NWA-PCUG Newsletter Article
Digital Subscriber Line (DSL)
by John M. Hunt, July 2000
(click to email author)

DSL (Digital Subscriber Line) is a relatively new method of achieving high speed internet access using the existing twisted copper pair telephone lines which currently provide conventional POTS (Plain Old Telephone Service) from the subscribers premises to the central office. A number of variants of DSL exist, a few of which are ADSL, IDSL, SDSL and DSL. The one of greatest interest locally is a slightly reduced performance version of ADSL called DSL- lite, which is now being offered in NW Arkansas by Southwestern Bell for $39.95 per month. DSL is a service that is functionally comparable (but technologically totally different from) cable Internet service offered by the entertainment TV cable providers such as Cox Communications (TCA) in this area. Much debate continues regarding the relative merits of the two services, such debate necessarily being largely dependent upon the specific offerings of the DSL and cable providers in a given area at a given time. High-speed data service over existing lines is possible because of the electrical characteristics of the copper line, which, if the line is not too long, permit useful signals to be transmitted over a band of frequencies as high as one megahertz. Conventional voice used only the frequencies below 4000 hz.

From a connection standpoint DSL is totally independent of the dialup telephone system and its immense digital switched connection infrastructure. DSL service is connected full time to a single digital information source, which the DSL provider offers in the basic connection package offer. This source is entirely analogous to existing Internet service providers such as IPA, TCA, AOL, etc., and provides the link to Internet web, email, and usenet services. In some areas but not locally to the best of my knowledge one of the existing services may be specified as Internet provider of choice for a new DSL installation.

DSL is still quite new; a year ago there were very few installations in the US. At the end of the first quarter of the year 2000 there were about 900,000 installations in North America. The Belgian company Alcatel, a leading manufacturer of DSL hardware, is now shipping central office equipment (called DSLAMs) at a rate sufficient to support four million new subscribers per year. DSL Technology is still emerging, but the level embodied by the current SWBell offering is relatively, mature, being subject to several International Telecommunications Union standards.

Surprisingly enough, DSL operates over your existing telephone line, sharing the single twisted pair for both POTS and DSL. Equally surprising, the voice telephone service can be used simultaneously with Internet activity with no adverse impact on either service. Three fourths of the current residential DSL installations are provided by the various local telephone companies, called ILECs in SDL jargon. A number of independent service suppliers called CLECs also exist (don't try to determine or remember what the individual letters pertain to-the SDL world and associated networking is filled with these acronyms). The CLECs offer the same service as do the ILECs, but use the telco subscriber copper pairs, as well as existing telco infrastructure for mass digital data distribution over fiber optics (invisible to the subscriber). They also rent space in the telco central offices for their DSLAMs, which are tied to the telco subscriber lines. Needless to say, they pay a substantial fee to the phone companies for use of their copper lines and other services. In some areas there are several CLECs all selling service over the lines of a single local telco, usually in direct competition with the service also offered directly by the same telco (ILEC). To the best of the writers knowledge there are no CLECs operating locally.

Unfortunately current technology restricts DSL-lite to subscribers within about 18000 feet of the telco central office to which the copper twisted pair line runs. SWBell is currently investing a very large amount of money in a program called Project Pronto which is addressing this problem by running fiber optics cables to outposts located nearer the distant subscribers. These outposts are little more than simple enclosures which shelter the DSLAMS needed for DSL service. Project Pronto is scheduled for completion at the end of the year 2001; hopefully individual elements will be activated as they are installed. Aside from Project Pronto there is no new technological innovation on the immediate horizon that offers a solution for the over 18000 feet crowd, at least at affordable prices.

The download speed that one can achieve from DSL is subject to wide variations. SWBell advertises 384 kilobits per second up to 1.5 megabits per second download, and 128 kilobits per second upload. Theoretically speeds as high as five million bits per second can be achieved, but rarely are. At this time there is virtually no guarantee of speed from any provider, especially for the modest price services. There are commercial and professional service offerings, which charge up to several hundred dollars a month and which either guarantee minimum connect speed or offer refunds for less satisfactory service. The speed is considerably dependent upon the length of the copper line to the telco central office, with speed dropping rapidly as distance exceeds ten thousand feet. In some instances connect speed is deliberately restricted for the lower priced services, and of course in any event heavy traffic on the internet backbone and interconnects can restrict speed just as we now experience with V90 modems. Few people seem to report speeds less than 250,000 bits per second, which of course is at least a five to one improvement over V90. Anything above one megabit per second must be considered fortunate but certainly not unheard of. Additionally, the DSL modem establishes connection when powered up and remains connected at all times with most DSL providers, although this is not invariably the case. Accordingly, no dialup is required, with the elimination of the annoying time interval arising from handshaking and ISP verification procedures we are accustomed to with older modems.

The early ADSL installations were designed to employ a device at the customer's premises called a splitter. The splitter is a passive electrical filter that separates the low frequency signals (from 0 to 4000 Hz) from the high frequency DSL signals. The splitter was installed by a telco technician commonly at the point where the phone line enters the subscriber's premises. Two separate copper pairs were then run from the splitter, one to a DSL modem and the other to all the conventional phones on the premises. Most homes built in the last twenty years have two pairs running to all phone outlets, although some are already using the second pair for a second line service purchased from the telco. Frequently the telco technician had to run a lengthy line through the house in the usual telco semi-neat manner to provide the second line. The splitter installation and possible additional wire run were significant labor costs to the telco.

DSL-lite was originally designed to avoid an expensive telco technician house call by eliminating the splitter. Instead, small individual filters are installed by the customer at the outlet for each conventional telephone, including fax machines, conventional modems, answering machines, etc. These filters prevent the high frequency DSL signals from bothering the conventional phone devices. Even more important, they prevent the conventional devices from bothering the DSL modem. DSL-lite also limited design of the modems to speeds of about 1,500,000 bits per second, thereby reducing hardware costs at both ends of the line. Many residential offerings, including those of SWBell, are relying on customer installation of the DSL and associated filters.

All hardware associated with the installation is provided free of charge in the current SWBell offering, subject only to agreeing to a one year service subscription. There are three types of modems offered to the customer, the three models varying only in the manner in which the digital output of the modem is fed into the computer to be used with the DSL service.

1) Internal PCI slot modems which plug directly into the computer.
2) External modems which connect via the USB bus.
3) External modems which connect via an Ethernet cable to a simple
Ethernet adapter card installed in the computer.

The external models have a modest advantage in that a telco service engineer can test system operation in the event of problems, by substituting a portable computer for the customer's computer. This eliminates a sizeable portion of the hardware and software from the mystery, always a big factor in simplifying the thought process involved in troubleshooting of complex systems. In the extremely active Internet DSL-oriented usegroup comp.dcom.xdsl there is a very strong apparent bias toward the use of the Ethernet interface. This appears to be due to marginal hardware or software implementation of the internal PCI and USB versions, along with the well known bungling in operating system and motherboard implementation of the USB bus in earlier products.

There are two impediments to the physical telephone copper twisted pair installations that, if not removed, may prevent DSL service on a particular phone line. Both of these can be detected electronically, as can the actual length of the line from subscriber to central office. However, some of the Bell operating systems (I don't know about SWBell) have been attempting to rely on paper records regarding the history of the individual lines, and these records over the years have understandably departed from reality. The two offenders are:

    1) Loading coils. These are inductors that were sometimes connected in series with very long line runs to improve voice transmission. While they should not be present on lines short enough for DSL, they sometimes, over the years, found their way into short runs made by chopping up older long runs. Loading coils for long distance lines were the stuff of legend in the thirties, in which era it didn't take much in the way of high technology to really impress us.

    2) Bridged taps. These are parallel (of necessity, currently unused) branches which fork off from the subscribers line. The classis example for the reasoning that justified this apparently strange practice is embodied in the task stringing many new lines from central office to a location of, for example, four new high-rise apartment buildings. Some of the lines can be inexpensively paralleled into all four buildings and left unconnected in the building basement. When a new line is required in any one building at a future date the subscriber is simply connected to the "bridged tap" which was fortuitously installed when line installation crews were available in the past. Needless to say, the three remaining "bridged taps" on this line will be abandoned for future use unless the new install is later not needed.

Both offenders play havoc with the broadband frequency response (out to about 1Mhz) needed for DSL. Neither caused any problems with simple voice circuits in the past. Needless to say the removal of either offender type can be costly to the telco in terms of manpower, and so introduces the temptation to avoid initiating DSL service to the loading coil/bridged tap crowd.

Judging from voluminous Internet reports the universal user consensus regarding DSL is that it can be a monstrously frustrating experience to deal with the telco throughout the installation process, as well as nursing the system through its early usage days. After surviving this trying initial period, everyone agrees that DSL is very much worth having. I suspect that freeing up the phone line for simultaneous voice use, combined with instant service without listening to and waiting for our old friends the modem handshaking sounds, play a bigger part in their apparent delight with the new service than pure download speed. The only time that fast download would be really irreplaceable, in my opinion, would be for downloading files larger than two megabytes, where V90 does get to be a bit of a bore.

Those of us who are not in the users group solely to discuss online trading and financial software (the latter don't read the newsletter and hopefully will not take offense) might be interested in learning how exactly, after many years of hearing that bit rates can go no higher than V90 (in fairness itself a miracle) we are suddenly blessed with a major increase in download speed along with simultaneous voice usage on the same old tired copper pair. If interest emerges I will attempt to address this mystery from a technology viewpoint in a future paper.

Click here to return to top