Resumen
The demand for multimedia services that can simultaneously support multiple image, video, voice, and data traffic has resulted in an explosive growth in the telecommunications industry. The consequence of this applications-driven market is an increasing need for connectivity and greater bandwidth. This need is driving the industry to provide high-speed data networks with bandwidth capacities to satisfy the full-scale delivery of video, voice and data. Electronic commerce, telemedicine, teleconferencing, online banking, knowledge acquisition through the electronic delivery of educational and training materials, video-on-demand or near-video-ondemand, home entertainment, electronic home shopping and a host of other applications are finding their way into data networks. Digital subscriber line (DSL) technology offers an attractive solution for providing high-speed communications over existing copper cables. These cables are already installed, cheap to acquire and affordable compared to more recent systems. These are the main reasons why DSL technology has an important share of current broadband markets worldwide. It is realized that the main impairments affecting the performance of data transmissions in copper cables is the crosstalk among the lines. Dynamic spectrum management (DSM) was developed to combat this obstacle by allowing adaptive allocation of spectrum to various users in a multiuser environment as a function of the physical channel or by joint coordination on the signal level, so called vectoring, taking into consideration channel conditions. This technology can contribute to the development of future data networks that will be more reliable, provide increased data rates and reduce energy consumption. Energy efficiency will become more important with emerging policies for environment protection against ’greenhouse’ gases, renewable energy sources and actions against climate change. This thesis presents a new approach to dynamic spectrum management (DSM) and investigates alternative techniques that can improve the performance of the DSL systems. Any new algorithm needs accurate knowledge of the transmission channel characteristics that can be obtained by measurements and modeling. Well established models, that explored cable characteristics up to 30 MHz, exist for conventional DSL systems. Therefore, we decided to focus our attention on measurements and modeling of copper cables for extended bandwidth. We performed measurements of transfer and coupling functions using bandwidth up to 200MHz and demonstrated that it is possible to exploit this bandwidth for short copper cables. Measurement was a complicated task due to number of measurements needed to perform and measurement set up preparations including calibration process for high frequencies. In order to provide models we extrapolated already existing models to incorporate high frequencies and showed that they match reasonably with the measurements results. Intending to improve the models by parameter fitting we developed a new approach that included phase modeling for coupling functions. This is very important for further evaluation vectoring techniques. Moreover, this investigation opened a path for further research and development of new generation of DSL systems using short copper cables already installed in homes, buildings and offices. Since the cables are mostly made of copper they are very good antennas, i.e, they radiate and pick up electromagnetic waves. Protecting the cables with shield can substantially relax the electromagnetic compatibility issues. Thus, we considered that it would be interesting to explore shielded twisted pair (STP) cables. Certainly that this cables have been investigated before, but for the state-of-the-art DSL technology that is using voltage difference between two wires of a twisted pair for signaling. Taking into consideration the shield as joint common to all the other wires in the cable we formed wire-shield common mode configuration. Applying the basic principles of the multiconductor transmission line theory (MTL) on this configuration we derived new model for a cable seen as multiple-input-multipleoutput (MIMO) channel that is becoming more important with emerging vectoring schemes. The model was verified with measurement results that included measuring of model parameters. The measurement process had complicated set-up and huge number of measurements were performed. Using this model we evaluated the capacity that can be achieved by applying MIMO techniques to such configuration in the presence of radio ingress. Comparing to conventional techniques revealed that the capacity can be doubled. Of course, the shield effectiveness has big impact on the capacity and with good shield this result can be obtained as demonstrated in this thesis. In upstream direction users that are using longer loops may experience severe performance degradation due to high power transmitted by users that are closer to the central office. The solution to this problem is proposed in the standard in the form of power back-off (PBO) where closer users reduce the power that they are transmitted. Actually, received power for all users should follow the reference power that is a parametrized function of frequency. However, there is a little guidance of how these parameters should be calculated. Previous works optimized these parameters for a country or region, assuming worst-case noise scenario and for each user. First approach results in bad performance and the later has high complexity since it needs the knowledge of all coupling functions. Therefore, we used the power back-off parameters optimized for a particular cable bundle to improve data rates. Furthermore, by noting that all users receive the same power we showed how normalized coupling functions can be used to omit the need for the knowledge of each coupling function. Moreover, the algorithm is capable to incorporate the coupling functions that are actually present in the cable and avoid model assumptions. Because the algorithm is taking into account the actual situation in the network it is performing DSM by using standardized parameters. Other DSM algorithms need additional guidance on their implementation in order to take into account the interoperability issues. Demonstrated by simulations the improvements are significant compared to worst-case scenarios. Allocating power to different sub-carriers in multi-carrier system by waterfilling policy is not optimal because the distribution of the input symbols is assumed to follow the Gaussian distribution. This is not the case for practical systems where inputs are obtained by different modulation techniques. Particularly, DSL systems use quadrature amplitude modulation (QAM) with equally probable symbols. Nevertheless, bit loading algorithms proposed for practical systems, although incorporates discrete nature of different modulations, depend on the capacity equation for Gaussian distribution corrected with Shanon gap approximation. These algorithms claim the optimality but it was recently demonstrated that mercury/waterfilling performs optimal power allocation knowing the input distribution ’a priori’. In this thesis we demonstrated that using the same bit distribution as obtained by bit loading algorithms based on capacity equation and applying mercury/waterfilling better performance in bit error rate (BER) can be achieved. This can be used for reducing the noise margin or making the system more reliable. Using this bit distribution as starting point and applying mercury/waterfilling policy we developed new bit loading algorithm that improves the system throughput by searching for a solution of a corresponding combinatorial optimization problem, constrained by the same power and BER restrictions as the previously developed. This optimization problem is very complex and hard to solve, therefore, we decided to use the greedy approach. This new algorithm is not restricted to the knowledge of the bit distribution in advance but rather uses the mercury/waterfilling with bit distribution obtained in each iteration. Consequently, the same problem can be formulated as power minimization and we have developed the algorithm that reduces the needed power while satisfying the same rate and BER constraint as algorithms using gap approximation. Simulations revealed that improvements are higher for longer cables. These algorithms can be beneficial to the operators in either improving the data rates that they can deliver or by reducing operating costs and thus, making their networks more energy efficient. Parts of this thesis were done within European projects BANITS2 and MUSE. During the stay at Ericsson AB IT Technologies, Stockholm, Sweden all the measurements were performed and the training in The Telecommunications Research Center Vienna (Das Forcshungszentrum Telekommunikation FTW), Viena, Austria contributed to better understanding of Dynamic Spectrum Management (DSM) algorithms. Colaborating with Telef´onica I+D within several internal projects led to the development of some chapters and better knowledge of operators needs. Furthermore, parts of this work has led to new European project named 4GBB that will deal with opened questions regarding new generation of DSL systems for short copper loops according to the Fiber To The Near Home (FTTNH) architecture.