====== MU-MIMO ====== Created Tuesday 10 June 2008 ===== C802.16mDL_MIMO-08/004_ZTE_ZTE-USA_r1 ===== ZTE The 802.16m system supports MS diversity combining techniques such as: selective combining, switched combining, equal gain combining, maximal-ratio combining and interference rejection combining. The 802.16m system supports Cyclic delay diversity (CDD). ===== C802.16m-08/399r1 ===== Toshiba corporation 11.1.3.3.1.1.2. Non-unitary precoding Vender dependent linear beamforming weight is used such as Zero forcing or MMSE. The channel state information used for calculating the weight is fed back from MS in both TDD and FDD mode. It is also obtained by the sounding in TDD mode. 11.x.3.3.1.2. Non-linear Precoding Vender dependent nonlinear beamforming is used such as Multiuser THP or Vector perturbation. Both the schemes use Modulo operation to combat with the increase in transmit power in the linear precoding. Therefore, the Modulo operation is also required at the MSs. The channel state information used for calculating the weight is fed back from MS in both TDD and FDD mode. It is also obtained by the sounding in TDD mode. ===== C802.16mDL_MIMO-08/004_Huawei ===== Huawei  Non-unitary MU-MIMO If the assemble precoding matrix could be unitary or non-unitary, it is defined as non-unitary MU-MIMO, which relaxes the orthoganlity limitation of the precoding vectors. The precoding vectors with lower cross-correlation could also be scheduled together. Since Non Unitary precoding is a more general and less restricted case of unitary precoding, the gains could possibly be higher, but possibly at the cost of higher control overhead. In non unitary based precoding where ZF-beamforming is one example, the MS reports a quantized version of the channel based on some known codebook. The BS then collects all the quantized channels and constructs a precoding matrix as a function of the quantized channels of the scheduled users. This could lead to large DL control overhead and this is why dedicated pilots are used to indicate this information. However, it is also true that under certain conditions the ZF action of the BS can be made transparent to the MS and no additional control overhead is needed. Furthermore, since the non-unitary MU-MIO is more sensitive to CSI quantization error, which can be partly solved by quantizing in an iterative manner. To fulfill the potential of MU-MIMO, the codebook specified for MU-MIMO shall be defined as [TBD]. To diminish unnecessary and redundant feedback, users grouping in MU-MIMO should be done in multiple stages. At each stage, BS should broadcast user grouping related information to prevent MS feeding back redundant and/or unnecessary CSI. To efficiently support large codebook size and to reduce CSI feedback overhead, hierarchical codebook should be supported. In this case, the codebook is composed of a base codebook and a localized codebook. The system shall follow the progressive refinement procedure to quantize the channel state information in hierarchical manner. The forward control channel shall include a field for feedback size and a field indicating the feedback type which could be full CSI or partial CSI.  Nonlinear MU-MIMO It assumes that the transmitted signal is generated by a non-linear combination of data symbols contained in a vector, which requires higher computational complexity and more signaling overhead but can provide a better performance than linear techniques. 11.x.4.2. MIMO for Multi-cast Broadcast Services 11.x.4.2.1. Simultaneous transmission of MBS and unicast data To improve system resource utilization and capacity on the downlink, the BS shall super-impose unicast data for a MS on MBS data in a sub-frame. This superposition of the unicast data with MBS data shall involve splitting the BS transmit power between the unicast and MBS data, as illustrated in Figure 2. The BS TX power shall be split such that the MBS data gets a greater proportion of the transmit power than the unicast data. This power ratio between MBS and unicast data powers shall be semi-statically configured depending on the available power and the desired rate. Once the power ratio is determined, the unicast symbols are linearly superimposed on the unicast symbols in the frequency-domain before IFFT at the BS. An example illustration is shown in Figure 2. The power scaling factors gb and gu shown in the figure shall be computed using the power ratio. Figure 2 Additional unicast traffic transmitted in parallel with E-MBS providing significant unicast capacity advantage with little or negligible reduction in E-MBS capacity ===== C802.16mDL_MIMO-08/004_Samsung ===== Samsung Figure 3 An example of MBS/Unicast superposition At the MS, the broadcast data is first demodulated and decoded using channel estimates via MBS reference signals. The successfully decoded MBS data is then reconstructed using the broadcast channel estimates. The reconstructed broadcast signal is cancelled from the overall received signal. The reconstruction of the broadcast signal using the overall composite channel estimate assures that all the broadcast interference including broadcast interference from neighboring cells to the unicast traffic is cancelled. The resulting clean unicast signal that is free from any broadcast interference is then further processed for unicast traffic demodulation and decoding. The MSs that subscribe to the MBS data alone will treat the underlying unicast data as noise to decode MBS data. 11.x.4.2.2. Spatial multiplexing of MBS and unicast data The unicast data and MBS data can be superposed and spatially multiplexed. Two illustrations for MIMO unicast/MBS superposition are shown in Figure 3 and Figure 4. The example shown in Figure 3 uses a single-stream transmission for the MBS traffic while two-stream MIMO transmission is considered for the unicast traffic. In the second example shown in Figure 4, 2-streams are employed for transmitting both MBS and the unicast traffic. ===== C80216mDL_MIMO-08_004_NEC ===== NEC Adding to "Draft 1: Table of Contents for SDD Text on Downlink MIMO Schemes C80216mDL_MIMO-08_004" Based on the contribution C80216m-08_333, we propose adding the following text to the draft1: ----------------------- beginning of the proposed text -------------------------------------------- 11.x.3.2.2.1.1 Multi-resolution precoding ----------------------- end of the proposed text ------------------------------------------