How does a DSL line work?

About the author

Michael Spalter

---

Michael Spalter has been a networking technician for over 30 years and has been the CEO of DrayTek in the UK since the company’s formation in 1997. He has written and lectured extensively on networking topics. If you have an idea for a blog or a topic you would like explored, please get in touch with us.

 

For an overview of the different types of DSL lines available, see our previous article G.Fast, VDSL2 and Supervectoring (35b).  The article explains how each type of line varies and how they are delivered to the consumer. In this article, we go a lot deeper technically and examine how a DSL line actually works. How does your data get turned into a signal that can travel down a regular wire phone line?

The Analogue Days

If you used the Internet as a consumer 20 years ago (I did!), you were likely connecting via an analogue modem. For Wide Area Networking (WAN) - connecting computer systems in different locations - using the existing telephone network which was already in place for voice calls provided a convenient medium without having to install any new infrastructure.

A standard phone line and the exchanges they were connected to carried voice calls. The lines were limited to carrying frequencies between 300 Hz to 3.4 kHz. An analogue modem would dial the phone number that a modem (or ISP) was connected to at the other end and the two modems would communicate using audio tones within that frequency range.

The earliest modems used very simple modulation techniques - a single tone (frequency) which was modulated up or down to indicate a 0 or a 1. This was Frequency Shift Keying (FSK). Data could be sent in both directions (send and received) so a different frequency was used in each direction so that they could send and receive simultaneously. These early methods were slow - sending data at between 75-300 bits per second (that compares to today's DSL lines which can be a million times faster (300bps vs 300,000,000 bps).

As time went on, new coding methods were developed - every couple of years there was a new standard, and you would have to buy a newer model. Eventually, analogue "dial-up" modems could reach speeds of up to 56Kb/s downstream.  Even at that speed, you still had to wait for websites to load and it was common to turn off 'images' in your browser so that just the text would be loaded, thus loading was much quicker.

The introduction of DSL

Although the limit of analogue voiceband transmissions had been reached, there was still demand for ever-increasing speed as more content-rich websites and other online services grew. ISDN, a fully digital line, had already been introduced but it required a dedicated line to be installed, was costly to the customer in installation and ongoing rental and still only provided 64Kbps (or 128Kbps in ML-PPP bonded mode). ISDN did provide guaranteed speeds and connected instantly but you were still billed for your calls on a per-minute basis.

The introduction of DSL lines meant that always-on data services could be overlaid onto existing voice (analogue) lines whilst still allowing voice calls to be made by using frequencies above the voice band and users installing microfilters to filter out the data signal from telephones to avoid any interference.  When ADSL was first introduced into the UK, it offered speeds of 512Kbps downstream.

So how does DSL actually work?

DSL lines work on very similar principles to dial-up modems but they use a wider range of frequencies and use complex coding methods to get the absolute maximum speed possible from the line.  The lower frequencies can travel further; the higher frequencies get attenuated by the line, which is why the maximum speed on a DSL line will relate to the line length.  If you're right next to the exchange (ADSL) or street cabinet (VDSL) you wll get the maximum DSL speed, assuming the line and wiring are also of decent quality.

A DSL line splits its bandwidth into a certain number of individual carriers or subcarriers, also known as bins, buckets or DMT tones. Those bins are spread evenly across the total available bandwidth (frequency range). A bin can carry up to 15 bits.  Each carrier is orthogonal to the adjacent one, which means it's at 90 degrees in phase (GCSE geometry!). Making the carriers relatively orthogonal reduces interference between them, reducing the need for large intervals (tone spacing), which would waste space.

Digital communication methods share many common systems - OFDM, DMT and QAM are used not only for DSL (here) but also in WiFi and I have explained them in depth in other articles.

How to calculate the maximum speed of a DSL line

VDSL2 (17a profile), as an example, uses frequencies spread over a 17Mhz wide 'spectrum'.

In the real world, during the DSL training sequence (when the connection is initiated) the modem assesses the line (attenuation and power levels) and the SNR (Signal to Noise Ratio) of each bin. The SNR (Signal to Noise Ratio) tends to be worse at the higher end of the frequency range. Every line will have a different profile (quality) and be impaired at different frequencies. This means that each line cannot make full use of every bin/carrier and each carrier may not carry the full 15 bits. Certain bins are not usable due to clash with other services (to avoid interference) and some bins are reserved for signalling (as opposed to user data). So, once you add in line quality/length, downstream/upstream splits and other restrictions, you get down to the real-world maximums of 100Mb/s downstream, then further reduced to 80Mb/s to ensure uniform/consistent service.

The cycle length (or symbol rate /baud rate) for G.Fast is 48000bps compared to 4000bps in VDSL2 so a symbol lasts 20μsec (VDSL2 symbols last 250μsec). The subcarrier spacing of G.Fast (48Khz) is much wider than VDSL2 (4.3Khz) so even though it has three times the bandwidth of VDSL, it doesn't have three times the number of bins as each bin is spaced further apart.  A tone doesn't turn on and off instantly; each is a gated periodic wave and the modem's processing, using a Fast Fourier Transform (the maths of converting a period sample into data), is imperfect. The faster the symbol rate, the more each bin may 'spread' into adjacent space. So, to avoid interference from adjacent bins, the subcarrier/bin spacing must be wider in G.Fast.

The frequencies used for G.Fast cross those used for FM radio, so the overlapping part of the spectrum is not used to avoid interference (88-108Khz). The avoidance of the clashing subcarrier frequencies is called notching (because a 'notch' is cut out of the band).

Your DSL router or model may have a diagnostic screen showing the line quality and bin usage such as shown below.  Note how the bandwidth is shared between Upstream (Top/Blue) and Downstream (Bottom/Green) but that they use different frequencies.  The highest frequencies are used for downstream as they are the ones most likely to be lost (due to line problems or length). Typically, one can afford that, whereas, as the reserved upstream bandwidth is much smaller, you will be less willing to lose any of that.

Even after initial training, your DSL modem will continuously monitor the line. A process called bitswapping can reallocate unused bit to replace bits within other bins which have become unusable. A previously 'good' bin might become unusable due to real-yime line condition variations, such as line noise or even a change in the weather. Bitswapping can therefore aid connection stability.

I hope you found this article interesting; if you did, you will probably also enjoy the forthcoming one on 802.11ax Wi-Fi which uses many of the same principles as DSL.  In any case, please do send us any suggestions of articles you would like to see and please do share the links to this one.

Tags