The advent of cable modems is not the only technological event with the potential to substantially improve bandwidth into the household. A number of schemes for substantially increasing the throughput through a telephone twisted pair cable are currently in the early phases of standardisation and in many instances development. These may be collectively described as advanced Digital Subscriber Line (DSL) technologies and if successfully pursued, promise some interesting alternatives to the Plain Old Telephone Service (POTS) we are accustomed to.

ISDN - The Path to Advanced DSL

The starting point for any discussion of DSL technologies is DSL / ISDN, or Integrated Services Digital Network. When initially conceived, ISDN was envisaged to be the digital replacement for the established POTS service. Every household was to have its analogue service replaced with ISDN, which would provide a pair of 64 kbit/s circuit switched synchronous channels (ie the 2B in 2B+D) for telephone and data services, and a single 32 or 16 kbit/s packet mode channel (the D in 2B+D) for the purposes of link management and data services. The D-channel uses the LAP-D protocol which is closely related to the established X.25 LAP-B protocol (LAP-A having been a failure by design).

ISDN was not the roaring success many expected it to be. Differences in local variants of ISDN equipment meant that interoperability became an issue across national boundaries, indeed many countries including Australia required customised versions of LAP-D management software to meet local interoperability and certification needs. Performance had also proven to be an issue, with many sites suffering arguably unacceptable quality of service. In many instances, the existing twisted pair wiring used for POTS simply could not be reused for ISDN as its electrical performance and/or length were out of the bounds of what the line protocols and line transceiver Silicon could handle.

While ISDN struggled to receive customer acceptance, voice band modems on POTS lines plugged the gap with line bit rates up to and beyond 28.8 kbit/s. The addition of some quite clever data compression techniques to the fast voiceband modem yielded under the right conditions throughput comparable to the single 64 kbit/s ISDN B channel. The attraction of a more expensive, in terms of ongoing rental costs, ISDN service was therefore debatable. As a result, ISDN has not proliferated to the scale the Telco community may have hoped for a decade ago.

The explosive growth in the Internet we have seen since the emergence of the HTTP protocol and its associated W3 service has produced a significant demand for services with substantially higher throughput than standard ISDN. It is therefore fair to say that ISDN missed the boat, in that the throughput it offers at a time of limited market penetration is not really adequate to the demands of the marketplace. Whether we see further growth in ISDN market penetration, on the world scene, will depend on how quickly its successors and competitors (eg CATV) manage to capture the customer base.

HDSL

The next step up in the DSL hierarchy is High data rate Digital Subscriber Line or HDSL, which is often described as a smarter way of transmitting the established US T.1 and European E.1 1.544 Mbit/s and 2.048 Mbit/s data rates.

The limitation in most established T.1 and E.1 circuits is imposed by the basic signal modulation technique, usually based on a variant of Alternate Mark Inversion schemes. These are hungry in terms of bandwidth, usually require a repeater to boost the signal every 1,000 or so metres, and can often cause serious crosstalk problems in large multipair cables. The T.1 and E.1 data rates fair a lot better when carried over optical fibres, but these are relatively expensive to install, in comparison with exploiting an existing wiring infrastructure which may not have paid for itself yet. Therefore the T.1 and E.1 services have been used primarily for connections between telephone switches and between exchanges and larger PABX systems. More recently, the service has been made available to customer premises, but is usually costly due the need to lay dedicated cables, be they coax, twisted pair or fibre, to the customer premises.

HDSL circumvents the limitations of older T.1/E.1 transmission techniques by using more sophisticated modulation techniques which allow it to squeeze the T.1/E.1 rate into 80 to 240 kHz, which in turn allows suitable twisted pair cable runs as long as 4,000 metres. HDSL is used primarily for inter-switch connections within telephone networks. It is expected that the service will be made available overseas for customer premises, until the newer SDSL and ADSL become available.

SDSL

The Single line Digital Subscriber Line (SDSL) service is a derivative of HDSL, designed to provide a T.1/E.1 service to customer premises over a single twisted pair of up to 3,300 metres length. Importantly, SDSL allows the concurrent use of the existing analogue POTS service over the same piece of wire. As SDSL is a symmetrical service, offering equal bandwidth in both directions, it is well suited to host-to-host network connections with similar traffic loads back and forth.

ADSL

The Assymetric Digital Subscriber Line (ADSL) service was designed to provide high data rate services directly to customer premises, unlike HDSL and SDSL which are aimed at internal network services. The assymetry in ADSL reflects two factors (the same as in cable modem technology). The first is that interference and crosstalk problems in large bundles of multipair cables make it much more difficult to send high data rates upstream. This is because the upstream signal is at its weakest the closer it get to the exchange, where it must share a large cable bundle with a large number of downstream signals which coming straight out of the exchange, are at full power levels. the second factor is that a large proportion of customer premises applications are inherently assymetrical in behaviour, where downstream traffic is much larger than upstream traffic. Video on demand and W3 browsing are good examples.

The ADSL service is designed to provide a range of transmission speeds, with speed traded off against distance. Nominal downstream rates are T.1 to 6,000 metres, E.1 to 5,000 metres, 6.312 Mbit/s (DS2) to 4,000 metres and 8.448 Mbit/s to 3,000 metres. Upstream rates range between 64 kbit/s and 640 kbit/s. The capability to concurrently support analogue POTS is retained.

The intent with ADSL is to support compressed video (MPEG or H series protocol) downstream, as well as circuit switched traffic and packet switched traffic, such as IP. Because all of these traffic types require different quality of service and hence error correction or detection protocols, the ADSL protocol is complex and this will reflect in complex modem designs.

VDSL

The Very high data rate Digital Subscriber Line (VDSL) service is the fastest and most sophisticated of the DSL family of protocols. As such it is the most immature and at the time of writing was still very much in the definition phase.

VDSL is intended to provide comparable performance to cable modems, out to customer premises, using a combination of optical fibre and twisted pair. While optical fibre is by all means the best possible medium for data transfer, most Telcos consider the economics of running fibre to customer sites en-masse to be unsupportable by the short term revenues it can generate. The compromise solution is similar to that being used by cable modem vendors, where optical fibre is used to distribute the service to a suburb, and copper twisted pair then to the customer's premises.

VDSL is conceptually similar to ADSL in that it is a fast assymetrical service running concurrently over a twisted pair with established slower services, in this instance POTS and ISDN. Where it differs is in that it runs several times faster, over correspondingly shorter distances, and will be designed to cleanly interface to an ATM network. By using ATM related protocols, it avoids much of the complexity of ADSL's application specific protocol suite.

It is envisaged that VDSL will use integer fractions of the ATM transmission rates. The nominal 155 Mbit/s stream splits into three 51.84 Mbit/s, six 25.92 Mbit/s and twelve 12.96 Mbit/s streams. VDSL is intended to support 12.96-13.8 Mbit/s up to a distance of 1,500 metres, 25.96-27.6 Mbit/s to 1,000 metres and 51.84-55.2 Mbit/s to 330 metres. Assymetrical upstream rates of 1.6-2.3 Mbit/s and 19.2 Mbit/s are envisaged, or a symmetrical rate equal to the downstream rate. It is expected that the higher assymetrical upstream rate, and the symmetrical upstream rate will only be available for short cable lengths. Early implementations of the VDSL protocols are expected to support only the lowest upstream rates.

A large part of VDSL's use will be the transmission of video services, and to this effect a forward error control (FEC) scheme will be used. Useful bandwidth will thus be somewhat lower than the nominal bandwidth. It is expected that Reed-Solomon coding similar to that in ADSL will be used.

VDSL is intended to support passive taps much like cable modems do, with active taps as an option. As a result, VDSL systems using the passive option will share a similar topology to CATV while also using similar strategies for multiplexing upstream traffic (this is indeed a generic problem with branched segmented topologies).

Before VDSL can be defined and deployed, it must cross a number of technical hurdles. One important issue is that of electromagnetic compatibility, or in simpler terms, its ability to operate without compromising radio and television services in close proximity. The other side of this equation is its ability to operate in the noisy electromagnetic environment of cable ducts and risers shared with electrical power and computer networking cables.

The second important issue is that of protocols, in that it must reconcile many of the technical issues which are currently plaguing the cable modem community, whilst still providing the ATM interface compatibility which is considered so desirable for seamless interconnection to the respective national backbones. The requirement to compete with CATV services imposed the need for passive taps, which in turn complicates many aspects of protocol design in comparison with the hub-to-terminal model of ISDN and HDSL/SDSL/ADSL.

The advent of the advanced DSL technologies promises significantly more bandwidth to the user. Whether these services can hold back the tide of the cable modem remains to be seen.