The PCN/cordless/mobile communications industry has experienced an extraordinary growth in the last few years, with prediction of a worldwide subscriber population of 900 million by the year 2010. To successfully cater for this huge market, new technologies must be developed to greatly improve the system capacity within the constraint of the limited availability of the RF spectrum. One way of increasing the system capacity is to adopt more sophisticated frequency re-use scheme via the use of smart directional base-station antennas that can automatically adapt themselves to the changing mobile environment to minimise the interference from adjacent channels, co-channels, etc. For a DSP-based phased array where the signal from each array element is downconverted to a very low frequency for digitisation and then directly processed to realise various beam-forming algorithms, a parallel-processing architecture represents the most choice. This paper focuses on the implementation of such an architecture based on the TMS320C40 platform.