XDSL FAQ

Table of contents

1) What is xDSL? 2) How fast is xDSL? 3) Where are the xDSL standards? 4) How does xDSL compare to other technologies? 5) Should I get xDSL? 6) How does xDSL work? 7) What are the various types of xDSL? 8) How much does xDSL cost? 9) Is xDSL available in my area? 10) Why are some variations of xDSL asymmetric? 11) What does a POTS splitter do and when do I need one? 12) What is QAM? 13) What is PCM? 14) What is PAM? 15) What is V.90? 16) What is CAP? 17) What is DMT?

What is xDSL?

xDSL is a generic abbreviation for the many flavors of DSL or Digital Subscriber Line technology. DSL refers to the technology used between a customer's premises and the telephone company, enabling more bandwidth over the already installed copper cabling than users have traditionally had.

How fast is xDSL?

The short answer is "it depends". Typically speeds start at about 128Kb/s and go up to 1.5Mb/s for most home users. Some installations may go as fast as 50Mb/s or more depending primarily on the equipment used, distances involved, cabling quality, encoding techniques, frequency spectrum available and even to some degree, end system configurations. Be aware that some xDSL is sold as asymmetric or "rate-adaptive". It is best to consult the providers in your area as to the access rates available in your area. Speeds can vary from provider to provider even if they are all servicing your area from the same central office.

Where are the xDSL standards?

From International Telecommunication Union (ITU) http://www.itu.int

• G.992.1 (G.dmt) standards information
• G.992.2 (G.lite) standards information

From American National Standards Institute (ANSI) http://www.ansi.org

• ANSI TI.413-1998 ($175.00 US)
• Asymmetric Digital Subscriber Line (ADSL) Metallic Interface

From Universal ADSL Working Group http://www.uawg.org

• G.lite standards information

From the Standards Committee T1-Telecommunications http://www.t1.org

• Many xDSL standards
• Relevant documents are from the T1E1.4 (Digital Subscriber Loop

Access) working group

From European Telecommunications Standards Institute (ETSI) http://www.etsi.org

• ADSL, VDSL and SDSL standards

From the Internet Engineering Task Force (IETF) http://www.ietf.org

• ADSL MIB working group

http://www.ietf.org/html-charters/adslmib-charter.html

How does xDSL compare to other technologies?

• Cable Modems

Cable modems are devices that attach to the cable TV network connection in a home. This broadband technology is being driven by the cable companies to provide services beyond traditional broadcast cable TV such as Internet access. Along with xDSL, it is still in the early stages of development. There are a number of challenges faced by this industry, including return path capabilities, customer service issues and standards. However, potential bandwidth estimates range upwards of 30Mb/s from the service provider to subscriber. Cable networks are inherently different in design than telephone networks. Cable networks are broadcast oriented, with each subscriber in an area receiving the same signals as all others in that area. xDSL is circuit oriented so that each connection is independent of all others. Cable networks are inherently hierarchical in nature and thus require two paths, one for downstream and one for upstream. This requires either a second cable plant for upstream or a second frequency band allocated onto the existing system.

• ISDN

ISDN is a telephone company technology that provides digital service typically in increments of 64Kb/s channels. ISDN has been around for many years, but it's popularity only recently began to increase due to the limitations of analog modems and the rise of Internet usage. ISDN requires the phone company to install services within their phone switches to support this digitally switched connection service. Roll out of this service initially got off to a slow start and was stalled by high costs, lack of standards and low acceptance rates by consumers. xDSL and other new high speed technologies have in many cases "leapfrogged" the ISDN market.

• T1

A T1 (E1 is the European near equivalent) line is a 1.544 Mb/s pulse code modulated (PCM) system compromised of 24 time division multiplexed (TDM) channels of 64 Kb/s each. A T1 defines a copper copper wire interface specification for transmission between a customer and provider. Not to be confused with a DS1, which is the digital signaling rate of the underlying carrier. Many people however use these terms interchangeably. T1/E1 lines have been used in voice and data networks throughout the world where highly available, high capacity networks needed to be built. In fact, DS1 (or T1) is just one step in hierarchy of systems with higher speeds (e.g. T3/DS3). In many cases, T1 lines have been installed for end users who require dedicated high speed bandwidth between their home and work (or Internet). T1/E1 cabling requirements are more stringent than that of xDSL with the setup costs reflecting the differences in the service. Still a popular solution for many organizations and individuals, typically you will find that this service is considerably more expensive for an end user than xDSL or cable modems. However, the service level for T1 lines is usually very high.

• Voiceband Modems

Voiceband modems (or just modems for short) use a telephone network as is. That is, there are no special provisions that are required to use modems in today's telephone networks. Modems allow digital data to flow over the telephone company's traditional telephone network by performing a digital to analog conversion for transmission onto the network and vice versa on the receiving end. The only requirement for modems is that each end of the call must have a compatible modem. In essence, this makes modem connections the most ubiquitous form of data communications available today. However, modems are limited by the telephone company's voice bandwidth service. Current voiceband modem technology is struggling to achieve rates of only 56Kb/s. With only a bandwidth of about 3,000 Hz, there is a extremely finite limit on the amount of data that can be encoded and sent reliably through this network. User requirements far outstrip what modems can obtain today.

• Wireless

There are a number of different wireless schemes proposed, planned and implemented throughout the world. Wireless access technology takes shape in a number of different forms such as via a satellite TV service provider or a cellular phone network. Wireless systems can provide ubiquitous access to a large number of subscribers in a relatively large area. Bandwidth can range from a few kilobits a second to many megabits and be either symmetrical or asymmetrical. Like all other technologies, there can be deployment issues which may include spectrum licensing, interference, line of sight requirements, noise problems or bandwidth limitations.

• xDSL

xDSL is technology backed by telephone companies to provide next generation high bandwidth services to the home and business using the existing telephone cabling infrastructure. xDSL to the home over existing phone lines promises bandwidths up to 9Mb/s or more, but distance limitations and line quality conditions can reduce what will actually be achievable. xDSL technologies will use a greater range of frequencies over the telephone cable than the traditional telephone services have used. This in turn allows for greater bandwidth with which to send and receive information. xDSL technology is still in the early stages of development with standards and products just getting under way. Driving this market is the competition from competing access providers and the pursuit of your Internet access dollar.

Should I get xDSL?

That depends on a number of answers to questions which you'll need to ask yourself. First and foremost you need to determine if DSL is even available in your area. You may not have a choice. By reading this FAQ, you can hopefully learn enough about xDSL and how to get more information to make an informed decision. Although there are merits to all competing technologies, we make no recommendation in this FAQ to specify which one is right for you.

How does xDSL work?

xDSL utilizes more of the bandwidth on copper phone lines than what is currently used for plain old telephone service (POTS). By utilizing frequencies above the telephone bandwidth (300Hz to 3,200Hz), xDSL can encode more data to achieve higher data rates than would otherwise be possible in the restricted frequency range of a POTS network. In order to utilize the frequencies above the voice audio spectrum, xDSL equipment must be installed on both ends and the copper wire in between must be able to sustain the higher frequencies for the entire route. This means that bandwidth limiting devices such as loading coils must be removed or avoided.

What are the various types of xDSL?

There are several forms of xDSL, each designed around specific goals and needs of the marketplace. Some forms of xDSL are proprietary, some are simply theoretical models and some are widely used standards. They may best be categorized within the modulation methods used to encode data. Below is a brief summary of some of the known types of xDSL technologies.

ADSL
Asymmetric Digital Subscriber Line (ADSL) is the most popular form of xDSL technology. The key to ADSL is that the upstream and downstream bandwidth is asymmetric, or uneven. In practice, the bandwidth from the provider to the user (downstream) will be the higher speed path. This is in part due to the limitation of the telephone cabling system and the desire to accommodate the typical Internet usage pattern where the majority of data is being sent to the user (programs, graphics, sounds and video) with minimal upload capacity required (keystrokes and mouse clicks). Downstream speeds typically range from 768 Kb/s to 9 Mb/s Upstream speeds typically range from 64Kb/s to 1.5Mb/s.

ADSL Lite (see G.lite)

CDSL
Consumer Digital Subscriber Line (CDSL) is a proprietary technology trademarked by Rockwell International.

CiDSL
Globespan's proprietary, splitterless Consumer-installable Digital Subscriber Line (CiDSL).

EtherLoop
EtherLoop is currently a proprietary technology from Nortel, short for Ethernet Local Loop. EtherLoop uses the advanced signal modulation techniques of DSL and combines them with the half-duplex "burst" packet nature of Ethernet. EtherLoop modems will only generate hi-frequency signals when there is something to send. The rest of the time, they will use only a low-frequency (ISDN-speed) management signal. EtherLoop can measure the ambient noise between packets. This will allow the ability to avoid interference on a packet-by-packet basis by shifting frequencies as necessary. Since EtherLoop will be half-duplex, it is capable of generating the same bandwidth rate in either the upstream or downstream direction, but not simultaneously. Nortel is initially planning for speeds ranging between 1.5Mb/s and 10Mb/s depending on line quality and distance limitations.

G.lite
A lower data rate version of Asymmetric Digital Subscriber Line (ADSL) was been proposed as an extension to ANSI standard T1.413 by the UAWG (Universal ADSL Working Group) led by Microsoft, Intel, and Compaq. This is known as G.992.2 in the ITU standards committee. It uses the same modulation scheme as ADSL (DMT), but eliminates the POTS splitter at the customer premises. As a result, the ADSL signal is carried over all of the house wiring which results in lower available bandwidth due to greater noise impairments. Often a misnomer, this technology is not splitterless per se. Instead of requiring a splitter at customer premises, the splitting of the signal is done at the local CO.

G.shdsl
G.shdsl is a ITU standard which offers a rich set of features (e.g. rate adaptive) and offers greater reach than many current standards. G.shdsl also allows for the negotiation of a number of framing protocols including ATM, T1, E1, ISDN and IP. G.shdsl is touted as being able to replace T1, E1, HDSL, SDSL HDSL2, ISDN and IDSL technologies.

HDSL
High Bit-rate Digital Subscriber Line (HDSL) is generally used as a substitute for T1/E1. HDSL is becoming popular as a way to provide full-duplex symmetric data communication at rates up to 1.544 Mb/s (2.048 Mb/s in Europe) over moderate distances via conventional telephone twisted-pair wires. Traditional T1 (E1 in Europe) requires repeaters every 6000 ft. to boost the signal strength. HDSL has a longer range than T1/E1 without the use of repeaters to allow transmission over distances up to 12,000 feet. It uses pulse amplitude modulation (PAM) on a 4-wire loop.

HDSL2
High Bit-rate Digital Subscriber Line 2 was designed to transport T1 signaling at 1.544 Mb/s over a single copper pair. HDSL2 uses overlapped phase Trellis-code interlocked spectrum (OPTIS).

IDSL
ISDN based DSL developed originally by Ascend Communications. IDSL uses 2B1Q line coding and typically supports data transfer rates of 128 Kb/s. Many end users have had to suffice with IDSL service when full speed ADSL was not available in their area. This technology is similar to ISDN, but uses the full bandwidth of two 64 Kb/s bearer channels plus one 16 Kb/s delta channel.

MDSL
Usually this stands for multi-rate Digital Subscriber Line (MDSL). It depends on the context of the acronym as to its meaning. It is either a proprietary scheme for SDSL or simply a generic alternative to the more common ADSL name In the former case, you may see the acronym MSDSL. There is also another proprietary scheme which stands for medium-bit-rate DSL.

RADSL
Rate Adaptive Digital Subscriber Line (RADSL) is any rate adaptive xDSL modem, but may specifically refer to a proprietary modulation standard designed by Globespan Semiconductor. It uses carrierless amplitude and phase modulation (CAP). T1.413 standard DMT modems are also technically RADSL, but generally not referred to as such. The uplink rate depends on the downlink rate, which is a function of line conditions and signal to noise ratio (SNR).

SDSL
Symmetric Digital Subscriber Line (SDSL) is a 2-wire implementation of HDSL. Supports T1/E1 on a single pair to a distance of 11,000 ft. The name has become more generic over time to refer to symmetric service at a variety of rates over a single loop.

UDSL
Universal DSL. See G.lite.

VDSL
Very High Bit-rate Digital Subscriber Line (VDSL) is proposed for shorter local loops, perhaps up to 3000 ft. Data rates exceed 10 Mb/s.

How much does xDSL cost?

It varies. xDSL service availability is still in the early stages, but pricing in some areas has been very aggressive. Prices can change overnight and differ significantly depending on the service provider and surrounding area. Local tariffs and government regulations may also play a role in determining end user cost. To complicate matters further, some providers are claiming to offer free xDSL service. In many of these cases however, it requires you to be subjected to directed marketing or to make long term commitments to their service. You should first determine what your needs and tolerances are. Do you want static IP addresses? How fast do you want to go? What level of service do you require? Do you need multiple email addresses? ...and so on. Your answers to these types of questions will help you narrow down your choices. To find out more about how much xDSL service may cost, check with the service providers listed in section [8.8] or ask in the newsgroup(s) or mailing list(s) for the most up to date information.

Is xDSL available in my area?

To find out, you can check a number of sources. First, you can check with your local telephone company to see if they are providing xDSL services. Second, check around with your local Internet Service Providers (ISPs). Thirdly, try the competitive local exchange companies (CLECs) in your area. A good resource for CLECs is at http://www.clec.com. Lastly, there are some sites which claim to tell you if DSL is available in your area simply by filling in a online form. Unfortunately you cannot rely upon these sites for 100% accuracy. Even if you're told xDSL is available in your area, you still might be not able to get it. Often providers will need to perform a "qualification test" to determine if they can send and receive a signal within their parameters. Long local loops and poor cabling plants are common reasons for failing a loop qualification test.

Why are some variations of xDSL asymmetric?

It is primarily due to near-end crosstalk (NEXT). The large bundle of wire at the CO is heavily susceptible to crosstalk, particularly with regards to the signal that travels from the far end (the end user). At the far end, there are fewer problems with NEXT so bandwidth is greater from the CO to the user.

High bit rates, or in this case, higher frequencies suffer a greater amount of attenuation. The reason that the upstream speed in ADSL is generally much less than the downstream rate is due to this fact. When the high frequencies have attenuated at the CO end, they are very susceptible to all the other signals in the binder group due to EMI.

In the downstream direction, the high frequencies still attenuate, but at the customer end, they have a better chance of avoiding crosstalk since most subscribers will not have large bundles of cables running into their premises.

What does a POTS splitter do and when do I need one?

A POTS splitter uses a low pass filter to separate the low end frequencies of the telephone audio spectrum from the higher frequencies of the xDSL signals. The splitter should be a passive device, not requiring power so that "life-line" voice service can be provided as has been in the past. This splitter allows for the traditional voice service that consumers are accustomed to. A splitter is required at both the customer premises and at the far end (CO). xDSL that does not use a POTS splitter on customer premises is termed "splitter-less xDSL". However, there really is no such thing as splitter-less xDSL. The splitter function in these cases is just performed at the provider, generally the CO. Whether a POTS splitter is required or not depends on the xDSL service being provided.

What is QAM?

Quadrature amplitude modulation (QAM) is a method for encoding data on a single carrier frequency. The modulation encodes data (or bits) as discrete phase plus amplitude changes of a carrier tone. The phase vectors are arranged in a pattern of points called a constellation from which the transmitted point is selected based on the data to be sent.

The modem sends the symbols as abrupt changes in phase and amplitude, but only as what emerges from a sharp cutoff filter which carefully limits the bandwidth. The transmitted signal occupies slightly more than ±1/2 the modulation rate either side of the carrier frequency. The excess bandwidth, perhaps as much as 10%, is required for recovering symbol timing within the remote receiver.

The receiver has to pick which point was transmitted with great reliability. It may employ adaptive equalization or other methods to reduce intersymbol interference to levels which are acceptable for discriminating the received point. The background noise level of the receiver limits the number of distinct constellation points which may be reliably determined, and hence limits the data rate for a given symbol rate.

QAM has become the dominate modulation for high speed voice band modems. Examples are V.22bis, V.27, V.29, V.32bis, V.34. About every 2/3 of a carrier cycle the phase or amplitude is changed to a new value. This signaling rate is known as the baud (or symbol) rate. The highest QAM baud rate in use today for telephone line modems is 10/7 of 2400 Hz or about 3429 baud on a 1920 Hz carrier in V.34. By encoding something between 9 to 10 bits per baud a final data rate of 33.6Kb/s is developed. To encode this number of bits, over 1000 different phase/amplitude values must be resolved by the receiver. This is a nontrivial process involving adaptive equalizers, trellis coding, and other highly sophisticated signal processing.

Transmit path: scrambler -> symbol generator -> 3x upsample (S1,0,0,S2,0,0,S3,...) -> complex transmit baseband FIR filter -> e^jwt carrier modulation -> scale real signal output -> DAC converter

The baseband filter is about 3 dB down at ±1/2 symbol rate, so for 3429 baud the signal out of the filter extends from -1715 Hz to +1715Hz. This is shifted by the positive 1920 Hz carrier to +205Hz to +3635Hz. One can see that this just fits in the frequency spectrum of the voice band telephone network. This filter, the analog electronics and the phone channel smear any given symbol over a 10msec period of the signal (about 32 symbols).

The scrambler is very important. It randomizes the signal so an adaptive equalizer in the remote modem can build the inverse channel response (including the transmit filter). The smearing (or intersymbol interference) is largely eliminated by dynamically adjusting adaptive equalizer coefficients with the goal of minimizing least square error in the received points. The major adaptation is done during the training phase, although the feedback loops remain active throughout the connection. Other impairments to be solved are gain normalization, timing recovery, carrier offset frequency, phase jitter removal, and echo cancellation.

What is PCM?

Pulse code modulation (PCM) is used in the phone network to reduce the data rate required for voice grade audio to less than 64Kb/s. It uses either u-law (North America) or A-law (Europe) as the compression method. Any given 8 kHz analog audio sample is converted to 4 bits of mantissa, 3 bits of exponent, and a sign bit. This code has a characteristic that quantization noise is proportional to signal amplitude and does not become objectionable to the average telephone user. For a conventional modem this noise floor limits the available dynamic range to about 36 dB which sets the maximum data rate. The least significant bit of the mantissa may be periodically stolen for signaling within the phone network (called robbed-bit signaling) further increasing the noise.

The 8-bit codes are processed through the telephone switching network in fixed time slots. There exists an ever increasing hierarchy of data rates to support this. A DS0 is a 64Kb/s time slot. 24 DS0s become a DS1. 4 DS1s become a DS2 (now obsolete). 7 DS2s become a DS3, etc.

The physical layer of a DS1 (T1) may be remodulated as alternate mark inversion for passing over a wire pair as a method to concentrate local loops. Repeaters regenerate the signal every 6000-9000'. These signals may coexist with xDSL in the same wire bundle.

What is PAM?

Pulse amplitude modulation (PAM) is the physical layer of an ISDN or HDSL connection. The modulation consists of sending discrete amplitude levels (symmetric about 0 volts) at a regular rate. Both use the two binary one quaternary (2B1Q) line code. Four analog voltages (called quaternary symbols) are used to represent the four possible combinations of two bits. These symbols are assigned the names +3, +1, -1, and -3. So each amplitude level being held for one symbol time communicates two bits.

The following diagram is typical of the 2B1Q waveform at the transmitter:

+3 =  2.64V +   .--.        .--.                    .--
            +   |  |        |  |                    |
+1 =  0.88V +   |  `--.  .--'  |        .--.        | 
            ++++|+++++|++|+++++|++++++++|++|++++++++|++++
-1 = -0.88V + --'     |  |     |  .-----'  `--.     |
            +         |  |     |  |           |     |
-3 = -2.64V +         `--'     `--'           `-----'

One might assume this is a digital signal relative to the definition in [4.2], but by the time the signal has reached the receiver these discrete levels have diffused into each other because of phone line induced amplitude and phase distortion. This is called intersymbol interference. Therefore an adaptive equalizer must be used to restore the levels to values which may be discriminated for recovering the data. The symbol timing is recovered by examining the squared signal energy for a tone at the modulation rate. Transitions between levels cause the instantaneous power to dip on average provided there is adequate excess bandwidth.

PAM differs from the other modulations in that it is baseband modulation and does not use a carrier. Some versions of HDSL increase the number of levels to 16 which communicates four bits per symbol in the same bandwidth.

What is V.90?

V.90 is actually a variant of PAM. It has 256 PCM levels from which to choose a more limited set. The spacing between levels is set by the u-law or A-law characteristic described in [6.2]. The inner levels become more closely spaced so some of these must be excluded for reasons of limited signal-to-noise ratio. In addition, outer codes are excluded to keep transmit power on the local loop below -12dBm, a formal limit established by the FCC. V.90 includes a spectral shaping algorithm to prevent sending signal at DC.

V.90 bypasses the problems associated with a conventional modem. It recognizes that with enough signal processing the original PCM samples sent by the phone company may be resolved as individual levels using a 16-bit A/D converter on the receiving end. Audio is sent through the digital network as 8-bit u-law or A-law samples. Of course, the telco D/A converter, reconstruction filter, and phone network blur the levels into one continuous signal, so it's up to the receiver to reconstruct what was sent. An additional problem is recovering symbol (i.e. PCM sample) timing information which must be inferred from the residue of modulation at a frequency around 4 kHz. By just selecting a limited set of codes with say 64 levels, 6 bits per 8 kHz symbol may be sent for a data rate of 48Kb/s. More levels, more data, but a maximum of about 53.3Kb/s is a practical limit.

What is CAP?

Carrierless amplitude and phase (CAP) modulation is a proprietary standard implemented by Globespan Semiconductor. While the name specifies that the modulation is "carrierless" an actual carrier is imposed by the transmit band shaping filter through which the outbound symbols are filtered. Hence CAP is algorithmically identical to QAM. The upstream symbol rate is 136K baud on a 113.2KHz carrier, while the downstream symbol rate is 340K baud on a 435.5KHz carrier, 680K baud on a 631KHz carrier, or 952K baud on a 787.5KHz carrier. This allows the modem to be symbol rate adaptive to varying line conditions (see RADSL). The QAM modulation is also rate adaptive by varying the number of bits per symbol.

One advantage CAP claims to have is a lower peak-to-average signal power ratio relative to DMT. This means that the drivers and receivers may operate at lower power than DMT because they are not required to have the peak signal capacity that is required in the DMT circuitry. This is mitigated by the infrequency of the really high signal peaks in DMT which may be just considered to be another form of noise if they happen to clip.

CAP's principle advantage is its installed base of modems. It is actively being deployed in many trial markets and is available from several manufacturers.

What is DMT?

Discrete multitone (DMT) modulation is a method by which the usable frequency range is separated into 256 frequency bands (or channels) of 4.3125KHz each. These are intimately connected to the FFT (fast Fourier transform) algorithm which DMT uses as its modulator and demodulator. The FFT is not perfect in separating the frequencies into individual bands, but it does well enough, and it generates spectra which are fully separable on the receiving end. By dividing the frequency spectrum into multiple channels DMT is thought to perform better in the presence of interference sources such as AM radio transmitters. It is also better able to focus its transmit power on those portions of the spectrum in which it is profitable to send data.

The assignment of channels is less flexible, but typical settings might be channels 6-31 for upstream (24KHz-136KHz), 32-250 for downstream (136KHz-1.1MHz). The modulation used on any given frequency channel is QAM. Channels 16 and 64 are reserved for pilot tones which are used to recover timing. The number of bits per symbol within each channel may be independently selected allowing the modem to be rate adaptive.

The use of the FFT is considered to be somewhat substandard to other orthogonal transformations such as the discrete wavelet transform which do a better job of isolating the individual frequency spectra. The FFT is chosen for its computational efficiency.

While DMT is off to a slow start in the marketplace, it is expected to dominate for two reasons: it is thought to perform better for technical reasons and there is an ANSI standard behind it (not to mention Intel/Microsoft support).