Manager, Communications and High Tech
Accenture
Some observers have classified Fiber to the Home (FTTH) as potentially the best choice for new networks-one that can deal with all bandwidth and service problems, and which also offers the greatest flexibility for companies on the fast track to high performance.

The actual situation is somewhat more complex than that. To make the right decisions regarding their access networks, communications providers need to understand and compare all the relevant variables: the ability of a particular access network option to handle all needed services, as well as its scalability and cost.

Is Fiber to the Home (FTTH) the best choice for a service provider's access network? Many believe so. Fiber optic cabling offers high bandwidth with a small physical size, as well as very low attenuation, especially at 1550nm.

The advantages of fiber are not as compelling, however, when cost is factored into the equation. Fiber comes with high costs for terminating and connecting, as well as for the electrical interface hardware. This means that FTTH is extremely suitable for long-haul trunk use but not necessarily for access networks.

Communications providers must take into account a number of considerations when weighing their options. Essentially there are two basic fiber physical layer access topologies based on international standards:

1. A "metro Ethernet" type solution, serving a city, town or village, which uses two dedicated fiber cores from a router to each user. One fiber is for the forward path signals (to the user) and the second is from the user. These fibers are configured in a star topology back to distribution point or hub, where a router with optical line cards is located. Each pair of fibers can be configured separately to provide between 10Mbit/s and 1Gbit/s per connection.

2. A passive optical network (PON) which uses a single fiber to serve a number of users (typically a maximum of 32). This fiber uses optical splitters and operates in a two-way mode over a single fiber core. Terminating hardware (OLTs and ONUs) is currently expensive, although predicted to become cheaper. There are a number of different flavors of PON (see Figure 1). Bandwidth is shared among all connected users. For example, with E-PON, 1Gbit/s is shared among all user connections giving a theoretical uncontended capacity of 30Mbit/s per user.

   There are other proprietary possibilities-for example, a direct fiber extension of the Hybrid Fiber Coaxial (HFC) network-but these is outside the scope of this document.

   Another issue to be considered is how much of an advantage is actually found in the high bandwidth capability of fiber. There are approximately 100 standard wavelengths allocated by the ITU for use on fiber. Using all of these, each fiber can handle in excess of 1,000Gbit/s of traffic, a large number of analog (radio frequency or RF) signals or a combination of both.

   This appears to equate to an enormous theoretical capacity, yet all is not as it seems. Because of impairments in non-dispersion corrected fiber at 1550nm (the majority of fiber in use), single wavelength speeds in excess of 2.4Gbit/s can incur major costs, because of the need to regenerate signals or use external dispersion correctors. These high speeds and the use of multiple wavelengths are only feasible on trunk and super trunk fiber links.

   For an access network, is this speed usable? Let's assume a situation with six video streams (broadcast and on-demand) at between 2.5Mbits and 7Mbits per stream (MPEG4 SDTV-HDTV), as well as one or two Voice over IP services at 64kbit/s symmetrical, together with high-speed data at 2- 6Mbit/s downstream and 1-4Mbit/s upstream. Given these assumptions, the resulting downstream/upstream requirement is between 20/2 Mbit/s and 37/4.5 Mbit/s per user, with no contention. This should be more than adequate for single subscriber use. It means that the speeds offered by fiber to the home, while impressive, may not have real value in the access network for domestic customers.

How does fiber to the home compare to an HFC network when it comes to capacity? Let's assume that an HFC network uses all channels between 108MHz (above band 2) and 750MHz (the upper design operating frequency) for digital services and that all channels are 6MHz wide and modulated at 256QAM. This means that the total capacity available is approximately 100 channels, each with a payload of 39Mbit/s (post error correction). These parameters equate to a total downstream data capacity of 3.9Gbit/s.

If these channels are carrying IP DOCSIS traffic then, depending on the segment size, companies could design their networks for any traffic profile. For example, with 390 users, each could have 10Mbit/s uncontended. However if the RF broadcast television channels are carried separately, then the total traffic requirements for on-demand television, high-speed data and VoIP are somewhat reduced. If we say that broadcast television occupies 30 channels (300 individual programs) then 70 DOCSIS channels (2.7Gbit/s) would still provide approximately 7Mbit/s uncontended per user. That means that the traffic could accommodate two standard "on-demand" television channels, high-speed data and VoIP. Even allowing for some contention, this would translate into a very high-value service offering.

Upstream capacity of HFC networks can be addressed by using different upstream segment sizes to the downstream, permitting channel re-use. Keeping segment sizing smaller makes possible a further increase in upstream capacity. Using smaller upstream segments results in a lower return path carrier-to-noise ratio, making it possible to use higher-order modulation schemes.

What about HFC network flexibility? The ratio of RF broadcast channels to DOCSIS channels can be adjusted, without great difficulty, depending on the service offerings and customer requirements. With this arrangement, and using re-segmentation to cater to variations in user profiles, companies can provide a very flexible network.

In principle there is no reason, except cost, for not re-segmenting the HFC network to a lower number of homes passed, thus reducing the number of customers that are covered per segment. This way the extra bandwidth that is available per segment will be shared with a smaller number of customers, enabling a company to offer a higher bandwidth to each one.

New technologies such as wideband DOCSIS will offer significant additional flexibility. When added to the HFC environment this allows greater bandwidth for users, should the operator so wish. In addition, this flexibility will enable better management of utilization factors, since not all customers will require the same amount of bandwidth at the same time.

The costs of the access network and customer premises equipment are by far the largest capital costs of an operator's network. A headend that costs $30 million and serves 5 million homes equates to $6 per home passed. At a 25 percent penetration, this is still only $24 per customer. The installation of a single coaxial drop cable can be four times this figure (approximately $100 per connected customer). When installing fiber to the home, the fiber drop installation can be five or six times greater than for a coaxial connection, due the cost of terminating the fibers and breaking out individual cores. The cost of a twisted pair drop cable for xDSL is slightly lower than that for a coaxial drop but not significantly; labor cost is the biggest single factor.

Taking out the cost of civil groundwork and construction-which can be 70 percent of the total network capital cost-the cost of the network build for an HFC network is approximately 50 percent of the CPE costs. Most of these costs are due to the set-top box and cable modem costs. With a fiber to the home network, the build costs are similar; however, optical fiber signals need to be converted to two way electrical signals for interfacing onto standard domestic equipment. These interfaces can add between $50 and $500 per subscriber to the cost of installing a fiber drop cable.

xDSL is an alternative technology which has a significant deployment worldwide. This uses the telephone twisted pair as a medium for the delivery of two-way, low-frequency RF data channels to the home. It is added here as another option for cost comparison purposes. With xDSL the CPE cost is low but the cost of the associated exchange hardware (DSLAMs) is still quite high.

Based on this comparison of access network options, here are a few conclusions that can be drawn:

1. Don't be misled by the speed issue. Be careful not to base decisions on a single parameter like bandwidth or speed. Instead, consider all the services to be delivered (broadcast, VoIP, etc.) and their dependence on other performance specifications.

2. Because of the cost of FTTH, a good compromise may be fiber to the curb, with a coaxial drop cable. This solution leverages the high speed and high bandwidth of the fiber and brings it close to the end user. Other types of access are more suitable for specific services. Twisted pair cable, for example, is more suited to basic phone service (POTS) without the need for expensive interfaces.

3. There is no obvious advantage in delivering broadcast channels via fiber to the home. Broadcast RF channels on an FTTH network add significant cost and complexity; a separate optical wavelength is required, as well as the use of optical amplifiers and wave division multiplexing. A single coaxial cable for the broadcast RF channels is low cost, easy to install and does not require expensive media converters. A composite cable-one that might combine fiber, coaxial cable and twisted pair-will carry specific services using different media more suited to the different services.

4. xDSL is a good option. xDSL is a good and relatively low-cost option for operators such as telecoms operators who do not have an HFC network. However, such a solution may be restricted by distance and/or speed. xDSL was designed as a means of carrying two way data signals on twisted pair cables, using an RF carrier above the voice signals. However, telephone cables were originally intended only for basic telephone service. Because of the poor performance of telephone cables above the normal voice frequency range (due to slope, cross talk and attenuation) there can be severe limitations in using this approach to provide required services.

   More recent developments in DSL (e.g., VDSL and DSL2) have addressed these shortcomings but long-drop cables still have a relatively low speed limit. High speeds (several Mbit/s) may be possible close to the exchange but this falls to less than 1 Mbit/s typically at 5 to 7kM (3 to 5 miles).

5. FTTH is an important option if cost is not a restriction. If cost is not a significant factor, then FTTH should be considered. However, remember that FTTH offers only minimal advantages over a good HFC network with multiple DOCSIS channels, which is also capable of being re-segmented.

   These marginal advantages are high bandwidth (even if not fully used), small size (always attractive) and low attenuation (which is not relevant on short access cables). Terminating and CPE costs are significantly higher with FTTH compared with standard HFC- CPE equipment. A coaxial drop cable is preferred if it will deliver all the necessary services at lower cost and with easier maintenance requirements.

6. The power costs of an HFC network are probably higher than those for FTTH and xDSL. An HFC network typically consumes one watt of power for each home passed. Over a period of one year this equates to 8,760 watt hours. If we presume a cost of 10 cents (US) per kWh, this gives us an annual cost of nearly $1 per home passed. Scaling this for the total number of homes passed can run into several million dollars per year. All of this power is dissipated as heat in the cables and active equipment, (such as amplifiers and optical nodes). Although an all-fiber network may have a much lower annual powering cost, the power needs of optical to electrical interfaces will be higher. However, this cost is usually passed on to the end user.

   Analysis of the advantages and disadvantages of fiber to the home shows that the theoretical bandwidth discussions in relation to fiber technologies are not as helpful as an understanding of what is really needed by companies in practice. HFC technology today has significant flexibility at a low cost that can support diversity of customer segments fairly easily. Although fiber offers more bandwidth theoretically, in practice HFC can be made to offer similar or even higher useful bandwidth than fiber at a cost that customers and service providers are more willing to bear.

David Woollard is a London-based manager for Accenture's Communications and High Tech operating group. He represents the UK on CENELEC TC209/WG6 which is involved with the CATV return paths. He is the inventor of the European variant of the DOCSIS standard (Euro-DOCSIS) and is involved with a number of international working groups and standards bodies.

David has an extensive career in the cable industry. He was involved with the installation of the first fiber optic link in the UK carrying television, and carried out extensive engineering on satellite television transmission (pre-broadcast). He is a Chartered Engineer, a Member of the Institution of Electrical Engineers, A European Engineer and a Fellow of the Society of Cable Telecommunication Engineers (SCTE) - UK. For the past few years he has been the Chairman of the Technical Committee of the UK- SCTE.