11.1 Introduction

Different factors have pushed the evolution of the access network from its copper-based architecture to a fiber-based one.

Besides bandwidth increase impelled by the introduction of new services (see Chapter 2, the penetration of ultra-broadband connections stresses the capacity of existing copper cables tosupport a high number of very fast xDSLs. In many cases, copper cables deployed before the diffusion of broadband access presents a non-negligible pair-to-pair interference when xDSL is used. Often, this phenomenon sets a limit to the number of connections possible in a given area, depending on the quality and length of the cables [1].

Moreover, the need of the main carriers to reduce operational cost (OPEX) also drives toward a fiber-optic-based access network. As a matter of fact, optical transmission allows a completely passive access infrastructure to be implemented; much longer spans can be realized, and even the possibility of a delayering of the peripheral part of the network appears Chapter 3

Completely replacing the access infrastructure is perhaps the most CAPEX-intensive operation that an incumbent carrier can face. Such carriers have several million terminations (e.g., in a country like Italy, the network of Telecom Italia has about 20 million terminations) and their copper network reaches almost everywhere, both in cities and in the countryside.

In this situation, it is clear that the deployment cost is the main parameter characterizing a given access solution, and a huge effort is directed toward cost reduction.

In case civil work is needed to deploy new tubes where fiber cables have to be installed, this is by far the most expensive part of the deployment of the new access infrastructure. However, this situation is not so common in North America, Europe, and the most advanced Far East countries, especially in the most populated areas where network deployment is likely to start.

Over the last 20 years, fiber deployment activity has been constant in the major cities [2]. Moreover, even when fibers are not available, tubes are already in the ground, greatly facilitating fiber deployment. For these reasons, we will not deal with civil work in this chapter; readers interested in this aspect can find a starting point in the bibliography [3,4].

Beyond the capacity delivered to the end user and the cost, a variety of other merit parameters characterize an optical access network. Due to the huge investment needed for the deployment, next generation access infrastructure has to be future proof. Thus, the access infrastructure has to accommodate changes in the service bundle offered to the end user and the related evolution of the network equipment. Nevertheless, the investment needed to deploy the new network cannot be made available in a short time.

The convergence of all the telecommunication services toward the Internet paradigm opens the possibility of unifying data transport on the same infrastructure and collecting traditional wireline residential and business users, together with mobile users. Such a convergence is possible to connect the access points of the mobile network (the base stations) to the infrastructure of the wireline. This requires a further penetration to the fiber infrastructure, besides the capability of supporting a large variety of termination equipment.

From the bandwidth management point of view, the copper network assures physical separation among signals belonging to different users in the whole access segment: each user has his own twisted pair. This is a good situation from several points of view. The privacy of the user signal is completely guaranteed because it is impossible to detect the signal belonging to a user from another user terminal. Moreover, the concept of unbundling is strictly related to this characteristic of the copper network, assuring, in this case, perfect separation among signals belonging to different carriers. For these reasons, it would be desirable to reproduce this situation as much as possible in the next generation access network.

11.2 TDMA and TDM Overlay Passive Optical Network

As we have seen in Chapter 3, both due to the need of a careful use of already deployed fibers and tubes and to reflect the broadcast nature of a few key services like IP-TV, passive optical network (PON) is an almost universal choice for the deployment of the fiber access infrastructure.

Using a PON means that the fiber is deployed in a tree topology, often called the optical distribution network (ODN)

Since part of this medium is shared among the different channels, a shared medium access strategy is needed.

While almost all the possible strategies have been proposed, TDMA is by far the most used and all the PON products on the market are based on TDMA.

Recently WDMA has also been considered and a couple of products appeared on the market using this technique.

Differently from TDM PONs that are standardized in detail by ITU-T and IEEE, no consolidated standard still exists for WDM-PON, and this is an obstacle for the development of this kind of access system. In any case, ITU-T also works on WDM-PON standards and a first set of elements have been already defined, even if we are far from a rich standard like those for TDM PONs.

11.2.3 NG-PON Project and the GPON WDM Overlay

One of the main targets of tier one carriers is to burst individual fruition of services with respect to broadcast fruition (that in the field of television means VoD with respect to Broadcast IpTV)

This is because broadcast can also be done, even with a smaller bandwidth, by means of satellites and traditional terrestrial television, and the customers are used to having the first for steadily decreasing prices, and the second for free.

Moreover, where a diffused cable TV exists, like in the United States, the IpTV is perceived by the customers as something they already have and it is difficult to obtain from it the marginal increase that the carriers need.

Finally, the importance that pair-to-pair customer activities are assuming, is something that puts value in the intrinsic bidirectional nature of the telecommunication network; thus, it can be a good opportunity for carriers.

This consideration also points the attention toward the weaker aspect of the GPON, as we have seen in the example of the last section.

Thus, a great amount of work is being undertaken in the standardization bodies (mainly in the ITU-T FSAN group) to standardize one or more versions of the so-called NG-PON (Next Generation PON)

Since the deployment of the GPON in vast areas is considered certain, all the work is based on the assumption that the new access transmission equipment, while providing a much greater capacity, exploits the same ODN deployed for GPON.

Besides this simple, but important requirement, a group of carriers in ITU-T have proposed that the entire standardization of the NG-PON be based on a much more severe set of requirements targeting a smooth and gradual transition from GPON to NG-PON. This set of requirements, that can be called Compatibility Conditions (CC), can be synthesized as follows [16]:

• Coexistence:NG-PON systems need to coexist on the same fiber with today’s GPON

  • Allows existing GPON subscribers to be individually migrated to NG-PON on an as-needed basis without disrupting other users on the PON

  • Relies on the deployment of G.984.5 compliant ONTs today and NG-PON systems using G.984.5 enhancement band wavelengths

• Digital capacity: At a minimum, the capacity of NG-PON systems should be 10 Gbit/s downstream and 2.5 Gbit/s upstream. Alternatively, a system that overlays multiple GPON systems on the same ODN is also considered an acceptable solution.

• Loss budget: At a minimum, NG-PON is expected to operate over “Class C” ODNs (30 dB loss) Class C++ (>32 dB) operation with optical pre/post optical amplification is also considered a possible solution for long reach access.

• Split ratio: A minimum of 64 ways split should be supported, but for some applications (e.g., office consolidation) a 256-way (or higher) split may be needed. In the last case, optical amplification is suggested to overcome power budget limitations.

• Reach: A 20 km physical reach through a passive ODN is required and at least a 60 km reach using optical amplification is recommended.

If the CCs greatly simplify the carrier problem in evolving from one PON generation to the next, they pose a difficult problem to the equipment vendor.

The simpler way to respect the CCs seems to be through the exploitation of WDM technology starting from the ITU-T standard allocation of the fiber bandwidth in the access area summarized in Figure 11.1.

The idea is to use the empty bandwidths to allocate, besides existing GPONs, other GPONs that differ from the standard today only for the value of the upstream and the downstream wavelengths.

An example is the use of the 1505–1545 bandwidth (i.e., the great part of the C band) to convey a certain number of “colored” GPON on the wavelengths of the DWDM standard.

The network architecture coming out from this idea is plotted in Figure 11.7.

In order to try to smooth the traditional difficulties that are always present when designing a DWDM bidirectional system, the considered bandwidth has to be divided into two sub-bands (upper and lower wavelengths) and a guard interval is maintained between the bandwidths.

Let us assume that a 200 GHz spacing is used to simplify the optical system design, and that a guard interval of 400 GHz is maintained between the upstream and the downstream optical bandwidths.

Thus, it results that we can put 10 channels in the upstream bandwidth and 10 channels in the downstream bandwidth, using exactly the same ODN and without in any way disturbing the already existing GPON.

This solution is very good for the fidelity to (CCs), that means to guarantee an easy adoption by the carriers, and this, for sure, provides a good way of upgrading the network in pace with the penetration of broadband services.

On the other hand, functionally, the capacity delivered to a single user is the same as the GPON, unless more than one ONU is allocated at the user’s premises.

Moreover, and definitely much more important, the overall cost of this solution heavily depends on the cost of the DWDM colored interfaces and on their market.

As a matter of fact, while a traditional PON has a single OLT interface for each served ODN, this version of the NG-PON has one interface for each served wavelength.

A way to alleviate the problem related to the interface cost is to adopt CWDM, whose standard grid is reported in Figure 11.8..As is evident from the figure, the water peak is unfortunately located around 1400 nm in the SSMF, almost the only fibers that are installed in the access area, which is just in the middle of the upgrade band (see Figure 11.1)

This renders the CWDM almost useless for use, since the number of frequencies that can be used is really small.

If the condition to interact with an already existing GPON in the same ODN is removed, the situation changes completely. In this case, the use of CWDM is quite effective, allowing eight colored GPONs (16 wavelengths) to be superimposed on the same ODN so creating a simple and powerful structure that takes advantage of the simplicity of the transmitter and receiver that are sold in great volumes for datacom and other applications.

However, the main problem, in term of cost of this architecture, is the need for colored sources both at the OLT and at the ONU. As a matter of fact, not only colored (and temperature stabilized in the DWDM case) sources have a much higher cost with respect to the simple sources used in standard GPON, but the management of the spare parts that have to be available for fast maintenance in case of failure is more complex, increasing the OPEX related to these equipments.

This problem can be removed in two ways. The first is to introduce a sufficiently cheap tunable laser suitable for this application.

In this case, the management of spare parts is the same as in the gray case. Today, tunable lasers, existing only for DWDM applications, are still too costly, but the situation could change.

As a matter of fact, on one side VCSELs seem to be almost ready for application in the third window, and tunable VCSELs have been obtained and tested in labs experiments [17].

On the other hand, the technology of lasers bars Chapter 5 could open the way to the realization of a CWDM tunable laser, which is an impossible task if designed with traditional tunable architectures.

The second method to solving the problem of having colored interfaces at the ONU is more on the system side.

At the OLT, 2N lasers (one for each different ODN terminating on the OLT) generate 2N wavelengths that are divided into upstream and downstream bandwidths. Downstream wavelengths are directly modulated and transmit the signal toward the end users.

Upstream wavelengths are transmitted unmodulated up to the end user. Every ONU is equipped with an identical, colorless Fabry–Perot Laser.

When an unmodulated external optical beam is injected into the cavity of an FP Laser, it will experience the internal gain from the semiconductor material. This process results in a reflected and amplified signal that contains the data modulation imparted by the pump current.

Since the injected signal can be much larger than the internally generated spontaneous noise, the majority of the optical output power will be at the wavelength of the injected signal. The normal multiwavelength spectrum of the FP is transformed into a quasi single-wavelength spectrum similar to that of a DFB laser. This narrowband output signal can then be efficiently transmitted through a WDM communication channel. The external injected wavelength is called the “locking” or “seeding” wavelength.

This mechanism is shown in the plot of Figure 11.9.

It is to be noted that since the upstream signal is amplified by the FP gain, it is possible to design the system so that it is not disturbed by small reflections of the seed arriving in the downstream direction. Last, but not least, a single AWG can be used in this system, processing the signal in two directions. This is possible with the so-called cyclical AWGs [18] that are devised to contemporary multiplex one bandwidth in one direction and demultiplex the other in the other direction.

In this way, a colorless FP is forced to emit a colored signal by the incoming wavelength containing the wanted modulation and constituting the upstream signal coming from the ONU. It can be noted that in this WDM-PON architecture also, every used wavelength is emitted by a colored source. However, all the sources are collected in the OLT, and the ONU is colorless.

Thus, the management of the spare parts and in general the system maintenance is simplified.

11.2.4 XG-PON

Another straightforward idea in preparing the network to meet a great increase in bandwidth request is simply to increase the bit rate from 2.5/2.25 Gbit/s to, for example, 10/2.5 Gbit/s.

This is the way that has been preferred for standardization by FSAN.

After finalizing the XG-PON1 basic specification, FSAN successfully carried it to ITU-T and the set of requirements was standardized as in the G 987 series.

The important choice in anticipating the standardization of the architecture relying on an higher bit-rate to achieve a greater capacity was not done arbitrarily, on the contrary, all the solutions presented by FSAN members or known in literature have been evaluated especially from the point of view of the asymmetry between the real per user bandwidth, and the XG-PON has been evaluated as the most suitable with the most common bundles of services [19].

Moreover, the design has been built on the requirements of the CCs, so that an XG-PON can be installed on an ODN where a normal GPON is running without disrupting the GPON operation.

These requirements have conditioned the choice of the spectral plan of the XG-PON, bringing us to the solution reported in Figure 11.10.

Three classes of XG-PON are foreseen even if the standardization mainly concerns the first class, which is considered the first to be deployed.

1. XG-PON1

   1. NRZ both upstream and downstream

   2. Downstream, R = 9.95328 Gbit/s

   3. Upstream, R = 2.48832 Gbit/s

   4. Maximum fiber distance: 20 km, with options of 40 km (in case with optical amplification)

   5. Split ratio up to 1:256

   6. Loss budgets: 29 and 31 dB

   7. FEC in the frame to help correct transmission both upstream (RS(248, 232) and downstream (RS(248, 216)

2. XG-PON1 Extended

   1. As XG1-PON, but with a longer reach; this solution will probably need the so-called extenders, which are amplifiers suitable for remote power feeding

3. XG-PON2

   1. Has to target a downstream capacity of 40 Gbit/s

All the main elements that are present in the GPON are reproduced and adapted for the XG-PON1; thus, the adaptation for higher levels packets is the GEM, suitably adapted to work with the different TDM frames of the XG-PON1 (e.g., it contains an FEC that was not present in the GPON frame), ranging protocol managed by the OLT to prevent collision in the splitting points, DBA both in upstream and in downstream.

On top of these characteristics, the XG-PON1 standard presents some aspects of novelty with respect to the GPON; among those a much more efficient security, with internal mechanisms for ONU and OLT registration, and a great attention to power consumption.

11.3 WDM Passive Optical Network

During the first phase of the NG-PON project, FSAN has also considered the possibility of accessing the shred medium through wavelength division multiplexing.

A simple architecture of a WDM-PON is shown in Figure 11.11..The OLT, placed in the local exchange, is constituted by an Ethernet switch managing a certain number of PONs and allowing grooming and QoS protocol management, and by a set of optical interfaces, one for each end customer. A different wavelength is associated to each end customer, so that the OLT must have WDM-capable interfaces.

In order to avoid problems with reflections from the fiber infrastructure, unidirectional transmission is generally considered. Thus, a fiber pair is needed in each branch of the PON infrastructure. A couple of athermal mux/demux is placed in the branching point, and every signal is terminated at the final user site with an ONU tuned on the user wavelength.

The wavelength plan used in this architecture depends on the adopted WDM technique. If the CWDM standard is used, generally, no more than 16 wavelengths can be adopted.

This is due to the fact that common fibers have an absorption peak around 1390 nm, due to OH ions, and as shown in Figure 11.8, where the absorption profile of an SSMF fiber (ITU-T G.652) is superimposed on the CWDM frequency standard. Assuming a sensitivity of −18 dBm of the receiving PIN, a transmitted signal power of 0 dBm and a loss of 5 dB from the mux/demux, an ODN attenuation smaller than 0.65 dB/km is needed in order to achieve a reach of 20 km. This requirement is not fulfilled by the CWDM channels that are around 1390 and 1410 nm, thereby reducing the number of useful channels to 16.

Adoption of the CWDM standard has several advantages, from the possibility to use low cost, robust transceivers, which are produced in very high volumes mainly for datacom applications, to the availability of low cost athermal mux/demux.

If a high wavelength number is to be used, DWDM is needed. In this case, it is easy to reach the number of 32 or 64 wavelengths in the third fiber transmission windows, at the expense of a higher interface cost.

One of the potential problems in deploying the WDM-PON architecture depicted in Figure 11.11, is the presence of colored ONU optical interfaces. As a matter of fact, every ONU receives and transmits a different wavelength. The management of 16 different ONUs during network deployment (in the case of CWDM-PON) can increase operational costs, due both to spare parts management and in-field maintenance.

Naturally, in this case also, the need of colored interfaces can be eliminated either by using tunable lasers or by injection locking in FP Lasers.

11.4 WDM-PON versus GPON and XG-PON Performance Comparison

At this point, it is interesting to compare the WDM-PON and GPON on one side, where WDM-PON is based on CWDM and Gigabit Ethernet (GbE) channels on each wavelength, and WDM-PON and XG-PON on the other side, where WDM-PON is based on CWDM again, but each wavelength carries a 10 GbE channel, which we will call XWDM-PON.

In this section, we will limit our analysis to an architectural and performance comparison since an economic comparison is critically dependent on commercial and industrial strategies, and it is quite impossible to do if on one side there is a commercial product and on the other, a system under prototyping.

However, even without an explicit calculation, the main point driving the cost structure of WDM-PON and GPON/XG-PON will be clear.

In order to correctly carry out the comparison, we have to concentrate our attention on a single WDM-PON architecture; here we will consider in both cases the simple architecture of Figure 11.11, where colored ONUs and unidirectional transmission are used.

• Capacity: GPON versus WDM-PON

  The capacity per user of a WDM-PON is easily evaluated: a single wavelength is dedicated to each end user. In general, a GbE signal is transmitted on each wavelength, assigning a capacity of 1.25 Gbit/s to each end user. It is worth noting that the WDM-PON has no particular advantage if part of the signal is constituted by pure broadcast (e.g., conventional IP-TV): the broadcast signal has to be replicated by the OLT on every wavelength and independently sent to each user.

  The evaluation of the GPON capacity per user is not so simple as we have seen in the dedicated section.

  Using the example we have cited there when we calculated the GPON expansion capacity, let us try to carry out the calculation also for the WDM-PON.

  Since we have to feed 400 customers through 32 ONUs placed in the basements of the buildings of the area, we use the fiber to the building (FTTB) configuration. In the case of WDM-PON, we will have 16 ONUs, thus every ONU will have twice the traffic load.

  Since the WDM-PON is symmetric, the evaluation is only for the downstream, which is the most critical direction.

  Repeating the traffic evaluation in this case, we will find that each ONU that is served by a GbE channel has to manage a traffic constituted by 500 Mbit/s of broadcast IpTV, and a total of 73.5 Mbit/s of point-to-point traffic constituted by 18 kbit/s of Ip phone (this is naturally an average value since it is lower that the base Ip phone rate), 17.5 Mbit/s of web browsing, and 56 Mbit/s due to individual video programs.

  The total amount of traffic is about 574 Mbit/s.

  We have to take into account that the efficiency of the GbE channel should not be pushed to the limit as it might risk congesting the far end switch. Let us image reaching a 68% efficiency of the GbE, which is a net capacity of about 0.81 Gbit/s.

  The free capacity per ONU is in this case 236 Mbit/s and globally 3.7 Gbit/s. If we want to evaluate how many VoD channels we can allocate, it would be misleading to divide 3.7 Gbit/s per 16 Mbit/s, since in this case it is not possible to distribute in a dynamical way, the bandwidth among different groups of users as is possible in the case of a GPON.

  The only result that can be considered is that with the WDM-PON, it is possible to open 14 VoD channels for each of the 16 ONUs. The result is that the comparison between the two architectures is quite difficult.

  If the service penetration is uniform, and corresponds to an area covered by the uniform central office from a social point of view, the WDM-PON is much more effective in delivering point-to-point services that have high activity. If the penetration of the service is very inhomogeneous, the WDM-PON advantage gets smaller and smaller.

• Capacity: XG-PON1 versus XWDM-PON

  Under a qualitative point of view, the situation is exactly the same as we described in the previous section. The relevant difference is that if such very high capacity systems have been adopted, this means that the penetration of high capacity and high peak bandwidth services are already here.

  In this case, the much higher total capacity of the XWDM-PON gets, for sure, a better result from a capacity point of view.

  However, we must not forget that this is not for free, and whatever technology solution will be found to decrease the cost of the XWDM-PON interfaces, a certain price difference should always exist between the two categories of systems.

• Security and unbundling: Both cases

  WDM-PON assigns to each user, a dedicated wavelength. This is not exactly the same as the dedicated physical carrier that is assigned to the user in the copper network, but the situation is quite similar. Due to the presence of mux/demux in the branching point, every ONU receives only its own signal, so that a user cannot gain access to signals directed to other users. Moreover, a wrong working of a single ONU cannot influence the signals of other users, thus a good degree of security is guaranteed.

  As far as unbundling is concerned, different wavelengths could be assigned to different carriers if a suitable mechanism is implemented in the OLT. This is not a physical separation exactly, because the physical layer is common among different carriers and data security is guaranteed by wavelength separation.

  In the case of GPON, each user receives the signal directed to all the users so that it is possible to gain access to the signal directed to another user by simply working on the ONU. Moreover, if the ranging protocol of a single ONU does not work, the wrongly synchronized signal interferes with the signals from other ONUs in the branching point damaging other users.

  As far as unbundling is concerned, it is practically impossible using GPON, unless virtual unbundling is considered. In this case, an amount of bandwidth in the shared downstream channel is assigned to a competing carrier under a defined SLA.

• Fiber utilization: Both cases

  Due to the fact that bidirectional transmission is used in the GPON case, while in our example WDM-PON unidirectional transmission is adopted, the fiber infrastructure is clearly better exploited by the GPON. As shown in Section 11.3, unidirectional transmission can be used in WDM-PON, but comes at some cost. As a matter of fact, in order to achieve a sufficient branching ratio, DWDM is needed, for example, 32 channels with a channel spacing of 100 GHz. A possible design can individuate two different bandwidths to be used upstream and downstream. They can be separated by a gap of about 800 GHz to prevent destructive interference from reflections. In this way, a branching ratio of 16 can be achieved. However, 100 GHz channel spacing requires cooled DFB lasers to be used both in the ONU and in the OLT. This fact, besides the greater cost of the mux/demux, clearly influences the cost of the system. In order to cope with this problem, the use of a WDM comb derived from the filtering of a single broadband noise source [20].has been proposed, but it is not clear yet if real cost advantage is achieved.

• Optical link budget: WDM-PON versus GPON

  The transmission scheme of WDM-PON is quite simple: attenuation is given by the loss of the mux/demux and by fiber propagation (taking into account connectors, patch panels, and other signal losing elements that can be present in the access infrastructure)

  Focalizing on CWDM-PON, standard CWDM optics can assure a transmitted power of 0 dBm, while the receiver sensitivity depends on the used detector. Using a PIN, the sensitivity at 1.25 Gbit/s (assuming that a GbE is transmitted) can be about.−18 dBm. This number increases to about.−28 dBm using an APD.

  Assuming the use of 16 wavelengths, the worst channel experiences an attenuation of about 0.9 dB km. Adding the other contributions and a system margin, we can assume a loss of 1.3 dB/km. Inserting these numbers in a very simple link budget evaluation, we obtain 10 km using a PIN at the receiver and about 22 km with an APD.

  In the case of GPON, the link budgets are standardized in different GPON classes and are reported in the GPON description section. In any case, classes B and C power budgets are of the same order of magnitude as the budgets evaluated in the case of the WDM-PON.

  There is no relevant difference from this point of view, as it was intuitive from the fact that fiber propagation and splitting losses are not strongly dependent on the signal wavelength in the band relevant for fiber access.

• Optical link budget: XWDM-PON versus XG-PON1

  Standardization prescribes for the XG-PON1, a link budget sufficient to have a reach compatible with GPON B+ and GPON C. Taking into account the slightly higher losses experimented by XG-PON1 wavelengths with respect to the GPON wavelength and a set of other differences in the transmission line between GPON and XG-PON1, a budget of 29 and 31 dB, depending on the comparison with GPON B+ or GPON C, results.

  A similar prescription does not exist for XWDM-PON, and a real power budget will be available only after that the first industrial product is produced in volumes.

  Presently, Ethernet components and equipment implementing 10 GbE are designed for top class Ethernet applications, if not for the carrier core network as backhauling of the core machines. Thus, these systems can provide good transmission performances. For example, a 10GbE XFP designed for the switches placed at the core of the private network of a large corporate or for the public Ethernet switches of the core metro network are equipped with APD photodiodes and narrow line width DFB lasers, and generally assures a power budget in between.−30 and.−34 dBm.

  It is to be noted that an XWDM-PON equipped with interfaces that guarantees a power budget of −32 dBm can cover a reach of about 25 km. This is not for free naturally, since we have to remember that different from the XG-PON that has a single transmitter into the OLT, an XWDM-PON has a maximum of 16 transmitters of different colors.

11.5 Enabling Technologies for Gbit/s Capacity Access

The problem that optical and electronics technology has to solve in the case of next generation access is to allow gigabit class equipment to be realized and deployed in huge volumes at a cost comparable with today’s megabit class equipment (e.g., xDSL DSLAMs) The first step to individuate the key technology challenges is to assess the cost structure of the new access network.

In Figure 11.12,.the cost structure of an FTTH and an FTTB network is shown assuming deployment in the years 2004 and 2010 [21,22]. In both cases, a GPON installation is assumed, exploiting B class GPON with a 32 splitting ratio. However, the results would be not so different if WDM-PON would be considered, but for a greater impact of the OLT equipment, that is almost comparable with the ONU.

In order to evaluate the infrastructure costs, a medium European city is assumed as reference and the OPEX cost is evaluated for a period of 10 years. As assumed everywhere, civil work is not included, and it is assumed that the fibers are to be deployed in already existing tubes.

From the figure, it is clear that the greater part of the CAPEX is due to the optical network termination: the building equipment in the FTTB case and the ONU in the FTTH case. On the other hand, the impact of the VDSL CPE cost in the FTTB case has been quite reduced passing from year 2004 to year 2008. This last figure is the effect of a great cost reduction of xDSL chips in the last few years.

The cost of a WDM-PON ONU is almost always determined by the optical interface, while in the case of the GPON ONU, two elements determine the cost of the equipment: the optical interface and the GPON chip, including the GPON physical frame processor, the ranging protocol processor, and the Ethernet switch, all in the same chip.

The importance of the ONU would be even more evident if we were to make a similar plot for XG-PON1 and XWDM-PON. Here, the challenge of reducing, essentially to a commodity, a 10 Gbit/s interface supporting a system power budget of about 30 dBm, is really formidable.

Another key point in the case of XG-PON1 is the realization of a suitable burst mode receiver both for the ONU and the OLT problem that is an example of the more general problem of the 10 Gbit/s electronics for the access receivers.

From this brief discussion, it is evident that optical interface technology is a key to the deployment of next generation access networks, and we will be devoting an important part of the discussion to these components.