The DSLAM at the Telco collects the digital signals from its many modem ports and combines them into one signal, via multiplexing.
Depending on the product, DSLAMs connect DSL lines with some combination of Asynchronous Transfer Mode (ATM), frame relay or Internet Protocol networks (a.k.a. IP-DSLAM that use the PTM-TC stack).
In terms of the OSI 7 Layer Model, the DSLAM acts like a massive network switch, since its functionality is purely Layer 2.
The aggregated signal then loads onto backbone switching equipment, traveling through an access network (AN) — also known as a Network Service Provider (NSP) — at speeds of up to 10 Gbit/s and connecting to the Internet-backbone.
The DSLAM, functioning as a switch, collects the ADSL modem data (connected to it via twisted or non-twisted pair copper wire) and multiplexes this data via the gigabit link that physically plugs into the DSLAM itself, into the Telco’s backbone.
A DSLAM is not always located in the telephone company central office, but may also serve customers within a neighborhood Serving Area Interface (SAI), sometimes in association with a digital loop carrier. DSLAMs are also used by hotels, lodges, residential neighbourhoods and other corporations setting up their own private telephone exchange.
Besides being a data switch and multiplexer, a DSLAM is also a large number of modems, each modem on the aggregation card communicating with a subscriber’s DSL modem. This modem function being inside the DSLAM rather than separate hardware, and being wideband rather than voiceband, it isn’t often called a modem. Like voiceband modems of standard v.32 and later, it has the ability to probe the line and train itself to compensate for forward echoes and other impairments, in order to move data at the maximum rate the line allows. This is also why twisted pair DSL services have a longer range than physically similar unshielded twisted pair (UTP) Ethernet.