BER Performance Improvement in UWA Communication via Spatial Diversity

Abstract

In present era while wireless communication has become an integral part of our life, the advancements in underwater communications (UWA) is still seem farfetched. Underwater communication is typically essential because of its ability to collect information from remote undersea locations. It don’t use radio signals for signal transmission as they can propagate over extremely short distance because of degradation in signal strength due to salinity of water, rather it uses acoustic waves. The underwater acoustic channel has many characteristics which makes receivers very difficult to be realized. Some of the characteristics are frequency dependent propagation loss, severe Doppler spread multipath, low speed of sound. Due to motion of transmitter and receiver the time variability and multipath makes underwater channel very difficult to be estimated. There are various channel estimation techniques to find out channel impulse response but in this thesis we have considered a flat slow fading channel modeled by Nakagami-m distribution. Noise in underwater communication channel is frequency dependent in nature as for a particular range of frequency of operation one among the various noise sources will be dominant. Here they don’t necessarily follow Gaussian statistics rather follows Generalized Gaussian statistics with decaying power spectral density. The flexible parametric form of this statistics makes it useful to fit any source of underwater noise source. In this thesis we have gone through two step approach. In the first step, we have considered transmission of information in presence of noise only and designed a suboptimal maximum likelihood detector. We have compared the performance of this proposed detector with the conventional Gaussian detector where decision is taken based on a single threshold value and the threshold value is calculated by using various techniques. Here it is being observed that the ML detector outperforms the Gaussian detectors and the performance can be improved further by exploiting the multipath components. In the second step we have considered channel along with noise and have designed a ML detector where we have considered the receiver is supplied with two copies of the same transmitted signal and have gone through a two-dimensional analysis. Again we compared the performance with conventional maximal ratio combiner where we can observe the ML detector performance is better. Further we have incorporated selection combining along with these detectors and compared the performance. Simulation results shows that the proposed detector always outperforms the existing detectors in terms of error performance.