Login| Sign Up| Help| Contact|

Patent Searching and Data


Title:
METHOD AND PRECODER INFORMATION FEEDBACK IN MULTI-ANTENNA WIRELESS COMMUNICATION SYSTEMS
Document Type and Number:
WIPO Patent Application WO/2011/140062
Kind Code:
A1
Abstract:
A method for two component feedback in wireless communication systems is disclosed, with a wireless communication device sending a first representation of a first matrix chosen from a first codebook, wherein the first matrix has at least two columns, the wireless communication device sending a second representation of a second matrix chosen from a second codebook, and the base station obtaining a precoder from the first representation and the second representation.

Inventors:
SAYANA KRISHNA KAMAL (US)
ZHUANG XIANGYANG (US)
Application Number:
PCT/US2011/034959
Publication Date:
November 10, 2011
Filing Date:
May 03, 2011
Export Citation:
Click for automatic bibliography generation   Help
Assignee:
MOTOROLA MOBILITY INC (US)
SAYANA KRISHNA KAMAL (US)
ZHUANG XIANGYANG (US)
International Classes:
H04L25/03; H04B7/06
Foreign References:
US61331818A
Other References:
LG ELECTRONICS: "DL Codebook design for 8Tx precoding", 3GPP DRAFT; R1-102380 DL CODEBOOK DESIGN FOR 8TX PRECODING, 3RD GENERATION PARTNERSHIP PROJECT (3GPP), MOBILE COMPETENCE CENTRE ; 650, ROUTE DES LUCIOLES ; F-06921 SOPHIA-ANTIPOLIS CEDEX ; FRANCE, vol. RAN WG1, no. Beijing, china; 20100412, 8 April 2010 (2010-04-08), XP050419735
ALCATEL-LUCENT SHANGHAI BELL ET AL: "Development of two-stage feedback framework for Rel-10", 3GPP DRAFT; R1-101859, 3RD GENERATION PARTNERSHIP PROJECT (3GPP), MOBILE COMPETENCE CENTRE ; 650, ROUTE DES LUCIOLES ; F-06921 SOPHIA-ANTIPOLIS CEDEX ; FRANCE, vol. RAN WG1, no. Beijing, china; 20100412, 8 April 2010 (2010-04-08), XP050419713
HUAWEI: "Extensions to Rel-8 type CQI/PMI/RI feedback using double codebook structure", 3GPP DRAFT; R1-100251 EXTENSIONS TO REL-8 CQIPMIRI FEEDBACK USING DOUBLE CODEBOOK STRUCTURE, 3RD GENERATION PARTNERSHIP PROJECT (3GPP), MOBILE COMPETENCE CENTRE ; 650, ROUTE DES LUCIOLES ; F-06921 SOPHIA-ANTIPOLIS CEDEX ; FRANCE, vol. RAN WG1, no. Valencia, Spain; 20100118, 12 January 2010 (2010-01-12), XP050417935
MOTOROLA: "3GPP Draft; R1-103328, Two Component Feedback Design and Codebooks", 10 May 2010 (2010-05-10), Montreal, Canada, pages 1 - 7, XP002655795, Retrieved from the Internet [retrieved on 20110803]
Attorney, Agent or Firm:
BOWLER, Roland K., et al. (Libertyville, Illinois, US)
Download PDF:
Claims:
C L A I M S

1. A method for a wireless communication device to send a precoder matrix information to a base station, the method comprising:

the wireless communication device sending a first representation of a first matrix chosen from a first codebook, wherein the first matrix has at least two column vectors;

the wireless communication device sending a second representation of a second matrix chosen from a second codebook;

wherein the first representation and the second representation together convey a precoder matrix of one or more vectors associated with one or more spatial layers, the precoder matrix comprises a first sub-precoder matrix including a first set of weights on a first subset of transmit antennas of the base station, and a second sub-precoder matrix including a second set of weights on a second subset of transmit antennas of the base station, wherein the first and second subsets of transmit antennas are non-overlapping,

wherein the first sub-precoder matrix is one or more column vectors of the first matrix corresponding to the first representation, multiplied by one or more entries of the second matrix corresponding to the second representation, and the second sub-precoder matrix is one or more column vectors of the first matrix corresponding to the first representation, multiplied by one or more entries of the second matrix corresponding to the second representation.

2. The method in Claim 1, wherein the first matrix has at least two column vectors, the at least two column vectors are different.

3. The method in Claim 1, the wireless communication device sending the first representation of the first matrix chosen from the first codebook, wherein the first representation is an index of the chosen matrix in the first codebook.

4. The method in Claim 1, the wireless communication device sending the second representation of the second matrix chosen from the second codebook, wherein the second representation is an index of the chosen matrix in the second codebook.

5. The method of Claim 1, the wireless communication device sending the first representation of the first matrix chosen from the first codebook, wherein the first codebook includes a set of matrices whose column vectors are DFT vectors.

6. The method of Claim 1, the wireless communication device sending the second representation of the second matrix chosen from the second codebook, wherein the second matrix has at least two entries that have different and non-zero amplitude.

7. The method of Claim 1, the wireless communication device sending the second representation of the second matrix chosen from the second codebook, wherein the second matrix has the same vector norm of each row of the second matrix.

8. The method of Claim 1, the wireless communication device sending the second representation of the second matrix chosen from the second codebook, wherein the second matrix has the same vector norm of each column of the second matrix.

9. The method of Claim 1, the wireless communication device sending the second representation of the second matrix chosen from the second codebook, wherein the second matrix comprises entries that are co- phasing factors associated with the first sub-precoder matrix and the second sub-precoder matrix.

10. The method of Claim 9, wherein the co-phasing factors are based on a DFT codebook.

11. The method of Claim 1, where the second matrix is represented as , where p is a real value and Θ is angle in

radians in the range [0, 2π]

12. The method in Claim 1 further comprising

the wireless communication device selecting one column vector of the first matrix;

the first sub-precoder matrix associated with one spatial layer is obtained by multiplying the selected column vector with a first element in a first row of the second matrix;

the second sub-precoder matrix associated with one spatial layer is obtained by multiplying the selected column vector with a first element in a second row of the second matrix.

13. The method in Claim 1 further comprising the wireless communication device selecting one column vector of the first matrix;

the first sub-precoder matrix associated with two spatial layers is obtained by multiplying the selected column vector with first and second elements in a first row of the second matrix;

the second sub-precoder matrix associated with two spatial layers is obtained by multiplying the selected column vector with first and second elements in a second row of the second matrix.

14. The method in Claim 1 further comprising

the wireless communication device selecting first and second column vectors of the first matrix;

the first sub-precoder matrix associated with two spatial layers is obtained by multiplying the first column vector with a first element in a first row of the second matrix and multiplying the second column vector with a second element in a first row of the second matrix;

the second sub-precoder matrix associated with two spatial layers is obtained by multiplying the first column vector with a first element in a second row of the second matrix and multiplying the second column vector with a second element in a second row of the second matrix.

15. The method in Claim 1 further comprising

the wireless communication device selecting first and second column vectors of the first matrix;

the first sub-precoder matrix associated with two spatial layers is obtained by multiplying the first column vector with a first element in a first row of the second matrix and multiplying the second column vector with a second element in a first row of the second matrix; the second sub-precoder matrix associated with two spatial layers is obtained by multiplying the second column vector with a first element in a second row of the second matrix and multiplying the first column vector with a second element in a second row of the second matrix.

16. The method in Claim 1, further comprising

forming an alternative form of the second matrix from entries of the second matrix;

obtaining a product matrix from multiplication of the first matrix and the alternative form of the second matrix;

obtaining the precoder matrix by mapping columns of the product matrix to the first and second sub-precoder matrix.

17. The method in Claim 16, where the alternative form of the second matrix is obtained by mapping entries of the second matrix such that one entry is present in each column of the alternative form of the second matrix, with remaining entries of each column being zero.

18. The method in Claim 1 further comprising rotating at least one vector of the first matrix according to a rotation matrix selected from a rotation codebook.

19. The methods of Claim 18, wherein the rotation codebook is a DFT codebook.

20. The method of claim 1, the wireless communication device sending the first representation of the first matrix chosen from the first codebook, wherein the first matrix is chosen from a subset of the first codebook.

21. The method of claim 1, the wireless communication device sending the second representation of the second matrix chosen from the codebook, wherein the second matrix is chosen from a subset of the second codebook.

22. The method of claim 1 further comprising the wireless communication device determining a number of column vectors in the precoder matrix, wherein the number of column vectors corresponds to a number of spatial layers.

23. The method of Claim 1 further comprising

the wireless communication device sending the first representation of the first matrix, wherein the first matrix corresponds to a frequency band;

the wireless communication device sending the second representation of the second matrix, wherein the second matrix corresponds to a frequency sub-band within the frequency band;

wherein the first representation and the second representation together convey the precoder matrix of one or more vectors associated with one or more spatial layers, and the precoder matrix corresponds to the frequency sub-band.

24. The method of Claim 18 further comprising the wireless communication device rotating at least one column vector of the first matrix according to a rotation matrix chosen from a rotation codebook, wherein the rotation matrix corresponds to a sub-band.

Description:
METHOD OF PRECODER INFORMATION FEEDBACK IN MULTI-ANTENNA WIRELESS COMMUNICATION SYSTEMS

CROSS REFERENCE TO RELATED APPLICATIONS

[0001] The present application is a non-provisional application of U.S. provisional Application No. 61/331,818 filed on 5 May 2010, the contents of which are incorporated herein by reference and from which benefits are claimed under 35 U.S.C. 119.

FIELD OF THE DISCLOSURE

[0002] The present disclosure relates generally to wireless communications and, more particularly, to a feedback framework in wireless communication systems.

BACKGROUND

[0003] In wireless communication systems, channel state information at a transmitter, for example, at a base station, is important for beam-forming transmissions (also referred to as precoding) that deliver more power to a targeted user while minimizing interference on other users. Precoding operations can be in the context of single-user multiple input multiple output (SU-MIMO) or multi-user MIMO (MU-MIMO), where two or more users are served by a single base station. An eNB needs accurate spatial channel information in order to perform a high rank transmission to a single UE or to perform precoding to two or more UEs simultaneously so that the mutual interference among multiple transmissions can be minimized at each UE. [0004] Precoding operations may also be in the context of SU/MU-

MIMO users served by coordinated multi-point (CoMP) transmissions where antennas belonging to different eNBs, rather than to the same eNB, can coordinate their precoding to serve multiple users simultaneously. Further support for up to eight transmit antennas is enabled in the next generation cellular standards like 3GPP LTE Release-10. Due to such a relatively large number of antennas (4-Tx or 8-Tx) involved in such transmissions, it is desirable that the UE feedback be designed efficiently with good performance overhead trade-off, so that feedback does not scale linearly with the increasing number of antennas.

[0005] The antenna configurations which support a large number of antennas in practice must allow large beamforming gains and also larger spatial multiplexing gains achieved from higher rank transmission. Beamforming allows efficient support for low geometry users and also for multi-user transmission thereby improving cell-edge and cell-average throughput with larger number of users in the system, while spatial multiplexing allows higher peak spectral efficiency. A typical antenna configuration to achieve this would be to have groups of antennas where each group is a set of correlated antennas and each group is uncorrelated with the other groups. A cross-polarized antenna configuration is one such setup. The correlated antenna elements provide the required beamforming gains and the uncorrelated antenna elements enable high rank transmissions.

[0006] The above structure in the antennas has some unique spatial characteristics that can be exploited. For example, the correlation among correlated antennas changes slowly and is confined to a smaller vector space on an average. This can be used to feedback the correlated and uncorrelated channel characteristics, i.e., two components, at different rates and/ or with different levels of quantization/ overhead in time and frequency to reduce feedback overhead. One of the components representing the correlated channel characteristics can be fed back on a wideband basis and/ or slowly in time, while the other component is fed back on a subband basis and / or more frequently in time.

[0007] However, one of the key challenges in designing such a two component feedback system is identifying the parameters used in the two components and the construction of the final precoder matrix as a function of the two components.

[0008] The various aspects, features and advantages of the invention will become more fully apparent to those having ordinary skill in the art upon a careful consideration of the following Detailed Description thereof with the accompanying drawings described below. The drawings may have been simplified for clarity and are not necessarily drawn to scale.

BRIEF DESCRIPTION OF THE DRAWINGS

[0009] FIG. 1 illustrates a wireless communication system.

[00010] FIG.2 illustrates an embodiment with a base station transmitting to a device.

[00011] FIG.3 illustrates an example of a frame structure used in the 3GPP LTE Release-8 (Rel-8) specification and different reference symbols. [00012] FIG. 4 illustrates exemplary antenna configurations at a base unit.

[00013] FIG. 5 illustrates a first subset of antennas and a second subset of antennas transmitting two spatial layers to a device.

[00014] FIG. 6 illustrates a wideband and subbands, each of which is further composed of contiguous subcarriers.

DETAILED DESCRIPTION

[00015] In FIG. 1, a wireless communication system 100 comprises one or more fixed base infrastructure units 110 andl20 forming a network distributed over a geographical region for serving remote units in the time and/ or frequency domain. The base infrastructure unit may also be referred to as the transmitter, access point (AP), access terminal (AT), base, base station (BS), base unit (BU), Node-B (NB), enhanced Node-B (eNB), Home Node-B (HNB), Home eNB (HeNB) or by other terminology used in the art. The base units are generally part of a radio access network that includes one or more controllers communicably coupled to one or more corresponding base units. The access network is generally communicably coupled to one or more core networks, which may be coupled to other packet or data networks, like the Internet, and to public switched telephone networks (PSTN), among other networks. These and other elements of access and core networks are not illustrated but they are well known generally by those having ordinary skill in the art.

[00016] The one or more base units each comprise one or more transmitters for downlink transmissions and one or more receivers for receiving uplink transmissions from the remote units as described further below. The one or more base units serve a number of remote units, for example, remote unit 102 and 104 in FIG. 1, within a corresponding serving area, for example, a cell or a cell sector of the base unit, via a wireless communication link. The remote units may be fixed units or wireless communication devices. The remote unit may also be referred to as a receiver, subscriber station (SS), mobile, mobile station (MS), mobile terminal, user, terminals, user equipment (UE), user terminal (UT) or by other terminology used in the art. The remote units also comprise one or more transmitters and one or more receivers. In FIG. 1, the base unit 110 transmits downlink communication signals to serve remote unit 102 in the time and/ or frequency domain. The remote unit 102 communicates directly with base unit 110 via uplink communication signals.

[00017] The term "transmitter" is used herein to refer to a source of a transmission intended for receipt by a user or receiver. A transmitter may have multiple co-located antennas each of which emits, possibly different, waveforms based on the same information source. In FIG. 1, for example, antennas 112 and 114 are co-located. A transmitter is typically associated with a cell or a cell sector in the case of a base unit having or serving multiple sectors. Also, if a base unit has geographically separated antennas (i.e., distributed antennas with remote radio heads), the scenario is also referred to as "a transmitter". Thus generally one or more base units transmit information from multiple antennas for reception by a remote unit.

[00018] In the diagram 200 of FIG. 2, at 210, a base unit transmits from a plurality of antennas. Also in FIG.2, a remote unit receives transmissions from a plurality of antennas, which may or may not be co-located. In a typical embodiment, a base unit may be associated with a cell-ID, by which it identifies itself to a remote unit. As a conventional mode of operation, also sometimes referred to as a single-point transmission scheme, a remote unit 240 receives transmissions from a plurality of antennas of a single base unit 210. Such a base unit is also referred to as a serving cell (or serving base unit) to the user device / remote unit.

[00019] In one implementation, the wireless communication system is compliant with the Third Generation Partnership Project (3GPP) Universal Mobile Telecommunications System (UMTS) Long Term Evolution protocol, also referred to as Evolved Universal Terrestrial Radio Access (EUTRA), or some future generation thereof, wherein the base unit transmits using an orthogonal frequency division multiplexing (OFDM) modulation scheme on the downlink and the user terminals transmit on the uplink using a single carrier frequency division multiple access (SC-FDMA) scheme. In another implementation, the wireless communication system is compliant with the IEEE 802.16 protocol or a future generation thereof. More generally, however, the wireless communication system may implement some other open or proprietary communication protocol where channel feedback is useful or desired. Thus the disclosure is not intended to be limited to or by the implementation of any particular wireless communication system architecture or protocol. The teachings herein are more generally applicable to any system or operation that utilizes multiple antennas in a transmission, whether the multiple antennas belong to a single base unit or to multiple base units or whether the multiple antennas are geographically co-located (e.g., belong to a single base unit) or distributed (belong to either remote radio heads or multiple cells). [00020] In a general embodiment, pilots or reference symbols are sent from each antenna in a transmitter. These pilots occupy the operational bandwidth to allow users to estimate the channel state information (CSI) of the entire bandwidth. Typically the pilots from different antennas are orthogonal so the pilots do not interfere with each other. Such orthogonality can be ensured if the pilots are sent using different time and / or frequency resources or code resources. For example, in systems based on OFDM technology, the pilots can occupy different subcarriers in frequency or different OFDM symbols in time or share the same set of resources, but different code sequences.

[00021] In FIG. 3 illustrates a frame structure used in the 3GPP LTE Release-8 (Rel-8) protocol to illustrate a possible reference symbol (RS) pattern in an OFDM system. A subframe 310 in a radio frame 302 spans 14 OFDM symbols in time. Further a subframe 310 contains multiple resource blocks 312, each spanning 12 consecutive subcarriers in frequency. In typical OFDM based systems like 3GPP LTE, a block of consecutive OFDM symbols are referred to as a subframe. Each sub-carrier location in each of the OFDM symbols is referred to as a resource element (RE), since a single data modulation symbol can be mapped to such a resource element. A resource block (RB) is defined as a block of REs comprising a set of consecutive sub- carrier locations in frequency and a set of symbols. In LTE Rel-8, a slot is defined to span 7 symbols and each subframe is made of two slots, and hence 14 symbols. A minimum resource unit allocated to a user is the two RBs corresponding to two slots in a subframe for a total of 2x12x7 REs. A resource block may be more generally defined as a set of resource elements/OFDM subcarrier resources in time and frequency domain. [00022] Some of the REs in a RB are reserved for reference symbols (also referred to as pilots) to help in the demodulation and other measurements at the UE. These reference symbols, as defined in Release 8 specification of LTE can be further divided into two types. The first type is cell-specific reference symbols, which are cell-specific and "common" to all users, and are transmitted in all the RBs. A common reference symbol (CRS) may or may not correspond to actual physical antennas of the transmitter, but CRSs are associated with one or more antenna "ports", either physical or virtual. In FIG. 3, as an example only, RE 304, 305, 306, 307, 308 and 309 may be a CRS. The second type is user-specific or a dedicated reference symbol (DRS), which are user-specific and hence applicable only to that user, and allocated in the RB's allocated to that user's data. Furthermore, DRS typically correspond to "precoded" or beam-formed RSs, which can be directly used by a user for the demodulation of the data streams. The precoding operation is explained later. In FIG. 4, as an example only, RE 320, 325, 330, 335, 340, 345, 350 and 355 may be a DRS. In LTE Release-10, a new spare RS, namely CSI-RS are defined to enable channel measurements, while DRSs are primarily relied upon for demodulation. These can be used similar to CRSs in LTE Release- 8 to derive channel feedback information.

[00023] The location of the reference symbols is known to the UE from higher layer configurations. For example, depending on the number of antenna ports as configured by a transmission unit, UE knows the location of all the reference symbols corresponding to all configured antenna ports. As another example, when a UE is instructed to use a DRS, the UE also knows the DRS locations, which may depend on the user identification. [00024] In typical FDD operation of a LTE Rel-8 system, CRSs are used for both channel related measurements at the UE and also for demodulation. If eNB employs a precoder at the transmitter, such information is made available to the UE, which allows it to construct the channel for demodulation based on the CRSs. In a FDD operation of a future LTE Rel-10 system, CSI-RS (and possibly CRSs that may still be available) may be used for channel related measurements, while DRSs are used for demodulation. Hence an eNB may apply precoder which are not exactly the same as the UE feedback, and does not have to signal the precoder explicitly. This is further described in detail later.

[00025] The "precoding" operation is explained in the following. The base station transmits a signal via weighting each antenna signal with a complex value, an operation referred to as precoding, which may be mathematically represented by the matrix equation:

[00026] Y=HVs + n

[00027] in which, when transmitting one spatial layer of data, or rank-1, may be represented as:

[00029] in which, when transmitting two spatial layers of data, or rank-2, may be represented as:

[00031] where y x ... y NR may be the received data at the UE receive antenna

#1 to #NR, respectively. In the example with a rank-1 transmission, or a transmission with one data stream denoted as "s", the Matrix V may be a precoding vector with weights v u ... ν Ντ 1 for base station transmit antenna #1 to #NT respectively. In an embodiment with a rank-2 transmission, or a transmission with two data streams si and s2 on the same subcarrier, V may be a precoding matrix. Precoding vector and precoding matrix can be referred to as precoding matrix given vector is a degenerated case of matrix.

[00032] Matrix H may be the propagation channel matrix between transmit antennas and receive antennas with entry representing a channel between the jth transmit and ith receive antennas. Value n may represent noise and interference. The precoding weights V, either a vector or matrix, may be determined by the base station, typically based on the channel particular to the UE or can be UE-specific and may also take into account a preference indicated by feedback from the UE. Further the matrix HV can be referred to as the effective channel between a user's data streams and its receivers. The effective channel, instead of the propagation channel H, is all a UE needs for demodulation purposes. The precoding weights may or may not be constrained to a predefined codebook that consists of a set of pre-defined vectors or matrices. In an embodiment with constrained precoding, the precoding matrix may be signaled by the base unit efficiently with a precoding matrix index (PMI) or with an index to a precoding matrix within a predefined codebook. The term "matrix" in this context may include the degenerated special case of vector, which applies to single stream transmission. In the most generic sense, the term "precoding" refers to any possible transmission scheme that may be deemed as mapping a set of data streams to an antenna set using a matrix V.

[00033] The applied precoding could be based on corresponding feedback from the UE or channel measurements at a base station. In a simple single- user single base unit scheme, one set of DRSs could be defined corresponding to the effective precoded channel (i.e., "HV" in the above equation). If two streams are transmitted to a user in a rank-2 transmission, then only 2 DRS ports (i.e., 2 subsets of DRS each corresponding to a precoded antenna port) are sufficient, even though the actual signal transmission may come from all the NT antennas at the base unit where Ντ can be greater than 2. In FIG. 3, as an example only, RE 320, 340, 330 and 350 may correspond to one DRS port while RE 325, 345, 335 and 355 may correspond to another DRS port.

[00034] In a future migration of a system, for example in 3GPP LTE Release 10 and beyond, user-specific RS (or DRS) are expected to be used widely with advanced Multiple-Input Multiple-Output (MIMO) modes like Coordinated Multipoint transmission (CoMP) and multi-user (MU) MIMO modes described earlier. As described earlier, DRSs are sufficient to enable demodulation. This is also helpful since an eNB is not required to signal exact transmission parameters like precoders, co-ordinating points, etc. However, an estimate of the actual (un-precoded or explicit) channel is required at the eNB to derive such transmission parameters. So as mentioned before, feedback measurements for this purpose are enabled in LTE Release-10 by defining lower density reference signals specifically for the purpose of feedback measurements (CSI-RS). Since they do not need to support demodulation, like CRS in LTE Release 8, a lower density is sufficient. Further, with CoMP, CSI-RS may be setup to enable measurements at the user device on a plurality of antennas from multiple base units. In FIG. 3, as an example only, RE 304, 305, 306, 307, 308 and 309 may also be CSI-RS.

[00035] From either CRS or CSI-RS, the remote unit receiver can estimate the CSI. For the OFDM example, the receiver estimates CSI at each subcarrier between each receiver antenna and each transmitter antenna. The CSI may be denoted as a channel matrix on a sub-carrier k represented by

[00036]

h h NrNt

[00037] where h l} is the channel matrix from j th transmit antenna to the i th receive antenna.

[00038] A correlation between antenna port i and antenna port j may be computed as follows

[00039] , s

AreS

[00040] where h ki is the channel measured corresponding to antenna port i on subcarrier k , S is a set of subcarriers, typically corresponding to the whole operational bandwidth (denoted as RWB) or a sub-band / narrowband (denoted as RNB). [00041] More generally, an antenna correlation matrix that represents the spatial covariance among a plurality of transmit antennas can be computed as follows

R R

[00042]

ieS

R Nt,l R Nt,Nt

[00043] The Eigen Decomposition of R may be expressed in a well-defined format as

VDV" (1) where V is a unitary matrix of Eigen vectors, where the first column is the most dominant vector, the second column the second dominant vector and so on. D is a diagonal matrix with diagonal entries as Eigen values of R. The full knowledge of R at the transmitter will enable advanced beamforming/ precoding techniques that will improve spectral efficiency and system throughput. However, the overhead may be large and approximations suitable to the transmission mode are applied.

[00044] For SU-MIMO precoding, the Eigen space information as represented by V above can be viewed as optimal precoding transmission weights in a capacity maximizing sense.

[00045] Existing 4 th Generation (4G) air interfaces (i.e., 3GPP LTE and IEEE 802.16e) already support beamforming operation via the precoding operation as described earlier. To support precoding operation from the base station, a user terminal will be reporting back to the base station a preferred Precoding Matrix Index (PMI) which is an index to a set of predetermined precoding matrices. The recommended precoding matrix is obtained at the user terminal based on a certain metric such as maximizing the post-precoding link quality or throughput and is selected from one of the quantized codebook entries, wherein the codebook is known to the transmitter and the receiver. Specifically, the standard requires the UE to feedback the PMI that supports a MCS (modulation and coding scheme) with the highest rate, while satisfying a probability if block error target. In future releases, different or more explicit definitions of PMI may be defined. However, in general, the preferred PMI approximately represents a vector quantization of the dominant Eigenspace of R. Further PMI is feedback with an associated rank and as such PMI is an quantized approximation of V(l:r), where V is the rank.

[00046] FIG. 4 illustrates some exemplary antenna configurations at a base unit. A closely spaced ULA, with a typical spacing of 0.5 to 1 wavelengths, is illustrated in 410. A large spaced ULA with typical spacing of 4 to 10 wavelengths is illustrated in 420. A cross-polarized configuration with two sets of cross-poles each with two antennas at +/- 45 polarizations is illustrated in 440. Depending on the configuration, the correlation between different antenna elements may have a certain structure. Some exemplary cases are described herein.

[00047] We now illustrate how the structure of the antenna configuration can be used to develop efficient precoder structures.

[00048] One of the structures that can be exploited is a Kronecker based approximation of the channel covariance. For example, an 8x8 long term covariance matrix corresponding to 8 antennas, as in FIG. 4 at 460, for the transmitter can be approximated as a Kronecker product of a 4x4 correlation matrix corresponding to the ULA and a 2x2 correlation matrix, corresponding to the cross-polarized component i.e.,

[00049] R = R XPOL ® R ULA

[00050] Conceptually, the ULA Kronecker component R ULA captures the correlation submatrix between two non-overlapping subsets of antennas with similar ULA configuration, which in FIG. 4 at 460 are antenna sets (461-464) and (465-468). The polarization Kronecker component R p 0 i captures the correlation submatrix between subsets with similar cross-polarized configuration, namely antenna subsets (461,465), (462,466), (463,467) and (464,468) in FIG. 4. More generally, the spacing/ location and polarization of antenna elements introduce some redundant structure in the antenna correlation, which lead to good Kronecker approximations and can be used as effective compression schemes for feedback overhead reduction. The above representation in the covariance matrix also translates to similar structure for the precoder.

[00051] Even for ULA, the transmit antennas can also be divided into two non-overlapping subsets of antennas. An example is shown in FIG. 4 for subset 431 and 432.

[00052] The final precoder for SU-MIMO rank-r may be computed as

V = SVD(R XP0L^b ® R ULAWB ) (0.2) [00053] The principal Eigen vectors and Eigen values of the constructed matrix are related to that of the Kronecker components as

D = permute(D x p ® D ULA )

(0.3) V = permute{V XP ® V ULA )

where the "permute" operation performs re-ordering of Eigen values.

[00054] We can further illustrate how the reordering influences the structure of the precoder for a 4 Tx cross-pole as an example, where both the ULA and cross-pole sub-matrices are of size 2x2, i.e.,

and

K 0

(0.5) 0 r .

Let us consider a rank-2 SU-MIMO transmission as a further example. Typically the cross-pole covariance matrix is highly rank-2 and ULA covariance can be approximated as rank 1. To express it quantitatively, if the λ K

two Eigen values ratios satisfy— >— , then the rank-2 SU-MIMO precoder, after correspondin re-ordering, can be approximated as

(0.6) On the other hand, in case of — L <— (which is less likely but could occur on short-term basis, like a subband of contiguous subcarriers), then the rank-2 precoder may be a roximated as

[00055] As can be seen, two structures are shown which allow expressing the overall precoder as a Kronecker product of two precoders. Further, the ULA component of the precoder may be feedback at a different time/ frequency granularity than the cross-pole component of the precoder, and allows two component feedback schemes.

[00056] Though the Kronecker representation leads to an elegant separation to two component precoders and is one way to achieve two- component feedback, it also imposes some limitations, where either the ULA or the cross-pol component is assumed to be rank 1 for deriving an overall rank-2 precoder. In general, however a more general two-component precoder structure is useful for higher ranks, which will be further discussed below.

[00057] For the purpose of discussion, we will assume the long- term/ correlated component corresponds to a wide frequency band such as the whole system bandwidth and the short-tem component corresponds to a subband / narrowband that is composed of a set of contiguous subcarriers and is a part of the wideband. [00058] The optimal precoding vector V (optimal in an information theoretic capacity maximizing sense) can be obtained from the Eigen decomposition of the narrowband covariance matrix for band indexed 'b' as follows.

For the rank-2 or 2-layer precoder, the ideal precoder is simply the first two columns of V NBJb . Let us denote the rank-2 Eigen decomposition based precoder as follows (a partition based representation of V NBJb )

where \ tj is a 4x1 (assuming 8-Tx eNB) vector. Clearly, each block corresponds to a vector of weights applied on a subset of antennas (e.g., ULA subset) corresponding to one spatial layer of data stream.

[00059] In a preferred embodiment, we approximate or otherwise represent V m b as

and then we can impose the constraint v 22 1|= 1 and ^ are real values and θ θ 2 e [0,2;r] .

[00060] Clearly, the above precoder representation is based on a matrix with a block of sub-matrices, where each sub-matrix is represented with a vector multiplied with a scalar. More importantly, each sub-matrix corresponds to transmission from a subgroup of antennas, and as one special case, where they all have the same polarization.

[00061] FIG. 5 further describes the above precoding operation from two subsets of antennas. Two non-overlapping subsets of antennas 510, 520 are weighted by a first and a second sub-precoder matrix, respectively. Each sub- precoder matrix corresponds to one or more spatial layers of data transmission, for example in FIG 5, the first sub-precoder is for spatial layer 1 (530) and layer-2 (540). Similarly for the second-precoder, it corresponds to two spatial layers. Mathematically, as an example with eight antennas composed of two groups of 4 antennas, where the first subgroup is number 1- 4, and second subgroup numbered 5-8, a rank r precoder may be expressed as follows

"21

"41 (11)

"52

"61

In the above the first sub-precoder is the top 4 rows(l-4) and the second sub- precoder is the bottom 4 rows (5-8)

[00062] A precoder matrix of one or more vectors associated with one or more spatial layers consists of a first sub-precoder matrix, which comprises of a first set of weights on a first subsets of transmit antennas of the base station, and a second sub-precoder matrix which comprises of a second set of weights on a second subset of transmit antennas of the base station as illustrated above. The set of weights here can be for one or more spatial layers of transmission. [00063] In the final precoder matrix, the first sub-precoder matrix is one or more column vectors, which are of length equal to the number of antennas in the first subgroup, multiplied by one or more scalar s. Similarly for the second sub-precoder.

[00064] For practical reasons, it is often preferred to have the precoder satisfy two constraints, namely i) Full power utilization on each transmit antenna, for maximum Power Amplifier (PA) use and ii) Equal power on each transmitted stream. These constraints can be imposed on the precoder structure above [00059]. To satisfy equal power constraint on each transmit stream, we can impose additional constraint of γ + χ 21 = γ η + γ 22 . To satisfy full power utilization on each individual transmit antenna, we could impose as a sufficient condition, that y + / l2 = / 2l + / 22 and that v !y are constant modulus vectors. With these constraints, we have another preferred embodiment of the precoder structure as follows

[00065] ε [0,2 p 2 e [0, 1]

[00066] The above discussion on the precoder structure is tied to feedback method in this invention. In the feedback scheme for a wireless communication device to send a precoder matrix information to a base station, the wireless communication device sends a first representation of a first matrix chosen from a first codebook, wherein the first matrix has at least two column vectors. The wireless communication device sends a second representation of a second matrix chosen from a second codebook, wherein the first representation and the second representation together convey a precoder matrix of one or more vectors associated with one or more spatial layers. The precoder matrix comprises a first sub-precoder matrix including a first set of weights on a first subsets of transmit antennas of the base station and a second sub-precoder matrix including a second set of weights on a second subset of transmit antennas of the base station. The first sub-precoder matrix is one or more column vectors of the first matrix corresponding to the first representation, multiplied by one or more entries of the second matrix corresponding to the second representation, and the second sub-precoder matrix is one ore more column vectors of the first matrix corresponding to the first representation, multiplied by one or more entries of the second matrix corresponding to the second representation.

[00067] The two-component feedback conveys the information of a precoder matrix recommended by the user to the base station. The actual precoder used by the base station may be different from the suggested precoder, but the actual precoder is derived from the recommended feedback.

[00068] We can use the embodiment above to describe a particular example here. A user feeds back a representation of a first chosen matrix that has a set of vectors v u ,v 12 , v 21 ,v 22 / wherein the representation can be just the index of the chosen matrix within a codebook. Then, the user feeds back a representation of a second matrix of the form of P ^ 7 ? , also as index

pe*

within a codebook format. The precoder information conveyed by these two representations corresponds to a final precoder matrix defined by

[00070] In one embodiment, the vector y could be selected from a constant modulus codebook. One embodiment is a codebook of DFT vectors described below. Alternatively, one could consider a codebook expanded to include some non-DFT vectors.

[00071] A DFT codebook can be described as

ν ϋ ≡{[\ β Νι e Nl ....e Nl f ,n = 0,\,..N 1 = 2 Bl - 1} , which requires Bl bits for representation. L could be Nt/2 for the cross-pole configuration described.

[00072] Further, each column of the second matrix can be considered as a co-phasing vector that aligns the phase of the transmission using the first sub- precoder and the second sub-precoder. In the case of p 2 = o.5 which results in the same amplitude of the entries of the second matrix, we can define another codebook from which the co-phasing vectors are allowed to choose, and an example of the co-phasing codebook is

[1 e A , n = 0,1,.. N 2 = 2 Bl - 1} , with B2 bits of representation.

[00073] More importantly, the above structure can also take advantage of wideband properties. For example, one can observe that v i} are constrained in vector space to a set of entries in the codebook that are highly related to the Eigen space of the ULA component (more specifically the covariance matrix R ULA among the correlated antennas). Similarly the amplitude factor p is also related to the Eigen Spread I . The co-phasing is expected to change more on a subband basis as it relates to the phase alignment between cross-pol groups of antennas. [00074] A preferred embodiment of implementation of two-component feedback is described by adapting v u , v 12 , v 21 , v 22 . The general adaptation can take the following steps of i) Feedback of a set of vectors c wb wb wb wb

l^i i ' ¼2 ' V 2i ' V 22 J in the form of codebook entries on a wideband basis, and ii) Feed back a representation of a rotation matrix to rotate the wideband vector to arrive at the subband vector (i.e., sub-precoder vector that is suitable for the particular sub-band of interest). The rotation matrix is also in the form of a codebook and is typically smaller in size. The co-phasing and the amplitude scalar p are feedback on a subband basis.

[00075] An exemplary embodiment is further described below, where a DFT codebook is used for the wideband vectors and an oversampled DFT based codebook is used for the rotation codebook. As a special case, K can be set to one.

v nb = co.. v wb

[00076]

η = -2 Β * - ι = -Ν 2 12, -(Ν 2 12 - 1), ..0.., Ν 2 12 - 1}

[00077] Another alternate representation of the above is

rf = <Hag(y†)a>,

[00078] L l K i- l n l K i- Ln

^ G C¾ = {([l e

n = -2 B> - 1 = -N 2 12, -(N 2 / 2 - 1), ..0.., N 2 / 2 - 1} [00079] In a variation of the above embodiment, the codebook for the wideband vectors and / or rotation matrix may include some non-DFT vectors in addition to DFT vectors.

[00080] In a more general embodiment, the rotation codebook may also be further defined dependent on the original codebook entry or a combination of dependent and independent entries. Note that above embodiment uses an independent and fixed codebook.

[00081] The main objective of the above representation is to use more feedback bits to capture the wideband information and used fewer bits to capture subband information. An exemplary embodiment is to use a 2 bit codebook for subband vectors and a 4 or 6 bit codebook to represent the wideband entries. In general, the selection of (B l ,B 2 ) may depend on the feedback mode. More specifically, it could depend on the number of sub- bands the feedback is requested on.

[00082] As a further refinement of the above embodiments, we can further reduce the number of wideband vectors to two from 4, which are then represented as (v b ,v b ) ior rank-2. Reducing the wideband basis vectors from 4 in the most generic case to 2 is based on the observation that the spatial correlation of the ULA sub-array can be approximated well by the first two dominant eigenvectors in general and is sufficient for rank 2.

[00083] We describe another preferred representation of precoder based on above simplifications, which is based on the steps 1) a user feeds back a representation of a first chosen matrix (ν^, ν^) which could be seen as representing the wideband property of the ULA sub-array; and 2) the user may optionally adapt (vj" , ν ) on a subband basis with a rotation codebook CBi, as v nb = f(v w ' b , CB i ) ; 3) the user feeds back a representation of a second matrix of the form of P 1 ? , also as index within a codebook format.

pe 2

The precoder information conveyed by these two representations corresponds to a final precoder matrix defined by one of the three structures below (the user is allowed to select which structure to use)

vibP v lb - P

[00085] The above three structures capture three different cases of interaction between ULA spatial components and cross-pol components. Mainly, they correspond approximately to the following three cases i) ULA is strictly rank 1 and cross-pole is used to achieve the rank-2 spatial separation, ii) ULA is highly rank 2 and can be used mainly for spatial separation, and iii) all other cases, where ULA and cross-pole both show rank 2 behavior.

[00086] In the first structure, the wireless communication device selecting one column vector of the first matrix. The first sub-precoder matrix associated with two spatial layers is obtained then by multiplying the selected column vector with first and second elements in a first row of the second matrix, and the second sub-precoder matrix associated with two spatial layers is obtained by multiplying the selected column vector with first and second elements in a second row of the second matrix. [00087] In the second structure, the wireless communication device selecting first and second column vectors of the first matrix. The first sub- precoder matrix associated with two spatial layers is obtained by multiplying the first column vector with a first element in a first row of the second matrix and multiplying the second column vector with a second element in a first row of the second matrix, and the second sub-precoder matrix associated with two spatial layers is obtained by multiplying the first column vector with a first element in a first second row of the second matrix and multiplying the second column vector with a second element in a second row of the second matrix.

[00088] In the third structure, the wireless communication device selecting first and second column vectors of the first matrix. The first sub- precoder matrix associated with two spatial layers is obtained by multiplying the first column vector with a first element in a first row of the second matrix and multiplying the a second column vector with a second element in a first row of the second matrix, and the second sub-precoder matrix associated with two spatial layers is obtained by multiplying the second column vector with a first element in a first second row of the second matrix and multiplying a the first column vector with a second element in a second row of the second matrix.

[00089] The illustration above is based on rank 2 or 2-layer precoding. For rank 1 or one-layer precoding, a similar generalized construction of the final precoder could be A simplified structure that is found

sufficient for 8 Tx cross-pol antenna configuration rank-1 transmission is = f(v w l b , CB i ) . Similar generalization/ construction can be

applied to higher ranks using a combination of co-phasing vectors, additional wideband precoder components and additional amplitude scalars.

[00090] In other embodiments, we provide further description of the component codebooks, and will use rank 2 for illustration, as similar approach applies to other ranks.

[00091] In a preferred embodiment, a product matrix based method of precoder matrix representation can be used to describe the two-component feedback.

[00092] First component which describes a wideband codebook is expressed as

1 1

2 π jn x

[00093] W l « ! , ¾ e {0, 1, 2, ..N ! = 2 ¾ - 1}

2 π jLn x 2π]∑ η 2 which is essentially to capture [v * v ] . A different size may be chosen for nl and n2 and further only particular pairs may be allowed to reduce codebook size. In addition, some non-DFT codebook entries may be added to represent each column.

[00094] The second codebook may be represented as one of the three forms: ,or

2π jk 2

P pe

} , where

p 2 e {0.5,0.75,0.9,1}, k,,k 2 e {0,1,2,3}

[00096] The final precoder is obtained as

W = reshape W^V 2 )

where a "reshaping" operation is to re-arrange the columns of the product matrix to form the precoder matrix. For example, for 8 Tx with a 4-by-2 W x and a 2-by-4 W 2 and 4-by-4 product matrix can be column-wise reshaped to 8x2 precoder W as follows. reshape(X = [X V X 2 X X A ]) =

X 2 X 4

A more general "reshaping" operation on a matrix represented as a row of vectors can be defined as follows for higher ranks and arbitrary number of antennas.

X l X N+l X, (N

X,

Γ β5 Ηαρ β ([Χ γ Χ 2 Χ,Χ ...Χ ΝχΜ ] , Ν,Μ) =

Y Y X,

[00097] In another embodiment, additional joint coding of parameters p, k y , k 2 may be performed to restrict the size of the subband codebook by selecting a subset of combinations enumerated above. Further different entries may also be chosen for the codebook than shown in the embodiment. The three rows in the codebook are chosen to correspond to the three structures described in Error! Reference source not found.. In the first row, the amplitude scalar is set to 1 and co-phasing is reduced to 4 instead of 16, to ensure orthogonality. Similar joint coding can be performed to reduce the entries in other rows corresponding to structures 2 and 3, which may be based on further optimization.

[00098] In another embodiment, the subband codebook may be expanded allowing the incorporation of a rotation codebook, in which case the wideband and subband codebook could be represented mathematically as

W, = [ diag([\ e Nl ..e Nl ]), diag([\ e Nl ..e Nl ] ) ]

[00099]

1 0 0 1 0 0

0 e 1 0 0 e 1 0

/¾ ,/¾ e {0, l, ..Ni = 2 -1}

0 0 0 0 0 0

0 0 . e 1 0 0 . e 1

pu x 0 0 pe 4 u x

2njk } , where

0 ^l - p 2 e 4 u 2 p 2 u 2 0

;

u x ,u 2 e C5.. = {[1 e NlK . N,K - riT,

n =—2 Bn -l -N 2 12, -(N 2 / 2 - 1), ..0.., N2 / 2 - 1} where 0, 1 represent Lxl vectors of zeros and ones respectively.

[000101] The construction above presents a product design followed by reshaping to provide improved flexibility in the two component feedback. However, further overhead reduction can be configured by eNB, by enabling codebook subset restrictions at wideband and subband level. A codebook subset restriction is essentially where eNB signals the UE to only use a subset of codebook entries for the feedback. Subset restriction can be applied to either the first codebook, or the second codebook, or both.

[000102] Codebook subset restriction can be used for both wideband and subband components in a pre-configured and compatible manner, which we will further describe with some exemplary methods.

[000103] In one embodiment, a subset restriction can be imposed at wideband level to only transmit information regarding vl, by defining wideband codebook as follows 2π]η χ 2njLn x

W l = [ diag([\ e rtl ..e rtl ]) diag([\ 1 ..1] ) ]

With this restriction, only Bl bits of information need to be sent, and may be beneficial when ULA is highly rank 1 as in certain LOS cases, and only structure one may be used for all sub-bands included in the feedback. This would also implicitly result in following restriction for the subband codebook,

W 2 = { u x e , k x , k 2 e {0, 1, 2,3} ;

0 0 0

2π jn 2π j .2n 2π j.Ln

,K N,K -\T

[000104] = {[i e N 1 e N, 1 K -N 2 / 2, -(N 2 / 2 - l), ..0.., N 2 / 2 - l}

[000105] In another embodiment, at subband level, a subset restriction can be imposed to limit to one or more structures.

[000106] In another embodiment, a subset restriction can be used to signal wideband value of a parameter. For example, p may be signaled commonly on a wideband basis or as a fixed value common to all codebook entries corresponding to each structure. Similarly, co-phasing may be fixed if structure one is used, so that k Y = k 2 = 0 .

[000107] In another embodiment, subset restriction can enable/ disable further refinement using rotation i.e., { and u 2 can be set to [1 1 ... 1] T , which corresponds to no rotation of wideband vector. [000108] Additionally, codebook subset restrictions may be imposed in an inter-dependent manner over wideband and subband codebooks. In other words, an UE may receive an UE-specific or cell-specific higher layer subset- restriction parameter which can take a set of multiple values each of which maps to a 'codebook subset' pair for W x and W 2 . The mapping can be pre-defined and known to both the eNB and the UE. The construction of final precoder is simply based on

W = reshape(W l W 2 ) (0.12) once the component precoders are chosen from the restricted codebooks.

[000109] In general, the subset restriction may be useful under certain persistent long-term conditions like whether link is line of sight (LOS) or non line of sight (NLOS), the eNB antenna configuration if a general codebook is used for all configurations (say ULA/XPOL), UE geometry which could limit the uplink feedback link capacity and iv) additional feedback required for other transmission modes like multiuser MIMO which may provision fewer bits to higher-ranked SU transmission.

[000110] The two component feedback scheme described above is applicable to multiple feedback modes as further described below.

[000111] One embodiment is illustrated in Fig 6. The first component could correspond to a wideband matrix W™ corresponding to the whole bandwidth 610, which is composed of one or more sub-bands 620. Each subband 620 is composed of a set of contiguous subcarriers 630 or in LTE, a set of RBs, which is the minimum scheduling/ feedback granularity. A second matrix W 2 corresponds to a to a band 'b'. The precoder for a band 'b', represented as W b is obtained as a function of W™ and W 2 b . A preferred example of the function as described above is reshape (WlxW2). A UE is required to feedback the wideband precoder/ matrix for the whole band and the subband precoder/ matrix for each subband.

[000112] In one embodiment, the feedback could be defined along with an associated CQI feedback. The CQI feedback could correspond to a modulation and coding scheme (MCS) recommendation and / or a rank indication. The CQI could further consist of multiple CQI, one per each subband. In this case, the UE must select the components W™ and such that the subband precoders can support the CQI reported corresponding to the subband on the hypothesis UE is allocated that subband, while satisfying an error rate target. In another embodiment, a single CQI may be reported for the whole band, in which case the reported CQI must meet the error rate target, on the hypothesis that the UE is allocated the whole band, and the reported pre-coders are applied on each subband.

[000113] In another embodiment, the reported sub-bands may be a subset of the sub-bands that constitute the bandwidth and may or may not be equal in size as defined by number of subcarriers in each subband.

[000114] In another embodiment, the sub-bands to be reported may be signaled by the eNB explicitly from a set of predefined patterns.

[000115] In another embodiment, the sub-bands 620 may also be selected by the UE. The selection could be based on the best N (N<wholebandwidth) bands with ranking defined based on a performance criterion like sum rate supported or similar. [000116] In another embodiment, the first matrix may be further feedback at different time instances and/ or periodicities compared to the second matrix. This can, for example, be useful to alleviate the feedback load on the uplink in a single reporting instance like a subframe. Further, they may also be transmitted on different uplink physical channels like PUSCH and PUCCH defined in the current LTE Release-8.

[000117] In the above description, we have extensively used DFT based codebooks as preferred examples. If the antennas at the eNB are not calibrated, it is likely that the phase progression as in a DFT vector assumed for a ULA antenna sub-array may not hold. In the worst case scenario of un- calibrated antennas, the relative phase error between antennas may be random. However, it is also expected they are stable and slowly varying compared to time between feedback intervals. Other errors like relative timing errors between antennas may also introduce a relative phase shift that may not correspond to a ULA array. In such cases, one solution is for the UE to feedback a phase error correction vector relatively infrequently, which can be used to rotate the reported precoder at the eNB.

[000118] Further, another approach is to include some non-DFT but constant modulus vectors in the wideband codebook, which can compensate for the relative phase error. Note that the subband codebook may not be changed, since it is a relative rotation with respect to the wideband vector.

[000119] While the present disclosure and the best modes thereof have been described in a manner establishing possession and enabling those of ordinary skill to make and use the same, it will be understood and appreciated that there are equivalents to the exemplary embodiments disclosed herein and that modifications and variations may be made thereto without departing from the scope and spirit of the inventions, which are to be limited not by the exemplary embodiments but by the appended claims.

[000120] What is claimed is: