The choices

Written on 6:52 AM by ooe

Having established that to be successful a spectrum management plan must somehow restrict what can be deployed and where, the question remains how should it do this? There are probably as many ways of doing this as there are pairs in the target network to be managed, however there are some broad strategies that can be described:

1. limiting the specific types of system (e.g. by references to standards) and where they may be deployed;

2. constraining the spectrum of deployed systems with hard power spectral density (PSD) masks for given situations;

3. setting out an algorithm by which the interference properties of the deployment of a given DSL at a given location can be evaluated and the relative interference between it and other systems assessed in a consistent way and placing limits on the consequences (e.g. the amount of range reduction permitted by comparison with some reference case for all the systems concerned).

The first approach, (a), has the merit of being extremely simple to describe and lay down. However it is supremely inflexible. It says nothing about how systems other than those that are specifically described in the plan may be used and by implication such systems are excluded even though they may be very similar disturbers to other systems that have been described.

In the same way such a plan is almost certainly doomed to become dated very quickly as new DSL types emerge which were not known when the plan was devised.

But perhaps the most damning problem with such an approach is that it favours the described technologies over all others and so could be considered directly discriminatory against a technology that is not included in the plan but is in every way a more moderate disturber than another system which is. This could make such a plan open to accusations of being an unfair barrier to trade.

At the other extreme the last approach, (c), is ultimately flexible. By using the algorithm the limits that must be placed on a DSL of any given properties should be determinable. New, non-standard, modified or propriety systems can be evaluated without any need to revise the management plan. No system is unfairly excluded.

However, just because of its ultimate flexibility, it is almost inevitably also very complex. The algorithm will be complicated to develop and agree between the parties concerned. In order to be even approximately effective and fair to all concerned the resulting algorithm will be very complicated to evaluate or even interpret. This means that there will be much scope for errors in the algorithm itself or in its interpretation in a given case. And any attempt to police such a plan would be thwarted by the difficulty the policing authority would have in interpreting and applying the rules.

The second approach, (b), is a compromise between the two extremes. It does not explicitly include or exclude any particular technology. It is perfectly possible and in fact quite simple to determine if or where a new, non-standard or proprietary system can be deployed and so is reasonably future proof. And yet the basic method is reasonably simple to specify and understand and does not involve any complex algorithms.

It still has a trace of the disadvantages of method (a) in that it is very rigid; any failure to comply with the mask is a simple plan failure. There is no room for compromise as such a compromise would require some trading algorithm that would quickly lead in the direction of method (c). This inflexibility is actually not as severe a barrier as it may seem since minor transgressions of the plan masks can be addressed by reducing the transmit power of the offending system slightly; most modern DSL systems have the flexibility to do this.

Method (b) also has a trace of the disadvantage of (c) in that compliance with the plan is more difficult than simply quoting a standards reference. The PSD must be measured in a prescribed way and compared with the plan masks. This may sound simple but measuring PSD is a far from simple process especially if accurate calibration is required on which significant commercial risks are at stake.

Within the three broad categories of approach to spectrum management plans set out above there are a number of choices to be made about the scope of application of the plans.

Geographical application can be national, regional or even by exchange area or ultimately even by individual cable. In other words a single plan could be devised that would apply to all regions nationally, or separate plans could apply to different geographical parts of the network. Obviously there is a practical limit to which this can be taken but there might be justification in having regional plans based on the nature of the human geography of the area such as rural, urban, suburban, residential, commercial. For example a plan suitable for use in a commercial area might make use of bandwidth in a more symmetrical way (e.g. for general data exchange applications) while a plan for use in a residential area might have a more asymmetrical flavour (e.g. for broadcast type applications). Having developed just a few such plans a country or other geographic area could be divided up into parts and one of the predetermined plans assigned to each segment based on some measure of its geography.

Application to pairs in a network can also be either uniform or categorised across cables. Crosstalk between pairs in a cable is greatly influenced by their physical separation within the cable build. This means that if a subset of pairs is carefully chosen DSL deployments on those pairs could enjoy a lower resulting noise environment than if deployments were unplanned or allowed on all pairs. Any corresponding spectrum management plan could then take this lower level of crosstalk into account in its design. This mode of deployment though is dependent on network cable pair arrangements being maintained and recorded along the length of the cables, and does inherently restrict the penetration of DSL services. Once the spectrum plan was set any attempt to increase the plan's preset penetration at a later date would lead to difficulties with crosstalk into previously deployed systems.

Another way of categorising pairs would be by length. Different length pairs could then have different parts of the management plan assigned to them. This approach could help with the ‘band of uncertainty’ issue by allowing different uses of bandwidth on different length pairs.

A final approach that is worth mentioning is that of assigning various spectrum management plans at different times, e.g. at different times of day. This could address the commercial/residential issue by for example having a highly symmetric plan for the daytime and an asymmetrically biased plan during the evening. This has been discussed but not adopted anywhere to the authors' knowledge. 4.3

If you enjoyed this post Subscribe to our feed