JP5170535B2 - Data communication system, LLR calculation apparatus and method - Google Patents

Description

  The present invention relates to a data communication system suitable for wireless communication performing LDPC channel coding, and an LLR calculation apparatus and method.

  In wireless communication, SC-FDE (Single Carrier Frequency Domain Equalization) is a technology that has many advantages such as a low peak-to-average ratio and a strong carrier offset, and is a promising technology. In such an SC-FDE system, channel coding is indispensable.

LDPC (Low Density Parity Check codes), compared to other coding,
(1) High error correction capability
(2) Decoding is not complicated,
(3) Excellent extensibility,
Etc. show the advantages. Therefore, LDPC is suitable as channel coding in the SC-FDE system (Non-Patent Document 1 to Non-Patent Document 5).

In order to construct an SC-FDE system, in order to use LDPC as channel coding, it is necessary to calculate an LLR (Log-likelihood ratio). In order to obtain LLR, it is necessary to estimate equalization channel gain and noise variance. Conventionally, as shown in Non-Patent Document 4 and Non-Patent Document 5, an LLR is estimated using a channel estimation sequence embedded in a preamble segment.
RG Gallagar, Low-Density Parity-Check Codes, Cambridge, MA: MIT Press, 1963. DJC MacKay, "Good error-correcting codes based on very sparse matrices", IEEE Trans. Inform. Theory, vol. 45, no. 2, pp. 39931, March 1999. J. Hou, PH Siegel, and LB Milstein, "Performance analysis and code optimization of lowdensity parity-check codes on Rayleigh fading channels," IEEE J. Select. Areas Commun., Vol. 19, pp. 92434, May 2001. H. Harada et al., “CoMPA PHY proposal,” IEEE 802.15-07-0693-03-003c, May 2007. H. Harada et al., “Unified and flexible millimeter wave WPAN systems supported by common mode,” IEEE 802.15-07-0761-00-003c, July 2007.

  However, as described above, frequency domain equalization is performed in the SC-FDE system. As shown in Non-Patent Document 4 and Non-Patent Document 5, the preamble used for channel estimation in the conventional system is extracted before the frequency domain equalization processing and is not equalized together with the data. On the other hand, the channel gain and noise variance of data change due to frequency domain equalization. Therefore, it is difficult for the conventional system to correctly estimate the equalization channel gain and noise variance.

  In view of the above-described problems, the present invention provides a data communication system, an LLR calculation apparatus, and a method capable of accurately estimating channel gain and noise variance, obtaining LLR, and performing decoding without increasing overhead. For the purpose.

  In order to solve the above-mentioned problems, the data communication system of the present invention equalizes a transmitter that transmits data as a block structure in which a unique word is added to a data block and a received signal from the transmitter including the unique word. An equalization unit that calculates a channel gain from the received unique word after equalization processing, calculates a noise variance from the received unique word after equalization processing, and obtains a log likelihood ratio, and a calculation And a receiver having a channel decoding unit for performing channel decoding using the log likelihood ratio.

  Preferably, the unique word is a Golay series.

Preferably, channel decoding by the channel decoding unit is channel decoding by a low density parity check code .

  Preferably, the equalization unit performs an equalization process in a frequency domain.

  Preferably, the equalization unit performs an equalization process in a time domain.

  The LLR calculation apparatus of the present invention is a channel gain calculation that calculates a channel gain from a received unique word after equalization processing in an LLR calculation apparatus that calculates a log likelihood ratio by estimating a channel gain and noise variance of a received signal. And a noise variance calculation unit for calculating noise variance from the received unique word after equalization processing.

  Preferably, the channel gain calculation unit includes a first demultiplexer that divides the received unique word after equalization processing into first and second sub-sequences, and first and second sub-sequences from the first and second sub-sequences. First and second expected value calculation units for respectively obtaining the expected value, a first subtracting unit for subtracting the expected value of the first sub-sequence and the expected value of the second sub-sequence, and the first subtraction And a first division unit that divides the subtraction value of 2 by 2.

  Preferably, the noise variance calculation unit obtains an expected value from the second demultiplexer that divides the received unique word after equalization processing into the first and second sub-sequences, and the first and second sub-sequences, respectively. A third and fourth expected value calculation unit; a second subtraction unit that subtracts the expected value of the first sub-sequence from the first sub-sequence; and the expected value of the second sub-sequence and the second sub-sequence A third subtraction unit for subtracting the series, a first square expected value calculation unit for obtaining a square expected value of a subtraction value between the expected value of the first sub series and the first sub series, and a second sub A second expected square value calculation unit for obtaining a square expected value of a subtraction value between the expected value of the series and the second sub series; and a square of a subtraction value of the expected value of the first sub series and the first sub series An adder that adds the expected value, the expected value of the second sub-sequence, and the squared expected value of the subtraction value of the second sub-sequence; Characterized in that it comprises a second divider for dividing the addition value of the adder 2.

  The LLR calculation method of the present invention is a channel gain calculation for calculating a channel gain from a received unique word after equalization processing in an LLR calculation method for obtaining a log likelihood ratio by estimating a channel gain and noise variance of a received signal. And a noise variance calculating step of calculating noise variance from the received unique word after equalization processing.

  Preferably, the channel gain calculating step divides the received unique word after the equalization processing into first and second sub-sequences, obtains expected values of the first and second sub-sequences, And subtracting the expected value of one sub-sequence from the expected value of the second sub-sequence and dividing by two.

  Preferably, the noise variance calculation step includes a step of dividing the received unique word after equalization processing into first and second sub-sequences, a step of obtaining expected values of the first and second sub-sequences, Subtracting the first expected value from the first sub-sequence, subtracting the second expected value from the second sub-sequence, and the subtracted value between the first expected value and the first sub-sequence A step of obtaining a square expected value of the first step, a step of obtaining a square expected value of a subtraction value of the second expected value and the second sub-sequence, and a square of the subtraction value of the first expected value and the first sub-sequence A step of adding an expected value and a squared expected value of a subtraction value of the second expected value and the second sub-sequence and dividing by two.

  According to the present invention, a transmitter that transmits data as a block structure in which a unique word is added to a data block, an equalization unit that equalizes a received signal from the transmitter including the unique word, and a post-equalization process A channel gain is calculated from the received unique word and a noise variance is calculated from the received unique word after equalization processing to obtain a log likelihood ratio, and a channel decoding is performed using the calculated log likelihood ratio. A receiver having a channel decoding unit for performing channel decoding, LLR can be accurately calculated and channel decoding can be performed without increasing overhead.

  According to the present invention, the unique word is a Golay sequence, so that the channel gain is calculated from the received unique word after the equalization process and the noise variance is calculated from the received unique word after the equalization process to obtain the logarithm. A likelihood ratio can be obtained.

According to the present invention, since channel decoding by the channel decoding unit is channel decoding by a low density parity check code , high error correction capability can be secured without complicating the system.

  According to the present invention, the equalization unit can perform equalization with high performance on the transmission path of the multipath fading channel by performing equalization processing in the frequency domain.

  According to the present invention, the equalization unit can perform equalization in the transmission path of the multipath fading channel with a simple configuration by performing equalization processing in the time domain.

  According to the present invention, in an LLR calculation apparatus that estimates a channel gain and noise variance of a received signal and obtains a log likelihood ratio, a channel gain calculation unit that calculates a channel gain from a received unique word after equalization processing; Since the noise variance calculation unit for calculating the noise variance from the received unique word after the equalization processing is provided, the LLR can be accurately calculated without increasing the overhead.

  According to the present invention, the channel gain calculation unit includes a first demultiplexer that divides the received unique word after the equalization process into the first and second sub-sequences, and the first from the first and second sub-sequences. And a first expectation value calculator for obtaining the second expected value, a first subtractor for subtracting the expected value of the first sub-sequence and the expected value of the second sub-sequence, Since the first division unit for dividing the subtraction value of 1 subtraction unit by 2 is provided, the channel gain for obtaining the LLR can be accurately estimated.

  According to the present invention, the noise variance calculation unit includes the second demultiplexer that divides the received unique word after equalization processing into the first and second sub-sequences, and the expected value from the first and second sub-sequences. , A third subtraction unit for subtracting the expected value of the first sub-sequence and the first sub-sequence, an expected value of the second sub-sequence, A third subtraction unit that subtracts the second sub-sequence, a first expected-square value calculation unit that calculates a squared expected value of the subtraction value between the expected value of the first sub-sequence and the first sub-sequence, A second expected square value calculation unit for obtaining an expected square value of a subtraction value between the expected value of the second sub-sequence and the second sub-sequence; and subtraction of the expected value of the first sub-sequence and the first sub-sequence Addition of adding the squared expected value of the value and the squared expected value of the subtracted value of the second subsequence expected value and the second subsequence If so and a second division unit for dividing the added value of the adder 2 it can be estimated accurately noise variance for determining the LLR.

  According to the present invention, in the LLR calculation method for calculating the log likelihood ratio by estimating the channel gain and noise variance of the received signal, the channel gain calculating step for calculating the channel gain from the received unique word after equalization processing; The noise variance calculating step of calculating the noise variance from the received unique word after the equalization processing is included, so that the LLR can be accurately calculated without increasing the overhead.

  According to the present invention, the channel gain calculating step includes a step of dividing the received unique word after the equalization processing into first and second sub-sequences, and a step of obtaining expected values of the first and second sub-sequences, respectively. And subtracting the expected value of the first sub-sequence and the expected value of the second sub-sequence and dividing by 2, so that the channel gain for obtaining the LLR can be accurately estimated. Can do.

  According to the present invention, the noise variance calculation step includes a step of dividing the received unique word after equalization processing into first and second sub-sequences, and a step of obtaining expected values of the first and second sub-sequences, respectively. Subtracting the first expected value from the first sub-sequence, subtracting the second expected value from the second sub-sequence, the first expected value and the first sub-sequence, Obtaining a squared expected value of the subtracted value, obtaining a squared expected value of the subtracted value of the second expected value and the second sub-sequence, and subtracting the first expected value and the first sub-sequence A step of adding a square expected value of the value and a square expected value of the subtraction value of the second expected value and the second sub-sequence and dividing the result by two. Can be estimated accurately.

First embodiment.
Hereinafter, embodiments of the present invention will be described with reference to the drawings. FIG. 1 shows an outline of the SC-FDE system according to the first embodiment of the present invention.

  In FIG. 1, on the transmitter 1 side, LDPC coding is performed by a channel coding unit 11. The output of the channel coding unit 11 is modulated by the modulation unit 12 and the guard interval is inserted by the guard interval insertion unit 13. In the embodiment of the present invention, an FFT block is configured by adding a unique word to the data block at this time. As the unique word, for example, a Golay series is used. A signal to which the guard interval is added is transmitted from the transmission unit 14. A transmission signal from the transmitter 1 is received by the receiver 2 through the transmission path of the multipath fading channel.

  On the receiver 2 side, the signal from the transmitter 1 is received by the receiver 20, the guard interval is removed by the guard interval remover 21, and then sent to the equalizer 23. In the embodiment of the present invention, the unique word added to the data block in the FFT block among the signals added as the guard interval is sent to the equalization unit 23 together with the data without being removed.

  The equalization unit 23 converts the received signal in the time domain into the frequency domain by the FFT 31, performs equalization processing in the frequency domain by the frequency domain equalization unit 32, and then returns to the time domain again by the IFFT 33 and outputs it. The received signal that has been equalized in the frequency domain by the equalization unit 23 is sent to the LLR calculation unit 25.

  The LLR calculation unit 25 calculates an LLR. In the first embodiment of the present invention, the LLR is calculated using the received unique word after the frequency domain equalization processing.

  The output of the LLR calculation unit 25 is sent to the channel decoding unit 26. The channel decoding unit 26 performs decoding by LDPC. The received bit stream is decoded by the channel decoding unit 26, and the received bit stream is output from the channel decoding unit 26.

  As described above, in the first embodiment of the present invention, the LLR calculator 25 obtains the LLR using the received unique word after the frequency domain equalization processing. The process for calculating the LLR will be described using the model shown in FIG.

  FIG. 2 is a model of the above-described SC-FDE system. In FIG. 2, a point A from the output of the modulation unit 12 on the transmitter 1 side to a point C of the input to the LLR calculation unit 25 on the receiver 2 side is shown as a virtual channel 40. In the virtual channel 40, processing of the guard interval insertion unit 13 on the transmission side, multipath fading from the transmission unit 14 to the reception unit 20 on the receiver 2 side through the multipath transmission path, The channel 41 and the receiver 2 side include processing of converting to the frequency domain by the FFT 31, performing equalization processing in the frequency domain by the equalization unit 32, and then returning to the time domain again by the IFFT 33 and outputting. .

  The multipath fading channel 41 is a model of a channel of a signal transmitted from the transmission side to the reception side, and is a multipath fading channel assumed to be quasi-static.

  If the original transmission signal on the transmitter 1 side is x (t), the reception signal r (t) on the receiver 2 side can be described as follows.

Here, h (t) is a channel impulse response, n (t) is white Gaussian noise added with variance σ ² , and an operator between x (t) and h (t) indicates a linear convolution operation. .

  As shown in FIG. 3, the equation (1) is obtained by convolving a channel impulse response indicated by h (t) with respect to the transmission signal x (t) and adding noise n (t) to the received signal. This is a signal model of a multipath fading channel.

  On the receiver 2 side, after the guard interval is removed by the guard interval removing unit 21, the received signal r (t) is converted into the frequency domain by the FFT unit 31, and the frequency domain equalizing unit 32 performs frequency domain equalization. Processing is performed. The equalized signal is returned to the time domain by the IFFT unit 33. Then, the LLR calculation unit 25 obtains the LLR, and the channel decoding unit 26 performs LDPC decoding.

  In order to calculate the LLR by the LLR calculator 25, it is necessary to obtain the channel gain and the noise variance as will be described later. Intersymbol interference is suppressed by the frequency domain equalization unit 32. However, by performing such frequency domain equalization, the channel gain and the noise variance of the equalized signal r (t) are changed. It becomes difficult to estimate channel gain and noise variance.

  Therefore, in the first embodiment of the present invention, transmission is performed in a block structure that forms a FFT block by adding a unique word to a data block, and using the received unique word after frequency domain equalization processing, LLR is obtained.

  That is, since the intersymbol interference is suppressed by performing equalization in the frequency domain equalization unit 32, this signal model is flat Rayleigh fading and can be described as follows:

Here, α is equalization channel gain, and n (t) is noise having variance σ ² . In addition, the symbol which attach | subjects "~" means equalizing. As shown in FIG. 4, the flat Rayleigh fading signal model is obtained by multiplying the transmission signal x (t) by an equalization channel gain indicated by α and adding an equalized noise n (t). Thus, the received signal r (t) after equalization is obtained.

  After sampling, this discrete signal can be described as:

When the sample index is removed from the above equation, it can be further described as follows.

Compared with the signal model expressed by equation (1), the assumed channel model expressed by equation (4) simplifies the calculation of LLR. The calculation of LLR will be further described.

  As an example, if BPSK is used as the modulation method of the modulation unit 12,

The coded bits are mapped to modulation symbols. The LDPC encoded bitstream is a symmetric binary code,
Pr (s = 0) = Pr (s = 1) = 1/2
Is assumed. For an uncorrelated Rayleigh fading channel, the LLR is calculated as shown in equation (5).

In order to calculate the LLR from the equation (5), it is necessary to estimate the equalization channel gain α and the noise variance σ ² after equalization.

  In the conventional SC-FDE, the preamble used for channel estimation is not equalized with the data by the frequency domain equalization unit 32. For this reason, in the conventional SC-FDE, it is difficult to estimate the equalized channel gain and the noise variance after equalization.

  On the other hand, in the embodiment of the present invention, a known signal is embedded in the data block. If a known signal is embedded in the data block, this known signal is converted into the frequency domain by the FFT unit 31 together with the data, the frequency domain equalization unit 32 performs frequency domain equalization processing, and the IFFT unit 33 performs time domain processing. The process returned to is performed.

FIG. 5 shows a frame structure of SC-FDE. FIG. 5A shows a conventional frame structure based on a cyclic prefix, and FIG. 5B is based on a unique word. T _{G, T D,} and _{T FFT} is _{T G,} respectively, the time length of the guard interval, the time length of the data block, the time length of the FFT block.

The frame structure shown in FIG. 5A is a cyclic prefix and is copied from the last part of each data block. In FFT / IFFT operation, the FFT block length is exactly equal to the length of the data block (T _FFT = T _D ).

The frame structure shown in FIG. 5B is a unique word and is inserted between two adjacent data blocks. Each FFT block is composed of one block and a unique word added thereto (T _FFT = T _D + T _G ).

By using a frame composed of unique words as shown in FIG. 5B and performing frequency domain equalization together with the data block, the equalization channel gain α and the noise variance σ ² after equalization can be estimated. .

That is, the data block time length T _D and the unique word time length T _FFT in FIG. 5B are the T _FFT , and the data block plus the unique word are combined into one FFT block. It is sent to the FFT unit 31 to perform FFT, the frequency domain equalization unit 32 performs frequency domain equalization processing, and the IFFT unit 33 performs processing to return to the time domain. Then, the LLR is calculated using the received unique word after the frequency domain equalization processing.

  As an example, the Golay sequence is used as a unique word. The Golay series is a bipolar series having {+1, -1} elements. Depending on the original code (+ or −), the equalized Golay sequence symbols (point C in FIG. 2) can be classified into a “+” subsequence and a “−” subsequence.

From the above equation, the equalization channel gain α and the equalized noise variance σ ² can be estimated as follows.

In the above equation, E (•) is a mathematical expectation value, and E (•) ² is a mathematical square expectation value. The LLR for LDPC can be obtained from the equations (6) and (7). By using this method, no additional overhead is required for the calculation of the LLR.

  FIG. 6 shows a configuration of a channel gain calculation unit in the LLR calculation unit 25. This channel gain calculation unit calculates the channel gain based on the equation (6).

  In FIG. 6, the received unique word after the frequency domain equalization processing is divided into a “+” sub-sequence and a “−” sub-sequence by the demultiplexer 51 (first demultiplexer). The “+” sub-sequence is sent to the expected value calculator 52 (first expected value calculator), and the “−” sub-sequence is sent to the expected value calculator 53 (second expected value calculator). The subtraction unit 54 (first subtraction unit) subtracts the expected value of the “+” sub-sequence from the expected value of the “−” sub-sequence, and the division unit 55 (first division unit) The equalized channel gain is obtained by dividing the subtraction value between the expected value of the “+” sub-sequence and the expected value of the “−” sub-sequence by 2.

  FIG. 7 shows a configuration of a noise variance calculation unit in the LLR calculation unit 25. This noise variance calculation unit calculates the noise variance based on equation (7).

  In FIG. 7, the received unique word after the frequency domain equalization processing is divided into a “+” sub-sequence and a “−” sub-sequence by the demultiplexer 61 (second demultiplexer). The “+” sub-sequence is sent to the expected value calculator 62 (third expected value calculator), and the “−” sub-sequence is sent to the expected value calculator 63 (fourth expected value calculator).

  The subtraction unit 65 (second subtraction unit) subtracts the “+” sub-sequence and the expected value of the “+” sub-sequence. The subtraction unit 66 (third subtraction unit) subtracts the “−” sub-sequence and the expected value of the “−” sub-sequence.

  The expected square value calculation unit 67 (first expected square value calculation unit) calculates the expected square value for the subtraction value between the expected value of the “+” sub-sequence and the “+” sub-sequence. The expected square value calculation unit 68 (second expected square value calculation unit) calculates the expected square value for the subtraction value between the expected value of the “−” sub-sequence and the “−” sub-sequence.

  The addition unit 69 subtracts the squared expected value of the subtraction value between the expected value of the “+” sub-sequence and the expected value of the “+” sub-sequence, and the expected value of the “−” sub-sequence and the “−” sub-sequence. The square expected value for the value is added, and this is divided by 2 in the division unit 70 (second division unit), and the noise variance is calculated.

  FIG. 8 shows an evaluation method of an SNR (Signal to Noise Ratio) in an environment defined by IEEE802.15.3c (60 GHz WPAN (Wireless Personal Area Network). The effects of the first embodiment will be described, and the specifications are as shown in Fig. 9. In order to evaluate the LDPC, BPSK (Binary Phase Shift Keying) was used as the modulation. Covers coding rates 1/2, 3/4 and 7/8, the maximum number of decoding iterations was set to 64. The FFT / IFFT length was 256. Cyclic prefix or unique word guard interval Was assumed to be 64. The synchronization and channel estimation were ideal, and frequency domain equalization was the MMSE Minimum Mean-Square Error) algorithm. It is based on the arm. Each data packet carries 2K (2048 bytes) of raw data.

  The two calculation methods of the LLR calculation method according to the present invention and the conventional LLR calculation method were compared. Note that NUW-LLR (non-unique word based LLR calculation) is a conventional LLR calculation method, and channel gain and noise variance were estimated after a multipath fading channel (point B in FIG. 2). . Usually, a channel estimation sequence embedded in a preamble segment is used.

  UW-LLR (unique word based LLR calculation) is an LLR calculation method according to an embodiment of the present invention, and the LLR is calculated based on a unique word. Channel gain and noise variance are estimated after IFFT (point C in FIG. 2). As a unique word, a Golay series having a length of 64 was used.

FIG. 8 shows the performance of BER (Bit Error Rate) using NUW-LLR and UW-LLR coded with a length 576 LDPC code. Comparing UW-LLR and conventional NUW-LLR shows that UW-LLR has better performance. At a BER of 10 ⁻⁶ , UW-LLR gain performance is 1.7 dB, 0.6 dB, and (576,288), (576,432), and (576,504) LDPC codes, respectively. And 0.7 dB. However, as described above, the estimation of the channel gain and noise variance of NUW-LLR is assumed to be ideal. This means that the gain performance of UW-LLR may be even better than that shown in FIG.

  In the UW-LLR method according to the embodiment of the present invention, the LLR is calculated as shown in FIGS. There is no need for additional overhead. This promises to further improve the performance of the LDPC code of the SC-FDE system.

Second embodiment.
FIG. 10 shows a second embodiment of the present invention. The difference between the first embodiment and the second embodiment is whether equalization processing is performed in the frequency domain or in the time domain. In the first embodiment shown in FIG. 1, the equalization unit 23 performs an equalization process in the frequency domain. On the other hand, in the second embodiment shown in FIG. 10, the time domain equalization unit 123 performs equalization processing in the time domain. About another structure, it is the same as that of the above-mentioned 1st Embodiment, The description is abbreviate | omitted.

  As described above, according to the embodiment of the present invention, a unique word is added to a data block for transmission, and the overhead is increased by obtaining the LLR using the received unique word after frequency domain equalization processing. Therefore, it is possible to accurately estimate the channel gain and noise variance and perform LDPC decoding.

  The present invention is not limited to the above-described embodiments, and various modifications and applications can be made without departing from the gist of the present invention.

  The present invention can be widely used in communication systems that perform transmission using multipath fading channels in addition to wireless communication such as WPAN.

It is a block diagram used for description of the communication system of 1st Embodiment of this invention. It is a block diagram used for description of the channel model of the communication system of 1st Embodiment of this invention. It is a block diagram used for description of the transmission channel in the communication system of 1st Embodiment of this invention. It is a block diagram used for description of the transmission channel in the communication system of 1st Embodiment of this invention. It is explanatory drawing of the frame structure in the communication system of 1st Embodiment of this invention. It is a block diagram used for description of calculation of the channel gain in 1st Embodiment of this invention. It is a block diagram used for description of calculation of noise variance in a 1st embodiment of the present invention. It is a graph for demonstrating the effect of 1st Embodiment of this invention. It is explanatory drawing of the item of the simulation for demonstrating the effect of 1st Embodiment of this invention. It is a block diagram used for description of the communication system of 1st Embodiment of this invention.

Explanation of symbols

DESCRIPTION OF SYMBOLS 1 Transmitter 2 Receiver 11 Channel coding part 12 Modulation part 13 Guard interval insertion part 14 Transmission part 20 Reception part 21 Guard interval removal part 23 Equalization part 25 LLR calculation part 26 Channel decoding part 31 FFT part 32 Equalization part 33 IFFT unit 51 Demultiplexer 52 Expected value calculating unit 53 Expected value calculating unit 54 Subtracting unit 55 Dividing unit 61 Demultiplexer 62 Expected value calculating unit 63 Expected value calculating unit 65 Subtracting unit 66 Subtracting unit 67 Expected square value calculating unit 68 Expected square value Calculation unit 69 Addition unit 70 Division unit

Claims (11)

A transmitter for transmitting data as a block structure in which a unique word is added to a data block;
An equalization unit that equalizes the received signal from the transmitter including a unique word, and calculates a channel gain from the received unique word after the equalization process and noise from the received unique word after the equalization process A data communication comprising: an LLR calculation unit that calculates variance and obtains a log likelihood ratio; and a receiver that includes a channel decoding unit that performs channel decoding using the calculated log likelihood ratio. system.   The data communication system according to claim 1, wherein the unique word is a Golay sequence. 3. The data communication system according to claim 1, wherein channel decoding by the channel decoding unit is channel decoding by a low density parity check code .   4. The data communication system according to claim 1, wherein the equalization unit performs equalization processing in a frequency domain.   4. The data communication system according to claim 1, wherein the equalization unit performs an equalization process in a time domain. In an LLR calculation apparatus that calculates a log likelihood ratio by estimating a channel gain and noise variance of a received signal,
A channel gain calculation unit for calculating a channel gain from the received unique word after equalization processing;
An LLR calculation apparatus comprising: a noise variance calculation unit that calculates a noise variance from the received unique word after the equalization processing. The channel gain calculator is
A first demultiplexer for dividing the received unique word after the equalization processing into first and second sub-sequences;
First and second expected value calculators for obtaining first and second expected values from the first and second sub-sequences, respectively;
A first subtraction unit for subtracting the expected value of the first sub-sequence and the expected value of the second sub-sequence;
The LLR calculation apparatus according to claim 6, further comprising: a first division unit that divides a subtraction value of the first subtraction unit by two. The noise variance calculation unit
A second demultiplexer that divides the received unique word after equalization into first and second subsequences;
Third and fourth expected value calculation units for obtaining expected values from the first and second subsequences, respectively;
A second subtraction unit for subtracting the expected value of the first sub-sequence and the first sub-sequence;
A third subtracting unit for subtracting the expected value of the second sub-sequence and the second sub-sequence;
A first expected square value calculation unit for obtaining an expected square value of a subtraction value between the expected value of the first sub-sequence and the first sub-sequence;
A second expected square value calculation unit for obtaining an expected square value of a subtraction value between the expected value of the second sub-sequence and the second sub-sequence;
Expected square value of subtraction value between expected value of first sub-sequence and first sub-sequence, and expected square value of subtraction value of expected value of second sub-sequence and second sub-sequence An adder for adding and
The LLR calculation apparatus according to claim 6, further comprising: a second division unit that divides the addition value of the addition unit by two. In the LLR calculation method for calculating the log likelihood ratio by estimating the channel gain and noise variance of the received signal,
A channel gain calculating step for calculating a channel gain from the received unique word after equalization processing;
And a noise variance calculating step of calculating a noise variance from the received unique word after the equalization processing. The channel gain calculation step includes:
Dividing the received unique word after equalization processing into first and second subsequences;
Obtaining expected values of the first and second sub-sequences, respectively;
The method of claim 9, further comprising: subtracting the expected value of the first sub-sequence from the expected value of the second sub-sequence and dividing by two. The noise variance calculating step includes:
Dividing the received unique word after equalization processing into first and second subsequences;
Obtaining expected values of the first and second sub-sequences, respectively;
Subtracting the first expected value from the first sub-sequence;
Subtracting the second expected value from the second sub-sequence;
Obtaining a squared expected value of a subtraction value between the first expected value and the first sub-sequence;
Obtaining a squared expected value of a subtraction value between the second expected value and the second sub-sequence;
The squared expected value of the subtraction value between the first expected value and the first sub-sequence and the squared expected value of the subtracted value between the second expected value and the second sub-sequence are added to obtain 2 The LLR calculation method according to claim 9, further comprising: dividing by.