The key achievements of the trials were: demonstration of the technical feasibility of fibre to small customers, with PON systems in particular being shown to be practical under real operating conditions; and delivery of key inputs on a number of important specification issues, regarding both transmission systems and optical outdoor plant. The trials also allowed accurate cost estimates for future fibre deployments to be made, advanced the understanding of UK industry as potential suppliers, and served as a test bed for some early interactive video service initiatives. Current Pilots From the outputs of the LLOFT trials, and from many in depth technical and strategic studies, BT became sufficiently confident that PON systems would play a significant role in the evolution of its access network to place contracts in 1993 for supply of equipment intended for volume deployment. These supply contracts involve operational pilots which are in progress now, followed by ‘business as usual’ deployment, subject to success in the pilots. This approach is significantly different from that of other telcos who may be trialling fibre systems in volume to gain experience and understand issues. BT’s intention is that the systems in pilot now will be deployed as a standard planning option with little further technical development, and will deliver operating cost benefits when compared to existing platforms. This confidence is possible thanks to the LLOFT exercise above, and to the level of study and detail which went into the procurement specifications. This has ensured that aspects such as O&M requirements (including the critical network management interfacing requirements), and a fully deployable service interface feature set, were covered adequately from day one. Additionally the thorough economic studies, which drove the demanding dimensioning and flexibility requirements, give confidence that the present system architectures will prove cost effective in widespread deployment. It is clear at this stage that the biggest challenge to be addressed by the pilots will be the integration of operational processes to achieve the key low operating costs. Even with the carefully specified and highly sophisticated network management architecture, there are still some procedures which will need to be tuned to be as efficient with PON as they are for the existing copper network. There seems little serious doubt remaining over whether BT can run PONs in its access network and so the challenge over the next few months will be to prove that it can be done at equal or lower cost than the already highly tuned processes for copper. If PON can enter at parity with copper, then there can be good confidence that operating costs will tumble with experience and further tuning, thanks to the inherent O&M advantages discussed above. The PON products in the current pilots The PON may be planned with up to 32 way split and a maximum reach of 18km is possible. The fibre network is duplicated from the serving exchange site down to the first splitter point. An OLT at the exchange site terminates four PONs and provides full switching capability between any customer line presentation at any ONU, and any timeslot in any of 32 x 2Mbit/s ports to the BT switch and core network platforms. The ONU is 32 line capacity, plus a capability to support 2Mbit/s services in addition. Two variants of the ONU are used, one for indoor installation at business sites (and apartment blocks) and one in a street cabinet to serve residential and very small businesses. The service range supported is comprehensive, since it is intended wherever possible for the PON to be the single delivery mechanism into the customer’s premises. The range includes: PSTN (with business variants, 001 etc.), ISDN BRA, ISDN PRA (1421 and BT’s DASS), analogue private circuits, digital private circuits (BT’s KiloStream range up to 64kbit/s and BT’s MegaStream at 2Mbit/s). For network management, BT has specified an element manager to be supplied with the transmission equipment which handles all management requirements and provides a TMN ‘Q’ interface to BT’s own operational support systems. This is technically a highly demanding area and in order to secure best early benefit within reasonable development risk and timescales the first implementation will use the Q interface for ‘test’ and ‘event’ features, i.e. those relevant to efficient maintenance of equipment and services already configured but a direct terminal interface to the element manager for configuration and assignment purposes. One possible drawback of this approach could be the proliferation of terminals in operations offices or the non-availability of the correct terminal at all offices who might need access. To avoid these risks, the terminal interface specified is text based conforming to a BT standard style, and is delivered to users on the LAN used by BT’s current operational systems. This mean that any operations office can have access, and the method of driving the terminal will be familiar to users of BT’s current systems. Beyond the element manager, BT’s operational support systems for test and event, the ‘Test Access Controller’ and ‘Access Fault Manager’, themselves provide terminal interfaces for users as well as links to the support systems used by customer facing offices, and to other systems for fault correlation. Note that the TAC and AFM also provide the interface to element managers of other technologies besides PON. Future upgrade opportunities The final area to address in this paper is to take a look at how the PON systems in pilot now might develop to meet the anticipated customer needs for evolving services and increased bandwidth. From a long term planning perspective the provision of fibre out to business sites or to street cabinets represents an enduring network capability and asset. The deployment of upgraded electronics and possibly the extension of the fibre network through further splits could enable the provision of higher bit rate services to all customers be they business or residentia