JP2011139294A - Transmitter and receiver - Google Patents

Abstract

<P>PROBLEM TO BE SOLVED: To provide a technology for exchanging knowledge of interference between blocks to combine energy distributed in the past or future. <P>SOLUTION: A first equalization part 42a equalizes an input signal by a first block of a predetermined period. A second equalization part 42b equalizes the signal by a second block of the past. A third equalization part 42c equalizes the signal by a third block of the future. A first turbo decoding part 44a feeds back a logarithmic likelihood ratio acquired by decoding an equalization result in the first equalization part 42a, as a first logarithmic likelihood ratio. A second turbo decoding part 44b feeds back a logarithmic likelihood ratio acquired by decoding an equalization result in the second equalization part 42b, as a second logarithmic likelihood ratio. A third turbo decoding part 44c feeds back a logarithmic likelihood ratio acquired by decoding an equalization result in the third equalization part 42c, as a third logarithmic likelihood ratio. <P>COPYRIGHT: (C)2011,JPO&INPIT

Description

  The present invention relates to communication technology, and more particularly to a transmission device and a reception device that execute block transmission.

  Block transmission systems that use guard intervals (GI), such as single carrier block transmission (SC-CP) with cyclic prefix (CP) and orthogonal frequency division multiplexing (OFDM) with CP, have frequency selectivity. It is a powerful method against fading. In recent years, single carrier frequency division multiple access (SC-FDMA), which is a mixture of these, has been proposed as a candidate for a radio access scheme in next-generation radio communication. A common premise for these block transmission systems is that the length of the GI is sufficiently long compared to the length of the channel impulse response (CIR).

  However, if the GI is shorter than the CIR length, the system is affected by inter-block interference (IBI) and inter-symbol interference (ISI), so even if the signal-to-noise ratio (SNR) is sufficiently high. An error floor occurs in the bit error rate (BER). In order to reduce the interference caused by the insufficient length of the GI, several techniques have been proposed as in the case of not transmitting the CP. By using turbo equalization, better characteristics can be obtained than when turbo equalization is not used (see, for example, Non-Patent Documents 1 to 5).

S. Suyama et al., “An OFDM receiver further turbo equalization for multipath environments with delay difference greater than the guard interval”, IEEE VTC2003-Spring, Jan. 2005, 1, p. 632-636 K. Hayashi et al., “Interference cancellation scheme for single carrier block transmission with insufficient cyclic prefix”, EURASHIP Journal on Wireless Communications and Networking, Oct. 2007, 2008, p. 1-12 K. Shimezawa et al., “A novel sc / mmse turbo equalization for multicarrier systems with insufficient cyclic prefix”, IEEE PIMRC2008, Sept. 2008, p. 1-5 D. Wang et al., “Low complexity turbo equalization for single carrier systems without cyclic prefix”, IEEE ICCS 2008, Nov. 2008, p. 1091-1095 Z. Chen et al., “A turbo fde technique for ofdm system without cyclic prefix”, IEEE VTC2009-Fall, Sept. 2009, p. 1-5

  In order to suppress a decrease in spectrum use efficiency, it is desirable not to transmit GI or CP. On the other hand, even when GI and CP are not transmitted, it is desired to suppress deterioration of characteristics. Furthermore, it is also desired that the complexity of the device configuration is at a level that can be realized. Under such circumstances, the present inventor has come to recognize the following problems. In order to obtain good characteristics while limiting the device configuration to a realizable complexity, it is necessary to exchange knowledge of interference between blocks and combine energy dispersed in the past and future.

  The present invention has been made in view of such circumstances, and an object thereof is to provide a technique for exchanging knowledge of interference between blocks and combining energy dispersed in the past and the future.

  In order to solve the above-described problem, a receiving apparatus according to an aspect of the present invention includes an input unit that sequentially inputs encoded signals, and a first block that equalizes a signal input at the input unit in a first block for a predetermined period. A first equalization unit, a second equalization unit that equalizes the signal in the second block that is past the first block, and a third equalization unit that equalizes the signal in the third block that is future than the first block Then, the log likelihood ratio obtained by decoding the equalization result in the first equalization unit is returned to the first equalization unit, the second equalization unit, and the third equalization unit as the first log likelihood ratio. The log likelihood ratio acquired by decoding the equalization result in the first decoding unit and the second equalization unit is fed back to the first equalization unit and the second equalization unit as the second log likelihood ratio. The log likelihood ratio acquired by decoding the equalization result in the second decoding unit and the third equalization unit is set as the third log likelihood ratio. 1 equalizer, and a third decryption unit for feeding back to the third equalizer, and an output unit for outputting the decoded result in the first decoding portion. The first equalization unit reflects the first log likelihood ratio, the second log likelihood ratio, and the third log likelihood ratio at the time of equalization, and the second equalization unit at the time of equalization, The first log likelihood ratio and the second log likelihood ratio are reflected, and the third equalization unit reflects the first log likelihood ratio and the third log likelihood ratio at the time of equalization.

  Another aspect of the present invention is a transmission device. The apparatus includes a generation unit that sequentially generates a block signal formed by a combination of a first partial block including data and a second partial block, and a transmission unit that transmits the block signal sequentially generated by the generation unit. The generation unit is operable in a first mode in which the guard interval is included in the second partial block and a second mode in which an error correction code is included in the second partial block instead of the guard interval. A control unit for selecting a mode.

  It should be noted that any combination of the above-described constituent elements and a conversion of the expression of the present invention between a method, an apparatus, a system, a recording medium, a computer program, etc. are also effective as an aspect of the present invention.

  According to the present invention, knowledge of interference can be exchanged between blocks, and energy dispersed in the past and the future can be combined.

It is a figure which shows the structure of the communication system which concerns on the Example of this invention. It is a figure which shows the structure of the block influenced by the frequency selective fading channel in the communication system of FIG. It is a figure which shows the concept of CHARUE in the communication system of FIG. It is a figure which shows the detail of LLR exchange of the posterior probability in the communication system of FIG. It is a figure which shows the detail of FD / SC-MMSE for the CHARUTE equalization part of FIG. FIGS. 6A to 6D are diagrams showing strict covariance matrices X of the present, past, and future in the CHARUE equalization unit of FIG. 1 (I _{a, c} = I _{a, p} = I _{a, f} = 0.5). FIGS. 7A to 7D are diagrams showing the current, past, and future covariance matrix X ⁻¹ in the CHARUE equalization unit of FIG. 1 (I _{a, c} = I _{a, p} = I _{a, f} = 0.5). It is a figure which shows an EXIT analysis in case of SNR = 2dB. It is a figure which shows EXIT analysis in case of SNR = 4dB. It is a figure which shows the parameter of the simulation with respect to the communication system of FIG. It is a figure which shows the BER characteristic when using exact | strict X- ¹ matrix. It is a figure which shows the BER characteristic when using the approximated X- ¹ matrix. It is a figure which shows the structure of the receiver of FIG. It is a figure which shows the structure of the 1st equalization part of FIG. 13, and a 1st turbo decoding part. It is a figure which shows the structure of the transmitter of FIG. FIGS. 16A and 16B are diagrams illustrating signals transmitted from the transmission apparatus in FIG.

  Before describing the present invention in detail, an outline will be described. If GI or CP is transmitted, the spectrum utilization efficiency decreases, but the embodiment of the present invention does not transmit GI or CP in order to improve the decrease in spectrum utilization efficiency. Furthermore, in order to reduce the complexity of the device configuration, it is desirable that the channel matrix in the block is a cyclic configuration. For this purpose, an irregular permutation matrix J is introduced. By using the matrix J, frequency domain signal processing becomes possible. As a result, it is possible to use turbo equalization technology, frequency domain soft cancellation technology, and least square error (FD / SC-MMSE). They are known as the most effective equalization techniques in terms of characteristics and complexity. Here, CHARUE (CHAined Turbo Equalization) which is a new frequency domain turbo equalization technique is proposed. This takes advantage of the frequency domain calculation.

  Furthermore, received signals of several adjacent blocks in the time domain, for example past blocks, current blocks, future blocks are stored. The equalizers corresponding to the respective received signals exchange the posterior information of the symbols that cause the IBI. The posterior information is obtained from the decoder and is used to soft cancel the IBI component of the adjacent block. However, MMSE filtering is required to further suppress the residual IBI component after soft cancellation. By using the construction of the residual IBI covariance matrix in the frequency domain, the complexity is reduced. Further, this embodiment proposes an approximation method. This approximation method reduces the computational complexity of the algorithm.

In the following, a scalar is indicated by a normal notation such as a. The vector is shown with a subscript “v” added, such as “a _V” , or in a normal notation, such as “a”. The matrix is shown in uppercase letters like A. Past blocks and future blocks relative to the current block are denoted as a ′, a ″. The operator diag (a) represents a diagonal matrix having vector elements as diagonal components. Estimation result variable, as a _E, shown is added subscript E. Hermitian matrix A is represented as A ^H. The identity matrix is denoted as I, and unless otherwise indicated, the size of I is assumed to be K × K. The matrix transpose operator is denoted as ^AT . E [a] indicates the expected value of a. Unless otherwise indicated, L _a indicates _a prior LLR and L _p indicates a posterior LLR.

1. System Model This embodiment assumes a single-input single-output (SISO) antenna system. However, this embodiment can be easily extended to a multiple-input multiple-output (MIMO) configuration. FIG. 1 shows a configuration of a communication system 100 according to an embodiment of the present invention. The communication system 100 includes a transmission device 10 and a reception device 20. The transmission device 10 includes an encoding unit 12, an interleaving unit 14, and a modulation unit 16, and the receiving device 20 includes a CHARUE equalization unit 22, a deinterleaving unit 24, a decoding unit 26, an adding unit 28, and an interleaving unit 30. In the transmission apparatus 10, the encoding unit 12 encodes the bits to be transmitted, and the interleaving unit 14 performs random interleaving on the encoded bits. Symbol s _V modulated by the modulator 16 is shown as follows.

This is shown as a block. K is the length of the block, and s _V ′, s _V , and s _V ″ indicate the past block, the current block, and the future block, respectively. Here, it is assumed that these are transmitted from the same transmitter 10, but when transmitted from different transmitters 10, this corresponds to time division multiple access (TDMA). These blocks are transmitted via a frequency selective fading channel H having frequency selectivity for the multipath transmission line. It is assumed that the fading channel gain is constant over one block period and varies from block to block. H ′, H, and H ″ correspond to the past block channel, the current block channel, and the future block channel.

When CP is added on the transmission side and CP is removed on the reception side, the channel matrix H becomes cyclic. However, in this embodiment, since CP transmission is not assumed, the channel matrix for the current block has a Toeplitz configuration as follows.
The channel matrix for the interference component from the past block is shown as follows.
The channel matrix for the interference components from future blocks is shown as follows:
The receiving device 20 includes a deinterleaving unit 24, a decoding unit 26, and an adding unit 28.

FIG. 2 shows a configuration of a block affected by a frequency selective fading channel in the communication system 100. The fading channels H _p , H _c , and H _f are assumed to have distributed Rayleigh fading channel characteristics that are independent of each other while the average power is equal. They also vary from block to block. Since the CP is not transmitted, the observed current block is affected by fading H _c and cannot avoid IBI components from previous blocks affected by fading channel H _p ′. This is shown as interference 200 from past blocks in FIG. Furthermore, if the sampling continues to the end of the impulse response for the last symbol of the current block, another IBI component from a future block affected by H _f ′ cannot be avoided. This is shown in FIG. 2 as interference 202 from future blocks. In FIG. 2, the current block length 204 corresponds to a period for detecting a block. The current block length 204 is set as K + L. Returning to FIG.

In the receiving device 20, the received signal y of the current block is shown as follows.
Here, s _V , s _V ′, and s _V ″ denote a current block, a past block, and a future block, respectively, and n denotes a zero-mean complex added white Gaussian noise vector. Also, the covariance E {nn ^H } = σ ² I is established. σ ² is the covariance of noise, and is expressed as follows in decibel (dB) notation by the SNR specified for each antenna.

2. Outline of CHARUE equalization unit Introducing the J matrix The J matrix is introduced in order to convert the channel matrix H, which has become a Toeplitz matrix because no GI has been inserted, into a cyclic matrix. Due to this cyclic nature, the latest version of the FD / SC-MMSE algorithm is used. The J matrix functions as an insertion such as a CP as follows in order to guarantee the cyclicity of the channel.

By applying the J matrix to the received signal y, the current received block is shown as follows:
Note that the matrix of past and future interference components is not a Toeplitz matrix as shown in Equations (3) and (4), and therefore does not have cyclicity even if a J matrix is introduced. Meanwhile, the current channel H _c being the main component, has a cyclicity.

B. Removal of IBI The received signal is subjected to strong interference from the past block JH _p ′ S _V ′ and the future block JH _f ″ S _V ″, as shown in equation (8). Here, the removal of IBI will be described. The interference is caused by the last L-1 symbol of the past block and the first L-1 symbol of the future block. They are shown as follows:
Here, K is the length of the block, and L is the length of the CIR.

By using soft symbol estimates of past S _VE ′ and future S _VE ″, interference is reduced. The soft replica of the received signal shown in the equation (8) is shown as follows.
Here, the k-th soft symbol estimation value of S _VE ′ is expressed as follows.
In addition, the k-th soft symbol estimation value of S _VE ″ is expressed as follows.
Here, L _a ′ and L _a ″ indicate prior LLRs of past and future blocks. They are derived by interleaving the posterior LLR output from the decoder. Here, S _{VE 'to} compute, the external LLR from its own decoder is used, S _VE' to calculate the S _{VE ''} and, the posterior LLR from past and future decoder used. As a result, good characteristics and quick convergence are obtained. Therefore, the internal iteration is performed independently at each equalizer, and the final result of the internal iteration is used between the equalizers.

FIG. 3 shows the concept of CHARUE in the communication system 100. This illustrates the principle of ex-post LLR exchange between blocks. The equalizer and decoder (E & D2) correspond to the current block, E & D1 corresponds to the past block, and E & D3 corresponds to the future block. E & D2 receives L _{p, D1} ′, L _{p, D3} ″ from D1 and D3 of the past and future blocks, and at the same time provides L _{p, D2} ′ of the posterior LLR to the future block. Provide postmortem LLR L _{p, D2} ″.

FIG. 4 shows details of the LLR exchange of posterior probabilities in the communication system 100. This shows FIG. 3 in more detail, and shows in more detail the exchange of post-event information from other decoders and external information from external decoders. Pre LLR is _{L a, E2} 'and _{L a, E2'} 'is used to remove the IBI, _{L a} is a pre LLR obtained from the external LLR by its decoder, normal FD / SC- Similar to MMSE, it is used to reduce intersymbol interference within a block. As shown in FIG. 4, the external LLR is obtained after subtracting the prior information from E2.

By obtaining the soft replica of equation (11), interference from past blocks and future blocks is removed, and the kth symbol is estimated as follows.
Here, h _c (k) represents a column vector of the current channel matrix H _{c in} the equation (2), and r _Vr = r _V −r _VE represents a residual error of the soft replica with respect to the received signal r _V.

C. Modification of FD / SC-MMSE Algorithm Since noise is not estimated in equation (11), equation (14) includes noise. Such noise can potentially increase errors. Furthermore, processing for reducing errors is required. Here, the MMSE algorithm is applied as follows to reduce errors.
In addition, the following equation is solved.
The weight of MMSE filtering is shown as follows.

Since the interleaver performs randomization sufficiently, it is reasonable to assume that the soft symbols S _VE ′, S _VE , S _VE ″ have no correlation, and as a result, the soft of the ISI component remaining in the current block. The level covariance matrix is diagonalized as follows.
Assuming BPSK modulation, equations (18a) to (18c) are expressed as follows.
Using estimation from equation (14), the output z (k) of FD / SC-MMSE is shown as follows:

When the equation (20) is transformed into a block expression, the following relationship is shown.
Here, the following relationship is shown.
This is an expression by a block of γ (k) in the equation (21). The following relationship is also shown.
Since equalization is done in the frequency domain, the beneficial properties of the circulant matrix JH _c are shown as follows.

This is used to reduce the computational complexity of the equalization process. This is because Φ = FJH _c F ^H becomes a diagonal matrix by simple processing. As a result, Σ in the equation (17) is expressed as follows.
This is shown in the frequency domain as:
Here, the following relationship is shown.
The final output z _V of the CHARUE equalizer is shown as follows:

Here, together with Φ in the equation (25), Γ in the equation (23) is expressed in the frequency domain as follows.
The following relationship is also shown.
The output z _V is expressed as follows.
Moreover, the following relationship is shown.
Here, S _V ^* is a complex conjugate of S _V. For BPSK modulation, [| S _V | ² ] = 1, and ν is an equivalent noise vector with dispersion as follows.
The conversion from CHARUE output to external LLR is shown as follows.
Here, R (z) is a real component of the complex number z. FIG. 5 shows details of the FD / SC-MMSE for the CHARUE equalization unit 22. This corresponds to an example of the operation outline of the CHARUE algorithm.

3. Reduction of Computational Complexity As shown in equation (29), the computational complexity of a CHARUTE equalizer is based on the computation of inverse matrices Σ ⁻¹ and (I + ΓS) ⁻¹ . These are based on the inverse matrix Σ ⁻¹ as shown in equation (30). Therefore, complexity can be reduced by performing approximation using the following relationship.
Here, it is assumed that the FFT size of F is sufficiently large, for example, K >> 16.

A. Approximate 1
The first approximation is performed on the estimated symbols for soft cancellation of the current block as follows:
In the case of BPSK, the calculation of Λ in equation (27) can be reduced, so that

B. Approximation 2
According to equation (28), an approximation of Σ ⁻¹ is obtained by an approximation of X ⁻² . Furthermore, the complexity of X- ¹ depends on the structure of X itself. If X is a diagonalization matrix, the amount of computation of X ⁻¹ can be ignored. From the equation (38) and the matrix components shown in FIGS. 6A to 6D, it is appropriate to approximate X to a diagonal matrix as follows. Here, FIGS. 6A to 6D show the current, past, and future strict covariance matrices X in the CHARUE equalization unit 22 (I _{a, c} = I _{a, p} = I _{a, f} = 0.5).

Approximated component _{X A,} as shown in FIG. 7 (a) - shown in (d). FIGS. 7A to 7D show the covariance matrix X ⁻¹ approximated by diagonalization of the current, past, and future in the CHARUE equalization unit 22 (I _{a, c} = I _{a, p} = I _{a, f} = 0.5). These have only a diagonal component. Since the off-diagonal component is replaced with zero and the diagonal component has an equivalent value, the calculation of the inverse matrix is simplified.

C. Approximate 3
Finally, the approximation is made as follows.

4). EXIT chart analysis In order to evaluate the characteristics of the CHARUE equalizer and observe the convergence characteristics, the EXIT chart is analyzed here. The receiving device 20 is mainly composed of a HATUE equalization unit 22 and a decoding unit 26. CHATUE equalizer 22 has _{L a} is a three pre LLR ', _{L a'} ', the input of _{L a,} the _{L e,} an output of the _E is one of the external LLR. It is shown as follows.
If SNR is fixed, it is shown as follows.
Here, I _{e, E} indicates mutual information between s (k) and the external LLR corresponding to the equalizer output. Also, I _{a, E} , I _p ′, I _p ″ correspond to the mutual information calculated from the external LLR of the current block and the mutual information calculated from the posterior LLRs of the past and future blocks.

The decoder has one input, _{La, D,} and one output, _{Le, D.} It does not depend on the SNR because it does not connect directly to the channel. The EXIT curve, which is the transfer function for the decoder, is shown as follows:
The external LLR is obtained as follows.
This is shown in FIG. However, the EXIT curve of the decoder may be shown as follows:

FIG. 8 shows the EXIT analysis when SNR = 2 dB, and FIG. 9 shows the EXIT analysis when SNR = 4 dB. Since there are three LLR inputs and one LLR output, the EXIT function of the equalizer is four-dimensional (4D). In order to simplify the notation, the upper and lower limits of the EXIT curve are shown here. The lower limit is calculated as follows by setting I _{a, E} ′ = I _{a, E} ″ = 0.
The upper limit of the EXIT curve is calculated as follows by setting I _{a, E} ′ = I _{a, E} ″ = 1.

The decoder's EXIT curve is calculated by measuring the histogram of the decoder's output LLR. The output LLR of the decoder is obtained by using BCJR (Bahl-Cocke-Jelinek-Raviv). The convolutional encoder is assumed to have a limit length of 3 and a generator polynomial G = [75]. The locus of decoding is plotted by exchanging mutual information obtained from the equations (43) and (46). Here, the repetition index for a fixed SNR is n. For n = 0, the iteration starts with I _{a, E, 0} = 0 _, which is zero prior knowledge. When the iteration is n, the external mutual information at the equalizer output is denoted as I _{e, E, n} = T _EQ (I _{a, E, n} ). Then, in order to obtain I _{e, D, n} = TD (I _{a, D, n} ), I _{a, D, n} = I _{e, E, n} . For the next iteration n + 1, I _{a, E, n + 1} = I _{e, D, n} . Finally, the decoding trajectory is plotted after performing several iterations.

  As is apparent from the trajectories of FIGS. 8 and 9, the adjacent LLRs raise the EXIT curve to avoid the intersection at point A and shift the intersection to point B. By such processing, the characteristics of the equalizer are improved. The improvement shown by the upper EXIT curve as shown in FIG. 9 is greater at SNR = 4 dB. Therefore, the BER performance is improved by exchanging interference knowledge between adjacent blocks as well as increasing the SNR.

5. Characteristic Evaluation A computer simulation is performed to clarify the characteristics of the CHARUE equalizer. FIG. 10 shows simulation parameters for the communication system 100. Single carrier block transmission (SCBT) is assumed. Further, CP is not added, and information is subjected to BPSK (Binary Phase Shift Keying) modulation. The block length is 256. The frequency selective fading channel has 64 paths of equal average power. Further, the electric power varies from block to block while being constant in the block. All interleavers are 256 in length and random.

  The encoder to consider is the same as the encoder in EXIT analysis. It is a non-recursive system convolutional code (NRSC) with a limit length of 3 and a generator polynomial of G = [75]. Since no punching is performed, the resulting coding rate is R = 1/2. In order to simulate many channels, the fading channel is assumed to be constant within a block but varies between blocks. The CIR length is 64 paths and has a uniform average power. Channel estimation and synchronization are assumed to be complete upon reception.

  A block diagram of a chained simulation performed to evaluate the characteristics of a CHARUE equalizer is shown in FIG. Each CHARUTE equalizer is coupled to an equalizer corresponding to a past or future block. The mutual information (MI) provided to block i + 1 by the decoder for block (i + 2) should be considered in the same way as the mutual information from block (i-2) to (i-1). To avoid the unrealizable complexity due to the chain configuration, the following assumptions are invoked. The block length is large and the channel variation is sufficiently random from block to block. Therefore, the posterior mutual information between each symbol of the block (i-2) and the corresponding posterior LLR is the same as between each symbol of the block (i + 1) and the corresponding posterior LLR of the decoder feedback. It is. This assumption is reasonable because the mutual information already means average.

In conventional FD / SC-MMSE, complexity is reduced by performing the approximation in Fairamudafai ^H of equation (40). Therefore, here, the approximation of Fairamudafai ^H is always assumed. In the case of the CHARUE algorithm, FJH _f ′ ^H Λ ″ ^H _p ′ ^H J ^H F ^H , FJH _f ″ ^H Λ ″ ^H _f ″ ^H J ^H F ^H , Fσ ² JJ ^H F ^H The characteristics are examined in detail.

FIG. 11 shows the BER characteristics when using an exact X- ¹ matrix. The repetition improves the BER characteristics until it concentrates at a position about 1 dB away from the lower limit of BER 10 ^-4 after three repetitions. The lower limit is the BER curve for block transmission on the AWGN channel when using the same encoding / decoding parameters. On the other hand, the upper limit is the theoretical BER curve of an uncoded system on a Rayleigh fading channel.

FIG. 12 shows the BER characteristics when using an approximate X- ¹ matrix. Finally, as with the reduced complexity version, the CHARUE algorithm exhibits good characteristics in block transmission systems that do not add GI / CP. The result of the approximation is the same as the result when strict matrix calculation is performed for past and future interference. This means that the approximate version is sufficiently practical.

6). Configuration of Receiving Device Here, a specific configuration of the receiving device 20 including the CHARUE algorithm described so far will be described. FIG. 13 shows the configuration of the receiving device 20. The receiving device 20 includes a delay unit 40, an equalization unit 42, and a turbo decoding unit 44. The delay unit 40 includes a first delay unit 40a and a second delay unit 40b, and the equalization unit 42 includes a first equalization unit 42a, a second equalization unit 42b, and a third equalization unit 42c. The unit 44 includes a first turbo decoding unit 44a, a second turbo decoding unit 44b, and a third turbo decoding unit 44c. Note that the first equalization unit 42a, the second equalization unit 42b, and the third equalization unit 42c may be collectively referred to as the equalization unit 42, and the first turbo decoding unit 44a, the second turbo decoding unit 44b, The third turbo decoding unit 44c may be collectively referred to as the turbo decoding unit 44.

  The delay unit 40 inputs signals sequentially. The signal is transmitted from a transmission device 10 (not shown) and is encoded. Here, turbo coding is the object of description as coding. The signal is a baseband digital signal. The delay unit 40 outputs the input signal to the third equalization unit 42c. The first delay unit 40a delays the input signal for a predetermined period, and outputs the delayed signal to the first equalization unit 42a and the second delay unit 40b. Here, the predetermined period corresponds to K in FIG. 2, that is, a period corresponding to one block.

  The second delay unit 40b delays the signal from the first delay unit 40a over a predetermined period, and outputs the delayed signal to the second equalization unit 42b. The predetermined period in the second delay unit 40b is the same as the predetermined period in the first delay unit 40a. The signal input to the third equalization unit 42c, the signal input to the first equalization unit 42a, and the signal input to the second equalization unit 42b are signals delayed by a period corresponding to the block. That is, the delay unit 40 generates a signal corresponding to the future block, a signal corresponding to the current block, and a signal corresponding to the past block.

  The first equalizer 42a corresponds to the CHARUE equalizer 22 in FIG. 1, E2 in FIG. 4, and FIG. 5, and is an equalizer corresponding to the current block. That is, the first equalization unit 42a equalizes the signal input from the first delay unit 40a with the current block. The period of the current block is K, but the first equalization unit 42a performs processing on the current block length 204 shown in FIG. The processing of the first equalization unit 42a is based on the CHARUE algorithm described so far, and a specific configuration will be described later. As described above, the first equalization unit 42a performs prior LLR from the first turbo decoding unit 44a (hereinafter referred to as “first LLR”) and prior LLR from the second turbo decoding unit 44b during equalization. (Hereinafter referred to as “second LLR”), the prior LLR (hereinafter referred to as “third LLR”) from the third turbo decoding unit 44c is reflected. The first equalization unit 42a outputs the equalization result, that is, the estimated symbol, to the first turbo decoding unit 44a.

  The second equalizer 42b corresponds to E1 in FIG. 4 and is an equalizer corresponding to a past block. The second equalization unit 42b equalizes the signal input from the second delay unit 40b with a past block. The second equalization unit 42b performs the same process as the first equalization unit 42a, but the processing target blocks are different. Further, unlike the first equalization unit 42a, the second equalization unit 42b does not consider the influence from the past block, and thus the corresponding processing is omitted. Therefore, the second equalization unit 42b reflects the first LLR and the second LLR at the time of equalization. The second equalization unit 42b outputs the equalization result, that is, the estimated symbol, to the second turbo decoding unit 44b.

  The third equalizer 42c corresponds to E3 in FIG. 4 and is an equalizer corresponding to a future block. The third equalization unit 42c equalizes the signal input from the delay unit 40 with a future block. The third equalization unit 42c also performs the same process as the first equalization unit 42a, but the processing target blocks are different. Further, unlike the first equalization unit 42a, the third equalization unit 42c does not consider the influence from the future block, and thus the corresponding processing is omitted. Therefore, the third equalization unit 42c reflects the first LLR and the second LLR at the time of equalization. The third equalization unit 42c outputs the equalization result, that is, the estimated symbol, to the third turbo decoding unit 44c.

  The first turbo decoding unit 44a corresponds to the combination of the deinterleaving unit 24, the decoding unit 26, the adding unit 28, and the interleaving unit 30 in FIG. 1, D2 in FIG. 4, and FIG. The first turbo decoding unit 44a decodes the equalization result in the first equalization unit 42a. Decoding corresponds to turbo coding. Also, the first turbo decoding unit 44a feeds back the first LLR acquired in the decoding to the first equalizing unit 42a, the second equalizing unit 42b, and the third equalizing unit 42c. The first turbo decoding unit 44a outputs a decoding result.

  The second turbo decoding unit 44b corresponds to D1 in FIG. The second turbo decoding unit 44b decodes the equalization result in the second equalization unit 42b. In addition, the second turbo decoding unit 44b feeds back the second LLR acquired by decoding to the first equalization unit 42a and the second equalization unit 42b. The third turbo decoding unit 44c corresponds to D3 in FIG. The third turbo decoding unit 44c decodes the equalization result in the third equalization unit 42c. Further, the third turbo decoding unit 44c feeds back the third LLR obtained by decoding to the first equalization unit 42a and the third equalization unit 42c. Here, the 2nd turbo decoding part 44b and the 3rd turbo decoding part 44c are comprised similarly to the 1st turbo decoding part 44a.

  FIG. 14 shows the configuration of the first equalization unit 42a and the first turbo decoding unit 44a. The first equalization unit 42a includes an extraction unit 50, an estimation unit 52, a filter unit 54, a generation unit 56, a holding unit 58, an acquisition unit 60, a derivation unit 62, a conversion unit 64, and a first conversion unit 64a, The generation unit 56 includes a first conversion unit 66a, a second multiplication unit 66b, a third multiplication unit 66c, and a multiplication unit 68, which are collectively referred to as a multiplication unit 66. A first multiplier 68a, a second multiplier 68b, a third multiplier 68c, and an adder 70 are included. The first turbo decoding unit 44 a includes a deinterleaving unit 72, a decoding unit 74, an adding unit 76, and an interleaving unit 78.

The acquisition unit 60 acquires a first transmission path matrix corresponding to the current block, a second transmission path matrix corresponding to the past block, and a third transmission path matrix corresponding to the future block. These correspond to the aforementioned H _c , H _p ′, and H _f ″, respectively. In addition, since a well-known technique should just be used for estimation of a 1st transmission line matrix, a 2nd transmission line matrix, and a 3rd transmission line matrix, description is abbreviate | omitted here. The holding unit 58 holds a permutation matrix for converting the first transmission path matrix into a cyclic matrix. The permutation matrix corresponds to J described above.

The extraction unit 50 extracts a signal included in the current block by calculating a permutation matrix on the signal input from the first delay unit 40a (not shown). The processing of the extraction unit 50 corresponds to the above-described equation (8). Here, the input signal corresponds to y, the extracted signal corresponds to r _V. The first converter 64a acquires the first LLR, the second converter 64b acquires the second LLR, and the third converter 64c acquires the third LLR. The conversion unit 64 estimates the symbol by performing the tanh operation after dividing the LLR by 2. The processing in the conversion unit 64 corresponds to the above-described equations (12) and (13). The conversion unit 64 outputs the estimated symbol to the multiplication unit 66.

  The multiplication unit 66 multiplies the symbol from the conversion unit 64 by the transmission path matrix from the acquisition unit 60, and the multiplication unit 68 multiplies the multiplication result by the multiplication unit 66 and the permutation matrix from the holding unit 58. . The processing of the multiplication unit 66 and the multiplication unit 68 is performed for each block, and the addition unit 70 generates a replica by adding the multiplication results of all the blocks. The replica generated in the adding unit 70 corresponds to the above-described equation (11). Multiplication unit 66 outputs the replica to estimation unit 52.

The estimation unit 52 acquires the signal extracted by the extraction unit 50, the replica generated by the addition unit 70, the first transmission path matrix from the acquisition unit 60, and the symbol estimated by the first conversion unit 64a. Based on the extracted signal, replica, first transmission line matrix, and estimated symbol, the estimation unit 52 estimates a signal in which the influence of signals included in past blocks and future blocks is reduced. The process in the estimation part 52 respond | corresponds to (14) Formula. The signal estimated by the estimation unit 52 corresponds to s _VE in the equation (14). The estimation unit 52 outputs the estimated signal to the filter unit 54.

  The deriving unit 62 acquires the permutation matrix from the holding unit 58, the first transmission path matrix, the second transmission path matrix, the third transmission path matrix, and the symbols estimated by the conversion unit 64 from the acquisition unit 60. The deriving unit 62 derives a tap coefficient for equalization by calculation in the frequency domain based on the permutation matrix, the first transmission line matrix, the second transmission line matrix, the third transmission line matrix, and the estimated symbols. . The tap coefficient corresponds to the aforementioned weight. The processing in the derivation unit 62 corresponds to the above-described equation (17) and equations (18a) to (18c). Further, the derivation unit 62 approximates the term including the second transmission line matrix and the permutation matrix and the term including the third transmission line matrix and the permutation matrix by diagonalization in the frequency domain calculation. May be. This approximation processing corresponds to the above-described equations (37) to (41).

  The filter unit 54 filters the signal estimated by the estimation unit 52 with the tap coefficient derived by the derivation unit 62. The processing in the filter unit 54 corresponds to equations (20) to (31). The filter unit 54 outputs the filtered result to the deinterleave unit 72 as an equalization result. The deinterleaver 72 outputs the deinterleave result to the decoder 74 and the adder 76. The decoding unit 74 decodes the deinterleave result, outputs the decoded result to the outside, and outputs it to the adding unit 76. The adding unit 76 derives the posterior LLR by subtracting the deinterleaving result in the deinterleaving unit 72 from the decoding result in the decoding unit 74. The interleaving unit 78 derives an a priori LLR by performing an interleaving process on the a posteriori LLR. The interleaving unit 78 outputs the prior LLR to the first conversion unit 64a.

  This configuration can be realized in terms of hardware by a CPU, memory, or other LSI of any computer, and in terms of software, it can be realized by a program loaded in the memory, but here it is realized by their cooperation. Draw functional blocks. Accordingly, those skilled in the art will understand that these functional blocks can be realized in various forms by hardware only, software only, or a combination thereof.

7). Configuration of Transmitting Device Here, a specific configuration of the transmitting device 10 shown in FIG. 1 will be described. 1 performs block transmission without adding a GI. The transmitting apparatus 10 described here also has a function of adding a GI, and switches between adding a GI and not adding a GI. In addition, the repetition cycle of blocks and symbols is the same regardless of whether or not GI is added. Therefore, processing in the transmission device 10 and the reception device is simplified.

  FIG. 15 shows the configuration of the transmission apparatus 10. The transmission device 10 includes a generation unit 80, a transmission unit 82, and a control unit 84. The generation unit 80 includes a first encoding unit 86, a modulation unit 88, a GI insertion unit 90, a selection unit 92, a second encoding unit 94, and a modulation unit 96.

  The first encoding unit 86 performs encoding on the input data. The encoding may be any of turbo encoding, block encoding, and convolutional encoding. Here, it is assumed that turbo coding is used. The first encoding unit 86 outputs the encoded signal to the modulation unit 88. Modulation section 88 modulates the encoded signal and outputs the modulated signal to GI insertion section 90. The GI insertion unit 90 inserts a GI into the signal modulated by the modulation unit 88. FIGS. 16A and 16B show signals transmitted from the transmission apparatus 10. FIG. 16A shows a signal in which the GI insertion unit 90 has inserted the GI. As illustrated, the block signal is formed by a combination of a first partial block including data and a second partial block. Moreover, the combination is repeated sequentially. Here, a guard interval is inserted in the second partial block. Returning to FIG.

  The second encoding unit 94 performs encoding on the input data. Here, it is assumed that the encoding is turbo encoding. It is assumed that the redundancy of turbo encoding in the second encoding unit 94 is higher than the redundancy of turbo encoding in the first encoding unit 86. That is, the encoding in the second encoding unit 94 is stronger than the encoding in the first encoding unit 86. The second encoding unit 94 outputs the encoded signal to the modulation unit 96. FIG. 16B shows a signal modulated by the modulation unit 96. Similar to FIG. 16A, the block signal is formed by a combination of a first partial block including data and a second partial block. Moreover, the combination is repeated sequentially. The combination period, the first partial block period, and the second block period are the same as those in FIG. Here, the second partial block includes the error correction code generated in the second encoding unit 94 instead of the guard interval. Returning to FIG.

  The control unit 84 gives an instruction to select one of the first mode for operating the GI insertion unit 90 from the first encoding unit 86 and the second mode for operating the second encoding unit 94 and the modulation unit 96. Accept from outside. The control unit 84 executes the first mode or the second mode according to the instruction. The transmission unit 82 transmits the block signals sequentially generated by the generation unit 80.

  According to the embodiment of the present invention, block transmission without adding CP or GI can be realized by using CHARUE. Further, block transmission without adding CP or GI can be realized by using a version with reduced complexity of CHARUE. Also, IBI and ISI can be reduced by exchanging knowledge between adjacent blocks. In addition, IBI and ISI can be reduced in block transmission without adding CP or GI. Also, if the number of repetitions is sufficient, the influence of interference can be reduced by the approximation method. In addition, the complexity of processing can be reduced by using an approximation method. Further, since the block period is made uniform regardless of whether or not the GI is added, even when the GI is added or not, an increase in processing complexity can be suppressed. Further, since an error correction code is inserted instead of GI, the characteristics can be improved.

  In the above, this invention was demonstrated based on the Example. This embodiment is an exemplification, and it will be understood by those skilled in the art that various modifications can be made to the combination of each component and each processing process, and such modifications are also within the scope of the present invention. .

  In the embodiment of the present invention, the receiving device 20 receives a signal from a single transmitting device 10. However, the present invention is not limited to this. For example, the receiving device 20 may receive signals from a plurality of transmitting devices 10. At that time, the signal from each transmitting apparatus 10 is switched in units of blocks. The receiving device 20 may use a transmission path matrix or the like for each of the plurality of transmitting devices 10. According to this modification, the communication system 100 can be applied to TDMA.

In the embodiment of the present invention, the case where the modulation scheme used in the transmission device 10 and the reception device 20 is BPSK is illustrated. However, the present invention is not limited to this, and general QAM such as QPSK, 16QAM, 64QAM, etc. may be used as the modulation method. Up to now, the equations (19a) to (19c) correspond to BPSK. Here, the expression (12), the expression (19a) to the expression (19c) are replaced with the following expression.
Here, the following relationship is shown.
Note that b _m (s _i ) indicates the m-th bit of the symbol s _i , and M indicates the modulation multilevel number. Specifically, M = 2 in QPSK, M = 4 in 16QAM, and M = 6 in 64QAM. According to this modification, the present invention can be applied to various modulation schemes.

  10 transmitting device, 12 encoding unit, 14 interleaving unit, 16 modulating unit, 20 receiving device, 22 CHARUE equalizing unit, 24 deinterleaving unit, 26 decoding unit, 28 adding unit, 30 interleaving unit, 40 delay unit, 42 etc. Conversion unit, 44 turbo decoding unit, 50 extraction unit, 52 estimation unit, 54 filter unit, 56 generation unit, 58 holding unit, 60 acquisition unit, 62 derivation unit, 64 conversion unit, 66,68 multiplication unit, 70 addition unit, 72 Deinterleaving unit, 74 decoding unit, 76 adding unit, 78 interleaving unit, 80 generating unit, 82 transmitting unit, 84 control unit, 86 first encoding unit, 88 modulating unit, 90 GI inserting unit, 92 selecting unit, 94 A second encoding unit, 96 modulation unit, 100 communication system.

Claims (4)

An input unit for sequentially inputting the encoded signals;
A first equalization unit for equalizing a signal input in the input unit in a first block of a predetermined period;
A second equalization unit for equalizing a signal in a second block that is past the first block;
A third equalization unit for equalizing the signal in the third block in the future rather than the first block;
The first equalization unit, the second equalization unit, and the third equalization unit, with the log likelihood ratio obtained by decoding the equalization result in the first equalization unit as a first log likelihood ratio. A first decoding unit for returning to
A second decoding unit that feeds back a log likelihood ratio obtained by decoding the equalization result in the second equalization unit to the first equalization unit and the second equalization unit as a second log likelihood ratio When,
A third decoding unit that feeds back the log likelihood ratio obtained by decoding the equalization result in the third equalization unit to the first equalization unit and the third equalization unit as a third log likelihood ratio When,
An output unit for outputting a decoding result in the first decoding unit,
The first equalization unit reflects the first log likelihood ratio, the second log likelihood ratio, and the third log likelihood ratio at the time of equalization,
The second equalization unit reflects the first log likelihood ratio and the second log likelihood ratio at the time of equalization,
The third equalization unit reflects the first log likelihood ratio and the third log likelihood ratio at the time of equalization. The first equalization unit includes:
An acquisition unit for acquiring a first transmission path matrix corresponding to the first block, a second transmission path matrix corresponding to the second block, and a third transmission path matrix corresponding to the third block;
A holding unit that holds a permutation matrix for converting the first transmission line matrix into a cyclic matrix;
An extraction unit that extracts a signal included in the first block by calculating a permutation matrix on the signal input in the input unit;
A replica is generated based on the first log likelihood ratio, the second log likelihood ratio, the third log likelihood ratio, the first transmission path matrix, the second transmission path matrix, the third transmission path matrix, and the permutation matrix. A generator,
Based on the signal extracted by the extraction unit, the replica, the first transmission line matrix, and the first log likelihood ratio, a signal in which the influence of the signals included in the second block and the third block is reduced is estimated. An estimation unit;
A derivation unit for deriving a tap coefficient for equalization by frequency domain calculation based on the first transmission line matrix, the second transmission line matrix, the third transmission line matrix, and the permutation matrix;
The receiving apparatus according to claim 1, further comprising: a filter unit that filters the signal estimated by the estimation unit based on the tap coefficient derived by the deriving unit.   The deriving unit approximates a term including the second transmission path matrix and the permutation matrix and a term including the third transmission path matrix and the permutation matrix by diagonalization in the frequency domain calculation. The receiving device according to claim 2. A generator that sequentially generates a block signal formed by a combination of a first partial block including data and a second partial block;
A transmission unit for transmitting the block signals sequentially generated in the generation unit;
The generating unit is operable in a first mode in which a guard interval is included in the second partial block and a second mode in which an error correction code is included in the second partial block instead of the guard interval. A control section for selecting a mode;
A transmission device comprising: