DSL Spectrum Management—The UK Approach

Written on 6:50 AM by ooe

The issues

DSL systems do not live in isolation from their neighbours. We have seen in Chapter 3 that in fact there is crosstalk between systems operating in the same part of the access network, and that this crosstalk can be categorised into two important types, near end crosstalk (NEXT) and far end crosstalk (FEXT). The distinction is important because these two types of crosstalk limit available transmission capacity in different ways. Duplex transmission suffers from the more severe NEXT but benefits from duplicate use of bandwidth for each direction of transmission while the less severe FEXT allows a higher throughput but only in one direction at a time or frequency.

It was shown that a critical frequency could be found below which it was favourable to transmit duplex against NEXT and above which it was favourable to transmit dual simplex against FEXT.

Crucially, though, this critical frequency was found to be loop-length dependent. The longer the loop the lower the critical frequency becomes. This is important because in most real access networks not all loops are the same length and indeed loops of different lengths often share the same cables and so crosstalk between them may well occur. This fact blurs the concept of the critical frequency to cover a band. The band stretches from the critical frequency for the longest loops in the cable, below which duplex transmission is beneficial for all, to the critical frequency for the shortest practicable loops above which dual simplex transmission is beneficial for all. Between these two limits is a band of uncertainty where optimal transmission for some loops may cause sub-optimal transmission for others in the same cable.

Transmission capacity would be wasted if duplex transmission were used in the band best for dual simplex or vice versa. However, any plan will inevitably be a compromise for the loops' different lengths, and in the band of uncertainty some waste is inevitable, at least for some of the loops.

In the duplex band the rate of transmission is substantially symmetrical as all of the bandwidth is available in both directions all of the time. However, in the dual simplex band a choice has to be made for the direction of transmission for which each part of the band will be used. It would be possible to choose that all the band is used for transmission in just a single direction, making transmission very asymmetric. Alternatively interleaved sub-bands of similar width in each direction could be used so that the net effect is nearly symmetric transmission. But once this choice is made in a cable, all the DSL systems in that cable at least would have to use the same band allocations else where there is conflict NEXT will be created and transmission compromised. The choice defines the symmetry of all transmissions on the cable, and once set cannot be easily changed unless all the services in the cable are changed at once. Clearly the choice is a far reaching one and once made is unlikely to be altered in the future as it would affect so many customers.

A way to look at the complete picture for the cable is that the set of all transmission systems operating in the cable jointly create the dominant common noise environment (due to crosstalk) they each suffer. Introducing a new system type into the mix of deployed systems can change the joint noise environment and affect all the systems in the cable. In fact even a single variant system could have a drastic adverse effect if it causes crosstalk in a frequency range that had previously been comparatively quiet and exploited by already deployed systems for receiving weak signals.

The expected noise environment obviously affects both the viability of existing deployed systems and also the deployability of future systems. Changes to the noise environment could deny service to existing customers where systems had previously been deployed with a different noise environment expectation. Such changes could also adversely affect the deployability of future systems and hence even the commercial viability of a whole class of technology. A careful plan is required to control ongoing deployments to safeguard existing services and preserve the commercial value of deployments still being planned or envisaged.

It is interesting to consider what would happen if such a plan did not exist. Suppose that any kind of transmission at any frequency were permitted at any point in the access network. This would mean that no system operator could realistically expect to operate any part of the spectrum in a dual simplex, FEXT-limited mode, since he would have to plan for the likely event that other pairs in the network would be used to transmit in the opposite direction at the same frequency, causing NEXT. As was seen in Chapter 3, FEXT limited transmission is preferred on long pairs (~5km) at all frequencies above about 50 kHz and above about 200 kHz there is no NEXT limited capacity at all on these pairs. This would effectively condemn the user with a long wire pair to very low transmission speeds. Meanwhile, on a very short wire pair, NEXT limited transmission is effective up to 1 MHz or so and much greater transmission capacity is available. So the effect of the lack of a plan is detrimental to all users of the cable, but disproportionately detrimental to those that happen to have longer pairs. While there will always be an advantage in receiving service via a shorter pair, a good plan helps to maintain a better compromise between short and long pair transmission possibilities, as well as giving potentially greater throughput for all.

In fact a complete lack of any kind of spectrum plan could have an even more drastic adverse effect in that the greatest transmission capacity would go to those systems that used the loudest transmissions. This is because they gain a signal to noise ratio advantage, through having the largest transmit signal, while still only suffering the common crosstalk noise environment. This advantage is obtained at the expense of all the neighbouring systems since the loud transmitter raises the common noise environment disproportionately, and their weaker transmissions lose signal to noise ratio. Retaliation is possible through the affected systems also raising their transmit signal levels, further raising crosstalk levels. This process is sometimes known as the ‘cocktail party effect’. Signal levels all go up but no net gain is obtained, just a lot of chaos as systems that previously functioned cease functioning because they can no longer transmit loud enough.

Breaking this cycle requires a plan. The plan has in some way to constrain what transmission levels and frequencies may be used in what parts of the access network. This will inevitably itself restrict the deployability of some system types in some locations, but the plan will provide some level of confidence in the levels of crosstalk noise that may in the future exist in any part of the access network and so banish the chaos and the cocktail party effect and replace it with a comparatively stable and predictable environment that operators can use to plan their deployments and so that customers can enjoy continued and uninterrupted service.

This plan is spectrum management. Spectrum management must somehow constrain what kinds of signal can be injected into the access network, and hence the systems that inject them, and where they are located. If it does not do this it will not be effective.

It should be noted that once such a plan exists it will rapidly become immutable, or nearly so. This is because the existing plan, as has been described, sets the effective noise environment for all existing systems and the business cases for future systems. Any attempt to change the plan will meet with resistance from nearly all existing players whose deployments or planned deployments would or could be adversely affected. As a result it seems unlikely that the plan could ever be significantly changed once agreed and acted upon, unless there were some overriding outside reason to do so that affected the majority of the players.

A technology that can have a profound effect on spectrum management is that of the repeater. A repeater is an active device that is inserted in a pair to amplify the signal being sent so as to overcome the effects of signal attenuation along the loop. By doing this the adverse effect of crosstalk from other nearby systems can be drastically reduced. The disadvantage of the repeater is that although it offers an advantage to the repeatered pair, systems operating on adjacent pairs without repeaters are disadvantaged by the increased crosstalk caused by the repeater itself. So to deploy repeaters is to need many of them. From a spectrum management perspective repeaters cause a shift in the critical frequencies of the repeatered pair favouring duplex transmission but also introduce a large number of extra points in the access network where signals are being injected. As we have seen that spectrum management is about controlling and planning signals injected into the access network plainly the use of repeaters drastically complicates spectrum management planning.

One thing a spectrum management plan does not do is set out the deployment rules per se that must be used by operators. It may bound the deployment options through the limits it sets but as crosstalk noise is statistical in nature, variously in terms of the distribution of systems being used in a given cable, in terms of the individual crosstalk coupling between pairs, and even in terms of the imprecise estimates of loop characteristics on which deployment decisions must be made, the crosstalk expected even in a spectrally managed network is still a statistical quantity. In order to obtain a very low risk that crosstalk noise will adversely affect service deployment it would be necessary to be very conservative with deployment rules, while a more bullish approach might deploy with a higher risk of failure on much longer loops. This choice and the deployment rules derived from it remain the domain of the operators.

If you enjoyed this post Subscribe to our feed