Low-complexity MLSE balancing method and system based on mode sharing lookup table

Abstract

The invention discloses a low-complexity MLSE balancing method and system based on a mode sharing lookup table, and is applied to the technical field of short-distance optical communication. The method comprises the following steps: a transmitting end converts a pseudo-random binary sequence into a PAM4 signal through Gray mapping; the arbitrary waveform generator carries out pre-emphasis on the PAM4 signal; the PAM4 signal is amplified by an electric amplifier and modulated by a Mach-Zehnder modulator; the modulated PAM4 signal is transmitted by a standard single-mode optical fiber, detected by a photodiode and captured by a real-time oscilloscope; and the receiving end eliminates the inter-symbol interference of the signal through feed-forward equalization and MLSE equalization based on the mode sharing lookup table. Compared with the traditional LUT and MLSE, the method has the advantages that the calculation complexity is greatly reduced, and the system performance cost can be ignored.

Description

Technical Field

[0001] The present invention relates to the technical field of short-distance optical communication, and more specifically, to a low-complexity MLSE equalization method and system based on a pattern-sharing lookup table. Background Art

[0002] With the rapid development of modern technology, the number of Internet users continues to grow, and society has entered the digital and information age. Under this background, efficiently utilizing limited spectrum resources to achieve high-speed and large-capacity data transmission has become the central issue of concern in the industry. Short-distance optical communication has received increasing attention because it can meet the requirements of high-speed transmission. However, with the growing demand for bandwidth efficiency and power efficiency, it is necessary to control costs while ensuring transmission efficiency. Intensity Modulation / Direct Detection (IM / DD) systems have been widely used in short-distance optical communication due to their simple structure and low cost. However, when the inter-symbol-interference (ISI) is relatively severe, the performance of IM / DD systems can no longer meet the requirements.

[0003] To effectively solve ISI, existing technical solutions usually adopt advanced Digital Signal Processing (DSP) equalization algorithms. At the receiving end, feed-forward equalization (FFE) is the most commonly used solution to solve ISI. The signal processed by FFE may over-compensate for noise in the high-frequency region. When ISI becomes more severe, FFE cannot meet the performance requirements. To further improve the system performance, a solution combining FFE, a postfilter (PF), and Maximum Likelihood Sequence Estimation (MLSE) has been proposed to more effectively solve the ISI problem. PF is a two-tap filter that effectively reduces the enhanced noise but also introduces known ISI. MLSE solves the remaining ISI after FFE processing and the ISI introduced by PF. When the IM / DD system faces severe ISI, MLSE will lose some non-linear ISI when calculating the Euclidean distance. To solve this problem, a training-based lookup table (LUT) is used after PF to calculate the non-linear response, further improving the system performance. Since the complexity of MLSE is proportional to the memory length, the computational complexity of MLSE also increases after the non-linear response is recalculated by LUT, limiting its performance in practical applications.

[0004] Currently, in response to the problem of high computational complexity of MLSE, researchers at home and abroad have also carried out various studies, mainly divided into two types. One is the approximation algorithm, which simplifies the Viterbi algorithm. By restricting the number of possible paths, the state space is simplified, and the path calculation at each step is reduced. Common methods include restricting the number of paths in the Viterbi algorithm to a fixed constant, or using soft decisions instead of hard decisions to reduce the number of path comparisons. By selecting a partial state subset to perform MLSE instead of calculating all state spaces. It can be done in a dynamic programming way. Based on known prior information or channel models, it is speculated which states are more likely to occur, thereby reducing the computational amount. Another method combines hard decisions and soft decisions. In traditional MLSE, hard decisions are usually used to classify and select signals, but hard decisions may lead to larger errors. Therefore, a method of combining hard decisions and soft decisions is proposed. Soft decision MLSE calculates based on the reliability of signals by introducing soft decision signals, rather than relying only on hard decision values. Soft decisions provide more information and can more accurately reflect the reliability of signals, which can reduce the number of paths and the computational complexity. Hard decision optimization can effectively reduce the computational complexity of MLSE by restricting the number of possible symbol combinations and the number of path updates without increasing significant complexity.

[0005] In summary, in the above technologies, the computational complexity of MLSE is reduced, but it is accompanied by a significant loss of system performance. Therefore, how to provide a low-complexity MLSE equalization method and system based on a pattern-sharing lookup table is an urgent problem to be solved by those skilled in the art. Summary of the Invention

[0006] In view of this, the present invention provides a low-complexity MLSE equalization method and system based on a pattern-sharing lookup table. Utilizing the characteristic that inter-symbol interference is related to adjacent front and rear code elements, when generating the lookup table, the lookup table patterns with the same front and rear symbols are integrated. The obtained pattern-integrated lookup table is given to MLSE to calculate the Euclidean distance, greatly reducing the addition and path comparisons required for calculation, and realizing the reduction of the lookup table size and the complexity of MLSE while ensuring that the system performance meets the requirements.

[0007] To achieve the above object, the present invention provides the following technical solutions:

[0008] A low-complexity MLSE equalization method based on a pattern-sharing lookup table, comprising the following steps:

[0009] S1. The transmitting end converts the pseudo-random binary sequence into a PAM4 signal through Gray mapping;

[0010] S2. The arbitrary waveform generator pre-emphasizes the PAM4 signal;

[0011] S3. The PAM4 signal is amplified by an electrical amplifier and modulated by a Mach-Zehnder modulator;

[0012] S4. After being transmitted through a standard single-mode fiber, the modulated PAM4 signal is detected by a photodiode and captured by a real-time oscilloscope;

[0013] S5. At the receiving end, the inter-symbol interference of the signal is eliminated through feed-forward equalization and MLSE equalization based on a pattern-sharing lookup table.

[0014] Optionally, S5 is specifically as follows:

[0015] S51. The received signal is processed using feed-forward equalization and a post-filter;

[0016] S52. The output signal of the post-filter and the transmitted training sequence share intermediate symbols, only considering the inter-symbol interference of adjacent symbols before and after, and integrating patterns with the same adjacent symbols;

[0017] S53. The MLSE algorithm based on the Viterbi algorithm is used to eliminate the remaining inter-symbol interference;

[0018] S54. The performance is evaluated by calculating the bit error rate.

[0019] Optionally, S51 is specifically as follows:

[0020] Clock recovery is performed through a digital square algorithm, and the recovery process includes normalization and resampling; feed-forward equalization is used to suppress inter-symbol interference caused by bandwidth limitation and dispersion; a post-filter is used to process the high-frequency noise introduced by the feed-forward equalization.

[0021] Optionally, S52 is specifically as follows:

[0022] The output signal of the post-filter is:

[0023] R(i) = r(i) + α * r(i + 1)

[0024] In the formula, r(i) represents the input signal of the post-filter, α is the second tap coefficient of the post-filter, and i represents the current time;

[0025] A lookup table for state integration is obtained using the training sequence:

[0026]

[0027] LUT(i) = LUT(i) + e(i)

[0028] e(i) = R(i) - X(i)

[0029] Where, e(i) is the error corresponding to each occurrence of the fixed channel response pattern, R(i) is the output of the post-filter, X(i) is the transmitted signal at the corresponding moment, i is the current moment, LUT(i) is the cumulative error of the fixed channel response pattern, N is the number of occurrences of the fixed channel, and LUT (abc) is the value of the fixed channel in the look-up table;

[0030] After obtaining the look-up table, since the inter-symbol interference only affects the symbols before and after the current symbol, only the value of the current symbol is considered, and the patterns with the same adjacent symbols are integrated to reduce the number of states of the look-up table, obtaining a pattern-sharing look-up table.

[0031] Optionally, S53 is specifically:

[0032] Calculate the MLSE Euclidean distance based on the pattern-sharing look-up table:

[0033] d(k) = (R(k) - LUT (abc) - x(k)) 2

[0034] Where, d(k) represents the MLSE Euclidean distance; compare the MLSE Euclidean distances at each time to identify the shortest branch metric path, and set the threshold using the pattern-sharing look-up table:

[0035] threshold = LUT(i) - R(i)

[0036] When the threshold is less than -2, the value corresponding to the shortest path at the current moment is 3; when the threshold is set between -2 and 0, the value corresponding to the shortest path at the current moment is 1; when the threshold is between 0 and 2, the value corresponding to the shortest path at the current moment is -1; in other cases, the value corresponding to the shortest path at the current moment is -3.

[0037] A low-complexity MLSE equalization system based on a pattern-sharing look-up table executes the above-mentioned low-complexity MLSE equalization method based on a pattern-sharing look-up table, including a transmitter, an arbitrary waveform generator, an electrical amplifier, a Mach-Zehnder modulator, a standard single-mode fiber, a photodiode, a real-time oscilloscope, and a receiver connected in sequence. The Mach-Zehnder modulator is also connected to a laser; wherein, the transmitter converts the pseudo-random binary sequence into a PAM4 signal through Gray mapping; the arbitrary waveform generator performs pre-emphasis on the PAM4 signal; the PAM4 signal is amplified by the electrical amplifier and modulated by the Mach-Zehnder modulator; after the modulated PAM4 signal is transmitted through the standard single-mode fiber, it is detected by the photodiode and captured by the real-time oscilloscope; the receiver eliminates the inter-symbol interference of the signal.

[0038] Optionally, the receiving end includes a feed-forward equalization unit, a post-filter, a pattern sharing lookup table unit, and an MLSE unit connected in sequence; among them, the feed-forward equalization unit is used to suppress the inter-symbol interference caused by bandwidth limitation and dispersion, the post-filter is used to process the high-frequency noise introduced by the feed-forward equalization, the pattern sharing lookup table unit is used to integrate patterns with the same adjacent symbols, and the MLSE unit is used to eliminate the remaining inter-symbol interference.

[0039] As can be seen from the above technical solutions, compared with the prior art, the present invention provides a low-complexity MLSE equalization method and system based on a pattern sharing lookup table, which has the following beneficial effects: The present invention shares intermediate symbols through the output signal of the post-filter and the transmitted training sequence, only considers the inter-symbol interference of adjacent symbols before and after, and integrates patterns with the same adjacent symbols, so that the size of the obtained lookup table is greatly reduced; the integrated lookup table is given to the MLSE, which greatly reduces the number of addition operations when the MLSE calculates the Euclidean distance. At the same time, the MLSE sets a threshold when comparing paths, and the number of comparisons is also greatly reduced. Compared with the traditional lookup table, the present invention greatly reduces the computational complexity while the system performance cost paid can be ignored. Description of the Drawings

[0040] In order to more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the following will briefly introduce the drawings required for the description of the embodiments or the prior art. Obviously, the drawings in the following description are only the embodiments of the present invention. For those of ordinary skill in the art, other drawings can be obtained according to the provided drawings without creative work.

[0041] Figure 1 It is a flowchart of the low-complexity MLSE equalization method of the present invention;

[0042] Figure 2 It is a flowchart of the digital signal processing at the receiving end of the present invention;

[0043] Figure 3 It is a schematic diagram of the low-complexity MLSE equalization system of the present invention. Detailed Embodiments

[0044] The following will clearly and completely describe the technical solutions in the embodiments of the present invention with reference to the drawings in the embodiments of the present invention. Obviously, the described embodiments are only a part of the embodiments of the present invention, rather than all of the embodiments. All other embodiments obtained by those of ordinary skill in the art based on the embodiments of the present invention without creative work belong to the scope of protection of the present invention.

[0045] The embodiments of the present invention disclose a low-complexity MLSE equalization method based on a pattern sharing lookup table, as Figure 1As shown in the figure, it includes the following steps:

[0046] S1. The transmitting end converts the pseudo-random binary sequence into a PAM4 signal through Gray mapping;

[0047] S2. The arbitrary waveform generator pre-emphasizes the PAM4 signal;

[0048] S3. The PAM4 signal is amplified by an electrical amplifier and modulated by a Mach-Zehnder modulator;

[0049] S4. After the modulated PAM4 signal is transmitted through a standard single-mode fiber, it is detected by a photodiode and captured by a real-time oscilloscope;

[0050] S5. The receiving end eliminates the inter-symbol interference of the signal through feed-forward equalization and MLSE equalization based on a pattern-sharing lookup table.

[0051] Furthermore, as Figure 2 shown in the figure, S5 is specifically as follows:

[0052] S51. Process the received signal using feed-forward equalization and a post-filter;

[0053] S52. Share the intermediate symbols through the output signal of the post-filter and the transmitted training sequence, only consider the inter-symbol interference of the adjacent symbols before and after, and integrate the patterns with the same adjacent symbols;

[0054] S53. Use the MLSE algorithm based on the Viterbi algorithm to eliminate the remaining inter-symbol interference;

[0055] S54. Evaluate the performance by calculating the bit error rate.

[0056] Furthermore, S51 is specifically as follows:

[0057] Perform clock recovery through the digital square algorithm, and the recovery process includes normalization and resampling; use feed-forward equalization to suppress the inter-symbol interference caused by bandwidth limitation and dispersion; use a post-filter to process the high-frequency noise introduced by feed-forward equalization.

[0058] After feed-forward equalization, although most of the inter-symbol interference has been eliminated, it also leads to an increase in noise. In order to further suppress this increased noise, in the embodiment of the present invention, the signal quality is improved through a two-tap post-filter, and the post-filter can effectively reduce the influence of noise, and its transfer function is:

[0059] H(z) = 1 + αz -1

[0060] In the formula, z represents the Laplace transform.

[0061] Furthermore, S52 is specifically as follows:

[0062] The output signal of the post-filter is:

[0063] R(i) = r(i) + α * r(i + 1)

[0064] Wherein, r(i) represents the input signal of the post-filter, α is the second tap coefficient of the post-filter, and i represents the current time;

[0065] Obtain a lookup table for state integration using the training sequence:

[0066]

[0067] LUT(i) = LUT(i) + e(i)

[0068] e(i) = R(i) - X(i)

[0069] Wherein, e(i) is the error corresponding to each occurrence of the fixed channel response pattern, R(i) is the output of the post-filter, X(i) is the transmitted signal at the corresponding time, i is the current time, LUT(i) is the cumulative error of the fixed channel response pattern, N is the number of occurrences of the fixed channel, and LUT (abc) is the value of the fixed channel in the lookup table;

[0070] After obtaining the lookup table, since the inter-symbol interference only affects the symbols before and after the current symbol, only the value of the current symbol is considered, and the patterns with the same adjacent symbols are integrated to reduce the number of states in the lookup table, thereby obtaining a pattern-sharing lookup table.

[0071] Further, S53 is specifically:

[0072] Calculate the MLSE Euclidean distance based on the pattern-sharing lookup table:

[0073] d(i) = (R(i) - LUT (abc) - x(i)) 2

[0074] Wherein, d(i) represents the MLSE Euclidean distance; compare the MLSE Euclidean distances at each time to identify the shortest branch metric path, and set the threshold using the pattern-sharing lookup table:

[0075] threshold = LUT(i) - R(i)

[0076] When the threshold is less than -2, the value corresponding to the shortest path at the current time is 3; when the threshold is set between -2 and 0, the value corresponding to the shortest path at the current time is 1; when the threshold is between 0 and 2, the value corresponding to the shortest path at the current time is -1; in other cases, the value corresponding to the shortest path at the current time is -3.

[0077] In an embodiment of the present invention, when the signal is PAM4 and the MLSE memory length is 1, the corresponding possible paths are 4 2 = 16, because there are 4 possible states (-3, -1, 1, 3) corresponding to each moment of PAM4, and there is a set of corresponding distances between every two states. The final surviving path is expressed as:

[0078] d(i) = min{d(i - 1)+(R(i)-LUT (abc) -x(i)) 2}

[0079] When the intersymbol interference is severe, as the channel response increases, the computational complexity of the lookup table-based MLSE increases accordingly. To further optimize the algorithm and reduce the computational complexity of the algorithm, it is necessary to significantly reduce the size of the lookup table and the corresponding MLSE branch metrics. The schematic diagram of the MLSE based on the pattern-integrated lookup table is as Figure 2 shown. In the figure, (a) represents the computational flow of the lookup table-based MLSE, and (b) represents the computational flow of the MLSE based on the pattern-integrated lookup table. In Figure 2 , for the case where the memory length is 3, there are 64 possibilities for the lookup table. Since the intersymbol interference only affects the symbols before and after the current symbol, the value of the current symbol is not considered in the embodiment of the present invention. The stages (3, 3, 3), (3, 1, 3), (3, 1, 3), (3, 3, 3) are regarded as (3, x, 3). For the lookup table with a length of 3, the number of states of the lookup table can be optimized by 75%, from 64 to 16. Similarly, for the lookup table with a length of 5, the number of states can be reduced from 1024 to 254. When the MLSE is looking for the shortest path, it is necessary to compare the Euclidean distances at each time. For example, in the lookup table-based MLSE algorithm, when the next moment is -3, the shortest branch metric path is identified among "(-3, -3, -3), (-3, -1, -3), (-3, 1, -3), (-3, 3, -3)". At a certain moment, 16 paths are compared 12 times. By setting a threshold using the pattern-integrated lookup table, 16 paths only need to be compared 4 times. Compared with the traditional lookup table-based MLSE, the number of comparators is reduced by 75%.

[0080] Corresponding to Figure 1 the method described above, an embodiment of the present invention also discloses a low-complexity MLSE equalization system based on a pattern-sharing lookup table, which executes the above-mentioned low-complexity MLSE equalization method based on the pattern-sharing lookup table, as Figure 3As shown in the figure, it includes a transmitting end, an arbitrary waveform generator, an electrical amplifier, a Mach-Zehnder modulator, a standard single-mode optical fiber, a photodiode, a real-time oscilloscope, and a receiving end connected in sequence. The Mach-Zehnder modulator is also connected to a laser; among them, the transmitting end converts the pseudo-random binary sequence into a PAM4 signal through Gray mapping; the arbitrary waveform generator performs pre-emphasis on the PAM4 signal; the PAM4 signal is amplified by the electrical amplifier and modulated by the Mach-Zehnder modulator; after the modulated PAM4 signal is transmitted through the standard single-mode optical fiber, it is detected by the photodiode and captured by the real-time oscilloscope; the receiving end eliminates the inter-symbol interference of the signal.

[0081] Furthermore, the receiving end includes a feed-forward equalization unit, a post-filter, a pattern sharing lookup table unit, and an MLSE unit connected in sequence; among them, the feed-forward equalization unit is used to suppress the inter-symbol interference caused by bandwidth limitation and dispersion, the post-filter is used to process the high-frequency noise introduced by the feed-forward equalization, the pattern sharing lookup table unit is used to integrate the patterns with the same adjacent symbols, and the MLSE unit is used to eliminate the remaining inter-symbol interference.

[0082] In this specification, each embodiment is described in a progressive manner. The key point of each embodiment is to illustrate the differences from other embodiments. The same or similar parts among the embodiments can be referred to each other. For the system disclosed in the embodiments, since it corresponds to the method disclosed in the embodiments, the description is relatively simple. For the relevant parts, please refer to the description in the method section.

[0083] The above description of the disclosed embodiments enables those skilled in the art to implement or use the present invention. Various modifications to these embodiments will be obvious to those skilled in the art. The general principles defined herein can be implemented in other embodiments without departing from the spirit or scope of the present invention. Therefore, the present invention will not be limited to these embodiments shown herein, but will be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A low-complexity MLSE equalization method based on a pattern-sharing lookup table, characterized in that, It includes the following steps: S1. The transmitting end converts the pseudo-random binary sequence into a PAM4 signal through Gray mapping; S2. The arbitrary waveform generator pre-emphasizes the PAM4 signal; S3. The PAM4 signal is amplified by an electrical amplifier and modulated by a Mach-Zehnder modulator; S4. After the modulated PAM4 signal is transmitted through a standard single-mode optical fiber, it is detected by a photodiode and captured by a real-time oscilloscope; S5. The receiving end eliminates the inter-symbol interference of the signal through feed-forward equalization and MLSE equalization based on a pattern-sharing lookup table.

2. The low-complexity MLSE equalization method based on a pattern-sharing lookup table according to claim 1, wherein S5 specifically is: S51. Process the received signal using feed-forward equalization and a post-filter; S52. Share the intermediate symbols through the output signal of the post-filter and the transmitted training sequence, only consider the inter-symbol interference of the adjacent symbols before and after, and integrate the patterns with the same adjacent symbols; S53. Use the MLSE algorithm based on the Viterbi algorithm to eliminate the remaining inter-symbol interference; S54. Evaluate the performance by calculating the bit error rate.

3. A low-complexity MLSE equalization method based on a pattern-sharing lookup table according to claim 2, characterized in that S51 specifically is: Perform clock recovery through the digital square algorithm, and the recovery process includes normalization and resampling; use feed-forward equalization to suppress the inter-symbol interference caused by bandwidth limitation and dispersion; use a post-filter to process the high-frequency noise introduced by feed-forward equalization.

4. A low-complexity MLSE equalization method based on a pattern-sharing lookup table according to claim 2, characterized in that, S52 specifically is: The output signal of the post-filter is: R(i) = r(i) + α * r(i + 1) In the formula, r(i) represents the input signal of the post-filter, α is the second tap coefficient of the post-filter, and i represents the current time; Obtain the lookup table for state integration using the training sequence: LUT(i) = LUT(i) + e(i) e(i) = R(i) - X(i) where e(i) is the error corresponding to each occurrence of the fixed channel response pattern, R(i) is the output of the post-filter at the current moment, X(i) is the transmitted signal at the corresponding moment, i is the current moment, LUT(i) is the cumulative error of the fixed channel response pattern, N is the number of occurrences of the fixed channel, and LUT (abc) is the value of the fixed channel in the look-up table; After obtaining the lookup table, since the inter-symbol interference only affects the symbols before and after the current symbol, only consider the value of the current symbol, integrate the patterns with the same adjacent symbols, reduce the number of states of the lookup table, and obtain the pattern-sharing lookup table.

5. A low-complexity MLSE equalization method based on a pattern-sharing lookup table according to claim 3, characterized in that, S53 specifically is: Calculate the MLSE Euclidean distance based on the pattern-sharing lookup table: d(i) = (R(i) - LUT (abc) - x(i)) 2 In the formula, d(k) represents the MLSE Euclidean distance; compare the MLSE Euclidean distances at each time, identify the shortest branch metric path, and set the threshold using the pattern-sharing lookup table: threshold = LUT(i) - R(i) When the threshold is less than -2, the value corresponding to the shortest path at the current time is 3; when the threshold is set between -2 and 0, the value corresponding to the shortest path at the current time is 1; when the threshold is between 0 and 2, the value corresponding to the shortest path at the current time is -1; in other cases, the value corresponding to the shortest path at the current time is -3.

6. A low-complexity MLSE equalization system based on a pattern-sharing lookup table, characterized in that, Implement the low-complexity MLSE equalization method based on a pattern-sharing lookup table according to any one of claims 1-5, including a transmitter, an arbitrary waveform generator, an electrical amplifier, a Mach-Zehnder modulator, a standard single-mode fiber, a photodiode, a real-time oscilloscope, and a receiver connected in sequence. The Mach-Zehnder modulator is also connected to a laser; wherein, the transmitter converts the pseudo-random binary sequence into a PAM4 signal through Gray mapping; the arbitrary waveform generator performs pre-emphasis on the PAM4 signal; the PAM4 signal is amplified by the electrical amplifier and modulated by the Mach-Zehnder modulator; after the modulated PAM4 signal is transmitted through the standard single-mode fiber, it is detected by the photodiode and captured by the real-time oscilloscope; the receiver eliminates the inter-symbol interference of the signal.

7. A low-complexity MLSE equalization system based on a pattern-sharing lookup table according to claim 6, wherein The receiver includes a feed-forward equalization unit, a post-filter, a pattern-sharing lookup table unit, and an MLSE unit connected in sequence; wherein, the feed-forward equalization unit is used to suppress the inter-symbol interference caused by bandwidth limitation and dispersion, the post-filter is used to process the high-frequency noise introduced by the feed-forward equalization, the pattern-sharing lookup table unit is used to integrate patterns with the same adjacent symbols, and the MLSE unit is used to eliminate the remaining inter-symbol interference.

Images