The capacity of VDSL2 technology – constraints and potential
As IPTV becomes a reality and the demand for higher access rates increases, operators and end users often ask themselves what are the true limits of existing copper infrastructure and what rates are achievable over an ordinary twisted pair?
This article elaborates on the constraints and potential of the latest DSL technology, VDSL2 (ITU G.993.2), including extensions and successors. It also discusses how new technology can handle most of today’s limiting factors to achieve even higher rates on available copper networks.
Gigabit-per-second speeds are already available on copper wires (for instance, Gigabit Ethernet) but the transmission is limited to very short distances and requires four copper pairs of CAT5 or better quality. A good summation would be “the rate depends on the wire” as someone once cleverly stated; yet let us approach the problem in a more structured way.
From information theory we know that the channel capacity, as defined by Claude Elwood Shannon, ultimately sets the theoretical transmission limits for any particular channel or media. This means that signal strength, bandwidth and noise floor are determining factors of the attainable rate.
If we limit the problem to twisted copper pairs available for DSL (VDSL2 being the latest standard) in our public networks, we find that the copper network and the standards have specific properties that put additional limitations on system performance.
In a public copper-access network, we find cable bundles of twisted wire pairs of up to several kilometers in length and with a diameter of 0.4-0.5mm. They generally also have a plastic coating of polyethylene. Therefore, besides signal strength, bandwidth and noise floor, we must also consider attenuation, crosstalk in the cable and immunity to external noise.
The Shannon-Hartley theorem, which applies to an additive white Gaussian noise channel with B Hz of bandwidth, and a signal-to-noise ratio S/N, gives: C = B log2 (1 + S/N), where C (capacity) is the rate in bits per second, assuming B (bandwidth) in Hertz and a S/N (signal-to-noise ratio) expressing a power ratio in dB. The theorem tells us that available bandwidth, together with the strength of the signal and the strength of the noise, will determine the attainable rate (capacity). Let us see how this translates into the copper channel.
Available bandwidth and standards
Bandwidth – that is, the available spectrum for transmission – is given by standards and regulation as well as the physical properties of the copper cable. The transceiver standard for VDSL2 is based on discrete multitone (DMT – similar to orthogonal frequency-division multiplexing, OFDM) technology, where data is carried by a given number of tones with constant spacing. The duplex method is frequency-division duplex (FDD). This means that the upstream and downstream directions (comparable to a wireless uplink and downlink) are divided into different frequency bands. The standard stipulates that each tone can carry up to 15 bits of information. Several profiles have been defined for the allocation of spectrum, resulting in a variety of ratios (symmetries) for the upstream and downstream directions. The 17MHz VDSL2 profile (17a) uses 4095 tones, with an effective symbol rate of 4kHz. This gives the following total system capacity: 4000 DMT symbols/s * 4095 tones/symbol * 15 bits/tone = 246Mbps to be shared between the two directions. In the same way, the VDSL2 profile for 30MHz (30a) has an effective symbol rate of 8kHz using 3478 tones, which gives 417Mbps. To comply with the VDSL2 standard, the required minimum system capability of bi-directional net data rate is 100Mbps for a 17MHz system and 200Mbps for a 30MHz system. Whether capable or not, real systems have natural constraints and impairments that lower system capacity.
Signal strength and attenuation
As stated by the Shannon-Hartley theorem, the strength of the received signal influences capacity. The power of the signal sent out on the available spectrum determines how much of the signal will remain after attenuation by the wire pair. Standards and regulation set limits by defining the maximum allowed total transmit power (dBm) and allowed power spectral density (PSD). This ensures system capacity without disturbing other systems that use the same or similar frequencies (“self” or “alien” systems that could be on adjacent wire pairs). From an environmental aspect, power should not be wasted, and power efficiency is essential.
Attenuation is a property of any transmission media. Attenuation on twisted pairs is dependent on the thickness (gauge or diameter) and the length of the wire pair (American Wire Gauge 24 is equivalent to European wires of 0.5mm in diameter; AWG26 to 0.4mm). Thicker and shorter wires have lower attenuation. Attenuation is also frequency-dependent; higher frequencies experience more attenuation than lower frequencies (tones). This means that the distance to the customer will determine whether some of the allocated tones will be attenuated below a useful level (below the noise) and can carry information. Higher frequency bands are the first to suffer from increased length, because attenuation reduces the available spectrum. In other words, the signal strength is given by
• maximum allowed signal power; and
• how the signal is attenuated by the media in the respective spectrum.
For VDSL2 systems that transmit over normal polyethylene-insulated pairs, there is a good approximation of the attenuation A (in dB) at a certain frequency f:
A = k * l * sqrt(f), where k is a wire-dependant constant and l is the length of the wire.
The available spectrum increases distinctly the closer you are to the end user. Whereas the ADSL2 standard (ITU G.992.3) is designed for longer distances (up to approximately 6km, working up to 2MHz), the VDSL2 standard is designed for shorter distances (up to approximately 2km, working up to 30MHz).
Fiber-deep architectures benefit from decreased attenuation of the copper lines, due to their shorter length. Copper wires close to the end user are usually also thicker than present-day Central Office-based deployments. Deployments from the cabinet (300-800m from the end user) are generally wires with a diameter of 0.5mm (AWG24).
It is not unusual for two or four pairs to be pulled out from the cabinet for potential end users; bonding technologies (ITU-T G.998) have been standardized to aggregate rates from several physical pairs.
Physical properties are difficult to change and the general view today is that distance is the limiting factor for copper lines. However, in most cases, the limiting factor is actually noise.
Signal-to-noise ratio (SNR), noise sources and noise mitigation
Noise can be regarded as external signals that interfere with the transmitted signal.
SNR is a measure of transmission quality. The higher the ratio, the less obtrusive the noise. Together with the Shannon capacity theorem, this can be interpreted as follows: the higher the signal relative to noise, the more room there is for transmission. But if the signal is attenuated excessively or there is too much noise, no practical data transmission is possible.
There are numerous noise sources. Some affect the entire spectrum, some affect parts of the spectrum, some are stationary and some are non-stationary with impulsive appearance (impulse noise).
The source of impulse noise can be anything from home appliances to elevators or electrical fences. Impulse noise is difficult to handle because it appears suddenly and often affects a large part of the spectrum, giving rise to burst errors. Impulse noises are often characterized by their duration, by being repetitive or by other properties. Experiments show that new technologies, such as physical retransmission (PhyR), will handle most of these problems.
Stationary noise relates to more stationary sources, such as
• crosstalk (x-talk) from adjacent lines in the cable binder;
• radio frequency interference (RFI); or
• background noise (white Gaussian noise) – for example, thermal noise induced by components in the receiver (typical level about -140 dBm/Hz).
The relative stability of the source of disturbance often makes it possible to compensate for different types of stationary noise, such as RFI, by employing different notching techniques, for example, to avoid “bad” parts of the spectrum. Ordinarily, the most severe stationary noise is crosstalk, of which there are two kinds depending on the relative locations of disturbing transmitter and receiver:
• near-end crosstalk (NEXT); and
• far-end crosstalk (FEXT).
NEXT can be described as trying to listen to someone who is far away while your neighbor (standing right next to you) is shouting to someone else. In most cases, NEXT can be handled with frequency-division duplex (FDD), by separating the downstream and upstream frequencies.
FEXT can be described as trying to listen to someone who is far away while that person’s neighbor is shouting in the same direction, to your neighbor. The FEXT signal is subject to frequency-dependent attenuation by the cable; FEXT is thus an increasing problem over shorter distances with decreased attenuation, and when higher frequencies are used. Fortunately, crosstalk-cancellation technology is being developed that effectively mitigates most of this noise. A variety of names are used for this technology, including vectoring, dynamic spectrum-management level 3 (DSM L3) and MIMO for DSL. The upcoming standard for this technology in ITU-T is called G.vector.
There are several additional standardized techniques to manage spectrum and allocate power – for example, dynamic spectrum-management (DSM L1 & L2), virtual noise (VN), bit-swapping, and seamless rate adaptation (SRA). Indeed, new technologies are being developed and standardized to handle almost any harmful noise situation for DSL.
Future capacity on DSL lines
While it is always difficult to predict what the next step in the evolution of DSL will be, Ericsson is driving a European research initiative to find out. Ericsson wants to understand what systems in combination with fiber can reach distances of 20-200m from the end-user. What are the suitable transmission schemes? What does one gain from additional spectrum (which then becomes available for transmission, see Figure 1) and what is the best way to use this spectrum provided one can handle the crosstalk?
|Figure 1. Simulation with the following assumptions:|
Max bits: 15 bits/tone. Frequency: < 300MHz. Noise floor: -130 dBm/Hz. PSD: -60 dBm/Hz up to 30 MHz and -80 dBm/Hz above 30 MHz, no crosstalk. Max power: 10 dBm.