Characterization and Modeling of Power Line Communication Channels

Nowadays, we live in a highly interconnected world, where the information exchange is the milestone that has pushed and drives the incredible growth experienced by our society. This huge amount of generated data traffic will be expected to grow even further in the next few years. In order to meet these future requirements and the customers needs, the telecommunications systems must evolve and improve. Although twisted pair copper and wireless technologies represent an established and widespread solution, extensively adopted for today's communication systems, they cannot achieve these goals alone. The desire to provide a new technology that would be able to exploit what was already existing and deployed, has led to the development of the power line communication (PLC) systems. The PLC technology exploits the existing power delivery infrastructure in order to deliver high speed and reliable data communication. Although the energy distribution grid is worldwide widespread, being able to reach each consumer around the world, it has not been conceived for data transmission at high frequencies. This leads to a challenging transmission environment, characterized by strong attenuation, high selectivity, with significant fading effects, and a great deal of noise. Despite these detrimental effects, a reliable and secure high speed communication can be established. However, in order to design and develop the best next generation devices, able to overcome the environment and medium limitations, a thoroughly knowledge and a comprehensive analysis of the PLC channel is fundamental. All the collected information enables the development of new and effective models, able to faithfully describe a real communication scenario, allowing to save money and time in the development process. The aim of this thesis is to provide a detailed PLC channel characterization, considering several environments belonging to both the indoor and the outdoor environments. Among the indoor scenarios, that identify a confined communication environment, the in-home, the in-car and the in-ship networks are analyzed. Furthermore, concerning the outdoor scenario, both the low voltage (LV) and medium voltage (MV) distribution grids are investigated. The properties of the noise that typically affects the different PLC environments are assessed. The study mainly focuses on the broadband frequency spectrum (BB-FS), although some comparisons with the narrowband frequency spectrum (NB-FS) are also tackled. The extensive analysis of the PLC network, performed for each considered scenario, allows to highlight the main properties and relationships that enable the development of new and effective channel models, herein described. The first part of this work deals with the characterization of the in-home single-input single-output (SISO) PLC channel, considering a large database of measurements that was carried out during a measurement campaign performed in Italy. A statistical analysis is presented in terms of normality test, channel frequency response (CFR) distribution and phase behavior. Moreover, the main and most commonly used statistical metrics, namely average channel gain (ACG), root-mean-square delay spread (RMS-DS) and coherence bandwidth (CB), are assessed and compared to each other in order to highlight the typical relationships. Then, the performance that can be achieved are computed, inferring the relation between the geometrical distance and the maximum achievable rate, as well as the improvements due to the bandwidth extension. Furthermore, the correlation that can be experienced by channels belonging to the same network, but connecting different nodes, is discussed. Finally, the line impedance is investigated, identifying the main existing relationships between the resistive and the reactive part. Then, the attention is moved to the multiple-input multiple-output (MIMO) transmission scheme, still within the in-home environment. The signal transmission over all the three commonly available wires, typically deployed in domestic premises, is discussed and explained. The CFR statistics and characteristics are assessed basing on the channels collected by the special task force (STF) 410 of the ETSI during an experimental measurement campaign across Europe. Also the noise properties are discussed, providing a novel method to simulate the noise correlation among the spatial receiving modes, which is based on the actual noise measurements provided by the STF-410. In addition, the general processing system is discussed, describing the optimal transmission and reception scheme under two hypothesis, full channel state information (CSI) knowledge and no CSI knowledge. The performance, in terms of maximum achievable rate, are computed for both the cases. Afterwards, the focus is turned to in-vehicle communications, where two typical application scenarios, namely car and ship, are discussed and compared. In particular, a fundamental distinction is made among the electric car (EC) and the conventional car (CC) scenarios. Both channel and noise measurements that we carried out on a compact electric car are analyzed. The results are compared to an online available database of measurements concerning a conventional fuel car. A similar analysis is tackled considering measurements that we acquired on a large cruise ship. Finally, the two environments are compared in terms of statistical metrics relationship and average performance. Later, the outdoor communication context is addressed, investigating and comparing LV and MV distribution grids in both the narrow-band (NB) and broad-band (BB) spectra. The characterization of this transmission scenario is of fundamental importance since it has become very attractive, especially in recent times, for smart grid applications. The properties of one of the most known outdoor LV database and two MV set of measurements, that were carried out in two completely different network sites, are firstly discussed. Then, a comparison in terms of path loss, line impedance and background noise is tackled for the LV and MV networks in both the NB and BB frequency ranges. Some features of the main network devices usually deployed in the outdoor networks are also discussed. The final comparison, in terms of achievable rate, aims to assess if some improvements can be obtained by exploiting the BB-FS instead of the NB-FS. In order to provide an overall overview and a general comparison among all the previously discussed environments, the main channel properties and the statistical metrics relationships are shown together and distinctly discussed. Also the differences in terms of background noise properties are summarized. The channel and noise knowledge allows the computation of the SISO channel capacity distribution for all the environments, which is later compared to the MIMO channel capacity distribution for the in-home scenario. Two different noise types are considered for the MIMO capacity computation, namely spatially uncorrelated and correlated, relying on the ETSI STF-410 noise measurements. The second part of this work focuses on the channel modeling, which represents a quick and easy testing tool for the development and the simulation of standards and devices. This translates into a considerable saving in terms of costs and time, avoiding on-field measurements. All the detailed information previously collected, as well as the highlighted relationships, are the foundations on which we develop an extremely synthetic channel model, able to faithfully emulate a real MIMO PLC channel with a reduced set of parameters. The proposed model consists of a pure top-down approach, without any physical connection. It is extremely synthetic since it is able to numerically generate channel realizations, that are equivalent to the measurements, simply basing on the CFR amplitude and phase statistics and on their corresponding properties, such as the correlation exhibited among the frequencies, as well as between the different spatial modes. A final fundamental aspect is also taken into account. In a world based on a massive and constant flow of information, where usually the communications are private and confidential, ensuring the security of the exchanged data is of paramount importance. Towards this end, the concept of secure data communication is introduced focusing on the secrecy granted at the physical level, named physical layer security (PLS). As it is discussed, this concept completely differs from the secrecy ensured at the application layer, thus, exploiting encryption protocols through the use of keys. The main concepts underlying the secure data communication are discussed. In this respect, the wiretap channel is defined, evaluating the performance that can be achieved among a transmitter and an intended receiver, without releasing any information to a third counterpart. This quantity is known as secrecy capacity. The main differences between the conventional capacity and the secrecy capacity, and among the wireless and PLC scenarios, are highlighted. Furthermore, the detrimental influence on the secrecy capacity due to the typical PLC channel phenomena, such as the frequency selectivity and the multiple user (MU) correlation, as well as to the network topology, which gives rise to what is known as keyhole effect, are assessed. Afterwards, a typical MU broadcast scenario is considered, discussing and computing the achievable secrecy rate region under a total power and a quality of service (QoS) constraints. Finally, in an attempt to overcome all the PLC scenario limitations, a MIMO transmission scheme, a bandwidth extension, as well as a more fair background noise assumption are assessed, relying on experimental channel and noise measurements. The achievable secrecy rate is computed considering both an alternating optimization (AO) algorithm and a uniform power allocation approach. The results are compared to the SISO case, as well as to the conventional capacity, without secrecy constraints