JP2014140151A - Radio communication system and weight calculation method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To reduce overhead on channel estimation which increases with an increase in the number of antennas, in communication using an OFDM modulation system.SOLUTION: In a radio communication system, a communication station estimates a weight channel matrix which is information considering a number Nof channel estimation weights estimated on the basis of received OFDM packets for each frequency channel, and transmits feedback information calculated on the basis of the weight channel matrix to a transmission source communication station having transmitted the OFDM packets. The communication station having transmitted the OFDM packets restores a channel matrix of each frequency channel fed back from the communication target on the basis of the feedback information, makes a plurality of restored channel matrices of frequency channels into one matrix to generate an aggregation matrix whose size is larger than the original channel matrices, and calculates a weight used for communication for each frequency channel on the basis of the generated aggregation matrix and the channel estimation weight.

Description

  The present invention relates to a radio communication system that communicates using an orthogonal frequency division multiplexing modulation system, and a weight calculation method in the radio communication system.

  In recent years, as the high-speed wireless access system using the 2.4 GHz band or the 5 GHz band, the IEEE802.11g standard, the IEEE802.11a standard, and the like are remarkable. These systems use an Orthogonal Frequency Division Multiplexing (OFDM) modulation scheme, which is a technique for stabilizing characteristics in a multipath fading environment, and realize a physical layer transmission rate of 54 Mbps at the maximum. Yes.

  On the other hand, in the world of wired LANs, the provision of high-speed lines exceeding 100 Mbps has spread due to the widespread use of Ethernet (registered trademark) 100Base-T interface and FTTH (Fiber to the home) using optical fiber in each home. In the world of wireless LAN, further increase in transmission speed is demanded.

  As a technique for this purpose, in IEEE 802.11n, a MIMO (Multiple Input Multiple Output) technique has been introduced as a spatial multiplexing transmission technique. Further, in IEEE 802.11ac, a multi-user MIMO (MU-MIMO) communication method has been studied (for example, see Non-Patent Document 1). MU-MIMO communication has the potential to increase the throughput in the physical layer to several times the number of transmission antennas, but in order to obtain the transmission diversity effect by many transmission antennas, the transmission apparatus requires channel information for each antenna for the terminal. . However, there is a problem that overhead due to the signal for estimating the channel information and feedback information is large.

  FIG. 14 is a diagram illustrating an example of a sequence for acquiring communication channel information using the OFDM modulation scheme. The figure shows an example in which a base station (AP: Access point) acquires channel information for K terminals (STA: Station). K is an integer of 1 or more. In the figure, the horizontal axis represents time, and signals S1, S2, S3-1 to S3-K, S4-1 to S4- transmitted from the base station 91 and the K terminals 92-1 to 92-K, respectively. K is shown.

  The signal S1 is an announcement signal (NDPA: Null Data Packet Announce) for notifying transmission of a channel estimation signal. Signal 2 is a pilot signal (NDP: Null Data Packet) for channel estimation. Signals S3-1 to S3-K are channel information feedback signals (CSIFB: Channel State Information Feed Back). Signals S4-1 to S4-K are polling signals (Polling) instructing transmission of a response signal from a specific communication partner (terminal 92). As shown in the figure, after the base station 91 transmits the NDPA signal S1, the pilot signal S2 is transmitted, and each terminal 92 sequentially transmits feedback signals S3-1 to S3-K to the base station 91. Thereby, the base station 91 acquires feedback information for each antenna for each terminal 92.

  In addition, the figure shows the configuration of pilot signal S2 (NDP). The NDP is configured to perform channel estimation corresponding to the first pilot signal part (Column Symbols) S2-0, the last pilot signal part (Column Symbols), and N transmission antennas included in the base station 91. VHT-LTF (Very High Throughput-Long Training Frame). For example, in order to acquire channel information for 8 (N = 8) transmission antennas, it is necessary to transmit this VHT-LTF from the base station 91 to the terminal 92 for 8 OFDM symbols. Here, the signal Sk of the k-th frequency channel of VHT-LTF is, for example, the expressions (19-11), (19-12), (19-23), (19-24) described in Non-Patent Document 3. Can be determined as follows.

FIG. 15 is a schematic block diagram illustrating configurations of a base station (AP) 91 and a terminal (STA) 92 as wireless communication apparatuses in a wireless communication system using the OFDM modulation scheme. In addition, although the one terminal 92-i is shown in the same figure, the radio | wireless communications system may have several terminal 92-i (1 <= i <= I; I> 2).
The base station 91 includes a transmission signal generation circuit 912, a radio signal transmission / reception circuit 913, N antennas 914-1 to 914-N, a reception signal demodulation circuit 915, a feedback information extraction circuit 916, and a channel information acquisition circuit 917. ing. Terminal 92-i is the transmission signal generating circuit 922-i, provided with a radio signal transmitter 923-i, _{M i} antennas, channel information compression circuit 925-i, and the feedback information generating circuit 926-i.

  When the base station 91 determines a terminal 92-i from which channel information is desired to be acquired, the NDPA and NDP shown in FIG. The radio signal transmission / reception circuit 913 performs conversion of the NDPA and NDP generated by the transmission signal generation circuit 912 into an analog signal, frequency conversion to a carrier frequency, signal amplification, and the like, and performs antennas 914-1 to 914-N. To send through.

Terminal 92-i receives a signal including a NDPA and NDP through the antenna 924-i-1~924-i- M i. The radio signal transmission / reception circuit 923-i estimates channel information of each subcarrier in OFDM based on NDP included in the received signal. The channel information compression circuit 925-i compresses the channel information estimated by the radio signal transmission / reception circuit 923-i into a form suitable for feedback. The feedback information generation circuit 926-i stores the channel information compressed by the channel information compression circuit 925-i as bit information.

When the terminal 92-i acquires the transmission right, the transmission signal generation circuit 922-i generates a transmission signal including feedback information stored in the feedback information generation circuit 926-i. The radio signal transmission / reception circuit 923-i performs conversion of the transmission signal generated by the transmission signal generation circuit 922-i into an analog signal, frequency conversion to a carrier frequency, signal amplification, and the like, and performs antenna 924-i- 1-924-i-M _i is transmitted via at least one.

  When a signal including channel information is transmitted from terminals 92-1 to 92-K by feedback signals S3-1 to S3-K (CSIFB) shown in FIG. 14, base station 91 receives antennas 914-1 to 914-N. The feedback signals S3-1 to S3-K are received via at least one of them. The radio signal transmission / reception circuit 913 outputs a digital signal corresponding to the received feedback signals S3-1 to S3-K to the reception signal demodulation circuit 915. The reception signal demodulating circuit 915 obtains information acquired from any of the terminals 92-1 to 92-K by synchronizing with the digital signal and using channel information. Feedback information extraction circuit 916 extracts a feedback portion of channel information included in CSIFB from information obtained by reception signal demodulation circuit 915. The channel information acquisition circuit 917 acquires and stores channel information in each frequency channel from the feedback part extracted by the feedback information extraction circuit 916.

  As described in Non-Patent Document 2, the feedback of the channel information described above is a part of the channel information of the OFDM frequency channel even if the V matrix is compressed by expressing the angles φ and ψ. May be obtained. When the feedback information is compressed, the channel information acquisition circuit 917 estimates and stores the original channel information by converting or complementing the feedback information.

Here, an example of a feedback method of channel information is shown. In the j-th VHT-LTF (S2-j) in FIG. 14, the known signal used for the k-th frequency channel (subcarrier) is S _k, and the transmission weight used for the VHT-LTF (S2-j) is N ×. 1 vector _{w k,} and _j, it is a matrix _{W k} of N × L indicating a transmission weight used for the VHT-LTF (S2-1) to VHT-LTF (S2-L) representing the following formula (1) it can.

Here, it is assumed that L ≧ _{M i.} Against VHT-LTF received at _{M i-number} of antennas of the terminal 92-i (S2-1~S2-L) , synchronization, removal of guard interval, by performing such as frequency conversion to obtain the reception signal. A received signal matrix Y _k that is a M _i × L matrix indicating the received signal can be expressed by the following equation (2).

In Equation (2), the matrix H _{k, i} is a M _i × N channel matrix representing the propagation coefficient between the transmitting antenna and the receiving antenna in the k-th frequency channel between the base station 91 and the terminal 92-i. It is. The matrix D is an L × L transformation matrix, and the matrix N _{k, i} is a M _i × L noise matrix corresponding to thermal noise. Note that the transformation matrix D can be omitted, or a diagonal matrix having “1” as a diagonal element can be used (in this case, L = N). Also, by making the transformation matrix D unitary, it is possible to prevent the VHT-LTF from receiving a large difference in reception level and to prevent the channel information from being deteriorated due to a quantization error or the like. The conversion matrix D, D H ^{D =} orthogonal matrix satisfying I is desired (I is a unit and 1 diagonal element matrix). The superscript H is a complex conjugate transpose operator. Matrix D ^H where the superscript "H" is attached to the right shoulder of the matrix D is a complex conjugate transposed matrix of the matrix D (Hermitian).
In the terminal 92-i, the channel matrix G _{k, i} including the transmission weight can be estimated using the known transformation matrix D and the known signal S _k as represented by the following equation (3).

In equation (3), the matrix N ′ _{k, i} represents the channel estimation error. Here, a channel matrix G _{k, i including} a transmission weight is defined as a weight channel matrix. If there is no estimation error, the weight channel matrix G _{k, i} = H _{k, i} W _k , and if the diagonal matrix with a diagonal element of 1 as the transmission weight matrix W _k is used, G _{k, i} = H _{k, i} .
The weight channel matrix G _{k, i} obtained in this way can be quantized, converted into bits, and fed back to the base station 91.

Here, a compression method using a compressed beamforming feedback matrix described in Non-Patent Document 2 is shown. A V matrix (basis vector) is calculated by the orthogonalization method or singular value decomposition for the obtained weight channel matrix G _{k, i} . When performing singular value decomposition, the matrix V _{S, k, i} expressed by the following equation (4) is fed back.

In equation (4), the matrix U _{k, i} is a M _i × M _i singular vector, Σ _{k, i} is a M _i × M _i diagonal matrix with singular values as diagonal elements, and 0 is M _i × (L−M _i ) zero matrix, and (V _{S, k, i} V _{N, k, i} ) is a right singular vector. In the right singular vector (V _{S, k, i} V _{N, k, i} ), V _{S, k, i} corresponds to a singular value and V _{N, k, i} corresponds to a zero matrix.
Compressed beamforming feedback matrix is a singular value obtained by converting the imaginary number of the last term of the column vector of V _{S, k, i,} which is a part of the unitary matrix, to 0 and converting it as expressed by the following equation (5). V _{S, k, i} can be expressed by φ and ψ.

  In this way, the amount of bits of information to be fed back is reduced by performing compression using the characteristics of the unitary matrix, and the delay time from the start of channel estimation to the transmission of data is suppressed by suppressing overhead in communication. Can do.

  However, when the number of antennas N included in the base station 91 is large, in order to obtain an effect commensurate with the number of antennas provided, the number L of VHT-LTFs in NDP needs to be close to the number N of antennas. Since the increase in VHT-LTF is accompanied by an increase in overhead and an increase in the bit amount of the feedback signal, it causes a decrease in throughput in the wireless communication system.

IEEE, "Proposed specification framework for TGac," doc.:IEEE 802.11-09 / 0992r21, Jan.2011. IEEE, "IEEE P802.11nTM / D11.0," pp.55-57, June.2009. IEEE, "IEEE P802.11-REVmbTM / D6.0," pp.1597,1606,1635-1636, June.2011.

    As described above, in a base station having a large number of transmission antennas, channel information between each terminal serving as a communication partner and each transmission antenna is required. However, since it is necessary to transmit OFDM symbols corresponding to the number of transmission antennas, feedback information for channel estimation increases as the number of antennas increases. In addition, since the amount of information of the feedback information packet (CSI-FB) from the terminal to the base station increases, there is a problem that communication overhead increases and throughput decreases.

  The present invention has been made in view of such a situation, and in communication using the OFDM modulation scheme, a weight that can improve overhead by reducing overhead related to channel estimation that increases with an increase in the number of antennas. An object is to provide a calculation method and a wireless communication system.

In order to solve the above problems, the present invention is a wireless communication system comprising a plurality of communication stations performing communications using an orthogonal frequency division multiplexing modulation scheme, N _P number of training symbols to be transmitted for each frequency channel a channel estimation weight determining means for determining a channel estimation weight corresponding to each of, multiplied by the channel estimation weight to generate N _P-number of long training symbols for each frequency channel training symbols corresponding to the channel estimation weight, a training symbol sending means for sending to the other of the communication stations of the OFDM packet as the communication partner including the generated N _P-number of long training symbol, on the basis of the OFDM packet received for each frequency channel N of _P or by Include channel estimation weights Weight channel matrix estimation means for estimating a weight channel matrix that is channel information; feedback information transmission means for transmitting feedback information calculated based on the weight channel matrix to a transmission source communication station that has transmitted the OFDM packet; A channel information matrix of each frequency channel fed back from the communication partner based on feedback information is restored, and a channel information matrix of a plurality of frequency channels corresponding to one of the communication partners, or each corresponding to a plurality of the communication partners A set matrix having a size larger than that of the original channel information matrix is generated by using channel information matrices of a plurality of frequency channels as one matrix, and the weight used for communication is set to the frequency channel based on the generated set matrix and the channel estimation weight. C to calculate for each A wireless communication system, characterized by comprising: a site calculating means.

  Further, the present invention is the invention as described above, wherein the feedback information transmitting means includes the weight channel matrix, a right singular vector obtained from the weight channel matrix, or a basis vector obtained from the weight channel matrix. Feedback information is transmitted.

  Further, the present invention is the invention as described above, wherein the feedback information transmitting means includes a vector corresponding to a singular value among right singular vectors of the weight channel matrix and a Hermitian matrix of a left singular value vector of the weight channel matrix. The feedback information including the weight channel unitary matrix obtained by multiplying the frequency channel unitary matrix is transmitted, and the weight calculation means restores the weight channel unitary matrix of each frequency channel fed back from the communication partner based on the feedback information. A plurality of weight channel unitary matrices of a plurality of frequency channels corresponding to one communication partner, or a plurality of weight channel unitary matrices of a plurality of frequency channels corresponding to the plurality of communication partners, respectively. The One of the matrix to produce a large size aggregate matrix than the original weight channel unitary matrix as the generated aggregate matrix on the basis of the channel estimation weight and the weight used for communication, and calculates for each frequency channel.

  In the invention described above, the feedback information transmitting unit calculates a V matrix that is a basis vector having a high correlation with a right singular vector corresponding to a singular value of the weight channel matrix. A weight obtained by multiplying the matrix obtained by multiplying the weight matrix by the weight channel matrix from the right side by using the orthogonalization method and calculating the U matrix which is a unitary matrix and the Hermitian matrix of the U matrix and the V matrix. The feedback information including a channel unitary matrix is transmitted, and the weight calculation unit restores a weight channel unitary matrix of each frequency channel fed back from the communication partner based on the feedback information, and corresponds to one communication partner. Weight channel unitary matrix of multiple frequency channels to be processed, or multiple previous Select a weight channel unitary matrix for each of multiple frequency channels corresponding to the communication partner, and generate a set matrix that is larger in size than the original weight channel unitary matrix using the selected multiple weight channel unitary matrices as one matrix A weight used for communication is calculated for each frequency channel based on the set matrix and the channel estimation weight.

  Further, according to the present invention, in the invention described above, the weight calculation unit restores the weight channel unitary matrix of each frequency channel fed back from the communication partner based on the feedback information, A corresponding weight channel unitary matrix of a plurality of frequency channels or a plurality of weight channel unitary matrices of a plurality of frequency channels corresponding to the plurality of communication partners are selected, and the selected plurality of weight channel unitary matrices are respectively multiplied by weighting factors. A set matrix larger in size than the original weight channel unitary matrix is generated as one matrix above, and a weight used for communication is calculated for each frequency channel based on the generated set matrix and the channel estimation weight. And

  Further, the present invention is the invention described in the above, wherein the feedback information transmitting means generates a set matrix having a size larger than the original weight channel matrix by using the plurality of weight channel matrices as one matrix, and the set matrix, The feedback information including a right singular vector obtained from the set matrix or a basis vector obtained from the set matrix is transmitted, and the weight calculation unit is configured to send each frequency channel fed back from the communication partner based on the feedback information. The channel information matrix is restored, and the weight used for communication is calculated for each frequency channel.

  Further, the present invention is characterized in that, in the above-described invention, the channel estimation weight determining means determines the weight calculated by the weight calculating means as the channel estimation weight.

  In the invention described above, the present invention is characterized in that the feedback information transmitting means calculates the feedback information based on a weight channel matrix corresponding to a part of frequency channels.

Further, in the present invention described above, when the channel estimation weight determination unit obtains feedback from the communication partner and the weight calculation unit calculates L _ALL weights as channel estimation weights, a channel estimation weight, and wherein the _{L ALL} number of weights and, _{n the add} from the training weight matrix _{predetermined, n (however, n add, n = n P} -L ALL) to select the number of weights To do.

Further, in the present invention described above, when the channel estimation weight determination unit obtains feedback from the communication partner and the weight calculation unit calculates L _ALL weights as channel estimation weights, L the _ALL number of weights are divided into two or more groups, the number of weights to be distributed to the subcarriers are a weight that is divided into groups When N _SL, as the channel estimation weight of the sub-carrier, said weight N _SL weights calculated by the calculation means and N _{add, n} (where N _{add, n} = N _P −N _SL ) weights are selected from the training weight matrix.

Further, the present invention is the above-described invention, wherein the channel estimation weight determining means selects all the weights included in the training weight matrix when selecting N _{add, n} weights from the training weight matrix. It is characterized by minimizing the number of overlapping weights in the weights assigned to the channels and assigned to a predetermined number of consecutive frequency channels.

In order to solve the above problem, the present invention provides a weight calculation method in a wireless communication system including a plurality of communication stations that perform communication using an orthogonal frequency division multiplexing modulation method, and transmits each frequency channel. a channel estimation weight determining step of determining a channel estimation weight corresponding to each of N _P number of training symbols, N _P number of long per frequency channel by multiplying the channel estimation weight to training symbols corresponding to the channel estimation weight generates a training symbol, it generated a training symbol transmitting step of transmitting to another of the communication stations of the OFDM packet as the communication partner comprising N _P-number of long training symbol, on the basis of the OFDM packet received, each frequency channel _NP pieces each A weight channel matrix estimation step for estimating a weight channel matrix, which is channel information including estimated channel estimation weights, and feedback information calculated based on the weight channel matrix is transmitted to a transmission source communication station that has transmitted the OFDM packet. Feedback information transmission step, and a channel information matrix of each frequency channel fed back from the communication partner based on the feedback information, and a plurality of frequency channel channel information matrices corresponding to the one communication partner, or a plurality Selecting a channel information matrix of each of a plurality of frequency channels corresponding to the communication partner, generating a set matrix having a size larger than the original channel information matrix by using the selected plurality of channel information matrices as one matrix, and generating the generated set Procession and the above Based on the channel estimation weight, a weight calculation method characterized by having a a weight calculating step of calculating for each frequency channel weights to be used for communication.

  According to this invention, the weight calculation means generates a set matrix having a larger size than the original channel matrix by using a plurality of restored frequency channel channel matrices as one matrix, and the generated set matrix and the channel estimation weight Based on the above, by calculating the weight used for communication for each frequency channel, it is possible to calculate the weight with high accuracy from a small amount of feedback information, compared with the case of simply feeding back channel information. As a result, the number of training symbols transmitted for channel estimation can be reduced and the amount of information for feedback can be reduced, so that overhead in communication can be reduced and throughput can be improved.

1 is a schematic block diagram showing a configuration of a radio communication system 100 according to the present invention. It is a figure which shows an example of the combination of the channel estimation weight used with each subcarrier in the same embodiment. It is a figure which shows an example which allocated the channel estimation weight to each subcarrier according to Formula (7). It is a figure which shows an example of the relationship between a subcarrier and subcarrier group (alpha) _k . It is a graph which shows the effect acquired by applying the 2nd feedback method in the embodiment. It is a graph which shows the result of having analyzed the characteristic at the time of applying the 3rd feedback method in the embodiment by computer simulation. It is a graph which shows the result of having analyzed the characteristic at the time of applying the 2nd feedback method in the embodiment by computer simulation. It is a flowchart of communication in the 1st and 2nd feedback methods in the embodiment. It is a communication flowchart in the 3rd feedback method in the embodiment. It is a figure which shows an example which allocated the channel estimation weight to each carrier according to Formula (57). It is a graph which shows the effect of the radio communications system in the embodiment. In the wireless communication system of the embodiment is a graph showing the setting of the N _S during first channel estimation characteristics of the throughput performance for the setting of the N _S after convergence. It is a graph which shows the throughput characteristic in the model C of delay spread 30 [ns]. It is a figure which shows an example of the sequence which acquires the channel information of communication using an OFDM modulation system. It is a schematic block diagram which shows each structure of the base station (AP) 91 and the terminal (STA) 92 as a radio | wireless communication apparatus in the radio | wireless communications system using an OFDM modulation system.

  Hereinafter, a weight calculation method and a wireless communication system according to an embodiment of the present invention will be described with reference to the drawings. FIG. 1 is a schematic block diagram showing a configuration of a wireless communication system 100 according to the present invention. The wireless communication system 100 includes a base station 1 that performs communication using an OFDM modulation scheme and a plurality of terminals 2-i (1 ≦ i ≦ I; I ≧ 2). Moreover, the figure has shown the structure of the base station 1 and the terminal 2-i. Below, the structure which the base station 1 acquires the channel information of the radio area of the OFDM signal between terminal 2-i is demonstrated. In addition, although the case where the wireless communication system 100 includes one base station 1 will be described, a plurality of base stations 1 may be included.

  The base station 1 includes a channel estimation weight determination circuit 11, a transmission signal generation circuit 12, a radio signal transmission / reception circuit 13, N antennas 14-1 to 14-N, a reception signal demodulation circuit 15, a feedback information extraction circuit 16, and A channel information acquisition circuit 17 is provided.

The channel estimation weight determining circuit 11 stores training weights in advance, and determines channel estimation weights for training symbols used for each OFDM subcarrier. In other words, the channel estimation weight determination circuit 11 determines a channel estimation weight to be used as a training signal to be transmitted in each frequency channel from the training weight stored in advance. Alternatively, when the channel estimation weight determination circuit 11 stores a previously estimated channel estimation weight, a new channel estimation is performed from the previously stored training weight and the stored channel estimation weight. Determine the weight.
The transmission signal generation circuit 12 generates an OFDM packet using the channel estimation weight determined by the channel estimation weight determination circuit 11. For example, the transmission signal generation circuit 12 calculates a long training symbol for each subcarrier by multiplying the determined channel estimation weight by a training symbol corresponding to the channel estimation weight. The transmission signal generation circuit 12 generates an OFDM packet including the calculated long training symbol.

The radio signal transmission / reception circuit 13 performs transmission processing including provision of a guard interval, conversion to an analog signal, conversion to a carrier frequency, signal amplification, etc., on the OFDM packet generated by the transmission signal generation circuit 12, and performs antenna processing. It transmits to each terminal 2 via 14-1 to 14-N.
The radio signal transmission / reception circuit 13 receives a signal including feedback information transmitted from each terminal 2, and amplifies the signal, converts the frequency to a baseband frequency, converts to a digital signal, Reception processing including removal of guard intervals is performed. The radio signal transmission / reception circuit 13 outputs the digital signal obtained by the reception process to the reception signal demodulation circuit 15.

The reception signal demodulation circuit 15 synchronizes with the digital signal output from the radio signal transmission / reception circuit 13 and demodulates the information transmitted by each terminal 2.
The feedback information extraction circuit 16 extracts a feedback portion of channel information included in the CSIFB from the information demodulated by the reception signal demodulation circuit 15. The feedback information extraction circuit 16 outputs the extracted feedback portion to the channel estimation weight determination circuit 11 and the transmission signal generation circuit 12.

The terminal 2-i includes a transmission signal generation circuit 22-i, a radio signal transmission / reception circuit 23-i, M _i antennas 24-i-1 to 24-i-M _i , a channel information compression circuit 25-i, and A feedback information generation circuit 26-i is provided.
Radio signal transceiver circuit 23-i, based on a signal received via the antenna 24-i-1~24-i- M i, estimates channel information of each subcarrier. The channel information compression circuit 25-i compresses the channel information estimated by the radio signal transmission / reception circuit 23-i into a form suitable for feedback to the base station 1. The feedback information generation circuit 26-i generates feedback information from the channel information compressed by the channel information compression circuit 25-i, and stores the generated feedback information.

When the terminal 2-i acquires the transmission right, the transmission signal generation circuit 22-i generates a transmission signal including feedback information stored in the feedback information generation circuit 26-i. The radio signal transmission / reception circuit 23-i includes a transmission process including a guard interval, conversion to an analog signal, conversion to a carrier frequency, signal amplification, and the like for the transmission signal generated by the transmission signal generation circuit 22-i. It was carried out, and transmits via the at least one antenna 24-i-1~24-i- M i.

Hereinafter, the processing performed by each circuit will be specifically described.
Channel estimation weight decision circuit 11 stores in advance the N _B number of training weights used for transmitting a training signal. Channel estimation weight determining circuit 11, the base station 1 when transmitting a training signal, determining whether the channel estimation weight to be used in OFDM for each subcarrier any training weights among the N _B number of training weights. At this time, the channel estimation weight determination circuit 11 may select all the channel estimation weights from the training weights, and some of the channel estimation weights are calculated based on the channel information fed back from the communication partner. Also good. Here, _{N B} number of training weights _{t 1, t 2, ...,} training weight matrix _{T B} showing the _{t NB} is a matrix of N × _{N B,} expressed as the following formula (6).

In the case of using the N _P-number of long training symbols will select the N _P number of training weights (channel estimation weights) from among the N _B column vectors included in this training weight matrix T _B. For example, for each subcarrier as shown in FIG. 2, it can be assigned a beam corresponding to N _P number of channel estimation weight. FIG. 2 is a diagram illustrating an example of combinations of channel estimation weights used in each subcarrier in the present embodiment. In the figure, the beam allocation is shown N _P number for each subcarrier.

B (i, k) in the figure represents the (number indicating what number of column vectors in matrix T _B) k-th subcarrier of the i-th beam and comprising channel estimation weight number. b (i, k) can be expressed by, for example, the following formula (7).

In Expression (7), the function “mod (A, B)” is a function for calculating a remainder when A is divided by B. Here, N _S is the number corresponding to either shift number the number of training weights corresponding to each long training, for example, be a N S ₌ integer _(N B _/ N _P). Here, the function “integer (A)” is a function that converts a positive number A to an integer by rounding, rounding up, or rounding down. By using the N _S to be N B _/ N _P, by using the received information in the Ns subcarriers on the receiving side, the channel information is obtained corresponding to approximately N _B number of total transmit beam. Here, the number of channel estimation weights ₁₆ (N B = 16), the number of long training symbols ₄ (N P = 4), if the _{N S} is 4, the number of subcarriers is 52, the formula (7) An example in which channel estimation weights are assigned according to FIG. 3 is shown. FIG. 3 is a diagram illustrating an example in which channel estimation weights are assigned to each subcarrier according to Equation (7). As shown in the figure, assuming that N _S = N _B / N _P in equation (7), when channel information is estimated for N _S consecutive subcarriers on the receiving side, N _B transmission beams Channel information corresponding to all can be acquired.

  When the channel estimation weight determination circuit 11 determines a combination of channel estimation weights for each subcarrier as shown in FIG. 2, for example, the transmission signal generation circuit 12 generates an OFDM packet using the combination. The radio signal transmission / reception circuit 13 performs transmission processing on the generated OFDM packet and transmits it via the antennas 14-1 to 14-N.

Also in this embodiment, the channel information acquisition sequence in the base station 1 and each terminal 2-i can be expressed in the same manner as in FIG. However, it is L = _{N P.} The channel estimation weight matrix W _k (N × N _P matrix) used for VHT-LTF (S2-1 to S2-N _P ) can be expressed as the following equation (8).

In the terminal 2-i, an antenna 24-i-1~24-i- M i via wireless signal transmitting and receiving circuit 23-i receives a signal transmitted from the base station 1. The M _i × N _P received signal matrix Y _{k, i} indicating the received signal of the k-th subcarrier obtained by subjecting the signal received by the radio signal transmitting / receiving circuit 23-i to reception processing is given by the following equation (9). Can be written.

In equation (9), the matrix H _{k, i} is a M _i × N channel matrix representing the propagation coefficient from the transmitting antenna to the receiving antenna in the k-th frequency channel between the base station 1 and the terminal 2-i. . Matrix D is a transform matrix of _{N P × N P, N k} , i corresponds to the thermal noise is the noise matrix _{M i} × _{N P.} S _k is a known signal (training signal) transmitted by the base station 1.

Note that the transformation matrix D can be omitted, or a diagonal matrix having 1 as a diagonal element can be used for the transformation matrix D. By using a unitary matrix as the transformation matrix D, it is possible to prevent the reception level of each VHT-LTF from being greatly different and to prevent the channel information from being deteriorated due to a quantization error or the like. The transformation matrix D is preferably an orthogonal matrix that satisfies D ^H D = γI (the matrix I is a unit matrix having a diagonal element of 1 and γ is an arbitrary coefficient).

Here, the formula (9) is different from equation (2), the matrix _{Y k, i} is a matrix of _{M i} × _{N P.} In the terminal 2-i, since the transform matrix D and the known signal _{S k} is known, to estimate the channel matrix _{G k} inclusive transmission weight to the following equation (10), i and _{(M i} × _{N P} matrix) Can do.

  Here, three feedback methods for compressing and feeding back channel information in the terminal 2-i are shown. First, processing in the terminal 2-i will be described.

<First feedback method>
In the first feedback scheme, estimated M _i × N _P weights channel matrix G _{k, i} or part of column vectors or row vector channel information compression circuit 25-i is feedback information generating circuit of the matrix 26- output to i. The feedback information generation circuit 26-i quantizes the information output from the channel information compression circuit 25-i, converts it into feedback information, and stores it. Transmission signal generating circuit 22-i generates a transmission signal including the feedback information, the radio signal transmitting and receiving circuit 23-i transmits a transmission signal via the antenna 24-i-1~24-i- M i.

A right singular vector obtained by performing singular value decomposition on the weight channel matrix G _{k, i} or a weight channel matrix G _{k, i is used as} a partial column vector or row vector of the weight channel matrix G _{k, i.} The channel information compression circuit 25-i may output the basis vector obtained from the above to the feedback information generation circuit 26-i.
Further, the weight channel matrix G _{k, i} may be converted into time-series channel information by Fourier transform and quantized. Further, the dimension of the weight channel matrix to be fed back can be reduced from M _i to L _i assuming a reception weight of M _i × L _i matrix. In this case, the channel matrix to be fed back can be expressed as G ′ _{k, i} = W _R ^H G _{k, i} . W The method of determining the _R described in the second feedback method, it can also be simply or feedback to select any antenna, only the information corresponding to the large L _i receive antennas of the received power.

<Second feedback method>
In the second feedback method, the singular value decomposition expressed by the following equation (11) is performed on the weight channel matrix G _{k, i} .

In Equation (11), the matrix U _{k, i} is a left singular vector of M _i × M _i , Σ _{k, i} is a diagonal matrix of M _i × M _i with singular values as diagonal elements, and 0 Is a zero matrix of M _i × (N _P −M _i ), and (V _{S, k, i} V _{N, k, i} ) is a right singular vector, of which V _{S, k, i} is a singular value. The corresponding N _P × M _i matrix, V _{S, k, i} is the N _P × (N _P −M _i ) matrix corresponding to the zero matrix. Here, instead of feeding back V _{S, k, i} , terminal 2-i is based on V _{S, k, i} U _{k, i} ^H (or U _{k, i} V _{S, k, i} ^H ). Feedback to station 1. This is preparation for enabling the base station 1 to expand the transmission dimension.

Here, a weight channel unitary matrix A _{k, i} is defined as represented by the following equation (12).

The weight channel unitary matrix A _{k, i} is an N _P × M _i matrix. Since the matrix V _{S, k, i} is part of the unitary matrix and the matrix U _{k, i} is a full rank unitary matrix, the weight channel unitary matrix A _{k, i} is also part of the unitary matrix, and each column vector Are orthogonal. The weight channel unitary matrix A _{k, i} can be calculated by using the following equation (13) without using the equation (11) with a large calculation load. Further, by repeating QR decomposition, a matrix having a high correlation with Expression (12) and Expression (13) can be obtained.

For example, QR decomposition is performed on the weight channel matrix G _{k, i} as shown in the following equation (14) to calculate the matrix Q _{S, k, i} .

The matrix Q _{S , u, k} is obtained by further performing QR decomposition on the matrix obtained by multiplying the weight channel matrix G _{k, i} by the matrix Q _{S, k, i} as shown in the following equation (15). _{, I} is calculated.

The weight channel unitary matrix A _{k, i} can also be calculated using the matrix Q _{S, u, k, i} as shown in the following equation (16).

In this calculation method, QR decomposition can be repeated as follows. For example, the weight channel matrix G _{k, i} is further multiplied by the matrix Q _{S, u, k, i} obtained by the equation (15), and QR decomposition is performed as in the following equation (17).

Furthermore, using the obtained matrix Q _{S, k, i} , iterative QR decomposition is performed so as to calculate the matrix Q _{S, u, k, i} of Expression (15), thereby obtaining Expression (16) It can be asymptotic to the results of (12) and (13).

In the present embodiment, in the compressed beamforming matrix used in the wireless LAN, it is necessary to maintain this phase where the imaginary number of the last term of the column vector of the unitary matrix is 0. For this reason, as many pieces of phase information as the number of column vectors to be fed back are required. That is, the phase rotation matrix E _{k, i in} which the imaginary number of the last term of the column vector of the weight channel unitary matrix A _{k, i} is 0 is defined to be expressed by the following equation (18).

The N _P × M _i matrix A ′ _{k, i} obtained in this way can be converted so that it can be expressed by φ and ψ as in the equation (5), and the information can be compressed. In the second feedback method, it is necessary to feed back ζ in addition to φ and ψ. By feeding back φ, ψ, and ζ, the base station 1 can restore the weight channel unitary matrix A _{k, i} .

In the second feedback method described above, the number of column vectors to be fed back needs to be the same as the number of receiving antennas M _i, and thus a method for feeding back arbitrary L _i column vectors is shown below. .
First, define a reception weight matrix _{W R.} Receiving weight matrix _{W R} is a matrix of _{M i} × _{L i.} A value equal to or _smaller than the number of antennas M _i included in the terminal 2-i is set as the number of column vectors L _i . In the reception weight matrix W _R, column vectors are orthogonal to each other. The reception weight matrix W _R may also be expressed as a selection of vectors of rows of L _i of the unitary matrix M _i × M _i.

Reception weight matrix W _R, since commonly used by all subcarriers can determine more effective transmission weights when selected to satisfy the following equation (19) and equation (20).

In Expressions (19) and (20), || · || represents the Frobenius norm, and w _i is an arbitrary column vector of the reception / transmission weight matrix W.
For example, among the eigenvectors obtained for equation (21), it may be to obtain a reception weight matrix W _R by selecting L _i number of vectors corresponding to a higher eigenvalues, simply equation (21) it may be orthogonalized method matrix obtained from the orthogonal vector obtained by using the obtained reception weight matrix W _R in.

Thus the matrix G _'k of L _i × N _P obtained as the following equation (22) after setting the entire reception weight matrix W _R and _{the i} aforementioned weight channel matrix G _k, it is used in place of the _i in, it is possible to obtain _{N P} × _{L i} wait channel unitary matrix _{a k, i} of.

Specifically, singular value decomposition is performed on the matrix G ′ _{k, i} as shown in the following equation (23).

In Equation (23), the matrix U ′ _{k, i} is a left singular vector of L _i × L _i , and Σ ′ _{k, i} is a diagonal matrix of L _i × L _i with singular values as diagonal elements. , 0 is a zero matrix of L _i × (N _P −L _i ). (V ′ _{S, k, i} V ′ _{N, k, i} ) is a right singular vector, of which V ′ _{S, k, i} is an N _P × L _i matrix corresponding to the singular value, and V ′ _{N, k, i} are N _P × (N _P −L _i ) matrices corresponding to the zero matrix.
Therefore, an N _P × L _i matrix A _{k, i} can be obtained by the following equation (24) _, and a smaller number of column vectors than the number of receiving antennas can be fed back to the base station 1.

Also, the matrix A _{k, i} can be obtained by using the following equation (25).

As described above, in the second feedback method, the channel information compression circuit 25-i has the singular value of the right singular vectors (V _{S, k, i} V _{N, k, i} ) of the weight channel matrix G _{k, i.} corresponding vector V _{S in, k, i} and, weight channel matrix G _k, the left singular vectors U _{k, i} Hermitian matrix and the multiplication to wait channel unitary matrix a _k obtained in _{i, i} feedback information generation The signal is output to the circuit 26-i and fed back to the base station 1. Or channel information compression circuit 25-i is the weight channel matrix _{G k,} the right singular vector corresponding to the singular values of _{i (V S, k, i} V N, k, i) base vector having a high correlation with _{Q S , K, i} , and a matrix Q _S, which is a unitary matrix using an orthogonalization method to a matrix obtained by multiplying the weight channel matrix G _{k, i} from the right by the calculated basis vector Q _{S, k, i . u, k, i} is calculated, and a weight channel unitary matrix A _{k, i} obtained by multiplying the basis vector Q _{S, k, i} by the matrix (Q _{S, u, k, i} ) ^H is used as a feedback information generation circuit. 26-i and feed back to the base station 1.

<Third feedback method>
Unlike the first and second feedback methods, the third feedback method performs transmission dimension expansion in the terminal 2-i. For this reason, the amount of feedback information is larger than that in the second feedback method. In the third feedback method, the terminal 2-i defines and stores a subcarrier group for each subcarrier. Here, the subcarrier group for the _kth subcarrier is α _k . α _k is a set of numbers indicating subcarrier numbers. The number of subcarriers included in α _k may be the same for all subcarriers, or may be different for each subcarrier. For example, by reducing the number of subcarriers belonging to α _k in the subcarrier having the highest or lowest frequency, it is possible to prevent using reception information of subcarriers that are separated in frequency from the subcarrier.

as the number of subcarriers included in alpha _k, using the reception signals of the N _S subcarriers with information for the aforementioned Using N _S, respectively N _P number of transmission beams, N B ₍₌ N _S × Channel information for N _P ) transmit beams can be obtained. Here, the case is the N _S subcarriers near the alpha _k to the k-th subcarrier.

FIG. 4 is a diagram illustrating an example of a relationship between a subcarrier and a subcarrier group α _k . The figure shows subcarrier groups α _k = {α _k (1), α _k (2), α _k (3), _where F = 52, N _S = 4, N _P = 4, and N _B = 16. An example of α _k (4)} is shown. Nearby the N _S subcarriers including itself on each subcarrier is defined in the sub-carrier group. Using the subcarrier group information obtained in this way, an aggregate channel matrix GA _{, k, i} for _{the kth} subcarrier is defined as in the following equation (26).

Set channel matrix G _{A, k, i} is by using the received signals of a plurality of subcarriers, and has a matrix of M _i × N _B. While in the first and second feedback method the number of column vectors of the channel matrix was N _P, the third feedback method the number of column vectors of the channel matrix becomes a N _B. In the third feedback method, feedback information is obtained by performing signal processing on the aggregate channel matrix GA _{, k, i} . When singular value decomposition is performed on the aggregate channel matrix GA _{, k, i} , it can be expressed as the following equation (27).

In Equation (27), the matrix U _{A, k, i} is the left singular vector of M _i × M _i , and the matrix Σ _{A, k, i} is the diagonal of M _i × M _i with singular values as diagonal elements. 0 is a zero matrix of M _i (N _B −M _i ). (V _{S, A, k, i} V _{N, A, k, i} ) is a right singular vector, of which the matrix V _{S, A, k, i} is an N _B × L _i matrix corresponding to a singular value. Yes, the matrices V _{N, A, k, i} are N _B × (N _B −M _i ) matrices corresponding to the zero matrix.

In the third feedback method, L _i column vectors are selected from the matrices V _{S, A, k, i} , and the selected L _i column vectors are fed back to the base station 1. As a method for compressing information, the compressed beamforming feedback matrix shown in Expression (5) can be used. In addition, the imaginary number of the last term of the column vector of the matrix V _{S, A, k, i} is set to 0 and converted as in the following equation (28), so that the matrix V _{S, A, k, i} becomes φ and It can be expressed by ψ.

As described above, in the third feedback method, the channel information compression circuit 25-i uses a plurality of weight channel matrices G _{k, i} as one matrix and has a larger aggregate channel than the original weight channel matrix G _{k, i. A} matrix G _{A, k, i} is generated and the right singular vector (V _{S, A, k, i} V _{N, A, k, i} ) or aggregate channel matrix G _A obtained from the aggregate channel matrix G _{A, k, i , K, i} , L _i basis vectors selected from the matrixes V _{S, A, k,} i are output to the feedback information generating circuit 26-i and fed back to the base station 1. Here, Equation (26) shows an example of using the N _S subcarrier information, the number of subcarriers to be used is also less than N _S, no problem even if more than N _S. Further, in the equation (26), setting the _{N S N S0 (<N S} ) and, although the to estimate the small number of channel information than the number of transmitting antennas, but still uses only one subcarrier information It is possible to estimate channel information for more transmission dimensions. Further, by setting the N _S and N _{S0 (>} N _S), thereby to estimate the channel information to the transmission beam number greater than the number of transmit antennas. In this case, although the efficiency is poor, the channel information for the same transmission beam can be estimated with different subcarriers, so that the estimation accuracy can be improved. The size of the matrix V _{S, A, k, i} obtained by Expression (27) is (N _P × N _S0 ) × L _i .

Using any one of the first to third feedback methods described above, the channel information compression circuit 25-i uses the compressed channel to be fed back from the channel information acquired by the radio signal transmission / reception circuit 23-i. Generate information. The feedback information generation circuit 26-i converts the channel information generated by the channel information compression circuit 25-i into bit information. Transmission signal generating circuit 22-i generates feedback information including the bit information, radio signal transceiver circuit 23-i transmits the feedback information through the antenna 24-i-1~24-i- M i.

  In the base station 1, when the radio signal transmission / reception circuit 13 receives a signal including feedback information via the antennas 14-1 to 14-N, the received signal demodulation circuit 15 demodulates the received information. When the feedback information extraction circuit 16 detects that feedback information is included in the demodulated signal, the feedback information extraction circuit 16 extracts the feedback information. The channel information acquisition circuit 17 acquires channel information from the extracted feedback information, and calculates a transmission weight based on the acquired channel information. The transmission weight calculated by the channel information acquisition circuit 17 may be used as a data transmission weight used when transmitting data to the terminal 2-i or may be used as a training weight. Moreover, you may use as a weight at the time of receiving the signal which the terminal 2-i transmitted.

Hereinafter, handling of channel information by the channel information acquisition circuit 17 will be described for each of the first to third feedback methods.
<First feedback method>
The channel information acquisition circuit 17 acquires the weight channel matrix G _{k, i} of each fed back subcarrier. In the first feedback method, two types of weights can be calculated using the obtained weight channel matrix G _{k, i} . One is a first transmission weight calculation method with a high estimation accuracy but a narrow signal space region to be estimated, and the other is a second transmission weight calculation method with a slightly low estimation accuracy but a wide signal space region to be estimated. Since the first transmission weight calculation method has high accuracy, it is useful as a method for calculating the transmission weight of multiuser MIMO. The second transmission weight calculation method is suitable for a channel estimation weight or a transmission weight for a single user because a signal space area to be estimated is wide.

The first transmission weight calculation method uses the feedback weight channel matrix G _{k, i} . The matrix V _{S, k, i} obtained by singular value decomposition of the weight channel matrix G _{k, i} is used as an eigenvector transmission weight, or multi-user MIMO transmission is performed from the weight channel matrix G _{k, i} for a plurality of terminals 2. The weight can be calculated.
In the first transmission weight calculation method, the transmission weight for the i-th terminal 2-i in the k-th subcarrier is set as a matrix WP _{, k, i} . An antenna transmission weight W _{T, k, i} to be multiplied by a signal transmitted from each transmission antenna and transmitted to the i-th terminal 2-i in the k-th subcarrier is expressed by the following equation (29). Can be obtained.

Matrix _{N × N P (t b (} 1, k), ..., t b (NP, k)) by multiplying the matrix _{W P, k,} to _i, the antenna transmission weight a matrix of N × _{M i} W _{T, k, i} can be obtained.

The second transmission weight calculation method expands the signal space by using channel information of a plurality of subcarriers fed back. That is, the signal processing performed at the terminal 2 in the third feedback method is performed at the base station 1. A description will be given using the subcarrier group α _k in the example shown in FIG. The aggregate channel matrix GA _{, k, i} for the kth subcarrier is obtained as in the following equation (30).

In Equation (30), N _{S, k} is the number of subcarriers belonging to the subcarrier group α _k of the _kth subcarrier, and can be set to N _{S, k} = N _S or set to a different number. You can also Here, the aggregate channel matrix G _{A, k, i} is a matrix of M _i × (N _{S, k} × N _P ) by using the received signals of a plurality of subcarriers. While weight channel matrix G _{k, i} number column vector of was N _P, set the channel matrix G _{A, k,} the number of column vectors of _i is extended N _{S, k} times.

Signal processing is performed on the aggregate channel matrix G _{A, k, i} to obtain a transmission weight. For example, the set channel matrix _{G A, k,} corresponding to the singular values obtained performs singular value decomposition on _{i (N S, k × N} P) × M i matrix _{V S of, A, k,} a _i using Training weight and transmission weight.
When determining the transmission weight of multi-user MIMO, the transmission weight for each terminal 2 can be determined using the matrix V _{S, A, k, i} for the plurality of terminals 2. Since the matrix V _{S, A, k, i} is obtained from the channel matrix for the transmission beam using the channel estimation weight, it is composed of column vectors of (N _{S, k} × N _P ) × 1. Therefore, it is necessary to obtain a transmission weight for the antenna as in Expression (29).

When the transmission weight for the i-th terminal 2-i obtained in the second transmission weight calculation method is expressed as a matrix W _{P, k, i} of (N _{S, k} × N _P ) × L _i , the k-th sub An antenna transmission weight W _{T, k, i that} is an N × L _i matrix for the i-th terminal in the carrier can be obtained as the following equation (31).

The antenna transmission weights WT _{, k, i} can be used as channel estimation weights or can be used as transmission weights for data transmission. Here, the matrix (tb _{(1, αk (1))} ,... _{, Tb (NP, αk (NS, k))} ) is an N × ( _{NS , k} × N _P ) matrix. In this case, the transmission weight corresponding to the transmission antenna can be obtained by converting the matrix WP _{, k, i} using the channel estimation weight of the entire subcarrier group. In the first feedback scheme, 'If feedback of _{k, i,} wait a channel matrix in equation (30) G _k, the matrix G instead of _i' matrix G, such as using the assumptions of the reception weights _{k, i} Is used. In this case, the obtained aggregate channel matrix G _{A, k, i has} a matrix size L _i × (N _{S, k} × N _P ), but the obtained (N _{S, k} × N _P ) × L _i matrix Using V _{S, A, k, i} , channel estimation weights and transmission weights can be obtained in the same manner as described above. In addition, whether the weight channel matrix G _{k, i} or the matrix G ′ _{k, i} is used, the right singular matrix is not necessarily used as the matrix V _{S, A, k, i} , and the collective channel matrix is not necessarily used. _A matrix calculated from GA _{, k, i} or GA _{, k, i} can be used. For example, a specified vector obtained by orthogonalization or QR decomposition can be used for the aggregate channel matrix GA _{, k, i} .

<Second feedback method>
The channel information acquisition circuit 17 obtains N _P × L _i weight channel unitary matrices A _{k, i} in each fed back subcarrier. As described above _{, the} weight channel unitary matrix A _{k, i} is _obtained by multiplying the right singular matrix for the signal space by the left singular matrix that is a unitary matrix. Therefore, the weight channel unitary matrix A _{k, i} can be treated as a right singular matrix.

In the first transmission weight calculation method with high estimation accuracy, the weight channel unitary matrix A _{k, i} may be used as the transmission weight as it is, or may be used as the right singular matrix for the calculation of the transmission weight of multiuser MIMO. However, the obtained transmission weight is the transmission weight for the transmission beam multiplied by the channel estimation weight. Therefore, the transmission weight for the transmission antenna needs to be converted by the equation (29) with respect to the transmission weight WP _{, k, i} calculated from the weight channel unitary matrix A _{k, i} .

Similar to the first feedback method, the second transmission weight calculation method extends the signal space by using channel information of a plurality of subcarriers fed back. A description will be given using the subcarrier group α _k in the example shown in FIG. The set matrix A _{A, k, i} for the kth subcarrier is obtained as in the following equation (32).

Matrix _{A k} of the _{N P} × _{L i} for different channel estimation _{weights, i} a _{N S, by the k} arranging the set matrix _{A A, k, i} is _{(N P × N S, k} ) in matrix × _{L i} Has been extended. The superscript T is the transpose matrix operator. The set matrix A _{A, k, i} obtained in this way is a right singular matrix of a channel matrix that can be defined between a transmission beam generated by (N _P × N _{S, k} ) channel estimation weights and a receiving antenna. And has a high correlation.

Therefore, a right singular matrix corresponding to the signal space is further calculated from the set matrix obtained by Expression (32) by singular value decomposition expressed by the following Expression (33), or an orthogonal matrix is extracted by QR decomposition. Thus, a vector having a high correlation with the right singular vector corresponding to the signal space of the channel matrix that can be defined between the transmission beam generated by (N _P × N _{S, k} ) channel estimation weights and the receiving antenna is obtained. Can be obtained.

When expressed using singular value decomposition, the use is obtained as the following equation _{(33) (N P × N} S, k) × L i matrix _{V S of, A, k,} the column vector _i to the calculation of the transmission weight Can do.

In the QR decomposition, the column vector of the matrix Q _{S, A, k, i} of (N _P × N _{S, k} ) × L _i obtained as the following equation (34) is used for transmission weight calculation. .

Matrix _{V S} in equation _(33), which is _{A, k,} the matrix _{Q S} at _{the i} and equation _{(34), A, k,} another representation of exactly the same signal space to _i. The part expressed using singular value decomposition in the description so far can be replaced with the operation of the orthogonalization method. The matrix V _{S, A, k, i} obtained by the equation (33) or the matrix Q _{S, A, k, i} obtained by the equation (34) is used as a transmission weight as it is, or multi-user MIMO or the like is performed by linear calculation. The transmission weight can be calculated.

The transmission weight obtained here is (N N × N _{S, k} ) × 1 because the matrix V _{S, A, k, i} or the matrix Q _{S, A, k, i} is composed of (N _P × N _{S, k} ) × 1 column vectors. A transmission weight vector of ( _P × N _{S, k} ) × 1 is obtained. Since this is a transmission weight for the transmission beam generated by the (N _P × N _{S, k} ) channel estimation weights described above, the channel estimation weight is used to convert the transmission weight to the transmission weight for the transmission antenna. Need to be converted to the next dimension.

The determined number of spatial streams of transmission weights for the i-th terminal 2-i is L _i, and the transmission weight for the transmission beam of the k-th subcarrier for the i-th terminal 2-i is (N _P × N _{S, k} ) × L _i matrix W _{P, k, i} , the transmission weight W _{T, k, i that} is the N × L _i matrix for the transmission antenna can be obtained as the following equation (35).

In the equation (35), (t _{b (1, αk (1))} ,..., T _{b (NP, αk (Ns, k))} ) is an N × (N _{S, k} × N _P ) matrix. In this way, it is possible to calculate transmission weights for transmission antennas using channel information for many transmission beams.
Furthermore, in order to increase the accuracy of the signal space vector of the channel between the transmission beam and the receiving antenna generated by (N _P × N _{S, k} ) channel estimation weights, a weighting factor can be further introduced. In the calculation of the set matrix A _{A, k, i} in Expression (32), it can be defined as the following Expression (36).

In Equation (36), _{pk, j, i} are weighting factors used for the _jth frequency channel matrix _{Aj, i} in the kth frequency channel for the terminal 2-i. As a weighting coefficient, it can determine as following Formula (37), for example.

In the formula (37), the function _f LOSS _{(D S)} is a function for the subcarrier spacing _{D S,} it can be set to be smaller as the subcarrier spacing _{D S} increases. For example, it can be defined as the following equation (38).

In the equation (38), ν can be a number smaller than 1 and larger than 0, for example, ν = 0.9. Alternatively, the function f _LOSS can be a constant value. In this case, regardless of which subcarrier is used, the weight can be determined as follows regardless of k.

In equation (39), || H _{j, i} || may be fed back from terminal 2-i, or || G _{j, i} || or || G instead of || H _{j, i} || ' _{j, i} || and a value correlated with the power of the channel can be fed back and used.
Further, in the time domain multiple access (TDMA), the weighting factor p can be determined without feedback of the channel power. This is because the power information of the channel information estimated in the uplink can be used. Even if there is no calibration mechanism, the information about the power of the channel can be expected to be highly accurate. The power of the j-th frequency channel (or the square root of the power value) in the access channel from the terminal 2-i to the base station 1 can be used as the weighting coefficient p _{k, j, i} .

<Third feedback method>
In the third feedback method, channel information with an expanded transmission dimension is fed back to the base station 1, so the first transmission weight calculation method cannot be used, and the second transmission weight calculation method is used. become.
The channel information acquisition circuit 17 obtains a set matrix A _{A, k, i} of (N _{S, k} × N _P ) × L _i for each fed back subcarrier. As described above _{, the} set matrix A _{A, k, i} is a base vector for the channel estimation weight of (N _{S, k} × N _P ), and this is used as a transmission weight as it is or a transmission weight is calculated by calculation. Can do.

The determined number of spatial streams of transmission weights for the i-th terminal 2-i is L _i, and the transmission weight for the transmission beam of the k-th subcarrier for the i-th terminal 2-i is (N _P × N _{S, k} ) × L _i matrix W _{P, k, i} , the transmission weight W _{T, k, i that} is the N × L _i matrix for the transmission antenna can be obtained as the following equation (40).

Here, the matrix (t _{b (1, αk (1))} ,..., T _{b (NP, αk (Ns, k))} ) is an N × (NS _{, k} × N _P ) matrix. In this way, transmission weights using channel information for many transmission beams can be calculated.

  In this way, using the first or second feedback method, the channel information acquisition circuit 17 acquires the transmission weight by using one or both of the first transmission weight calculation method and the second transmission weight calculation method. To do. Alternatively, using the third feedback method, the channel information acquisition circuit 17 acquires the transmission weight using the second transmission weight method.

  When the acquired transmission weight is used as the data transmission weight, the transmission weight is output to the transmission signal generation circuit 12 and used as a transmission weight for the data packet. When the acquired transmission weight is used as a training weight set, the transmission weight is output to the channel estimation weight determination circuit 11 and can be used as a training weight or stored separately as a channel estimation weight based on channel information. You can also keep it.

Incidentally, it may be updated some of acquired training weights T _B at the transmission weights may be separately stored training weights obtained from the channel information as the second training weights T _'B. When the second training weight T ′ _B is used, there are a method of storing the second training weight T ′ _B for each terminal 2 and a method of storing for each group of the terminals 2.

For example, the column vectors of the matrix V _{S, A, k, i} , the matrix Q _{S, A, k, i} and the set matrix A _{A, k, i} are set to the k-th subcarrier of the i-th terminal 2-i. it can be stored as a second training weights T _'B. Further, when storing for each group of terminals 2, the group of terminals 2 is defined as Ω, and the aggregate channel matrix G _{A, k, i} , matrix V _{S, A, k, i} , A set matrix can be defined using the matrix A _{A, k, i} (∀i∈Ω), and a vector corresponding to the signal space can be stored. When the number of terminals 2 belonging to the group Ω is N _STA and each terminal 2 is represented by Ω (1) to Ω (N _STA ), a vector corresponding to the signal space can be expressed as follows. The vector corresponding to the signal space can be expressed using either QR decomposition or singular value decomposition. When expressed using singular value decomposition, the following equation (41) is obtained.

The second training weight can be selected from the column vector of the matrix V _{S, Ω, k} in Equation (37). Alternatively, the second training weight may be selected from the matrix V _{S, Ω, k} expressed as the following equation (42) or the following equation (43) by singular value decomposition.

When the weights calculated as described above with the first training weights, N _B number of basis vectors containing one or more vectors of N × 1 calculated can be recalculated.
Alternatively, in Equation (41), Equation (42), and Equation (43), an orthogonal vector obtained by performing QR decomposition on the Hermitian transpose matrix on the left side, Equation (41), Equation (42), Equation (43) such as basis vectors obtained by using the orthogonal method column vector on the left of the Hermitian transposed matrix, the matrix V _{S, Omega, k} and high a possible matrix to have a correlation matrix V _{S, Omega,} be used as _k Can do.
Or Formula (41), Formula (42), and Formula (43) can be expressed by mathematical formulas as follows. Matrix _{G k} of _{M i} × _{N P} fed back from the terminal, i, _{L i} × _{N P} matrix _{G 'k, i, N P} × L i wait channel unitary matrix _{A k of,} from one of _i, the matrix V _{S, Ω, k} is calculated. An example using a matrix G ′ _{k, i} and a weight channel unitary matrix A _{k, i} is shown. A set signal space matrix B _{Ω, k that} is an (N _{S, k} × N _P ) × L _ALL matrix for the set Ω of the terminal 2 is _obtained as the following Expression (44), Expression (45), or Expression (46). It is done.

Here, _{pk, j, i} are weighting factors used for the _jth frequency channel matrix _{Aj, i} in the kth frequency channel for the terminal 2-i. L _ALL is the sum of the fed back vectors, and L _ALL = L ₁ + L ₂ +... + L _NSTA . These correspond to the left side of Expression (41) to Expression (43). Singular value decomposition (SVD) and QR decomposition can be performed on the collective signal space matrix B _{Ω, k} thus obtained as follows. The singular value decomposition can be obtained as the following equation (47).

In Expression (47), the matrix U _{B, Ω, k} is a left singular matrix of L _ALL × L _ALL . Matrixes Σ _{B, Ω, k} are diagonal matrices with singular values of L _ALL × L _ALL as diagonal components, and matrices V _{BS, Ω, k} are singular of (NS _{, k} × N _P ) × L _ALL . Is a right singular matrix corresponding to the value, and the matrix V _{BN, Ω, k} is a right singular matrix corresponding to a zero matrix of (N _{S, k} × N _P ) × (N _{S, k} × N _P −L _ALL ). is there. The thus obtained matrix V _{BS, Omega,} with _k, the antenna transmission weight W _T for a set of terminal 2 _{Omega, Omega} and may be obtained as the following equation (48).

_{Alternatively,} the matrix Q _{BS, Ω, k} of (NS _{, k} × N _P ) × L _ALL corresponding to the upper triangular matrix R _{B, Ω, k} obtained by using QR decomposition with the following equation (49) is used. When the antenna transmission weight W _T for Omega set of _terminals, it is possible to obtain the _Omega.

The antenna transmission weight W _{T, Ω} is an N × L _ALL matrix and can be obtained as the following equation (49).

The antenna transmission weights WT _{and Ω} may be used as training weights or data transmission weights.
Further, when the antenna transmission weights WT _{and Ω} thus obtained are used as channel estimation weights, the channel estimation weight matrix W _k can be expressed by the following equation (51).

  Alternatively, the channel estimation weight matrix is expressed by the following equation (52) using the antenna transmission weights calculated by the equations (31), (35), and (40) for the terminals belonging to the set Ω of the terminal 2. You can also.

Further, the SVD, QR method, or orthogonalization method is used for the matrix on the right side of Equation (51) or Equation (52) to convert the column vectors of the channel estimation weight matrix W _k into conditions that are orthogonal to each other. Thus, the channel estimation weight matrix W _k may be used.
The channel estimation weight matrix W _k is an N × N _P matrix, and the matrix W _{T, Ω} is an N × L _ALL matrix W _{T, Q} calculated using channel information _, and N × (N _P − L _ALL ) matrix, which is composed of additional channel estimation weights W _{0, Ω} .
Additional channel estimation weights _{W 0, Omega} is or determined by selecting a column vector from the training weight _{T B (N P -L ALL),} from the training weights _{T B} the column vectors of _{(N P -L ALL)} selected, W _T, using the orthogonalization method to satisfy the orthogonality condition to _Omega, it is also possible to use a column vector of the resulting (N P _{-L ALL).} The column vector of (N _P −L _ALL ) that is the source of the additional channel estimation weight W _{0, Ω} can be selected as follows.

In the formula (53), b '(i , k) represents the (number indicating what number of column vectors in matrix T _B) k-th subcarrier of the i-th beam and comprising additional channel estimation weight number. b ′ (i, k) can be represented by the following equation (54), for example.

Here, n is a coefficient indicating how many times the training packet is transmitted after starting to use WT _{and Q} as channel estimation weights, and N ′ _S corresponds to how many training weight numbers corresponding to each long training are shifted. It is a number to do. N ″ _S is a number corresponding to how many the corresponding training weight numbers are shifted each time a channel estimation sequence as shown in FIG. 14 is performed. By so setting that, it is possible to collect (N P _{-L ALL)} Even were very small number, the channel information corresponding to the gradually N _B number is the training weight.

In the first to third feedback methods, the signal processing method for each subcarrier has been described. However, feedback information need not be fed back to the base station 1 in all subcarriers. For example, by feeding back the feedback information of one sub-carrier to the base station 1 for each sub-carrier of N _F or by, in the base station 1, based against the N _F sub-carriers, the feedback feedback information data A transmission weight or a training weight may be determined. For subcarriers that do not have corresponding feedback information, the data transmission weight or training weight may be determined by interpolation or extrapolation based on the feedback information fed back to the base station 1.

Using the second feedback method, the terminal 2-i feeds back the weight channel unitary matrix A _{k, i} (= V _{S, k, i} U _{k, i} ^H ) to the base station 1, and the base station 1 collects the set matrix. The effect of acquiring A _{A, k, i} and determining the weight is shown by the result of computer simulation.
FIG. 5 is a graph showing an effect obtained by applying the second feedback method in the present embodiment. As a comparative example, a case where a matrix V _{S, k, i} is fed back is shown. In the figure, the horizontal axis indicates Achievable bit rate [bit / sec / Hz], and the horizontal axis indicates the cumulative distribution function (CDF) [%].

  In this computer simulation, a MIMO channel having 32 transmitting antennas and 2 receiving antennas is generated so that each element has a complex Gaussian distribution. The average of each element of the channel matrix is set to 0, and the square value of the variance value is set to 100. Further, for simplification, there is no frequency selective fading, and channel information in all subcarriers of OFDM is set with the same 2 × 32 channel matrix H.

The case where the number of training symbols was 16 (N _P = 16) and the case where the number of training symbols was 8 (N _P = 8) were examined. When it is assumed that the training symbol can be received without error in the terminal 2-i, a 16 × 2 matrix (V _{S, k, i} U _{k, i} ^H ) is fed back to the base station 1 when N _P = 16. In addition, when N _P = 8, an 8 × 2 matrix (V _{S, k, i} U _{k, i} ^H ) is fed back to the base station 1. For simplicity, the quantization error is not considered.

When the matrix (V _{S, k, i} U _{k, i} ^H ) is fed back to the base station 1, the 32 × 2 matrix V _{S, A, k, i} or the matrix is _{obtained from the} equation (33) or the equation (34). Q _{S, k, i} is restored. At this time, it was verified how much the transmission characteristics (Achievable bit rate) deteriorated compared with the case where the right singular vector V of the actual 32 × 2 channel matrix H was used as the transmission weight. Matrix _{V S, k, i U k} , instead of feeding back the ^{i H,} the matrix _{V S, k,} by feeding back the _i matrix in equation (33) or formula (34) _{V S, k,} the case of using the _i We also calculated the transmission characteristics. The Achievable bit rate (C) is defined as in the following equation (55).

  In Equation (55), the matrix W is a 32 × 2 weight matrix. The matrix W is determined so that the Frobenius norm of each column vector is equal (equal power distribution) and the square value of the Frobenius norm is 1. It is assumed that the square value of the variance value of the thermal noise is 1, and the thermal noise term is omitted in the equation (42). In FIG. 5, the graph (rightmost) described as “Capacity” uses the right singular vector V of the channel matrix H as the weight matrix W. This graph shows the condition value of Achievable bit rate. This graph corresponds to a case where a 32 × 2 channel matrix is fed back for each subcarrier.

On the other hand, a 16 × 2 matrix V _{S, k, i} or a matrix (V _{S, k, i} U _{k, i} ^H ) is fed back in each subcarrier, and information on two subcarriers is transmitted in the base station 1. The results when a 32 × 2 weight matrix is calculated using V2 and VU2 are shown. When the matrix V _{S, k, i} is fed back (graph V2), it can be seen that the characteristic is greatly deteriorated with respect to the graph of “Capacity”. If channel information for each of 16 training weights is estimated and only the right singular vector is fed back, they cannot be combined again. On the other hand, when a 16 × 2 matrix (V _{S, k, i} U _{k, i} ^H ) is fed back, it can be seen that the transmission weight W can be obtained with almost no deterioration (graph VU2). Deterioration is only 0.05 [bit / sec / Hz] at the median.

Similarly, in each subcarrier, an 8 × 2 matrix V _{S, k, i} or a matrix (V _{S, k, i} U _{k, i} ^H ) is fed back, and information on four subcarriers is used for the base station The result of calculating a 32 × 2 weight matrix with 1 is shown by a graph of V4 and VU4. As shown in these two graphs, when the matrix V _{S, k, i} is fed back (graph V4), the characteristics are remarkably degraded (decrease of 2.9 [bit / sec / Hz]). On the other hand, when the matrix (V _{S, k, i} U _{k, i} ^H ) is fed back (graph VU4), there is only a decrease of 0.15 [bit / sec / Hz] at the median, and high characteristics are obtained. It can be confirmed that

Then, the result of having analyzed the characteristic of the 3rd feedback method is shown.
FIG. 6 is a graph showing the result of analyzing the characteristics when the third feedback method in the present embodiment is applied by computer simulation. In this computer simulation, a channel model determined by IEEE 802.11n is used. Also, the Achievable bit rate by the third feedback method is evaluated by the channel generated by Model F. Also, in this evaluation, estimation errors due to thermal noise and channel compression quantization errors due to compressed beamforming matrix are included in the evaluation.

  In Model F, the number of transmission antennas is 16, the number of reception antennas is 4, a 64-point FFT is used for a bandwidth of 20 [MHz], and 52 subcarriers are used as data subcarriers. As an Achievable bit rate, Formula (55) is evaluated for each subcarrier, and the average value is used. “Full search” shown in FIG. 6 is a comparative example for the third feedback method of the present embodiment. In “Full search”, the same number of training symbols as the number of transmission antennas (N = 16, N = 8) are used.

In Model F, the average value of the delay spread is about 57 [ns]. In Model F, unlike the case shown in FIG. 5, the actual characteristic that the correlation between subcarriers decreases as the frequency between subcarriers increases.
The number of subcarriers N _S used for channel estimation by the _{terminal, k} is common to all subcarriers, and the N _s. Also, it is assumed that all transmission beams for 16 training weights are estimated by each subcarrier. When N _S = 8, the number of training symbols is 2 (N _P = 2), and when N _S = 4. The number of training symbols is 4 (N _P = 4).

Further, FIG. 6 shows two types of characteristics, Type I and Type II, in the third feedback method. In Type I, channel information in all subcarriers is fed back. In Type II, channel information in one subcarrier is fed back for every N _S subcarriers (N _F = N _S ). The the N _S each sub-carrier are grouped, a common transmission weight is used for each group of subcarriers.

When four training symbols are used (N _P = 4), although only a small number of four training symbols is used with respect to the number of transmission antennas 16, there is little deterioration in characteristics. For example, at the median, Type I is a decrease of 0.18 [bit / sec / Hz], and Type II is a decrease of 0.27 [bit / sec / Hz].
Further, even when the number of training symbols to be used is small and two training symbols are used (N _P = 2), that is, when two long preambles having a significantly smaller number than the number of transmission antennas 16 are used, the deterioration of the characteristics is not improved. It is slight. Specifically, at the median, Type I is a decrease of 0.72 [bit / sec / Hz], and Type II is a decrease of 0.99 [bit / sec / Hz].

A decrease in the Achievable bit rate caused by feeding back channel information of one sub-carrier for each the _{N S} subcarriers is negligible. Despite the fact that the amount of feedback information to the base station 1 is (1 / N _S ), high transmission characteristics are obtained. That is, the feedback method in the present embodiment and the reduction of subcarriers to which channel information is fed back have high affinity. Therefore, further improvement in characteristics can be expected by acquiring information on subcarriers that are not fed back by interpolating between subcarriers.

  Further, FIG. 7 is a graph showing the result of analyzing the characteristics when the second feedback method in the present embodiment is applied by computer simulation. Also in this computer simulation, a channel model determined by IEEE 802.11n is used. Also, the Achievable bit rate by the second feedback method is evaluated by the channel generated by Model F. Also, in this evaluation, estimation errors due to thermal noise and channel compression quantization errors due to compressed beamforming matrix are included in the evaluation.

In Model F, the number of transmitting antennas is 32, the number of receiving antennas is 2, the number of terminals is 3 (N _STA = 3), a 64-point FFT is used for a bandwidth of 20 [MHz], and 52 subcarriers are used as data subcarriers. (F = 52). The number of training symbols is 8 (N _P = 8) except for the full search, and the transmission weight is assigned to each user by using block diagnosis in the multi-user MIMO technology in which three users transmit simultaneously in the same frequency band. Calculated. Using the transmission weights W _{t, k,} i for the k-th frequency channel for the terminal 2-i, the Achiable bit rate was defined as the following equation (56).

“Full search” shown in FIG. 7 is a comparative example for the second feedback method of the present embodiment. In “Full search”, the same number of training symbols (N _P = 32) as the number of transmission antennas (N = 32) are used. For comparison, a conventional method (Conv.) Is also included. As a conventional method, eight channel estimation weights are simply selected from training weights at random, the determined transmission weights orthogonal to each other are determined for each subcarrier, and channel information can be used without using information on a plurality of subcarriers. Orthogonal vectors corresponding to the six signal spaces are derived by the first transmission weight calculation method with high estimation accuracy, and this is used as WT _{, k} in the equation (51), and further from the training weight as W0 _{, k.} The channel estimation weights are selected at random, 8 training symbols are newly transmitted, and the transmission weight of the multiuser MIMO is derived from the obtained feedback information by the first transmission weight calculation method. Equation (56) It is the result of evaluating Achievable bit rate by.

On the other hand, the proposed method (Proposal) uses the weight channel unitary matrix A _{k, i} fed back by transmitting 8 training symbols _, and uses the information of NS frequency _k frequency channel to calculate the second transmission weight. The channel estimation weight is derived by using as WT _{, k} in Equation (51) _, and two channel estimation weights are randomly selected from the training weights as W0 _{, k} , and eight new training symbols are transmitted. Then, from the obtained feedback information, the transmission weight of multi-user MIMO is derived by the first transmission weight calculation method. That is, the difference between “Conv.” And “Proposal” is whether or not information of a plurality of frequency channels is used for channel information fed back at the first training symbol transmission.

In Model F, the average value of the delay spread is about 57 [ns]. In Model F, unlike the case shown in FIG. 5, the actual characteristic that the correlation between subcarriers decreases as the frequency between subcarriers increases. As training weights, orthogonal basis vectors having 32 transmission antennas were used (N _B = 32).
The number of subcarriers used for channel estimation at the terminal is assumed to be common to all subcarriers and can be expressed as N _{s, k} . In the pilot design used for transmission, N _S = 4 is set in Equation (7) and 8 training symbols are used. Therefore, if four frequency channels are used, channel information for all training weights can be obtained. (N _S × N _P = N _B ). FIG. 7 shows the result when the number N _{S, k} of frequency channels used for determining the channel estimation weight of one frequency channel in the base station 1 is 3, 4, and 7.

In the second feedback method, since singular value information is not fed back, 91.5%, 94.3%, and 95.0% Achievable bit rates are obtained with respect to “Full search”. It can be seen that the characteristics can be greatly improved from the result of “Conv.” Which does not use the frequency channel information. Furthermore, NS, if 4 to k, despite the channel information are aligned with respect to all training weights, further N _S, the transmission characteristics by increasing the _k can be confirmed also be slightly improved. This is considered to be because the estimation accuracy can be improved by estimating channel information corresponding to the same training weight with another frequency channel.

FIG. 8 is a flowchart of communication in the first and second feedback methods in the present embodiment.
When communication is started, the base station 1 first determines a terminal 2-i to be a communication partner that performs channel estimation or data transmission (step S001).
The channel estimation weight determination circuit 11 determines whether or not the channel estimation weight calculated from the significant channel information for the terminal 2-i as the communication partner is stored (step S002). Here, the significant channel estimation weight can be determined, for example, by storing channel estimation weights using channel information estimated within a predetermined time.

If you do not store the channel estimation weight (step S002: NO), the channel estimation weight determining circuit 11, from the training weight T _B stored in advance, the channel estimation weight matrix W _k to be used for each frequency channel and each training symbol Is selected (step S003), and the process proceeds to step S005.
On the other hand, when the channel estimation weight is stored (step S002: YES), the channel estimation weight determination circuit 11 stores in advance the channel estimation weight matrix W _k stored as in Expression (51). determining a new channel estimation weight matrix _{W k} by using both the training weight _{T B} (step S004).

The transmission signal generation circuit 12 generates an OFDM packet using the channel estimation weight matrix W _k determined by the channel estimation weight determination circuit 11 in step S003 or step S004, and the radio signal transmission / reception circuit 13 transmits the OFDM packet ( Step S005).
The terminal 2-i receives a training symbol transmitted using the channel estimation weight matrix W _k , estimates a weight channel matrix (step S006), and channel information including a channel matrix, a right singular matrix, or a similar basis vector Is fed back to the base station 1 (step S006).

In the base station 1, the channel information acquisition circuit 17 determines the channel information matrix (H _{k, i} , A _{k, i} , G _k, which is a channel matrix, a right singular matrix, or a similar basis vector from the fed back channel information _{. i} , G ′ _{k, i} ), the dimension of the channel matrix as shown in Equation (30), Equation (32), Equation (36), Equation (44), Equation (45), and Equation (46). Expansion is performed (step S009). The channel information acquisition circuit 17 uses the obtained set matrix to perform weight calculation as shown in Equation (31), Equation (35), Equation (40), Equation (48), and Equation (50). S010), the process is returned to step S001, and the processes from step S001 to step S010 are repeated. At this time, when the channel estimation weight is calculated, the obtained weight is stored in the channel estimation weight determination circuit 11.
The channel information acquisition circuit 17 calculates the weight based on the single frequency channel information based on the first transmission weight calculation method in parallel with step S009 and step S010 (step S007), and performs processing. Returning to step S001, steps S001 to S010 are repeated.

FIG. 9 is a communication flowchart in the third feedback method according to this embodiment.
When communication is started, the base station 1 first determines a terminal 2-i to be a communication partner that performs channel estimation or data transmission (step S101).
The channel estimation weight determination circuit 11 determines whether or not the channel estimation weight calculated from the significant channel information for the terminal 2-i as the communication partner is stored (step S102). Here, the significant channel estimation weight can be determined, for example, by storing channel estimation weights using channel information estimated within a predetermined time.

If you do not store the channel estimation weight (step S102: NO), the channel estimation weight determining circuit 11, from the training weight T _B stored in advance, and select the channel estimation weight to be used for each frequency channel and each training symbol (Step S103), the process proceeds to Step s105.
On the other hand, when the channel estimation weight is stored (step S102: YES), the channel estimation weight determination circuit 11 stores in advance the channel estimation weight matrix W _k stored as in equation (51). determining a new channel estimation weight matrix _{W k} by using both the training weight _{T B} (step S104).

The transmission signal generation circuit 12 generates an OFDM packet using the channel estimation weight matrix W _k determined by the channel estimation weight determination circuit 11 in step S103 or step S104, and the radio signal transmission / reception circuit 13 transmits the OFDM packet ( Step S105).
The terminal 2-i receives the training symbol transmitted using the channel estimation weight matrix W _k and estimates the weight channel matrix (step S106). In the terminal 2-i, the channel information compression circuit 25-i receives a channel matrix (H _{k, i} , A _{k, i} , G _{k, i} , G ′) that is a channel matrix, a right singular matrix, or a similar basis vector. _{k, i} ) is used to expand the dimension of the channel matrix as shown in Equation (30), Equation (32), Equation (36), Equation (44), Equation (45), and Equation (46) (step) S107).
The channel information compression circuit 25-i feeds back channel information including the obtained set matrix, the right singular matrix of the set matrix, or a similar basis vector to the base station 1 through the radio signal transmission / reception circuit 23-i (step S108). ).

  In the base station 1, the channel information acquisition circuit 17 calculates the weight calculation from the fed back channel information as shown in Equation (31), Equation (35), Equation (36), Equation (48), and Equation (50). (Step S109), the process is returned to Step S101, and the process from Step S101 to Step S109 is repeated. At this time, when the channel estimation weight is calculated, the obtained weight is stored in the channel estimation weight determination circuit 11.

Hereinafter, a method for determining the channel estimation weight in the channel estimation weight determination circuit 11 will be described. Channel estimation weight determining circuit 11, the n-th channel estimation, N _{the add} from the training weight matrix T _B stored in _advance, the _n vectors selected for each frequency channel, be used as a channel estimation weight it can. Simply, N _{add, 1} = N _P at _{the first} channel estimation, and N _{add, n} = N _P -L _{ALL at the} subsequent time. At this time, L _ALL calculated by the channel information acquisition circuit 17 (the sum of the fed back vectors, that is, the sum of the spatial multiplexing numbers of the terminal 2-i) is used in addition to the N _{add, n} weights. Note that L _ALL or less weights may be used from the weights calculated by the channel information acquisition circuit 17.

At this time, the rules for selecting a vector from the training weight matrix T _{B (codebook),} a vector for the N _S frequency channels to be combined in the base station 1 N _S × N _{add, n-number} of code Choose to minimize vector overlap in the book vector. As a result, the effect of combining channel matrices as obtained from equations (32) to (33) can be enhanced. Further, regarding the second and subsequent channel estimation, by repeating the channel estimation step, it is possible to improve long-term channel estimation accuracy by selecting all the codebook vectors in each frequency channel.

For example, it is possible to perform also the determination of the codebook in Equation (7), this time, when the number N _S of the vector number N _B and the frequency channels of the codebook will have a common divisor other than 1, wherein (7) the number obtained in the not all of the numbers in between 1 and N _B. If the common divisor is 2, the number obtained by Equation (7) is either odd or even. For information on their frequency channel is the most accurate, estimating all vectors of the training weight matrix T _B in the frequency channel, is very important in improving the accuracy of channel estimation. When setting the number of frequency channels _{NS, n} to be independently coupled to the nth channel estimation, for example, the following vector (57) can be used to designate a vector to be selected for NS _{, n} .

By using Equation (57), it is possible to avoid an increase in duplication when combining _{NS, n} frequency channel information. Further _{N S,} that _n and _{N B} is set to have no common divisor other than 1, by a sufficient interval _{N S,} the _n constant, b (m k) is _{N B} number of numbers outputs all can acquire the channel information for all vectors (channel estimation weight) training weight matrix T _B in the frequency channel.

In the equation (57), the first long training frame can be set as b (1, k) = 1 + (k−1 mod N _B ), for example. In Expression (57), m is a total number of long training symbols using the initial channel estimation and the additional channel estimation weight in the subsequent channel estimation. When the channel estimation interval becomes large and the accuracy of the past channel information becomes low, it can be reset to m = 1. An example of equation (57) is shown in FIG. FIG. 10 is a diagram showing an example in which channel estimation weights are assigned to each carrier according to Expression (57). Here, 32 a vector number of the training weight matrix _{T B, N S,} the _n 5, the number of sub-carrier (frequency channel) is set to 52.

It can be selected vector (channel estimation weight) from the training weight matrix T _B as described above but, N _{the add,} as _n is large, can reduce the time it takes until the channel estimation for all vectors. However, channel estimation weights W _{T, Ω} calculated from the channel information are used for the long training symbol of the vector number L _ALL fed back from the communication partner, and the channel estimation accuracy becomes better. Thus, while obtaining channel information for L _ALL pieces of channel estimation weights, N _{the add,} for increasing _n, it may be distributed to L _ALL number of vectors into a plurality of frequency channels.

Here, the number of frequency channels belonging to a subcarrier group that uses channel estimation weights based on channel information is defined as _NSG . Equation (48), _{W T} of the formula _{(51), Omega} is calculated at only one frequency channel to _{N SG} pieces. Each resulting _{L ALL} pieces of N × 1 vector _{N SG} subcarriers _are distributed _{N SL (1) ~N SL (} N SG) or by, a channel estimation weight. Note that L _ALL = N _SL (1) + N _SL (2) +... + N _SL (N _SG ).

In this way, the long training symbol number _{N P} of the channel estimation is assumed to be _{constant, N the add,} the number of _{n N P -N SL (γ)} (γ = 1,2, ..., N SG ), And the number of additional channel estimation weights can be increased as compared to N _P -L _ALL when N _SG = 1 and no subcarrier group is used, and all weight matrix vectors in a certain frequency channel It is possible to reduce the time taken to estimate In other words, it assigns all vectors of the training weight matrix T _B in the frequency channel, by minimizing the number of overlapping weight of the weight to be assigned in consecutive frequency channels of a predetermined number (N _S), all the frequency channels The time taken to estimate the weight matrix vector can be reduced.

FIG. 11 is a graph showing the effect of the wireless communication system in the present embodiment. In the figure, the horizontal axis represents time (number of channel estimations), and the vertical axis represents Achievable bit rate [Gb / s]. The IEEE 802.11n model F is used as the channel model. Expected value of the delay spread 100 [ns], the terminal 2-i and the distance between the base station 1 20 [m], the number of transmitting antennas and the training weight matrix _{T B} of the number of vector 32, the terminal 2-i of the receiving antenna The number was calculated as 2, the number of users as communication partners was 3, the bandwidth was 20 [MHz], the number of data subcarriers was 52 (FFT points: 64), and the Doppler spread was 0.41 [Hz]. The long training frame number N _P is fixed at 8, wait channel unitary matrix A _k of 8 × 2 from each terminal _2-i, and must obtain _i. The second and subsequent communications use the channel estimation weight calculated from the channel estimation result of L _ALL = 2 × 3 = 6 and two additional channel estimation weights (N _{add, n} = 2). The channel estimation is performed every 6 [ms], and the transmission weight calculated based on the estimated channel information is used to calculate the throughput C defined by the following equation (58).

As the transmission weights multiplied by the data symbols are only those of the frequency channel, the throughput in the first communication is low. This is the first channel estimate weights do not include the channel estimation weight calculated by the channel information is for that consists of selected vector from all weight matrix, simply N _P number of Since the channel for the vector is estimated, the characteristics are close to those when the number of antennas is eight.

In the case where the N _S always 1, it takes time to convergence characteristics, by the N S _[1] to 7, using the estimation results for 4 × 8 = 32 pieces of vector from the data communication for the second time Therefore, it can be confirmed that high throughput characteristics can be obtained after the second time. On how to set the N _S includes a first channel estimation, each optimum value in the case of channel estimation after convergence different.

12, in the radio communication system of the present embodiment is a graph showing the setting of the N _S during first channel estimation characteristics of the throughput performance for the setting of the N _S after convergence. N _S is used to first round of channel estimation (for second transmission data) has a maximum upon four. However, there is no large change is also a characteristic set of 4 greater than the N _S. It can be seen that it is necessary to set a number of 4 or more that can use all the weight matrix vectors. However, if you set the like 8, 16 into _{N S} characteristics decreases significantly. This is because channel information for the same vector is estimated with the same frequency channel. For example, in the case of N _S = 8, if vector selection according to Equation (57) is used, the 1st to 4th and the 5th to 6th out of 8 long training symbols are exactly the same combination. When N _P × N _{add, 1} exceeds N _B (= 32), it is important not to perform channel estimation for the same vector multiple times on the same frequency channel.

As described above, when n> 1, that is, in the second and subsequent channel information estimation, the base station 1 already has channel estimation results for all weight matrix vectors or most vectors. It is required to obtain channel information with higher accuracy. As shown in FIG. 12, the optimum value of N _S is the transmission after convergence, and has a 1 or 3. The reason why 1 is a high characteristic is that the calculation of the combination of channel information in a plurality of frequency channels is not performed, so that it is not affected by channel information deterioration due to frequency selective fading. Higher N _S, since it is possible to reduce the time required to collect the channel information for all vectors, although it is possible to prevent an increase in the channel estimation error due to time variation of the channel, the error is increased due to frequency selective fading . It is necessary to consider the influence of these two characteristics. Since the model F used in the study with a significant frequency selective fading delay spread 100 [ns], but the deterioration of the characteristics by increasing the N _S it is assumed strong, but still N _{S =} 3 is higher throughput It can be confirmed that it has characteristics. The effect due to the Doppler frequency is increased N _S The greater the stronger, the optimum value is larger value.

For reference, FIG. 13 shows the result of model C with a delay spread of 30 [ns] (distance 5 m between terminal 2-i and base station 1). FIG. 13 is a graph showing throughput characteristics in model C with a delay spread of 30 [ns]. Here, it can be confirmed that the optimum value is N _S = 3, which is higher than N _S = 1. 12 and 13, transmission characteristics after convergence can confirm the N _S is lowered during the even. This is because having a common divisor other than N _B and 1, at a certain frequency channel is because can not estimate the channel information for all of the weight matrix vector.

  Note that a program for realizing the functions of the base station 1 and the terminal 2-i is recorded on a computer-readable recording medium, the program recorded on the recording medium is read into a computer system, and the program is executed. Thus, wireless communication processing including feedback of channel information may be performed. Here, the “computer system” includes an OS and hardware such as peripheral devices. The “computer-readable recording medium” refers to a storage device such as a flexible medium, a magneto-optical disk, a portable medium such as a ROM and a CD-ROM, and a hard disk incorporated in a computer system. Further, the “computer-readable recording medium” refers to a volatile memory (RAM) in a computer system that becomes a server or a client when a program is transmitted via a network such as the Internet or a communication line such as a telephone line. In addition, those holding programs for a certain period of time are also included.

  The program may be transmitted from a computer system storing the program in a storage device or the like to another computer system via a transmission medium or by a transmission wave in the transmission medium. Here, the “transmission medium” for transmitting the program refers to a medium having a function of transmitting information, such as a network (communication network) such as the Internet or a communication line (communication line) such as a telephone line. The program may be for realizing a part of the functions described above. Furthermore, what can implement | achieve the function mentioned above in combination with the program already recorded on the computer system, and what is called a difference file (difference program) may be sufficient.

The embodiments according to the present invention have been described above with reference to the drawings. The above-described embodiment is merely an example, and the present invention is not limited to the configuration of the embodiment. Accordingly, additions, omissions, and substitutions of configurations may be made without departing from the technical idea of the present invention.
For example, in the above-described embodiment, the base station 1 determines a training weight or a data transmission weight based on information fed back from the terminal 2-i. However, the terminal 2-i may determine a training weight or a data transmission weight based on information fed back from the base station 1. Information may be fed back between the plurality of base stations 1 to determine a training weight or a transmission weight.

  Further, the channel estimation weight determination circuit 11 may determine the number of training symbols to be transmitted based on the reception quality (for example, SNR) at the terminal 2-i. In this case, the channel estimation weight determination circuit 11 determines a channel estimation weight corresponding to the determined number of training symbols. Further, the channel estimation weight determination circuit 11 may determine the number of training symbols to be transmitted according to the data packet length transmitted by the base station 1. Further, the channel estimation weight determination circuit 11 may determine the number of training symbols to be transmitted according to the combination of reception quality and data packet length at the terminal 2-i. When the data packet length is short, a reduction in throughput due to an increase in overhead can be reduced by reducing the number of training symbols.

  In wireless communication using the OFDM modulation scheme, it is possible to reduce the number of OFDM symbols used for channel information estimation and to reduce the amount of information in the transmission of estimated channel information, and to improve MAC efficiency by suppressing communication overhead Thus, a wireless communication system with improved throughput can be realized.

DESCRIPTION OF SYMBOLS 1 Base station 2, 2-i terminal 11 Channel estimation weight determination circuit 12 Transmission signal generation circuit 13 Radio signal transmission / reception circuit 14-1, 14-2, 14-N Antenna 15 Reception signal demodulation circuit 16 Feedback information extraction circuit 17 Channel information Acquisition circuit 22-i Transmission signal generation circuit 23-i Radio signal transmission / reception circuit 24-i-1, 24-i-2, 24-i-M _i antenna 25-i Channel information compression circuit 26-i Feedback information generation circuit 100 Wireless communication system 91 Base station 92, 92-i Terminal 912 Transmission signal generation circuit 913 Radio signal transmission / reception circuit 914-1, 914-2, 914-N Antenna 915 Reception signal demodulation circuit 916 Feedback information extraction circuit 917 Channel information acquisition circuit 922 -I transmission signal generation circuit 923-i wireless signal transmission / reception 924-i-1, 924-i-2, 924-i-M _i antenna 925-i channel information compression circuit 926-i feedback information generation circuit

Claims (12)

A wireless communication system comprising a plurality of communication stations that perform communication using an orthogonal frequency division multiplexing modulation system,
A channel estimation weight determining means for determining a channel estimation weight corresponding to each of N _P number of training symbols to be transmitted for each frequency channel,
Generates N _P-number of long training symbols for each frequency channel by multiplying the channel estimation weight to training symbols corresponding to the channel estimation weight, the OFDM packet including the generated N _P-number of long training symbol as a communications partner Training symbol transmitting means for transmitting to the other communication station;
Based on the received OFDM packet, and the weight channel matrix estimation means for estimating the weight channel matrix is estimated channel estimation weight inclusive channel information by N _P entries per frequency channel,
Feedback information transmitting means for transmitting feedback information calculated based on the weight channel matrix to a communication station of a transmission source that has transmitted the OFDM packet;
A channel information matrix of each frequency channel fed back from the communication partner based on the feedback information is restored, and a channel information matrix of a plurality of frequency channels corresponding to one of the communication partners, or corresponding to a plurality of the communication partners. A set matrix having a larger size than the original channel information matrix is generated by using channel information matrices of a plurality of frequency channels as one matrix, and the weight used for communication is set to a frequency based on the generated set matrix and the channel estimation weight. Weight calculation means for calculating for each channel;
A wireless communication system comprising: The wireless communication system according to claim 1,
The feedback information transmitting means includes
The wireless communication system, wherein the feedback information including the weight channel matrix, a right singular vector obtained from the weight channel matrix, or a basis vector obtained from the weight channel matrix is transmitted. The wireless communication system according to claim 1,
The feedback information transmitting means includes
Transmitting the feedback information including a weight channel unitary matrix obtained by multiplying a vector corresponding to a singular value of the right singular vector of the weight channel matrix and a Hermitian matrix of the left singular value vector of the weight channel matrix;
The weight calculation means includes
Based on the feedback information, the weight channel unitary matrix of each frequency channel fed back from the communication partner is restored, the weight channel unitary matrix of a plurality of frequency channels corresponding to one of the communication partners, or a plurality of the communication partners Select a corresponding weight channel unitary matrix for each of a plurality of frequency channels, and generate a set matrix having a size larger than the original weight channel unitary matrix using the selected plurality of weight channel unitary matrices as one matrix, A wireless communication system, wherein a weight used for communication is calculated for each frequency channel based on the channel estimation weight. The wireless communication system according to claim 1,
The feedback information transmitting means includes
A V matrix that is a basis vector having a high correlation with the right singular vector corresponding to the singular value of the weight channel matrix is calculated, and an orthogonalization method is applied to the matrix obtained by multiplying the calculated V matrix from the right side by the weight channel matrix. A U matrix that is a unitary matrix is used to transmit the feedback information including a weight channel unitary matrix obtained by multiplying the Hermitian matrix of the U matrix and the V matrix;
The weight calculation means includes
Based on the feedback information, the weight channel unitary matrix of each frequency channel fed back from the communication partner is restored, the weight channel unitary matrix of a plurality of frequency channels corresponding to one of the communication partners, or a plurality of the communication partners Select a corresponding weight channel unitary matrix for each of a plurality of frequency channels, and generate a set matrix having a size larger than the original weight channel unitary matrix using the selected plurality of weight channel unitary matrices as one matrix, A wireless communication system, wherein a weight used for communication is calculated for each frequency channel based on the channel estimation weight. The wireless communication system according to claim 3 or claim 4,
The weight calculation means includes
Based on the feedback information, the weight channel unitary matrix of each frequency channel fed back from the communication partner is restored, the weight channel unitary matrix of a plurality of frequency channels corresponding to one of the communication partners, or a plurality of the communication partners Select the corresponding weight channel unitary matrix for multiple frequency channels, multiply each selected multiple weight channel unitary matrix by the weighting factor, and then create a set matrix that is larger than the original weight channel unitary matrix as one matrix A wireless communication system, characterized in that a weight used for communication is calculated for each frequency channel based on the generated set matrix and the channel estimation weight. The wireless communication system according to claim 1,
The feedback information transmitting means includes
A plurality of weight channel matrices are used as one matrix to generate a set matrix larger in size than the original weight channel matrix, and the set matrix, a right singular vector obtained from the set matrix, or a basis vector obtained from the set matrix Including the feedback information including,
The weight calculation means includes
A wireless communication system, wherein a channel information matrix of each frequency channel fed back from the communication partner is restored based on the feedback information, and a weight used for communication is calculated for each frequency channel. A wireless communication system according to any one of claims 1 to 4,
The channel estimation weight determining means includes
The wireless communication system, wherein the weight calculated by the weight calculation means is determined as the channel estimation weight. The wireless communication system according to any one of claims 1 to 6,
The feedback information transmitting means includes
The wireless communication system, wherein the feedback information is calculated based on a weight channel matrix corresponding to a part of frequency channels. The wireless communication system according to any one of claims 1 to 8,
The channel estimation weight determining means includes
When feedback is obtained from the communication partner and the weight calculation means calculates L _ALL weights as channel estimation weights, the channel estimation weights are calculated from the L _ALL weights and a predetermined training weight matrix. n _{add, n (however, n add, n = n P} -L ALL) wireless communication system, characterized by selecting the number of weights. The wireless communication system according to claim 9, wherein
The channel estimation weight determining means includes
When feedback is obtained from the communication partner and the weight calculation means calculates L _ALL weights as channel estimation weights, the L _ALL weights are divided into two or more groups, and the weights divided into groups are obtained. When the number of weights to be distributed to the subcarriers and n _SL are a at, as a channel estimation weight of the sub-carrier, and n _SL number of weights the weight calculating means is calculated, n _{the add} from the training weight _{matrix, n (where, n add, n = n P} -N SL) wireless communication system, characterized by selecting the number of weights. A wireless communication system according to claim 9 or 10, wherein
The channel estimation weight determining means includes
When N _{add, n} weights are selected from the training weight matrix, all weights included in the training weight matrix are assigned to frequency channels, and weights that are duplicated in weights assigned to a predetermined number of consecutive frequency channels A wireless communication system characterized by minimizing the number of signals. A weight calculation method in a wireless communication system including a plurality of communication stations that perform communication using an orthogonal frequency division multiplexing modulation system,
A channel estimation weight determining step of determining a channel estimation weight corresponding to each of N _P number of training symbols to be transmitted for each frequency channel,
Generates N _P-number of long training symbols for each frequency channel by multiplying the channel estimation weight to training symbols corresponding to the channel estimation weight, the OFDM packet including the generated N _P-number of long training symbol as a communications partner A training symbol transmission step of transmitting to the other communication station;
Based on the received OFDM packet, and the weight channel matrix estimation step of estimating the weight channel matrix is estimated channel estimation weight inclusive channel information by N _P entries per frequency channel,
A feedback information transmission step of transmitting feedback information calculated based on the weight channel matrix to a communication station of a transmission source that has transmitted the OFDM packet;
A channel information matrix of each frequency channel fed back from the communication partner based on the feedback information is restored, and a channel information matrix of a plurality of frequency channels corresponding to one of the communication partners, or corresponding to a plurality of the communication partners. A channel information matrix of each of a plurality of frequency channels is selected, a set matrix having a size larger than the original channel information matrix is generated using the selected plurality of channel information matrices as one matrix, and the generated set matrix and the channel estimation weight are generated. A weight calculating step for calculating the weight used for communication for each frequency channel,
The weight calculation method characterized by having.

Images