Downlink eigenbeamformer with combining of eigenbeams

For W-CDMA several closed-loop Tx diversity concepts employing 4 antenna elements at the base station (BS) are currently discussed to be included in 3GPP release 5. Among these, the downlink eigenbeamformer effectively uses long-term as well as short-term channel properties. More specifically, it is based on an eigenanalysis of the long-term spatial covariance matrix at the mobile station (MS). The eigenvectors (vector of complex weighting factors for all antenna elements) with the largest eigenvalues (largest average signal power) are determined and fed back step by step to the BS. This process takes place on the same time scale as the MS physical movement. Accordingly, required operations in the MS for eigenanalysis as well as required long-term feedback bits for representation of eigenvectors are distributed over a very large number of transmission slots. As an extension of simple switching between eigenbeams a linear combination of several eigenbeams is considered. This combining is performed using constellation rotation and progressive refinement techniques applied to the eigenbeam space. Performance of combining methods is compared with an upper bound (beamforming based on perfect short-term knowledge of the downlink channel) and uplink based beamforming methods where only the long-term spatial covariance matrix is known at the BS. Although combining methods require more short-term feedback information then simple switching and thus are more sensitive to feedback delay and restricted feedback rate, they prove to be beneficial up to a vehicle speed of 40 km/h