Welcome, Guest. Please login or register.
DSL Broadband consists of data transmitted as analogue signals over a range of radio frequencies down a telephone line, about 95% of verbal communication happens around the 300-4000Hz range. Telephone networks use 0-4000Hz which covers human speech and signaling in the phone network. ADSL leaves a bit of space (white space) between it and the telephone frequencies then uses the remaining ones for your broadband, these are inbetween 28kHz and 1.1mHz for ADSL1/ADSL2 connections, inbetween 28kHz and 2.2mHz for ADSL2+ connections and inbetween 28kHz and 30mHz for VDSL2 connections.

Here we see a typical diagram showing the use of frequencies for ADSL1/ADSL2 over a telephone line:
File:ADSL frequency plan.svg

Local Telephone Exchange
Your broadband signal is routed through your ISP to the internet, a basic description of the signals journey woud be:-
Internet -> ISP -> High Speed Network -> Telephone Exchange -> Patch Point -> Customers Home
It is actually a lot more complicated than that but this is just to keep it nice and simple and easy to understand. When your broadband signal reaches the Exchange it's routed through a DSLAM which Multiplexes/De-Multiplexes other customers signals, signals outbound to the internet are 'Multiplexed' together with other customers broadband for easier transmission as one large signal and broadband signals inbound from the internet are 'De-Multiplexed' to be sent off to each customer, once the customers broadband signal leaves the DSLAM its routed through to a 'switch' which add's the customers Telephone Signal to it from the PSTN (Public Switched Telephone Network) and directs the signal off to the customer, the signal travels along E-Side (Exchange Side) cables to a Patch Point (youve probably seen these, usually green cabinets outside of housing or Industrial estates, these are a sort of halfway point from the Telephone Exchange to the customers premises), from the Patch Point the signal travels along D-Side (Distribution Side) Cable to the customers premises.

Here we see an old Patch Point being replaced with a nice shiny new one.

Filtering the Frequencies
When the broadband and telephone signals reach the customers home its splits into two signals (one for telephone and one for broadband) each filter out the frequencies they need, the telephone gets its frequencies by using a Filter that you plug into the telephone socket, it runs a low-pass filter on the entire spectrum which filters out the low frequencys for the telephone, these low frequencies then leave the filter and on to the telephone, the second signal (which is still both phone and broadband signal) is routed to the modem which has a high-pass filter built-in to filter out the high frequencies for your broadband, filtering allows a device to hear only a selected range of the frequencies. Higher frequencies attenuate more than lower frequencies, in other words, low frequencies can travel longer distances, thats why when you hear music coming from a local club in the distance you hear more of the bass-lines and drum kicks than the higher frequency sounds, because of this there are limitations on how many frequencies can be used to transmit data over distance on copper telephone wires and thus speeds acheived.

Here we see a diagram of how a typical broadband signal is filtered:
(some filters may do both the high and low pass filtering)

Coping with Background Noise on Lines
Your Broadband signal comes from your ISP (Internet Service Provider) usually over BT's backbone network and through to your local telephone exchange, there the Line Card (or DSLam) sets the sync rate (connection speed) between it and your modem, it does this by working out the SNR (Signal-to-Noise Ratio - how much background noise is on the line compared to the signal itself) and how much Attenuation (loss of signal over length) there is and 'adjusts' the SNR Margin and Power Output which in turn determines the connection rate (speed) on the line. The 'SNR Margin' is used to keep the connection live (afloat, if you will) from the noise on the line, if the noise becomes to great it exceeds the margin threshold and the line drops (disconnects).

Noise on lines is caused by crosstalk from other lines and/or interference in the same radio spectrum as the broadband signal, this can be due to other electrical devices emitting radio waves in the same or nearby frequencies that the broadband is transmitted over which 'block' the communication over those wave bands and thus give slower speeds.

Think of it like shouting at someone over a crowded hall, the further away they are and the noisier the hall, the louder you have to shout (but you can only shout so loud) and sometimes the hall is very noisy and sometimes its not - so the Line Card (or DSLAM) works out an 'average' of how noisy the hall is and set's a 'margin' so that any sudden increases in noise - the other person can still hear you (or in this case the router still receives a signal), the only downfall is that the bigger the margin (using more frequencies) the lower the connection speed and thats why longer and noisier lines have lower speeds.

Attenuation (Loss of Signal over Distance)
One of the issues with DSL technology over the existing infrastructure is the fact that it loses speed over distance, this is because of a few factors...

  • Attenuation or 'loss-of-signal' due to the length of the line itself
  • The Type, Quality and Guage of the wiring used
  • Noise introduced from neighbouring wires or environment

If you live close to your telephone exchange then you will probably get a good speed, but the further you are away the more the signal degrades and the more chance there is of your connection being run across a poor batch of wiring on its journey to your home (e.g. Aluminium wiring, low gauge copper, bad/rusty connections), and the more chance of cross-talk and background noise being introduced into your line, again, affecting your speed. You can tell what speed you should be getting from your line currently as your modem will display the Line Attenuation figure and from that you can work out a rough estimation of what your speeds should be and if they are low do something about it, you can do all this using our guides.

Software Modems (Bins / Tones)
and Bandwidth Blocks
Your modem that connects to the exchange actually uses 'hundreds' of software modems called Carrier Bins, it uses a Tone over the Bin to send and receive data (256 on ADSL1 and 512 on ADSL2+). Each Bin handles a 4.2kHz frequency block (or 'channel'), from the example image above you can see that ADSL1 uses frequencies from 26kHz to 1100kHz, this means the total available bandwidth of frequencies for ADSL1 is around 1078kHz (1.078Mhz) - divide that by 256 and you get 4.2kHz, 4kHz is used for the actual payload (bandwidth) and 0.2kHz is used as white space between each channel, each 4kHz channel has a "theoretical" bandwidth of 56Kbps (we say 'Theoretical' as the actual delivery payload will be slightly less).

ADSL1/2 uses 256 Bins like so:-
223 Bins for downstream (223 x 56k = 12Mbps max 'Theoretical' speed) + 25 for upstream (25 x 56k = 1.3Mbps) and 8 Bins are used as white space between the upload and download streams to prevent crosstalk. These are theoretical speeds and depend on standards and equipment used.

ADSL2+ uses 512 Bins like so:-
478 Bins for downstream (478 x 56k = 26Mbps max 'Theoretical' speed) + 25 for upstream (25 x 56k = 1.3Mbps) and 8 Bins are used as white space between the upload and download streams to prevent crosstalk.These are theoretical speeds and depend on standards and equipment used.

Extended Upload ADSL2+ (Annex M) takes bins/tones from the download frequencies to increase the upload speed to 3.3Mbps - namely the upload Bin count is increased from 25 to 59 and download bin count is decreased from 478 to 444, the upload/download frequency split is shifted from 138 kHz to 276 kHz.
To work out the above on Extended Upload ADSL2+ you do the math 276 - 26 (as under 26 kHz is used for white space and telephone) = 249 kHz, this is total frequency bandwidth used for Annex M upload, you then divide that by 4.2 because each channel is 4.2 kHz wide, this equals to 59, you then times 59 by 56 = 3304 which is 3.3Mbps.

Bit-Loading & Bit-Swapping
Bit-Loading refers to how many 'bits' of data each bin/tone can send, this is determined from the Signal-to-Noise Ratio on the frequencies the bin is using, where the SNR is low the power will have to be boosted but because each bin/tone is subject to an overall power limit then they carry less bits than that of a bin that has a higher SNR. When noise spikes are introduced into your line then certain bins SNR will drop and they will be able to send less bits (if at all), 'Bit-Swapping' is when the modem can dynamically swap bits with other bins that have some spare and/or can take power from now disabled bins and apply it to other bins to increase their carrying power and move the bits to those.

Error Correction and Interleaving
Due to noise spikes on the line from interference of somekind and/or dropped packets it is always likely that a percentage of data is lost on its way to its destination, continually re-transmitting lost data would cause significant delays and also increase overall traffic across the network so we use Error Correction and Interleaving to get around this. There are two types of Error Correction used, Trellis Code Modulation (TCM) and Reed Solomon.

Trellis Code Modulation (TCM)
TCM defines the sequences that symbols are transmitted in, if data is lost by the time it reaches the recipient then it can be caculated from the values before and after the missing value, for e.g. if we transmit 1,1,2,3,5,8,13 and the recipient gets 1,1,2,3,?,8,13 then it knows the missing value is 5 as 8-3=5.

Reed Solomon (FEC)
RS is a form of Forward-Error-Correction and is used to transmit the data layer, it works by adding a total to the end of the block of digits so that if data is lost it can still work out the missing value from the block total and remaing values, e.g. if it sends 1,4,7,12 (1+4+7=12) and recipient gets 1,?,7,12 it knows that 12-7-1=4 (which is the missing value).

Using the Error-Correction above is all good and proper but what if more than one value is missing from blocks of data? it would be impossible to calculate the missing values and a re-transmit would have to occurr (exactly what we are trying to avoid), this is where Interleaving comes in, what it does is 'mixes' up (interleaves) all the blocks before sending and then 'un-mixes' (De-Interleaves) the blocks at the recipients end so that if a contiguous piece of data was lost in the sequence it would end up being one value from a few blocks of data rather than a few values from one block of data. If we transmit 1,2,3,6 - 3,2,7,12 - 6,1,4,11 - 2,2,4,8 Interleaving would transmit it as 1,3,6,2 (all the 1st digits) - 2,2,1,2 (all 2nd digits) - 3,7,4,4 (all 3rd digits) - 6,12,11,8 (all 4th digits)
so the whole interleaved sequence would be:- 1,3,6,2,2,2,1,2,3,7,4,4,6,12,11,8
so for e.g. lets say the recipient gets 1,3,6,2,2,2,1,?,?,?,4,4,6,12,11,8 because it knows the first four numbers are the first digits and second lot of four numbers are the second digits (so on and so forth) it De-Interleaves the sequence to:
1,2,?,6 - 3,2,?,12 - 6,1,4,11 - 2,?,4,8

as you can see, even though the interleaved signal lost a contiguous (in-a-row) piece of data it is able to correct the data at the recipients end as now only one value is missing per block and it is able to calculate the missing value from the rest of the numbers, in the first block of four digits from the sequence above there you can see that 6 (which is the total as its the last number) minus 2 minus 1 equals 3 which was the missing value, through both Interleaving and Error-Correction data can be delivered more efficiently, the only cost to this is a slight delay in response times but this is only around 20 milliseconds (the time it takes for the data to be Interleaved by the sender and De-Interleaved by the recipient).

Some ISPs allow for Interleaving to be disabled which improves response times, only those living close to their telephone exchanges with short lines can benefit from this as there is less noise on the line and less likely that errors will occur, the only real benefits this brings is for online gaming where response time matters, web pages will load a fraction quicker too but there really isnt much in it, online gaming can be played just as well with Interleaving turned on (turning off Interleaving can bring down reponse times from 30-40ms to 5-15ms) - if one had a noisey line and wanted to have low pings then you could possibly trade off download speed for faster response times for gaming, it may be possible to set a really high SNR Margin (15-20db) and ask for Interleaving to be disabled (if your ISP allows it), the higher margin (and reduced connection speed) should allow for more stable 'error-free' line with faster reponse times but slower download speeds, this is because the high SNRM keeps your broadband signal further away from the noise on the line (if you will) , basically if your reducing or turning off error correction then its best to try and reduce the noise, I dont know if this works in practice as ive just read it elsewhere and thought I would mention it..

Interleaving also has a 'depth' value, in otherwords how much Interleaving to apply to the data, noisier lines (where theres more chance of errors occuring) can benefit from an increased Interleaving Depth, this essentially increases the amount of interleaving so that if "more than average" data is lost it can still arrive and work out the missing data, the cost again is very slight increase in latency but not very much if at all, These Depth values seem to range from 1 (off) up to 511.

Back to top 

Powered by EzPortal