Capacity improvement of CDMA systems by adaptive beam forming using GRNN techniques

Abstract

The objective of this work is to demonstrate a method of improving the capacity of a Code Division Multiple Access (CDMA) system by employing Neural Network based Adaptive Beam forming. A Generalized Regression Neural Network was used for this purpose. First, the Neural Network was designed, which could accurately predict the phase angle of the feed current to a ten-element antenna array in order to form a directional beam towards a given signal source direction while forming a null towards a given interfering source. Then, the model was developed to form multiple beams to wards different signal sources. Next, using the multiple beam forming technique, a new Space Division Multiple Access (SDMA) model was developed to improve the capacity of an existing CDMA system that has already become saturated. This SDMA model is based on the statistical distribution of the mobile users within a sector. It assumes that the user distribution within a sector is non-uniform. More densely populated areas within a sector are identified and the users in a particular area are grouped together. One such group is called a cluster. Similarly, a number of clusters are selected and the three most populated clusters are served with three isolated directional beams operating on the same frequency achieving SDMA. It was observed that this new SDMA model could improve the capacity of existing CDMA systems up to a maximum of 20% with three directional beams.