EE535 - PROJECT 3
VDSL - Very High-Rate Digital Subscriber Lines

As mentioned in the introduction, VDSL is meant to provide the final link between a fiber network and the premises. VDSL is the technology that permits the transmission of data in a certain fashion, over some physical medium. The physical medium used is independent of VDSL. One possibility is to use the existing infrastructure of local loop wiring.

Although ADSL is quite likely to become widely used within the next few years, its use is mainly targeted at providing broadband service to the home over POTS wiring, over relatively long distances (18,000 feet on 25 AGW TP). VDSL on the other hand will operate at much shorter distances, and will in turn provide much higher data rates. VDSL is planned to be used together with a fiber network. The fiber network will be extended up to close to the residential areas. From there, the plain old telephone service wiring is used (thanks to VDSL) to transmit the information finally down to the home. Figure 2.1 shows a diagram of the configuration of a VDSL connection.

Figure 2.1. VDSL Setup.

Like all other xDSL technologies, VDSL provides a downstream channel and a upstream channel. These two channels do not use the same bit rate because it would cause echo. The downstream channel has usually a much bigger bit rate, which is also appropriate for the kind of applications that xDSL technologies will be used for: to provide a high data rate stream to the home.

VDSL will be quite similar to ADSL, but with higher data rates. ADSL has to face some problems that the concept of VDSL eliminates. These include the larger dynamic ranges that ADSL has to deal with, and the longer distances. For these and other reasons, the design of ADSL makes it more complex than VDSL. Telecommunications operators have pointed out that the cost is an important requirement. Therefore VDSL will be less complex and thus less costly.

This section is not titled "The VDSL Signaling Method" because there is no VDSL standard as of this date. There are several methods being considered to use with VDSL:

• CAP - Carrierless AM/PM, a version of suppressed carrier QAM.

  Carrierless amplitude modulation/phase modulation is based on Quadrature Amplitude Modulation and it works very similarly to QAM. A QAM receiver needs an input signal with the same spectral and phase relationships as the transmitted signal. Regular telephone lines do not guarantee this quality of delivery and therefore a QAM implementation for use with xDSL has to include adaptive equalizers that can measure the characteristics of the line and perform compensation for the distortion introduced in the twisted pair.

  A variation of QAM denominated CAP was developed by AT&T [Max]. It works very similarly to QAM: A transmit waveform is generated by applying each half-rate bit stream to two digital transversal bandpass filters with equal amplitudes but with a Pi/2 difference in their phase responses. The result is the same spectral shape as QAM. CAP seems to be somewhat more efficient compared to QAM with digital implementation. The references to CAP are scarce in the literature and it was not possible to obtain detailed description of CAP.

• DMT - Discrete Multitone, a multicarrier system using Discrete Fourier Transforms to create and demodulate individual carriers.

  This line code divides the available bandwidth in smaller units. These individual bands are tested to determine if they can be used to transmit information. This scheme is advantageous due to the wide range of line characteristics that can be found in the existing installation of twisted pair wire. Every installation can present differences in the quality and length of the line, and interference like crosstalk, and AM and HAM radio can affect the signal on these lines. DMT overcomes this problem by using those parts of the spectrum that offer less attenuation and interference.

  
  Fig 2.2. Discrete Multitone Spectrum.
  The line is tested to determine which frequency bands are available and how many bits can be transmitted per unit bandwidth [Max]. The bits are encoded in the transmitter with inverse FFTs and then passed to a D/A converter. At the receiving end, the signal is processed with an FFT to decode the incoming string of bits. ADSL also uses this line code, and splits the downstream channel into 256 tones of 4 KHz bandwidth and the upstream into 32 subchannels [Saa]. Each subchannel can carry a different number of bits, depending on the quality of the subchannel. DMT can operate in fixed-rate or adaptive-rate mode, i.e., it can use a constant data rate or it can modify the data rate during operation as a response to the characteristics of the line. However, the DMT suffers from sub-optimal subchannel isolation, as illustrated in Figure 2.3. The use of Fourier transforms introduces additional harmonics that do not carry information. The DWMT tackles this problem.
  
  Fig 2.3. Actual Discrete Multitone Spectrum.

• DWMT - Discrete Wavelet Multitone, a modified version of DMT.

  The Discrete Wavelet Multitone encoding scheme is based on the same idea as DMT, that is, to divide the channel into subchannels to make use of the sections of the frequency spectrum that are unaffected by interference. While DMT utilizes fast Fourier transforms to encode the bits in each subchannel, the DWMT uses wavelet transforms. The use of the digital Fourier transform to encode bits in the DMT algorithm generates harmonics along with the main lobes that represent the actual subchannel. This imposes greater restrictions in the design of a DMT receiver. However, the wavelet transform produces harmonics of lower energy (see Figure 2.4), which make it a simpler task to detect the encoded signal at the receiving end.

  
  Fig 2.4. Discrete Wavelet Multitone Spectrum.
  The SNR achieved with DWMT can be in the order of 43 dB, whereas DMT has a SNR of about 13 dB [Aware]. With DWMT, most of the energy is contained in the actual subchannels and is not lost so much in the additional harmonics that result from the transform operation.

• SLC - Simple Line Code, a version of four-level baseband signaling that filters the based band and restores it at the receiver.

The first VDSL modem by Orckit Communications uses DMT as the signaling method. DMT is the code being studied by the European and American VDSL committees as a potential standard. Orckit Communications is however pushing for the use of CAP in their newer versions of VDSL. DMT is the method used in ADSL, although several companies are asking the standards committees to release a second version of ADSL that uses CAP. DMT is the standard for ADSL because it can operate on more types of lines and over longer distances than CAP. It remains to be seen which signaling method will become predominant in the VDSL specification.

One aspect of the VDSL specification that is being studied is the bandwidth of the system. If the line code used for VDSL is CAP (a variant of QAM), then the bandwidth of the system maps directly to some value for a symbol rate [Kri]. The maximum throughput of the system can be maximized by selecting different symbol rates. The bit rate be given by the type of QAM used.

The noise in the channel imposes a limit on the symbol rate and the bits per symbol that can be used. The study done at GTE ([Kri]) assumes an asymmetric system, with a ratio of 10:1 in the data rates (downstream/upstream). In this scenario, the noise model assumed takes into consideration mainly the far-end crosstalk (FEXT). This noise source is a consequence of capacitive coupling among several twisted pairs in one same multipair cable. The other important noise source present in this medium is Gaussian noise, with an assumed two-sided spectral height of -140 dBm/Hz. Radio frequency interference (RFI) is also to be taken into account, although it is not clear how to quantify its impact on the transmission line. The assumed noise on the channel is approximated by the following equation:

Sn(f) = S10FEXT(f) = |P(f)|2 |C(f)|2 l f2 3-20  +  SAWGN(f)

where 10 neighboring twisted pairs are assumed. l is the line length in feet. P(f) is a filter applied to the symbols before transmitted and C(f) is the channel transfer function. Under this assumption, Krinsky et al. derive the SNR needed for a 16-QAM for an error rate of 10^-7 to be 21.5 dB, with an additional +/- 3 dB of SNR for every increase/decrease of 1 bit/symbol.

There are two different cases to consider in determining the symbol rate: short and long lines. Let us look at the short lines (300 - 600 meters) first. As the symbol rate is increased, the SNR decreases, but it does so in a way that the symbol rate increase is more relevant than the decrease in SNR. In these short loops, symbol rates of up to 15 Mbaud can be achieved. In the case of the longer loops (1 Km), however, as the symbol rate is increased, the decrease in SNR becomes the dominant factor for performance and rates of only 7.5 Mbaud are possible.

Therefore, using 16-QAM with 13 Mbaud over a 300 meter loop, the data rate obtained is 52 Mbps. For longer loops, however, it is not appropriate to lower the bits/symbol and maintain the same symbol rate. This baud rate is too high for long loops. For the 1Km loop it is best to use 16-QAM with 6.5 Mbaud, for instance, which provides 26 Mbps. The problem here is that a transceiver would need to implement two (or more) different symbol rates, which is a less attractive option than implementing 16-QAM and 4-QAM, for instance.

Reference [Sch] discusses several methods to improve the quality (and thus the throughput) of the transmission by introducing echo caller circuits and also noise cancelation to minimize the effects of NEXT (near-end crosstalk) and FEXT (far-end crosstalk). The authors of this paper consider two VDSL implementations that use CAP (Carrierless Amplitude Modulation) and PAM (Pulse Amplitude Modulation) respectively. The choice of PAM has the advantage that this baseband transmission scheme makes use of the lower frequency bands, which are less subject to noise (attenuation and crosstalk). On the other hand, CAP can allow to use POTS (voice service) or ISDN simultaneously with VDSL. The following graph shows a comparison of the transmission capacity of VDSL using CAP and PAM. The number of disturbers is the number of twisted pairs in the same multipair cable that can be interfering with each other if carrying VDSL signals as well.

Figure 2.2. Comparison of transmission capacity for PAM and CAP based VDSL.

Back to the Very High Speed Digital Subscriber Lines
Project Homepage. HW3 - Spring 97.

Back to the WPI EE535
Telecommunications and Transmission Technologies
Course Homepage