Megabit Data Networks to the Home CSE 588 May 1, 1997 Introduction The demand for megabit data networks to the home is increasing rapidly. The popularity explosion of the Internet and World Wide Web in the last few years is the major force behind this demand. Other important applications driving this demand include Home Shopping, Telecommuting, Interactive Video, and Video on Demand. There are two emerging technologies that provide a megabit data network to the home. The first is Digital Subscriber Loop (xDSL), and the second is Hybrid Fiber Cable (HFC). This paper compares and contrasts these two approaches. Background: Digital Subscriber Loop Technologies The infrastructure that xDSL modems utilize is the existing subscriber loop of telephone networks. The subscriber loop is the length of wire that connects the customer to the Telephone Company's Central Office (CO). This wire is typically unshielded twisted pair (UTP) copper wire. The characteristics of the subscriber loop installed base, varies greatly depending on how close the customer is to the CO. A lucky customer will have less than a mile of copper connecting them to the CO, while an unlucky customer could be as far away as 3-4 miles. There exist three broad classifications of telephone lines in the United States: Line lengths greater than 18,000 feet with load coils (15%) Line lengths less than 18,000 feet with no load coils (70%) Line lengths less than 12,000 feet with no load coils (15%) The load coils are used to extend the range of the physical connection. These coils will not work with xDSL, as they effectively filter out any high frequency signal. Customers with loaded lines will not be able to use xDSL technologies without rewiring their connection to the CO. The basic idea with xDSL technologies is to use the full bandwidth available over the UTP copper wire that is installed. The telephone utilizes only a small portion of the bandwidth available, using the frequency ranges of 0-4Khz. The xDSL technologies provide a couple of orders of magnitude better performance over analog modems due the fact that they use the frequencies up to 1Mhz. Some xDSL technologies go as high as 30Mhz (VDSL). The different types of available xDSL Technology are: ADSL - Asymmetric DSL RADSL - Rate Adaptive DSL VDSL - Very High Speed DSL SDSL - Synchronous DSL HDSL - High Speed DSL ADSL gets its name from the asymmetric split of the available bandwidth. Approximately 90% of the bandwidth is allocated to downstream traffic, with 10% allocated to the upstream traffic. The asymmetric split minimizes the effect that crosstalk has on the data rate, and thus maximizes the data rate. VDSL is really just ASDL, but over a much shorter wire. VDSL exploits the shorter cable lengths by utilizing more of the available frequency range of 0-30Mhz. Shorter cable lengths exhibit less crosstalk and attenuation problems, thus allowing use of a greater frequency range. The following table lists the available data rates vs. the cable length for both ADSL and VDSL. Rates vs. Distance vs. xDSL Technology 1.5Mb/sec 18,000ft (ADSL) 2.0Mb/sec 16,000ft (ADSL) 6.0Mb/sec 12,000ft (ADSL) 1.5Mb/sec 12,000ft (HDSL) 9.0Mb/sec 9,000ft (ADSL) 13.0Mb/sec 4,500ft (VDSL) 26.0Mb/sec 3,000ft (VDSL) 52.0Mb/sec 1,000ft (VDSL) RADSL is just ADSL that adapts its data rate to the particular line conditions. RADSL dynamically adapts to the particular data line it is connected to, as well as the line's environment (noise). The other problem besides crosstalk and line attenuation is noise. The worse type of noise is impulse noise. The telephone system is fairly resistant to impulse noise, so there have been very few studies on it. The few that have been done have shown that a typical impulse noise event last 1/2 millisecond and affect a narrow frequency band. SDSL splits the bandwidth equally between the upstream and downstream channels. Crosstalk between other lines is a much larger problem when the bandwidth is split symmetrically, so SDSL has lower data rates compared to SDSL. SDSL's data rates are on the order of 700Kb/sec. HDSL also has a symmetric split between upstream and downstream channels. It achieves a higher bit rate because it uses two UTP pairs. HDSL has the largest installed base of any xDSL technology currently. HDSL has been used to support T-1 lines when the Telephone Company's CO is located within 12,000 feet of the customer. HDSL has mainly been used for connections to businesses so far. Background: Hybrid Fiber Cable Technologies The infrastructure that HFC utilizes is the cable television network. Cable networks were designed with one-way broadcast communication in mind, so utilizing them in two-way data network is definitely a challenge. The current state of cable television networks is an adhoc branching structure with the root at the cable office. The first branches from the cable office are usually fiber optic lines. These first branches are called "trunks". The last branch to the home is a shared coaxial cable that typically serves anywhere from 50-200 homes. HFC gets its name from the hybridization of fiber and coaxial cable in the network. Current cable television systems use the frequency range of 42Mhz - 750Mhz. Each channel carried by the system occupies 6Mhz. A single downstream HCF channel will occupy the allotted bandwidth for one cable channel (6Mhz). The data rate that one downstream channel supports can be as high as 40Mb/sec, although it is typically 10Mb/sec. For the most part, the downstream channel does not have significant problems with noise. The upstream channel will occupy the frequency range from 5-40Mzh. The upstream channel in the cable systems face a lot of noise just like the xDSL technologies. Impulse noise also effects the upstream channel in the cable systems. The most common source of impulse noise is from home appliances such as vacuum cleaners, televisions, microwaves, etc. Beyond the impulse noise, there is also lot of noise from radio hobbyists such as HAM radio and CB radio. The data rate for the upstream channel is much smaller due to the noise in the frequencies it occupies. The upstream data rate can be as high as 3Mb/sec, although typically it is around 300Kb/sec. Key Issues and Comparison The fundamental difference between the two strategies is the access method. Cable Modems have a shared physical connection to the network, while xDSL modems have a dedicated connection. The shared physical wire that Cable Modems use, increases the complexity of their design, and can decrease performance. All customers on the same physical wire must share the available bandwidth. Without changing the existing infrastructure, the data rates a customer will achieve could be significantly lower than the rates with a xDSL Modem. A Cable Modem system with 100 simultaneous users would see a data rate of only 100Kb/sec, slower than the rate of today's 128Kb/sec ISDN modems. Hopefully the cable companies will be able to rewire the areas that have a lot of customers, so the data rates can be kept high. The customer's Cable Modems must also arbitrate for the upstream channel, because they are sharing it with up to 200 other users. The traditional media access control (MAC) protocol is carrier-sense multiple-access collision-detect (CSMA/CD). The problem with running CSMA/CD over the cable network is the length of the network. The cable network may be up to 50 miles in length. This length requires a minimum wait of almost one millisecond before a collision can be detected. In Ethernet, the maximum network length is approximately 1.5 miles. This longer wait causes many problems for the CSMA/CD protocol, including lower utilization. The simplest method to share the channel would be to allocate time slots for each customer. This method wastes a lot of bandwidth when some customers are no using it. The most plausible solution is a hybridization of both approaches. At certain "contention" times everyone will be allowed to submit their request for the next time period. The protocol used during this "contention" period would be CSMA/CD. Once the "contention" period was complete, each customer would use the time slots allotted to it in the "contention" period. The xDSL Modems on the other hand do not share their wire so they do not need any of the special features just described about. The xDSL Modems don't need a physical layer address, or MAC controls. This simplifies the physical layer of the network greatly. The number of customers in the area does not effect the bandwidth available to xDSL Modems, either. Security is affected by the cabling as well. Cable Modems need to use some sort of encryption in a higher level of the network protocol, because any customer on the same cable will receive all other's data. In contrast, a system based on xDSL a customer will only see the data that is addressed to them. For someone who wants to tap into the network illegally, the cable system would be the easier target. They would only need to be in the same neighborhood as their target, and finding the cable system's wire would be easy. Finding the telephone wire of the target is only easy near the target's location. This is usually not very inconspicuous. Once the target's wire joins the hundreds of others heading to the CO, it becomes very hard to locate. Authentication is also important in the cable system. The cable companies would like to know who is using their system, so they can charge them. They would also like to keep non-subscribers from using their system. Authentication at a higher network level solves this problem as well, but yet again adds to the complexity of the cable modem. The xDSL system does not have as great a need for authentication because each customer has its own dedicated connection. Authentication does provide protection for the customer, though. Both systems have multiple standards defined. Most of the standards are not compatible. The lowest level standardization issue for both systems is the type of modulation used on the wire. There are several competing techniques. Currently there is not a clear winner, so a customer just needs to make sure their equipment follows the same standard as their provider. This lack of one standard does not hurt either industry, as both techniques are closed systems. The simplest device a consumer would need to use either technology would be a modem, which gives the consumer one node (address) on the network. This modem would plug into the consumers PC into the system bus, which would most likely be the PCI Bus. More complex products would include a modem with an Ethernet connection to the customer's home/business LAN. The jump from the simple device to the more complex one necessitates adding a router to the device as well. These devices are expected to be significantly more expensive. Another big issue for both systems is what transport level protocol to use. Should this protocol be ATM or should it be IP? It seems as though the answer will depend on the Quality of Service (QoS) requirements of the end users. If the QoS enhancements in IPv6 turn out to be good enough for most applications, then IP will surely win. IP has a lot of mass behind it in the Internet and World Wide Web. Most of the support for ATM comes from the lack of QoS in the current version of IP, version four. Many of the planned applications for both systems need QoS to function properly, such as Interactive Video. The United States is one of few countries around the world that has a large cable television network installed. This is a further knock against HFC Modems, as the market for HFC Modems is much smaller than xDSL Modems. In comparison, there are over 800 million copper lines worldwide, which constitutes a major market advantage for the xDSL technologies. Conclusion It seems as though the economic and technical winner should be the xDSL solutions. The xDSL modems are less complex to design and manufacture due their simpler design. At most layers in the network protocol, the xDSL modem has a simpler design than the HFC modem. The only disadvantage for xDSL modems is that 15% of customers in the United States have loaded lines. The phone companies will have to reengineer these customers' lines before they will be able to use xDSL. The market for megabit data networks to the home has just begun to develop, and there are big companies backing both solutions. It will be interesting to see what happens in the coming years. References Lee Goldberg, "Cable Modems: The Journey From Hype to Hardware" Electronic Design, vol.44, no.8. pp. 65-6, 68, 70, 75-6, 78, 80. 15 April 1996. George T. Hawley, "Systems Considerations for the Use of xDSL Technology for Data Access" IEEE Communications Magazine, vol.35, no.3. pp. 56-60. March 1997. Enrique J. Hernandez-Valencia, "Architectures for Broadband Residential IP Services Over CATV Networks", IEEE Network, vol.11, no.1. pp. 36-43. Jan/Feb 1997. Marlis Humphrey, and John Freeman, "How xDSL Supports Broadband Services to the Home", IEEE Network, vol.11, no.1. pp. 14-23. Jan/Feb 1997. Leif Aarthun Ims, Dagfinn Myhre, and Borgar Torre Olsen, "Economics of Residential Broadband Access Network Technologies and Strategies", IEEE Network, vol.11, no.1. pp. 51-7. Jan/Feb 1997 Bhumip Khasnabish, "Broadband to the Home (BTTH): Architectures, Access Methods, and the Appetite for It", IEEE Network, vol.11, no.1. pp. 58-69. Jan/Feb 1997. Kim Maxwell, "Asymmetric Digital Subscriber Line: Interim Technology for the Next Forty Years" IEEE Communications Magazine, vol.34, no.10. pp. 100-6. October 1996. Gavin Young, Kevin T. Foster, and John W. Cook, "Broadband multimedia delivery over copper" Electronics & Communication Engineering Journal, vol.8, no.1. pp. 25-36. February 1996.