JP2014075679A - Wireless communication system, wireless transmitter and wireless communication method - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a wireless communication system which suppresses an increase in communication overhead caused by feedback of channel state information and furthermore can achieve a good spectrum efficiency, in a MIMO communication system.SOLUTION: The wireless communication system performs wireless communication of a signal modulated by an orthogonal frequency division multiplex system, by MIMO. A base station receives channel state information for each subband fed back from a terminal, and calculates a perturbation vector in a VP method (S104). Further, time-frequency processing for obtaining the channel state information on the predetermined number of subcarriers is performed with respect to the channel state information for each subband (S106 and S108), and a weighting coefficient for controlling the antenna directivity is calculated (S110), and linear interpolation of the weighting coefficient with respect to all subcarriers is performed (S114).

Description

  The present invention relates to a radio communication system in which a base station having a plurality of antennas and a terminal device exist, and more particularly to a transmission beamforming technique in a radio communication system of multi-user MIMO (Multiple Input Multiple Output) communication. The present invention relates to a wireless communication system, a wireless transmission device, and a wireless communication method.

  Conventionally, multi-user MIMO technology has been proposed (Patent Document 1, Patent Document 2, and Patent Document 3). Multi-user MIMO has a large number of antenna elements on the base station (or access point) side and a relatively small number of antenna elements on the terminal side, so that a virtual MIMO channel can be formed simultaneously between the base station and a plurality of terminals. To do.

  In other words, the multi-user MIMO transmission technology is such that, on the transmitting station side, independent signals different from each other at the same frequency and the same timing are transmitted from a plurality of transmission antennas to a plurality of communication counterpart devices, and the entire reception antennas of the plurality of communication counterpart devices are enormous. This is a technique for improving the downlink throughput as a receiving array.

  On the other hand, technologies such as the IEEE802.11g standard and the IEEE802.11a standard are prevalent as high-speed wireless access systems using the 2.4 GHz band or the 5 GHz band. In these wireless communication systems, an Orthogonal Frequency Division Multiplexing (OFDM) system, which is a technique for stabilizing characteristics in a multipath fading environment, is used.

  Also in such a high-speed wireless access technology, a multi-user MIMO transmission technology has attracted attention as a technology for increasing the capacity.

  Furthermore, the multi-user MIMO technology is also adopted in cellular phone standards such as LTE (Long Term Evolution) and LTE-A (Long Term Evolution Advanced: LTE).

  FIG. 30 is a conceptual diagram showing the configuration of such a multiuser MIMO communication system.

As shown in FIG. 30, the base station BS transmits transmission signals x _{1 to} x _M from antennas _{1 to M} , respectively. Terminals UE _{1 to} UE _k are each provided with two antennas, for example. Terminal UE ₁ receives transmission signals x _{1 to} x _M from the base station via its two antennas and receives signals y ₁ and y ₂ , respectively. Similarly, other terminals UE ₂ ~UE _k, by each of the two antennas, receives a transmission signal x ₁ ~x _M from the base station, the signal y ₃ and _{y 4, ..., y M-} 1 and to receive y _M.

  That is, the example of FIG. 30 illustrates a case where (the number of transmission antennas of the base station) = (the number of reception antennas on one terminal side) × (the number of terminals).

At this time, a received signal vector Y in which the received signals y _{1 to} y _M at the antennas at the receiving side terminals UE _{1 to} UE _k are collected is expressed by the following expression.

Here, each element of the matrix H corresponds to an impulse response of a transmission path from each antenna on the transmission side to each antenna on the reception side, and the matrix H is called a “channel response matrix” (or “transmission path matrix”). . The vector X is a vector in which transmission signals to each antenna on the transmitter side are arranged. The vector N is an array of noise components included in signals received by the receiving antennas.

  As a transmission beam forming (BF: Beam Forming) method in a conventional multi-user MIMO downlink, a ZF (Zero Forcing) method for forming nulls for all receiving antennas of all terminals other than the own terminal, Various beam forming methods based on the MMSE (Minimum Mean Square Error) method have been devised. The beam forming method based on the MMSE method does not form nulls for all receiving antennas of all terminals other than its own terminal, but allows a certain amount of leakage.

  In the following, it is assumed that the number M of transmission antennas is M = Nt = Nr × K. Nr is the number of antennas on the receiving side (mobile terminal), and K is the number of mobile terminals.

(Configuration of beam forming by MMSE method)
Here, in the case of a single carrier, a method of estimating a “weighting coefficient” (beam forming weight matrix) for combining transmission signals to be given to each antenna by the MMSE method is disclosed in Japanese Patent Application Laid-Open No. 2007-110664. No.).

  Hereinafter, beam formation by the MMSE method will be briefly described using mathematical expressions.

  FIG. 31 is a conceptual diagram showing a beam forming procedure by the MMSE method.

  Here, in the transmitter BS, the number of antennas used for transmission is Nt, and on the receiving side, a configuration is described in which a signal is received by Nt (= Nr × K) antennas.

  The original signal to be transmitted is represented by a vector X as follows.

At this time, for example, the channel response matrix H is assumed to be estimated from information from the reception side (receiver UEs).

  In such a situation, the signal vector X hat after multiplying the transmission beam weight (the one in which “^” is added on X is expressed in the following manner in the text) is expressed by the following equation. It is.

Here, superscript H indicates a Hermitian matrix after matrix conjugate transposition operation, α indicates a normalization coefficient, and I _Nt indicates a unit matrix of (Nt × Nt).

  The signal Z transmitted from the transmission antenna is expressed by the following equation.

Here, γ is a coefficient for normalizing the total transmission power. However, || represents the Euclidean norm.

After the transmitted signal is transmitted through the channel (ie, equivalent to the signal Z multiplied by the channel response matrix), on the receiving side, √γ is given to the signal with additive white Gaussian noise (AWGN). Each is multiplied to separate the signal x _i hat.

  As described above, MIMO communication is realized by the beam forming method using the MMSE method.

  However, multi-user MIMO is not only used outdoors, but is increasingly used indoors such as in offices and homes.

  For example, for near field communication systems such as picocell or femtocell systems, the spectral efficiency (SE) of the system is multi-user MIMO (with accurate channel state information (CSI) at the transmitter or base station BS side ( It is known to be improved using the technology of the (MU-MIMO) algorithm.

  However, in such a case, each receiver is spatially close, and as a result, the correlation between MIMO channels is high.

  In such a case, in the above-described MMSE method, for the channels highly correlated with each other, the value of the normalization coefficient γ becomes too large, resulting in a channel in which the normalized signal power is suppressed more than necessary. Therefore, there is a problem that part of the stream data may be lost.

(Nonlinear precoding method: VP method)
For such a problem, for example, a VP (Vector Perturbation) method of a nonlinear precoding method is known. Such a “non-linear precoding method” is also disclosed in Patent Document 4, for example.

  In a non-linear precoding method such as the VP method, a precoding process is performed on a transmission signal addressed to a receiving device that receives interference between receiving devices, and an interference component is thinned out in advance.

  In a multi-user MIMO (VP MU-MIMO) system using the VP method, a terminal device performs an operation called a modulo operation (modulo operation) on a received signal. The modulo operation is a process in which, on the signal point plane, points translated by an integral multiple of a predetermined interval called a modulo width are regarded as the same signal point.

  When the terminal apparatus regards a plurality of signal points as the same, it means that the base station apparatus has a higher degree of freedom when selecting a transmission signal. The base station apparatus uses this degree of freedom to select and transmit points that can be transmitted with lower power.

  The base station BS generates a transmission signal according to the following two procedures when performing VP MU-MIMO communication.

  1. Add a signal (perturbation vector) that is an integer multiple of the appropriately selected modulo width (hereinafter “modulo width”) to the desired signal.

  2. The interference between mobile terminals is removed by the same process as the linear precoding.

  At this time, the signal to be added is called a perturbation vector. When this perturbation vector is appropriately selected in consideration of signals addressed to all mobile terminals and propagation path conditions, transmission power can be suppressed. Although the VP MU-MIMO method uses the same linear filter as linear precoding, it is classified as nonlinear processing because it performs nonlinear processing called addition of perturbation vectors.

  That is, the modulo arithmetic process is performed on the signal before the precoding process. “Modulo arithmetic processing” is equivalent to adding a perturbation signal vector having a value that is an integer multiple of a specified value to the real part and imaginary part of the real part and imaginary part of the input signal. A signal whose transmission power is suppressed within a specified value by modulo calculation is transmitted by beam forming.

  On the other hand, in the reception processing, the modulo arithmetic processing is performed again on the reception signal in which the above-described perturbation signal is added to the desired signal (information signal including data), the perturbation signal component is removed, and the information signal is It is taken out.

  FIG. 32 is a conceptual diagram for explaining processing of the VP method which is a nonlinear precoding method.

  Referring to FIG. 32, first, in transmitter BS, perturbation vector addition operation (hereinafter referred to as “DPC MOD processing”) based on Dirty Pay Percoding (DPC) is performed on transmitted original signal vector X. And the vector Xmod is calculated as follows. Such a VP method is disclosed in Non-Patent Document 1.

If the beam forming operation is performed on such a vector Xmod by the MMSE method, the transmission signal Z is calculated as follows in the same manner as the simple MMSE method described above.

In the UEs on the receiving side, the signal added with the additive white Gaussian noise (AWGN) is multiplied by √γ, and further DPC and modulo operation (hereinafter abbreviated as “DPC DeMOD processing”) are performed. The signal x _i hat is separated by the following equation:

Note that the operation represented by the following expression represents a process of rounding to a negative integer.

If such a nonlinear precoding method, for example, the VP method is used, the problem that a part of stream data is lost due to the increase of the coefficient γ due to the presence of a highly correlated channel as described above is avoided. Is possible.

  By the way, in a precoding method such as the VP method, it is necessary to obtain channel state information between transmitting and receiving antenna elements on the transmission side.

  For example, in the case of time division duplex (TDD) using the same frequency band for transmission / reception, the terminal station transmits in advance a known signal between the transmission / reception antenna elements to the transmission side, thereby This can be done by an open loop method in which channel state information between transmitting and receiving antenna elements is estimated.

  However, in the case of frequency division duplex (FDD), since the above open loop method cannot be used, it is necessary to feed back channel state information from the terminal station to the base station. Even in the case of TDD, if the calibration function is not provided in each of the base station and the terminal station, the channel state information becomes asymmetric in uplink / downlink. Need feedback.

  At this time, the channel state information (transfer function) is estimated in advance on the terminal station side in order to obtain channel state information on the base station side before transmitting signals to a plurality of transmission partners using the multi-user MIMO transmission technique. Channel state information is fed back to the base station. When the channel state information is quantized (digitized) according to a certain reference and fed back, if the total amount of data to be fed back increases, communication overhead increases and the bandwidth available for communication decreases. When the communication band is occupied by the information to be fed back, the throughput of the entire wireless communication system is reduced.

  In order to reduce the total amount of data required for feedback of channel state information, it is possible to reduce the total amount of data to be fed back by reducing the number of quantization bits, but in this case, the quantization error becomes large and the base station side Transmit beamforming cannot be formed correctly. As a result, there is a problem that the transmission quality at the terminal station deteriorates and the throughput in communication decreases.

  Therefore, it is necessary to provide some restrictions on the feedback of the channel state information.

  In order to reduce the feedback load for limited feedback resources, the MIMO OFDM scheme often uses the correlation of channel state information CSI between adjacent subcarriers.

  In that case, for example, the subcarriers are all divided into several subbands. Here, each subband includes a predetermined number of continuous subcarriers.

  Then, the channel state information CSI of the center subcarrier of each subband is fed back to the base station BS.

  Therefore, the feedback load can be adjusted by the number of subbands.

  Thereafter, the channel state information CSI of the remaining subcarriers is reconstructed using the feedback channel state information CSI.

  Since the frequency domain channel state information CSI is obtained by sampling the transfer function of the MIMO channel, the matrix of the channel state information CSI can be treated as a Laurent polynomial (LP) matrix and re-processed by time domain and frequency domain transform processing. It can be configured (see Non-Patent Document 2).

  Then, for each subcarrier, the transmitter executing the VP algorithm calculates a channel state information CSI related matrix, finds the optimal perturbation vector, and cancels the interference in the transmitted stream in advance. Generate a precoding matrix or spatial filter.

JP 2005-328312 A JP 2007-110664 A JP 2009-177616 A JP 2010-154320 A

Hochwald, BM et al., “A vector-perturbation technique for near-capacity multiantenna multiuser communication-part II: perturbation”, IEEE Transactions on Communications, vol.53, no.3, pp. 537-544, March 2005; Kusume et al. (2007) A. Ancora, C. Bona, and DTM Slock, “Down-sampled impulse response least-squares channel estimation for LTE OFDMA”, in Proc. 2007 IEEE ICASSP, vol. 3, pp. 293-296, 15-20 Apr. 2007.

  However, such a scheme requires a high calculation amount (calculation load) in hardware.

  A simple way to reduce the computational load of the system is to use the correlation of channel state information CSI of adjacent subcarriers. That is, the channel state information CSI of all subcarriers in one subband is assumed to be the same.

  Under such assumptions, a valid precoding matrix or spatial filter need only be calculated once for all subcarriers in this subband.

  This method is called “simple subband mode”.

  However, due to errors in the channel state information CSI between the center and end subcarriers of each subband, this simple subband mode makes a small number of subbands and limited channel state information CSI feedback. Depending on the case, it may cause a loss of spectrum efficiency.

  Therefore, conventionally, in the MIMO communication system, there has been a problem that it is not clear how to achieve good spectrum efficiency while suppressing an increase in communication overhead due to feedback of channel state information. In particular, such a problem becomes prominent in the multi-user MIMO communication system.

  The present invention has been made to solve the above-described problems, and the object thereof is to suppress an increase in communication overhead due to feedback of channel state information in a MIMO communication system, and at the same time, improve It is to provide a wireless communication method, a wireless communication system, and a wireless communication apparatus that realize a high spectrum efficiency.

  According to one aspect of the present invention, an orthogonal frequency division multiplexing scheme using a plurality of subcarriers is provided between a first communication device including a plurality of antenna elements and a second communication device communicating with the first communication device. A wireless communication method for wirelessly communicating a signal modulated in (2) by MIMO (Multiple Input Multiple Output), wherein a second communication device is provided in an antenna element provided in the device itself and in the first communication device. Estimating for each subband channel state information with each of the plurality of antenna elements, wherein the subband includes a predetermined number of adjacent subcarriers of the plurality of subcarriers, The communication device feeds back the estimated channel state information for each subband to the first communication device and transmits it. The first communication device calculates a perturbation vector to be added to the transmission signal from the first communication device based on the channel state information for each subband fed back, the first communication device, A first interpolation process for obtaining channel state information of a predetermined number of subcarriers is performed on the fed back channel state information for each subband, and an antenna is obtained based on the channel state information subjected to the first interpolation process. A step of calculating a weighting coefficient for controlling directivity; and a first communication apparatus determines a second weighting coefficient for the remaining subcarriers for MIMO communication based on the weighting coefficient for a predetermined number of subcarriers. And the first communication device calculates the weighting coefficient calculated by the second interpolation process as a perturbation vector. It multiplied against the added transmission signal, and a step of transmitting a plurality of transmit antenna elements.

  Preferably, in the first interpolation process, the channel state information for each subband fed back is converted from the frequency domain to the time domain, and the channel state information in the time domain is cut with a response up to a predetermined delay amount. And reconverting from the time domain to the frequency domain.

  Preferably, the step of transforming from the frequency domain to the time domain transforms the fed back channel state information for each subband from the frequency domain to the time domain based on the least square error criterion.

  Preferably, in the first interpolation process, the number of channel state information converted from the frequency domain to the time domain is larger than the number of subbands and smaller than the number of subcarriers.

  According to another aspect of the present invention, there is provided a wireless communication system that wirelessly communicates a signal modulated by an orthogonal frequency division multiplexing system using a plurality of subcarriers by MIMO, the first communication apparatus including the first communication apparatus. Includes a plurality of first antenna elements and a feedback information receiving unit that receives channel state information for each subband fed back from a communication partner, and the subband is a predetermined number of adjacent subcarriers. And a control unit that calculates a perturbation vector to be added to the transmission signal and a weighting factor for transmitting the transmission signal from the plurality of antenna elements according to the MIMO scheme. A perturbation vector calculating means for calculating a perturbation vector to be added to the transmission signal based on the information, and feedback A first interpolation process for obtaining channel state information of a predetermined number of subcarriers is performed on the channel state information for each subband, and antenna orientation is performed based on the channel state information subjected to the first interpolation process. A first coefficient calculating means for calculating a weighting coefficient for controlling the characteristics and a weighting coefficient for the remaining number of subcarriers for MIMO communication based on the weighting coefficient for the predetermined number of subcarriers. And a second coefficient calculation means for calculating the transmission coefficient by multiplying the transmission signal to which the perturbation vector is added by the weighting coefficient calculated by the second interpolation processing, further comprising: And the second communication device supports channel state information between the second antenna element, the second antenna element, and each of the plurality of first antenna elements. It includes a channel response estimation unit which estimates for each band, the channel state information for each estimated sub-band, and a channel state information transmission processing unit for transmitting the feedback to the first communication device.

  Preferably, in the first interpolation processing, the first coefficient calculation means converts the fed back channel state information for each subband from the frequency domain to the time domain, and converts the channel state information in the time domain to a predetermined delay. After cutting with a response up to the amount, retransform from the time domain to the frequency domain.

  Preferably, when converting from the frequency domain to the time domain, the first coefficient calculation unit converts the fed back channel state information for each subband from the frequency domain to the time domain based on the least square error criterion. .

  Preferably, in the first interpolation process, the number of channel state information converted from the frequency domain to the time domain is larger than the number of subbands and smaller than the number of subcarriers.

  According to still another aspect of the present invention, there is provided a wireless communication apparatus that wirelessly communicates a signal modulated by an orthogonal frequency division multiplexing system using a plurality of subcarriers using MIMO, the plurality of first antenna elements and a communication partner. A feedback information receiving unit that receives channel state information for each subband fed back, and the subband has a predetermined number of adjacent subcarriers among a plurality of subcarriers and is added to a transmission signal. And a control unit for calculating a weighting factor for transmitting the transmission signal from the plurality of antenna elements by the MIMO method, and the control unit calculates a perturbation vector to be added to the transmission signal based on the fed back channel state information. Perturbation vector calculation means and channel state information for each fed back subband On the other hand, a first interpolation process for obtaining channel state information of a predetermined number of subcarriers is performed, and a weighting coefficient for controlling antenna directivity is calculated based on the channel state information subjected to the first interpolation process. And a second coefficient calculation means for calculating weighting coefficients for the remaining subcarriers for MIMO communication by a second interpolation process based on weighting coefficients for a predetermined number of subcarriers. , The transmission coefficient to which the perturbation vector is added is multiplied by the weighting coefficient calculated by the second interpolation process, and further includes a transmission unit from a plurality of transmission antenna elements.

  Preferably, in the first interpolation processing, the first coefficient calculation means converts the fed back channel state information for each subband from the frequency domain to the time domain, and converts the channel state information in the time domain to a predetermined delay. After cutting with a response up to the amount, retransform from the time domain to the frequency domain.

  Preferably, when converting from the frequency domain to the time domain, the first coefficient calculation unit converts the fed back channel state information for each subband from the frequency domain to the time domain based on the least square error criterion. .

  Preferably, in the first interpolation process, the number of channel state information converted from the frequency domain to the time domain is larger than the number of subbands and smaller than the number of subcarriers.

  According to the present invention, in the MIMO communication system, it is possible to realize good spectrum efficiency while suppressing an increase in communication overhead due to feedback of channel state information.

It is a conceptual diagram for demonstrating the state of communication in the multiuser MIMO of embodiment. It is a block diagram which shows the structure of the radio | wireless transmitter 1000 in the radio | wireless communications system of embodiment. It is a block diagram which shows the structure of the terminal device 1100 in the radio | wireless communications system of embodiment. It is a conceptual diagram which shows the procedure of the inverse Fourier transform of the OFDM symbol in LTE. It is a figure which shows the order of the calculation load for a different matrix calculation. It is a conceptual diagram for explaining perturbation vector search processing by VP method and spatial filter matrix calculation processing for precoding. It is a figure explaining a simulation parameter. It is a figure which shows the channel parameter in simulation. It is a figure which shows the simulated result. It is a figure which shows the simulated result. It is a 1st figure for demonstrating the concept of the linear interpolation scheme of this Embodiment. It is a 2nd figure for demonstrating the concept of the linear interpolation scheme of this Embodiment. It is a figure for demonstrating the estimator which performs a time and frequency domain conversion process. It is a conceptual diagram explaining the calculation procedure of the weighting factor of a spatial filter. It is a flowchart explaining the calculation procedure of the weighting factor of a spatial filter. It is a figure which compares calculation load amount. It is a figure which shows the parameter used for simulation. It is a 1st figure which shows the result of having simulated the spectrum efficiency about the TFT interpolation method in case of _NSB = 50, and the simulation of a simple subband mode. It is a 2nd figure which shows the result of having simulated the spectrum efficiency about the TFT interpolation method in case of _NSB = 50, and the simulation of a simple subband mode. It is a figure for demonstrating allocation of the OFDM subcarrier in IFFT process in the case of LTE-A. Conversely FFT (IFFT) is a diagram illustrating a matrix F ^H for processing. It is a figure for demonstrating the inverse FFT process in a TFT interpolation process. It is a figure for demonstrating the division | segmentation into a subband. It is a figure for demonstrating the feedback channel state information matrix used for a TFT process. It is a diagram for explaining the matrix F _p ^H for use in TFT process. It is a figure for demonstrating the matrix calculation in a TFT process. Of TFT process is a diagram for explaining the process of multiplying the matrix F _L. Is a diagram illustrating the configuration of a matrix F _L used in FIG. It is a figure explaining the channel state information produced | generated by TFT process. It is a conceptual diagram which shows the structure of a multiuser MIMO communication system. It is a conceptual diagram which shows the procedure of the beam formation by MMSE method. It is a conceptual diagram for demonstrating the process of VP method which is a nonlinear precoding method.

  Hereinafter, a radio communication system according to an embodiment of the present invention will be described with reference to the drawings. In the following embodiments, components and processing steps given the same reference numerals are the same or equivalent, and the description thereof will not be repeated unless necessary.

  FIG. 1 is a conceptual diagram for explaining a state of communication in multi-user MIMO according to the embodiment.

  FIG. 1A shows a state in which a base station BS and a plurality of terminal apparatuses UEi are communicating in a relatively wide area.

  When the terminal apparatus UEi is dispersed in a wide area, it is possible to form a beam in which each communication channel is separated.

  On the other hand, FIG. 1B shows a state in which the base station BS and a plurality of terminal devices UEi are communicating in a relatively narrow area such as indoors.

  When the terminal device UEi is present in a narrow area, it is difficult to form a beam in which each communication channel is separated. This corresponds to a high correlation between channels.

  Also, in this situation, in the channel response matrix, a certain terminal may be slightly farther from the base station than the other terminal, compared to the norm (size) of the elements in other rows. Thus, extremely small norm rows (corresponding to antennas on the receiver side) may occur.

  FIG. 2 is a block diagram illustrating a configuration of the wireless transmission device 1000 in the wireless communication system according to the embodiment.

  In FIG. 2, it is assumed that a signal to be transmitted by MIMO is OFDM-modulated.

Then, with respect to the NS streams (NS / 2 users), sub-carriers for OFDM signal transmission is Nc number, transmitting antennas is assumed to be the _{N T.} As will be described later, the number of antennas on each receiver side is generally Nr, but here, it is assumed that there are two antennas as an example.

  Referring to FIG. 2, radio transmitting apparatus 1000 includes encoding section 20 that performs error correction encoding such as forward error correction (FEC) on a digital signal given from input node 10; A serial / parallel converter 22 for serial-parallel conversion of the signal from the encoding unit 20 to form a group of Nc parallel signals, and Nc groups for each parallel-converted signal. Includes modulation units 30-1 to 30-NS for modulating with a predetermined modulation method.

  Such error correction is configured to correct errors up to a predetermined number of bits in the transmitted OFDM frame, as will be described later. Such error correction is performed together with interleaving processing and the like.

  Note that a transmission symbol may include an in-phase component and an orthogonal component, but in FIG. 2, both are represented by a single signal line.

  Also, the modulation units 30-1 to 30-NS control the outputs from the modulation units 30-1 to 30-NS from the control unit 50 in order to perform the above-described nonlinear precoding as precoding processing. The modulo arithmetic processing for adding the perturbation vector is executed.

  In addition to such a modulo operation, DPC (Dirty Paper Coding) can be used as an encoding technique for interference cancellation.

  Based on the channel state information, radio transmitting apparatus 1000 has a control unit 50 for calculating a precoding matrix, and coefficients in the precoding matrix from control unit 50 from modulation units 30-1 to 30-NS. Weighting processing units 40-1 to 40-Nc for multiplying signals. That is, in OFDM modulation, corresponding to the fact that there are Nc subcarriers, the weighting processing units 40-1 to 40-Nc each perform weighting processing on the signal component of one corresponding subcarrier.

  Based on the channel state information (CSI) sent from the receiving side and received by the feedback information receiving unit 80, the control unit 50 calculates a weighting factor for each OFDM subcarrier, as will be described later. The coefficient for the attaching process is calculated. Here, the channel state information (CSI) is fed back for each subband described later.

  The channel state information fed back for each subband may be frequency domain information or time domain information. In the following description, as an example, it is assumed that the information is frequency domain information.

  Further, the storage unit 52 stores values of matrix elements for “time / frequency domain conversion” described later, which are calculated in advance for various parameters, and the control unit 50 Performs the “time / frequency domain conversion process” using the data stored in the storage unit 52.

The weighting processing unit 40-1 multiplies the signal from the modulation unit 30-1 by a coefficient in the precoding matrix to generate a signal to be transmitted from each of the antennas 100-1 to 100- _NT. Multipliers 42-11-42-N _T 1 are included. The weighting processing unit 40-1 includes the same configuration as that corresponding to the modulation unit 30-1, corresponding to the modulation units 30-2 to 30-NS. For example, the weighting processing unit 40-1 multiplies the signal from the modulation unit 30-NS by a coefficient in the precoding matrix and transmits the signals from the antennas 100-1 to 100- _NT , respectively. Includes multipliers 42-1NS-42-N _T NS. Furthermore, the weighting processing unit 40-1 integrates signals from the multipliers 42-11 to 42-1NS corresponding to the antenna 100-1 and generates an adder 44-1 that generates a signal for the antenna 100-1. Including. It is compatible with other antennas 100 - 2 to _{100-N T,} similar adder 44-2 (not shown) includes a ~ _{44-N T.} Here, the other weighting processing units 40-2 to 40-Nc that perform weighting processing on signals from the modulation units 30-1 to 30-NS also have the same configuration as the weighting processing unit 40-1.

Radio transmitting apparatus 1000 further includes up-converters 60-1 to 60- _NT that receive the outputs of weighting processing sections 40-1 to 40-Nc, respectively, and convert them into signals to be transmitted by radio.

The up-converter 60-1 receives the outputs of the weighting processing units 40-1 to 40-Nc and performs an inverse Fourier transform unit 62 for performing an inverse Fourier transform, a DA conversion unit 64 for performing digital-analog conversion, and a DA conversion. Frequency converter 66 for frequency converting the signal of unit 64 based on the signal from local transmitter 70, and power amplifying unit 68 for amplifying the output of frequency converter 66 and supplying it to antenna 100-1. Including. The other up-converters 60-2 to 60- _NT have the same configuration.

  FIG. 3 is a block diagram showing a configuration of terminal apparatus 1100 in the wireless communication system of the embodiment.

  Referring to FIG. 3, terminal device 1100 includes antennas 200-1 and 200-2. The wireless communication system includes four terminals corresponding to the exemplary configuration of the wireless transmission device 1000 of FIG. 2, (total number of transmission antennas) = (number of antennas per terminal) × ( The following description will be made assuming that the relationship of the number of terminals) is established. However, the number of antennas of the terminal device on the receiving side is not limited to such a configuration, and may be one, for example.

  Terminal apparatus 1100 includes down-converters 220-1 and 220-2 for down-converting received signals from antennas 200-1 and 200-2. The down converter 220-1 converts the frequency of the output of the low noise amplification unit 232 for amplifying the reception signal from the antenna 200-1 and the low noise amplification unit 222 by the local transmission signal from the local transmission unit 230. A frequency conversion unit 224 for performing the analog-to-digital conversion on the output of the frequency conversion unit 224, and a Fourier transform unit 228 for OFDM demodulation. The down converter 220-2 has the same configuration.

  The terminal device 1100 further includes a weighting processing unit 240 for receiving signals from the down converters 220-1 and 220-2 and executing weighting processing under the control of the control unit 250. The weighting processing unit 240 adds and combines the signals from the multipliers 242-1 and 242-2 for multiplying the weighting coefficients from the control unit 250 and the multipliers 242-1 and 242-2, respectively, and And adder 244 for selectively separating received signals from channels corresponding to apparatus 1100.

  In the terminal device 1100 side as well, when nonlinear precoding as described above is performed as precoding processing on the transmission side, modulo arithmetic processing is performed on the outputs of the down converters 220-1 and 220-2. The Such modulo arithmetic processing is also executed under the control of the control unit 250. In addition to such a modulo operation, when DPC (Dirty Paper Coding) is used as an encoding technique for interference cancellation, a decoding process for this is also executed.

  The output of the weighting processing unit 240 is demodulated by the demodulating unit 270, then parallel-serial converted by the parallel-serial converting unit 272, error-corrected by the decoder 280, and received as a received signal from the node 300. Is output.

  In the calculation processing of the weighting coefficient in the control unit 250, the channel response matrix (transmission path matrix) estimation result for the terminal device 1100 in the channel response estimation unit 260 is used. Note that, for such channel response matrix estimation processing, processing similar to that disclosed in the above-described prior art can be used.

  Also, the channel response matrix for terminal apparatus 1100 estimated by channel response estimation section 260 is subbanded from antennas 200-1 and 200-2 to radio transmission apparatus 1000 by channel state information transmission processing section 262. Each time it is sent as feedback information.

For channel estimation for subcarriers, for example, a known pilot signal is transmitted on a known subcarrier on the transmitting side and the receiving side. On the receiver side, channel estimation is performed for the subcarrier on which the pilot signal is transmitted. The channel information estimated in this way is fed back to the transmitter as CSI information.
[1. Evaluation of VP-based algorithm and its computational complexity]
In the following, as a premise for evaluating the calculation amount and the spectrum efficiency when the channel state information feedback method and the beamforming method based on the channel state information feedback method in the present embodiment are adopted, first, the algorithm based on the VP method will be discussed later. As is appropriate, the calculation load is evaluated again in the “simple subband mode”. Further, after describing the calculation method in the spatial filter matrix (weighting coefficient matrix) for directivity control in the present embodiment, the calculation load in the present embodiment is compared with the calculation load in the “simple subband mode”. To do.

  The following simulation shows that most of the spectral efficiency loss of the VP algorithm is generated from a non-conforming spatial filter.

  Therefore, in the present embodiment, for the spatial filter of the algorithm of the VP method, interpolation with lower calculation complexity is executed as compared with the calculation of the spatial filter matrix for antenna directivity control.

  The interpolation scheme of this embodiment reconstructs partial subcarrier channel state information CSI from feedback channel state information CSI for each subband by time-frequency domain transformation.

  These partial subcarrier spatial filter matrices are then used to generate a spatial filter matrix for all subcarriers by linear interpolation.

Based on the calculation result of the spectrum efficiency by the analysis of the calculation load and the simulation, the interpolation scheme of this embodiment has a slight increase in the calculation amount, and the spectrum of the algorithm of the VP method compared with the simple subband mode. The efficiency can be greatly improved.
(1.1 Perturbation vector (VP) algorithm)
As a general model for considering the MU-MIMO system, there are K mobile terminals each having Nr receive antennas and a base station BS having Nt (= Nr × K) transmit antennas. Think.

In the following, as a notation method, (.) ^T and (.) ^H indicate a transposed matrix and a conjugate transposed matrix (Hermitian matrix). E {.} And Var {.} Are expected values and distributed operators.

  The corresponding vector equation is:

The Nt × Nt matrix H is composed of complex elements indicating the channel response of the MIMO channel.

In the equation (1), it is assumed that E {nn ^* } = σ ² I _Nt and E {| x | ² } = 1 is assumed as the power condition.

I _Nt is an Nt × Nt identity matrix.

  A channel is estimated based on the MMSE, and an MMSE precoding matrix or an MMSE spatial filter is calculated.

  First, assuming that the base station BS side has the complete channel state information CSI H, the MMSE filter W is set as follows, and the base station BS sets the transmission signal s as follows: To do.

Here, α = Ntσ ² , and σ ² is a signal-to-noise ratio (SNR) value of the downlink receiving antenna, and is fed back from the receiving side.

  The transmission symbol is a standardized transmission symbol u, and finally the symbol u as shown below is transmitted to all users on the MU-MIMO channel H.

However, if the channel response H is a highly correlated channel matrix, the normalization parameter γ is greatly increased. All data X is lost.

  In order to alleviate the result, as described above, in the VP method, the transmission symbol X is perturbed by a data vector.

  In the simplest case, when τ is a positive real number and l is an Nt-dimensional complex perturbation vector (a + ib), the perturbed transmission vector is set as Xmod = X + τl. Here, a and b are integers.

The signal power normalization coefficient (scalar) γ _{1 in} the process of searching for the perturbation vector is calculated as follows.

here,

It is.

  Further, P is defined as follows, and Expression (4) is simply expressed.

From equations (2) and (4), if ρ = 1, the MMSE filter W (weight coefficient matrix) for antenna directivity control is the same as the matrix P.

  However, as described in Non-Patent Document 1 described above, when the value of ρ is less than 1, the VP algorithm can achieve better performance.

  Therefore, in the following embodiment, although not particularly limited, for example, when Nt = 8, ρ = 0.125 is set.

  The transmission signal after the perturbation vector search is as follows.

The transmission symbol u is transmitted to the mobile terminal via the MU-MIMO channel H and is expressed as follows.

On the receiving side, the scalar τ is selected so that the following modulo operation is possible.

By using modulo arithmetic, the receiver can demodulate the original data.

The central point of the design of the VP method is to optimally select the perturbation vector l so as to minimize the scalar γ ₁ .

This is a least square problem of an integer lattice of Nt dimensions that can be solved by using a so-called “sphere decoder” described in Document 1 below.

Reference 1: B. Hassibi and H. Vikalo, “On sphere decoding algorithm. I. expected complexity”, IEEE Trans. On Signal Processing, vol. 53, no. 8, pp. 2806-2818, Aug. 2005.
However, in general, since the computational load of the sphere decoder is large, an efficient sphere decoder algorithm is able to trade off between the computational load and the degree of perturbation vector diversity. This is realized by using E).

  Such “QR decomposition M-algorithm encoder (QRDM-E)” is disclosed in Document 2 below.

Reference 2: Manar Mohaisen, Bing Hui, KyungHi Chang, Seunghwan Ji, and Jinsoup Joung, “Fixed-complexity vector perturbation with block diagonalization for MU-MIMO systems”, Proc. 2009 IEEE MICC, pp.238-243, 15-17 , Dec. 2009
Describing the outline of QRDM-E, the integer values a and b of the perturbation vector l are limited to a symmetric integer set A. Here, A = {− T, −T + 1,... −1, 0, 1, T−1, T}. By adopting such a set A, the range is limited to a smaller value T, and the search complexity is reduced.

  On the other hand, QRDM-E decomposes matrix P in equation (8) into the product of unitary matrix Q and upper triangular matrix R, so the search problem in equation (8) is as shown in equation (9) below. To be simplified.

For each iteration that performs equation (9), the best branch with the least cumulative metric is selected for the next level at each encoding level.

  In the following description, although not particularly limited, the MU-MIMO of the OFDM modulation scheme in LTE will be considered as an example.

  FIG. 4 is a conceptual diagram showing the procedure of inverse Fourier transform of an OFDM symbol in such LTE.

  As shown in FIG. 4, the OFDM symbol in each frame includes as many samples as the number of signal transmission subcarriers. Each sample in FIG. 4 is, for example, a signal subjected to BPSK modulation, QPSK modulation, QAM modulation, or the like.

In the example of FIG. 4, samples [X ¹ ₁ , X ² ₁ , X ³ ₁ , X ⁴ ₁ ] for a plurality of frames respectively corresponding to a plurality of streams correspond to the first subcarrier. The same applies to other subcarriers.

The above-described channel response matrix H _k for each subcarrier (1 ≦ k ≦ Nc) corresponds to a channel response matrix when a transmission path from the transmitting antenna to the receiving antenna is assumed for each subcarrier. To do. Nc is the number of signal transmission subcarriers.

On the other hand, in LTE, the number of all subcarriers N _FFT (FFT size: including all subcarriers of (guardband + signal transmission + DC)) is 1024 or 2048. In the description, a case where N _FFT = 2048 is taken as an example.
(1.2 Calculation load of VP algorithm)
In the following, in order to estimate the calculation load of the algorithm of the VP method, the order of the sum of matrix product operation and sum operation is used.

  FIG. 5 is a diagram illustrating an order of calculation load for different matrix operations.

  Therefore, using FIG. 5, the order Cvp of the calculation load of the VP algorithm for one subcarrier is roughly as follows.

Here, C _MMSE includes a product operation of two (Nt × Nt) complex matrices and an inverse matrix operation of (Nt × Nt) complex matrices as shown in Equation (2), and 12N _t Equal to ³ .
(1.3 MU-MIMO performance of VP method with feedback of limited channel state information)
(1.3-1 MU-MIMO OFDM subband)
For an OFDM scheme with N _C-number of subcarriers for data transmission, N _C-number of subcarriers are divided into N _SB subbands, each subband is continuous N _SC = N _C / N _SB Suppose that the subcarrier is included.

  Here, it is assumed that only the channel state information CSI of the center subcarrier of each subband is fed back to the transmitter.

Therefore, when _NSB is reduced, the feedback load is reduced, but the quality of the feedback channel state information CSI is degraded.

When the OFDM scheme has N _C subcarriers for data transmission, the total computational load for one OFDM symbol increases to NcCvp for systems using the VP method.
(Reference method 1: Calculation load in simple subband mode)
In the simple subband mode (hereinafter referred to as reference scheme 1), it is assumed that the frequency domain channel state information CSI of all subcarriers of each subband is the same as that of the center subcarrier of this subband.

  Therefore, the total computational load for the simple subband mode is calculated as follows:

The computational load can be reduced by keeping _NSB at a small value.

However, the error of the channel state information CSI between the center subcarrier and the end subcarrier of the subband increases when _NSB is reduced, and causes a loss of spectrum efficiency.
(1.3-2 Influence of nonconforming spatial filter on performance of MIMO OFDM by VP method)
FIG. 6 is a conceptual diagram for explaining perturbation vector search processing by the VP method and spatial filter matrix calculation processing for precoding.

  The VP algorithm has two main calculation processes, a perturbation vector search process and a spatial filter calculation process.

Without accurate channel state information CSI, the smallest normalization coefficient γ ₁ of equation (4) cannot be obtained in the perturbation vector search process, so an inefficient perturbation vector is selected.

  Furthermore, a non-conforming spatial filter cannot suppress interference between streams.

  Therefore, in the following, as a premise for explaining the “perturbation vector search process and spatial filter calculation process” of the present embodiment, the “incompatible spatial filter” for the performance of the MU-MIMO OFDM system based on the VP method is described. Assess the impact.

  As shown in FIG. 6, MMSE spatial filters are configured for all subcarriers for each subband, and two schemes for searching for perturbation vectors are compared by simulation.

  Reference scheme 1 is a simple subband mode, as described above, for the same MMSE spatial filter and perturbation vector search process for all subcarriers in the subband, with the same subcarrier in the center of each subband. Channel state information CSI.

  On the other hand, in the reference method 2, in the calculation process of the MMSE spatial filter, complete channel state information is fed back for all subcarriers, and this is used, but each subband is used for the perturbation vector search process. For all of the subcarriers, the same channel state information CSI of the subcarrier at the center of the subband is used.

  FIG. 7 is a diagram illustrating simulation parameters.

  FIG. 8 is a diagram showing channel parameters in the simulation.

  As shown in FIG. 8, two types of channel models were chosen to evaluate the impact of the integrity of channel state information.

  Such channel parameters are the same as those disclosed in Document 3 below.

Reference 3: WINNER II D1.1.2, “WINNER II channel models,” https://www.ist-winner.org/deliverables.html, Sep. 2007.
9 and 10 are diagrams showing simulated results.

  FIG. 9 employs A1 LOS as a channel model and is simulated by 64QAM, and FIG. 10 employs C1 NLOS as a channel model and is simulated by 64QAM.

  9 and 10, the number Nc of signal transmission subcarriers is assumed to be 1200.

Also simulated by changing the number N _SB subbands.

Therefore, if N _SB = 1200, the number of subcarriers N _SC for each subband is 1, and in this case, the channel state information is fed back for each subband. Indicates that the complete channel state information (for all subcarriers) is being fed back.

“N _SB = 1200, N _SC = 1, VP” means that the complete channel state information is fed back in both the perturbation vector search process in the VP method and the spatial filter for antenna directivity control. Indicates what is happening.

“N _SB = 1200, N _SC = 1, MMSE” means that non-linear MIMO processing is not performed, complete channel state information is fed back, and only a spatial filter calculation for antenna directivity control by the MMSE method is performed. Indicates what is happening.

“N _SB = 50, N _SC = 24, VP Simple subband mode” means that channel state information for one subcarrier at the center of each of 50 subbands (24 subcarriers per subband). In both the perturbation vector search process in the VP method and the spatial filter for antenna directivity control, the calculation is performed assuming that the subcarriers in each subband are the same as the central channel state information. Indicates that

“N _SB = 50, N _SC = 24, VP Perfect MMSE filter” means that in the perturbation vector search process in the VP method, each of 50 subbands (24 subcarriers per subband) The channel state information for one subcarrier is fed back, and the subcarriers in each subband are calculated as being the same as the central channel state information, while the spatial filter for antenna directivity control is performed. This calculation processing indicates that calculation is performed on the assumption that complete channel state information is fed back for all subcarriers in each subband.

  From the simulated results of FIGS. 9 and 10, most of the performance loss of the VP algorithm algorithm is due to the incomplete “feedback” channel state information of the spatial filter for antenna directivity control. It can be seen that it is generated from "calculation".

  For example, consider the case of a C1 NLOS channel model having an average 234 [ns] delay spread as shown in FIG.

If the MMSE spatial filter uses complete information for all subcarriers in a subband, it is assumed that only using the channel state information CSI of the central subcarrier to search for the perturbation vector, N _{Even when SC} = 48, that is, when the number of subbands is small, a large-scale performance loss is not generated.

However, if there is no perfect MMSE spatial filter, even if N _SC = 24, that is, if the number of subbands is large, there will be a significant performance loss.

The reason for this is that the inefficient perturbation vector only generates a large value of the normalization factor γ ₁ (equation (4)) that reduces the SNR value of each stream. In contrast, an incompatible MMSE spatial filter causes interference between streams and greatly degrades system performance.
(2. Interpolation scheme of VP MU-MIMO OFDM system of this embodiment)
An effective method for reducing the spectrum efficiency loss for the VP MU-MIMO OFDM scheme is to generate a complete MMSE spatial filter for each subcarrier, as described in FIG. 9 and FIG. is there.

  Therefore, in the present embodiment, a simple interpolation scheme for generating a more appropriate MMSE spatial filter with a small calculation load is adopted.

  11 and 12 are diagrams for explaining the concept of the interpolation scheme of the present embodiment.

Referring to FIG. 11, the interpolation scheme of the spatial filter for antenna directivity control in the present embodiment is based on feedback channel state information CSI (N) by time and frequency domain transformation (TFT transformation). _{From fb} (= N _SB )), channel state information CSI of N _TFT subcarriers at equal intervals is reconstructed by interpolation processing described later. This is called “reconstruction interpolation processing”.

In reconstruction interpolation processing, by adjusting the value of N _TFT, it is possible to trade off the amount of calculation interpolation performance. Furthermore, from the channel state information of the N _TFT subcarriers reconstructed interpolated by the weight calculation of the spatial filter, to calculate the N _TFT pieces of weighting coefficients.

Furthermore, by performing interpolation processing, for example, linear interpolation processing, on N _TFT weight coefficients, the weight coefficients of spatial filters for all subcarriers (N _all ) are calculated.

As shown in FIG. 12, in the reconstruction interpolation process, as described later, from the channel state information H _fb in the frequency domain that has been fed back, using the least square error criterion, Reconfigure channel state information. In the channel state information of the reconstructed partial subcarrier, the channel response is cut at a component up to a predetermined response time (sampling time is set as L).

  Here, L is assumed to use a value larger than the maximum value of the path delay.

Then, the channel state information of the cut time domain, again, is converted into the channel state information H _TFT in the frequency domain.

Here, a matrix used to reconstruct partial subcarrier time domain channel state information from the fed back frequency domain channel state information H _fb using the least square error criterion. the matrix F _L that are used to convert the F _P and partial re frequency domain channel state information in the time domain subcarrier, to be described later.

Thereafter, as described in FIG. 11, using the spatial filter matrix (weight coefficient matrix) calculated using the channel state information for the N _TFT partial subcarriers thus obtained, The spatial filter matrix for all subcarriers is calculated by linear interpolation. This is called “linear interpolation processing”. Here, as an example, the calculation of the spatial filter matrix is performed by MMSE. However, the interpolation processing here is not necessarily limited to linear interpolation, and other interpolation methods may be used.

On the other hand, in the process of searching for the perturbation vector in the VP method, as in the simple subband mode, the channel state information of the center subcarrier of each subband is the other channel state information in the subband. Are also assumed to have the same value, search processing is executed. More specifically, when QR decomposition is performed using QRDM-E as described above, the channel state information of the subcarriers of each subband includes the sub-center at the center of each subband. The perturbation vector is searched using the matrix R calculated as being the same as the channel state information of the carrier.
(2.1 Interpolation method for time / frequency domain transformation for MMSE spatial filter)
A linear interpolation method with time-frequency domain transform has been used for channel estimation by a pilot signal as an estimator based on a normalized least square error criterion. Such channel estimation is disclosed in Document 4 below, for example.

Reference 4: A. Ancora, C. Bona, and DTM Slock, “Down-sampled impulse response least-squares channel estimation for LTE OFDMA”, in Proc. 2007 IEEE ICASSP, vol. 3, pp. 293-296, 15- 20 Apr. 2007.
Here, this estimator is used as an interpolation function for reconstructing the channel state information CSI of N _TFT subcarriers.

  FIG. 13 is a diagram for explaining an estimator that performs such time / frequency domain conversion processing.

  The estimator is expressed as:

Here, the following values are the frequency domain channel state information CSI fed back between the mth transmitting antenna and the nth receiving antenna.

Also, the following values are the reconfigured frequency domain channel state information CSI between the mth transmit antenna and the nth receive antenna.

L is the effective channel length.

In Expression (12), the matrix _FP is a row corresponding to the subcarrier position to be fed back from the (2048 × 2048) FFT matrix used in the orthogonal frequency division multiplexing method, taking the case of LTE as an example. And (N _SB × L) matrix obtained by selecting the first L columns of the FFT matrix. N _SB is equal to the number of subbands, that is, the number N _fb of channel state information to be fed back.

Where α ₂ is Lσ ² and σ ² is the SNR value of the downlink receive antenna.

In the formula (12), the matrix F _L from among the (2048 × 2048) FFT matrix, obtained from N _TFT rows and the first L columns corresponding to the position of the subcarrier interval constant N _TFT subcarriers (N _TFT × L) matrix.

The matrix I _P is an (L × L) identity matrix.

The M _TFT of the (N _TFT × N _SB ) matrix can be calculated in advance for several different values of σ ² and the selection parameter L and stored in the storage unit 52.

Using equation (12), a matrix H _TFT (i) (i = 0,... N _TFT −1) of N _TFT types of channel state information CSI is generated for all transmission antennas and reception antennas. Later, the base station BS calculates the weight W _TFT (i) (i = 0,... N _TFT −1) of the MMSE spatial filter as follows:

Finally, the base station BS calculates MMSE spatial filter weights for all subcarriers by linear interpolation as follows:

FIG. 14 is a conceptual diagram for explaining the procedure for calculating the weighting factor of such a spatial filter, and FIG. 15 is a flowchart for explaining the procedure for calculating the weighting factor of the spatial filter.

  Referring to FIGS. 14 and 15, first, in mobile terminal 1100 on the receiving side, channel response estimation section 260 calculates a channel response in the frequency domain of one subcarrier for one subband (S100). The channel state information (CSI) transmission processing unit 262 of the mobile terminal 1100 feeds back the calculated “frequency domain channel response” for each subband to the base station 1000 on the transmission side (S102).

  In the base station 1000, the control unit 50 performs perturbation vector calculation processing in the VP method based on the fed back “frequency domain channel response” (S104).

Furthermore, the control unit 50 of the base station 1000 performs the TFT process based on the information of the matrix M _TFT stored in the storage unit 52 (TFT process in FIG. 14). Specifically, the control unit 50 first converts the “frequency-domain channel response” received as feedback information into the time domain, which corresponds to IFFT and is initially processed by the square error criterion (S106). Subsequently, the control unit 50 cuts the response up to a predetermined delay amount (greater than or equal to the maximum delay amount) out of the “time-domain channel response obtained by conversion”, and again performs frequency domain processing by processing corresponding to FFT. (S108).

  Subsequently, the control unit 50 performs MMSE weight calculation processing by spatial filter calculation processing, for example, “frequency domain channel response” subjected to TFT processing (S110, MMSE weight calculation in FIG. 14).

  Further, the control unit 50 linearly interpolates the calculated MMSE weights to calculate the MMSE weights for all subcarriers (S114, linear interpolation in FIG. 14).

Next, in the base station 1000, the weighting processing units 40-1 to 40-Nc multiply the transmission signal vector after the VP processing by the MMSE weight and transmit from the multiple antennas 100-1 to 100- _NT (S116). .
(2.2 Computation amount of interpolation method of this embodiment provided with TFT processing)
In the TFT processing (Equation (12)) and the linear interpolation method (Equation (14)), the order of their calculation amounts is the calculation load of O [4N _TFT N _SB N _t ² ] and O [2N _TFT N _t ² ]. Real number multiplication and addition operations.

   Compared with equation (11), the order of calculation load for the interpolation method of the present embodiment for performing TFT processing is calculated as follows:

FIG. 16 is a diagram for comparing calculation load amounts.

In FIG. 16, the TFT interpolation method as described above and the calculation load of the simple subband mode are compared with N _SB = 25 and N _SB = 50, respectively.

  In the figure, the display with VP means that the perturbation vector is added by the VP method and then the spatial filter processing is performed, and the display with MMSE indicates that the spatial filter processing is performed without adding the perturbation vector. Means that

  As shown in FIG. 16, the simple subband mode has the least computational load compared to applying the TFT interpolation method to both the MMSE and VP algorithm.

The calculation load of TFT interpolation increases by increasing the value of N _TFT .

When N _TFT = 1200, the computational load of the algorithm using the VP method is about 436% for N _SB = 25 and 575 for N _SB = 50 when compared to the simple subband mode. %, Greatly increasing.

However, for N _TFT = 100 and N _TFT = 200, the computational load of the algorithm using the VP method is about 22% for N _SB = 25 compared to the simple subband mode. And only 56% and 40% and 80% for N _SB = 50, respectively.

Therefore, when using N _TFT = 100 and _NTFT = 200, compared to that of the simple subband mode, the computational load of TFT interpolation is suppressed to an increase within an acceptable range.
(2.3 Spectrum efficiency for TFT interpolation)
In the following, simulation results for spectrum efficiency when the TFT interpolation method is used will be described.

  FIG. 17 is a diagram illustrating parameters used in the simulation. The parameters other than those in FIG. 17 are the same as those in FIG.

  Modulation and coding type. Type 1 means 16QAM, code rate is 7/8, and codeword length is 5397 bits. Type 2 means 64QAM, code rate is 3/4, codeword. The length is 4197 bits. Type 3 is 64QAM, the code rate is 7/8, and the codeword length is 6297 bits.

FIGS. 18 and 19 are diagrams showing the results of simulating the spectrum efficiency for the TFT interpolation method and the simple subband mode simulation when N _SB = 50, respectively.

FIG. 18 shows the results of simulating the spectrum efficiency for the TFT interpolation method and simple subband mode simulation for N _SB = 50 on the WINNER II A1 LOS channel.

FIG. 19 shows the results of simulating spectral efficiency for TFT interpolation and simple subband mode simulation on the C1 NLOS channel with N _SB = 50.

  The value of L in equation (12) is set to 40.

In addition, to compare the spectral efficiency loss of both methods, we also simulated the algorithm of the VP method where N _SB = 1200, which corresponds to each subcarrier getting complete channel state information CSI.

  As shown in FIGS. 18 and 19, the TFT interpolation method greatly improves the spectrum effect of the simple subband mode, especially for 64QAM with a coding rate of 7/8.

By comparing with the optimum channel state information CSI for each subcarrier, the TFT interpolation method for N _TFT = 100 and N _TFT = 200 can achieve almost maximum spectral efficiency performance.

On the other hand, the spectral efficiency difference between _NTFT = 100 and _NTFT = 1200 is small for Type 1 and Type 2.

Therefore, under the condition of N _TFT = 100, the TFT interpolation method can reduce the loss of spectrum efficiency while suppressing an increase in calculation amount.

For Type 3, the TFT interpolation method when N _TFT = 200 uses the complete channel state information CSI of each subcarrier, and can greatly reduce the difference in spectrum efficiency.

Therefore, compared with the simple subband mode, the TFT interpolation method can significantly improve the spectrum efficiency in the MIMO spatial filter operation algorithm using the VP method with a slight increase in the amount of calculation. .
(Specific example of TFT treatment)
Hereinafter, the TFT process will be described in more detail by taking the case of LTE-A as an example.

  FIG. 20 is a diagram for explaining OFDM subcarrier allocation in IFFT processing in LTE-A.

  There are a total of 2048 subcarriers.

  Subcarriers 1025-1448 and subcarriers 602-1024 are used as guard bands.

  Subcarrier 1 is a DC subcarrier and is not used.

  A total of 1200 subcarriers of subcarrier 2-601 and subcarriers 1049-2048 are used for data transmission.

FIG. 21 is a diagram for explaining a matrix F ^H for inverse FFT (IFFT) processing.

First, the element in the m-th row and the n-th column of the matrix F ^H is expressed as follows using ω.

Here, ω is expressed as follows.

Therefore, the matrix F ^H is written as follows.

Further, the data vector X that is used for the inverse FFT processing is expressed as follows.

As described above, among the elements of the data vector X, the values of the elements x (602) to x (1448) corresponding to the subcarriers 602 to 1448 corresponding to the guard band are set to 0.

  FIG. 22 is a diagram for explaining the inverse FFT process in the TFT interpolation process.

  As described above, it is assumed that data is 0 for the guard band.

  Therefore, in the data vector X, the data vector X1 is obtained by deleting the element corresponding to the guard band.

On the other hand, in the inverse FFT matrix F ^H , the matrix F ^H ₁ is obtained by deleting the elements corresponding to the guard band.

The matrix F ^H ₁ and the data vector X1 are written down as follows.

In other words, in the actual TFT processing calculation, calculation is not performed for unused subcarriers. However, in the following, for the specific convenience of the position of the row and column of the inverse FFT matrix, it will be described how which row and column are used in the TFT operation in the inverse FFT matrix F ^H. .

  FIG. 23 is a diagram for explaining division into subbands.

  For example, 50 subbands are set for 1200 subcarriers for data transmission. 24 subcarriers are included per subband.

  For example, the first subband SB1 includes subcarriers SC2 to SC25. Similarly, subcarriers are assigned to other subbands in the same manner, and the last subband B50 includes subcarriers SC2025 to SC2048.

  Subbands SB1 to SB25 correspond to the first 600 subcarriers, and subbands SB26 to SB50 correspond to the second 600 subcarriers.

  Then, the channel state information CSI of the subcarrier at the center of each subband is fed back to the transmitter side.

  For example, for subband SB1, channel state information H (14) (estimated channel response) for subcarrier SC14 is fed back. The same applies to the other subbands.

  Each channel state information H (k) (14, 38,..., 2037) is an (Nt × Nt) matrix indicating an estimated channel response for a corresponding subcarrier in the frequency domain.

  FIG. 24 is a diagram for explaining a feedback channel state information matrix used for TFT processing.

In FIG. 24 and subsequent figures, it is assumed that N _TFT = 100.

  Extracting the fed back frequency domain channel state information between the m th transmitting antenna and the n th receiving antenna from the fed back channel state information H (k) (14, 38,..., 2037). Thus, the following feedback channel state information matrix is constructed.

FIG. 25 is a diagram for explaining a matrix F _p ^H used in TFT processing.

That is, the matrix F _p ^H is a matrix that is calculated from the fed back Nfb (= 50) channel state information, and is a sub-channel of the fed back channel state information from the (2048 × 2048) inverse FFT matrix. (L × Nfb) matrix in which elements from the first row to the Lth row corresponding to the time to cut off the response are extracted from the columns corresponding to the carriers (SC14, SC38,..., SC590, SC1461, SC1485, SC2037). It is.

In the case where the time-domain channel state information is fed back, there is no particular limitation, but only the time-domain channel state information up to the cutoff time is fed back, and this matrix F _P ^H is multiplied. Can be omitted.

  FIG. 26 is a diagram for explaining matrix calculation in TFT processing.

The process of multiplying the frequency domain channel state information fed back between the m th transmit antenna and the n th receive antenna by the matrix F _p ^H is an inverse FFT operation and corresponds to the transform from the frequency domain to the time domain To do. After the multiplication process of the matrix F _p ^H , the filter process is performed. Finally, the multiplication process of the matrix F _L corresponding to the FFT process is performed, and the channel state information becomes a frequency domain component again.

  The filtering process corresponds to the least square error criterion, and has an effect of removing the influence from the unused area corresponding to the unmodulated subcarrier.

27, of the TFT process is a diagram for explaining a process of multiplying the matrix F _L.

N _TFT / 2 frequency domain channel state information is generated corresponding to the subcarriers SC1 to SC601 excluding the guard band. This N _TFT / 2 frequency domain channel state information is all equally spaced.

  The subcarrier number k to be generated is as follows.

k = 8 + 12 × (i−1) (i = 1,..., N _TFT / 2)
Similarly, N _TFT / 2 frequency domain channel state information is generated corresponding to subcarriers S1449 to SC2048. This N _TFT / 2 frequency domain channel state information is also equally spaced.

  The subcarrier number k to be generated is as follows.

k = 1455 + 12 × (i−1) (i = 1,..., N _TFT / 2)
Figure 28 is a diagram illustrating the configuration of a matrix F _L used in FIG.

Matrix F _L executes FFT processing, it is necessary to generate channel state information of the N _TFT frequency-domain as described in FIG. 27.

Therefore, as an example, for the case of N _TFT = 100, it is as follows expose to write down.

FIG. 29 is a diagram for explaining channel state information generated by TFT processing.

From the 50 frequency domain channel state information fed back by the TFT processing, 100 frequency domain channel state information H _TFT (7 + 12 × i) and H _TFT (7 + 12 × k) (i, k = 0,... 49) is generated.

  In the above description, FFT is performed in a (2048 × 2048) matrix, 1200 subcarriers for data transmission are used, and frequency domain channel state information to be fed back is a 50 matrix. Although the description has been given assuming that the frequency domain channel state information generated by the configuration interpolation is 100 matrices, the beamforming process of the present embodiment is not limited to such a case, and other parameters It can also be applied to values.

  As described above, according to the radio communication system, radio transmission apparatus, and radio communication method of the present embodiment, in the MIMO communication system, while suppressing an increase in communication overhead due to feedback of channel state information, at the same time, good Spectrum efficiency can be realized.

  Embodiment disclosed this time is an illustration of the structure for implementing this invention concretely, Comprising: The technical scope of this invention is not restrict | limited. The technical scope of the present invention is shown not by the description of the embodiment but by the scope of the claims, and includes modifications within the wording and equivalent meanings of the scope of the claims. Is intended.

  10 input nodes, 20 serial / parallel conversion units, 30-1 to 30-8 modulation units, 40-1 to 40-8 weighting processing units, 50 control units, 52 storage units, 60-1 to 60-8 upconverters, 80 Feedback information receiving unit, 100-1 to 100-Nt antenna, CSI channel state information, 1000 wireless transmission device, 1100 terminal device.

Claims (12)

A signal modulated by orthogonal frequency division multiplexing using a plurality of subcarriers is transmitted between a first communication device including a plurality of antenna elements and a second communication device communicating with the first communication device. A wireless communication method for performing wireless communication using Multiple Input Multiple Output),
The second communication apparatus estimates, for each subband, channel state information between an antenna element provided in the own apparatus and each of the plurality of antenna elements provided in the first communication apparatus. The subband includes a predetermined number of adjacent subcarriers of the plurality of subcarriers;
The second communication device feeds back the estimated channel state information for each subband to the first communication device; and
The first communication device calculates a perturbation vector to be added to a transmission signal from the first communication device based on the channel state information for each subband fed back;
The first communication apparatus performs a first interpolation process for obtaining channel state information of a predetermined number of subcarriers on the channel state information for each subband fed back, and the first interpolation Calculating a weighting factor for controlling antenna directivity based on the processed channel state information;
The first communication device calculates a weighting coefficient for the remaining subcarriers for MIMO communication by a second interpolation process based on the weighting coefficient for the predetermined number of subcarriers;
The first communication device multiplies the transmission signal to which the perturbation vector is added by the weighting coefficient calculated by the second interpolation processing, and transmits the multiplication signal from the plurality of transmission antenna elements. A wireless communication method. The first interpolation process includes:
Transforming the fed back channel state information for each subband from frequency domain to time domain;
The wireless communication method according to claim 1, further comprising: reconverting the time domain channel state information from a time domain to a frequency domain after cutting the channel state information with a response up to a predetermined delay amount.   The wireless communication according to claim 2, wherein the step of transforming from the frequency domain to the time domain transforms the fed back channel state information for each subband from the frequency domain to the time domain based on a least square error criterion. Method.   3. The radio communication method according to claim 2, wherein, in the first interpolation processing, the number of the channel state information converted from the frequency domain to the time domain is larger than the number of the subbands and smaller than the number of the subcarriers. . A wireless communication system that wirelessly communicates a signal modulated by orthogonal frequency division multiplexing with a plurality of subcarriers using MIMO,
Comprising a first communication device;
The first communication device is:
A plurality of first antenna elements;
A feedback information receiving unit that receives the channel state information for each subband fed back from a communication partner, and the subband has a predetermined number of adjacent subcarriers among the plurality of subcarriers;
A control unit that calculates a perturbation vector to be added to the transmission signal and a weighting factor for transmitting the transmission signal from the plurality of antenna elements in a MIMO scheme;
The controller is
Perturbation vector calculation means for calculating the perturbation vector to be added to the transmission signal based on the channel state information fed back;
A first interpolation process for obtaining channel state information of a predetermined number of subcarriers is performed on the fed back channel state information for each subband, and based on the channel state information subjected to the first interpolation process. First coefficient calculating means for calculating a weighting coefficient for controlling the antenna directivity;
Second coefficient calculation means for calculating weighting coefficients for the remaining subcarriers for MIMO communication by a second interpolation process based on the weighting coefficients for the predetermined number of subcarriers;
Multiplying the transmission signal added with the perturbation vector by the weighting coefficient calculated by the second interpolation processing, further including a transmission unit from the plurality of transmission antenna elements,
The second communication device is:
A second antenna element;
A channel response estimator for estimating channel state information between the second antenna element and each of the plurality of first antenna elements for each subband;
A wireless communication system, comprising: a channel state information transmission processing unit that feeds back and transmits the estimated channel state information for each subband to the first communication device.   In the first interpolation process, the first coefficient calculation means converts the fed back channel state information for each subband from a frequency domain to a time domain, and converts the channel state information in the time domain to a predetermined value. The wireless communication system according to claim 5, wherein the wireless communication system re-converts from the time domain to the frequency domain after being cut with a response up to the delay amount.   When converting from the frequency domain to the time domain, the first coefficient calculating unit converts the fed back channel state information for each subband from the frequency domain to the time domain based on a least square error criterion. The wireless communication system according to claim 6.   The radio communication system according to claim 6, wherein, in the first interpolation process, the number of the channel state information converted from the frequency domain to the time domain is larger than the number of subbands and smaller than the number of subcarriers. . A wireless communication apparatus that wirelessly communicates a signal modulated by orthogonal frequency division multiplexing with a plurality of subcarriers using MIMO,
A plurality of first antenna elements;
A feedback information receiving unit that receives the channel state information for each subband fed back from a communication partner, and the subband has a predetermined number of adjacent subcarriers among the plurality of subcarriers;
A control unit that calculates a perturbation vector to be added to the transmission signal and a weighting factor for transmitting the transmission signal from the plurality of antenna elements in a MIMO scheme;
The controller is
Perturbation vector calculation means for calculating the perturbation vector to be added to the transmission signal based on the channel state information fed back;
A first interpolation process for obtaining channel state information of a predetermined number of subcarriers is performed on the fed back channel state information for each subband, and based on the channel state information subjected to the first interpolation process. First coefficient calculating means for calculating a weighting coefficient for controlling the antenna directivity;
Second coefficient calculation means for calculating a weighting coefficient for the remaining subcarriers for MIMO communication based on the weighting coefficient for the predetermined number of subcarriers by a second interpolation process;
A wireless communication apparatus, further comprising: a transmission unit that multiplies the transmission signal to which the perturbation vector is added by the weighting coefficient calculated by the second interpolation process, and a plurality of transmission antenna elements.   In the first interpolation process, the first coefficient calculation means converts the fed back channel state information for each subband from a frequency domain to a time domain, and converts the channel state information in the time domain to a predetermined value. The wireless communication apparatus according to claim 9, wherein the wireless communication apparatus performs re-conversion from the time domain to the frequency domain after being cut with a response up to a delay amount.   When converting from the frequency domain to the time domain, the first coefficient calculating unit converts the fed back channel state information for each subband from the frequency domain to the time domain based on a least square error criterion. The wireless communication apparatus according to claim 10.   The radio communication apparatus according to claim 10, wherein, in the first interpolation processing, the number of the channel state information converted from the frequency domain to the time domain is larger than the number of subbands and smaller than the number of subcarriers. .