Uplink to downlink spatial covariance matrix transformation methods for downlink beamforming of a UMTS-FDD system

We analyze the effect of different uplink to downlink spatial covariance transformation methods on the system level performance of a smart-antenna based UMTS-FDD system. We propose a new transformation technique that provides robustness against uplink spatial signature estimation errors in a frequency selective fading environment. Our approach leads to a matrix robust least squares (RLS) problem that can be formulated in a convex form as a second order cone (SOC) program and solved efficiently using the well-established interior point methods. The performance of the proposed method is compared with those of K. Hugl et al. (see IEEE Globecom´99, p.2097-101, 1999) and T. Aste et al. (see Proc. ICASSP. p.3313-16, 1998) using a dynamic system level simulator developed for investigating the capacity improvement of the UMTS-FDD system with smart antennas at the base station (BS). Downlink capacity is used as a measure of system performance by finding the number of users for which a certain quality of service (QoS) requirement can be fulfilled. System level simulations carried out for 4-element uniform circular arrays (UCA) show that our method outperforms both Hugl et al. and Aste et al. without any significant degradation in performance when compared to Aste et al. with a known uplink spatial signature.