R-FurtherperformanceevaluationforCSIfeedbackupdatedvclear.docVIP

PAGE20 / NUMPAGES20 - 3GPP TSG RAN WG1 #62 Meeting R1-105017 Madrid, Spain, 23rd August –27th August 2010 Title: Further performance evaluation for CSI feedback enhancement proposals  MACROBUTTON MTEditEquationSection2 Equation Section 1 SEQ MTEqn \r \h \* MERGEFORMAT  SEQ MTSec \r 1 \h \* MERGEFORMAT Source: Alcatel-Lucent Shanghai Bell, Alcatel-Lucent Agenda Item: 6.3.3 Document for: Discussion 1. Introduction Implementing two-stage feedback for Rel-10 downlink MIMO is agreed in RAN1#60bis [1] and RAN1#61 [2]. Various two-stage feedback solutions have been proposed in recent LTE-A discussions. Basically, there are three two-stage feedback approaches: Adaptive feedback (see for example [3]), i.e. spatial correlation matrix based feedback: The main idea of this approach is to match the quantization codebook to specific propagation scenario. Differential feedback (see for example [4]): targets for aligning the signals towards the direction of the user, while is a diagonal matrix performing one-step differential. Cluster based feedback [5, 12-14]: In this approach, is a block diagonal matrix denoting a cluster composed of 4 DFT beams chosen over wideband and executes beam selection and co-phasing for each subband. We believe these approaches should be compared by simulation results in different propagation scenarios before taking the final decision. In this contribution, system-level evaluation results are provided to compare these feedback approaches for both 4Tx and 8Tx with cross-polarized and co-polarized antenna configurations. 2. Two-stage feedback approaches 2.1 Adaptive feedback Adaptive feedback, i.e. spatial correlation matrix based feedback, was firstly proposed in [6, 7] and introduced to 3GPP RAN1 discussion on UE feedback in [8]. Many companies have proposed various solutions for implementing two-stage adaptive feedback in Rel-10, e.g. [3, 9-11]. In the following, we clarify our solution of how to implement two-stage adaptive feedb