Spectrum management for revenue maximisation in DSL

Digital Subscriber Line technologies use existing telephone cables to provide high speed data communications. Spectrum Management in DSL is a set of policies to contain the damage caused by interference between modems. Current practice in Spectrum Management controls this damage by placing constraints on power and spectrum. But recent advances in research define the constraints directly in terms of data loss. The spectrum adopted by interferers is allowed to be much more flexible, as long as the data rate targets are respected. These technologies are collectively called Dynamic Spectrum Management and have achieved huge gains in achievable data rates ([Yu (2002)] [Cendrillon et, al. (2003)]). The enforcement of data rate targets is done via discovering the set of feasible sets of data rates, the data rate region. But there remains a-problem left open on how to decide between alternative feasible sets of data rates. The algorithms in DSM typically find the best spectrum for a new user, given that pre-existing users have already been allocated data rate targets. The question of how to allocate these data: rate targets over time has been left unanswered. An additional complication is that the DSL channel is a MIMO channel that is only partially known, and gradually discovered. The crosstalk functions between different users and the direct channels are only discovered as new users request DSL, and this occurs gradually over time. This thesis introduces an economic focus to solve both problems. Among alternative, feasible set of data rates, the one with the best revenue potential is chosen. And the uncertainty about the MIMO channel and user take-up is included in a estimate of expected revenue. This is the first main contribution of this thesis, Economic Spectrum Management. The second contribution is a new, fast and near-optimal spectrum allocation algorithm, the Noise Floor Mask (NFM) algorithm. It has been developed as a faster alternative to the algorithm that achieves optimal data rates, Optimum Spectrum Management (OSM). OSM involves a more computationally expensive function optimisation and may present exponential complexity on the number of users, under certain conditions. Even for small number of users, NFM is fast and light, and thus useful for conducting Monte Carlo projections of the data rate region, given statistics on the telephone cable. These projections axe necessary in Economic Spectrum Management to compute expected revenues, given that the MIMO channel is not fully known. The NFM algorithm is based on a common insight among engineers about the "noise floor" as a relevant milestone when assessing how damaging a particular source of interference is. The rule of thumb is that interference sources less powerful than the background noise are significantly less relevant than those "above the noise floor". This intuition was developed and quantified, to form the basis of NFM.EThOS - Electronic Theses Online ServiceGBUnited Kingdo