JP2009159037A - Receiver - Google Patents

Abstract

PROBLEM TO BE SOLVED: To provide a receiving device capable of soft decision decoding of a received signal without assuming a communication channel and without using an AD converter that outputs a plurality of bits.
An oversampling unit determines binary data of each sampling point obtained by oversampling a received signal, and sequentially outputs a binary data string. The reference table 45 defines the relationship between the sampling data yn obtained by dividing the binary data string corresponding to the received signal and the likelihood λn indicating the likelihood of the received signal with respect to the transmission signal xn. The likelihood determining unit 48 determines the likelihood using. The error correction decoding unit 25 performs error correction decoding on the received signal based on the determined likelihood.
[Selection] Figure 9

Description

  The present invention relates to a receiving apparatus that performs soft decision decoding by oversampling a received signal.

As a technique for realizing a long distance and large capacity communication system at low cost, error correction (FEC) plays a major role.
In the past, Receiving sensitivity was improved by using Read Solomon and other hard decision error correction techniques, but in recent years, soft decision error correction that can further improve reception sensitivity has been incorporated as a standard specification. (For example, IEEE802.16e for broadband wireless access, DVB-S2 for digital broadcasting standard, IEEE802.3an for 10G Base-T standard).

  Typical examples of the soft decision error correction code include a turbo code and a low-density parity check (LDPC) code. The correction capability is expressed by expressing the received data using the certainty of the received data (soft decision) instead of definitely expressing (hard decision) as a binary bit string of “0” and “1”. Has improved.

  For example, as described in Non-Patent Document 1, it is assumed that additive white Gaussian noise is applied to a system in which a transmission symbol (0, 1) is modulated to (+1, −1) and transmitted. In this case, the probability of the received value yn of the received nth symbol is that the transmission data xn is originally “0” and received as yn, and the transmission data xn is originally “1” and yn. The natural number of received probability ratios can be taken and expressed as:

上式において、λｎは対数尤度比と呼ばれ、σ²は伝送路で印加される雑音の分散値である。

In the above equation, λn is called a log likelihood ratio, and σ ² is a variance value of noise applied in the transmission line.

On the other hand, Non-Patent Document 2 does not sample received data with binary values, but uses an AD converter that can output an output value of a plurality of bits from received data, and samples with multiple values of two or more values. A method for softly determining received data is disclosed.
Wadayama Tadashi “Low Density Parity Check Codes and Decoding Methods” IEICE Technical Report, MR2001-83, December 2001 Ouchi-Ei et al., "Study on threshold interval control in soft decision FEC for WDM systems", IEICE General Conference, 2004, p.469

In Non-Patent Document 1, it is assumed that a communication channel (also referred to as a transmission channel) is an additive white Gaussian channel. However, since the influence of reception jitter is not taken into consideration, this assumption is not used in a normal communication system. It may not hold. For example, in the case where there is a delayed wave (reflected wave) from a building in wireless communication or in the case of optical communication using a multimode fiber, the above assumption is often not satisfied.
For this reason, on the receiver side, by using an equalizer to reduce the effects other than white noise and converting them into a state where the effects can be ignored, the entire communication path is regarded as an additive white Gaussian communication path. It is normal to use the above log likelihood ratio equation. Therefore, in this case, two of an equalizer and a log likelihood ratio calculator are required independently.

On the other hand, in the case of Non-Patent Document 2, since the communication path is not assumed to be an additive white Gaussian communication path, the above problem does not occur, but the AD converter outputs a plurality of bits. There is a problem that the processing of the AD converter cannot catch up as the transmission speed increases.
For example, in the case of high-speed data transmission exceeding 1 Gbps, the feasibility itself of an AD converter that outputs a plurality of bits is difficult, and even the latest technology trend is compatible with a sampling frequency of 3 GHz. It is said that there is.

  In addition, for example, in the case of a PON (Passive Optical Network) system in which the communication system shares an optical communication path, since the distance and the number of branches of the optical fiber from the station side device to each home side device are different, the station side device The received light level differs depending on which home-side device has transmitted. For this reason, the station side device needs to set the threshold value of the AD converter in accordance with the received light level, and the processing becomes complicated.

  In view of such circumstances, the present invention provides a receiving apparatus capable of soft-decision decoding of a received signal without assuming a communication channel and without using an AD converter that outputs a plurality of bits. For the purpose.

  A receiving apparatus according to the present invention (Claim 1) includes: an oversampling unit that determines binary data of each sampling point obtained by oversampling a received signal and sequentially outputs a binary data string; and the binary data The likelihood is determined using a reference table that defines a relationship between sampling data obtained by dividing a column corresponding to the received signal and a likelihood indicating the likelihood of the received signal with respect to a transmission signal, and the reference table. A likelihood determining unit and a decoding unit that performs error correction decoding on the received signal based on the determined likelihood are provided.

  According to the receiving apparatus of the present invention, the oversampling unit sequentially outputs a binary data sequence at each sampling point obtained by oversampling the received signal, and the likelihood determining unit outputs the binary data sequence to the received signal. The likelihood is determined using a reference table that defines the relationship between the sampling data divided corresponding to the likelihood and the likelihood indicating the likelihood of the received signal, and the decoding unit receives based on the determined likelihood. Since error correction decoding is performed on the signal, the received signal can be soft-decision decoded without assuming a communication channel and without using an AD converter that outputs a plurality of bits.

  Further, according to the receiving apparatus of the present invention, the likelihood determining unit determines the likelihood using a reference table that predefines the relationship between the sampling data and the likelihood, so that the likelihood can be accurately determined in a short time. Can be determined.

Incidentally, as will be described in detail in a later embodiment, it has been clarified that if the sampling timing in the oversampling unit is appropriately set, symmetry appears in the likelihood result corresponding to each sampling data.
Therefore, if the sampling timing in the oversampling unit is set so that symmetry appears in the likelihood result (claim 2), the table size required for the reference table can be reduced, and the memory of the receiving apparatus can be reduced. Can be used effectively, and the observation time for creating the reference table can be shortened.

In the receiving apparatus of the present invention, a scaling unit for scaling the likelihood determined by the likelihood determining unit is provided in order to decode the received signals having a plurality of signal-to-noise ratios by one reference table. (Claim 3).
In this case, by scaling the likelihood determined using one reference table, it becomes possible to decode received signals having other signal-to-noise ratios, so that the decoding performance of the receiving apparatus can be further improved. .

  However, as will be described in detail later, the decoding unit performs error correction decoding based on a decoding method (for example, a minisum decoding method) in which the sum product decoding method is simplified with respect to encoded data composed of LDPC codes. If it is to be performed, the decoding result does not fluctuate even if the likelihood (specifically, the log likelihood ratio) is scaled. Therefore, as the likelihood determining unit, the received signal having a plurality of signal-to-noise ratios is added. On the other hand, what determines the likelihood based on one reference table may be adopted.

Furthermore, in the receiving apparatus of the present invention, a transmission line monitoring unit that monitors the state of the transmission line and a sampling rate control unit that changes the number of samples in the oversampling unit corresponding to the state of the transmission line are provided. (Claim 5).
In this case, since the sampling rate control unit changes the number of samplings in the oversampling unit corresponding to the state of the transmission path, even if the trajectory of the received signal changes more complicatedly due to the change in the state of the transmission path, the sampling rate The received signal can be appropriately decoded by changing the number.

  According to the present invention, it is possible to perform soft decision decoding of a received signal without assuming a communication path and without using an AD converter that outputs a plurality of bits.

Hereinafter, embodiments of the present invention will be described with reference to the drawings.
[Configuration of communication system]
FIG. 1 is a topology diagram showing a configuration example of a communication system suitable for implementing the present invention.
As shown in FIG. 1, this communication system is called a PON (Passive Optical Network) system, and includes one station-side device 1 and a plurality of home-side devices that perform bidirectional communication with the station-side device 1. 2a, 2b, ..., 2i, 2j are connected by an optical fiber 5. In such a PON system, the optical fiber 5 is branched using the optical splitters 3 and 4 so that one optical fiber 5 can be shared by a plurality of home-side devices.

In the PON system of FIG. 1, the distance of the optical fiber 5 to the station side device 1 and the home side devices 2a, 2b,..., 2i, 2j and the number of branches of the optical fiber 5 are different depending on each home side device 2. ing. Specifically, the optical fiber 5 a connected to the station side device 1 is branched into four by a 1 × 4 splitter 3.
The optical fiber 5b branched into four by the 1 × 4 splitter 3 is further branched by the 1 × 4 splitter 4a and connected to the home-side devices 2a and 2b, and further branched by the 1 × 8 splitter 4d to be the home-side devices 2i, 2i, 2j.

Therefore, in the PON system of FIG. 1, the number of branches of the optical fiber 5 from the station side device 1 to the home side devices 2a and 2b is 4 × 4 = 16 branches. The number of branches of the optical fiber 5 is 4 × 8 = 32 branches.
The distance from the station side device to the 1 × 4 splitter 3 is 1 km, and the distance from the 1 × 4 splitter 3 to the 1 × 4 splitter 4a is 5 km. The distance from the 1 × 4 splitter 4a to the home device 2a is 6 km, and the distance from the 1 × 4 splitter 4a to the home device 2b is 7 km.

Further, the distance from the 1 × 4 splitter 3 to the 1 × 4 splitter 4d is 9 km. The distance from the 1 × 4 splitter 4d to the home device 2i is 8 km, and the distance from the 1 × 4 splitter 4d to the home device 2j is 7 km.
The light transmitted from any home-side device 2 depending on the distance of the optical fiber 5 from the station-side device 1 to the home-side devices 2a, 2b, ..., 2i, 2j and the difference in the number of branches of the optical fiber 5. Depending on the signal, the power of the optical signal received by the station side device 1 differs.

For example, when the optical signal transmitted by the home apparatus 2a is received by the station apparatus 1, the transmitted optical signal is 1/16 times the power due to 16 branches, and the transmission distance is 12 km (= 1 km + 5 km + 6 km). Power further decreases.
When the station side apparatus 1 receives the optical signal transmitted from the home apparatus 2i, the transmitted optical signal has a power of 1/32 times by 32 branches and a transmission distance of 18 km (= 1 km + 9 km + 8 km). Power further decreases.

Thus, since the power of the optical signal received by the station side device 1 is different, when the station side device 1 makes a soft decision on the received signal by the AD converter as in Non-Patent Document 2, the AD converter is provided for each home device 2. This threshold value must be prepared, and the processing may be complicated.
In such a PON system, bidirectional communication is possible between the station side device 1 and the home side devices 2a, 2b,..., 2i, 2j, but in the present invention, the power of the optical signal is the home side device 2a. , 2b,..., 2i, 2j is particularly effective in the case of uplink transmission to the station side device 1, and in the following, the home side devices 2a, 2b,. The embodiment of the present invention will be described with the transmission side (transmission device) and the station side device 1 as the reception side (reception device).

[Configuration of home device (transmission device)]
FIG. 2 is a configuration diagram of a home side apparatus (transmission apparatus) 2a which is a transmission side of the communication system of FIG. The other home side devices 2b,..., 2i, 2j are similar to this.
As shown in FIG. 2, the home device 2 a includes an error correction encoding unit 11 and an electro-optical conversion unit 12. Of these, the error correction encoding unit 11 divides the transmission data into K bits.

This error correction encoding unit 11 adds M redundant bits calculated from K information bits and S synchronization bits to K information bits divided by itself, and adds L LDPC code (low density parity check code) data of (= K + M + S) bits is generated. Note that the sum of the number of information bits K and the number of redundant bits M is N.
On the other hand, the electro-optical converter 12 converts the LDPC code data (0 or 1 time-series data) output from the error correction encoder 11 into an optical signal and outputs the optical signal to the optical fiber 5 serving as a communication path. .

[Configuration of station side device (receiving device)]
FIG. 3 is a configuration diagram of a station-side device (receiving device) 1 that is the receiving side of the communication system of FIG.
As illustrated in FIG. 3, the station-side device 1 according to the present embodiment includes an opto-electric converter 21, a TIA (Trans-Impedance Amplifier) 22, an LA (Limiting Amplifier) 23, a log likelihood ratio calculation unit 24, The error correction decoding unit 25 is provided.

Among these, the photoelectric converter 21 outputs a current corresponding to the optical level of the received signal received from the optical fiber 5, and the TIA 22 converts the current output from the photoelectric converter 21 into an analog voltage. To do. The LA 23 amplifies the analog voltage output from the TIA 22.
Further, the log likelihood ratio calculation unit 24 calculates the log likelihood ratio λ from the voltage output from the LA 23, and the error correction decoding unit 25 uses the calculated log likelihood ratio λ, Error correction decoding is performed in a unit of code length N by a sum-product decoding method or a decoding method (for example, a min-sum decoding method) obtained by simplifying the sum-product decoding method.

[Configuration of error correction decoding unit]
FIG. 4 is a configuration diagram of the error correction decoding unit 25.
As shown in FIG. 4, the error correction decoding unit 25 includes a row processing unit 27 that performs row processing of the parity check matrix, a column processing unit 28 that performs column processing of the parity check matrix, and a log likelihood ratio calculation unit 24. And a loop determination unit 26 that generates a decoded word according to the output bit (external value logarithmic ratio) αmn of the row processing unit 27.

[Sum product decoding method]
When the decoding method is the sum product decoding method, the row processing unit 27 and the column processing unit 28 perform arithmetic processing according to the following equations (1) and (2), respectively, and each element in the row of the parity check matrix The process (row process) and the process for each column element (column process) are executed.
Specifically, the row processing unit 27 updates the external value log ratio αmn according to the prior value log ratio βmn and the log likelihood ratio λn given from the column processing unit 28. The column processing unit 28 calculates the prior value logarithmic ratio βmn according to the external value logarithmic ratio αmn given from the row processing unit 27.

In the above formulas (1) and (2), n′∈A (m) \ n and m′∈B (n) \ m mean elements other than themselves. For the external value log ratio αmn, n ′ ≠ n, and for the prior value log ratio βmn, m ′ ≠ m.
In addition, the subscript “mn” indicating the position in the matrix of α and β is normally indicated by a subscript, but in the present specification, it is indicated by “horizontal characters” for readability.

The function f (β) in the equation (1) is the Gallager f function and is defined by the equation below the equation (1). In addition, the function sign (x) in the equation (1) is defined by the following equation (3).

The sets A (m) and B (n) are set [1, N] = {1, when H = [Hmn] of the binary M · N matrix is a parity check matrix of the LDPC code to be decoded. 2,..., N}.
A (m) = {n: Hmn = 1} (4)
B (n) = {m: Hmn = 1} (5)
That is, the subset A (m) means a set of column indexes where 1 stands in the m-th row of the parity check matrix H, and the subset B (n) is 1 in the n-th column of the parity check matrix H. Indicates the set of row indexes where

[Minisum decoding method]
On the other hand, when the decoding method is the minisum decoding method, the row processing unit 27 and the column processing unit 28 perform arithmetic processing according to the following equations (6) and (7).

As is clear from the comparison of Equations (1) and (6), the minisum decoding method replaces the term related to the Galaga f function with an approximate value in the computation of the external value logarithmic ratio αmn. The load is reduced. Therefore, the minisum decoding method is one simple implementation format of the sum product decoding method.
In equation (6), the function min is an operator for obtaining the minimum value. Also, the equation (2) of the sum product decoding method and the equation (7) of the minisum decoding method are the same.

[Parity check by loop judgment unit]
The loop determination unit 26 generates primary estimated words based on the arithmetic processing results in the row processing unit 27 and the column processing unit 28, and checks whether these primary estimated words constitute a code word.
If the syndrome does not become “0” at the time of this parity check, the processing is repeatedly executed again. When the number of repetitions of this process reaches a predetermined value, the primary estimated word at that time is output as a decoded word.

FIG. 5 is a flowchart showing the processing operation of the loop determination unit 26. Hereinafter, with reference to FIG. 5, the processing change of the loop determination unit 26 will be described.
First, as an initial operation, initial setting of the number of loops and the prior value log ratio βmn is performed. This number of loops indicates the number of operations in a loop in which the row value processing unit 27 generates the external value logarithmic ratio αmn again using the prior value logarithmic ratio βmn generated by the column processing unit 28. A maximum value is predetermined for the number of loops. The prior value log ratio βmn is initially set to “0” (step SP1).

Next, according to the received symbol sequence, an approximate log likelihood ratio λn and an external value log ratio αmn are generated by the log likelihood ratio calculation unit 24 and the row processing unit 27, respectively, and are supplied to the loop determination unit 26 (step). SP2).
The loop determination unit 26 calculates the estimated received word Qn by performing the operation Qn = λn + Σαmn according to the supplied approximate log likelihood ratio λn and external value log ratio αmn (step SP3). Here, the summation Σ is executed for the element m of the subset B (n).

The sign of the value Qn calculated in step SP3 is determined (step SP4), and a primary estimation code Cn is generated (step SP5).
In this sign positive / negative determination, for example, when the estimated received word Qn is displayed in two's complement, positive / negative determination can be performed by looking at the bit value of the most significant bit (sign bit). it can.

When all estimated codes Cn are generated and codewords (C1,..., CN) are generated, a parity check is then performed (step SP6).
In this parity check, (C1,..., CN) · H ^t = 0 is calculated using the transposed matrix of the previous check matrix H. If the syndrome generated by this calculation is 0, K-bit primary estimated words (C1,..., CK) are output as estimated words (step SP9).

On the other hand, if the generated syndrome is different from 0, it is determined whether the loop count is the maximum value (step SP7).
That is, the number of generations of the primary estimation word is counted, and when the generation number reaches a predetermined maximum number, the calculation for the code is stopped and the currently generated primary estimation word is output as a decoded word. (Step SP9). This prevents unnecessary processing time from being spent on noisy codes with poor convergence.

If it is determined in step SP7 that the loop count has not reached the maximum value, the loop count is incremented by 1 (step SP8), and processing in the row processing section 27 and the column processing section 28 is further started. The process from step SP2 is executed again.
The series of processes by the error correction decoding unit 25 is the operation of the LDPC sum product decoding method (including the minisum decoding method which is a simplified format).

[Configuration of Log Likelihood Calculation Unit]
FIG. 6 is a configuration diagram of the log likelihood ratio calculation unit 24 according to the first embodiment.
As shown in FIG. 6, the log likelihood ratio calculation unit 24 of the present embodiment includes an oversampling unit 42, a phase specifying unit 44, a reference table 45, a bit division unit 46, a synchronization control unit 47, a likelihood. A degree determining unit 48, a communication path monitoring unit 49, and a scaling unit 50 are included.

Among these, the oversampling unit 42 is a signal (voltage) output from the LA 23 at a frequency P times (P is a natural number of 2 or more) the transmission frequency (1 / T) of the transmission signal transmitted from the home-side device 2. For example, a sampling circuit constituted by a plurality of comparators. Here, T represents the transmission cycle of the transmission signal.
When oversampling is performed at P times the signal frequency of the transmission signal, there are (P + 1) sampling points including the reception time point, and when viewed in terms of period, the transmission period T of the transmission signal is divided by P. The value is sampled at the sampling cycle.

Therefore, in the following, oversampling the transmission frequency (1 / T) by P times (P is a natural number of 2 or more) will be referred to as P times oversampling of the transmission period T.
Note that the sampling interval in the oversampling unit 42 is normally set at a constant interval, but may be a non-uniform interval in which the intervals between the samplings are different.
The oversampling unit 42 outputs “1” when the value sampled at the predetermined sampling frequency is equal to or higher than a predetermined threshold (eg, “0”), and “0” when the sampled value is less than the predetermined threshold. Is output. Therefore, the oversampling unit 42 outputs (P + 1) -bit data for one bit of transmission data transmitted from the home side device.

The phase specifying unit 44 includes a circuit that detects a point at which the value of binary data sequence (binary data of 0 or 1) sequentially output from the oversampling unit 42 changes. According to the detection result, the phase change point of the binary data string is specified, and this phase change point is notified to the bit dividing unit 46.
In accordance with the phase change point notified from the phase specifying unit 44, the bit dividing unit 46 performs grouping using P bits in the continuous binary data string output from the oversampling unit 42 as one unit.

Therefore, the grouped data becomes sampling data obtained by dividing a continuous binary data sequence sequentially output from the oversampling unit 42 in accordance with the period T of the received signal.
The likelihood determination unit 48 is based on the grouped P-bit sampling data yn output from the bit division unit 46, and the log likelihood ratio λn indicating the probability of the received signal with respect to the sampling data yn and the transmission signal xn. Is used to determine the log likelihood ratio λn for each sampling data yn.

Here, FIG. 10A shows the eye pattern of the received signal when oversampling is performed at four times the transmission cycle T (when P = 5). In this case, the values of the grouped sampling data yn are yn = (a, b, c, d, e) in time series order, and a to e each take a value of “0” or “1”.
Therefore, in the present embodiment, when the reference table 45 is created, the probabilities P0 and P1 at which the respective sampling data yn appear with respect to the transmission signal xn (= 0 or 1) are observed in advance for a specific communication path through experiments or the like. To do. Note that P0 is the probability of becoming yn when xn = 0, and P1 is the probability of becoming yn when xn = 1.

Then, as shown in FIG. 10B, based on the observed probabilities P0 and P1, the log likelihood ratio λn for each sampling data yn is calculated as λn (yn) = ln (P0 / P1), and this yn And λn are stored in the memory as a reference table 45 in advance.
Therefore, for example, as shown in FIG. 11 (a), when P = 5, the sampling data yn has a total power of 2 in the fifth power of 2, and there are usually 32 logarithms for each of the 32 yn. A reference table 45 for assigning the degree ratio λn (yn) is required.

However, in this embodiment, by setting the sampling timing in the oversampling unit 42 so that the logarithmic likelihood ratio λn is symmetrical, the table size of the reference table 45 is reduced as shown in FIG. Halved.
That is, in the reference table 45 of the present embodiment, log likelihood ratios λn corresponding to a total of 16 sampling data yn (= (00000) to (01111)) in the fourth power of 2 are described. The log likelihood ratio λn corresponding to the sampling data yn (= (10000) to (11111)) is not described.

  Therefore, the likelihood determining unit 48 determines the log likelihood ratio λn using the reference table 45 as it is for the description data yn described in the reference table 45, and sampling that is not described in the reference table 45. For the data yn (= (10000) to (11111)), the description data yn corresponding to the undescribed data yn is specified, and the log likelihood ratio λn corresponding to the description data yn is determined from the reference table 45. It comes to pull out.

The reason why the log likelihood ratio λn has symmetry will be described later.
The likelihood determining unit 48 further quantizes the log likelihood ratio λn determined using the reference table 45 into an R-bit value and outputs the result.

Returning to FIG. 6, the synchronization control unit 47 uses the data of the central bit in each group output from the bit dividing unit 46 (for example, the third data when there are five values in the group), The position of the synchronization bit of S bits in the encoded data of bits is detected, and the position of the start bit of the encoded data of L bits is notified to the error correction decoding unit 25.
The error correction decoding unit 25 receives the start position of the L-bit encoded data notified from the synchronization control circuit 47, and sets the log likelihood ratio λn sent from the likelihood determining unit 48 via the scaling unit 50 to L The error correction decoding is performed using N (= L−S) log likelihood ratios λn, which are divided into pieces and excluding S log likelihood ratios λn corresponding to the synchronization bits.

The communication channel monitoring unit 49 constantly calculates a bit error rate (BER) based on the correction result performed by the error correction decoding unit 25, thereby constantly monitoring the communication channel state. Further, the communication channel monitoring unit 49 obtains a signal-to-noise ratio (SNR) of the communication channel based on the bit error rate, and outputs this signal-to-noise ratio to the scaling unit 50.
The scaling unit 50 determines whether or not to multiply the log likelihood ratio λn from the likelihood determining unit 48 by a predetermined magnification based on the signal-to-noise ratio received from the communication path monitoring unit 49, and to multiply Multiplies the log likelihood ratio λn by the predetermined magnification and outputs the result to the error correction decoding unit 25.

The relationship between the signal-to-noise ratio and the log likelihood ratio λn will be described later.
However, when the error correction decoding unit 25 performs error correction decoding on the encoded data composed of the LDPC code based on, for example, the minisum decoding method, which is a simplified sum product decoding method, the log likelihood ratio λn is set. Scaling does not affect the decoding result.
Therefore, in this case, the scaling unit 50 does not perform scaling with respect to the log likelihood ratio λn, and the log likelihood ratio λn determined by the pair likelihood determination unit 48 is an error as it is regardless of fluctuations in the signal-to-noise ratio. The data is input to the correction decoding unit 25.

[Specific example 1 of bit division]
FIG. 7 is a time series diagram showing an example of the relationship between oversampling and bit division.
As shown in FIG. 7, when P = 4, the oversampling unit 42 samples the signal (voltage) output from the LA 23 at a frequency four times the transmission frequency of the transmission signal, and the sampling value is 0 or more. “1” is output to “0”, and “0” is output when the sampling value is less than zero.

The phase specifying unit 44 detects a point at which the value of the bit string (0 or 1 binary data) sequentially output from the oversampling unit 42 changes. In the example of FIG. 7, the values of the bit strings change between e2 and a3, between e3 and a4, and between e4 and a5, respectively.
Therefore, in this case, the phase specifying unit 44 notifies the bit dividing unit 46 of a phase change point between eX and eY (X is an arbitrary number, Y = X + 1).
Then, the bit division unit 46 divides the continuous bit string output from the oversampling unit 42 between aX and eX, and groups the five bits as one unit.

[Specific example 2 of bit division]
FIG. 8 is a time-series diagram showing another example of the relationship between oversampling and bit division.
In the case of FIG. 8, the value of the bit string changes between a3 and b3, between e3 and a4, and between e4 and a5. In this case, the phase specifying unit 44 sets the most probable point as the phase change point.
That is, the phase specifying unit 44 sets a phase change point between aX and bX (X is an arbitrary number), or a phase change point between eX and aY (X is an arbitrary number, Y = X + 1). Decide whether to do so by majority vote. As a result, the interval between eX and aY is determined as the phase change point and notified to the bit division unit 46.

  According to the station side device (reception device) 1 of the first embodiment having the above configuration, the received signal is oversampled, the data at each sampling point is binary-determined, and the binary data string obtained by this determination is received. Based on the sampling data yn divided in accordance with the signal, the log likelihood ratio λn is determined using the reference table 45 that defines the relationship between the data yn and the log likelihood ratio λn, and Since error correction decoding is performed on the received signal based on the determined log likelihood ratio λn, the received signal is softened without assuming a communication path and without using an AD converter that outputs a plurality of bits. Decision decoding can be performed.

Further, according to the station side device (receiving device) 1 of the first embodiment, the likelihood determining unit 48 uses the reference table 45 that predefines the relationship between the sampling data yn and the log likelihood ratio λn, and uses the log likelihood. Since the frequency ratio λn is determined, the log likelihood ratio λn can be determined accurately in a short time, and the decoding speed can be improved.
Furthermore, according to the station side device (receiving device) 1 of the first embodiment, the reference table 45 is set by setting the sampling timing in the oversampling unit 42 so that the result of the log likelihood ratio λn is symmetric. Since the necessary table size is reduced, the memory of the receiving device 1 can be used effectively, and the observation time for creating the reference table 45 can be shortened.

  Further, according to the station side device (receiving device) 1 of the first embodiment, when the error correction decoding unit 25 performs decoding based on the sum product decoding method, the log likelihood ratio determined by the likelihood determining unit 48 is used. Since λn is scaled by the scaling unit 50, a single reference table 45 can decode a plurality of received signals with a signal-to-noise ratio, and the decoding performance of the receiving apparatus 1 can be further improved. .

[Second Embodiment]
FIG. 9 is a configuration diagram of the log likelihood ratio calculation unit 24 according to the second embodiment.
As illustrated in FIG. 9, the log likelihood ratio calculation unit 24 of the present embodiment includes an oversampling unit 42, a phase specifying unit 44, a reference table 45, a bit division unit 46, a synchronization control unit 47, a likelihood. It is the same as that of the first embodiment (FIG. 6) in that the degree determination unit 48, the communication path monitoring unit 49, and the scaling unit 50 are included.
The log likelihood ratio calculation unit 24 of the present embodiment (FIG. 9) is different from that of the first embodiment (FIG. 6) in that it further includes an appearance frequency measurement unit 51 and a sampling rate control unit 52.

The appearance frequency measuring unit 51 constantly monitors the appearance frequency of the sampling data yn (= (00000) to (11111)) output from the bit dividing unit 46, and thereby measures the distribution tendency generated in the received signal. .
The distribution tendency of the received signal is input to the communication path monitoring unit 49. Based on the bit error rate and signal-to-noise ratio obtained by the communication channel monitoring unit 49 and the distribution tendency of the received signal from the appearance frequency measuring unit 51, the change in the distribution trend of the received signal becomes the change in the state of the transmission channel. If it is determined whether the transmission path state is the cause, a rate change signal is output to the sampling rate control unit 52.

When the sampling rate control unit 52 receives the rate change signal from the communication path monitoring unit 49, the sampling rate control unit 52 changes the sampling rate in the oversampling unit 42.
For example, the sampling rate control unit 52 increases the number of samplings in the oversampling unit 42 when the reception level in the station side device 1 becomes low and the state of the transmission path deteriorates. When the reception level at 1 becomes high and the state of the transmission path is improved, the number of samples in the oversampling unit 42 is decreased.

Thus, according to the station side apparatus (reception apparatus) 1 of 2nd Embodiment, since the sampling rate control part 52 changes the sampling number in the oversampling part 42 according to the state of a transmission line, transmission line Even if the trajectory of the received signal changes more complicatedly with the state change, the received signal can be appropriately decoded by changing the sampling number.
Therefore, a function as an equalizer, that is, a function as an adaptive likelihood block is realized for the receiving apparatus 1 that performs decoding using the log likelihood ratio λ determined by oversampling the received signal. can do.

[Symmetry of log-likelihood ratio]
FIG. 12A shows sampling data yn (horizontal axis) and log-likelihood ratio λn (vertical) when oversampling shown in FIG. 10A is performed on a received signal assuming an additive white Gaussian channel. It is a graph which shows the result of trial experiment by the numerical calculation about the relationship with an axis | shaft.
As shown in the graph of FIG. 12A, when the log likelihood ratio λn is specified based on the sampling data yn of a considerable number of observations, the graph of the log likelihood ratio λn has a center point C Symmetric with respect to.

  As shown in FIG. 10A, the symmetry of the log-likelihood ratio λn is such that an upward convex curve S1 within the range of the transmission period T and a downward convex curve S2 as a corresponding curve. This occurs when the appearance frequencies are equal (when the value on the vertical axis is axisymmetric with respect to the horizontal axis). Further, the symmetry of the log likelihood ratio λn has been found from a detailed observation with the bit pattern of yn. That is, the graph of the log likelihood ratio λn is point-symmetric only after using a considerable number of observations.

Therefore, even if yn is sampled by the number of observations which is not necessarily a statistically significant number, and the reference table 45 is created based on this sampling, the value of λn is corrected so as to be point-symmetric from the table 45. Thus, the observation time can be shortened.
Further, by shortening the observation time, it is possible to improve the followability to the fluctuation even when the communication path fluctuates. Further, if the symmetry of the log likelihood value λn is used, the table size of the reference table 45 can be reduced as described above.

[Relationship between signal-to-noise ratio and log-likelihood ratio]
FIG. 12B shows a graph of FIG. 12A, which is a result of a trial experiment by numerical calculation of the relationship between the sampling data yn (horizontal axis) and the log likelihood ratio λn (vertical axis). 3 is a three-dimensional graph showing the results obtained for the noise to noise ratio (SNR).
As shown in FIG. 12B, the likelihood value at a certain signal-to-noise ratio (for example, SNR = 6 dB) is a multiple of the likelihood value of another signal-to-noise ratio (for example, SNR = 3 dB). It became clear by trial experiment.

Therefore, by recording only the likelihood value of a specific SNR in the reference table 45 and multiplying the likelihood value by several times (scaling) according to the SNR, one table value for a plurality of SNRs. Can respond.
Note that a fixed magnification may be used as the scaling multiple. In this case, by recording this multiple value in the reference table 45 of the log likelihood ratio λn, it is commonly used for all SNRs. be able to.

[Correction effect]
FIG. 13 is a graph showing a correction effect when error correction decoding is performed by obtaining a log likelihood ratio by hard decision by oversampling.
The decoding method in this case is the sum product decoding method. Further, in the conventional method, decoding is performed with a log likelihood ratio calculated using a conventional multi-value output AD converter, and in the oversampling method, the received signal is oversampled at four times the transmission period T (FIG. 10A). ) To calculate the log likelihood ratio.

As shown in FIG. 13, even if the decoding process by the sum product decoding method is performed using the log likelihood ratio based on the oversampling method of the present invention, the correction effect is only degraded by about 2 dB compared to the conventional method, It can be seen that it can withstand practical use.
Therefore, even in the case of high-speed data transmission where it is difficult to implement an AD converter, soft decision decoding of an LDPC code is possible by determining the log likelihood ratio based on the oversampling method of the present invention as an alternative method. It becomes.

[Minisum decoding method]
Incidentally, when the decoding process by the error correction decoding unit 25 is the minisum decoding method, it is known that the scaling process regarding the likelihood does not affect the decoding result as shown in FIG. For this reason, in the original sum product decoding method, normalization by noise information corresponding to the SNR is required, but in the minisum decoding method, such normalization is not necessary.
In other words, in the case of the minisum decoding method, the likelihood value for a specific SNR is used as the reference table 45, which means that it can be used for all SNRs. For this reason, it is very compatible with the fixed magnification reference table 45 described above.

FIG. 15 is a graph of experimental results demonstrating the effect of SNR differences on the correction effect.
In FIG. 15, a likelihood table is created based on the log-likelihood ratio when SNR = 3 dB obtained by the combination of 4 times oversampling (FIG. 10A) and hard decision, and the value is set to 0 for each. The behavior of the correction capability is illustrated for the cases of .5 times, 1.0 times, and 2.0 times. In addition, it is not made to change according to SNR.
As is apparent from FIG. 15, the minisum decoding method is not affected at all by scaling, and conversely in the case of the sum product decoding method, when a scaling value of about twice is used, the AD converter is used. It can be seen that the characteristics are asymptotic when using.

[About changing the sampling rate]
FIG. 16A shows an eye pattern of the received signal when the sampling timing is intentionally shifted from the Nyquist sampling point. FIG. 17 is a graph showing the result of verifying the correction capability in that case.
As can be seen from FIG. 17, soft decision decoding by LDPC may be possible without including Nyquist sampling points as sampling points when the received signal is oversampled.

For example, as shown in FIG. 16B, the transmission signal xn is “0” in the case of the eye pattern along the solid line arrow, and the transmission signal xn in the case of the eye pattern along the broken line arrow. Can be estimated as “1”.
That is, if a plurality of sampling points are determined along the trajectory of the eye pattern, the value of the transmission signal xn can be estimated approximately deterministically, and the likelihood by oversampling can be obtained. The graph of FIG. 17 shows the result of creating a likelihood table at such sampling timing and performing error correction.

The reception situation shown in FIG. 16 (a) is a situation where the reception signal has received interference from previous and subsequent data, and an equalizer is essentially required. The reason that reception is possible even at such a shifted timing is that the likelihood is calculated based on the trajectory that the received signal can take and its appearance frequency.
For this reason, if the transmission path becomes more complex, the trajectory that the received signal can take is expected to become more complicated. However, by changing the oversampling number together with the transmission path according to the fluctuation state, the transmission path The decoding capability can be made to follow the state fluctuation of

That is, by incorporating adaptive control of the sampling rate into the receiving device 1, it is possible to realize an adaptive likelihood block function that can more flexibly cope with the transmission path.
In the log likelihood ratio calculation unit 24 (FIG. 9) of the second embodiment described above, the sampling rate control unit 52 that changes the sampling number in the oversampling unit 42 corresponding to the state of the transmission path is provided. This is based on this viewpoint.

  The above embodiment is an exemplification of the present invention and does not limit the present invention. The technical scope of the present invention is specified by the scope of claims for patent, and includes all modifications that are equivalent in scope and configuration. That is, the present invention includes the following modifications, for example.

[Modification of the present invention]
(1) Synchronization in the oversampling unit In the above embodiment, each sampling of the oversampling unit 42 has been described on the assumption that synchronization is established. However, sampling is performed after synchronization is established using a synchronization bit. You may decide.
(2) Likelihood In the above embodiment, the likelihood of the received signal is represented by the log likelihood ratio λn. However, the likelihood referred to in the present invention is not limited to the logarithmic expression.

(3) Error correction decoding method In the above embodiment, transmission data is LDPC-encoded on the transmission side, and reception data is decoded on the reception side by the sum product decoding method or minisum decoding method. However, the present invention is not limited to this. is not. For example, decoding methods such as Normalized-BP, Offset-BP, Δ-Min, and FUMP-APP can be adopted as other decoding methods that simplify the sum product decoding method.
In addition, the present invention is not limited to LDPC, and encoding may be performed with a turbo code on the transmission side, and error correction decoding for decoding the turbo code may be used on the reception side.

(4) Communication System In the above embodiment, the PON system is exemplified as the communication system. However, the present invention can be applied to a mobile communication system, WiMAX, and other wireless communication systems.

It is a topology diagram showing a configuration example of a communication system suitable for implementing the present invention. It is a block diagram of a home side apparatus (transmission apparatus). It is a block diagram of a station side apparatus (receiving apparatus). It is a block diagram of an error correction decoding part. It is a flowchart which shows the processing operation of a loop determination part. It is a block diagram of the log likelihood ratio calculation part which concerns on 1st Embodiment. It is a time series figure which shows an example of the relationship between oversampling and bit division. It is a time series figure which shows the other example of the relationship between oversampling and bit division. It is a block diagram of the log likelihood ratio calculation part which concerns on 2nd Embodiment. (A) is an eye pattern of a received signal when oversampling is performed at four times the transmission cycle (when P = 5), and (b) is a calculation formula showing a method of calculating a log likelihood ratio. (A) is a figure which shows an example of a reference table, (b) is a figure which shows the other example of a reference table. (A) is a graph which shows the result of the trial experiment by numerical calculation about the relationship between sampling data (horizontal axis) and logarithmic likelihood ratio (vertical axis), (b) is a graph of (a) 3 is a three-dimensional graph showing the results obtained for the signal-to-noise ratio (SNR). It is a graph which shows the correction effect at the time of calculating | requiring a log likelihood ratio by the hard decision by oversampling, and performing error correction decoding. It is a figure which shows the influence which it has on a temporary estimated word at the time of dividing or multiplying each variable used by a decoding process. It is a graph of the experimental result which demonstrates the influence which the difference of SNR has on the correction effect. (A) is an eye pattern of the received signal when the sampling timing is intentionally shifted from the Nyquist sampling point, and (b) is a diagram showing a relationship between the eye pattern and the transmission signal. It is a graph which shows the result of having verified correction ability at the time of sampling at the timing of Drawing 16 (a).

Explanation of symbols

1: Station side device 2 (2a to 2j): Home side device 3: Splitter 4 (4a to 4d): Splitter 5 (5a, 5b): Optical fiber 11: Error correction encoding unit 12: Electro-optical converter 21: Photoelectric converter 22: TIA 23: LA 24: log likelihood ratio calculation unit 25: error correction decoding unit 26: loop determination unit 27: row processing unit 28: column processing unit 42: oversampling unit 44: phase specifying unit 45 : Bit division unit 47: Synchronization control unit 48: Likelihood determination unit 49: Transmission path monitoring unit 50: Scaling unit 51: Appearance frequency measurement unit 52: Sampling rate control unit

Claims (5)

An oversampling unit for determining data of each sampling point obtained by oversampling the received signal by binary and sequentially outputting a binary data string;
A reference table defining a relationship between sampling data obtained by dividing the binary data sequence corresponding to the received signal and likelihood indicating the likelihood of the received signal with respect to a transmission signal;
A likelihood determining unit that determines the likelihood using the reference table;
And a decoding unit that performs error correction decoding on the received signal based on the determined likelihood.   The receiving apparatus according to claim 1, wherein a sampling timing in the oversampling unit is set so that symmetry appears in the likelihood result corresponding to each sampling data.   The scaling part which scales the said likelihood determined by the said likelihood determination part is provided in order to decode the said received signal of several signal-to-noise ratio with one said reference table. The receiving device described. The decoding unit performs error correction decoding based on a decoding scheme obtained by simplifying the sum product decoding method for encoded data composed of LDPC codes,
The receiving apparatus according to claim 1, wherein the likelihood determining unit determines the likelihood based on one reference table for the received signals having a plurality of signal-to-noise ratios.   The transmission line monitoring part which monitors the state of a transmission line, and the sampling rate control part which changes the sampling number in the said oversampling part corresponding to the state of the said transmission line are provided. The receiving device according to item.

Images