Symbol determination device and symbol determination method

Abstract

This symbol determination device comprises: a provisional determination unit that performs provisional determination using a determination threshold value for determining that a symbol sequence has a value close to the nearest symbol; a transmission path estimation unit that generates an estimated reception symbol sequence for each transmission path state on the basis of a plurality of symbol sequences indicating transmission path states and an estimated transfer function of a transmission path; a sequence estimation algorithm processing unit that selects, by a predetermined estimation algorithm, a most likely estimated reception symbol sequence on the basis of a symbol sequence obtained from a reception signal sequence, a metric obtained on the basis of each of the estimated reception symbol sequences, a provisionally determined provisional determination symbol, and a symbol in the vicinity of the provisional determination symbol determined to be used; and a path traceback determination unit that determines a transmission symbol sequence by tracing back a trellis path on the basis of the most likely estimated reception symbol sequence. The sequence estimation algorithm processing unit generates the plurality of symbol sequences indicating the transmission path states within a range of the provisional determination symbol and the symbol in the vicinity of the provisional determination symbol.　

Description

Symbol determination device and symbol determination method

[0001] The present invention relates to a symbol determination device and a symbol determination method.

[0002] In recent years, the rapid proliferation of smartphones and tablets, along with the increase in rich content such as high-definition video streaming services, has led to a continuous increase in the traffic transmitted through the internet's backbone network. Furthermore, the use of cloud services by businesses is also advancing. As a result, it is predicted that network traffic within and between data centers (hereinafter referred to as "DCs") will increase at an annual rate of approximately 1.3 times.

[0003] Currently, Ethernet (registered trademark) is primarily used for connections within and between data centers (DCs). With the increasing communication traffic, it is anticipated that scaling up DCs at single locations will become increasingly difficult. Therefore, the need for inter-DC collaboration will become even greater, and a further increase in traffic transmitted and received between DCs is expected. To address this situation, the establishment of low-cost, high-capacity short-distance optical transmission technology is required.

[0004] Under the current Ethernet® standard, optical fiber communication is used for transmission lines up to 40 km, with the exception of 10 GbE (Gigabit Ethernet®)-ZR. Furthermore, up to 100 GbE, intensity modulation is used, assigning binary information to the on / off states of light. The receiving side consists only of a photodetector, resulting in a less expensive configuration than the coherent receiving method used for long-distance transmission.

[0005] In 100GbE, a transmission capacity of 100Gb / s is achieved by multiplexing four NRZ (Non-return-to-zero) signals with a modulation speed of 25GBd (GigaBaud) and an information content of 1 bit / symbol per symbol.

[0006] In the standardization of 400GbE, the next generation after 100GbE, the 2-bit / symbol PAM4 (4-level pulse-amplitude-modulation) was adopted for the first time, taking into consideration the maintenance of the economical device configuration used in 100GbE and the efficiency of signal bandwidth utilization. As a result, 400GbE achieves a transmission capacity of 400Gb / s by multiplexing eight 50Gb / s signals. Examples of 400GbE standards include 400GBASE-FR8 and LR8.

[0007] To accommodate further increases in traffic in the future, standardization of 800 GbE and 1.6 TbE is planned from 2020 onwards. These communication speeds are planned to be achieved, for example, by employing 200 Gb / s PAM4 with a modulation speed of 100 Gbaud and multiplexing signals across 4 to 8 wavelengths, as shown in Figure 15, or by employing 200 Gb / s PAM8 with a modulation speed of 75 Gbaud and multiplexing signals across 4 to 8 wavelengths.

[0008] As a challenge to further increasing transmission capacity, signal quality degradation occurs with increasing transmission capacity, specifically the effects of device bandwidth limitations and chromatic dispersion. For example, as shown in Figure 16, when transmission capacity increases and the usable bandwidth increases, a problem arises where the frequency domain 501 (hatched area with diagonal lines) is lost due to device bandwidth limitations. Also, as shown in Figure 17, as transmission capacity increases, the effects of chromatic dispersion become greater, and the interfering region 502 increases.

[0009] Methods to solve these problems include using DACs (Digital to Analog Converters) or ADCs (Analog to Digital Converters) that support high communication speeds, or using dispersion compensation modules to compensate for wavelength dispersion. However, such equipment is expensive, and the cost of the equipment is high, so from an economic standpoint, these methods are to be avoided. From an economic standpoint, the desired method is to maintain the configuration of conventional transceivers and improve multi-level control, bandwidth limiting tolerance, and wavelength dispersion tolerance, and utilize low-cost narrowband devices. However, when using low-cost narrowband devices, it is necessary to solve the problems caused by bandwidth limiting and intersymbol interference due to wavelength dispersion that arise with the increased communication speeds mentioned above.

[0010] Maximum Likelihood Sequential Estimation (MLSE) is known as the most effective equalization method for obtaining correct transmission data from received signal waveforms distorted by problems such as intersymbol interference (see, for example, Non-Patent Document 1).

[0011] For example, Figure 18 is a block diagram of a conventional communication system 100 configured using the low-cost narrowband device described above. The communication system 100 comprises a transmitting signal generator 1, a transmission line 2, and a receiving symbol determination device 90. The signal generator 1 takes in an m-value data signal provided from an external source and generates an electrical signal transmission signal sequence {s ｔ This generates {s}. Here, m is the symbol multi-level index and is an integer greater than or equal to 2. Also, t is an identification number that identifies the transmitted signal sequence, and the transmitted signal sequence {s} is generated. ｔ If the number of symbols contained in} is N, then integer values ​​such as 1, 2, 3, ..., N are assigned to them.

[0012] The intensity modulator 2-2 of the transmission line 2 transmits the electrical signal sequence {s} output by the signal generator 1. ｔ The} is captured, and the transmitted signal sequence of the captured electrical signal {s ｔ The light emitted from light source 2-1 is modulated by}, and the transmission signal sequence of the optical signal {sｔ}. The optical fiber 2-3 transmits the transmission signal sequence {s ｔ} of the optical signal generated by the intensity modulator 2-2. The light receiver 2-4 receives the transmission signal sequence {s ｔ} of the optical signal transmitted by the optical fiber 2-3 as a reception signal sequence {r ｔ} of the optical signal, converts it into a reception signal sequence {r ｔ} of an electric signal, and outputs the converted sequence.

[0013] At this time, when the transmission path 2 is represented by an equalization circuit, it has a configuration as shown in FIG. 19. In FIG. 19, delay units 82-1 to 82-2L capture and store input symbols, and output the stored input symbols after a time "T" has elapsed. Here, "T" is the symbol interval, and the calculation timing for each symbol is "tT".

[0014] The delay unit 81 captures and stores an input symbol, and outputs the stored input symbol after a time "-LT" has elapsed. Since the delay amount is marked with a negative sign, the delay unit 81 provides a negative delay of "LT". Here, in the transmission path 2, assuming that intersymbol interference occurs for L symbols before and after a code at time t, the delay unit 81 processes the transmission signal sequence {s ｔ} element s ｔ for L symbols before and after to the transfer function unit 83.

[0015] The transfer function unit 83 applies a transfer function (H) to the symbol sequences output from the delay unit 81 and the delay units 82-1 to 82-2L. The adder 85 adds the noise component ω ｔ to the output value of the transfer function unit 83, thereby generating the reception signal sequence {r ｔ}. ω ｔ is a mutually independent Gaussian random sequence with a mean of 0 and variance δ ２ .

[0016] The reception signal sequence {r ｔ} generated by the equalization circuit in FIG. 19 is expressed by the following formula (1). In formula (1), t=1, 2, ..., N.

[0017]

[0018] Figure 20 shows the received signal sequence {r} output by transmission line 2. ｔ This diagram shows the internal configuration of the symbol determination device 90, which is an identification circuit that identifies the transmitted symbol sequence by MLSE based on {r}. The transfer function (H) in the transmission line 2 is an unknown function. Therefore, the symbol determination device 90 estimates the transfer function (H) of the transmission line 2 and generates a replica of the received symbol sequence using the estimated transfer function (H') (hereinafter referred to as "estimated transfer function (H')"). Hereinafter, the received symbol sequence replicated by the estimated transfer function (H') will be called the estimated received symbol sequence. The symbol determination device 90 then compares the generated estimated received symbol sequence with the received signal sequence {r}. ｔ The symbol sequence obtained from} is compared with the most likely estimated received symbol sequence, and the result is determined to be the result of the determination.

[0019] In MLSE, the conditional joint probability density function p Ｎ ( {r Ｎ} {s' Ｎ Transmit signal sequence {s'} that maximizes}) ｔ Symbol determination is performed by searching for}. Conditional joint probability density function p Ｎ ( {r Ｎ} {s' Ｎ}) is a transmission signal sequence {s'} of sequence length N generated from m-value data through transmission line 2. ｔ When} is transmitted, the received signal sequence {r ｔ The probability of receiving} is expressed by the following equation (2).

[0020]

[0021] Conditional joint probability density function p Ｎ ( {r Ｎ} {s' Ｎ To maximize}), the distance function d shown in equation (3) is obtained. Ｎ This is equivalent to minimizing the value. Note that in equation (3), (p-1) / 2 = L.

[0022]

[0023] (s' in equation (3) ｔ－（ｐ－１）／２ ,…, s' ｔ ,…, s' ｔ＋（ｐ－１）／２) is the state of transmission path 2 at time t μ ｔ (Hereafter referred to as "Transmission Path Condition μ") ｔ This indicates that when the symbol sequence length is "p", the modulation symbol I = [i １ , i ２ , ..., i ｍ The total number of combinations of ] is "m ｐ In this case, transmission line 2 is m ｐ It can be considered as a finite state machine with a finite number of transmission path states. Since it can be considered as a finite state machine, for example, a maximum likelihood sequence estimation method such as the Viterbi algorithm can be used to estimate the received signal sequence {r ｔ Each step of the calculation is performed to calculate the distance function d. Ｎ It is possible to calculate this.

[0024] At time t, the transmission path state μ ｔ Distance function d to reach ｔ ( {μ ｔ}) is the distance function d at time t-1. ｔ－１ ( {μ ｔ－１}) and the likelihood associated with the state transition at time t, i.e., metric b(r ｔ μ ｔ－１ →μ ｔ It can be expressed by the following equation (4) using ).

[0025]

[0026] metric b(r ｔ μ ｔ－１ →μ ｔ ) is expressed as equation (5) below using the estimated transfer function (H').

[0027]

[0028] The metric b at time t depends only on the state transition from t-1 to t, and not on any previous state transitions. Here, the transmission path state μ ｔ The minimum value of the distance function that reaches [the specified value] is d_min. ｔ－１ (μ ｔ－１ ) and the total state transitions corresponding to this are all transmission path states μ at time t-1. ｔ－１ Assume that this is known in [location].

[0029] Under this assumption, the transmission path state μ ｔ Distance function d to reach ｔ ( {μ ｔ When finding the minimum value of}), the distance function d corresponds to all state transitions. ｔ ( {μ ｔ There is no need to determine the transmission path state μ. ｔ All transmission path states {μ} that can transition to ｔ－１ Regarding}, d_min ｔ－１ (μ ｔ－１ ) + b(r ｔ μ ｔ－１ →μ ｔ ) Calculate the transmission path state μ, and if you find the minimum value among them, that value will be the transmission path state μ ｔ All distance functions d that reach ｔ ( {μ ｔ The minimum value of}) is d_min ｔ (μ ｔ This can be expressed as equation (6) below.

[0030]

[0031] Returning to Figure 20, in the symbol determination device 90, the delayers 92-1 to 92-(p-1) capture and store the input symbol, and output the stored input symbol after a time interval of "T". Therefore, the estimated transfer function unit 93 receives the transmission path state μ of the transmission path 2 at time t. ｔ (s') ｔ－（ｐ－１）／２ ,…, s' ｔ ,…, s' ｔ＋（ｐ－１）／２ A symbol sequence of {r} is given. The estimated transfer function unit 93 applies the estimated transfer function (H') to this symbol sequence. The subtractor 94 receives the signal sequence {r}. ｔ The output value of the estimated transfer function unit 93 is subtracted from}. The absolute value unit 95 calculates the absolute value of the output value of the subtractor unit 94, and the calculated absolute value becomes the metric b shown in equation (5).

[0032] The addition comparison selection unit 91 determines the transmission path state μ ｔ All transmission path states {μ} that can transition to ｔ－１ Regarding}, in equation (6) d_min ｔ－１ (μ ｔ－１ ) + b(rｔ ; μ ｔ－１ →μ ｔ ), calculate the minimum value among the calculated values as the distance function d ｔ ({μ ｔ}), which is d_min, the minimum value ｔ (μ ｔ ).

[0033] The path backtracking determination unit 96 backtracks the trellis path of the Viterbi algorithm based on the minimum value of the distance function d ｔ ({μ ｔ}), obtains an estimated value of the m-value data acquired by the signal generation device 1, and outputs the obtained estimated value as a determination result.

[0034] As described above, in the Viterbi algorithm, when obtaining the minimum value of the distance function d ｔ that reaches the transmission channel state μ ｔ ({μ ｔ}), instead of obtaining the distance functions d ｔ ({μ ｔ}) corresponding to all state transitions, for all transmission channel states {μ ｔ that can transition to the transmission channel state μ ｔ－１}, d_min ｔ－１ (μ ｔ－１ ) + b(r ｔ ; μ ｔ－１ →μ ｔ ) is calculated. Accordingly, in the Viterbi algorithm, the amount of computation that increases exponentially with the sequence length can be suppressed to a linear increase. Therefore, in MLSE, by using the Viterbi algorithm, maximum likelihood sequence estimation with a reduced amount of computation is enabled.

[0035] Furthermore, in MLSE, high equalization performance is achieved by estimating and reproducing intersymbol interference imposed on a transmission signal waveform through digital signal processing on the receiver side. Therefore, the higher the estimation accuracy, the more it can suppress code errors caused by intersymbol interference, making it possible to obtain correct transmission data from a reception signal waveform distorted by intersymbol interference. Signal quality degradation suppression technology using MLSE is also being studied for the increase in capacity of the aforementioned Ethernet (registered trademark).

[0036] As described above, the Viterbi algorithm performs the operation shown in equation (6) above. If we set the memory length (constraint length) to "3", the combinations of the operations to be performed in equation (6) can be shown in the trellis diagram as shown in Figure 21. For example, in the case of the top edge, [i １ i １ i １ ] is (s' in equation (5) ｔ－（ｐ－１）／２ ,…, s' ｔ ,…, s' ｔ＋（ｐ－１）／２ ) corresponds to m as shown in Figure 21. ２ The transmission path state μ ｔ Since m branches grow from each of them, the total number of branches is m ３ It will become a book.

[0037] For example, if the memory length is "3" and m = 8, the transmission path state μ is as shown in Figure 22. ｔ The number of is "64", and the number of branches is "512". In the Viterbi algorithm, the transmission path state μ ｔ The metric calculation needs to be performed for each branch of the trellis. Therefore, as the number of m, i.e., the symbolic multi-level index, increases and the number of states increases, the number of trellis branches also increases, leading to a problem where the computational complexity of the Viterbi algorithm increases.

[0038] To solve the problems described above, several improved Viterbi algorithms have been proposed (see, for example, Patent Document 1 and Non-Patent Document 2).

[0039] Japanese Patent Publication No. 2020-184696

[0040] DD FALCONER and FR MAGEE, JR. “Adaptive Channel Memory Truncation for Maximum Likelihood Sequence Estimation”, The Bell System Technical Journal Vol. 52, No. 9, pp. 1541-1562, November, 1973.Yukui Yu, Yi Che, Tianwai Bo, Daeho Kim, and Hoon Kim, “Reduced-state MLSE for an IM / DD system using PAM modulation," Opt. Express 28, 38505-38515 (2020).

[0041] The method described in Patent Document 1 above is a Viterbi algorithm based on a trellis diagram that considers a combination of three symbols in total: the judgment symbol and the upper and lower levels. Furthermore, Non-Patent Document 2 proposes a Viterbi algorithm based on a trellis diagram that considers a combination of two symbols in the upper and lower levels near the filter output of a feedforward equalizer (FFE) for outputting the judgment symbol. These methods reduce the number of paths in the trellis diagram used in the Viterbi algorithm, thus reducing the computational complexity of the MLSE. However, if the output of the preceding digital filter F is sufficiently close to the symbol value, calculations considering combinations of multiple neighboring symbols are likely unnecessary. Therefore, there was room for further reduction in computational complexity in the conventional method.

[0042] In view of the above circumstances, the present invention aims to provide a technology that can reduce the amount of computation compared to conventional methods.

[0043] One aspect of the present invention includes: a provisional determination unit that generates a symbol sequence by performing adaptive equalization of a received signal sequence taken from a transmission line using the estimated inverse transfer function of the transmission line, and performs a provisional determination using a determination threshold to determine if the symbol sequence has a value close to the nearest symbol; a transmission line estimation unit that generates an estimated received symbol sequence for each transmission line state based on a plurality of symbol sequences indicating the transmission line state and the estimated transfer function of the transmission line; a symbol sequence obtained from the received signal sequence, a metric obtained based on each of the estimated received symbol sequences, and a provisional determination symbol provisionally determined by the provisional determination unit; The symbol determination device comprises: a sequence estimation algorithm processing unit that selects the most likely estimated received symbol sequence using a predetermined estimation algorithm based on the provisional determination symbol determined to be used by provisional determination using the determination threshold and its neighboring symbols; and a path trace determination unit that determines the transmitted symbol sequence by tracing back the trellis path based on the most likely estimated received symbol sequence, wherein the sequence estimation algorithm processing unit generates a plurality of symbol sequences indicating the transmission path state within the range of the provisional determination symbol and the neighboring symbols of the provisional determination symbol and outputs them to the transmission path estimation unit.

[0044] One aspect of the present invention is a symbol determination method that generates a symbol sequence by performing adaptive equalization on a received signal sequence taken from a transmission line using an estimated inverse transfer function of the transmission line; performs a preliminary determination on the symbol sequence using a determination threshold to determine if it has a value close to the nearest symbol; generates a plurality of symbol sequences indicating the transmission line state within the range of symbols near the preliminary determination symbol determined to be used by the preliminary determination using the determination threshold; generates an estimated received symbol sequence for each of the transmission line states based on the plurality of symbol sequences indicating the transmission line state and the estimated transfer function of the transmission line; selects the most likely estimated received symbol sequence using a predetermined estimation algorithm based on the symbol sequence obtained from the received signal sequence, a metric obtained based on each of the estimated received symbol sequences, the preliminary determination symbol, and symbols near the preliminary determination symbol determined to be used by the preliminary determination using the determination threshold; and determines the transmitted symbol sequence by tracing back the trellis path based on the most likely estimated received symbol sequence.

[0045] This invention makes it possible to reduce the amount of computation compared to conventional methods.

[0046] This is a block diagram showing the configuration of the communication system S in the first embodiment. This is a block diagram showing the internal configuration of the symbol determination device in the first embodiment. This is a block diagram showing the detailed internal configuration of the symbol determination device in the first embodiment. This is a diagram illustrating the sequence captured by the adaptive filter unit in the first embodiment. This is a diagram illustrating the pulse width compression performed by the adaptive filter unit of the sequence estimation unit in the first embodiment. This is a diagram for illustrating the processing by the determination processing unit 302 in the first embodiment. This is a diagram showing an example of a trellis diagram based on the provisional determination result shown in Figure 6. This is a diagram for illustrating the method of setting the determination threshold in the first embodiment. This is a diagram for illustrating the method of setting the determination threshold in the first embodiment. This is a flowchart illustrating the processing flow by the provisional determination unit of the symbol determination device in the first embodiment. This is a flowchart illustrating the processing flow by the sequence estimation unit of the symbol determination device in the first embodiment. This is a block diagram showing the internal configuration of the symbol determination device in the second embodiment. This is a block diagram showing the detailed internal configuration of the symbol determination device in the second embodiment. This is a diagram showing the system configuration during the experiment. This is a diagram showing the experimental results. This is a diagram showing the experimental results. This is a block diagram showing the configuration of a conventional communication system. This is a block diagram of the transmission path equalization circuit. This is a block diagram illustrating the internal configuration of a conventional symbol determination device. This is a trellis diagram showing the state that is the target of calculation by the Viterbi algorithm. Figure 21 shows a trellis diagram when the symbol multi-level index is set to 8.

[0047] One embodiment of the present invention will be described below with reference to the drawings.

[0048] (First Embodiment) Figure 1 is a block diagram showing the configuration of the communication system S in the first embodiment. The communication system S comprises a signal generation device 1, a transmission line 2, and a symbol determination device 3.

[0049] The signal generator 1 and transmission line 2 have the same configuration as the signal generator 1 and transmission line 2 in the conventional communication system 100 shown in Figure 18. The signal generator 1 receives an m-value data signal from an external source. m is the symbol multi-level, and is, for example, an integer of 2 or more. Each symbol is represented by a number or a symbol. For example, when m = 8, each symbol is represented by the numbers [0, 1, 2, 3, 4, 5, 6, 7].

[0050] The signal generator 1 uses the acquired m-value data signal as a transmission symbol sequence, and the transmission signal sequence of the electrical signal including the transmission symbol sequence {s ｔ Generates {s}. t is an identification number that identifies the transmitted signal sequence. Transmitted signal sequence {s ｔ If the number of symbols in} is N, then integer values ​​such as t = 1, 2, 3, ..., N are assigned to them.

[0051] In the transmission line 2, the intensity modulator 2-2 transmits the electrical signal sequence {s} output by the signal generator 1. ｔ The intensity modulator 2-2 receives the signal sequence {s} of the electrical signal. ｔ Based on the symbol sequence of m values ​​contained in {s}, the light emitted by the light source 2-1 is modulated. As a result, the intensity modulator 2-2 transmits the signal sequence {s} of the optical signal representing the symbol sequence of m values. ｔ Generates}.

[0052] The optical fiber 2-3 transmits the signal sequence {s} of the optical signal generated by the intensity modulator 2-2. ｔ} is transmitted.

[0053] The light receiver 2-4 receives the transmission signal sequence {s} of the optical signal transmitted by the optical fiber 2-3. ｔ} is the received signal sequence of optical signals {r ｔ The light is received as {r}. The light receiver 2-4 receives the received signal sequence of the received light signal {r}. ｔ} is the received signal sequence of electrical signals {r ｔ It is converted to} and output. The light receivers 2-4 are, for example, photodiodes.

[0054] The symbol determination device 3 receives the signal sequence {r} output from the transmission line 2. ｔThis is an identification circuit that identifies the transmitted symbol sequence based on {r}. The symbol determination device 3 has the internal configuration shown in Figure 2. The symbol determination device 3 includes a provisional determination unit 30 and a sequence estimation unit 40. For example, an FFE (Feed Forward Equalizer) is applied to the provisional determination unit 30. The provisional determination unit 30 determines the received signal sequence {r} by using a function that estimates the inverse transfer function (hereinafter referred to as the "estimated inverse transfer function"). ｔ The} is adapted and equalized to perform a hard judgment and then a preliminary determination of the transmitted symbol sequence is made.

[0055] The sequence estimation unit 40 applies the estimated transfer function (H') to the symbol sequence representing the transmission path state to generate an estimated received symbol sequence. The sequence estimation unit 40 then compares the generated estimated received symbol sequence with the received signal sequence {r}, which has a compressed pulse width. ｔ The sequence estimation unit 40 calculates a metric based on the above. The sequence estimation unit 40 then uses the calculated metric to execute the Viterbi algorithm within the range of neighboring symbols centered on each of the symbols that the provisional determination unit 30 has provisionally determined. As a result, the sequence estimation unit 40 obtains an estimated value of the transmitted symbol sequence, that is, an estimated value of the m-value data signal acquired by the signal generator 1.

[0056] The provisional determination unit 30 comprises an adaptive filter unit 301, a determination processing unit 302, and an update processing unit 303. The adaptive filter unit 301 is, for example, a linear transversal filter as shown in Figure 3. The adaptive filter unit 301 adaptively equalizes the input signal using an estimated inverse transfer function obtained by estimating the inverse transfer function of the transfer function (H) of the transmission line 2.

[0057] As shown in Figure 3, the adaptive filter unit 301 includes delays 31, 32-1 to 32-(u-1), taps 33-1 to 33-u, and an adder 34. As shown in Figure 4, the delays 31 include a sequence of N received signals {r ｔ The delay unit 31 captures u symbol sequences, which are part of the}. The delay unit 31 stores the captured symbol sequences and outputs them after a time elapsed of (u-1)T / 2, i.e., after a delay of (u-1) / 2 symbols.

[0058] Tap 33-1 contains the r symbol, which is delayed by the (u-1) / 2 symbol output by the delay unit 31. ｔ＋（ｕ－１）／２ It is given.

[0059] Each of the delay units 32-1 to 32-(u-1) takes in and stores one symbol, and outputs the taken symbol after a time interval of "T", i.e., after a delay of one symbol. For example, the first delay unit 32-1 outputs the r delayed by "(u-3) / 2" symbols. ｔ＋（ｕ－３）／２ The symbol is output. The last delay unit 32-(u-1) delays the r by "(u-1) / 2" symbols. ｔ＋（ｕ－１）／２ The symbol is output. As a result, taps 33-1 to 33-u are supplied with a signal containing a symbol sequence of sequence length u, as shown in equation (7).

[0060]

[0061] Each of the taps 33-1 to 33-u has a so-called filter coefficient value f １ , f ２ , ..., f （ｕ＋１）／２ , ..., f ｕ The tap gain value f is set. １ ~f ｕ This represents the estimated inverse transfer function.

[0062] Taps 33-1 to 33-u each have a tap gain value f for the symbol they are assigned to. １ ~f ｕ Multiply by and output. Adder 34 sums the output values ​​of taps 33-1 to 33-u and outputs the result. Equation (7) is the (u+1) / 2th element r ｔ Since this can be described as a series centered around this, the output value of adder 34 can be expressed as shown in equation (8) below.

[0063]

[0064] In the adaptive filter unit 301, a delay of "(u-1)T / 2" is applied by the delay unit 31. Therefore, the sequence of u symbols, which is the input information corresponding to the output value of the calculation timing tT of the linear digital filtering by the adaptive filter unit 301, is delayed by "(u-1)T / 2".

[0065] The determination processing unit 302 performs a preliminary determination using a hard judgment on the output value of the adaptive filter unit 301 and obtains an estimated value of the transmission symbol corresponding to the output value. The determination processing unit 302 outputs the obtained estimated value, the preliminary determination symbol A', as the preliminary determination result. At this time, the determination processing unit 302 refers to the output value of the adaptive filter unit 301 and determines whether or not to use neighbor symbols as preliminary determination results for each value in the output value. That is, the determination processing unit 302 refers to the output value of the adaptive filter unit 301 and determines whether or not to include neighbor symbols in the estimated value of the transmission symbol for each value included in the output value.

[0066] If the determination processing unit 302 determines that neighbor symbols should be used (i.e., determines that neighbor symbols should be included in the estimated value of the transmitted symbols), it outputs a provisional determination result (provisional determination symbol A') that includes neighbor symbols for the values ​​in which it determined that neighbor symbols should be used. On the other hand, if it determines that neighbor symbols should not be used (i.e., determines that neighbor symbols should not be included in the estimated value of the transmitted symbols), it outputs a provisional determination result (provisional determination symbol A') that does not include neighbor symbols for the values ​​in which it determined that neighbor symbols should not be used. The determination processing unit 302 may also perform nonlinear filtering on the provisional determination result (provisional determination symbol A').

[0067] The determination of whether or not to use a neighbor symbol by the determination processing unit 302 is achieved by setting a determination threshold for determining whether the output value of the adaptive filter unit 301 is close to the value of the nearest symbol. The determination threshold is set, for example, to ±1 for 100%. In a situation where the determination threshold is set to 20% (±0.2), if the output value of the adaptive filter unit 301 is within 20% (±0.2) of the nearest symbol, the determination processing unit 302 considers the output value that is within 20% (±0.2) of the nearest symbol as the value of the nearest symbol and determines that the neighbor symbol will not be used.

[0068] On the other hand, if the output value of the adaptive filter unit 301 is not within 20% (±0.2) of the nearest symbol, the determination processing unit 302 considers the output value that is not within 20% (±0.2) of the nearest symbol to be potentially not the value of the nearest symbol and determines to use the nearest symbol. By performing the determination in this way, the number of provisional determination results based on the output value can be reduced. As a result, the amount of computation by the Viterbi algorithm in the subsequent stage can be reduced.

[0069] The update processing unit 303 uses the target value of the output value of the adaptive filter unit 301 as a provisional determination symbol A' output by the determination processing unit 302, and the respective tap gain values ​​f of taps 33-1 to 33-u of the adaptive filter unit 301. １ ~f ｕ The update processing unit 303 calculates the updated value of the tap gain value f. １ ~f ｕ The updated value, i.e., the estimated inverse transfer function, is calculated using a predetermined update algorithm.

[0070] As shown in Figure 3, the update processing unit 303 comprises a filter update algorithm processing unit 35, a subtractor 36, and an arithmetic unit 37. In the update processing unit 303, the subtractor 36 subtracts the output value of the adaptive filter unit 301 from the provisional judgment symbol A' output by the judgment processing unit 302, and outputs the resulting subtracted value as an error to the filter update algorithm processing unit 35 and the arithmetic unit 37.

[0071] The filter update algorithm processing unit 35 reduces the tap gain value f based on the error output by the subtractor 36 using a predetermined update algorithm. １ ~f ｕ The filter update algorithm processing unit 35 calculates the updated value of the tap gain value f. １ ~f ｕ Set this to taps 33-1 to 33-u, and set the tap gain value f １ ~f ｕ Perform the update.

[0072] The calculation unit 37 calculates the signal noise (SN) ratio based on the error output by the subtractor 36. The calculation unit 37 outputs a value indicating the calculated SN ratio to the judgment processing unit 302. The judgment processing unit 302 sets a judgment threshold corresponding to the SN ratio output from the calculation unit 37. For example, the judgment processing unit 302 may determine the judgment threshold to be used for the above judgment by referring to a table that associates the SN ratio with the judgment threshold. However, the method for determining the judgment threshold according to the SN ratio is not limited to this. The calculation unit 37 is one embodiment of the quality calculation unit.

[0073] The sequence estimation unit 40 comprises an adaptive filter unit 401, a sequence estimation algorithm processing unit 402, a transmission path estimation unit 403, an update processing unit 404, and a path trace determination unit 405. The adaptive filter unit 401 is, for example, a linear transversal filter as shown in Figure 3, and in order to reduce the memory length of the transmission path estimation unit 403, it receives the sequence {r ｔ The impulse response of {r} is compressed. Here, impulse response compression refers to compressing the time-spread received signal sequence {r}, which is compressed due to bandwidth limiting or wavelength dispersion, as shown in Figure 5. ｔ This involves compressing the pulse width of the symbol, which reduces interference between symbols.

[0074] As shown in Figure 3, the adaptive filter unit 401 includes delays 41, 42-1 to 42-(v-1), taps 43-1 to 43-v, and an adder 44. The delay 41, similar to the delay 31, is configured as shown in Figure 4, to process a sequence of N received signals {r ｔ The sequence of v symbols, which is part of}, is incorporated. Here, v may be the same value as u or a different value.

[0075] The delay unit 41 stores the captured symbol sequence and outputs the captured symbol sequence after a time elapsed of "(v-1)T / 2", i.e., "(v-1) / 2" symbols. Tap 43-1 contains the symbol r output by the delay unit 41. ｔ＋（ｖ－１）／２ It is given.

[0076] Each of the delay units 42-1 to 42-(v-1) takes in and stores one symbol, and outputs the taken symbol after a time interval of "T", i.e., after a delay of one symbol. For example, the first delay unit 42-1 outputs the r delayed by "(v-3) / 2" symbols. ｔ＋（ｖ－３）／２ The symbol is output. The last delay unit 42-(v-1) delays the r by "(v-1) / 2" symbols. ｔ＋（ｖ－１）／２ The symbol is output. As a result, taps 43-1 to 43-v are supplied with a signal containing a symbol sequence of sequence length v, as shown in equation (9).

[0077]

[0078] Each of the taps 43-1 to 43-v has a so-called filter coefficient value c. １ , c ２ , ..., c （ｖ＋１）／２ , ..., c ｖ The tap gain values ​​are set. Taps 43-1 to 43-v multiply the symbol assigned to each tap by its respective tap gain value and output the result. Adder 44 sums the output values ​​of taps 43-1 to 43-v and outputs the result. Equation (9) is the (v+1) / 2th element r ｔ Since this can be described as a series centered around this, the output value of adder 44 can be expressed by the following equation (10).

[0079]

[0080] As can be seen from equation (10), the adaptive filter section 401 controls the tap gain value c １ , c ２ , ..., c （ｖ＋１）／２ , ..., c ｖ Although the degree of influence is adjusted, it effectively outputs a single output symbol that compresses the information content of v symbol sequences.

[0081] It is known that the computational load of MLSE increases exponentially with respect to the pulse width, but the increase in computational load can be suppressed by compressing the pulse width with the adaptive filter unit 401.

[0082] In the adaptive filter unit 401, a delay of "(v-1)T / 2" is applied by the delay unit 41. Therefore, the v symbol sequences, which are input information corresponding to the output value of the calculation timing tT of the linear digital filtering by the adaptive filter unit 401, are delayed by "(v-1)T / 2".

[0083] The transmission path estimation unit 403 records the transmission path state μ of transmission path 2 at time t. ｔ From the symbol sequences representing the symbols, the provisional determination unit 30 provides a range of symbols representing the neighboring symbols centered on each of the symbols it has provisionally determined. The transmission path estimation unit 403 applies the estimated transfer function (H') to each of the symbol sequences provided by the neighbor addition comparison selection unit 51 to generate an estimated received symbol sequence for each symbol sequence.

[0084] The transmission path estimation unit 403 is, for example, a linear transversal filter as shown in Figure 3. The transmission path estimation unit 403 comprises delay units 62-1 to 62-(x-1), taps 61-1 to 61-x, and an adder 63. Each of the delay units 62-1 to 62-(x-1) captures and stores one symbol, and outputs the captured symbol after a time elapsed of "T", i.e., after a delay of one symbol. The transmission path state μ of the transmission path 2 at time t given by the neighboring sum comparison selection unit 51 ｔ This is the symbol sequence {s'} shown in the following equation (11). ｔ It can be expressed as}. Note that x corresponds to the memory length, and if the memory length is "3", then x = 3.

[0085]

[0086] Each of the taps 61-1 to 61-x contains the symbol sequence {s'} of equation (11). ｔ Each of the symbols contained in} is given. Each of the taps 61-1 to 61-x is the coefficient value h of the estimated transfer function (H'), which is the so-called filter coefficient value. １ , h ２ , ..., h （ｘ＋１）／２ , ..., h ｘ This is set. For example, tap 61-1 is h １ ×s' ｔ－（ｘ－１）／２The calculation is performed. Similarly, taps 61-2 to 61-x multiply their respective coefficient values ​​by the given symbols and output the multiplication result to adder 63. Adder 63 outputs the sum of the multiplication results. The output values ​​of adder 63 become the symbols that constitute the estimated received symbol sequence and are expressed as equation (12) below.

[0087]

[0088] The sequence estimation algorithm processing unit 402 processes the transmission path state μ ｔ The sequence estimation algorithm processing unit 402 calculates the metric for each. ｔ Using the respective metrics, the Viterbi algorithm is executed within the range of neighboring symbols centered on each of the symbols that the provisional determination unit 30 has provisionally determined.

[0089] The sequence estimation algorithm processing unit 402 includes a subtractor 54, an absolute value unit 53, and a neighbor addition comparison selection unit 51. The subtractor 54 subtracts the output value of the transmission path estimation unit 403 shown in equation (12) from the output value of the adaptive filter unit 401 shown in equation (10). The subtractor 54 outputs the subtracted value obtained by subtraction to the absolute value unit 53. The absolute value unit 53 calculates the absolute value of the subtracted value received from the subtractor 54. The absolute value calculated by the absolute value unit 53 is the metric and is expressed as equation (13).

[0090]

[0091] The neighboring addition comparison selection unit 51 takes in a plurality of provisional judgment symbols A' output by the provisional judgment unit 30 as provisional judgment results. The neighboring addition comparison selection unit 51 uses the plurality of provisional judgment symbols A' to create a provisional judgment symbol sequence {A'} whose sequence length is predetermined to be the length of the memory length. ｔ The nearest neighbor addition comparison selection unit 51 generates the transmission path state μ of the transmission path 2 at time t. ｔ Among the symbol sequences representing the provisional determination symbol sequence {A' ｔ Multiple symbol sequences {s'} are represented by the range of neighboring symbols centered on each of the symbols contained within}. ｔ The neighbor addition comparison selection unit 51 generates a sequence of symbols {s'}. ｔThe output is sent to the transmission path estimation unit 403.

[0092] The neighboring summation comparison selection unit 51 uses the transmission path state μ output by the absolute value meter 53. ｔ Using the metric for each, the provisional determination symbol sequence {A' ｔ The Viterbi algorithm is executed within the range of neighboring symbols centered on each of the symbols contained in}. The neighbor addition comparison selection unit 51 performs the Viterbi algorithm to determine the likelihood of the estimated received symbol sequence using the distance function d ｔ ( {μ ｔ}) calculate the calculated distance function d ｔ ( {μ ｔ The minimum value of}) is detected. The estimated received symbol sequence corresponding to this minimum value becomes the most likely estimated received symbol sequence.

[0093] The path trace determination unit 405 uses the distance function d detected by the nearest neighbor addition comparison selection unit 51. ｔ ( {μ ｔ Based on the minimum value of}), the trellis path is traced back to determine the transmitted symbol. The starting point of the traced path is the transmission path state μ at time t. ｔ Distance function d when reaching ｔ ( {μ ｔ This is the transmission path state where}) is the minimum value. In addition, the number of times "w" is traced back when the path trace determination unit 405 traces back the path is predetermined, and by fixing the number of times "w" is fixed, the amount of computation required to determine the path can be reduced. It is known that the path converges when traced back by a number of times several times the memory length of the transmission path estimation unit 403.

[0094] The path trace determination unit 405 determines each symbol obtained by symbol determination as determination symbol A. The path trace determination unit 405 outputs determination symbol A as the determination result. The determination symbol sequence {A} is obtained by arranging the determination symbols A sequentially determined by the path trace determination unit 405 in a sequence. ｔ This value represents the estimated value of the transmitted symbol sequence, i.e., the estimated value of the m-value data acquired by the signal generator 1.

[0095] The update processing unit 404 adjusts the tap gain value c of the adaptive filter unit 401. １ ~c ｖAnd the coefficient value h of the estimated transfer function of the filter update transmission path estimation unit 70 １ ~h ｘ Calculate.

[0096] The update processing unit 404 adjusts the tap gain value c of the adaptive filter unit 401. １ ~c ｖ When calculating this, the target value of the output value of the adaptive filter unit 401 is determined by the judgment result judgment symbol sequence {A ｔ The output value of the filter update transmission path estimation unit 70, which takes} as input information, is set. The update processing unit 404 sets the tap gain value c of each tap 43-1 to 43-v of the adaptive filter unit 401 to the target value. １ ~c ｖ The updated value is calculated using a predetermined update algorithm.

[0097] Furthermore, the update processing unit 404 calculates the coefficient value h of the estimated transfer function of the filter update transmission path estimation unit 70. １ ~h ｘ When calculating, the target value of the output value of the filter update transmission path estimation unit 70 is set to the output value of the adaptive filter unit 401. The update processing unit 404 sets the coefficient values ​​h of each tap 71-1 to 71-x of the filter update transmission path estimation unit 70 so that the target value is reached. １ ~h ｘ The update value is calculated by a predetermined update algorithm. The coefficient value h calculated by the update processing unit 404 １ ~h ｘ The updated values ​​are also applied to the taps 61-1 to 61-x of the transmission path estimation unit 403.

[0098] As shown in Figure 3, the update processing unit 404 includes a filter update transmission path estimation unit 70, a filter update algorithm processing unit 75, a delay unit 76, and a subtractor 77. The configuration of the filter update transmission path estimation unit 70 corresponds to the configuration of the transmission path estimation unit 403, with taps 61-1 to 61-x corresponding to taps 71- to 71-x, delay units 62-1 to 62-(x-1) corresponding to delay units 72-1 to 72-(x-1), and adder 63 corresponding to adder 73.

[0099] The delay unit 76 delays the output value of the adaptive filter unit 401 by a time of "-wT" before outputting it to the subtractor 77. The delay of "-wT" is due to the processing in the path trace determination unit 405, which causes a delay of "-wT". The delay of "-wT" by the delay unit 76 ensures that the timing of the output value of the filter update transmission path estimation unit 70 and the output value of the adaptive filter unit 401 coincide.

[0100] The subtractor 77 subtracts the output value of the adaptive filter unit 401, which has been delayed by "-wT" time, from the output value of the filter update transmission path estimation unit 70, and outputs the error obtained by the subtraction to the filter update algorithm processing unit 75.

[0101] The filter update algorithm processing unit 75 reduces the tap gain value c based on the error output by the subtractor 77 using a predetermined update algorithm. １ ~c ｖ The filter update algorithm processing unit 75 calculates the updated value of h based on the error output by the subtractor 36, using a predetermined update algorithm to reduce the error. １ ~h ｘ Calculate the updated value.

[0102] The filter update algorithm processing unit 75 calculates the tap gain value c. １ ~c ｖ Set the taps to 43-1 to 43-v, and set the tap gain value c １ ~c ｖ The filter update algorithm processing unit 75 performs the update of the calculated coefficient value h. １ ~h ｘ Set this to taps 71-1 to 71-x and taps 61-1 to 61-x of the transmission path estimation unit 403, and set the coefficient value h １ ~h ｘ Perform the update.

[0103] (Processing by the determination processing unit 302 in the first embodiment) Next, the processing by the determination processing unit 302 in the first embodiment will be described with reference to Figures 6 to 10. Figure 6 is a diagram for explaining the processing by the determination processing unit 302 in the first embodiment. In Figure 6, the processing of a provisional determination using a determination threshold will be explained. As an example, assume that the signal is a quaternary signal (PAM4) and the transmission sequence is [-1, +1, +3, -3, -1, +3]. Assume that the determination threshold is set to 20% (±0.2). Under these circumstances, assume that the output values ​​of the adaptive filter unit 301 are [-0.9, +1.1, +2.5, -3.2, -1.4, +2.4]. In Figure 6, the output values ​​of the adaptive filter unit 301 are shown as P1 (-0.9), P2 (+1.1), P3 (+2.5), P4 (-3.2), P5 (-1.4), and P6 (+2.4).

[0104] Here, the determination processing unit 302 determines whether each of the output values ​​P1 to P6 is within the range based on the value of the nearest symbol and the determination threshold. For example, output value P1 is -0.9 and the value of the nearest symbol is -1. Therefore, the determination processing unit 302 determines whether output value P1 is within the range (-0.8 to -1.2) based on the value of the nearest symbol (-1) and the determination threshold (±0.2). Since output value P1 is within the range based on the value of the nearest symbol and the determination threshold, the determination processing unit 302 determines that the nearest symbol will not be used for output value P1.

[0105] Similarly, the output value P2 is +1.1, and the value of the nearest symbol is +1. Therefore, the determination processing unit 302 determines whether the output value P2 is within the range (+1.2 to +1.0) based on the value of the nearest symbol (+1) and the determination threshold (±0.2). Since the output value P2 is within the range based on the value of the nearest symbol and the determination threshold, the determination processing unit 302 determines that the nearest symbol is not used for the output value P2 either.

[0106] Similarly, the output value P3 is +2.5, and the value of the nearest symbol is +3. Therefore, the determination processing unit 302 determines whether the output value P3 is within the range (+3.2 to +2.8) based on the value of the nearest symbol (+3) and the determination threshold (±0.2). Since the output value P3 is not within the range based on the value of the nearest symbol and the determination threshold, the determination processing unit 302 determines that the nearest symbol should be used for the output value P3. The nearest symbol for the output value P3 is +3, and the nearest symbol is +1. Therefore, the determination processing unit 302 uses +1 and +3 as the provisional determination results for the output value P3.

[0107] The judgment processing unit 302 performs the above-described process for each of the output values ​​P1 to P6. Then, for output values ​​that the judgment processing unit 302 has determined to use neighbor symbols based on the above-described process, it outputs a provisional judgment symbol A', which is an estimated value of the transmitted symbol including neighbor symbols, as a provisional judgment result. In the example shown in Figure 6, the judgment processing unit 302 outputs the provisional judgment symbol A'[-1, +1, (+1 or +3), -3, (-1 or -3), (+1 or +3)] as a provisional judgment result. In the above example, the same judgment threshold is shown for each of the output values ​​P1 to P6 (each symbol), but individual judgment thresholds may be set. That is, different judgment thresholds may be set for each symbol, or different judgment thresholds may be set for some symbols.

[0108] If the provisional judgment result is [-1, +1, +1 or +3, -3, -1 or -3, +1 or +3], then the trellis diagram, with a memory length of "2", will look like Figure 7. Figure 7 is an example of a trellis diagram based on the provisional judgment result shown in Figure 6. As shown in Figure 7, the trellis diagram has one, two, or four branches. When the memory length is L, the number of branches is a minimum of one and a maximum of two. Ｌ This is the basis for the book. In this way, by using a trellis diagram consisting only of provisional judgment results and neighboring symbols with a judgment threshold, it is possible to maintain estimation accuracy while further reducing the amount of computation and extending the memory length.

[0109] Next, the method for setting the judgment threshold will be explained using Figures 8 to 10. Figures 8 to 10 are diagrams illustrating the method for setting the judgment threshold in the first embodiment. Figures 8 to 10 illustrate the method for setting the judgment threshold in 16QAM. As Pattern 1, as shown in Figure 8, it is possible to set judgment thresholds that are independent or the same in the I-plane and the Q-plane. When judgment thresholds are set independently in the I-plane and the Q-plane, the judgment threshold α and the judgment threshold β will be different values.

[0110] As a second pattern, as shown in Figure 9, it is possible to set a determination threshold equal to the equidistant point γ from each symbol. In this case, symbols located inside the determination threshold can be considered as provisional determination results. As shown in Figure 10, outside the determination threshold, for the output value of the adaptive filter unit 301 located inside the region divided by a dashed line passing through the symbol point, symbols corresponding to the vertices of that region can be considered as provisional determination results.

[0111] (Processing by the symbol determination device in the first embodiment) Next, the processing by the symbol determination device 3 in the first embodiment will be described with reference to Figures 11 and 12.

[0112] (Processing by the provisional determination unit in the first embodiment) Figure 11 is a flowchart showing the processing flow by the provisional determination unit 30 of the symbol determination device 3.

[0113] The delay unit 31 of the adaptive filter unit 301 controls the received signal sequence {r ｔ A sequence of symbols with sequence length u is captured and stored from} (step Sa1). The delay unit 31 outputs the captured symbol sequence with a delay of "(u-1)T / 2" time. Each of the delay units 32-1 to 32-(u-1) captures and stores the symbols output sequentially by the delay unit 31, and outputs the stored symbols after "T" time has elapsed.

[0114] As a result, the received signal sequence {r} shown in equation (7) ｔ The symbol sequence} is given to taps 33-1 to 33-u. Taps 33-1 to 33-u are each given the symbol r ｔ－（ｕ－１）／２ ~r ｔ＋（ｕ－１）／２ And the tap gain value f set for each１ ~f ｕ The multiplication is performed, and the result of the multiplication is output to the adder 34. The adder 34 sums the multiplication results and calculates and outputs the output value shown in equation (8). This output value is the received signal sequence {r} which has been adapted and equalized by the estimated inverse transfer function. ｔ This becomes the symbol for} (Step Sa2).

[0115] The determination processing unit 302 performs a preliminary determination using a hard judgment on the output value of the adaptive filter unit 301 to obtain an estimated value of the transmission symbol. At this time, the determination processing unit 302 performs a preliminary determination based on the determination threshold described above. The specific process is as explained with reference to Figure 6. The determination processing unit 302 outputs the obtained estimated value, the preliminary determination symbol A', as the preliminary determination result (step Sa3).

[0116] The subtractor 36 subtracts the output value of the adaptive filter unit 301 from the provisional judgment symbol A' output by the judgment processing unit 302 and outputs the resulting subtracted value as an error to the filter update algorithm processing unit 35.

[0117] The filter update algorithm processing unit 35 reduces the tap gain value f based on the error output by the subtractor 36 using a predetermined update algorithm. １ ~f ｕ The updated value, i.e., the estimated inverse transfer function, is calculated. The filter update algorithm processing unit 35 then calculates the tap gain value f １ ~f ｕ Set this to taps 33-1 to 33-u, and set the tap gain value f １ ~f ｕ Perform the update (Step Sa4).

[0118] The delay unit 31 of the adaptive filter unit 301 shifts the symbol that is one symbol from the beginning of the symbol sequence of sequence length u acquired in the previous step Sa1 to the beginning of the received signal sequence {r ｔ If a symbol sequence of sequence length u can be acquired from {r} (step Sa5, Yes), the process of step Sa1 is performed. Meanwhile, the delay unit 31 starts with a symbol shifted by one symbol from the beginning of the symbol sequence of sequence length u acquired in the previous step Sa1, and then processes the received signal sequence {r} ｔIf a symbol sequence of sequence length u cannot be imported from} (step Sa5, No), the process is terminated.

[0119] (Processing by the sequence estimation unit in the first embodiment) Figure 12 is a flowchart showing the processing flow by the sequence estimation unit 40 of the symbol determination device 3. The delay unit 41 of the adaptive filter unit 401 determines the received signal sequence {r ｔ A sequence of symbols with sequence length v is captured and stored from} (step Sb1-1). The delay unit 41 outputs the captured symbol sequence with a delay of "(v-1)T / 2" time. Each of the delay units 42-1 to 42-(v-1) captures and stores the symbols output sequentially by the delay unit 41, and outputs the stored symbols after "T" time has elapsed.

[0120] As a result, the received signal sequence {r} shown in equation (9) ｔ The symbol sequence} is given to taps 43-1 to 43-v. Taps 43-1 to 43-v are each given the symbol r ｔ－（ｖ－１）／２ ~r ｔ＋（ｖ－１）／２ And the tap gain value c set for each １ ~c ｖ The multiplier is multiplied by the result of the multiplication and output the result to the adder 44. The adder 44 sums the results of the multiplication and calculates the output value shown by equation (10), and the calculated output value is added to the received signal sequence {r ｔ The information of} is compressed and output as a symbol (step Sb1-2).

[0121] In parallel with the processing in steps Sb1-1 and Sb1-2, the neighbor addition comparison selection unit 51 takes in a plurality of provisional judgment symbols A' output by the provisional judgment unit 30 as provisional judgment results (step Sb2-1). The neighbor addition comparison selection unit 51 takes a provisional judgment symbol sequence {A'} whose sequence length is a predetermined memory length. ｔ Generates}.

[0122] Each of the delay devices 62-1 to 62-(x-1) of the transmission path estimation unit 403 stores the captured symbol and outputs the captured symbol after a time of "T" has elapsed. Each of the taps 61-1 to 61-x contains the symbol sequence {s' ｔ Each of the symbols contained in} is given.

[0123] Taps 61-1 to 61-x represent the respective coefficient values ​​h. １ ~h ｘ The given symbol is multiplied by the given symbol, and the result of the multiplication is output to the adder 63. The adder 63 sums the multiplication results and calculates and outputs the output value shown by equation (12). The output values ​​output sequentially by the adder 63 become the symbols that constitute the estimated received symbol sequence (step Sb2-2).

[0124] The subtractor 54 subtracts the output value of the transmission path estimation unit 403, shown in equation (12), from the output value of the adaptive filter unit 401, shown in equation (10), and outputs the subtracted value obtained by the subtraction to the absolute value unit 53. The absolute value unit 53 calculates the absolute value of the subtracted value received from the subtractor 54. The absolute value calculated by the absolute value unit 53 becomes the metric expressed as equation (13) (step Sb3).

[0125] The nearest neighbor addition comparison selection unit 51 executes the Viterbi algorithm within the range of symbols included in the provisional determination symbol sequence. The nearest neighbor addition comparison selection unit 51 executes the Viterbi algorithm and determines the distance function d ｔ ( {μ ｔ}) calculate the calculated distance function d ｔ ( {μ ｔ}) The minimum value is detected (step Sb4).

[0126] The path trace determination unit 405 uses the distance function d calculated by the nearest neighbor addition comparison selection unit 51. ｔ ( {μ ｔ Based on the minimum value of {A}, the path of the trellis is traced back, a symbol determination is performed, and the determination symbol A is found. The path trace determination unit 405 outputs the determination symbol A as the determination result (step Sb5). The determination symbol sequence {A} is formed by arranging the determination symbols A that the path trace determination unit 405 determines sequentially. ｔ This value represents the estimated value of the transmitted symbol, i.e., the estimated value of the m-value data acquired by the signal generator 1.

[0127] Each of the delayers 72-1 to 72-(x-1) of the filter update transmission path estimation unit 70 takes in and stores the judgment symbol A that is sequentially output by the path trace determination unit 405, and outputs the stored judgment symbol A after the elapsed time of "T". Each of the taps 71-1 to 71-x contains the judgment symbol sequence {A ｔ Each of the symbols contained in} is given. Taps 71-1 to 71-x are the respective coefficient values ​​h １ ~h ｘ The given symbol is multiplied by the given symbol, and the result of the multiplication is output to the adder 73. The adder 73 calculates and outputs the sum of the multiplication results.

[0128] The delay unit 76 delays the output value of the adaptive filter unit 401 by a time of "-wT", i.e., by "-w" symbols, and outputs it to the subtractor 77. The subtractor 77 subtracts the output value of the adaptive filter unit 401, which has been delayed by "-wT", from the output value of the filter update transmission path estimation unit 70, and outputs the error obtained by the subtraction to the filter update algorithm processing unit 75.

[0129] The filter update algorithm processing unit 75 reduces the error based on the error output by the subtractor 77 by using a predetermined update algorithm to reduce the coefficient value h １ ~h ｘ The filter update algorithm processing unit 75 calculates the updated value of the coefficient h. １ ~h ｘ Set the corresponding taps 71-1 to 71-x and the taps 61-1 to 61-x of the transmission path estimation unit 403, and set the coefficient value h １ ~h ｘ Perform the update (step Sb6).

[0130] The filter update algorithm processing unit 75 reduces the tap gain value c based on the error output by the subtractor 77 using a predetermined update algorithm. １ ~c ｖ The filter update algorithm processing unit 75 calculates the updated value of the tap gain c. １ ~c ｖ Set the taps corresponding to each of them, 43-1 to 43-v, and set the tap gain value c １ ~c ｖPerform the update (step Sb7).

[0131] The delay unit 41 of the adaptive filter unit 401 shifts the symbol that is one symbol from the beginning of the symbol sequence of sequence length v acquired in the previous step Sb1-1 to the beginning of the received signal sequence {r ｔ If a symbol sequence of sequence length v can be acquired from {r} (step Sb8, Yes), the processes of steps Sb1-1 and Sb2-1 are performed. Meanwhile, the delay unit 41 starts with a symbol shifted by one symbol from the beginning of the symbol sequence of sequence length v acquired in the previous step Sb1-1, and then processes the received signal sequence {r} ｔ If a symbol sequence of sequence length v cannot be imported from} (step Sb8, No), the process is terminated.

[0132] Note that in the process shown in Figure 7 above, the order of steps Sb6 and Sb7 may be reversed.

[0133] In the configuration of the first embodiment described above, the symbol determination device 3 includes: a provisional determination unit 30 that generates a symbol sequence by performing adaptive equalization of the received signal sequence taken from the transmission path using the estimated inverse transfer function of the transmission path and performs a provisional determination on the symbol sequence using a determination threshold; a transmission path estimation unit 403 that generates an estimated received symbol sequence for each transmission path state based on a plurality of symbol sequences indicating the transmission path state and the estimated transfer function of the transmission path; a sequence estimation algorithm processing unit 402 that selects the most likely estimated received symbol sequence using a predetermined estimation algorithm based on the symbol sequence obtained from the received signal sequence, a metric obtained based on each of the estimated received symbol sequences, the provisional determination symbol provisionally determined by the provisional determination unit 30, and the symbols in the vicinity of the provisional determination symbol determined to be used by the provisional determination using the determination threshold; and a path trace determination unit 405 that determines the transmitted symbol sequence by tracing back the trellis path based on the most likely estimated received symbol sequence, wherein the sequence estimation algorithm processing unit 402 generates a plurality of symbol sequences indicating the transmission path state within the range of the provisional determination symbol and the symbols in the vicinity of the provisional determination symbol and outputs them to the transmission path estimation unit 403.

[0134] As described above, the symbol determination device 3 refers to the output value of the adaptive filter unit 301 and, using a determination threshold, determines whether or not to use neighbor symbols, then performs a predetermined estimation algorithm (e.g., the Viterbi algorithm) that restricts the trellis diagram. As a result, for symbols determined not to use neighbor symbols, only the symbols obtained as a provisional determination result are used, and for symbols determined to use neighbor symbols, the predetermined estimation algorithm (e.g., the Viterbi algorithm) is performed using both the symbols obtained as a provisional determination result and the neighbor symbols. Conventionally, neighbor symbols were used for all symbols obtained as a provisional determination result, whereas in the present invention, the calculation using neighbor symbols can be reduced for some core rags. Thus, it is possible to reduce the amount of computation compared to the conventional technology. Furthermore, since it is a sequential calculation, the size of the trellis diagram changes depending on the number of branches, but by creating a table of trellis diagram information (provisional determination result) and referring to it when calculating branch metrics and path metrics, the implementation difficulty in sequential calculations can be reduced.

[0135] The judgment threshold used by the provisional judgment unit 30 is set based on the signal-to-noise ratio (SNR) of the output value of the adaptive filter unit 301. This allows for setting an appropriate judgment threshold according to the usage conditions. Thus, the optimal reduction in computational load according to the present invention is correlated with the SNR of the received signal.

[0136] In addition, in the configuration of the first embodiment described above, the sequence estimation unit 40 may be configured without an adaptive filter unit 401. In this case, the subtractor 54 of the sequence estimation algorithm processing unit 402 uses the received signal sequence {r ｔ This will involve incorporating}.

[0137] (Second Embodiment) Figure 13 is a block diagram showing the internal configuration of the symbol determination device 3a in the second embodiment, and Figure 14 is a block diagram showing the detailed internal configuration of the symbol determination device 3a in the second embodiment. In the second embodiment, the same reference numerals are used for components that are the same as in the first embodiment, and the different components will be described below.

[0138] As shown in Figure 13, the symbol determination device 3a comprises a provisional determination unit 30 and a sequence estimation unit 40a. The sequence estimation unit 40a comprises an adaptive filter unit 401a, a sequence estimation algorithm processing unit 402, a transmission path estimation unit 403, an update processing unit 404, and a path trace determination unit 405.

[0139] The input terminal of the adaptive filter unit 401a is connected to the output terminal of the adaptive filter unit 301 of the provisional determination unit 30. More specifically, as shown in Figure 14, the output terminal of the adder 34 of the adaptive filter unit 301 is connected to the input terminal of the delay unit 41 of the adaptive filter unit 401a.

[0140] As described in the symbol determination device 3 of the first embodiment, the adaptive filter unit 301 performs adaptive equalization using the estimated inverse transfer function to determine the received signal sequence {r ｔ The objective is to bring the signal sequence closer to the transmitted signal sequence. Therefore, during adaptive equalization using the estimated inverse transfer function, noise in the high-frequency range is amplified. In the adaptive filter section 401a of the sequence estimation section 40a, by performing filtering with a small number of taps sufficient to suppress the high-frequency range noise amplified by the adaptive filter section 301, it is possible to suppress the high-frequency range noise and compress the pulse width. In other words, in the second embodiment, by utilizing the output signal of the adaptive filter section 301, the efficiency of the pulse width compression process in the adaptive filter section 401a can be improved.

[0141] In the second embodiment, as in the first embodiment, the sequence estimation unit 40a may be configured without the adaptive filter unit 401a. In this case, the subtractor 54 of the sequence estimation algorithm processing unit 402 will take the output value of the adaptive filter unit 301 instead of the output value of the adaptive filter unit 401a.

[0142] (Other configurations using the BCJR algorithm) In the first and second embodiments described above, the neighbor addition comparison selection unit 51 of the symbol determination devices 3 and 3a executed the Viterbi algorithm as an estimation algorithm for estimating the transmitted symbol sequence. However, the configuration of the present invention is not limited to these embodiments. For example, the neighbor addition comparison selection unit 51 may be configured to execute the BCJR algorithm, which is a maximum a posteriori probability (MAP) decoding method that is effective when different prior probabilities exist for each symbol. Note that the maximum likelihood decoding method using the Viterbi algorithm is equivalent to the MAP decoding method when the prior probabilities are equal for all symbols.

[0143] In the BCJR algorithm for LDPC (Low Density Parity Check) coding, a trellis diagram is used, similar to the Viterbi algorithm for MLSE. Therefore, when using the BCJR algorithm, there is no need to change the configuration of the symbol determination devices 3 and 3a. By replacing the algorithm executed by the neighbor addition comparison selection unit 51 from the Viterbi algorithm to the BCJR algorithm, it is possible to estimate the transmitted symbol sequence using the BCJR algorithm. The BCJR algorithm uses the log-likelihood ratio as the soft decision output, and there is a problem that the log-likelihood ratio becomes infinite when the conditional probability is 0%, but this problem can be addressed by the method described in Patent Document 1.

[0144] In the configurations of the first and second embodiments described above, linear transversal filters are applied to the adaptive filter unit 301, adaptive filter units 401, 401a, transmission path estimation unit 403, and filter update transmission path estimation unit 70. However, the configuration of the present invention is not limited to these embodiments. Filters other than linear transversal filters, such as other linear filters or nonlinear filters, may be applied to the adaptive filter unit 301, adaptive filter units 401, 401a, transmission path estimation unit 403, and filter update transmission path estimation unit 70.

[0145] Furthermore, in the configurations of the first and second embodiments described above, the filter update algorithm processing units 35 and 75 reduce the error by a predetermined update algorithm, thereby reducing the tap gain value f １ ~f ｕ , tap gain value c １ ~c ｖ , coefficient value h １ ~h ｘ The updated values ​​of filter coefficients such as are calculated. Here, as a predetermined update algorithm, for example, iterative approximation algorithms such as the LMS (Least Mean Square) algorithm or the RLS (Recursive Least Square) algorithm are applied.

[0146] When the LMS algorithm is applied, the updated filter coefficient value set for each tap is calculated, for example, by the following equation (14).

[0147] H(n) ｋ＋１ = H(n) ｋ +φ・E・U(n) ｋ ... (14)

[0148] In equation (14), n is the identifier of one of the multiple taps, and k is a value indicating the number of updates. Also, H(n) ｋ＋１ However, this is the updated value of the filter coefficient, H(n) ｋ However, this is the filter coefficient value before the update, U(n) ｋ However, this is the input signal given to the nth tap immediately before calculating the updated value. E is the error, and φ is a convergence constant that is determined as appropriate. Each time the filter update algorithm processing units 35 and 75 calculate a new filter coefficient value, for example, they write the calculated filter coefficient value to an internal memory area, and at the next update, they read it from the internal memory area to obtain the filter coefficient value H(n) before the update. ｋ It is used as such.

[0149] (Experiment) An experiment based on the above-described process was conducted. Specifically, a high-speed PAM60-band optical signal transmission experiment was performed using the configuration shown in Figure 15. Symbol determination was performed by applying the configuration shown in Figure 14 to the offline digital signal processing on the receiving side. In the configuration shown in Figure 14 applied here, the same determination threshold was set for all symbols, and the bit error rate was evaluated for each determination threshold. At the same time, the candidate sequence reduction rate for each determination threshold was also evaluated by counting the total number of candidate sequences used in the determination of the received sequence. The candidate sequence number used as the basis for the reduction rate was the number of candidate sequences when the determination threshold is 0%, that is, based on a trellis diagram that considers combinations of two symbols at the upper and lower levels near the output of the output value of the adaptive filter unit 301.

[0150] Figures 16 and 17 show the experimental results. The threshold rate shown in Figure 16 is the judgment threshold. As shown in Figure 16, when the output distribution of the adaptive filter unit 301 is spread out (i.e., the signal-to-noise ratio is poor), if the trellis is not branched when the filter output is in area A1, components spreading from adjacent symbols (areas enclosed in circles) will definitely result in errors. The same error occurs with FFE, but with TL-MLSE, error propagation occurs, resulting in the same bit error rate as FFE. There are threshold rate conditions where the bit error rate worsens compared to the bit error rate when the judgment threshold (threshold rate) is 100%. The lower the baud rate, the higher the threshold rate at which the bit error rate deteriorates, that is, the wider the area of ​​area A1 in Figure 17, so it becomes possible to optimize the computation reduction effect by setting the judgment threshold (threshold rate) according to the spread of the output distribution of the adaptive filter unit 301.

[0151] The symbol determination devices 3 and 3a in the above-described embodiment may be implemented using a computer. In that case, the program for implementing this function may be recorded on a computer-readable recording medium, and the program recorded on this recording medium may be loaded into a computer system and executed. Here, "computer system" includes hardware such as the OS and peripheral devices. Furthermore, "computer-readable recording medium" refers to portable media such as flexible disks, magneto-optical disks, ROMs, CD-ROMs, and storage devices such as hard disks built into a computer system. Moreover, "computer-readable recording medium" may also include those that dynamically hold programs for a short period of time, such as communication lines used when transmitting programs via networks such as the Internet or communication lines such as telephone lines, and those that hold programs for a certain period of time, such as volatile memory inside a computer system that acts as a server or client in such cases. Furthermore, the above-mentioned program may be for implementing a part of the above-mentioned function, or it may be a program that can implement the above-mentioned function in combination with a program already recorded in the computer system, or it may be implemented using a programmable logic device such as an FPGA (Field Programmable Gate Array).

[0152] While embodiments of this invention have been described in detail above with reference to the drawings, the specific configuration is not limited to these embodiments and includes designs and the like that do not depart from the spirit of this invention.

[0153] This invention is applicable to technology.

[0154] 3...Symbol determination device, 30...Provisional determination unit, 40...Sequence estimation unit, 301...Adaptive filter unit, 302...Determination processing unit, 303...Update processing unit, 401...Adaptive filter unit, 402...Sequence estimation algorithm processing unit, 403...Transmission path estimation unit, 404...Update processing unit, 405...Path trace determination unit

Claims

1. Symbol determination device comprising:

1. A symbol determination device comprising:

1. A symbol determination device that generates a symbol sequence by performing adaptive equalization of a received signal sequence taken from a transmission line using an estimated inverse transfer function of the transmission line, and performs a provisional determination on the symbol sequence using a determination threshold to determine if it has a value close to the nearest symbol; 2. A transmission line estimation unit that generates an estimated received symbol sequence for each of the transmission line states based on a plurality of symbol sequences indicating the transmission line state and an estimated transfer function of the transmission line; 3. A sequence estimation algorithm processing unit that selects the most likely estimated received symbol sequence using a predetermined estimation algorithm based on a symbol sequence obtained from the received signal sequence, a metric obtained based on each of the estimated received symbol sequences, a provisional determination symbol provisionally determined by the provisional determination unit, and a neighboring symbol of the provisional determination symbol determined to be used by the provisional determination using the determination threshold; and 4. A path trace determination unit that determines the transmitted symbol sequence by tracing back the path of the trellis based on the most likely estimated received symbol sequence, wherein the sequence estimation algorithm processing unit generates a plurality of symbol sequences indicating the transmission line state within the range of the provisional determination symbol and neighboring symbols of the provisional determination symbol and outputs them to the transmission line estimation unit.

2. The symbol determination device according to claim 1, wherein the provisional determination unit determines whether each of the symbol sequences is located within a range based on the nearest symbol and the determination threshold, determines that the value of the neighboring symbol will not be used for symbols included in the symbol sequence that are located within the range based on the nearest symbol and the determination threshold, and determines that the value of the neighboring symbol will be used for symbols included in the symbol sequence that are located outside the range based on the nearest symbol and the determination threshold, and outputs a provisional determination result.

3. The symbol determination device according to claim 1 or 2, wherein the provisional determination unit comprises a quality calculation unit that calculates quality based on the error between the symbol sequence and the provisional determination result, and performs a provisional determination using the determination threshold based on the quality calculation result.

4. A symbol determination method comprising: generating a symbol sequence by performing adaptive equalization on a received signal sequence taken from a transmission line using the estimated inverse transfer function of the transmission line; performing a preliminary determination on the symbol sequence using a determination threshold to determine if it has a value close to the nearest symbol; generating multiple symbol sequences indicating the transmission line state within the range of symbols near the preliminary determination symbol determined to be used by the preliminary determination using the determination threshold; generating an estimated received symbol sequence for each of the transmission line states based on the multiple symbol sequences indicating the transmission line state and the estimated transfer function of the transmission line; selecting the most likely estimated received symbol sequence using a predetermined estimation algorithm based on the symbol sequence obtained from the received signal sequence, a metric obtained based on each of the estimated received symbol sequences, the preliminary determination symbol, and the symbols near the preliminary determination symbol determined to be used by the preliminary determination using the determination threshold; and determining the transmitted symbol sequence by tracing back the trellis path based on the most likely estimated received symbol sequence.

Images