Internet ADSL2+ Speed

From Wikibooks, open books for an open world
Jump to: navigation, search

Introduction[edit]

ADSL2+ is the fastest ADSL internet variant. The term is short for Asymmetric Digital Subscriber Line. The subscriber circuits for this type of broadband use copper wires all the way from the exchange to the house. Such broadband internet connections are commonly sold as having a maximum rate of 17 megabits per second. This page explains some of the ideas in a moderately technical way.

User-friendly Overview[edit]

On first connection to the line network, an automatic adjustment or synchronisation begins. Although the link can be used during this period, it takes about ten days or so for the system to sync itself correctly and to give the best line rate. Perhaps surprisingly, the fastest rate is not necessarily the best, since fast links have less stability. The network provider establishes the best compromise between speed and stability, bearing in mind factors such as the length of the line and the amount of electrical noise.

After the initial synchronisation period the network continuously monitors the line for errors and noise, and slight rate adjustments are made, both up and down, depending on how many people are using the internet at the time. The network provider keeps some flexibility in his settings so that hourly variations in noise cannot interrupt the connection, but sometimes, in extreme cases, the line is re-synced, for example, after a line fault is repaired.

The User's Equipment[edit]

The telephone provider's line terminates within the user's premises at a small connection box. From there, the user can plug in his phone and router, or run extension cables and wiring. This main box, like the line, is on rent from the phone company. Other items include frequency filters, and the computer or other equipment that is to connect to the network.

Figure 1: A typical ADSL filter or band splitter. It connects to a telephone socket, and provides both a phone output and another for the router, modem, or hub.

Filters[edit]

These serve to separate the telephone and data signals. Telephone and data travel on the same phone line. The low frequencies, (below 4Khz or so), belong to the telephone, and the higher frequencies to the data. Two kinds of filter are are in use. In one, the filters are housed in the main telephone connection box, (see figure 3), and the user makes connections to separate sockets for phone and data; in the other, a filter accessory that can be connected to a telephone socket does the same thing. See figure 1. While the filter's phone sockets are common to all phones, those for the ADSL data are of a different type, called RJ11 data connectors.

Figure 2: A Typical Router/ Hub. The RJ11 line socket is shown on the left, and on its right are four RJ45 sockets (Ethernet) that connect to the local computers.

The Router or Hub[edit]

The equipment that manages the link is called a router or sometimes a hub. At times other terminology is used for the data handling equipment. Units that convert between line frequencies and digital signals are called modems. Sometimes the word hub is heard; it is the process that distributes the data stream among the various computers. The hub also handles the wireless channels. However, because most manufacturers combine the modem, the router, and the hub into one unit, for example in the BT Home Hubs, the terminology has become somewhat blurred. In any case, the router is connected to line via the filter using an ADSL cable, with an RJ11 plug at both ends. A router yellow socket is connected to a computer's Ethernet socket by using a cable with an RJ45 plug at both ends. Figure 2 shows a typical router's socket panel.

Figure 3: A main connection box with its own ADSL filters. It shows the line connection below, and the sockets for phone and data.

Extension Cables[edit]

Clearly, having the computer equipment installed beside the main telephone box is not always suitable. If the router is to be placed beside the telephone main connection box, then a long Ethernet cable is needed to reach the computer. Alternatively, a long ADSL cable could be run from the phone box to a router placed near the computer. When the connection is always to be wifi, then it might be thought that a cable is unnecessary. However, the use of wifi will not give as high a download rate as a cable run from the router, so having at least the option of a hardwired cable connection is always preferred, even for a laptop.

Whichever method is used, the user must consider the signal losses that might be incurred by the extension, so choose good quality items, and bear in mind that the losses and noise will increase with its length. For ADSL extension cables, if possible, choose the round ones marked as Cat5 rather than flat ones since the round cables contain twisted wire pairs to reduce noise. This cable is similar to that of the line itself, so in lengths under 10 meters or so will not cause any undue impairment of the data. ADSL cable extensions have connectors already fitted so it is just a question of choosing a length that is slightly longer than required.

If there is no filter in the main connection box, that is, if it is just an ordinary telephone connection, then a filter must be placed between that connection and the router. It is unimportant as to whether the filter is located before or after any ADSL extension cable, as long as the router is given a filtered data connection.

Signals and Noise[edit]

Figure 4: The ADSL2 Frequency Spectra. Using combined modulation methods, ADSL2+ combines the equivalent of 512 x 4 KHz telephone bands to produce a theoretical digital download rate of 24 Mbps. The upload maximum is just 1 Mbps.

The throughput, the signal to noise ratio, and digital error rate are all related. As transmission rates and the interference from other cable users increases, the signal to noise ratio, (SNR), is reduced, and the detection of data levels becomes less certain. As a result, digital error rates increase, as does the rework (the overhead), and in the extreme a link can be interrupted or lose sync. The converse is also true; reducing transmission rates can improve all of these properties. In fact, when errors arise in data, error correction methods are first attempted, but when data packets are too badly damaged the data must be sent again, and at times the link rate must be reduced to obtain the necessary stability. The functions that can change line quality exist in sets, with variations in each component. These sets are called Profiles. To adjust a link the system selects from a finite set of these profiles, in a logical sequence, until a satisfactory solution is obtained.

Signal to Noise Ratio[edit]

It is the ratio of the signal level to the noise level. Sometimes it is written as SNR or S/N. It is expressed in decibels (dB), because the logarithmic form allows convenient addition and subtraction of quantities instead of the multiplication and division of natural values. It is also useful in that it allows compression of what tend to be wide power and voltage ranges. Although of great use to the network provider, the SNR ratio itself is rarely known by the internet user, even from his router's connection statistics. Knowing it alone would not necessarily allow him any insight into the quality of transmission unless many other factors were also known. It is more useful to know how far from noise overload the current working system is. This quantity is called the noise margin.

Noise Margin[edit]

This is sometimes referred to as signal to noise ratio margin or just margin. It is written as SNM and it too is expressed in decibels (dB). This quantity is the amount of noise that would need to be added to the prevailing signal to noise ratio to worsen it to the lowest expected signal to noise ratio. The lowest expected SNR is defined here as the level that produces a bit error rate (BER) of 10-7, or one errored bit in ten millions. This is the highest error rate that the provider intends to be manageable. Note that the margin is estimated with all of the advantages of error correction and other gains in place, so that when any of these change, the margin changes too. In a fully rate adaptive and stable state the margin is just:

SNR Margin = SNRprevailing - SNRlowest

To protect against the effect of noise and other changes, links are run, not at their best extremes but by allowing this operational margin for error. This fact is true of the initial synchronisation process and also for normal running. When a link is initialised, or synchronised, the margin that is used is called the target margin. It is the margin used in setting up the line. The target margin is initially selected to be no larger than necessary, to keep the line speed high, and large enough to maintain stability. In most systems it will be given a preferred value of 3,6,9,12, or 15dB. Although it seems unlikely that a preferred value could ever be the best choice, it should be remembered that a degree of adjustment is also available from other parameters like signal power.

Ideally, links are synchronised only when the prevailing SNR is equal to the sum of the lowest SNR and the target margin. This makes sure that the link will work even when an additional noise equal in value to that of target margin is applied. However, when this method is used the sync process could be delayed for several hours while the system waits for the right SNR to present itself, so other starting point compromises are also in use. In any case, this near-minimum level of SNR is referred to as the Optimum SNR. That is:

SNRoptimum = SNRlowest + Target Noise Margin

The choice of sync rate and target margin affect both the stability and throughput of the line. If a link is unstable, a higher margin can be used, or interleaving applied. Higher margins however, mean slower rates. If a faster rate is needed, interleaving can be removed or a lower margin used. These two changes in turn produce a worsened stability. The process of synchronisation attempts to find the best compromise between the two. See figure 7 for a summary of these effects.

During normal running, the margin changes slightly to reflect both changes in the prevailing SNR and in coding gains from error correction. Continuous small changes to the rate and other parameters are made so that the margin is returned to a value near that of the target value. That is to say, when the prevailing SNR improves, the margin increases too, and the provider takes advantage of it to reduce his overheads or increase the rate; when it worsens, the margin worsens too, and he applies more resources, or decreases the rate. Clearly the closer that the prevailing SNR is to the optimum SNR, the one used for synchronisation, the fewer running changes will be needed to maintain the link. This situation can be approximated if in the selection of a target margin, the optimum SNR can be made to equal the SNR most likely found on the link for the busiest part of the working day. With preferred values of margin, this would of course require adjustment of some other parameter to make the set fit.

The most commonly found value of noise margin is about 6dB. Although lines might be re-synchronised with margins above 6dB for long (say, above 2Km), or noisy lines, lines synced with margins lower than 6dB are found only on good quality connections. When lines worsen too badly for effective control, or after the repair of a fault, a resync might still be needed. See figure 6 for an example of how the longer term choices are made.

Table 1: Typical One Hour Error Rates for a Standard Option
GOOD
OK
POOR
ES <= 14
720 > ES > 14
ES >= 720
(MTBE > = 250)
(250 > MTBE > 5)
(MTBE < = 5)

Digital Error Rates[edit]

There are many error types used to control a link. Some have less significance than others. For example, successes in error correction (FEC) are likely to be of less concern than those that require the re-sending of packets, (CRC and HEC). Error counts are recorded by the link software and are used, in part, to decide the best sync rates. The following error counts are of particular interest in descriptions of the subject.

  • FEC, Forward Error Correction. These are errors that have been successfully corrected using error correction methods. This implies that it was not necessary to re-send their data packets, so they did not slow throughput.
  • CRC, Cyclic Redundancy Check. These errors are those that could not be corrected by error correction. They required data packets to be resent. Most forms of CRC check detect error without correcting it, so such errors in large numbers can slow throughput.
  • HEC, Header Error Correction. Packet headers have their own error checking and can correct one faulty bit within it. If there is more than one error the data must be sent again, and the HEC count is incremented. Such errors in large numbers can also slow throughput.
  • ES, Errored Seconds, or Error Seconds. This is the number of seconds in which there is any error at all. It is used in the calculation of MTBE, the mean time between errors, which in turn is used to categorize the quality of a line. See Tables 1 and 2.
  • BER, Bit Error Rate. It is the proportion of bits sent that have any error at all. It is used to estimate SNR and margin. The target BER is the BER associated with the lowest SNR, the highest manageable error rate that the provider intends. It is commonly a value of 10-7 or one in ten million bits.
  • MTBE, Mean Time Between Errors. It is the ratio of the link Uptime in seconds divided by the Error Seconds. Notice that the other error quantities are not used directly in this ratio. It is used to categorize line quality. See Table 1
  • MTBR, Mean Time Between Retrains. Retrains essentially means profile realignments like speed changes, to avoid a loss of sync. It is the ratio of the link Uptime in seconds divided by the number of retrains. It is used to categorize line quality. See Table 2.

About rates[edit]

There is some confusion over internet rate terminology. To illustrate the point with a specific example, an ADSL2+ link might have an attainable rate of 24 Mbps, a maximum rate of 17 Mbps, a sync rate of 14.335 Mbps, an IP profile rate of 12.643 mbps, an actual download rate or throughput of 12.34 Mbps, and a Minimum Guaranteed Access Line Speed of 8 Mbps. All of these measures are quoted at times.

Attainable Rate[edit]

This is the rate that the technology is capable of provided that no service restrictions are applied (eg: 24Mbps). It is never available, but is of interest in systems design.

Maximum Rate[edit]

This is a term used to advertise broadband packages. It is perhaps the most vague. It is the attainable rate further modified to take account of the level of service. For example, ADSL2+ is commonly sold as having a maximum rate of 17 Mbps.

Sync Rate[edit]

This is also known as the downstream rate or the DSL connection rate. It is determined largely by the physical properties of the line and by any local noise that is characteristic of the location. It is always equal to or less than the advertised maximum rate. The sync rate is determined at the time of initial connection. At that time the sync process for ADSL2+ is fully rate adaptive, in that there is no imposed maximum rate. It takes about 10 days or more for the system to adjust the rate to a sustainable and stable value. The line can be resynced at certain times, for example, after a fault repair, and the IP profile rate is recalculated from the sync rate every time that it changes. The router always displays this figure though it may be simply listed in the broadband connection settings as Downstream and Upstream. At times the sync rate is limited to particular range, a so-called profile band.

IP Profile Rate[edit]

It is the maximum download rate using the current sync rate. This is also known as the bRAS profile. It is found initially by taking 88.2% of the sync rate. This fraction applies to all 21st Century Network (21CN) ADSL products.

Download Rate[edit]

This is also known as the throughput or data rate. It changes slightly from hour to hour because of traffic management and traffic congestion. Traffic prioritization on the other hand, the giving of preference to some traffic types over others, is not applied to broadband products by the main providers, so the throughput tends to be quite close to the IP Profile rate for most of the time.

Guaranteed Line Speed[edit]

The internet providers (ISPs) of some countries sign up to a voluntary and self-regulatory code of practice. It provides their clients some protection against what they feel are lower than advertised speeds. In the United Kingdom for example, the Minimum Guaranteed Access Line Speed is such a scheme. The speed for a particular client is found statistically from a set of those with similar line profiles. The sync rates of the lowest 10% of such a set is taken and the highest of these is the Minimum Guaranteed Access Line Speed for that client. Notice that it does not apply to the download rate, but to the sync or line connection rate. Those with speeds below this during the first 90 days of a contract have the right to nullify it, subject to giving reasonable opportunity to the provider to correct the matter. Most of the major providers are signed up to it.

Line Management[edit]

The DLM makes continuous changes to the operational profile. It makes use of error rate logs and other data that are continuously supplied. These allow an estimate of the ILQ or Indicative Line Quality. The process functions for synchronisation and for day to day running. The parameters controlled by the DLM include the Sync rate, the Target noise margin, Latency type, the balance of frequencies used, and the levels of each. ADSL2+ can also make use of profile banding where a link must sync within a restricted range.

Many line management functions make use of bit error rates (BER) rather than SNR. Because all of a link's parameters are known, there can exist a calculable relationship between the two. As such, for a fixed profile, a knowlege of the one allows a sufficient estimate of the other. In fact, the bit error rate is found by taking a sample, with a sufficient sample size to provide the necessary level of confidence.

Internet service providers (ISPs) must give service priority to speed, stability, or extreme stability, but it is difficult at the time of contract for users to know their ISP's intentions. Online gamers prefer speed and low latency values, whereas some critical applications prioritize in terms of stability.

For a detailed description of the initial synchronisation process itself, some will find the British Telecom PLC patent application entitled Optimised line synchronisation the most informative. See also the flow charts of figures 6 and 7 below.

In the determination of sync rates and other associated data rates, the history of link interruptions and errors provide a basis for its actions. Using these and a knowledge of the intended standard of service, the profiles might be changed immediately, changed after study, or not changed at all. The tables below show some of the decision making details.

MTBE and MTBR[edit]

The mean time between errors (MTBE) and the mean time between dropouts or retrains (MTBR) are both used to determine actions by the DLM. Despite the many available error types, only the Errored Seconds figure is used for the MTBE calculation. The two are calculated as:

Mean Time Between Errors (MTBE) = ADSL Uptime (secs) / Errored Second Count
Mean Time Between Retrains (MTBR) = ADSL Uptime (secs) / Retrain Count

BER Approximations[edit]

The BER or Bit Error Rate approximation depends on the size of the CRC block that is being error detected. Consequently there are two of them, one that assumes no interleaving and another that does. The two are calculated as:

For the Fast path (with maximum delay = S1):
Bit Error Rate (BER) = Number of bit errors / Number of transmitted bits
≡ 15 * CRC error count / data rate (bps) * ADSL Uptime (secs)

and For the Interleaved path (with maximum delay ≠ S1):
≡ 40 * CRC error count / data rate (bps) * ADSL Uptime (secs)

Determining Link Quality[edit]

Table 2 shows how, once the MTBE and MTBR have been calculated, the line status can be identified. The categories are arranged from good down to very poor. For example, a MTBE of at least 250 and a MTBR of at least 16800 are needed in a Standard network to qualify it as good. Notice that a so-called Stable link has a higher expectation, needing an MTBE of at least 6000 and a MTBR of at least 33600 to qualify for the good category. The Super-stable is more critical still. The categories have been color coded, and the same colors are used in Table 2 when identifying the relevant courses of action.

Table 2: Typical Line Categorization
Stability Option Metric Good OK Poor Very Poor
Agressive
(Standard)
Retrains mtb ≥ 16800 mtb < 16800 mtb < 8640 mtb > 10 per hour
Errors mtb ≥ 250 mtb < 250 mtb < 5  
Normal
(Stable)
Retrains mtb ≥ 33600 mtb < 33600 mtb < 16800 mtb > 10 per hour
Errors mtb ≥ 6000 mtb < 6000 mtb < 300  
Stable
Super Stable
Retrains mtb ≥ 67200 mtb < 67200 mtb < 33600 mtb > 10 per hour
Errors mtb ≥ 60000 mtb < 60000 mtb < 3600  

Padding.gif

DLM Courses of Action[edit]

The line quality levels determined in Table 2 are used here in Table 3, to identify the associated courses of action.

Table 3: Typical Courses of Action
Line Classification ILQ Status Action
Good - Performing beyond expectations Green Possible removal or reduction of any barriers.
Within acceptable parameters Amber No changes will be made to the DLM profile.
Poor MTBR / MTBE Red Applies a further step to increase stability.
Poor MTBE upstream Crimson Applies a further step to increase stability.
Rapid Retrains (Dropouts) Scarlet Additional line monitoring for immediate changes.
No DLM data Grey No action.
Insufficient DLM data Black Days uptime was less than 15 mins. No action.

Fixed and Adaptive Profiles[edit]

For a more detailed flow chart of DLM actions, refer to the drop box below. It shows the decision process used in switching between the profile levels of a fully adaptive rate and those of fixed profile bands.

Sync rates are sometimes limited. Although ADSL2+ is fully rate adaptive at startup, when there are constant link failures or noise the link might be assigned to one of the fixed profile bands, as described in the flow chart. This band is a range of rates within which the link must sync. The associated IP profile rates will be 88.2% of these. A list of the fixed profile bands used for ADSL2+ is also given below.

Figures 6 and 7: Typical Profile Decision Making
Figure 6: A typical decision tree for profile changes. Notice that both fixed profile bands and the fully rate adaptive (FRA) service are included.

Figure 7: The SNR Margin-Stability Compromise: Dotted lines tend toward lower stability while solid lines tend toward better stability.


Table3: Profile bands for ADSL2+
Lower Rate Upper Rate
160 kbps 288 kbps
288 kbps 576 kbps
576 kbps 1152 kbps
1152 kbps 2272 kbps
1472 kbps 3072 kbps
2272 kbps 4544 kbps
3328 kbps 6656 kbps
4864 kbps 9728 kbps
7168 kbps 14336 kbps

See Also[edit]

External Links[edit]