Wavelength dispersion compensation device and coefficient optimization method

Abstract

A wavelength dispersion compensation device is provided with: a block division unit that divides an input signal loaded with wavelength dispersion distortion into blocks of a certain length so as to overlap adjacent blocks by a predetermined length; a Fourier transform unit that transforms each of the divided blocks into a plurality of frequency domain signals by performing Fourier transform on each of the divided blocks; a coefficient multiplication unit that multiplies each of the plurality of converted frequency domain signals by a wavelength dispersion compensation coefficient and a window function or a value obtained by synthesizing the dispersion compensation coefficient and the window function, and generates a coefficient application completion block in which multiplication results are added; an inverse Fourier transform unit that performs an inverse Fourier transform on the generated coefficient-applied block; an overlap shearing unit that removes an overlapped portion from the converted coefficient-applied block and generates an output signal; and a weight coefficient optimization unit that optimizes a window function so as to minimize waveform distortion in a time domain on the basis of the generated output signal and a desired waveform in which wavelength dispersion compensation has been performed on the input signal.

Description

Technical Field

[0001] The present invention relates to a wavelength dispersion compensation device and a coefficient optimization method. Background Art

[0002] In digital coherent optical communication, the following research is carried out: compensating for signal distortion generated in an optical transmission device, an optical transmission path, and an optical reception device through digital signal processing, so as to be suitable for large-capacity and long-distance transmission. In particular, in the case of increasing the transmission distance, the waveform distortion generated in the optical transmission path becomes larger, and the load of digital signal processing performed by the optical reception device becomes heavier, and it cannot be processed at once. Therefore, in the optical reception device, the following signal processing method has been proposed: dividing the received signal into a plurality of block sizes, respectively transforming the received signals of the divided block sizes into signals in the frequency domain, respectively performing calculation processing of multiplying the signals in the frequency domain by a wavelength dispersion compensation and a window function, and then performing synthesis, thereby compensating for wavelength dispersion (for example, refer to Non-Patent Document 1).

[0003] Prior Art Documents Non-Patent Documents Non-Patent Document 1: K. Ishihara, T. Kobayashi, R. Kudo, Y. Takatori, A. Sano, E. Yamada, H. Masuda, M. Matsui, M. Mizoguchi and Y. Miyamoto, “Frequency-domain equalisation without guard interval for optical transmission systems”, ELECTRONICS LETTERS 4th December 2008 Vol.44 No.25. Summary of the Invention

[0004] Problems to be Solved by the Invention In the technique described in Non-Patent Document 1, the input digital waveform data is divided into a certain block length, and the divided data is respectively subjected to Fourier transform. At this time, each block is divided with overlap. And, in the technique described in Non-Patent Document 1, the inverse response function of wavelength dispersion is multiplied by the signals in the frequency domain of each block length, and through inverse Fourier transform, it returns to the data of the time waveform. After that, in the technique described in Non-Patent Document 1, after deleting the overlapping parts, they are connected and returned to the continuous time waveform. However, in the conventional signal processing method, if the signals in the frequency domain are not synthesized with the optimal weights respectively, there are problems that the compensation accuracy of wavelength dispersion is reduced and as a result, the signal quality is reduced.

[0005] In view of the above, an object of the present invention is to provide a technique capable of optimizing weight coefficients for wavelength dispersion compensation.

[0006] Solution to the problem One aspect of the present invention is a wavelength dispersion compensation device, which includes: a block division unit that divides an input signal loaded with wavelength dispersion distortion into blocks of a certain length in such a way that a predetermined length of overlap is generated with adjacent blocks; a Fourier transform unit that transforms each of the blocks divided by the block division unit into a plurality of frequency domain signals by performing Fourier transform on each block; a coefficient multiplication unit that multiplies each of the plurality of frequency domain signals transformed by the Fourier transform unit by each of a wavelength dispersion compensation coefficient and a window function or a value obtained by synthesizing the dispersion compensation coefficient and the window function, and generates a coefficient-applied block after summing up the multiplication results; an inverse Fourier transform unit that performs inverse Fourier transform on the coefficient-applied block generated by the coefficient multiplication unit; an overlap cutting unit that removes the overlapping part from the coefficient-applied block transformed by the inverse Fourier transform unit to generate an output signal; and a weight coefficient optimization unit that optimizes the window function in such a way that waveform distortion in the time domain is minimized based on the output signal generated by the overlap cutting unit and a desired waveform obtained by performing wavelength dispersion compensation on the input signal.

[0007] One aspect of the present invention is a coefficient optimization method, which includes: dividing an input signal loaded with wavelength dispersion distortion into blocks of a certain length in such a way that a predetermined length of overlap is generated with adjacent blocks; transforming each of the divided blocks into a plurality of frequency domain signals by performing Fourier transform on each block; multiplying each of the transformed plurality of frequency domain signals by each of a wavelength dispersion compensation coefficient and a window function or a value obtained by synthesizing the dispersion compensation coefficient and the window function; generating a coefficient-applied block after summing up the multiplication results; performing inverse Fourier transform on the coefficient-applied block; removing the overlapping part from the transformed coefficient-applied block to generate an output signal; and optimizing the window function in such a way that waveform distortion in the time domain is minimized based on the generated output signal and a desired waveform obtained by performing wavelength dispersion compensation on the input signal.

[0008] Advantageous effects of the invention According to the present invention, it is possible to optimize weight coefficients for wavelength dispersion compensation. Description of the drawings

[0009] Figure 1 It is a diagram showing a structural example of an optical receiving device in the first embodiment. Figure 2 It is a diagram showing a structural example of a wavelength dispersion compensation unit in the first embodiment. Figure 3This is a diagram showing a structural example of the multiplication unit in the first embodiment. Figure 4 This is a diagram showing an example of the window function in the first embodiment. Figure 5 This is a flowchart showing the processing flow of the wavelength dispersion compensation device in the first embodiment. Figure 6 This is a diagram for explaining the effects in the first embodiment. Figure 7 This is a diagram showing a structural example of the weight coefficient optimization device in a modified example of the first embodiment. Figure 8 This is a diagram for explaining the outline of the method for calculating the optimal window function in the second embodiment by linear approximation. Figure 9 This is a diagram showing an example of the window function calculated by linear approximation in the second embodiment. Detailed Embodiment

[0010] Hereinafter, an embodiment of the present invention will be described with reference to the drawings. (First Embodiment) Figure 1 This is a diagram showing a structural example of the optical receiving device 10 in the first embodiment. The optical receiving device 10 is connected to the optical transmitting device via an optical transmission path. The optical receiving device 10 receives the optical signal transmitted from the optical transmitting device. The optical receiving device 10 includes an optical receiving unit 11 and a digital signal processing unit 12. The digital signal processing unit 12 is composed of an ADC 13, a wavelength dispersion compensation device 14, an adaptive equalization unit 15, a demapping unit 16, and a decoding unit 17.

[0011] The optical receiving unit 11 interferes the optical signal input from the outside (hereinafter referred to as "input signal") with the local oscillation light, and converts the input signal into a baseband analog electrical signal. Here, wavelength dispersion generated in the optical transmission path is loaded in the input signal.

[0012] The ADC 13 converts the analog electrical signal output from the optical receiving unit 11 into a digital electrical signal.

[0013] The wavelength dispersion compensation device 14 compensates for the wavelength dispersion generated in the optical transmission path by frequency domain equalization. The wavelength dispersion compensation device 14 is composed of a wavelength dispersion compensation unit 18 and a weight coefficient optimization unit 19.

[0014] The wavelength dispersion compensation unit 18 compensates for the wavelength dispersion generated in the optical transmission path based on the wavelength dispersion compensation coefficient, window function, and digital electrical signal output from the ADC 13 set by the weight coefficient optimization unit 19. The wavelength dispersion compensation coefficient is estimated by the weight coefficient optimization unit 19. The window function is optimized by the weight coefficient optimization unit 19. The wavelength dispersion compensation coefficient and the window function are weight coefficients for wavelength dispersion compensation.

[0015] The weight coefficient optimization unit 19 estimates the wavelength dispersion compensation amount based on the digital electrical signal output from the ADC 13. The weight coefficient optimization unit 19 sets the wavelength dispersion compensation coefficient corresponding to the estimated wavelength dispersion compensation amount in the wavelength dispersion compensation unit 18. Additionally, when the wavelength dispersion compensation amount is known through other measurement methods, the weight coefficient optimization unit 19 may not estimate the wavelength dispersion compensation amount. In this case, the weight coefficient optimization unit 19 sets the wavelength dispersion compensation coefficient corresponding to the known wavelength dispersion compensation amount in the wavelength dispersion compensation unit 18.

[0016] Furthermore, the weight coefficient optimization unit 19 optimizes the window function based on the digital electrical signal output from the ADC 13. As methods for optimizing the window function, there are a method using the least squares method of a cost function and a method using linear approximation. In the first embodiment, the structure for optimizing the window function by the method using the least squares method of a cost function is described.

[0017] The adaptive equalization unit 15 dynamically estimates and equalizes the dynamically varying polarization wave or laser phase noise generated in the optical transmission path through digital signal processing such as an FIR (Finite Impulse Response) filter or frequency domain equalization.

[0018] The demapping unit 16 transforms the symbol information of the signal equalized by the adaptive equalization unit 15 into a bit sequence.

[0019] The decoding unit 17 corrects the bit sequence output from the demapping unit 16 by an error correction decoding method corresponding to the error correction coding performed by the optical transmission device.

[0020] Figure 2 is a diagram showing a structural example of the wavelength dispersion compensation unit 18 in the first embodiment. The wavelength dispersion compensation unit 18 includes a block division unit 31, a Fourier transform unit 32, a coefficient multiplication unit 33, an inverse Fourier transform unit 34, and an overlap and cut unit 35. In Figure 2 , the "N" shown on the connection line between each functional unit indicates that a block including N values is input and output. The same applies to other diagrams after Figure 2 and later.

[0021] The block division unit 31 divides the digital signal output from the ADC 13 into blocks such that M sample numbers in the block size N of the Fourier transform performed by the Fourier transform unit 32 are repeated with adjacent blocks. Here, N is a positive integer, and M is an integer where N > M. In many cases, both N and M are powers of 2. Thus, repeating with adjacent blocks in the time domain is called overlapping. When the block size N is 1024 and the overlapping rate is set to 50%, M = 512. The block division unit 31 performs a serial-to-parallel conversion on the divided blocks and outputs each block after the serial-to-parallel conversion.

[0022] The Fourier transform unit 32 performs a Fourier transform on the blocks sequentially output by the block division unit 31. That is, the Fourier transform unit 32 transforms a time-domain signal of N samples into a frequency-domain signal X of N frequency bins. When performing the Fourier transform, the Fourier transform unit 32 performs a discrete Fourier transform or a fast Fourier transform.

[0023] The coefficient multiplication unit 33 multiplies each value of each frequency bin of the frequency-domain signal X output by the Fourier transform unit 32 by a different coefficient. Specifically, the coefficient multiplication unit 33 multiplies each value of each frequency bin of the frequency-domain signal X by a different wavelength dispersion compensation coefficient and a different window function. As Figure 3 shown, the coefficient multiplication unit 33 includes a plurality of multiplication units 40-1 to 40-O, one or more storage units 50-1 to 50-(O-1), and an addition unit 60. In addition, O is an integer of 2 or more. In the following description, the case where O is 3 is taken as an example for explanation.

[0024] The storage units 50-1 to 50-2 respectively store the blocks output by the Fourier transform unit 32 per unit time. For example, the storage unit 50-1 stores the block at time t-1. The storage unit 50-2 stores the block at time t-2.

[0025] The multiplication units 40-1 to 40-3 include wavelength dispersion compensation coefficient multiplication units 41-1 to 41-3 and window function multiplication units 42-1 to 42-3. The wavelength dispersion compensation coefficient multiplication unit 41-1 multiplies the block output by the Fourier transform unit 32 by the wavelength dispersion compensation coefficient H1(f) set by the weight coefficient optimization unit 19. The window function multiplication unit 42-1 multiplies the block after multiplying by the wavelength dispersion compensation coefficient H1(f) by the window function C1(f).

[0026] Each of the multiplication units 40-2 to 40-3 imports the blocks stored in the storage units 50-1 to 50-2. Each of the wavelength dispersion compensation coefficient multiplication units 41-2 to 41-3 multiplies the imported blocks by the wavelength dispersion compensation coefficients H2(f) to H3(f) set by the weight coefficient optimization unit 19. Each of the window function multiplication units 42-2 to 42-3 multiplies the blocks after multiplying the wavelength dispersion compensation coefficients H2(f) to H3(f) by the window functions C2(f) to C3(f). Each of the multiplication units 40-1 to 40-3 outputs the multiplication result to the addition unit 60.

[0027] In addition, in each of the multiplication units 40 of the coefficient multiplication unit 33, the order of multiplying the wavelength dispersion compensation coefficient H(f) and the window function C(f) can be either first. Alternatively, the product of the wavelength dispersion compensation coefficient H(f) and the window function C(f) can be calculated in advance, and in each of the multiplication units 40 of the coefficient multiplication unit 33, the product of the wavelength dispersion compensation coefficient H(f) and the window function C(f) calculated in advance is multiplied. For example, the product of the wavelength dispersion compensation coefficient H1(f) and the window function C1(f) can be calculated in advance, and the multiplication unit 40-1 multiplies the product of the wavelength dispersion compensation coefficient H1(f) and the window function C1(f) calculated in advance by the block.

[0028] The addition unit 60 adds the multiplication results output from each of the multiplication units 40-2 to 40-3 for each frequency bin. Thus, the addition unit 60 calculates the output signal including N values.

[0029] The inverse Fourier transform unit 34 performs an inverse Fourier transform on each block output from the coefficient multiplication unit 33. That is, the inverse Fourier transform unit 34 transforms the frequency domain signal of N frequency bins into a time domain signal of N samples. When performing the inverse Fourier transform, the inverse Fourier transform unit 34 performs an inverse discrete Fourier transform or an inverse fast Fourier transform.

[0030] The overlap-splicing unit 35 cuts out the part of M samples as the overlapping part from the blocks output from the inverse Fourier transform unit 34, performs a parallel-to-serial conversion on the cut blocks, and outputs the time domain signal Y including (N - M) samples in length.

[0031] Figure 4 is a diagram showing an example of the window function in the first embodiment. In Figure 4 n 1s and n 1e represent the frequency range of the window function C1(f), n 2s and n 2e represent the frequency range of the window function C2(f), n 3s and n 3eRepresents the frequency range of the window function C3(f). As the window function in the first embodiment, the frequency overlap (FOL: Frequency Overlap) representing the overlap in the frequency domain is defined as a control parameter for power and performance, and has an overlap FOL on both sides centered on the FSI (FFT block Switching Index, FFT block switching index) as the frequency switching point. By increasing this frequency overlap, the dispersion compensation penalty can be suppressed. On the other hand, since the amount of signal to be processed increases, there is a tendency for power consumption to increase.

[0032] The waveform distortion caused by wavelength dispersion is a phenomenon in which the group delay varies according to the frequency of the signal spectrum. By using the signals in the frequency domain obtained from multiple blocks, the size of the Fourier transform can be suppressed. At this time, the output signals of the blocks used vary according to the frequency. If the blocks are switched abruptly at the switching point of the blocks in this frequency domain, a steep filter with a very large change in the frequency domain is applied, resulting in waveform distortion with a long response time in the time domain. Therefore, by giving a width to the switching of the blocks in the frequency domain, applying a window function to multiple associated blocks centered on the switching point, and synthesizing and using the signals of multiple blocks, the dispersion compensation penalty can be made close to the performance in an ideal large-scale fast Fourier transform.

[0033] Therefore, the weight coefficient optimization unit 19 in the present embodiment optimizes the window function in such a way as to minimize the frequency overlap as much as possible, so as to minimize the signal distortion with a limited frequency overlap.

[0034] Next, a method for the weight coefficient optimization unit 19 in the first embodiment to optimize the window function will be described. First, as a premise, the signal obtained by combining the signals in the frequency domain of multiple blocks used in the synthesis in the adder 60 is set as X ext . Here, when combining the signals in the frequency domain of multiple blocks, the combination is performed by unraveling and arranging the overlap of the signals in the frequency domain. Unraveling the overlap of the frequency domain signals means making the overlap of the regions where the adjacent frequency domain signals overlap with each other disappear. The combined frequency domain signal X ext is a way of combining signal vectors.

[0035] The time domain signal Y(k) output by the overlap clipping unit 35 is represented by the product of the inverse DFT (Discrete Fourier Transformation) matrix IDFT ext , X ext (k), the dispersion compensation coefficient H ext , and the window function C ext . In addition, k represents the block number.

[0036] [Equation 1] Y = IDFT ext *X ext ⊙H ext ⊙C ext … Equation (1)

[0037] In Equation (1), the *(asterisk) represents matrix multiplication, and 〇·(· in 〇) represents element-wise multiplication. The weight coefficient optimization unit 19 uses the signal obtained by compensating for wavelength dispersion of the digital electrical signal output from the ADC 13 as the reference signal R, and calculates the window function C that minimizes the error between the reference signal R(k) and the time-domain signal Y(k) output from the overlapping shearing unit 35 ext That's all. For example, the weight coefficient optimization unit 19 generates a reference signal R with wavelength dispersion compensation by multiplying the digital electrical signal output from the ADC 13 by a wavelength dispersion compensation coefficient corresponding to the estimated wavelength dispersion compensation amount. As shown in the following Equation (2), the optimized cost function L is represented by the squared error between the time-domain signal Y(k) output from the overlapping shearing unit 35 and the reference signal R(k).

[0038] [Equation 2]

[0039] Here, if the covariance matrix M defined by the following Equation (3) cov and the correlation vector V cor are defined, then the optimal window function C ext * can be expressed as follows by the stationary condition of the least squares method of the cost function L. The weight coefficient optimization unit 19 calculates the optimal window function C applied to each multiplication unit 40 based on the following Equation (4) ext *.

[0040] Here, an overview of Equation (3) and Equation (4) will be described. First, the weight coefficient optimization unit 19 generates a single frequency-domain signal X ext by decomposing and combining the overlaps of multiple frequency-domain signals. In addition, the weight coefficient optimization unit 19 generates a combined wavelength dispersion compensation coefficient H ext (combined coefficient vector) by decomposing and combining the overlaps of multiple wavelength dispersion compensation coefficients multiplied by multiple frequency-domain signals respectively. In addition, after multiplying multiple frequency-domain signals by wavelength dispersion compensation coefficients respectively, the weight coefficient optimization unit 19 generates a combined inverse discrete Fourier matrix IDFT ext by decomposing and combining the overlaps of matrices equivalent to the inverse Fourier transform performed by the inverse Fourier transform unit 34. Then, the weight coefficient optimization unit 19 calculates the covariance matrix of the frequency-domain signal X ext , the combined wavelength dispersion compensation coefficient H ext and the combined inverse discrete Fourier matrix IDFT extThe product of the elements of the covariance matrix. Then, the weight coefficient optimization unit 19 calculates the inverse matrix of the calculated product of the elements.

[0041] After the weight coefficient optimization unit 19 calculates the combined wavelength dispersion compensation coefficient H ext , the frequency domain signal X ext , the combined inverse discrete Fourier matrix IDFT ext and the matrix product of the complex conjugate and the reference signal R, it generates a correlation vector that obtains the product of the elements of the combined wavelength dispersion compensation coefficient H ext . The weight coefficient optimization unit 19 optimizes the window function by multiplying the generated correlation vector by the inverse matrix of the calculated covariance matrix.

[0042] [Equation 3]

[0043] [Equation 4]

[0044] In addition, in Equations (3) and (4), I Nc represents the Nc×Nc identity matrix, Nc represents the length of the window function, and λ represents the adjustment parameter of the normalization term. In addition, when λ = 0, it also becomes a condition without the identity matrix term.

[0045] Figure 5 is a flowchart showing the processing flow of the wavelength dispersion compensation device 14 in the first embodiment. The block division unit 31 imports the digital signal output by the ADC 13. The block division unit 31 offsets the block interval to generate an overlap with the adjacent block M / N, and divides the time domain signal into a plurality of blocks. The block division unit 31 buffers the generated plurality of blocks in the internal storage area and performs a serial-to-parallel conversion, thereby generating a plurality of blocks. The block division unit 31 outputs the generated blocks to the Fourier transform unit 32 for each block (step S102).

[0046] The Fourier transform unit 32 performs a Fourier transform on the blocks sequentially output by the block division unit 31 (step S102). The coefficient multiplication unit 33 performs the following processing on each block of the frequency domain signal of the blocks divided into each N frequency bin output by the Fourier transform unit 32. Specifically, the coefficient multiplication unit 33 multiplies by the wavelength dispersion compensation coefficient and the window function whose values are different for each time position and each frequency position. The storage units 50-1 to 50-2 respectively store the blocks output by the Fourier transform unit 32 per unit time. For example, when the block at time t output by the Fourier transform unit 32 is Xm(f), the storage unit 50-1 stores the block Xm-1(f) at time t-1. The storage unit 50-2 stores the block Xm-2(f) at time t-2.

[0047] Each of the multiplication units 40-1 to 40-3 imports a frequency-domain signal from each of the storage units 50-1 to 50-2. Each of the multiplication units 40-1 to 40-3 multiplies each of the imported frequency-domain signals by the wavelength dispersion compensation coefficients H1(f) to H3(f) and the window functions C1(f) to C3(f). Then, each of the multiplication units 40-1 to 40-3 outputs the multiplication result to the addition unit 60 (step S103).

[0048] The addition unit 60 adds the multiplication results output from the multiplication units 40-1 to 40-3 to calculate a frequency-domain signal (step S104). The inverse Fourier transform unit 34 performs an inverse Fourier transform on the frequency-domain signal calculated by the addition unit 60 to generate a time-domain signal (step S105). The inverse Fourier transform unit 34 outputs the generated time-domain signal to the overlapping and clipping unit 35.

[0049] The overlapping and clipping unit 35 clips, as an overlapping part, a part of M samples from each block included in the time-domain signal output from the inverse Fourier transform unit 34. The overlapping and clipping unit 35 performs a parallel-to-serial conversion on each block including (N - M) samples after removing the overlapping part, generates a serial signal, and outputs it (step S106).

[0050] The weight coefficient optimization unit 19 calculates an optimal window function based on the serial signal output from the overlapping and clipping unit 35 and a signal reference signal R obtained by performing wavelength dispersion compensation on the digital electrical signal output from the ADC 13 (step S107). The weight coefficient optimization unit 19 applies the calculated window function to the coefficient multiplication unit 33 (step S108). Specifically, the weight coefficient optimization unit 19 sets an optimal window function for each of the multiplication units 40-1 to 40-3 included in the coefficient multiplication unit 33. Thus, the window function multiplication unit 42-1 included in the multiplication unit 40-1 performs an operation using the newly set window function C1(f). Similarly, the window function multiplication units 42-2 to 42-3 included in the multiplication units 40-2 to 40-3 perform operations using the newly set window functions C2(f) to C3(f).

[0051] Figure 6 It is a diagram for explaining the effects in the first embodiment. Figure 6 The diagram shown is a diagram for optimizing each window function by changing the frequency overlap to evaluate the penalty of dispersion compensation. In Figure 6 it, the horizontal axis shows the wavelength dispersion, and the vertical axis shows the Q value representing the signal quality calculated based on the bit error rate. In addition, "FDE512" shown in the legend represents the block size of the Fourier transform, "DIV7" represents the maximum value of the division number, and "FOL" represents the frequency overlap.

[0052] Figure 6 The diagram shown was simulated under the following conditions. (Simulation conditions) Modulation method: 16QAM (uniform) Baud rate: 60 - 65 GBd Analog F characteristic: Ideal

[0053] It can be seen that CD = 0 is the reference signal quality. When FOL = 0, as CD increases, the Q value deteriorates and the penalty is large. According to Figure 6 the results shown, it can be seen that as FOL increases, the penalty decreases. In the method of this embodiment, since the optimal window function can be calculated by overlapping at each frequency, the penalty can be minimized. In addition, the frequency overlap is selected according to the allowed power, whereby the window function is optimized and the power consumption is suppressed.

[0054] According to the optical receiving device 10 configured as described above, it includes: a block dividing unit 31 that divides an electrical digital input signal obtained from an input optical signal into blocks of a certain length in such a way that a predetermined length of overlap is generated with adjacent blocks; a Fourier transform unit 32 that transforms each block into a plurality of frequency domain signals by performing a Fourier transform on each block; a coefficient multiplying unit 33 that multiplies each of the plurality of frequency domain signals by a wavelength dispersion compensation coefficient and a window function to generate a coefficient applied block after summing the multiplication results; an inverse Fourier transform unit 34 that performs an inverse Fourier transform on the coefficient applied block; an overlap cutting unit 35 that removes the overlapping part from the coefficient applied block to generate an output signal; and a weight coefficient optimization unit 19 that optimizes the window function based on the output signal and the input signal so as to minimize the waveform distortion in the time domain. Thus, the optimal window function for synthesis with the optimal weight can be calculated, and the compensation performance can be maximally exerted. For each frequency overlap, the optimal window function can be calculated, and thus, the trade-off between power consumption and compensation performance can be controlled. Therefore, the weight coefficient for wavelength dispersion compensation can be optimized.

[0055] (Variant 1) In the above embodiment, the structure in which the wavelength dispersion compensation device 14 optimizes the window function based on the received signal received from the optical transmitting device is shown. The wavelength dispersion compensation device 14 may also be configured to use the dispersion-loaded and dispersion-compensated waveforms in the Nyquist waveform with extended bandwidth to obtain the optimal window function to minimize the compensated error. In the case of such a configuration, the wavelength dispersion compensation device 14 may be provided, for example, in Figure 7 the weight coefficient optimization device 1 shown.

[0056] Figure 7This is a diagram showing a structural example of the weight coefficient optimization device 1 in a modified example of the first embodiment. The weight coefficient optimization device 1 is a device that optimizes a window function as a weight coefficient used in frequency domain equalization. The weight coefficient optimization device 1 includes a signal generation unit 2, a phase shift unit 3, a wavelength dispersion loading unit 4, and a wavelength dispersion compensation device 14.

[0057] The signal generation unit 2 generates a reference signal R for optimizing the weight coefficient. The reference signal R generated by the signal generation unit 2 is a signal having a spectrum equal to or wider than the signal band. Regarding this reference waveform R, a time-domain signal is assumed. The signal generation unit 2 outputs the generated reference signal R to the phase shift unit 3 and the wavelength dispersion compensation device 14.

[0058] The phase shift unit 3 generates a plurality of waveforms (hereinafter referred to as "phase-shifted waveforms") that are time-shifted in units of samples or less than sample units by time-shifting the reference signal R generated by the signal generation unit 2 in units of samples or less than sample units. The phase shift unit 3 outputs the plurality of phase-shifted waveforms to the wavelength dispersion loading unit 4.

[0059] The wavelength dispersion loading unit 4 loads wavelength dispersion on each of the plurality of phase-shifted waveforms generated by the phase shift unit 3. Thereby, the wavelength dispersion loading unit 4 simulates the wavelength dispersion generated in the optical transmission path. In addition, the wavelength dispersion loading unit 4 can also intentionally load noise on the plurality of phase-shifted waveforms as needed.

[0060] The wavelength dispersion compensation device 14 compensates for the wavelength dispersion loaded on the signal by frequency domain equalization. The wavelength dispersion compensation device 14 is composed of a wavelength dispersion compensation unit 18 and a weight coefficient optimization unit 19.

[0061] The wavelength dispersion compensation unit 18 compensates for the wavelength dispersion based on the wavelength dispersion compensation coefficient and window function set by the weight coefficient optimization unit 19, and the plurality of phase-shifted waveforms loaded with wavelength dispersion by the wavelength dispersion loading unit 4. Specifically, the wavelength dispersion compensation unit 18 divides the input signal by the block size N of the Fourier transform, and transforms each of the divided blocks into a frequency domain signal. At this time, the wavelength dispersion compensation unit 18 divides while overlapping in the time domain. The frequency domain signal is generated for each block, but instead of using all the frequencies of all the blocks, only the necessary parts are selected and used.

[0062] The weight coefficient optimization unit 19 obtains information on the wavelength dispersion respectively loaded on multiple phase shift waveforms from the wavelength dispersion loading unit 4. The weight coefficient optimization unit 19 estimates the wavelength dispersion compensation amount based on the obtained wavelength dispersion information. The weight coefficient optimization unit 19 sets a correction compensation coefficient corresponding to the estimated wavelength dispersion compensation amount in the wavelength dispersion compensation unit 18. Additionally, when the wavelength dispersion compensation amount is known through other measurement methods, the weight coefficient optimization unit 19 may not estimate the wavelength dispersion compensation amount. In this case, the weight coefficient optimization unit 19 sets a correction compensation coefficient corresponding to the known wavelength dispersion compensation amount in the wavelength dispersion compensation unit 18.

[0063] In addition, the weight coefficient optimization unit 19 optimizes the window function so that the error between the reference signal R output from the signal generation unit 2 and the serial signal output from the wavelength dispersion compensation unit 18 is minimized. The weight coefficient optimization unit 19 adds signals from multiple blocks for the same frequency, but arranges them by resolving their overlap. Thereby, the weight coefficient optimization unit 19 combines multiple frequency domain signals to generate a combined signal vector. The combined signal vector of the arranged signals is equivalent to the above-mentioned frequency domain signal X ext . The weight coefficient optimization unit 19 similarly ext resolves the overlap and arranges for the dispersion compensation coefficient H ext and the window function C ext . Thereby, the weight coefficient optimization unit 19 combines multiple wavelength dispersion compensation coefficients respectively multiplied by multiple frequency domain signals to generate a combined coefficient vector. The weight coefficient optimization unit 19 transforms into a time domain signal by multiplying the frequency domain signal X ext by the dispersion compensation coefficient H ext and the window function C ext , and multiplying the multiplication result by the inverse DFT matrix (equivalent to IFFT). The weight coefficient optimization unit 19 calculates the optimal window function C ext in such a way as to minimize the error between the transformed time domain signal and the reference waveform R.

[0064] (Second Embodiment) In the second embodiment, a structure for calculating the optimal window function by linear approximation is described. The functional structure of the optical receiving device 10 in the second embodiment is the same as that in the first embodiment, but the processing performed by the wavelength dispersion compensation device 14 is different. Hereinafter, the description will focus on the differences.

[0065] The weight coefficient optimization unit 19 uses the difference between the frequency index of the window function and the index equivalent to the FFT block switching point FSI as a variable, classifies the region according to the index difference, and optimizes the window function by polynomial approximation, first-order approximation, or approximation based on only slices, with the optimized values for each region as coefficients. Additionally, the index equivalent to the FFT block switching point FSI may also be a fractional value.

[0066] Figure 8 This is a diagram for explaining the outline of the method for calculating the optimal window function in the second embodiment by linear approximation. Here, as shown in the upper diagram of Figure 8 , the window functions C1(f), C2(f), and C3(f) are taken as examples for explanation. As shown in the upper diagram of Figure 8 , the region where the frequency ranges overlap, such as the window functions C2(f) and C3(f) for example, is called Frequency overlap. The weight coefficient optimization unit 19 divides the region using the relative relationship (i - fsi) between the frequency index (Frequencyindex) i and the FFT block switching point FSI. Specifically, as shown in the lower diagram of Figure 8 , the weight coefficient optimization unit 19 divides the region according to the value of (i - fsi) such that, for example, the range from -1 to 0 is called category "-1" and the range from +1 to +2 is called category "2".

[0067] The weight coefficient optimization unit 19 approximately calculates the synthesized window function c(i) based on the following equation (5).

[0068] [Equation 5] C(i) = a n (i - fsi) + b n , i ∈ {0,..., N - 1}, N: FFT size ··· Equation (5)

[0069] The region is divided using (i - fsi), and linear approximation coefficients a n , b n are set for each divided region. The linear approximation coefficients a n , b n depend on the frequency overlap but do not depend on the division delay.

[0070] In the linear approximation method, there is a tendency for the wavelength dispersion range in which the estimated linear coefficients can be applied to be narrow. In particular, it becomes narrower when the symbol rate increases. Therefore, there is a problem that multiple linear approximation coefficients need to be stored in advance. For this problem, by including a variable proportional to the wavelength dispersion amount in the linear approximation to make it two-dimensional linear approximation, the wavelength dispersion range in which the estimated linear approximation coefficients can be applied can be enlarged. Specifically, in addition to the relative relationship (i - fsi) between the frequency index i and FSI, the weight coefficient optimization unit 19 also uses a variable n div proportional to the wavelength dispersion, and calculates the optimal window function c(i) based on the linear approximation equation shown in the following equation (6). Using the variable n divThis is an example. As long as it is a variable proportional to wavelength dispersion, it can also be other variables. In Equation (6), N represents the block size of the Fourier transform, and k represents the index of the category. In order to improve the accuracy of the linear approximation, the weight coefficient optimization unit 19 classifies according to the value of (i - fsi) into categories, and optimizes the coefficients of the linear approximation in each category. For example, the value obtained by rounding (i - fsi) to an integer can also be used as the category. In addition, the content where the absolute value of (i - fsi) exceeds FOL becomes Figure 8 the flat area in the following figure, so there is little change, and they can all be of the same category.

[0071] [Equation 6] C k C(i) = a k (i - fsi) + b k + c k (i - fsi)n div + d k n div , i ∈ {0,..., N - 1} ···Equation (6)

[0072] The weight coefficient optimization unit 19 optimizes the coefficients a k , b k , c k and d k shown in the above Equation (6). In addition, for simplicity, there is also a method of fixing the coefficients a k and c k to zero. As a specific example of the variable proportional to wavelength dispersion, a variable proportional to the product of the square of the sample rate of f de for wavelength dispersion and divided by the block size of the Fourier transform is used.

[0073] In addition, in the case of performing an approximation based only on slices, the index difference is 0. In this case, since the weight coefficient optimization unit 19 does not use the variable (i - fsi), the region is divided within a certain range. Then, the weight coefficient optimization unit 19 calculates the optimal window function c(i) using different coefficients for each of the divided regions.

[0074] Figure 9 is a diagram showing an example of the window function calculated by the linear approximation in the second embodiment.

[0075] (Variant Example 1 Common to the First Embodiment and the Second Embodiment) In the above-described embodiments, a configuration is shown in which the optical receiving device 10 is provided with the wavelength dispersion compensating device 14. However, the wavelength dispersion compensating device 14 may also be configured as a single device. In such a case, the wavelength dispersion compensating device 14 may use, as an input signal, any one of the signals shown below or a signal obtained by combining the signals shown below and having wavelength dispersion distortion imposed thereon and converted into a digital signal. The signal obtained by combining the signals shown below may be any combination as long as it is a signal combining two or more signals shown below. · A signal obtained by convolving an impulse signal with an RRC (Root Raised Cosine) waveform · A signal of a sequence obtained by upsampling a random signal sequence with an RRC waveform · A signal having a bandwidth wider than that of the signal originally processed · A plurality of signals of a signal sequence in which phases are offset one by one · A plurality of signals of a signal sequence in which phases are offset by a fractional amount · A signal having its phase rotated by 90 degrees · A signal with noise added thereto.

[0076] Part of the functions of the optical receiving device 10 in the above-described embodiments may also be implemented by a computer. In this case, it may also be implemented by recording a program for implementing this function in a computer-readable recording medium and causing a computer system to read and execute the program recorded in this recording medium. Here, the “computer system” includes hardware such as an OS (Operating System) and peripheral devices.

[0077] In addition, the “computer-readable recording medium” refers to removable media such as floppy disks, magneto-optical disks, ROMs (Read Only Memories), CD-ROMs (Compact Disc Read Only Memories), and storage devices such as hard disks built into a computer system. Further, the “computer-readable recording medium” may also include a medium that dynamically holds a program for a short period of time, such as a communication line when a program is transmitted via a network such as the Internet or a communication line such as a telephone line, and a medium that holds a program for a certain period of time, such as a volatile memory inside a computer system that becomes a server or a client in this case. In addition, the above program may be a program for implementing a part of the above functions, or may also be a program that implements the above functions by combining with a program already recorded in a computer system, or may also be a program implemented using a programmable logic device such as an FPGA (Field Programmable Gate Array).

[0078] As described above, the embodiments of the present invention have been described in detail with reference to the accompanying drawings. However, the specific structure is not limited to this embodiment and also includes designs that do not depart from the gist of the present invention, etc.

[0079] Industrial Applicability The present invention can be applied to technologies for wavelength dispersion compensation.

[0080] Description of Reference Numerals 10... optical receiving device, 11... optical receiving section, 12... digital signal processing section, 13... ADC, 14... wavelength dispersion compensation device, 15... adaptive equalization section, 16... demapping section, 17... decoding section, 18... wavelength dispersion compensation section, 19... weight coefficient optimization section, 31... block partitioning section, 32... Fourier transform section, 33... coefficient multiplication section, 34... inverse Fourier transform section, 35... overlap and cut section, 40-1 to 40-O... multiplication sections, 41-1 to 41-O... multiplication sections, 42-1 to 42-O... window function multiplication sections, 50-1 to 50-(O-1)... storage sections, 60... addition section.

Claims

1. A wavelength dispersion compensation device, wherein, Comprising: a block segmentation unit that segments an input signal with wavelength dispersion distortion into blocks of a certain length in such a way that an overlap of a predetermined length is generated with adjacent blocks; a Fourier transform unit that transforms each of the blocks segmented by the block segmentation unit into a plurality of frequency domain signals by performing a Fourier transform on each block; a coefficient multiplication unit that multiplies each of the plurality of frequency domain signals transformed by the Fourier transform unit by each of a wavelength dispersion compensation coefficient and a window function or a value obtained by synthesizing the dispersion compensation coefficient and the window function, and generates a block with the multiplication results summed up and the coefficients applied; an inverse Fourier transform unit that performs an inverse Fourier transform on the block with the coefficients applied generated by the coefficient multiplication unit; an overlap cutting unit that removes the overlapping part from the block with the coefficients applied transformed by the inverse Fourier transform unit to generate an output signal; and a weight coefficient optimization unit that optimizes the window function in such a way that the waveform distortion in the time domain is minimized based on the output signal generated by the overlap cutting unit and a desired waveform with wavelength dispersion compensation applied to the input signal.

2. The wavelength dispersion compensation device according to claim 1, wherein the input signal is a test waveform with wavelength dispersion distortion loaded on a reference signal pattern having a frequency band of the signal to be processed or a higher frequency band, the weight coefficient optimization unit optimizes the window function by the least squares method in such a way that the error between the output signal obtained based on the test waveform and the reference signal pattern is minimized.

3. The wavelength dispersion compensation device according to claim 2, wherein, The weight coefficient optimization unit calculates the covariance matrix of the time domain signals obtained by applying the inverse Fourier transform of the inverse Fourier transform unit after multiplying each of the plurality of frequency domain signals by the wavelength dispersion compensation coefficient, calculates the correlation vectors of the wavelength dispersion compensation coefficient, the time domain signals, and the reference signal pattern, and optimizes the window function by applying the inverse matrix of the calculated covariance matrix to the correlation vectors.

4. The wavelength dispersion compensation device according to claim 3, wherein, The weight coefficient optimization unit: generates a combined signal vector by unraveling and combining the overlaps of the plurality of frequency domain signals, generates a combined coefficient vector by unraveling and combining the overlaps of the plurality of wavelength dispersion compensation coefficients multiplied by the plurality of frequency domain signals respectively, generates a combined inverse discrete Fourier matrix by unraveling and combining the overlaps of the matrix equivalent to the inverse Fourier transform, calculates the element product of the covariance matrix of the combined signal vector, the covariance matrix of the combined coefficient vector, and the covariance matrix of the combined inverse discrete Fourier matrix, calculates the inverse matrix of the element product, generates a correlation vector with the element product of the combined coefficient vector obtained after calculating the matrix product of the complex conjugate of the combined coefficient vector, the combined signal vector, the combined inverse discrete Fourier matrix and the reference signal pattern, and optimizes the window function by multiplying the correlation vector by the inverse matrix of the covariance matrix.

5. The wavelength dispersion compensation device according to claim 1, wherein, The weight coefficient optimization unit: Using the difference between the frequency index of the window function and an index that can be a small value equivalent to the switching frequency of the block as a variable, classify the regions according to the index difference, and optimize the window function by using polynomial approximation, first-order approximation, or approximation based on only slices, with the optimized values for each region as coefficients respectively.

6. The wavelength dispersion compensation device according to claim 5, wherein, The weight coefficient optimization unit: When calculating the window function by the polynomial approximation, also use a variable proportional to the wavelength dispersion amount to optimize the window function.

7. The wavelength dispersion compensation device according to any one of claims 1 to 6, wherein, The input signal is: Any one of a signal obtained by convolving an impulse signal with an RRC (Root Raised Cosine), a signal of a sequence obtained by upsampling a random signal sequence with an RRC waveform, a signal with an extended frequency band compared to the originally processed signal, multiple signal sequences with phase shifted by one sample each, multiple signal sequences with phase shifted by a fractional sample each, a signal with a phase rotated by 90 degrees, a signal with noise added thereto, or a test waveform with wavelength dispersion distortion imposed on a signal combining these signals.

8. A coefficient optimization method, wherein Divide an input signal with wavelength dispersion distortion into blocks of a certain length in such a way that adjacent blocks have a predetermined length of overlap; Transform each of the divided blocks into a plurality of frequency-domain signals by performing a Fourier transform on each block; multiply each of the transformed plurality of frequency-domain signals by each of a wavelength dispersion compensation coefficient and a window function or a value obtained by synthesizing the dispersion compensation coefficient and the window function; Generate a block with the coefficients applied after summing the multiplication results; Perform an inverse Fourier transform on the block with the coefficients applied; Remove the overlapping part from the transformed block with the coefficients applied to generate an output signal; Optimize the window function in such a way that the waveform distortion in the time domain is minimized based on the generated output signal and a desired waveform with wavelength dispersion compensation applied to the input signal.

Images