Research On Several Key Problems Of Ultra-wideband Systems

In recent years, ultra-wideband (UWB) communication has drawn considerable attention from both the research community and industry. Extremely short period pulses have been employed in this technology (usually as short as several nanoseconds or even several picoseconds), which results in the transmission capability of up to gigabits per second [1-4]. Because of the high transmission rate, UWB communications are best suited for wireless personal area networks (WPAN). Meanwhile, because of the characteristics of UWB, the performances of many traditional communication techniques (such as diversity, modulation and so on) change a lot when they are applied in the UWB communication systems, which results in the necessity of re-evaluating the performances of these techniques.The main contributions of this thesis are composed of the following three parts:1. A bit error rate (BER) analysis of an asynchronous DSSS-BPSK UWB system in dense multipath channels by means of standard Gaussian approximation (SGA) and simplified improved Gaussian approximation (SIGA) is presented. The standard UWB channel (proposed by IEEE 802.15.3a task group) is employed in the analysis to reflect the actual channel conditions. Several key parameters that affect the BER performance are obtained by means of SGA. On the basis of SGA, the validity of improved Gaussian approximation (IGA) method is established in our systems and the accuracy of the bit error rate expression obtained via SIGA is demonstrated with Monte-Carlo simulations. Besides, the impacts of several different kinds of transmitted pulses on BER performance are investigated with numerical examples. Further, we discuss the design of optimal pulse in the context of this thesis.2. According to the characteristics of UWB standard channel models, we propose the concept of"composite lognormal random variables (RVs)". Based on this new concept, a new and precise approach is proposed to approximate the statistical distribution of lognormal RVs'sum, which places no restriction on the distributional parameters and correlation between the two branches. Then, we...