Transformation Based Downlink Beamforming Techniques for Frequency Divison Duplex Systems

1. Introduction Adaptive antennas at the base station are one of the key-technologies to satisfy the increasing capacity demand of already existing and future cellular mobile communication networks. Several sophisticated uplink algorithms have been proposed for signal separation and detection. But downlink beamforming seems to be the pivotal point for the applicability of adaptive antennas in cellular mobile communication systems. Obviously, a similar performance gain is required in up-and downlink to achieve the possible capacity increase. The downlink beamforming process is very challenging, especially in frequency division duplex (FDD) systems, because of the uncorrelated fading in up-and downlink and the frequency dependent antenna array response [1]. Thus the direct reuse of the uplink antenna weights for downlink transmission is not advisable [2]. One well known beamforming approach for FDD systems uses the estimated discrete DOAs (direction-of-arrival) to calculate the complex antenna weights for downlink transmission [3]. But in case of a large angular spread (AS) as in urban environments, the robustness of high-resolution DOA estimation methods is substantially degraded [4]. Moreover, the nulls of the antenna pattern directed towards the interferers have to be broadened to keep the produced interference as low as possible. This requires adequate null broadening dependent on the angular spread of the interfering signals [5]. Beamforming algorithms based on spatial covariance matrices adjust the nulls of the antenna pattern perfectly to the actual propagation conditions and therefore outperform high-resolution DOA-based methods by far [6,7]. But the covariance matrices have to be transformed from uplink (receive) to downlink (transmit) frequency because of the different antenna array responses. In this paper compare our Azimuth Power Spectrum based Spatial Covariance Matrix Frequency Transformation (ASCOFT) [7] with other proposed methods [8, 9] that also transform the spatial information included in the spatial covariance matrix from receive to transmit frequencies.