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.
|
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
|
where si is a stream transmitted through a i-th transmitting antenna.
The streams s1, s2, . . . , and sNT can be transmitted with different transmission powers. When it is assumed that the transmission powers are P1, P2, . . . , and PNT, the streams which the transmission powers are adjusted can be expressed in a vector as:
ŝ=[ŝ1,ŝ2, . . . ,ŝN
Using a diagonal matrix P of which diagonal elements are the transmission powers, the power-adjusted stream
ŝ
can be expressed as:
Consider that a weight matrix W is applied to the power adjusted stream vector
ŝ
so that N transmission signals x1, x2, . . . , and xNT are formed. The transmission signals x1, x2, . . . , and xNT can be expressed by a vector x as:
where a weight wij (1≦i≦NT, 1≦j≦NT) is a weighting value of a j-th stream for an i-th transmitting antenna. The weight wij 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 yNR for the NR receiving antennas are written as:
y=[y1y2 . . . yN
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 hij. 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.
Referring to
hiT=[hi1,hi2, . . . ,hiN
where a superscript T denotes a transpose.
The channels from the NT transmitting antennas to the NR receiving antennas can be expressed in the form of a matrix by:
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 ni and AWGNs for all channels can be expressed in the form of a vector by:
n=[n1n2 . . . nN
The receiving signal y can be expressed by:
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 NR antennas. Or NR user equipments may concurrently receive downlink signals through one antenna. In any combination, the total number of receiving antennas can be kept NR. This technique is generally called a MIMO BC (Broadcast Channel) or an SDMA (Space Division Multiple Access).
Reversely, an user equipment having NT antennas may transmit uplink signals. Or NT user equipments having one antenna may concurrently transmit uplink signals. In any combination, the total number of transmitting antennas can be kept NT. 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.
Referring to
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=UΣVH 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 UH as shown in:
y=Hx+n=UΣVHx+n=UΣVH Wŝ+n=UΣVHVŝ+n=UΣŝ+n,
{tilde over (y)}=UHy=UH UΣŝ+UHn=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.
Referring to
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 hi be elements of the i-th column of H as shown in:
hiT=[hi1,hi2, . . . ,hiN
When the channel matrix H is the unitary matrix, inner product between two different column vectors hi and hj of H is zero. This relation can be written as:
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=HH. 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 hi 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 wNT 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 . . . wN
A receiving signal yi received by the i-th receiving antenna can be expressed by:
In the RBF, the i-th receiving antenna's SINR for a weight vector wm can be defined by:
The weight vector wm in which the SINRi,m of Equation (18) is the maximum can be found out. We call it a maximum weight vector wmax.
Assume that an a NT dimension complex vector space
CN
is spanned by weight vectors w1, w2 . . . , and wNT. The vector space
CN
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
CN
the inner product of hi and any one of weight vectors is zero as shown in:
hiHwi= . . . =hiHwN
Since the weight matrix W is assumed as a unitary matrix, the null space of
CN
is a zero space. But the channel hi does not belong to the zero space. The product cannot be zero as shown in:
hiHwk≠0 (k=1,2, . . . ,NT). 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 hi should belong to one of subspaces L(w1), L(w2), . . . , and L(wNT) of the vector spaces. Geographically, subspaces L(w1), L(w2), . . . , and L(wNT) denote lines in the NT dimension complex vector space
CN
using weight vectors w1, w2 . . . , and wNT as a basis of the vector space
CN
The basis of a vector space
CN
is a linearly independent subset of
CN
that spans (or generates)
CN
When the weight matrix W is the basis, the basis comprises column (row) vectors of the weight matrix W.
When the channel hi belongs to a subspace L(wm), it gives:
Accordingly, if we find out the weight vector wm which belongs to the subspace L(wm), we maximize the SINR (wm=wmax). The maximum weight vector wmax is a vector which has a maximum coefficient when the vector is projected onto the basis of the vector space
CN
The SINR of Equation (18) indicates how closely the corresponding channel is orthogonal to one of weight vectors w1, w2, . . . and wNT.
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 NT. 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 NT dimension vector space can be expressed by a basis including NT 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:
Using the standard basis, we need the total NT coefficients of h1, h2, . . . , and hNT to represent the channel h. The channel h can be reproduced by combining h1, h2, . . . , and hNT and the standard basis.
If the channel h=[h1, h2, . . . , and hNT]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:
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:
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 NT (where NT 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.
Referring to
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.
Referring to
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.
Referring to
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 NT×NT where NT 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 NT 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.
Referring to
The UE receives the pilot signal through a receiving antenna (S220). The received pilot signal can be expressed by yi 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 hij. A channel from the NT transmitting antennas to the i-th UE is expressed in the form of a vector by:
hiT=[hi1,hi2, . . . ,hiN
The receiving signal yi received by the i-th UE can be expressed as:
yi=hiHr+ni Equation (26)
where ni 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
ĥi
of a i-th user equipment as shown in:
ĥiT=[ĥi1,ĥi2, . . . ,ĥiN
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
ĥi
is a vector having NT components and belongs to a NT dimensional space. The weight matrixes W, W′ and W″ are unitary matrixes and serve as the basis. The estimated channel
ĥi
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 NT is 4.
The weight matrix W can be expressed as:
The weight matrix W′ can be expressed as:
where a matrix T is a matrix performing a linear transformation.
The weight matrix W″ can be expressed as:
Using a standard basis, the estimated channel
ĥi
can be expressed by:
The standard basis means a basis which comprises a set of column vectors of an identity matrix.
The basis of the estimated channel
ĥi
expressed by Equation (31) can be changed by using the column vectors of the weight matrix W as a basis, thereby obtaining
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
ĥi
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
ĥi
using column vectors of the weight matrix W′ of Equation (29) can be written as:
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
ĥi
using column vectors of the weight matrix W″ of Equation (30) can be written as:
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.
Referring to
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:
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:
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.
Referring to
ĥi
of the i-th UE can be expressed by:
The UE selects a basis representing the estimated channel
ĥi.
For example, the UE can select the column vectors of the weight matrix W as a basis. The estimated channel can be expressed by:
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.
Referring to
ĥi=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.
Referring to
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.
Referring to
The UE selects a representative vector from the set of bases (S520). When it is assumed that the number of transmitting antennas NT of the transmitter is 4, a i-th UE's estimated channel can be expressed by:
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
ĥi
using the weight matrix W1 as the basis can be written as:
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
ĥi
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
ĥi
most closely.
Transformation of the estimated channel
ĥi
using the weight matrix W2 as the basis can be written as:
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
ĥi
using the weight matrix W3 as the basis can be written as:
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.
Referring to
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.
Referring to
The basis may be a weight matrix of the BS. When the number of transmitting antennas NT is 4, the weight matrix W can be expressed by:
where wi 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′=ejθ
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:
(ejθiwi)Hejθiwi=e−jθiwiHejθiwi=wiHw≠0, (ejθiwi)Hejθjwj=e−jθiwiHejθjwj=e−j(θi−θj)(wiHwj)=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.
Patent | Priority | Assignee | Title |
10361733, | Dec 13 2017 | AT&T Intellectual Property I, L.P. | Low complexity transmitter structure for active antenna systems |
10693508, | Dec 13 2017 | AT&T Intellectual Property I, L.P. | Low complexity transmitter structure for active antenna systems |
11424773, | Dec 13 2017 | AT&T Intellectual Property I, L.P. | Low complexity transmitter structure for active antenna systems |
Patent | Priority | Assignee | Title |
7020110, | Jan 08 2002 | QUALCOMM INCORPORATED, A DELAWARE CORPORATION | Resource allocation for MIMO-OFDM communication systems |
7248638, | Mar 23 2001 | Intel Corporation | Transmit antenna multi-mode tracking |
7321563, | Jan 08 2003 | Samsung Electronics Co., Ltd. | Apparatus and method for estimating a channel condition of a forward link in a mobile communication system |
7327800, | May 24 2002 | Vecima Networks Inc | System and method for data detection in wireless communication systems |
7450532, | Dec 05 2003 | Qualcomm Incorporated | Apparatus and method for transmitting data by selected eigenvector in closed loop MIMO mobile communication system |
20030060236, | |||
20030128658, | |||
20030134605, | |||
20030139139, | |||
20040120411, | |||
20050031044, | |||
20050170782, | |||
20050287978, | |||
20060029157, | |||
20060071807, | |||
20060209980, | |||
20060268623, | |||
20070191043, | |||
20070258366, | |||
20080069031, | |||
20080108310, | |||
20090016463, | |||
20090106619, | |||
20090219838, | |||
EP1294120, | |||
WO2005055465, |
Executed on | Assignor | Assignee | Conveyance | Frame | Reel | Doc |
Apr 04 2007 | LG Electronics Inc. | (assignment on the face of the patent) | / | |||
Jan 21 2008 | ROH, DONG WOOK | LG Electronics Inc | ASSIGNMENT OF ASSIGNORS INTEREST SEE DOCUMENT FOR DETAILS | 028237 | /0156 |
Date | Maintenance Fee Events |
May 22 2013 | ASPN: Payor Number Assigned. |
Nov 25 2014 | M1551: Payment of Maintenance Fee, 4th Year, Large Entity. |
Jan 21 2019 | REM: Maintenance Fee Reminder Mailed. |
Jul 08 2019 | EXP: Patent Expired for Failure to Pay Maintenance Fees. |
Date | Maintenance Schedule |
Mar 19 2016 | 4 years fee payment window open |
Sep 19 2016 | 6 months grace period start (w surcharge) |
Mar 19 2017 | patent expiry (for year 4) |
Mar 19 2019 | 2 years to revive unintentionally abandoned end. (for year 4) |
Mar 19 2020 | 8 years fee payment window open |
Sep 19 2020 | 6 months grace period start (w surcharge) |
Mar 19 2021 | patent expiry (for year 8) |
Mar 19 2023 | 2 years to revive unintentionally abandoned end. (for year 8) |
Mar 19 2024 | 12 years fee payment window open |
Sep 19 2024 | 6 months grace period start (w surcharge) |
Mar 19 2025 | patent expiry (for year 12) |
Mar 19 2027 | 2 years to revive unintentionally abandoned end. (for year 12) |