Ever wondered how ADSL (or broadband internet access over telephone lines) works? I often have and, having worked
for an ISP in the UK (BT Retail) for about 6 months I can offer a slightly better explanation than when I started. ADSL stands for Asymmetric Digital Subscriber Line. It is asymmetric as a larger fracion of the available bandwidth on the copper telephone line is used for download throughput than it is for upload. On Max DSL enabled lines the downstream bandwidth can be up to 8 M bit / s, while the upstream is usually limited to 400 k bit / s.

The backbone data network in the UK is owned by BT Wholesale, who lease it's usage to the various service providers. Below I'll describe the path a connection takes to get from an end user's modem to their service provider.

The local telephone exchange

When the copper line reaches this point it first gets to the DSLAM (Digital Synchronous Line Access Multiplexer), which is a filter that separates the analog voice and DSL signals. The analog signals are digitised and put on the (PSTN network, while the DSL signals go into a rather large modem. For this modem to have any useful communication with the modem on the other end there needs to be DSL synchronisation between them, the full meaning of which is beyond my knowledge. An inability to gain sync (or the intermittant loss of it) is the most common cause of ADSL problems, which is mainly because the copper lines used can be several miles long and were not originally designed for this purpose. All being well though, the connection gets through to the ATM.

The ATM, RAS and Home Gateway

The ATM (Asynchronous Transfer Mode) network was built to allow more end users to connect to high bandwidth access points. Among other things it connects between the Colloses backbone network and the local telephone exchanges, effectively extending the reach of the backbone network. The aforementioned two netorks are both fibre optic based. Before the connection can pass from one to the other, it needs to go through the Remote Access Server (RAS). The RAS is responsible for authenticating the end user for network access, usually based on a username, password and the telephone number associated with the ADSL line. If the EU has a static IP address the RAS will cross check it's records with those of the Home Gateway Radius (HGR) server of the EU's service provider. Once authenticated the EU will be allocated an address on the network by the IP server.

NCR (No Connection to RAS) issues are another source of connection problems. The RAS has a database of which users are allowed to access network paths to which service providers (known as the RAS mapping). If this database becomes corrupted an EU may find their login details rejected, or the RAS may not even begin a PPP session with them. Save for annoying issues like this, the connection proceeds over the backbone network to the HGR from which requests to reach other network sites (websites etc.) are "served" to them. Happy days, hey?