A HPGS 16QAM communication optimization method based on spatial coverage

Abstract

The present invention relates to a spatial coverage-based HPGS16QAM communication optimization method, belonging to the field of communication technology. By applying the concept of spatial coverage to GS shaping, a geometrically shaped constellation-shaped 15-QAM signaling scheme is proposed for the first time. On this basis, a PS+GS+adjuster optimization method for 16-QAM optical fiber channels is innovatively proposed. When the signal-to-noise ratio (SNR) is greater than 10 dB, error-free transmission can be achieved, thereby improving the quality of optical fiber communication. Experimental results show that compared with PS-16QAM, GS-16QAM, and schemes without PS and GS, the present invention reduces the signal-to-noise ratio (OSNR) by 0.7 dB, 0.8 dB, and 1.2 dB, respectively, to achieve error-free transmission. This results in good shaping gain and better transmission performance at a lower cost and lower complexity.

Description

Technical Field

[0001] The invention belongs to the technical field of communications and relates to a HPGS 16QAM communication optimization method based on spatial coverage. Background Art

[0002] As one of the mainstays of contemporary communications, optical fiber communications plays a vital role in telecommunications networks. Optical fiber communications are a key means of transmitting information in the future information society. Optical fiber communications are a cutting-edge communications technology, representing a new technological revolution. They utilize light waves as carrier waves, using optical fibers as the medium for information transmission. Since the 21st century, the rapid development of the internet and the surge in audio, video, and multimedia applications have created an urgent need for increased capacity and transmission quality in optical fiber channels.

[0003] In recent years, probabilistic and geometric shaping technologies have emerged in optical fiber communications, demonstrating promising results for increasing the capacity of optical fiber systems and reducing bit error rates. However, a key issue is how to better combine the strengths of probabilistic and geometric shaping to further improve the quality of optical fiber communications: achieving higher spectral efficiency and transmission capacity, reducing bit error rates, and enhancing the flexibility and reliability of optical communication systems without increasing transmission power or increasing optical fiber system complexity. Summary of the Invention

[0004] In view of this, the purpose of the present invention is to provide a HPGS 16QAM communication optimization method based on spatial coverage. Based on the uniformly distributed 0-1 bit data type, the main technologies adopted in this scheme are geometric shaping (GS), probability shaping (PS), adjuster, FEC coding, etc. The GS part sets the maximization of GMI and minimization of BER as the goal, and uses the automatic encoder neural network training to obtain the optimal constellation coordinates (16-QAM) under a specific SNR. The PS part generates a total of 15 modulation symbols according to the tree code. Among them, the 8 modulation symbols with the 111 prefix (6bit / Symbol) account for 1 / 8, which correspond to the outer circle of the constellation points; the remaining modulation symbols are 7 types (3bit / Symbol), accounting for 7 / 8, which correspond to the inner circle of the constellation points. 15 symbols, corresponding to 16-QAM, will have 1 constellation point remaining. The essence of the adjuster is to fully cover the symbol space during modulation: For the eight modulation symbols with a 111 prefix, statistics are collected and the last symbol with the first occurrence is marked; during modulation, it is mapped to an unoccupied constellation point. Without FEC, high bit error rates would result. Therefore, RS(15,11) codes and CCSDS-standard LDPC codes were introduced, both of which can achieve zero bit error.

[0005] In order to achieve the above object, the present invention provides the following technical solutions:

[0006] A HPGS 16QAM communication optimization method based on spatial coverage, the method comprising the following steps:

[0007] Probabilistic shaping of 15QAM partitioning;

[0008] Geometric shaping based on Gaussian noise GN and nonlinear interference noise NLIN channel models, trained by autoencoder neural network AENN;

[0009] An algorithm based on spatial coverage and a regulator for optical signal modulation / demodulation is performed.

[0010] Optionally, the probability shaping of the 15QAM division is:

[0011] Assuming that the probability of '0' and '1' appearing is equal, the data stream generates a total of 15 modulation symbols according to the tree code; among them, the 8 modulation symbols with the 111 prefix account for 1 / 8, and they are 111000-111111; there are 7 other modulation symbols, accounting for 7 / 8, and they are 000–110.

[0012] Optionally, in the channel model based on Gaussian noise GN and nonlinear interference noise NLIN, the discrete NLIN model is described as follows:

[0013] y=c(x,P 3 ,κ,κ3)

[0014] =x+n ASE +n NLI N

[0015]

[0016]

[0017] Where y represents the received codeword, x represents the transmitted codeword, c(·) represents the channel model, and n ASE Represents variance Gaussian noise samples, n NLIN Represents variance Gaussian noise samples;

[0018] The NLIN model describes the effective signal-to-noise ratio of a fiber optic system as follows:

[0019]

[0020] Assume that all channels have the same optical transmission power, P txrepresents the optical transmission power and draws energy from the same constellation; where P tx represents the optical transmission power, represents the variance of the accumulated amplified spontaneous emission (ASE) noise, represents the variance of NLIN; μ4 represents the fourth-order moment of the constellation, assuming that all channels have the same optical transmission power and draw energy from the same constellation; μ6 represents the sixth-order moment of the constellation.

[0021] Optionally, the spatial coverage is specifically:

[0022] The symbol space S is defined as the set of all possible symbols that may appear in the modulated data D; let

[0023] S={s1,s2,s3,…,s n}

[0024] Card(S) = n, that is, there are n symbols in the modulated data D. The modulation process is regarded as the process of obtaining symbols one by one from a certain moment. A shortest data block B is obtained in D to satisfy:

[0025]

[0026] Among them, Freq(x B ) represents the frequency of occurrence of symbol x in data block B; the shortest data block B that satisfies the above formula completes the full coverage of the symbol space S;

[0027] Let M be the length of data block B, and B[-1] be the last symbol of B. When data block B is fully covered, we have:

[0028] M≥n#

[0029] If x=B[-1], then:

[0030] Freq(x B )=1#

[0031] The above formula is the conclusion of the spatial covering idea.

[0032] Optionally, the algorithm for performing the adjuster based on spatial coverage and for optical signal modulation / demodulation is specifically:

[0033] The additional constellation points are used to uniquely represent all symbols using the spatial coverage concept: during modulation and demodulation, it is the last symbol that appears for the first time and is unique.

[0034] Optionally, in the method, the process at the sending end is:

[0035] (1) Load data and constellation diagram;

[0036] (2) Perform PS symbol division on the data;

[0037] (3) Perform statistics and symbol replacement on the data according to the adjuster;

[0038] (4) Perform RS / LDPC encoding;

[0039] (5) constellation mapping to generate modulation data;

[0040] (6) sending modulated data to the optical fiber channel;

[0041] The processing at the receiving end and the processing at the sending end are inverse processes;

[0042] The processing flow at the receiving end is as follows:

[0043] (1) Receive data from the fiber channel;

[0044] (2) Demodulation;

[0045] (3) RS / LDPC decoding;

[0046] (4) counting and restoring the replaced symbols according to the adjuster;

[0047] (5) Restore the data format according to PS rules;

[0048] (6) Count and calculate the bit error rate.

[0049] Optionally, when the signal quality reaches a certain threshold, RS or LDPC error correction codes are introduced to achieve error-free transmission.

[0050] The beneficial effects of the present invention are: First, this scheme organically integrates the existing PS and GS technologies, so that each of them achieves the best effect. Among them, the PS part uses the form of tree code to generate 7 3-bit symbols and 8 6-bit symbols, and the GS part finds the best 16QAM constellation coordinates through AENN training. Secondly, based on the integration of PS and GS technologies, this scheme proposes the technology of the adjuster for the first time, which utilizes the "uniqueness of the last symbol when it first appears". During modulation and demodulation, it can further reduce the average power of the system and achieve the purpose of constellation shaping. Finally, in order to ensure error-free transmission, this scheme adopts the forward error correction technology (FEC) of RS / LDPC to effectively reduce the system bit error rate to 0. The effect of the constellation modulation of this scheme can be seen in Figure 8 The experimental configuration of this scheme is shown in Figure 9 In terms of experimental results, compared with other 16QAM optical fiber communication solutions, the present invention has better shaping effect, see Figure 11; The present invention demonstrates better performance in terms of reducing transmission power, increasing channel capacity, and reducing bit error rate.

[0051] Other advantages, objects, and features of the present invention will be described in part in the following description and, in part, will be apparent to those skilled in the art upon examination of the following description or may be learned from practice of the present invention. The objects and other advantages of the present invention may be realized and obtained through the following description. BRIEF DESCRIPTION OF THE DRAWINGS

[0052] In order to make the purpose, technical solutions and advantages of the present invention more clear, the present invention will be described in detail below with reference to the accompanying drawings, in which:

[0053] Figure 1 This is an example diagram of the tree code (binary tree) used in the PS part and all the symbols it forms;

[0054] Figure 2 This is the neural network autoencoder model diagram used for training in the GS part;

[0055] Figure 3 For the GS part, when SNR=15, the initial coordinate diagram of the 16-QAM constellation;

[0056] Figure 4 This is the coordinate diagram of the 16-QAM constellation obtained after training with AENN.

[0057] Figure 5 This is the constellation diagram after GS training;

[0058] Figure 6 This is the flow chart of the sending end;

[0059] Figure 7 This is the flow chart of the receiving end;

[0060] Figure 8 This is the effect diagram after constellation modulation;

[0061] Figure 9 This is a diagram of the experimental configuration of the present invention;

[0062] Figure 10 Specific parameters of SSMF (standard single-mode fiber);

[0063] Figure 11 This is a comparison chart of the effects of the present invention and other 16QAM optical fiber communication solutions. DETAILED DESCRIPTION

[0064] The following describes the embodiments of the present invention by means of specific examples, and those skilled in the art can easily understand other advantages and effects of the present invention from the contents disclosed in this specification. The present invention can also be implemented or applied through other different specific embodiments, and the details in this specification can also be modified or changed in various ways based on different viewpoints and applications without departing from the spirit of the present invention. It should be noted that the illustrations provided in the following embodiments are only schematic illustrations of the basic concept of the present invention, and the following embodiments and features in the embodiments can be combined with each other without conflict.

[0065] Among them, the accompanying drawings are only for illustrative purposes and represent only schematic diagrams rather than actual pictures, and should not be understood as limiting the present invention. In order to better illustrate the embodiments of the present invention, some parts of the accompanying drawings may be omitted, enlarged or reduced, and do not represent the dimensions of actual products. For those skilled in the art, it is understandable that some well-known structures and their descriptions may be omitted in the accompanying drawings.

[0066] The same or similar numbers in the drawings of the embodiments of the present invention correspond to the same or similar parts; in the description of the present invention, it should be understood that if there are terms such as "upper", "lower", "left", "right", "front", "back", etc. indicating directions or positional relationships, they are based on the directions or positional relationships shown in the drawings. They are only for the convenience of describing the present invention and simplifying the description, and do not indicate or imply that the device or element referred to must have a specific direction, be constructed and operate in a specific direction. Therefore, the terms describing the positional relationship in the drawings are only used for illustrative purposes and cannot be understood as limiting the present invention. For ordinary technicians in this field, the specific meanings of the above terms can be understood according to specific circumstances.

[0067] See also Figures 1 to 11Based on the uniformly distributed 0-1 bit data type, the main technologies adopted in the present invention include geometric shaping (GS), probabilistic shaping (PS), adjuster, FEC coding, etc. The GS part sets the goals of maximizing GMI and minimizing BER, and uses autoencoder neural network training to obtain the optimal constellation coordinates (16-QAM) under a specific SNR. The PS part generates a total of 15 modulation symbols according to the tree code. Among them, the 8 modulation symbols with a 111 prefix (6 bits / Symbol) account for 1 / 8, and they correspond to the outer circle of the constellation points; the remaining modulation symbols are 7 types (3 bits / Symbol), accounting for 7 / 8, and they correspond to the inner circle of the constellation points. 15 symbols correspond to 16-QAM, and there will be one constellation point remaining. The essence of the adjuster is the idea of full coverage of the symbol space during modulation: for the 8 modulation symbols with a 111 prefix, statistics are taken and the last symbol that appears for the first time is marked; during modulation, it is mapped to an idle constellation point. If FEC coding is not used, a high bit error phenomenon will occur. Therefore, the RS (15, 11) code and the LDPC code of the CCSDS standard are introduced, both of which can achieve zero bit error. The specific implementation details of the main technology of the present invention are described as follows:

[0068] PS part:

[0069] In this part, the probability is well explained in the form of tree code (binary tree), such as Figure 1 As shown. This scheme assumes that the data follows a uniform distribution, that is, the probability of '0' and '1' appearing is 1 / 2. Starting from the root node, the probability of the data appearing '0' (to the left) and '1' (to the right) is 1 / 2. As shown in the figure, a total of 15 modulation symbols (blue circles) are generated. The leaf nodes (blue circles) are the data units (symbols) that need to be modulated. Among them, the probability of the symbols '000'-'110' appearing is 1 / 8, and the total probability of the 6-bit symbol with the prefix '111' appearing is 1 / 8.

[0070] GS section:

[0071] Fiber channels are modeled using a nonlinear model, where channel impairments depend solely on the amplified spontaneous emission (ASE), the average channel power P, and the fourth- and sixth-order moments (κ and κ3) of the constellation. In this scheme, the main channel models considered are the Gaussian noise (GN) model and the nonlinear interference noise (NLIN) model.

[0072] This solution combines a channel model, a nonlinear interference noise (NLIN) model, or a Gaussian noise (GN) model with gradient-based optimization from machine learning. For the GN model, the advantage of this combination is that it eliminates the dependence on κ and κ3 and the nonlinear effects of modulation. A trainable autoencoder (AE) model is constructed using a neural network consisting of a channel model, an encoder, and a decoder. Through neural network (NN) operations on real vectors, the symbols x and y are converted into vectors of real and imaginary parts. The AENN model is defined by the following equations:

[0073]

[0074]

[0075]

[0076]

[0077] In the above formula, f(·) represents the encoder NN and g(·) represents the decoder NN. The goal is to achieve the following through the latent variables: (and its damaged version ) in the output Re-enter The goal is achieved by minimizing the loss function, which eventually tends to Let x→ represent an N-dimensional vector, and we get an N-dimensional constellation. We train the autoencoder to implement an M-order constellation using a one-hot encoding vector.

[0078]

[0079] in is a vector of all zeros except the i-th row, so |S| = M. The training of this model is exactly the same as that of the traditional autoencoder. The AENN model used for training is as follows: Figure 2 As shown in Figure 2. The objective function is to maximize the combination of Gaussian GMI, Sigmoid GMI, and BER. The training parameter batchSize = [48, 80, 112, 144, 176, 208, 240], and the configurable channel parameter is SNR. Taking SNR = 15 as an example, the initial coordinates of the 16-QAM constellation are as follows: Figure 3 After AENN training, the constellation coordinates are finally obtained, as shown in Figure 4 shown.

[0080] Regulator part:

[0081] After GS training, the constellation diagram has 8 points in the inner circle and 8 points in the outer circle, such as Figure 7As shown. The number of symbols is 15, the number of constellation points is 16, and one constellation point is idle. If the data (symbol stream) is counted, only the first appearance of the symbol is considered. When the 15th (last) symbol appears for the first time, it is unique: this uniqueness holds true during both modulation and demodulation. Therefore, the last symbol that appears for the first time can be modulated on an idle constellation point. In this scheme, the points in the inner circle represent 0 to 6 (code words are: 000-110), and the points in the outer circle represent 7 to 15 (code words are: 111000–111111). In this way, only 7 of the 8 points in the inner circle are used, and there is one remaining point. This remaining constellation point is used to spatially cover the 15 symbols: during modulation and demodulation, it can be interpreted as the last symbol that appears, and this interpretation is unique. The final effect is as follows Figure 8 In essence, the adjuster also achieves the goal of constellation shaping.

[0082] The experimental setup of this scheme is as follows Figure 9 As shown in the figure, the transmission distance of standard single-mode fiber (SSMF) is set to 100 km. The specific parameters of SSMF are as follows: Figure 10 As shown in Figure 2. First, a binary pseudo-random number generator is used to generate a random bit sequence. Following the tree diagram rules for the 16-QAM codewords described in Section 2.1, 12,800 random sequences with 16 different symbols are generated. Following the rules of the proposed adjuster, the eight modulation symbols prefixed with '111' are counted, the last symbol that appears for the first time is marked, and mapped to an idle constellation point. This results in a probabilistically shaped data source. The data source is then processed using the CCSDS-standard LDPC code, with a code rate set to 1 / 2. At the transmitting end of the optical line, a 1550nm laser with a power of 0dBm and a linewidth of 100kHz serves as the optical signal source. The optical signal source is split into two orthogonal polarization-state lightwaves by an optical beam splitter, which are then modulated using an I / Q modulator. The I / Q modulator is a cascade of two Mach-Zehnder modulators and a π / 2 phase modulator. The Vpp of the two amplified electrical signals in the Mach-Zehnder modulator is 0.4V, with a half-voltage, insertion loss, and extinction ratio of 3V, 5dB, and 30dB, respectively. The two orthogonal light waves are I / Q modulated and then combined by an optical combiner and transmitted into a G.652D standard single-mode fiber (SSMF). The SSMF has an attenuation of α = 0.2dB / km, a dispersion of D = 16.5ps / (nm·km), and a nonlinear coefficient of 1.3 (W·km). -1After being amplified by an EDFA and transmitted over a 100km SSMF, the optical signal passes through a variable optical attenuator (VOA) to simulate transmission and splitter losses. At the receiving end of the optical line, the laser LO generates a local oscillator (LO) with a power of 0dBm and a linewidth of 100kHz. This LO is then mixed with the signal light into a 90° mixer. The output signal is converted to a digital signal after passing through a photodetector and a digital-to-analog converter. The resulting digital signal is then processed through an appropriate dispersion compensation algorithm to obtain the original data.

[0083] At the Ts end (sending end), the data sending process is described as follows:

[0084] 1. Load the data and constellation diagram.

[0085] 2. Perform PS symbol division on the data.

[0086] 3. According to the adjuster, count and replace the symbols of the data.

[0087] 4. Perform RS / LDPC encoding.

[0088] 5. Constellation mapping to generate modulation data.

[0089] 6. Send the modulated data to the fiber channel.

[0090] The detailed flow chart of Ts side is as follows Figure 6 Show.

[0091] The processing process at the Rs end (receiving end) is almost the opposite of that at the sending end. The data receiving process is described as follows:

[0092] 1. Receive data from Fibre Channel.

[0093] 2. Demodulation (constellation point demapping).

[0094] 3. RS / LDPC decoding.

[0095] 4. According to the adjuster, count and restore the replaced symbols.

[0096] 5. Restore the data format according to PS rules.

[0097] 6. Statistics and calculation of bit error rate, etc.

[0098] The detailed flow chart of Rs side is as follows Figure 7 shown.

[0099] The HPGS-16QAM solution, the traditional PS-16QAM solution, the GS-16QAM solution, and the uniform solution are described as follows:

[0100] As OSNR increases, the BER values of the HPGS-16QAM scheme, the traditional PS-16QAM scheme, the traditional GS-16QAM scheme, and the traditional uniform scheme gradually decrease.

[0101] As OSNR increases, the HPGS-16QAM solution has a lower BER compared with the traditional PS-16QAM solution, the traditional GS-16QAM solution, and the traditional uniform-16QAM solution.

[0102] like Figure 11 As shown in the figure, the HPGS-16QAM scheme saves 0.7dB, 0.8dB and 1.2dB of signal-to-noise ratio (OSNR) compared with the traditional PS-16QAM scheme, the traditional GS-16QAM scheme and the traditional uniform-16QAM scheme, respectively.

[0103] A detailed comparison of the HPGS-16QAM solution with other traditional solutions, such as Figure 11 shown.

[0104] Finally, it should be noted that the above embodiments are only used to illustrate the technical solutions of the present invention and are not limiting. Although the present invention has been described in detail with reference to the preferred embodiments, those skilled in the art should understand that the technical solutions of the present invention can be modified or replaced by equivalents without departing from the purpose and scope of the technical solutions, which should all be included in the scope of the claims of the present invention.

Claims

1. A HPGS16QAM communication optimization method based on spatial coverage, characterized by: The method comprises the following steps: Probabilistic shaping of 15QAM partitioning; Geometric shaping based on Gaussian noise GN and nonlinear interference noise NLIN channel models, trained by autoencoder neural network AENN; Performing algorithms based on spatial coverage and regulators for optical signal modulation / demodulation; The spatial coverage is specifically: The symbol space S is defined as the set of all possible symbols that may appear in the modulated data D; let <h2 style=";text-align:left;direction:ltr">S = {s1,s2,s3,…,s<h2 style=";text-align:left;direction:ltr"> n <h2 style=";text-align:left;direction:ltr">} Card(S) = n, that is, there are n symbols in the modulated data D. The modulation process is regarded as the process of obtaining symbols one by one from a certain moment. A shortest data block B is obtained in D to satisfy: Among them, Freq(x B ) represents the frequency of occurrence of symbol x in data block B; the shortest data block B that satisfies the above formula completes the full coverage of the symbol space S; Let M be the length of data block B, and B[-1] be the last symbol of B. When data block B is fully covered, we have: M≥n If x=B[-1], then: Freq(x B )=1 This is the conclusion of the idea of spatial coverage; The algorithm for performing the adjuster based on spatial coverage and for optical signal modulation / demodulation is specifically: The additional constellation points are used to uniquely represent all symbols using the spatial covering concept: during modulation and demodulation, the symbol is the last symbol that appears for the first time in the current data block and is unique.

2. The HPGS16QAM communication optimization method based on spatial coverage according to claim 1, characterized in that: The probability shaping of the 15QAM division is: Assuming the probability of '0' and '1' appearing is equal, the data stream generates a total of 15 modulation symbols according to the tree code. Among them, the eight modulation symbols with the 111 prefix account for 1 / 8, and they are 111000-111111. The remaining seven modulation symbols account for 7 / 8, and they are 000-110.

3. The HPGS16QAM communication optimization method based on spatial coverage according to claim 1, characterized in that: In the channel model based on Gaussian noise GN and nonlinear interference noise NLIN, the discrete NLIN model is described as follows: Where y represents the received codeword, x represents the transmitted codeword, c(·) represents the channel model, and n ASE Represents variance Gaussian noise samples, n NLIN Represents variance Gaussian noise samples; The NLIN model describes the effective signal-to-noise ratio of a fiber optic system as follows: Assume that all channels have the same optical transmission power, P tx represents the optical transmission power and draws energy from the same constellation; represents the variance of the accumulated amplified spontaneous emission (ASE) noise, represents the variance of NLIN; μ4 represents the fourth-order moment of the constellation and draws energy from the same constellation, and μ6 represents the sixth-order moment of the constellation.

4. The HPGS16QAM communication optimization method based on spatial coverage according to claim 1, characterized in that: In the method, the process at the sending end is: (1) Load data and constellation diagram; (2) Perform PS symbol division on the data; (3) Perform statistics and symbol replacement on the data according to the adjuster; (4) Perform RS / LDPC encoding; (5) constellation mapping to generate modulation data; (6) sending modulated data to the optical fiber channel; The processing at the receiving end and the processing at the sending end are inverse processes; The processing flow at the receiving end is as follows: (1) Receive data from the fiber channel; (2) Demodulation; (3) RS / LDPC decoding; (4) counting and restoring the replaced symbols according to the adjuster; (5) Restore the data format according to PS rules; (6) Count and calculate the bit error rate.

5. The HPGS16QAM communication optimization method based on spatial coverage according to claim 1, characterized in that: The method further comprises: introducing RS or LDPC error correction codes to achieve error-free transmission.

Images