Method for transmitting channel state information in multiple antenna system

Abstract

There is provided a method for transmitting channel state information in a wireless communication system. The method includes receiving a pilot signal, obtaining an estimated channel using the pilot signal, obtaining a plurality of bases representing the estimated channel, wherein a basis comprises a plurality of vectors and transmitting a representative vector as channel state information, wherein the representative vector is selected from the plurality of bases.

Description

where s_i is a stream transmitted through a i-th transmitting antenna.

The streams s₁, s₂, . . . , and s_NT can be transmitted with different transmission powers. When it is assumed that the transmission powers are P₁, P₂, . . . , and P_NT, the streams which the transmission powers are adjusted can be expressed in a vector as:
ŝ=[ŝ₁,ŝ₂, . . . ,ŝ_NT]^T=[P₁s₁,P₂s₂, . . . ,P_NTs_NT]^T.  Equation (2)

Using a diagonal matrix P of which diagonal elements are the transmission powers, the power-adjusted stream

ŝ

can be expressed as:

$\begin{array}{cc}\hat{s}=[\begin{array}{cccc}{P}_1& & & 0\\ & P_2& & \\ & & \ddots & \\ 0& & & P_{N_T}\end{array}[[\begin{array}{c}{s}_1\\ s_2\\ ⋮\\ s_{N_T}\end{array}[\begin{array}{c}\\ \\ \\ {}_{}\end{array}=\mathrm{Ps}.& \mathrm{Equation}\left(3\right)\end{array}$

Consider that a weight matrix W is applied to the power adjusted stream vector

ŝ

so that N transmission signals x₁, x₂, . . . , and x_NT are formed. The transmission signals x₁, x₂, . . . , and x_NT can be expressed by a vector x as:

$\begin{array}{cc}\begin{array}{c}x=[\begin{array}{c}{x}_1\\ x_2\\ ⋮\\ x_i\\ ⋮\\ x_{N_T}\end{array}[\begin{array}{c}\\ \\ \\ \\ \\ \end{array}\\ =[\begin{array}{cccc}{w}_{11}& w_{12}& \cdots & w_{1N_T}\\ w_{21}& w_{22}& \cdots & w_{2N_T}\\ ⋮& & \ddots & \\ w_{i1}& w_{i2}& \cdots & w_{{\mathrm{iN}}_T}\\ ⋮& & \ddots & \\ w_{N_{T}1}& w_{N_{T}2}& \dots & w_{N_T{N}_T}\end{array}[[\begin{array}{c}{\hat{s}}_1\\ {\hat{s}}_2\\ ⋮\\ {\hat{s}}_j\\ ⋮\\ {\hat{s}}_{N_T}\end{array}[\\ =W\hat{s}\\ =\mathrm{WPs}\end{array}& \mathrm{Equation}\left(4\right)\end{array}$

where a weight w_ij (1≦i≦N_T, 1≦j≦N_T) is a weighting value of a j-th stream for an i-th transmitting antenna. The weight w_ij can be arranged in the weight matrix W. The weight matrix W can also be called as a preceding matrix.

The transmission signal x can be transmitted by using a spatial diversity or by using a spatial multiplexing. When the spatial diversity is used, the same signal is transmitted through multiple antennas and elements of the power-adjusted stream

ŝ

are all same. When the spatial multiplexing is used, the elements of the power-adjusted stream

ŝ

may be different to each other. The spatial diversity and the spatial multiplexing can be combined. Three transmitting antennas can transmit streams by using the spatial diversity and the other transmitting antennas can transmit streams by using the spatial multiplexing.

Consider received signals of the receiver 101. Received signals y₁, y₂, . . . , and y_NR for the N_R receiving antennas are written as:
y=[y₁y₂ . . . y_NR]^T.  Equation (5)

Assume that the channels are expressed by the use of indexes of the transmitting and receiving antennas. A channel from the j-th transmitting antenna to the i-th receiving antenna is expressed as h_ij. The channel may also be called as MIMO channel. A set of the channel for multiple antennas can be expressed in the form of a vector and a matrix.

FIG. 2 is a diagram illustrating channels from the N_T transmitting antennas to an i-th receiving antenna.

Referring to FIG. 2, channels from the transmitter 100 to a i-th receiving antenna can be expressed in the form of a vector as:
h_i^T=[h_i1,h_i2, . . . ,h_iNT]  Equation (6)

where a superscript T denotes a transpose.

The channels from the N_T transmitting antennas to the N_R receiving antennas can be expressed in the form of a matrix by:

$\begin{array}{cc}H=[\begin{array}{c}{h}_1^T\\ h_2^T\\ ⋮\\ h_i^T\\ ⋮\\ h_{N_R}^T\end{array}[=[\begin{array}{cccc}{h}_{11}& h_{12}& \cdots & h_{1N_T}\\ h_{21}& h_{22}& \cdots & h_{2N_T}\\ ⋮& & \ddots & \\ h_{i1}& h_{i2}& \cdots & h_{{\mathrm{iN}}_T}\\ ⋮& & \ddots & \\ h_{N_{R}1}& h_{N_{R}2}& \dots & h_{N_R{N}_T}\end{array}[.\begin{array}{c}\\ \\ \\ \\ \\ \end{array}& \mathrm{Equation}\left(7\right)\end{array}$

In the actual wireless system, an additive white Gaussian noise (AWGN) is added to the channel matrix H. The AWGN for an i-th channel can be expressed as n_i and AWGNs for all channels can be expressed in the form of a vector by:
n=[n₁n₂ . . . n_NR]^T.  Equation (8)

The receiving signal y can be expressed by:

$\begin{array}{cc}\begin{array}{c}y=[\begin{array}{c}{y}_1\\ y_2\\ ⋮\\ y_i\\ ⋮\\ y_{N_R}\end{array}[\begin{array}{c}\\ \\ \\ \\ \\ \end{array}\\ =[\begin{array}{cccc}{h}_{11}& h_{12}& \cdots & h_{1N_T}\\ h_{21}& h_{22}& \cdots & h_{2N_T}\\ ⋮& & \ddots & \\ h_{i1}& h_{i2}& \cdots & h_{{\mathrm{iN}}_T}\\ ⋮& & \ddots & \\ h_{N_{R}1}& h_{N_{R}2}& \dots & h_{N_R{N}_T}\end{array}[[\begin{array}{c}{x}_1\\ x_2\\ ⋮\\ x_i\\ ⋮\\ x_{N_T}\end{array}[+[\begin{array}{c}{n}_1\\ n_2\\ ⋮\\ n_i\\ ⋮\\ n_{N_R}\end{array}[\begin{array}{c}\\ \\ \\ \\ \\ \end{array}\\ =\mathrm{Hx}+n.\end{array}& \mathrm{Equation}\left(9\right)\end{array}$

The receiver 101 can be an user equipment having multiple antennas or a plurality of user equipments having at least one antenna. For example, an user equipment may receive downlink signals through N_R antennas. Or N_R user equipments may concurrently receive downlink signals through one antenna. In any combination, the total number of receiving antennas can be kept N_R. This technique is generally called a MIMO BC (Broadcast Channel) or an SDMA (Space Division Multiple Access).

Reversely, an user equipment having N_T antennas may transmit uplink signals. Or N_T user equipments having one antenna may concurrently transmit uplink signals. In any combination, the total number of transmitting antennas can be kept N_T. This technique is generally called a MIMO MAC (Multiple Access Channel).

Since uplink transmission and downlink transmission show duality, the technique used in one side can be used in the other side. Although the MIMO BC be mainly shown in the following description, skilled in the art can easily be applied to the MIMO MAC.

FIG. 3 is a diagram illustrating a single-user multiple antenna system.

Referring to FIG. 3, a base station (BS) 310 as the transmitter has multiple antennas and an user equipment (UE) 320, that is, an user, as the receiver has multiple antennas. It is possible to obtain the maximum channel capacity by using MIMO technique.

Assume that the BS 310 and the UE 320 knows channel information perfectly. Using SVD (Singular Value Decomposition), a channel matrix H can be expressed as:
H=UΣV^H  Equation (10)

where U and V are a unitary matrix, Σ is a diagonal matrix and a superscript H denotes a Hermitian matrix. In order to acquire the maximum gain in view of the channel capacity, V is selected as the weight matrix W. A receiving signal y is multiplied by U^H as shown in:
y=Hx+n=UΣV^Hx+n=UΣV^H Wŝ+n=UΣV^HVŝ+n=UΣŝ+n,
{tilde over (y)}=U^Hy=U^H UΣŝ+U^Hn=Σŝ+ñΣPs+ñ.  (Equation 11)

A transmission power matrix P in Equation (11) can be determined by a water filling technique so as to acquire the maximum channel capacity.

FIG. 4 is a diagram illustrating a multi-user multiple antenna system.

Referring to FIG. 4, a BS 410 has N_T antennas and the respective UEs 420-1, 420-2, . . . , and 420-N_R have one antenna. Similarly to the single user environment shown in FIG. 3, channel capacity can be maximized by the use of a transmission beam-forming process.

However, interference between the UEs should be considered. When the transmission beam-forming process is used, the weighting vector should be selected so as to minimize the interference between the UEs. The multi-user environment needs a system more complicated than the single user environment.

In the multi-user multiple antenna system, a dirty paper coding (DPC) is well known as the optimal scheme in view of the channel capacity. The complexity thereof is considerable and thus it is not easy to implement the DPC. A zero forcing beam-forming (ZFBF), a random beam-forming (RBF) and the like are known as the next optimal schemes. In such schemes, as the number of UEs increases, optimal performance can be achieved.

In ZFBF, a receiver uses an inverse channel matrix H⁻¹ to decompose a receiving signal. The receiving signal y can be expressed by Equation (12).
W=H⁻¹
y=Hx+n=HWŝ+n=HH⁻¹ŝ+n=ŝ+n=Ps+n  Equation (12)

A channel matrix H should be a non-singular so that the inverse channel matrix H⁻¹ always exists. We can make up the channel matrix H as a unitary matrix since the unitary matrix is the non-singular matrix.

In order to be the unitary matrix, different row (or column) vectors of the channel matrix H should be orthogonal to each other. Two vectors are orthogonal if inner product between the two vectors is zero.

Let a column vector h_i be elements of the i-th column of H as shown in:
h_i^T=[h_i1,h_i2, . . . ,h_iNT].  Equation (13)

When the channel matrix H is the unitary matrix, inner product between two different column vectors h_i and h_j of H is zero. This relation can be written as:

$\begin{array}{cc}\begin{array}{c}\{\begin{array}{c}{h}_i^H{h}_j=0\\ h_i^H{h}_i\ne 0\end{array}\\ \mathrm{when}i\ne j\end{array}& \mathrm{Equation}\left(14\right)\end{array}$

So we can call the channel matrix H the unitary matrix if the column vectors of the channel matrix H satisfy Equation 14. The unitary matrix H has another property as:
H⁻¹=H^H.  Equation (15)

The ZFBF has an advantage that it need not particular operation of the receiver to detect the receiving signal and the matrix calculation is little complex. However, since the channel matrix H should be the unitary matrix, the receiving antennas, that is, the UEs, should be selected so that all the MIMO channels are orthogonal. The i-th receiving antenna should be selected so that the channel h_i of the i-th receiving antenna is orthogonal to channels of the other receiving antennas. To select channels, all the channel information should be known.

A receiver generally measures a estimated channel based on a pilot signal (reference signal) transmitted from a transmitter. The transmitter should be aware of the estimated channel by receiving channel information from the receiver. When the transmitter knows all channels, the transmitter can adaptively perform a transmission. The ZFBF is one example. Since the channels vary with the lapse of time, the channel information should be transmitted from the receiver to the transmitter as often as possible to improve accuracy. But it may exhaust radio resources and limit data rates. Also, it is not easy to transmit the channel information accurately since a quantization error may exist.

The RBF is a beam-forming process without any channel information. The RBF is performed as follows. (1) A transmitter transmits a pilot signal through multiple antennas by the use of a weighting vector created randomly. (2) A receiver receiving the pilot signal estimates a signal-to-noise ratio (SNR) and transforms the SNR into a proper data rate or a signal-to-interference noise ratio (SINR). (3) The receiver feeds back the SINR to the transmitter. The receivers need not be aware whether the transmitter has multiple antennas. (4) The transmitter assigns the channel to the receiver having the highest SINR. (5) After a predetermined time passes, the transmitter transmits the pilot signal by the use of the weighting vector created randomly and repeats the step (2) again.

As the number of receivers (UEs) in a cell increases, the probability that the weighting vector created randomly is very close to the actual channel increases. Accordingly, the RBF can achieve the performance closer to the ZFBF as the number of receivers increases.

Let weight vectors w₁, w₂, . . . , and w_NT be column vectors of a unitary weight matrix W created randomly by the transmitter. The product of a transmission power matrix P and a transmission stream vector s is expressed as Ps. Without losing generality, we can choose the transmission power matrix P as an identity matrix I. A transmission signal x can be expressed by:
x=[w₁w₂ . . . w_NT]Ps=WPs=Ws.  Equation (16)

A receiving signal y_i received by the i-th receiving antenna can be expressed by:

$\begin{array}{cc}\begin{array}{c}{y}_i=\left[h_{i1},h_{i2},\dots h_{{\mathrm{iN}}_T}\right]x+n_i\\ =h_i^{H}x+n_i\\ =h_i^H\mathrm{Ws}+n_i\\ =h_i^H\left[w_1{w}_2\dots w_{N_T}\right]s+n_i\\ =\left[h_i^H{w}_1{h}_i^H{w}_2{h}_i^H{w}_{N_T}\right]s+n_i\\ =\sum _{k=1}^{N_T}{h}_i^H{w}_k{s}_k+n_i\end{array}& \mathrm{Equation}\left(17\right)\end{array}$

In the RBF, the i-th receiving antenna's SINR for a weight vector w_m can be defined by:

$\begin{array}{cc}{\mathrm{SINR}}_{i,m}=\frac{{h_i^H{w}_m{s}_m}^2}{1+\sum _{k=1,k\ne m}^{N_T}{h_i^H{w}_k{s}_k}^2}& \mathrm{Equation}\left(18\right)\end{array}$

The weight vector w_m in which the SINR_i,m of Equation (18) is the maximum can be found out. We call it a maximum weight vector w_max.

Assume that an a N_T dimension complex vector space

C^NT

is spanned by weight vectors w₁, w₂ . . . , and w_NT. The vector space

C^NT

is a column space of the weight matrix W. When a channel of the i-th receiving antenna hi belongs to a null space of

C^NT,

the inner product of h_i and any one of weight vectors is zero as shown in:
h_i^Hw_i= . . . =h_i^Hw_NT=0.  Equation (19)

Since the weight matrix W is assumed as a unitary matrix, the null space of

C^NT

is a zero space. But the channel h_i does not belong to the zero space. The product cannot be zero as shown in:
h_i^Hw_k≠0 (k=1,2, . . . ,N_T).  Equation (20)

Accordingly, the minimum value of the SINR of Equation (18) is always larger than zero.

To maximize the SINR of Equation (18), the channel h_i should belong to one of subspaces L(w₁), L(w₂), . . . , and L(w_NT) of the vector spaces. Geographically, subspaces L(w₁), L(w₂), . . . , and L(w_NT) denote lines in the N_T dimension complex vector space

C^NT

using weight vectors w₁, w₂ . . . , and w_NT as a basis of the vector space

C^NT.

The basis of a vector space

C^NT

is a linearly independent subset of

C^NT

that spans (or generates)

C^NT.

When the weight matrix W is the basis, the basis comprises column (row) vectors of the weight matrix W.

When the channel h_i belongs to a subspace L(w_m), it gives:

$\begin{array}{cc}\begin{array}{c}{h}_i\in L\left(w_m\right)\iff \{\begin{array}{c}{h}_i^H{w}_m\ne 0\\ h_i^H{w}_j=0\end{array}\\ \left(j\ne m\right)\end{array}& \mathrm{Equation}\left(21\right)\end{array}$

Accordingly, if we find out the weight vector w_m which belongs to the subspace L(w_m), we maximize the SINR (w_m=w_max). The maximum weight vector w_max is a vector which has a maximum coefficient when the vector is projected onto the basis of the vector space

C^NT.

The SINR of Equation (18) indicates how closely the corresponding channel is orthogonal to one of weight vectors w₁, w₂, . . . and w_NT.

In the beam-forming, not are streams for all users transmitted but streams for users corresponding to the maximum number of transmitting antennas (or receiving antenna) can be transmitted. One of user selection criterion is that channels between the selected users are orthogonal to each other.

In the ZFBF, user selection process can be as follows. (1) All users transmit channel information to the transmitter. (2) The transmitter selects users of which channels are orthogonal to each other using the channel information. (3) A channel matrix for selected users is made up and an inverse matrix of the channel matrix is selected as a weight matrix. (4) Signals for selected users are beam-formed using the weight matrix and transmitted.

In the RBF, (1) a transmitter selects a unitary weight matrix and transmits to users using the selected unitary weight matrix. (2) Every user selects a column vector having the maximum value from the weight matrix based on the received signal. A SINR is calculated according to the selected column vector. The index of the selected column vector and the SINR are transmitted to the transmitter. (3) The transmitter selects users having the channels closest to the column vectors of the weight matrix based on the index and SNIR. Since the weight matrix is an unitary matrix, channels of the selected users are orthogonal to each other. (4) Signals for selected users are beam-formed using the weight matrix and transmitted.

In the ZFBF, each user transmits its channel information without performing any process. On the contrary, in the RBF, each user calculates the degree of orthogonality using a column vector of a weight matrix and then transmits the maximum SINR and an index of the column vector corresponding to the maximum SNIR.

As shown above, as an channel becomes closer to one of the subspaces which are constituted by the basis, the channel has a larger SINR value in the RBF. An user to be transmitted is determined depending on how closely the user's channel is related to the column space of the weight matrix. The dimension of the column spaces of the weight matrix, that is, the number of vectors constituting the basis, is equal to the number of antennas N_T. The probability that the user has the maximum SINR is equal to the probability that the user's channel is matched with one of the subspaces. The probability is very low since the user's channel is time-varying and unpredictable. As a result, many users are required to increase the probability.

Mathematically, a vector in an N_T dimension vector space can be expressed by a basis including N_T vectors. A standard basis is expressed as a set of column vectors of an identity matrix. Using the standard basis, a channel h can be expressed by:

$\begin{array}{cc}\begin{array}{c}h=[\begin{array}{c}{h}_1\\ h_2\\ ⋮\\ h_i\\ ⋮\\ h_{N_T}\end{array}[\begin{array}{c}\\ \\ \\ \\ \\ \end{array}\\ =h_1[\begin{array}{c}1\\ 0\\ ⋮\\ 0\\ ⋮\\ 0\end{array}[+h_2[\begin{array}{c}0\\ 1\\ ⋮\\ 0\\ ⋮\\ 0\end{array}[+\dots +h_i[\begin{array}{c}0\\ 0\\ ⋮\\ 1\\ ⋮\\ 0\end{array}[+\dots +h_{N_T}[\begin{array}{c}0\\ 0\\ ⋮\\ 0\\ ⋮\\ 1\end{array}[\begin{array}{c}\\ \\ \\ \\ \\ \end{array}\end{array}& \mathrm{Equation}\left(22\right)\end{array}$

Using the standard basis, we need the total N_T coefficients of h₁, h₂, . . . , and h_NT to represent the channel h. The channel h can be reproduced by combining h₁, h₂, . . . , and h_NT and the standard basis.

If the channel h=[h₁, h₂, . . . , and h_NT]^T has a specific value, the number of coefficients required to transmit the channel vector h can be reduced by changing the basis. For example, consider a channel vector h=[−3, 3, −3, 3]^T. Under the standard basis, 4 coefficients of −3, 3, −3 and 3 should be needed to represent the channel as shown in:

$\begin{array}{cc}h=[\begin{array}{c}-3\\ 3\\ -3\\ 3\end{array}[=-3[\begin{array}{c}1\\ 0\\ 0\\ 0\end{array}[+3[\begin{array}{c}0\\ 1\\ 0\\ 0\end{array}[+\left(-3\right)[\begin{array}{c}0\\ 0\\ 1\\ 0\end{array}[+3[\begin{array}{c}0\\ 0\\ 0\\ 1\end{array}[\begin{array}{c}\\ \\ \\ \end{array}& \mathrm{Equation}\left(23\right)\end{array}$

Consider that we choose [1, 1, 1, 1]^T, [1, −1, 1, −1]^T, [1, 1, −1, −]^T and [1, −1, −1, 1]^T as a basis, only one coefficient of −3 is needed to represent the channel as shown in:

$\begin{array}{cc}h=[\begin{array}{c}-3\\ 3\\ -3\\ 3\end{array}[=0[\begin{array}{c}1\\ 1\\ 1\\ 1\end{array}[+\left(-3\right)[\begin{array}{c}1\\ -1\\ 1\\ -1\end{array}[+0[\begin{array}{c}1\\ 1\\ -1\\ -1\end{array}[+0[\begin{array}{c}1\\ -1\\ -1\\ 1\end{array}[\begin{array}{c}\\ \\ \\ \end{array}& \mathrm{Equation}\left(24\right)\end{array}$

The coefficient of −3 is enough to represent the channel vector h. The amount of transmission information can greatly be reduced by changing a basis.

Column vectors of an unitary matrix can be used as a basis. In the RBF, column vectors of a weight matrix become a basis. A column vector which has the largest coefficient among the column vectors is selected. There are several problems in applying the RBF to an actual wireless communication environment.

The first problem is that the RBF is not efficient in view of channel capacity. Channel information can be fed back only when a random unitary matrix is continuously transmitted. Continuous transmission of the random unitary matrix causes a waste of that the channel capacity. A transmission signal generally includes a pilot signal for various reasons for data demodulation, synchronization, and the like. Since the receiver can measure a estimated channel using the pilot signal, transmission of the random unitary matrix only for acquiring the channel information causes waste of the channel capacity.

The second problem is that the RBF can be effectively operated when the number of users in a cell is large enough. A randomly generated weight matrix may not be suitable for representing every user's actual channel. When all column vectors of the randomly generated weight matrix are used to represent a channel, the channel may be accurate. But only one column vector is used in the RBF. If one column vector cannot effectively represent an user's actual channel, the channel information transmitted from the user may be mismatched with the user's actual channel. In order to select an user exhibiting a high SINR, many user are necessary in probability.

The third problem is that the channel capacity is wasted by transmitting unwanted channel information. All users should transmit a SINR and an index. If a channel is not matched with the basis, the channel exhibits a low SINR. An user having the low SINR is hardly selected, but the user transmits the low SINR nevertheless. The channel capacity is wasted by transmitting unused channel information.

The fourth problem is that the users' interference cannot effectively be mitigated in the RBF. An user's channel not matched with a basis gives a low SINR. Although the user's channel does not match with the basis, the user may necessarily receive or transmit a signal. Examples thereof include a case where re-transmission is required in response to Hybrid Automatic Repeat Request (HARQ) scheme and a case where an user's buffer is full and thus uplink transmission is necessarily required. The RBF does not effectively mitigate the interference with users since the user having the low SINR cannot use the channel. This is a loss in view of the channel capacity.

The above-mentioned problems of the RBF can be summarized as follows. The first problem is due to the basis change. The second to fourth problems are due to the mismatch between the basis and the user's actual channel.

Now, the multi-user MIMO system is considered in which the number of transmitting antennas of a transmitter is N_T (where N_T is a natural number larger than 2), the number of receivers is k (where k is a natural number larger than 1) and each receiver receives a signal through one receiving antenna.

FIG. 5 is a block diagram illustrating a transmitter having a multiple transmitting antenna according to an embodiment of the invention.

Referring to FIG. 5, the transmitter 500 includes a data processor 510, a power allocator 520, a precoder 530, an RF (Radio Frequency) unit 540 and a controller 550.

The data processor 510 processes and modulates information bits in accordance with predetermined modulation and coding scheme (MCS) so as to create a transmission stream. When the AMC scheme is used, the data processor 510 can select one of pre-determined MCS sets in accordance with an AMC control signal sent from the controller 550. The transmission stream is denoted by s. The data processor 550 may include a channel encoder (not shown) and a mapper (not shown). The channel encoder encodes information bits in accordance with a predetermined coding scheme so as to form a encoded data. The mapper maps the encoded data on a signal constellation. The modulation scheme of the mapper is not particularly limited and may be an m-PSK (m-Phase Shift Keying) or an m-QAM (m-Quadrature Amplitude Modulation). For example, the m-PSK method may be a BPSK (Binary-PSK), a QPSK (Quadrature-PSK) or an 8-PSK. The m-QAM may be a 16-QAM, a 64-QAM or a 256-QAM.

The power allocator 520 allocates proper powers to the transmission stream. The power-adjusted transmission stream is denoted by

ŝ.

The precoder 530 generates a transmission signal vector x by applying the adjusted transmission

ŝ

to a weight matrix W. The transmission signal x is transmitted through the RF unit 540.

The controller 550 receives channel state information (CSI) and controls the precoder 530, the data processor 510 and the power allocator 520. The CSI can be any form which shows channel information. The CSI can be feedback information indicating a status of a downlink channel or an uplink channel. The controller 550 provides the precoder 530 with information of the weight matrix W. The controller 550 supplies a control signal to the data processor 510 and the power allocator 520. The control signal may be a MCS index or a power offset value which is transmitted as the CSI.

FIG. 6 is a block diagram illustrating a receiver according to an embodiment of the invention.

Referring to FIG. 6, the receiver 600 includes a data processor 610, a channel estimator 620, an equalizer 630, an RF unit 640 and a controller 650.

The RF unit 640 receives a signal through a receiving antenna and outputs a receiving signal. The receiving signal is expressed as y. The channel estimator 620 estimates a channel. The channel estimator 620 may estimate the channel using a pilot signal included in the receiving signal. An estimated channel is expressed as

ĥ.

The equalizer 630 equalizes the receiving signal using the estimated channel. The data processor 610 corresponds to the data processor 510 of the transmitter 500. The data processor 610 demaps and decodes an input signal the signal to reproduce the information bits.

The controller 650 receives the estimated channel from the channel estimator 620 and generates the CSI as feedback information. The controller 650 transmits feedback information to the transmitter 500.

FIG. 7 is a diagram illustrating a method for transmitting channel state information.

Referring to FIG. 7, a base station (BS) transmits a pilot signal and receives a CSI from a plurality of user equipments (UEs). The BS applies a weight matrix to a transmission stream to generate a transmission signal while changing the weight matrix. After a predetermined time elapses, the BS transmits the pilot signal again.

The BS uses weight matrixes W, W′ and W″ which are changed depending on the transmission stream. The weight matrixes W, W′ and W″ are unitary matrixes of N_T×N_T where N_T is the number of transmitting antenna at the BS. The column vectors of a unitary matrix are orthogonal to each other. The column vectors of the weight matrix are linearly independent to each other. The column vectors of the weight matrix can be considered as a basis of a N_T dimensional space. W′ and W″ can be obtained by transforming the weight matrix W in accordance with a predetermined rule. For example, W′ and W″ can be obtained by linearly transforming the weight matrix W.

The BS can pseudo-randomly transform the weight matrix. The weight matrix and the transforming pattern of the weight matrix are known by both UEs and the BS.

Although three weight matrixes have been described, the inventive concept is not limited to it. Two or more transformations of the weight matrix can be performed and three or more weight matrix can be used.

FIG. 8 is a flowchart illustrating a method for transmitting CSI according to an embodiment of the invention.

Referring to FIG. 8, a BS transmits a pilot signal (S210). The pilot signal is a signal known in advance by both the BS and a UE.

The UE receives the pilot signal through a receiving antenna (S220). The received pilot signal can be expressed by y_i where the subscript i means the i-th UE. A channel from a j-th transmitting antenna to an i-th UE is expressed by h_ij. A channel from the N_T transmitting antennas to the i-th UE is expressed in the form of a vector by:
h_i^T=[h_i1,h_i2, . . . ,h_iNT].  Equation (25)

The receiving signal y_i received by the i-th UE can be expressed as:
y_i=h_i^Hr+n_i  Equation (26)

where n_i denotes Additive White Gaussian Noise (AWGN) received by the i-th UE and r denotes the pilot signal.

The UE estimates the channel using the pilot signal and acquires an estimated channel (S230). Since the UE knows the pilot signal r, the channel can be easily obtained, which is called an estimated channel

ĥ_i

of a i-th user equipment as shown in:
ĥ_i^T=[ĥ_i1,ĥ_i2, . . . ,ĥ_iNT].  Equation (27)

The UE obtains a basis representing the estimated channel (S240). The basis may be one of weight matrixes W, W′ and W″. The weight matrixes W, W′ and W″ are the same as described with reference to FIG. 7. The estimated channel

ĥ_i

is a vector having N_T components and belongs to a N_T dimensional space. The weight matrixes W, W′ and W″ are unitary matrixes and serve as the basis. The estimated channel

ĥ_i

can be expressed as linear combination of the column vectors of any one of weight matrixes W, W′ and W″.

The process for acquiring a basis of the estimated channel will be described in detail. For clarity, it is assumed that the number of transmitting antennas N_T is 4.

The weight matrix W can be expressed as:

$\begin{array}{cc}W=\left[w_1{w}_2{w}_3{w}_4\right]=[\begin{array}{c}{w}_{11}{w}_{12}{w}_{13}{w}_{14}\\ w_{21}{w}_{22}{w}_{23}{w}_{24}\\ w_{31}{w}_{32}{w}_{33}{w}_{34}\\ w_{41}{w}_{42}{w}_{43}{w}_{44}\end{array}[.\begin{array}{c}\\ \\ \\ \end{array}& \mathrm{Equation}\left(28\right)\end{array}$

The weight matrix W′ can be expressed as:

$\begin{array}{cc}{W}^{\prime }=\mathrm{TW}=\left[w_1^{\prime }{w}_2^{\prime }{w}_3^{\prime }{w}_4^{\prime }\right]=[\begin{array}{c}{w}_{11}^{\prime }{w}_{12}^{\prime }{w}_{13}^{\prime }{w}_{14}^{\prime }\\ w_{21}^{\prime }{w}_{22}^{\prime }{w}_{23}^{\prime }{w}_{24}^{\prime }\\ w_{31}^{\prime }{w}_{32}^{\prime }{w}_{33}^{\prime }{w}_{34}^{\prime }\\ w_{41}^{\prime }{w}_{42}^{\prime }{w}_{43}^{\prime }{w}_{44}^{\prime }\end{array}[\begin{array}{c}\\ \\ \\ \end{array}& \mathrm{Equation}\left(29\right)\end{array}$

where a matrix T is a matrix performing a linear transformation.

The weight matrix W″ can be expressed as:

$\begin{array}{cc}{W}^{″}={\mathrm{TW}}^{\prime }=\left[w_1^{″}{w}_2^{″}{w}_3^{″}{w}_4^{″}\right]=[\begin{array}{c}{w}_{11}^{″}{w}_{12}^{″}{w}_{13}^{″}{w}_{14}^{″}\\ w_{21}^{″}{w}_{22}^{″}{w}_{23}^{″}{w}_{24}^{″}\\ w_{31}^{″}{w}_{32}^{″}{w}_{33}^{″}{w}_{34}^{″}\\ w_{41}^{″}{w}_{42}^{″}{w}_{43}^{″}{w}_{44}^{″}\end{array}[.\begin{array}{c}\\ \\ \\ \end{array}& \mathrm{Equation}\left(30\right)\end{array}$

Using a standard basis, the estimated channel

ĥ_i

can be expressed by:

$\begin{array}{cc}{\hat{h}}_i=[\begin{array}{c}{\hat{h}}_{i1}\\ {\hat{h}}_{i2}\\ {\hat{h}}_{i3}\\ {\hat{h}}_{i4}\end{array}[={\hat{h}}_{i1}[\begin{array}{c}1\\ 0\\ 0\\ 0\end{array}[+{\hat{h}}_{i2}[\begin{array}{c}0\\ 1\\ 0\\ 0\end{array}[+{\hat{h}}_{i3}[\begin{array}{c}0\\ 0\\ 1\\ 0\end{array}[+{\hat{h}}_{i4}[\begin{array}{c}0\\ 0\\ 0\\ 1\end{array}[.& \mathrm{Equation}\left(31\right)\end{array}$

The standard basis means a basis which comprises a set of column vectors of an identity matrix.

The basis of the estimated channel

ĥ_i

expressed by Equation (31) can be changed by using the column vectors of the weight matrix W as a basis, thereby obtaining

$\begin{array}{cc}\begin{array}{c}{\hat{h}}_i=[\begin{array}{c}{\hat{h}}_{i1}\\ {\hat{h}}_{i2}\\ {\hat{h}}_{i3}\\ {\hat{h}}_{i4}\end{array}[\\ ={\mathrm{aw}}_1+{\mathrm{bw}}_2+{\mathrm{cw}}_3+{\mathrm{dw}}_4\\ =a[\begin{array}{c}{w}_{11}\\ w_{21}\\ w_{31}\\ w_{41}\end{array}[+b[\begin{array}{c}{w}_{12}\\ w_{22}\\ w_{32}\\ w_{42}\end{array}[+c[\begin{array}{c}{w}_{13}\\ w_{23}\\ w_{33}\\ w_{43}\end{array}[+d[\begin{array}{c}{w}_{14}\\ w_{24}\\ w_{34}\\ w_{44}\end{array}[.\end{array}& \mathrm{Equation}\left(32\right)\end{array}$

Acquiring a basis of an estimated channel means that the standard basis of the estimated channel is changed to a basis which comprises the column vectors of the weight matrix.

The UE selects a representative vector (S250). The representative vector is a column vector having the largest coefficient. For example, the coefficient of the column vector for the weight matrix W is value of a, b, c and d in Equation (32) respectively. Each value of a, b, c and d represents the coefficient in the direction of the column vectors w₁, w₂, w₃ and w₄ respectively. Each value of a, b, c and d means a value obtained by projecting the estimated channel

ĥ_i

onto each column vectors w₁, w₂, w₃ and w₄. The column vector having the largest coefficient (having the largest value of a, b, c, and d) most similarly represents the estimated channel and becomes the representative vector. The UE acquires an index of the representative vector.

Representative vectors can also be obtained for each weight matrixes W′ and W″. The estimated channel

ĥ_i

using column vectors of the weight matrix W′ of Equation (29) can be written as:

$\begin{array}{cc}\begin{array}{c}{\hat{h}}_i=[\begin{array}{c}{\hat{h}}_{i1}\\ {\hat{h}}_{i2}\\ {\hat{h}}_{i3}\\ {\hat{h}}_{i4}\end{array}[\\ =a^{\prime }{w}_1^{\prime }+b^{\prime }{w}_2^{\prime }+c^{\prime }{w}_3^{\prime }+d^{\prime }{w}_4^{\prime }\\ =a^{\prime }[\begin{array}{c}{w}_{11}^{\prime }\\ w_{21}^{\prime }\\ w_{31}^{\prime }\\ w_{41}^{\prime }\end{array}[+b^{\prime }[\begin{array}{c}{w}_{12}^{\prime }\\ w_{22}^{\prime }\\ w_{32}^{\prime }\\ w_{42}^{\prime }\end{array}[+c^{\prime }[\begin{array}{c}{w}_{13}^{\prime }\\ w_{23}^{\prime }\\ w_{33}^{\prime }\\ w_{43}^{\prime }\end{array}[+d^{\prime }[\begin{array}{c}{w}_{14}^{\prime }\\ w_{24}^{\prime }\\ w_{34}^{\prime }\\ w_{44}^{\prime }\end{array}[.\end{array}& \mathrm{Equation}\left(33\right)\end{array}$

The UE acquires the values of a′, b′, c′ and d′ and an index of the column vector having the largest coefficient.

The estimated channel

ĥ_i

using column vectors of the weight matrix W″ of Equation (30) can be written as:

$\begin{array}{cc}\begin{array}{c}{\hat{h}}_i=[\begin{array}{c}{\hat{h}}_{i1}\\ {\hat{h}}_{i2}\\ {\hat{h}}_{i3}\\ {\hat{h}}_{i4}\end{array}[\\ =a^{″}{w}_1^{″}+b^{″}{w}_2^{″}+c^{″}{w}_3^{″}+d^{″}{w}_4^{″}\\ =a^{″}[\begin{array}{c}{w}_{11}^{″}\\ w_{21}^{″}\\ w_{31}^{″}\\ w_{41}^{″}\end{array}[+b^{″}[\begin{array}{c}{w}_{12}^{″}\\ w_{22}^{″}\\ w_{32}^{″}\\ w_{42}^{″}\end{array}[+c^{″}[\begin{array}{c}{w}_{13}^{″}\\ w_{23}^{\prime }\\ w_{33}^{″}\\ w_{43}^{″}\end{array}[+d^{″}[\begin{array}{c}{w}_{14}^{″}\\ w_{24}^{″}\\ w_{34}^{″}\\ w_{44}^{″}\end{array}[.\end{array}& \mathrm{Equation}\left(34\right)\end{array}$

The UE acquires the values of a″, b″, c″ and d″ and an index of the column vector having the largest coefficient.

The UE transmits representative vectors (S260). When there are three weight matrixes, three representative vectors can be obtained for each weight matrixes. For example, let indexes 1, 2, 3, and 4 be sequentially given to the column vectors w₁, w₂, w₃ and w₄ of W. Let indexes 5, 6, 7, and 8 be given to the column vectors w₁′, w₂′, w₃′ and w₄′ of W′. Let indexes 9, 10, 11, and 12 be given to the column vectors w₁″, w₂″, w₃″ and w₄″ of W″. Among the values acquired using Equations 32 to 34, it is assumed that a of a, b, c and d has the largest coefficient, b′ of a′, b′, c′ and d′ has the largest coefficient, and d″ of a″, b″, c″ and d″ has the largest coefficient. The vector w₁ of W is a representative vector of W, the vector w₂′ of W′ is a representative vector of W′ and the vector w₄″ of W″ is a representative vector of W″. Then, the UE can feedback the index 1 and the value a of the corresponding column vector, the index 6 and the value b′ of the corresponding column vector and the index 12 and the value d″ of the corresponding column vector. In the following description, the CSI fi is expressed as fi={(1,a), (6,b′), (12,d″)} for the purpose of convenient expression.

In this way, the UE can transmit the CSI based on the basis which spans vector space of the weight matrix. Since the UE transmits the CSI at a time by using a series of weight matrixes, it is not necessary to transmit the CSI every when the BS changes the weight matrix. It is possible to reduce feedback overhead.

FIG. 9 is a flowchart illustrating a method for transmitting data according to an embodiment of the invention.

Referring to FIG. 9, the BS transmits a pilot signal (S310). The BS receives information for a plurality of representative vectors from UEs (S320). The UE transmits the CSI to the BS. For example, it is assumed that weight matrixes sequentially used in the BS is W, W′ and W″. The CSI comprises indexes and coefficients of the representative vectors in Equations (32) to (34). The BS schedules the UEs to which the transmission stream should be transmitted based on the CSI. The criterion for scheduling UEs can vary. For example, the UEs can be selected so that the UEs are most closely orthogonal to each other. The UEs can be schedules depending on the basis. For example, it is assumed that the number of transmitting antennas N_T is 4 and the number of UEs in a cell is 6. The weight matrix W can be used as the weight matrix of UEs 1, 2, 3 and 4. The weight matrix W′ can be used as the weight matrix of UEs 1, 3, 4 and 5. The weight matrix W″ can be used as the weight matrix of UEs 2, 4, 5 and 6.

Signals are beam-formed and transmitted sequentially in accordance with the plurality of representative vectors (S330). For example, it is assumed that the CSI fi received from the i-th UE is fi={(1,a), (6,b′), (12,d″)} and the i-th UE is selected by (6, b′). This means that the channel state of the i-th UE can be represented best by the second column vector w₂′ when the weight matrix W is used as a basis. The BS transmits a transmission signal x by applying the weight matrix W′ to a transmission stream of the i-th UEs. The transmission signal x can be written as:

$\begin{array}{cc}x=\left[w_1^{\prime }{w}_2^{\prime }{w}_3^{\prime }{w}_4^{\prime }\right][\begin{array}{c}{s}_1\\ s\\ s_3\\ s_4\end{array}[& \mathrm{Equation}\left(35\right)\end{array}$

where s₁, s₃ and s₄ are transmission streams for the other UEs.

It is assumed that the CSI fj received from the j-th user equipment is fj={(3,c), (5,a′), (11,c″)} and the j-th UE is selected by (3, c). It is also assumed that the transmission stream of the j-th UE is q. Then, the BS transmits a transmission signal x which can be shown as:

$\begin{array}{cc}x=\left[w_1{w}_2{w}_3{w}_4\right][\begin{array}{c}{s}_1\\ s_2\\ q\\ s_4\end{array}[& \mathrm{Equation}\left(36\right)\end{array}$

where s₁, s₂ and s₄ are transmission streams for the other UEs.

In this way, since the BS need not transmit the weight matrix every time when the weight matrix is changed, it is possible to reduce feedback overhead.

FIG. 10 is a flowchart illustrating a method for transmitting CSI according to another embodiment of the invention.

Referring to FIG. 10, a UE prepares a basis representing the channel (S410). It is assumed that the number of transmission antennas N_T is 4. Then, the estimated channel

ĥ_i

of the i-th UE can be expressed by:

$\begin{array}{cc}{\hat{h}}_i=[\begin{array}{c}{\hat{h}}_{i1}\\ {\hat{h}}_{i2}\\ {\hat{h}}_{i3}\\ {\hat{h}}_{i4}\end{array}[.& \mathrm{Equation}\left(37\right)\end{array}$

The UE selects a basis representing the estimated channel

ĥ_i.

For example, the UE can select the column vectors of the weight matrix W as a basis. The estimated channel can be expressed by:

$\begin{array}{cc}\begin{array}{c}{\hat{h}}_i=[\begin{array}{c}{\hat{h}}_{i1}\\ {\hat{h}}_{i2}\\ {\hat{h}}_{i3}\\ {\hat{h}}_{i4}\end{array}[\\ ={\mathrm{aw}}_1+{\mathrm{bw}}_2+{\mathrm{cw}}_3+{\mathrm{dw}}_4\\ =a[\begin{array}{c}{w}_{11}\\ w_{21}\\ w_{31}\\ w_{41}\end{array}[+b[\begin{array}{c}{w}_{12}\\ w_{22}\\ w_{32}\\ w_{42}\end{array}[+c[\begin{array}{c}{w}_{13}\\ w_{23}\\ w_{33}\\ w_{43}\end{array}[+d[\begin{array}{c}{w}_{14}\\ w_{24}\\ w_{34}\\ w_{44}\end{array}[.\end{array}& \mathrm{Equation}\left(38\right)\end{array}$

The UE selects at least two representative vectors from the basis (S420). The criterion for selecting the representative vectors may be the order of coefficients. For example, it is assumed that the coefficient in Equation 38 decreases in the order of b, d, c and a. The vectors w₂ and w₄ are representative vectors. The column vector w₂ is a first representative vector of W and the column vector w₄ is a second representative vector of W. Compared with only one representative vector, it is possible to more accurately represent the estimated channel by choosing two or more representative vectors.

It is possible to completely transmit the estimated channel without any loss by using all column vectors constituting the weight matrix. However, such a method loses a meaning of the efficient CSI transmission. The fact that an UE uses all the column vectors means that the UE uses all the transmitting antennas. Thus the multi-user diversity has no gain. It is need that the CSI is transmitted using a proper number of column vectors among the column vectors constituting a basis.

The UE transmits the representative vectors as the CSI (S430). The CSI may be the indexes and the coefficients of the representative vectors. For example, when the indexes of the column vectors w₁, w₂, w₃, and w₄ of W in Equation (38) are sequentially 1, 2, 3 and 4 and w₂ and w₄ are selected as two representative vectors, the CSI fi={(2,b), (4,d)} can be transmitted.

Streams can be transmitted to the UE which requires necessarily a transmission while minimizing the interference. For example, an UE requires re-transmission in the HARQ or an UE requires the transmission due to the full of the buffer thereof. Since the channel information can accurately be represented, it is possible to effectively perform communications even when the number of UEs in a cell is small.

FIG. 11 is a diagram illustrating a basis and an estimated channel.

Referring to FIG. 11, it is assumed that x, y, and z vectors constituting the basis span three dimensional space. The estimated channel can be expressed by:
ĥ_i=αx+βy+γz.  Equation 39

For clarity, it is assumed that the order of coefficients is the order of α, β and γ and indexes of vectors x, y and z are sequentially 1, 2 and 3. Then, rather than the case where the estimated channel is represented using the index 1 of vector x and the coefficient α in the direction of vector x, the estimated channel can be accurately represented using the index 1 of vector x, the coefficient a in the direction of the vector x, the index 2 of vector y and the coefficient β in the direction of vector y.

The CSI may comprise an MCS index and a power offset value. The MCS index indicates one MCS selected from a predetermined MCS set. The SINR generally means a ratio obtained by selecting a vector having the largest coefficient and dividing the square of the coefficient of the vector by a value obtained by adding the sum of the squares of the coefficients of other vectors to the power of the noise. The continuous SINR can be optimized in the Gaussian modulation at the transmitter. However, the modulation is generally performed discretely and one modulation scheme among a set of several MCS is selected. More meaningful information of the actual transmitter may be a maximum MCS index and a power offset value for the maximum MCS index. The power offset value means an offset value which should be added to or subtracted from a reference transmission power in the transmitter. The CSI may not comprise the power offset value.

When the CSI is transmitted using two or more column vectors, various combinations are possible. For example, consider that three vectors from the given basis are selected. First, an UE can transmit the indexes of three vectors and the SINR values as the CSI. The number of actual values to be transmitted is 6 in total of 3 indexes of the column vectors and 3 values of the SINR. Second, the UE can transmit a MCS index, a power offset value and the indexes of three vectors as the CSI. The number of actual values to be transmitted is 5 in total of 3 indexes of the vectors, 1 MCS index and 1 power offset value. It can be seen that the number of values to be transmitted is reduced compared with the first scheme. Also, since range of the MCS index is generally narrower than that of the SINR, the less number of bits can be used. It is possible to reduce the actual data amount to be transmitted. When the MCS index and the power offset value are transmitted instead of the SINR, the interval for transmitting the CSI may be increase so that feedback overhead is reduced.

FIG. 12 is a diagram illustrating a method for transmitting CSI according to another aspect of the invention.

Referring to FIG. 12, a BS prepares a set of bases, not one basis. A basis is a weight. The BS uses unitary matrixes as weight matrixes.

For example, the number of bases in the set of bases may be three and the set of bases can be expressed as B={W¹, W², W³}. The bases W¹, W², W³ can be selected randomly. The selection range of bases, that is, weight matrixes, can be widened. The order of bases to be used may be the order of W², W³ and W¹ or the order of W³, W¹ and W². The BS and UEs are aware of the set of bases and the order of changing the bases in advance. Alternatively, the order of weight matrixes can be notified from the BS to UEs.

FIG. 13 is a flowchart illustrating operations of the embodiment of FIG. 12.

Referring to FIG. 13, a set of bases representing an UE's channel is prepared (S510). The set of bases may be known to the UE. The set of bases may be a set of weight matrixes used in the BS.

The UE selects a representative vector from the set of bases (S520). When it is assumed that the number of transmitting antennas N_T of the transmitter is 4, a i-th UE's estimated channel can be expressed by:

$\begin{array}{cc}{\hat{h}}_i=[\begin{array}{c}{\hat{h}}_{i1}\\ {\hat{h}}_{i2}\\ {\hat{h}}_{i3}\\ {\hat{h}}_{i4}\end{array}[& \mathrm{Equation}\left(40\right)\end{array}$

A representative vector best matched with the estimated channel is found out of the set of bases. To find out the representative vector, one basis is not used, but a plurality of bases is used. The probability that the estimated channel is matched with the basis well can be improved.

Transformation of the estimated channel

ĥ_i

using the weight matrix W¹ as the basis can be written as:

$\begin{array}{cc}\begin{array}{c}{\hat{h}}_i=[\begin{array}{c}{\hat{h}}_{i1}\\ {\hat{h}}_{i2}\\ {\hat{h}}_{i3}\\ {\hat{h}}_{i4}\end{array}[\\ ={\mathrm{aw}}_1^1+{\mathrm{bw}}_2^1+{\mathrm{cw}}_3^1+{\mathrm{dw}}_4^1\\ =a[\begin{array}{c}{w}_{11}^1\\ w_{21}^1\\ w_{31}^1\\ w_{41}^1\end{array}[+b[\begin{array}{c}{w}_{12}^1\\ w_{22}^1\\ w_{32}^1\\ w_{42}^1\end{array}[+c[\begin{array}{c}{w}_{13}^1\\ w_{23}^1\\ w_{33}^1\\ w_{43}^1\end{array}[+d[\begin{array}{c}{w}_{14}^1\\ w_{24}^1\\ w_{34}^1\\ w_{44}^1\end{array}[\end{array}& \mathrm{Equation}\left(41\right)\end{array}$

where w₁¹, w₂¹, w₃¹ and w₄¹ denote the column vectors of the weight matrix W¹. The UE acquires values of a, b, c and d and the index of the column vector having the largest coefficient. Values of a, b, c and d mean values obtained by projecting the estimated channel

ĥ_i

onto the column vectors w₁¹, w₂¹, w₃¹, and w₄¹ of the weight matrix W¹ constituting the basis. The largest coefficient among a, b, c and d represents the estimated channel

ĥ_i

most closely.

Transformation of the estimated channel

ĥ_i

using the weight matrix W² as the basis can be written as:

$\begin{array}{cc}\begin{array}{c}{\hat{h}}_i=[\begin{array}{c}{\hat{h}}_{i1}\\ {\hat{h}}_{i2}\\ {\hat{h}}_{i3}\\ {\hat{h}}_{i4}\end{array}[\\ =a^{\prime }{w}_1^2+b^{\prime }{w}_2^2+c^{\prime }{w}_3^2+d^{\prime }{w}_4^2\\ =a^{\prime }[\begin{array}{c}{w}_{11}^2\\ w_{21}^2\\ w_{31}^2\\ w_{41}^2\end{array}[+b^{\prime }[\begin{array}{c}{w}_{12}^2\\ w_{22}^2\\ w_{32}^2\\ w_{42}^2\end{array}[+c^{\prime }[\begin{array}{c}{w}_{13}^2\\ w_{23}^2\\ w_{33}^2\\ w_{43}^2\end{array}[+d^{\prime }[\begin{array}{c}{w}_{14}^2\\ w_{24}^2\\ w_{34}^2\\ w_{44}^2\end{array}[\end{array}& \mathrm{Equation}\left(42\right)\end{array}$

where w₁², w₂²,w₃² and w₄² denote column vectors of the weight matrix W². The UE acquires the values of a′, b′, c′ and d′ and the index of the column vector having the largest coefficient.

Transformation of the estimated channel

ĥ_i

using the weight matrix W³ as the basis can be written as:

$\begin{array}{cc}\begin{array}{c}{\hat{h}}_i=[\begin{array}{c}{\hat{h}}_{i1}\\ {\hat{h}}_{i2}\\ {\hat{h}}_{i3}\\ {\hat{h}}_{i4}\end{array}[\\ =a^{″}{w}_1^3+b^{″}{w}_2^3+c^{″}{w}_3^3+d^{″}{w}_4^3\\ =a^{″}[\begin{array}{c}{w}_{11}^3\\ w_{21}^3\\ w_{31}^3\\ w_{41}^3\end{array}[+b^{″}[\begin{array}{c}{w}_{12}^3\\ w_{22}^3\\ w_{32}^3\\ w_{42}^3\end{array}[+c^{″}[\begin{array}{c}{w}_{13}^3\\ w_{23}^3\\ w_{33}^3\\ w_{43}^3\end{array}[+d^{″}[\begin{array}{c}{w}_{14}^3\\ w_{24}^3\\ w_{34}^3\\ w_{44}^3\end{array}[\begin{array}{c}{}^{}\\ \\ \\ \underset{}{}\end{array}\end{array}& \mathrm{Equation}\left(43\right)\end{array}$

where w₁³, w₂³, w₃³ and w₄³ denote column vectors of the weight matrix W³. The UE acquires values of a″, b″, c″ and d″ and the index of the column vector having the largest coefficient.

The UE transmits an index of the representative vector and an index of the basis for the representative vectors (S530). The UE selects a representative vector having the largest coefficient from the set of bases. The UE transmits the index of the representative vector and the index of the basis for the representative vector to the BS. For example, if the fourth column vector of w₄¹ has the largest coefficient, the index of the basis for the column vector is 4 for the weight matrix W¹. The UE transmits the index 4 of the column vector and the index 1 of the basis for the column vector as the CSI. The CSI may include a coefficient of the representative vector.

The probability that the UE's channel is not matched with the basis is reduced in inverse proportion to the number of bases. The UE need not transmit the CSI every time when the basis used in the BS is changed. It is possible to prevent the waste of the channel capacity.

FIG. 14 is a flowchart illustrating a method for transmitting data according to another embodiment of the invention.

Referring to FIG. 14, a BS receives information based on a set of bases (S610). The information based on the set of bases may be an index of the basis best representing the UE's channel among the set of bases and an index of a representative vector of the basis. The information may also include a coefficient of the representative vector.

The BS schedules UEs to which a transmission stream should be transmitted based on the CSI (S620). The scheduling scheme may vary depending on a system. For example, the BS may select UEs which are orthogonal to each other most closely.

The transmission stream is sequentially beam-formed using the set of bases and transmitted to the selected UEs (S630). The transmission stream beam-formed using the weight matrix forms a multiplexed stream. The multiplexed stream is transmitted through a plurality of transmitting antennas. After a predetermined time elapses, the BS changes the basis and then selects UEs to which the transmission stream should be transmitted according the basis.

The probability that the UE's channel is not matched with the basis is reduced in inverse proportion to the number of bases. The BS need not transmit the weight matrix to the UE every time when the weight matrix is changed. It is possible to prevent the waste of the channel capacity.

FIG. 15 is a diagram illustrating a method for transmitting CSI according to still another embodiment of the invention. FIG. 16 is a flowchart illustrating operation of the embodiment of FIG. 15.

Referring to FIGS. 15 and 16, an UE receives a pilot signal (S710). The UE acquires an estimated channel by using the pilot signal (S720). It is assumed that the number of transmitting antennas N_T of the transmitter is 4. The estimated channel of the i-th terminal can be expressed by:

$\begin{array}{cc}{\hat{h}}_i=[\begin{array}{c}{\hat{h}}_{i1}\\ {\hat{h}}_{i2}\\ {\hat{h}}_{i3}\\ {\hat{h}}_{i4}\end{array}[.& \mathrm{Equation}\left(44\right)\end{array}$

The basis may be a weight matrix of the BS. When the number of transmitting antennas N_T is 4, the weight matrix W can be expressed by:

$\begin{array}{cc}w=\left[w_1{w}_2{w}_3{w}_4\right][\begin{array}{c}{w}_{11}{w}_{12}{w}_{13}{w}_{14}\\ w_{21}{w}_{22}{w}_{23}{w}_{24}\\ w_{31}{w}_{32}{w}_{33}{w}_{34}\\ w_{41}{w}_{42}{w}_{43}{w}_{44}\end{array}[& \mathrm{Equation}\left(45\right)\end{array}$

where w_i denotes a column vector. The weight matrix W is an unitary matrix.

The UE acquires a phase shift of the basis and a representative vector to represent the estimated channel (S730). Column vectors of the weight matrix W can be shifted in phase as:
W′=e^jθiW=e^jθi[w₁w₂w₃w₄]  Equation (46)

where θ denotes a phase shift value.

Although the unitary matrix is phase-shifted, property of the unitary matrix remains as can be seen as:
(e^jθiw_i)^He^jθiw_i=e^−jθiw_i^He^jθiw_i=w_i^Hw≠0, (e^jθiw_i)^He^jθjw_j=e^−jθiw_i^He^jθjw_j=e^−j(θ^i−θj)(w_i^Hw_j)=0, (i≠j).  Equation (47)

A phase-shifted weight matrix W′ can be also used as the basis. The phase shift value θ may be a predetermined value. Alternatively, the phase shift value θ may transmitted from the BS to the UE. The same phase shift value can be applied to all the column vectors constituting the basis. Or different phase shift value can be applied to each column vector.

When the phase shift is applied to the basis, the number of bases virtually increases as many as the number of bases which are shifted in phase. The UE selects not only a basis from the predetermined set of bases but also phased-shifted bases. As the selectable number of bases increases, the probability that the UE's channel is matched with the basis is increased.

The UE transmits an index of a representative vector of the basis and the phase shift of the basis as CSI (S740). The UE can feedback the index of the representative vector and the index of the phase change as the CSI to the BS.

It is possible to obtain various column vectors by phase shift. The probability of the mismatch between the actual channel and the column vectors can be reduced.

Here, we use column vectors to span a weight matrix which is a basis. But skilled in art can use row vectors to span the weight matrix. The column vectors are exemplary of vectors which span the basis.

A basis is directly not transmitted to an UE. The basis is predefined. It is possible to prevent a waste of the channel capacity due to the transmission of the basis. In addition, it is possible to reduce a mismatch between the actual channel and the basis. The waste of entire channel capacity can be prevented and the transmission efficiency can be enhanced.

As the present invention may be embodied in several forms without departing from the spirit or essential characteristics thereof, it should also be understood that the above-described embodiments are not limited by any of the details of the foregoing description, unless otherwise specified, but rather should be construed broadly within its spirit and scope as defined in the appended claims. Therefore, all changes and modifications that fall within the metes and bounds of the claims, or equivalence of such metes and bounds are intended to be embraced by the appended claims.

Claims

1. A method for transmitting data in a multiple antenna system, the method comprising:

preparing a plurality of weight matrices for a plurality of antennas;

receiving information for a plurality of representative vectors which are selected for each weight matrix and represent channel information;

beam-forming a signal in accordance with the plurality of representative vectors using the plurality of weight matrices; and

transmitting the signal through the plurality of antennas.

2. The method of claim 1, wherein the information for the plurality of representative vectors are indexes of the plurality of representative vectors.