A method, apparatus, device and medium for compensating for impairments of a coherent optical transmission signal

Abstract

The application discloses a kind of coherent light transmission signal damage compensation method, device, equipment and medium.The method includes obtaining the original damage digital signal of coherent receiver;Original damage digital signal is input to signal joint compensation model, and the online adaptive damage compensation processing of chromatic dispersion, polarization mode dispersion, optical nonlinearity and phase noise is carried out in parallel, and the digital signal after joint compensation is obtained;The quality of digital signal after joint compensation is evaluated by end-to-end loss function, and the quality evaluation result is obtained;If quality evaluation result does not satisfy preset threshold, then the internal parameters of signal joint compensation model are updated by the multi-module joint adaptive mechanism driven by meta-learning controller, and continue to carry out signal compensation processing by parallel execution in signal joint compensation model, until quality evaluation result satisfies preset threshold, and output target digital signal.The technical scheme, realize online real-time adaptive compensation, improve the transmission quality and compensation processing efficiency of signal.

Description

Technical Field

[0001] This invention relates to the field of signal processing technology, and in particular to a method, apparatus, device, and medium for compensating for impairments in coherent optical transmission signals. Background Technology

[0002] Coherent optical communication technology significantly improves the spectral efficiency of optical transmission systems by simultaneously modulating information onto the amplitude, phase, and polarization state of an optical carrier. It is a core supporting technology for building ultra-high-speed, ultra-large-capacity optical fiber transmission networks.

[0003] With the explosive growth of transmission demands from 5G, cloud computing, data center interconnection, and the future 6G, single-channel transmission rates have evolved from the early 10Gb / s to 400Gb / s, 1Tb / s, and even higher. To achieve these high-speed targets, high-order quadrature amplitude modulation (QAM) formats, including 16QAM, 64QAM, 256QAM, and even 1024QAM, are widely used in engineering practice. By encoding more bits in a single symbol slot, the spectral efficiency (SE) is significantly improved to over 4 bits / s / Hz. However, high-order QAM signals have extremely high constellation point density and extremely small Euclidean distances, making them extremely sensitive to various physical layer impairments in the transmission link. Taking 256QAM as an example, the normalized Euclidean distance between adjacent constellation points is only 1 / 8 of that in QPSK. Any residual impairment will lead to severe constellation point diffusion, a sharp deterioration in the bit error rate (BER), and the system performance will fail to meet the forward error correction (FEC) threshold requirements.

[0004] Existing high-speed optical transmission DSP compensation schemes mostly employ fixed parameters and static compensation, failing to adaptively respond to dynamic changes in the link. Parameters such as existing linear equalizer coefficients and BPS window lengths rely on offline design, unable to detect channel parameter changes caused by transmission path switching, temperature drift, and laser aging, lacking online adaptive capabilities and failing to meet the development requirements of software-defined and flexible optical networks. Furthermore, each impairment module is compensated independently, neglecting the coupling effects between impairments. Therefore, traditional step-by-step compensation schemes cannot achieve joint adaptive suppression of multiple types of transmission impairments, and their compensation accuracy and dynamic tracking capabilities are insufficient, making it difficult to meet the high-performance and high-reliability transmission requirements of ultra-high-speed, high-order QAM transmission systems. Summary of the Invention

[0005] This invention provides a method, apparatus, device, and medium for compensating for impairments in coherent optical transmission signals. It overcomes the shortcomings of existing technologies, such as the inability of fixed parameters to adapt and the loss of global performance due to independent compensation of each impairment module. It achieves online real-time joint adaptive compensation for chromatic dispersion, polarization mode dispersion, optical nonlinearity, and phase noise, thereby improving the signal transmission quality and compensation processing efficiency.

[0006] According to one aspect of the present invention, a method for compensating for impairments in coherent optical transmission signals is provided, comprising: Acquire the raw, impaired digital signal from the coherent receiver; The original damaged digital signal is input into the signal joint compensation model. The multiple compensation modules contained in the signal joint compensation model perform online adaptive damage compensation processing for chromatic dispersion, polarization mode dispersion, optical nonlinearity and phase noise in parallel to obtain the jointly compensated digital signal. The signal joint compensation model has a built-in meta-learning controller. The quality of the jointly compensated digital signal is evaluated using an end-to-end loss function to obtain the quality evaluation result. If the quality assessment result does not meet the preset threshold, the internal parameters of the signal joint compensation model are updated through the multi-module joint adaptive mechanism driven by the meta-learning controller, and the signal compensation processing continues to be performed in parallel by multiple compensation modules included in the signal joint compensation model until the quality assessment result meets the preset threshold and the target digital signal is output.

[0007] According to another aspect of the present invention, a coherent optical transmission signal impairment compensation device is provided, comprising: The signal acquisition module is used to acquire the raw impaired digital signal from the coherent receiver; The joint compensation module is used to input the original damaged digital signal into the signal joint compensation model, and to perform online adaptive damage compensation processing for chromatic dispersion, polarization mode dispersion, optical nonlinearity and phase noise in parallel through multiple compensation modules included in the signal joint compensation model, so as to obtain the jointly compensated digital signal; wherein, the signal joint compensation model has a built-in meta-learning controller; The quality assessment module is used to assess the quality of the jointly compensated digital signal using the end-to-end loss function, and obtain the quality assessment result. The target digital signal determination module is used to update the internal parameters of the signal joint compensation model through a multi-module joint adaptive mechanism driven by the meta-learning controller if the quality assessment result does not meet the preset threshold, and continue to perform signal compensation processing in parallel through multiple compensation modules included in the signal joint compensation model until the quality assessment result meets the preset threshold, and output the target digital signal.

[0008] According to another aspect of the present invention, an electronic device is provided, the electronic device comprising: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, which enables the at least one processor to perform the coherent optical transmission signal impairment compensation method according to any embodiment of the present invention.

[0009] According to another aspect of the present invention, a computer-readable storage medium is provided, the computer-readable storage medium storing computer instructions for causing a processor to execute and implement the coherent optical transmission signal impairment compensation method according to any embodiment of the present invention.

[0010] The technical solution of this invention involves acquiring the original damaged digital signal from a coherent receiver; inputting the original damaged digital signal into a signal joint compensation model; and performing online adaptive damage compensation processing for chromatic dispersion, polarization mode dispersion, optical nonlinearity, and phase noise in parallel through multiple compensation modules included in the signal joint compensation model to obtain a jointly compensated digital signal. The signal joint compensation model incorporates a meta-learning controller. A quality assessment is performed on the jointly compensated digital signal using an end-to-end loss function to obtain a quality assessment result. If the quality assessment result does not meet a preset threshold, the internal parameters of the signal joint compensation model are updated through a multi-module joint adaptive mechanism driven by the meta-learning controller, and signal compensation processing continues to be performed in parallel through the multiple compensation modules included in the signal joint compensation model until the quality assessment result meets the preset threshold, at which point the target digital signal is output. This technical solution overcomes the shortcomings of existing technologies, such as the inability of fixed parameters to adapt and the loss of global performance due to independent compensation of each damage module. It achieves online real-time joint adaptive compensation for chromatic dispersion, polarization mode dispersion, optical nonlinearity and phase noise. It overcomes the shortcomings of existing optical transmission signal damage compensation technologies, such as the inability of fixed parameters to adapt, the loss of global performance due to independent compensation of each damage module, poor convergence of high-order modulation formats and excessive computational complexity, thereby improving the signal transmission quality and compensation processing efficiency.

[0011] It should be understood that the description in this section is not intended to identify key or essential features of the embodiments of the present invention, nor is it intended to limit the scope of the invention. Other features of the invention will become readily apparent from the following description. Attached Figure Description

[0012] To more clearly illustrate the technical solutions in the embodiments of the present invention, the accompanying drawings used in the description of the embodiments will be briefly introduced below. Obviously, the accompanying drawings described below are only some embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on these drawings without creative effort.

[0013] Figure 1 This is a flowchart of a coherent optical transmission signal impairment compensation method provided in Embodiment 1 of the present invention; Figure 2 This is a flowchart of a coherent optical transmission signal impairment compensation method according to Embodiment 2 of the present invention; Figure 3 This is a schematic diagram of the overall architecture of the ultra-high-speed coherent optical transmission system provided in Embodiment 2 of the present invention; Figure 4 This is a schematic diagram of the structure of the multi-granularity phase noise compensation module provided in Embodiment 2 of the present invention; Figure 5 This is a schematic diagram of a coherent optical transmission signal impairment compensation device according to Embodiment 3 of the present invention; Figure 6 This is a schematic diagram of the structure of an electronic device provided according to Embodiment 4 of the present invention. Detailed Implementation

[0014] To enable those skilled in the art to better understand the present invention, the technical solutions of the present invention will be clearly and completely described below with reference to the accompanying drawings of the embodiments of the present invention. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort should fall within the scope of protection of the present invention.

[0015] It should be noted that the terms "original," "target," etc., used in the specification, claims, and accompanying drawings of this invention are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such data can be interchanged where appropriate so that the embodiments of the invention described herein can be implemented in orders other than those illustrated or described herein. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion; for example, a process, method, system, product, or apparatus that comprises a series of steps or units is not necessarily limited to those steps or units explicitly listed, but may include other steps or units not explicitly listed or inherent to such processes, methods, products, or apparatus.

[0016] Example 1 Figure 1This is a flowchart of a coherent optical transmission signal impairment compensation method according to Embodiment 1 of the present invention. This embodiment is applicable to situations where loss compensation is performed on optical signals transmitted by a coherent receiver. The method can be executed by a coherent optical transmission signal impairment compensation device, which can be implemented in hardware and / or software. This coherent optical transmission signal impairment compensation device can be configured in an electronic device with data processing capabilities. Figure 1 As shown, the method includes: S110: Acquire the original damaged digital signal from the coherent receiver.

[0017] The coherent receiver can be the receiving hardware device of a coherent optical communication system, used to coherently demodulate the received polarization-multiplexed modulated optical signal. The original impaired digital signal can refer to the baseband dual-polarization digital signal obtained after the coherent receiver completes photoelectric conversion and analog-to-digital sampling, including all physical layer impairments such as dispersion, polarization distortion, nonlinear distortion, and phase noise introduced by the link transmission.

[0018] In this embodiment, the received polarization-multiplexed optical signal is mixed, photodetected, and sampled and quantized by a high-speed analog-to-digital converter (ADC) through a coherent optical receiving link. The output is a time-domain original baseband digital signal containing complete transmission impairments. This signal contains various transmission impairments introduced by the optical fiber link, such as dispersion, polarization distortion, nonlinear distortion, and phase noise, which is the original impaired digital signal.

[0019] S120. Input the original damage digital signal into the signal joint compensation model. Perform online adaptive damage compensation processing on chromatic dispersion, polarization mode dispersion, optical nonlinearity and phase noise in parallel through multiple compensation modules contained in the signal joint compensation model to obtain the jointly compensated digital signal.

[0020] The signal joint compensation model incorporates a meta-learning controller. In this embodiment, the meta-learning controller can be a built-in global parameter scheduling and adaptive update control unit, which can be used to achieve rapid generalization and dynamic parameter tuning of the global network. The signal joint compensation model can refer to the overall model framework of the constructed multi-level adaptive DSP compensation network. This signal joint compensation model framework includes multiple compensation modules, specifically a cascaded compensation architecture consisting of a differentiable parameterizable dispersion compensation module, a polarization demultiplexing and high- and low-order polarization mode dispersion joint compensation module, a lightweight nonlinear compensation module, and a multi-granularity phase noise compensation module.

[0021] In this embodiment, the original damaged digital signal can be input into a pre-trained signal joint compensation model. The model performs multi-dimensional damage correction step by step according to the processing sequence: First, adaptive chromatic dispersion equalization compensation is performed on the signal to eliminate residual chromatic dispersion damage in the link; then, dual polarization demultiplexing and high-order polarization mode dispersion joint compensation are performed to correct dynamic polarization distortion; subsequently, self-phase and cross-phase optical nonlinear damage compensation is performed through a lightweight neural network; finally, linear and nonlinear phase noise joint correction is performed through a multi-granularity phase tracking structure, and the digital signal after preliminary joint compensation is output.

[0022] S130. The quality of the jointly compensated digital signal is evaluated using an end-to-end loss function to obtain the quality evaluation result.

[0023] The end-to-end loss function can be a loss function that evaluates the overall quality of the digital signal output by the joint compensation. The quality assessment result can be a quantitative indicator used to characterize the current signal compensation performance.

[0024] In this embodiment, the jointly compensated digital signal output by the signal joint compensation model is acquired, and an end-to-end loss function is introduced to conduct an overall quality assessment of the output signal. This end-to-end loss function compares the constellation points of the jointly compensated digital signal with an ideal distortion-free reference constellation, statistically analyzes the offset error and dispersion of the constellation points, and calculates the loss value as the signal quality assessment result. In this embodiment, the quality assessment result comprehensively reflects the overall residual damage caused by chromatic dispersion, polarization mode dispersion, optical nonlinearity, and phase noise superposition, without relying on the local outputs of intermediate modules, directly reflecting the final signal quality. The output quality assessment result can be further fed back to the meta-learning controller for online adjustment of the internal parameters of each compensation module, achieving collaborative adaptive optimization of each compensation module.

[0025] S140. If the quality assessment result does not meet the preset threshold, the internal parameters of the signal joint compensation model are updated through the multi-module joint adaptive mechanism driven by the meta-learning controller, and the signal compensation processing is continued to be performed in parallel by multiple compensation modules contained in the signal joint compensation model until the quality assessment result meets the preset threshold and the target digital signal is output.

[0026] The preset threshold can be a pre-set signal quality threshold. Shared meta-parameters can refer to the general channel prior meta-parameters obtained through offline pre-training of the meta-learning controller. The multi-module joint adaptive mechanism refers to the global parameter update strategy executed by the meta-learning controller, which does not adjust a single compensation module individually, but rather coordinates multiple modules—dispersion equalization, polarization mode dispersion compensation, nonlinear impairment compensation, and phase noise compensation—to collaboratively update their internal parameters. Model internal parameters can refer to learnable parameters such as filter taps, network weights, iteration step size, and filter covariance of each stage of the compensation network. The target digital signal can be the baseband digital signal output after the signal quality meets the preset threshold requirements and various transmission impairments are fully compensated.

[0027] In this embodiment, the real-time quality assessment result is compared with a preset signal quality threshold by a meta-learning controller. If the real-time quality assessment result is less than or equal to the preset signal quality threshold, the compensated signal quality is considered substandard. At this point, the multi-module joint adaptive mechanism driven by the meta-learning controller is triggered. Based on the quality assessment result, the internal parameters of the signal joint compensation model are updated in reverse. The adjustable parameters of each module, including dispersion equalization, polarization state demultiplexing and polarization mode dispersion joint compensation, nonlinear damage compensation, and multi-granularity phase noise compensation, are adjusted collaboratively. After the parameter update is completed, the compensated signal is input back into the signal joint compensation model. Each compensation module then re-executes the adaptive damage compensation processing for chromatic dispersion, polarization mode dispersion, optical nonlinearity, and phase noise to obtain the jointly compensated digital signal again, and the quality assessment is performed again. The compensation and quality assessment, as well as the parameter update process, are continuously cyclical until the quality assessment result meets the preset threshold. At this point, the parameter iteration update stops, and the target digital signal that meets the quality conditions is output.

[0028] In this embodiment, if the signal quality assessment result is greater than the preset signal quality threshold, the compensated signal quality is considered to meet the standard, and the current jointly compensated digital signal can be directly output as the target digital signal. The technical solution of this embodiment involves acquiring the original damaged digital signal from the coherent receiver; inputting the original damaged digital signal into the signal joint compensation model; and performing online adaptive damage compensation processing for chromatic dispersion, polarization mode dispersion, optical nonlinearity, and phase noise in parallel through multiple compensation modules included in the signal joint compensation model to obtain the jointly compensated digital signal. The signal joint compensation model has a built-in meta-learning controller. The quality of the jointly compensated digital signal is assessed using an end-to-end loss function to obtain a quality assessment result. If the quality assessment result does not meet the preset threshold, the internal parameters of the signal joint compensation model are updated through a multi-module joint adaptive mechanism driven by the meta-learning controller, and the signal compensation processing continues to be performed in parallel through multiple compensation modules included in the signal joint compensation model until the quality assessment result meets the preset threshold, at which point the target digital signal is output. This technical solution overcomes the shortcomings of existing technologies, such as the inability of fixed parameters to adapt and the loss of global performance due to independent compensation of each damage module. It achieves online real-time joint adaptive compensation for chromatic dispersion, polarization mode dispersion, optical nonlinearity and phase noise, thereby improving the signal transmission quality and compensation processing efficiency.

[0029] Example 2 Figure 2This is a flowchart of a coherent optical transmission signal impairment compensation method according to Embodiment 2 of the present invention. This embodiment is based on the above embodiment and optimized. Specifically, the optimization is as follows: the original impaired digital signal is input into the signal joint compensation model, and multiple compensation modules included in the signal joint compensation model perform online adaptive impairment compensation processing for chromatic dispersion, polarization mode dispersion, optical nonlinearity, and phase noise in parallel to obtain the jointly compensated digital signal. This includes: inputting the original impaired digital signal into the signal joint compensation model, performing modulation format identification processing and online estimation of channel impairment parameters on the original impaired digital signal through the channel sensing and modulation identification module to obtain the modulation format and multiple impairment prior parameters, and inputting the original impaired digital signal, the modulation format, and multiple impairment prior parameters in parallel into the piecewise adaptive dispersion equalization module, the polarization state demultiplexing and polarization mode dispersion joint compensation module, the nonlinear impairment compensation module, and the multi-granularity phase noise compensation module, respectively; and performing online adaptive impairment compensation processing for chromatic dispersion, polarization mode dispersion, optical nonlinearity, and phase noise through the piecewise adaptive dispersion equalization module. The system performs chromatic dispersion compensation on the original damaged digital signal by combining the modulation format and multiple prior damage parameters, resulting in a dispersion-compensated digital signal. Then, it performs polarization mode dispersion compensation on the original damaged digital signal using a polarization state demultiplexing and polarization mode dispersion joint compensation module, resulting in a polarization mode dispersion-compensated digital signal. Next, it performs optical nonlinear damage compensation on the original damaged digital signal using a nonlinear damage compensation module, resulting in a nonlinearly compensated digital signal. Finally, it performs phase noise compensation on the original damaged digital signal using a multi-granularity phase noise compensation module, resulting in a phase noise-compensated digital signal. Finally, it performs signal fusion processing on the dispersion-compensated digital signal, the polarization mode dispersion-compensated digital signal, the nonlinearly compensated digital signal, and the phase noise-compensated digital signal, resulting in a jointly compensated digital signal. Figure 2 As shown, the method includes: S210. Acquire the original damaged digital signal from the coherent receiver.

[0030] In this embodiment, the signal joint compensation model includes a channel sensing and modulation identification module (channel sensing and AMR module), a segmented adaptive dispersion equalization module (SA-FDE dispersion equalization module), a polarization state demultiplexing and polarization mode dispersion joint compensation module (DU-PMDC polarization demultiplexing and PMD compensation module), a nonlinear impairment compensation module (NN-NLC nonlinear compensation module), and a multi-granularity phase noise compensation module (MJPNC phase noise compensation module); the signal joint compensation model has a built-in meta-learning controller.

[0031] The technical solution of this embodiment can be implemented by an ultra-high-speed coherent optical transmission system, for example. Figure 3This is a schematic diagram of the overall architecture of an ultra-high-speed coherent optical transmission system according to Embodiment 2 of the present invention. It includes a transmitter 11 (containing a light source, modulator, DAC, and encoder), an optical fiber transmission channel 12 (containing optical fiber and amplifier), a receiver 13, and the connection relationships and data flow of five functional modules (channel sensing and AMR module, SA-FDE, DU-PMDC, NN-NLC, and MJPNC). The meta-learning controller connects each module with dashed bidirectional arrows, indicating unified management of meta-parameters and format-aware adaptive scheduling. Specifically, the receiver 13 includes a coherent receiver 131; a channel sensing and AMR module 132; an SA-FDE dispersion equalization module 133; a DU-PMDC polarization demultiplexing and PMD compensation module 134; an NN-NLC nonlinear compensation module 135; an MJPNC phase noise compensation module 136; and a meta-learning controller 137. Transmitter 11 performs signal encoding, digital-to-analog conversion, and optical modulation to output an optical signal. The optical signal is transmitted via fiber optic transmission channel 12 to form a damaged optical signal. Receiver 13 performs photoelectric conversion and analog-to-digital sampling on the damaged optical signal using a coherent receiver 131, outputting the original damaged digital signal. The original damaged digital signal is then fed into a channel sensing and AMR identification module 132, which performs modulation format identification and channel impairment parameter estimation. The modulation format and prior channel impairment parameters output by the channel sensing and AMR module 132 are distributed in parallel through a parameter access point to the SA-FDE dispersion equalization module (segmented adaptive frequency domain dispersion equalization module). 133, DU-PMDC polarization demultiplexing (polarization state demultiplexing and polarization mode dispersion joint compensation module), 134, NN-NLC nonlinear compensation module (nonlinear damage compensation module), 135, and MJPNC phase noise compensation module (multi-granularity phase noise compensation module) 136. Each module independently and in parallel completes the corresponding type of transmission damage compensation processing and outputs multiple compensation sub-signals. The multiple compensation sub-signals are fused to obtain the jointly compensated digital signal. The meta-learning controller 137 interacts with each level of compensation module to realize meta-parameter sharing, modulation format scheduling, online incremental training, joint loss function calculation, and format adaptive weight adjustment, dynamically optimizing the internal parameters of each level of module.

[0032] S220. The original damaged digital signal is input into the signal joint compensation model. The channel sensing and modulation identification module performs modulation format identification processing and online estimation of channel damage parameters on the original damaged digital signal to obtain the modulation format and multiple damage prior parameters. The original damaged digital signal, the modulation format, and multiple damage prior parameters are input in parallel into the segmented adaptive dispersion equalization module, the polarization state demultiplexing and polarization mode dispersion joint compensation module, the nonlinear damage compensation module, and the multi-granularity phase noise compensation module.

[0033] The modulation format can refer to the QAM modulation scheme used in the current transmitted signal. In this embodiment, the modulation format can include 16QAM, 64QAM, 256QAM, 1024QAM, etc. The impairment prior parameters can refer to the core physical parameters of the channel obtained through online real-time estimation. In this embodiment, the impairment prior parameters can include multiple parameters. For example, in this embodiment, the impairment prior parameters can include estimated parameters such as residual dispersion, differential group delay, laser linewidth, and nonlinear phase noise power, which can provide an adaptive parameter tuning reference for the subsequent compensation module.

[0034] In this embodiment, the original damaged digital signal is input into the signal joint compensation model. Constellation feature analysis and channel statistical feature extraction are performed on the received original damaged digital signal. The high-order QAM modulation format of the current signal is automatically identified. At the same time, the linear and nonlinear damage characteristics of the optical fiber transmission link are fitted in real time through the online channel parameter estimation algorithm. Key channel prior parameters such as residual dispersion, differential group delay, laser linewidth, and nonlinear phase noise power are output. Then, the original damaged digital signal, modulation format, and multiple damage prior parameters can be input in parallel into the segmented adaptive dispersion equalization module, the polarization state demultiplexing and polarization mode dispersion joint compensation module, the nonlinear damage compensation module, and the multi-granularity phase noise compensation module.

[0035] In this embodiment, optionally, the original damaged digital signal is copied to generate multiple parallel sub-signals. These multiple parallel sub-signals, modulation formats, and multiple prior damage parameters are then input in parallel to a segmented adaptive dispersion equalization module, a polarization state demultiplexing and polarization mode dispersion joint compensation module, a nonlinear damage compensation module, and a multi-granularity phase noise compensation module for compensation processing. The segmented adaptive dispersion equalization module performs chromatic dispersion damage compensation processing on one parallel sub-signal to obtain a dispersion-compensated sub-signal. The polarization state demultiplexing and polarization mode dispersion joint compensation module performs polarization mode dispersion damage compensation processing on another parallel sub-signal to obtain a polarization-compensated sub-signal. The nonlinear damage compensation module performs optical nonlinear damage compensation processing on yet another parallel sub-signal to obtain a nonlinear compensation sub-signal. The multi-granularity phase noise compensation module performs phase noise damage compensation processing on yet another parallel sub-signal to obtain a phase noise-compensated sub-signal. Finally, the dispersion-compensated sub-signal, polarization-compensated sub-signal, nonlinear compensation signal, and phase noise-compensated sub-signal are fused to obtain a jointly compensated digital signal.

[0036] In this embodiment, optionally, multiple prior damage parameters include dispersion estimation parameters, differential group delay estimation parameters, nonlinear phase noise power estimation parameters, and laser linewidth estimation parameters. The original damaged digital signal is processed by a channel sensing and modulation identification module to identify the modulation format and perform online estimation of channel damage parameters, resulting in the modulation format and multiple prior damage parameters. This includes: extracting lightweight high-order cumulant features from the original damaged digital signal using the channel sensing and modulation identification module and identifying the current modulation format using a multilayer perceptron classifier to obtain the corresponding modulation format and symbol rate; and performing frame-by-frame online estimation of the original damaged digital signal based on the modulation format, symbol rate, and pilot-assisted maximum likelihood estimation to obtain dispersion estimation parameters, differential group delay estimation parameters, nonlinear phase noise power estimation parameters, and laser linewidth estimation parameters.

[0037] The modulation format can refer to the high-order orthogonal amplitude modulation scheme currently used by the signal, including 16QAM, 64QAM, 256QAM, 1024QAM, etc. The symbol rate can refer to the single-symbol transmission rate of the optical transmission system. Pilot-assisted maximum likelihood estimation can be a high-precision parameter estimation algorithm that uses known pilot symbols within a frame as a reference and employs the maximum likelihood criterion to optimally solve for the channel's latent impairment parameters. Dispersion estimation parameters can be physical parameters characterizing the chromatic dispersion intensity of the fiber link, used for adaptive configuration of compensation weights in the subsequent dispersion equalization module. Differential group delay estimation parameters (DGD) can be core parameters characterizing the polarization mode dispersion difference of the dual polarization principal axis transmission delay, providing a priori reference for initializing the number of polarization compensation network layers, etc. Nonlinear phase noise power estimation parameters can be indices characterizing the intensity of nonlinear phase distortion caused by fiber self-phase modulation and cross-phase modulation. Laser linewidth estimation parameters can be parameters characterizing the spectral broadening of the transmitter and local oscillator lasers, directly determining the fluctuation intensity of the system's random phase noise, and can be used for dynamic updating of filter parameters in the phase compensation module.

[0038] In this embodiment, lightweight high-order cumulant feature extraction can be achieved through low-dimensional high-order statistical feature extraction processing based on the signal amplitude and phase statistical characteristics, which can distinguish different high-order QAM modulation constellation distributions. The multilayer perceptron classifier can be a lightweight feedforward neural network classification model, which achieves high-precision identification of different QAM modulation formats through feature mapping relationships.

[0039] In this embodiment, a lightweight high-order cumulant statistical feature extraction is performed on the input original damaged digital signal through a channel sensing and modulation identification module to mine the inherent constellation statistical differences in different modulation formats. The extracted low-dimensional features are input into a pre-trained multilayer perceptron classifier. Through network feature mapping and classification decision, the high-order QAM modulation format of the current transmitted signal is automatically identified, and the real-time symbol rate of the system is calculated simultaneously, providing the format and timing prior conditions for subsequent adaptive estimation of channel damage parameters. Then, in this embodiment, the identified modulation format and symbol rate are used as constraints for the estimation algorithm. The pilot-assisted maximum likelihood estimation algorithm is used to perform frame-by-frame online parameter fitting and optimal solution processing on the original damaged digital signal to obtain the real-time link chromatic dispersion parameters, differential group delay parameters, nonlinear phase noise power parameters, and laser linewidth parameters, respectively.

[0040] In this embodiment, the channel sensing and modulation identification module can include three sub-modules: an automatic modulation format identification sub-module (AMR), an online estimation sub-module for optical signal-to-noise ratio and channel impairment parameters, and a meta-learning controller. In this embodiment, the original impaired digital signal, i.e., the digital signal sampled by the receiver's ADC, is sent to the channel state sensing module. The AMR sub-module uses lightweight feature extraction based on high-order cumulants (HOS) and a multilayer perceptron (MLP) classifier to identify modulation formats such as 16QAM / 64QAM / 256QAM within 32 symbols after receiving the preamble sequence, and broadcasts the format identifier F and symbol rate Rs to subsequent modules. The online estimation sub-module for optical signal-to-noise ratio and channel impairment parameters outputs dispersion estimate DL, differential group delay (DGD) estimate τ, and nonlinear phase noise power σ per frame (frame length N=1024 symbols) based on pilot-aided maximum likelihood estimation (Pilot-Aided MLE). 2 The estimation results of NL and laser linewidth Δf provide prior parameter initialization for subsequent compensation modules.

[0041] In this embodiment, the meta-learning controller can adopt the Model-Agnostic Meta-Learning (MAML) framework to maintain a set of cross-format shared meta-parameters θ0. When a format switch or link mutation event is detected, the parameters of each compensation module can be quickly converged to the new state with only a small number of gradient updates, starting from θ0, thus achieving rapid adaptation.

[0042] In this embodiment, the lightweight high-order cumulants can be used to extract low-dimensional distinguishing features, and a small multilayer perceptron can be used to complete modulation classification. This enables accurate identification of 16 / 64 / 256 / 1024QAM multi-high-order modulation with low DSP hardware resource consumption, synchronous output of symbol rate, and improved channel parameter estimation accuracy through pilot-assisted maximum likelihood estimation.

[0043] S230. The original damaged digital signal is subjected to chromatic dispersion damage compensation processing by the segmented adaptive dispersion equalization module, which combines the modulation format and multiple damage prior parameters to obtain the dispersion-compensated digital signal.

[0044] Chromatic dispersion refers to signal pulse broadening and inter-symbol interference (ISI) caused by inconsistent transmission rates of optical signals of different frequencies in an optical fiber. It is a fundamental linear impairment in optical transmission. In this embodiment, chromatic dispersion impairment compensation processing can compensate for the chromatic dispersion loss in the optical transmission of the signal. The dispersion-compensated digital signal can be the digital signal obtained after compensating for the chromatic dispersion loss in optical transmission.

[0045] In this embodiment, the compensation accuracy can be adapted according to the identified modulation format, and the segmented equalization coefficient and compensation range can be dynamically configured in combination with the estimated residual dispersion prior parameters. The chromatic dispersion pulse broadening distortion of the original digital signal can be globally corrected to eliminate inter-symbol interference introduced by optical fiber transmission, complete the basic linear dispersion impairment compensation, and output the dispersion-compensated digital signal.

[0046] In this embodiment, optionally, a segmented adaptive dispersion equalization module is used to perform chromatic dispersion damage compensation processing on the original damaged digital signal by combining the modulation format and multiple prior damage parameters to obtain a dispersion-compensated digital signal. This includes: calculating an initial frequency domain equalization filter based on dispersion estimation parameters; performing coarse compensation on the original damaged digital signal based on the initial frequency domain equalization filter to obtain a digital signal with residual dispersion; adaptively determining the filter length of the finite impulse response filter bank based on the modulation format; using an adaptive iterative algorithm to perform fine compensation optimization on the digital signal with residual dispersion; and outputting a dispersion-compensated digital signal.

[0047] The initial frequency domain equalization filter can be a fixed frequency domain compensation filter determined by the estimated dispersion parameters, which can be used to cancel the main chromatic dispersion of the link in one step. Residual dispersion refers to the small, dynamic, and minute dispersion deviation that remains after global fixed compensation and cannot be completely canceled by the fixed filter. In this embodiment, the digital signal of residual dispersion can be the digital signal of residual dispersion obtained after coarse compensation of the original damaged digital signal by the initial frequency domain equalization filter. The finite impulse response filter bank (FIR) can be a time-domain adaptive equalization filter unit with dynamic weight update capability, used for fine correction of small residual damage. The filter length can refer to the number of taps in the FIR filter, which can be used to determine the correction range and compensation accuracy of fine equalization. Modulation format adaptation can be dynamically matched to the filtering accuracy according to the current QAM modulation order. Higher-order QAMs have smaller constellation spacing and are more sensitive to residual dispersion, so a longer filter length is matched. For example, longer filter taps can be configured for higher-order systems such as 1024QAM and 256QAM to improve fine correction capability, while the length can be appropriately reduced for lower-order QAM systems to reduce computational overhead. Adaptive iterative algorithms can be equalization algorithms that calculate signal errors in real time and iteratively update filter weights, dynamically tracking time-varying residual dispersion. Fine compensation optimization can perform high-precision dynamic equalization for weak dynamic residual dispersion that cannot be eliminated by coarse compensation, completely eliminating intersymbol interference.

[0048] In this embodiment, based on the link dispersion estimation parameters obtained online, the coefficients of the frequency domain equalization filter matching the overall dispersion characteristics of the current link are calculated. This initial frequency domain filter is used to perform global, large-scale inverse dispersion transformation compensation on the original signal, which can offset the main chromatic dispersion impairment introduced by fiber transmission. After coarse compensation, large-amplitude dispersion distortion in the signal is completely eliminated, leaving only a small amount of dynamic dispersion deviation, resulting in a digital signal with residual dispersion. Then, the filter length of the finite impulse response filter bank can be adaptively configured according to the currently identified high-order QAM modulation format. For the coarsely compensated digital signal with residual dispersion, an adaptive iterative algorithm is used to statistically analyze the signal error in real time and dynamically update the filter weights iteratively, continuously offsetting dynamic residual dispersion and minor inter-symbol interference, completing fine dispersion compensation optimization, and finally outputting a dispersion-compensated digital signal with completely eliminated chromatic dispersion impairment.

[0049] For example, in this embodiment, chromatic dispersion compensation employs an improved segmented adaptive frequency domain equalization (SA-FDE) method. Unlike traditional fixed-coefficient FDE, the SA-FDE in this embodiment includes a coarse compensation stage and a fine compensation stage; wherein, the coarse compensation stage may calculate the initial frequency domain equalization filter based on the dispersion estimation parameters DL provided by the channel-aware module. The coarse-compensation stage involves performing most of the dispersion compensation, keeping the residual dispersion within ±200 ps / nm to obtain a digital signal of the residual dispersion. The fine-compensation stage can be achieved by deploying a trainable finite impulse response (FIR) filter bank on the digital signal of the coarse-compensated dispersion. The filter length L is adaptive to the QAM format (e.g., 16QAM: L=31; 64QAM: L=63; 256QAM: L=127). The filter coefficients are iteratively updated using the minimum mean square error (MMSE) criterion, with the step size automatically adjusted by an adaptive step size control (NLMS) algorithm based on normalized power.

[0050] Furthermore, in this embodiment, when the transmission path switching causes a step change in DL, SA-FDE does not need to be reset. Instead, it completes the reconvergence of coarse compensation coefficients within 256 symbols through cumulative gradient estimation, which improves the switching response speed by more than 20 times compared with the traditional fixed parameter scheme.

[0051] In this embodiment, the configuration allows for the rapid elimination of most fixed dispersion by utilizing prior dispersion parameters, thus avoiding the problem of slow iterative convergence. Furthermore, it eliminates dynamic residual dispersion by modulating adaptive FIR filtering, resulting in high compensation accuracy, fast convergence speed, and reduced computational overhead.

[0052] S240. The original damaged digital signal is processed by polarization mode dispersion damage compensation through polarization state demultiplexing and polarization mode dispersion joint compensation module to obtain the digital signal after polarization mode dispersion compensation.

[0053] Polarization mode dispersion (PMD) refers to the polarization signal distortion caused by the inconsistency in the transmission delay of the dual polarization principal axes of an optical fiber. Specifically, it can include constant first-order PMD and frequency-varying higher-order PMD. Polarization demultiplexing can completely separate the X / Y dual polarization signals that are coupled and confused during transmission, eliminating polarization crosstalk. The digital signal after polarization mode dispersion compensation can be the digital signal output after eliminating dynamic polarization rotation and frequency-varying polarization damage.

[0054] In this embodiment, for the original damaged digital signal, the network size and initialization parameters are adaptively configured by combining the differential group delay prior parameters output by the pre-channel estimation. First, the dual polarization signal demultiplexing and first-order polarization mode dispersion compensation are completed by the butterfly-shaped complex FIR depth expansion network. Then, the frequency-dependent higher-order polarization distortion is fitted and corrected by the second-order frequency-varying PMD compensation network, suppressing dynamic polarization rotation and frequency-varying polarization damage, and outputting a digital signal with fully compensated polarization damage.

[0055] In this embodiment, optionally, the polarization state demultiplexing and polarization mode dispersion joint compensation module includes a butterfly-shaped complex finite impulse response (FIR) expansion layer and a second-order polarization mode dispersion compensation layer. The original damaged digital signal is processed by the polarization state demultiplexing and polarization mode dispersion joint compensation module to obtain a polarization mode dispersion compensated digital signal. This includes: adaptively determining the number of network layers and filter taps of the butterfly-shaped FIR expansion layer based on the modulation format and differential group delay estimation parameters; iteratively updating the complex weight matrix and adaptive learning step size of each network layer using multi-loop decision pointing loss as the optimization criterion; completing polarization demultiplexing of the dual-polarization signal and first-order polarization mode dispersion compensation to obtain a compensated intermediate signal; and through the second-order polarization mode dispersion compensation layer, combining the differential group delay estimation parameters at each frequency point, extracting the frequency-varying link equivalent Jones matrix of the compensated intermediate signal and constructing a frequency differential model to correct the higher-order polarization mode dispersion distortion caused by the variation of differential group delay with optical frequency, thus obtaining the polarization mode dispersion compensated digital signal.

[0056] The butterfly-shaped complex finite impulse response (FIR) expansion layer can be a complex domain deep equalization network used for decoupling dual-polarization optical signals. Its symmetrical structure adapts to X / Y dual-polarization coupled signals, enabling simultaneous polarization crosstalk removal and temporal delay equalization. The number of network layers can be a hyperparameter of the butterfly-shaped complex finite impulse response expansion layer; generally, a larger number of layers and taps results in a stronger fitting ability for scenarios with large delays and strong polarization coupling. The multi-loop decision pointing loss is a composite loss function jointly constructed from constellation decision convergence accuracy, polarization orthogonality deviation, and signal error residuals. The complex weight matrix can be an updatable complex equalization parameter for each layer of the FIR network, used to correct polarization crosstalk and temporal distortion. The adaptive learning step size refers to the step size parameter for network iteration updates, dynamically adjusted according to the link impairment strength to ensure no oscillation for weak impairments and fast convergence for strong impairments. First-order polarization mode dispersion (PMD) can be the mainstream polarization delay distortion caused by a fixed differential group delay, and is the most significant polarization impairment in optical fibers. The compensated intermediate signal can refer to the transition signal obtained after completing polarization demultiplexing and first-order PMD correction.

[0057] In this embodiment, the second-order polarization mode dispersion (PMD) compensation layer can be a high-order polarization equalization unit designed for frequency-varying polarization damage, specifically addressing high-order PMD distortions that cannot be covered by first-order compensation. The differential group delay estimation parameters at each frequency point can refer to real-time DGD values ​​for different frequency components, reflecting the frequency-varying characteristics of DGD with optical frequency. The equivalent Jones matrix of the frequency-varying link can be a matrix used to fully characterize the time-frequency domain polarization transmission characteristics, including all polarization features such as polarization state rotation, frequency-varying delay, and polarization-related loss. The frequency differential model can be a mathematical model characterizing the nonlinear variation of the differential group delay with frequency, used to fit the high-order PMD distortion law. High-order polarization mode dispersion distortion refers to the frequency-varying, nonlinear polarization distortion caused by DGD variations with optical frequency, which is a residual performance bottleneck in high-speed, high-bandwidth systems.

[0058] In this embodiment, based on the constellation density and impairment tolerance of the current modulation format, combined with the real-time estimated differential group delay impairment intensity, the optimal number of network layers and filter taps of the butterfly-shaped complex FIR expansion layer are adaptively matched; a multi-ring decision pointing composite loss function is constructed as a global optimization criterion to calculate the constellation deviation and polarization orthogonality error of the output signal in real time; through an adaptive iterative mechanism, the network complex weight matrix and iterative learning step size are updated layer by layer to continuously remove crosstalk coupling between the X and Y polarization channels, accurately compensate for the first-order polarization mode dispersion impairment corresponding to the fixed delay, realize the complete demultiplexing of the dual polarization signal, and output a compensated intermediate signal that eliminates the first-order polarization distortion. Then, for the compensated intermediate signal after first-order compensation, the frequency-varying equivalent Jones matrix of the current link is extracted in real time by combining the differential group delay estimation parameters of each frequency point in the full frequency domain with the second-order PMD compensation network; a frequency differential model is constructed based on the frequency domain differential principle to accurately fit the nonlinear variation law of differential group delay with optical frequency shift, and the higher-order polarization mode dispersion distortion component is analyzed; the higher-order polarization damage is corrected by the reverse distortion cancellation method to eliminate the polarization waveform distortion and constellation spread caused by frequency-varying DGD in the broadband signal, and finally, first-order and higher-order full-dimensional PMD compensation are completed simultaneously, and a high-quality digital signal with fully compensated polarization mode dispersion is output.

[0059] For example, this embodiment proposes a Deep Unfolding (DU) framework based on a Deep Unfolding Iterative Polarization State Demultiplexing and Polarization Mode Dispersion Joint Compensation Module (DU-PMDC). This framework combines traditional butterfly FIR adaptive equalization with a neural network. The network structure can be an iterative update process of a T-level butterfly-shaped complex FIR (Ntap per level) unfolded into a T-layer neural network. Each layer contains a complex weight matrix Wt (corresponding to butterfly filter coefficients) and a learnable step size parameter ηt (corresponding to the adaptive learning rate). The entire network is trained online in an end-to-end manner, and the loss function adopts the Multi-Ring DD Loss for high-order QAM. The weight function w(·) is automatically selected based on the ring radius distribution of the QAM scheme, solving the mismatch between the CMA constant mode assumption and higher-order QAM. Furthermore, a second-order polarization mode dispersion compensation layer is added after the T-layer base network. This layer compensates for the higher-order PMD components of the differential group delay (DGD) that vary with frequency by modeling the frequency derivative of the Jones matrix.

[0060] In this embodiment, upon detecting a rapid PMD rotation event, the meta-learning controller is triggered to perform gradient updates. Using pre-trained meta-parameters θ0 as the initial point, the parameter updates are completed within 10 symbol periods, enabling rapid tracking of sudden PMD events. Furthermore, this embodiment can dynamically adjust the network depth T and tap count Ntap based on the available computing resources (register / multiplier utilization) of the current DSP chip. A high-precision mode is used when resources are abundant, while a low-power mode is degraded when resources are limited, avoiding global latency overflow caused by resource contention.

[0061] In this embodiment, the network size can be adaptively adjusted according to the actual DGD strength and modulation format of the link, balancing computing power and accuracy. Multi-ring loss constraints are used to avoid equilibrium divergence, while frequency domain differential modeling is used to accurately suppress broadband high-order polarization distortion, which greatly improves the polarization equilibrium stability of high-order QAM signals under dynamic links.

[0062] S250. The original digital damage signal is processed by optical nonlinear damage compensation through the nonlinear damage compensation module to obtain the nonlinearly compensated digital signal.

[0063] Optical nonlinear damage can refer to signal amplitude compression and phase distortion caused by self-phase modulation under the influence of a strong optical field in an optical fiber. The digital signal after nonlinear compensation can be the digital signal output by eliminating signal amplitude compression and phase distortion through a nonlinear damage compensation module.

[0064] In this embodiment, the original damaged digital signal can be segmented into fixed-length time sequences to construct multi-symbol time-series input features. These features are then input into a lightweight time-series convolutional residual network to predict the real-time nonlinear phase shift and amplitude distortion factor of the signal. Regularization correction of the prediction parameters is performed based on the physical constraints of fiber nonlinear transmission, and inverse nonlinear compensation is applied to the original damaged digital signal. Simultaneously, the network parameters can be fine-tuned online based on soft-decision pseudo-labels to track slow time-varying nonlinear characteristics such as optical power drift and amplifier gain shift in real time, outputting a digital signal after nonlinear damage suppression.

[0065] In this embodiment, optionally, the original damaged digital signal is subjected to optical nonlinear damage compensation processing by a nonlinear damage compensation module to obtain a nonlinearly compensated digital signal. This includes: segmenting the polarization mode dispersion compensated digital signal into segments of fixed length to obtain an equalized complex signal sequence; inputting the equalized complex signal sequence into the nonlinear damage compensation module to predict the corresponding nonlinear phase estimation result and amplitude compression factor; and, based on the physical constraints of fiber nonlinear transmission and combined with the nonlinear phase noise power estimation parameters, performing regularization correction on the nonlinear phase estimation result and amplitude compression factor to obtain the corrected nonlinear phase estimation result and amplitude compression factor, and then performing inverse compensation on the polarization mode dispersion compensated digital signal to obtain the nonlinearly compensated digital signal.

[0066] The fixed length can be a pre-set signal length. For example, the fixed length can be the equalized complex signal of three adjacent segments, or it can be set according to actual needs. In this embodiment, the continuous time-domain signal can be segmented into frames according to a preset uniform symbol frame number to construct a time-series input sample with a uniform length. The equalized complex signal sequence can be a regular time-series signal frame sequence composed of complex constellation points after dispersion and polarization mode dispersion linear impairment compensation. The nonlinear phase estimation result can refer to the signal phase distortion offset caused by fiber self-phase modulation and cross-phase modulation. The amplitude compression factor can be the distortion coefficient of signal constellation amplitude attenuation and compression caused by fiber nonlinear effects, used to characterize the degree of amplitude distortion. The physical constraint of fiber nonlinear transmission can be a physical law constraint obtained based on the fiber nonlinear Schrödinger transmission mechanism, used to limit the reasonable range of nonlinear distortion values ​​and avoid model prediction distortion. Regularization correction can be a process that combines physical priors and measured channel estimation parameters to constrain and correct the model prediction values, eliminating unreasonable and abnormal prediction values. Reverse compensation can be a process that uses distortion cancellation operations that are opposite to the nonlinear distortion of the optical fiber to perform inverse distortion correction on the original signal.

[0067] In this embodiment, the time-domain complex digital signal that has undergone linear impairment compensation can be segmented into consecutive frames. The signal is segmented according to a preset fixed symbol length, converting the continuous data stream into multiple sets of time-series complex signal frames of uniform length, forming a regularized, equalized complex signal sequence. This segmented and regularized complex signal sequence can then be fed into the nonlinear impairment compensation module. Relying on the module's built-in time-series feature learning capability, the nonlinear coupling distortion patterns between symbols are discovered, and the nonlinear phase offset and amplitude compression distortion degree corresponding to the current signal frame are predicted simultaneously, yielding the original nonlinear phase estimation result and amplitude compression factor.

[0068] In this embodiment, the inherent physical laws of optical fiber nonlinear transmission are introduced as constraints. The nonlinear phase noise power obtained by online estimation is used to characterize the nonlinear impairment intensity of the current link. The original phase estimation result and amplitude compression factor output by the module are corrected by regularization constraints to eliminate the noise abnormal components generated by network prediction. This makes the distortion parameters completely fit the real nonlinear transmission characteristics of optical fiber, resulting in a high-precision corrected nonlinear phase estimation result and amplitude compression factor. By using the corrected high-precision nonlinear phase estimation result and amplitude compression factor, an inverse nonlinear distortion cancellation operation is applied to the digital signal after linear compensation. This inversely cancels the phase distortion and amplitude compression impairment generated during optical fiber transmission, suppresses the signal distortion caused by self-phase modulation and wavelength division multiplexing cross-phase modulation, and finally outputs a nonlinear compensated digital signal that eliminates optical nonlinear impairment.

[0069] For example, this embodiment proposes a nonlinear damage prediction and compensation (NN-NLC) method based on lightweight neural networks, which achieves compensation performance close to that of DBP with a computational complexity far lower than DBP. The network architecture can be a temporal convolutional network (TCN) combined with residual connections. The input is an equalized complex signal sequence of three adjacent segments (length M is 64 symbols), and the output is the nonlinear phase rotation estimate φ of the current symbol. NL And the amplitude compression factor â. A perturbation solution of the nonlinear Schrödinger equation (NLSE) can be introduced as a regularization term during network training to constrain the network output to satisfy the first-order perturbation form φ of the Manakov equation. NL ≈γL eff ·P(t), where γ is the nonlinear coefficient, L eff The effective action length and P(t) represent the instantaneous power of the signal, which significantly reduces the network generalization error.

[0070] In this embodiment, the soft information after the decision can also be used as pseudo-labels to fine-tune the TCN parameters online with micro-batch stochastic gradient descent of 1–5 steps per frame, so that the network can continuously track the slow time-varying characteristics of the nonlinear channel (such as transmit power drift and optical amplifier gain shift), and achieve adaptive nonlinear compensation without offline retraining.

[0071] Furthermore, in this embodiment, for WDM systems, the cross-correlation term of pilot symbols from adjacent channels can be introduced as an auxiliary input feature. This allows NN-NLC to achieve an additional Q-factor gain of approximately 0.3 dB compared to the single-channel scheme in scenarios where XPM is the dominant impairment. By extracting the cross-correlation feature of pilot symbols from adjacent channels as an auxiliary input for NN-NLC, XPM sensing compensation can be achieved in WDM scenarios without adding inter-channel communication interfaces, demonstrating significant engineering practicality.

[0072] In this embodiment, the two-dimensional coordinated compensation of phase and amplitude can be achieved through this setting, which improves the diffusion of high-order QAM constellations and implements reverse compensation based on the corrected distortion parameters to accurately offset optical nonlinear damage.

[0073] S260. The original damaged digital signal is processed by the multi-granularity phase noise compensation module to obtain the phase noise compensated digital signal.

[0074] Phase noise can include random rapid phase jitter caused by laser linewidth, link frequency offset, and low-frequency phase drift, and it is strongly coupled with fiber nonlinear damage. The phase noise-compensated digital signal can refer to the baseband complex sub-signal output after phase noise distortion has been eliminated by a multi-granularity phase noise compensation module.

[0075] In this embodiment, a dual-granularity compensation mechanism at the symbol level and frame level is adopted. The symbol-level loop adapts its bandwidth according to the modulation format to compensate for high-speed random phase fluctuations. The frame-level Kalman filter dynamically updates the filter coefficients based on the laser linewidth prior parameters to correct low-frequency phase drift and carrier frequency offset. The nonlinear phase estimate predicted by the fusion network and the linear phase noise component output by the two-stage loop are fused. The linear and nonlinear phase joint correction is completed through a unified phase rotation operation, avoiding error propagation caused by multi-stage serial compensation. Finally, a high-quality digital signal after full-impairment joint compensation is output.

[0076] In this embodiment, optionally, the multi-granularity phase noise compensation module includes a symbol-level fast phase tracking loop and a frame-level slow phase trend estimator. The multi-granularity phase noise compensation module performs phase noise damage compensation processing on the original damaged digital signal to obtain a phase noise-compensated digital signal. This includes: adaptively configuring the loop bandwidth according to the modulation format using the symbol-level fast phase tracking loop to compensate for high-speed random phase fluctuations and output instantaneous high-frequency linear phase components; dynamically adjusting the Kalman filter parameters based on intra-frame pilot symbols and combined with laser linewidth estimation parameters output from channel sensing using the frame-level slow phase trend estimator to correct low-frequency phase noise and frequency shift and output slowly varying linear phase components; determining the linear phase noise estimation result based on the high-frequency linear phase component and the slowly varying linear phase component; fusing the nonlinear phase estimation result and the linear phase noise estimation result and performing joint phase correction to output the phase noise-compensated digital signal.

[0077] The symbol-level fast phase tracking loop is a real-time phase tracking unit updated at the single-symbol granularity, responsible for capturing high-frequency and rapid random phase fluctuations. The frame-level slow phase trend estimator is a filtering estimation unit updated at the data frame granularity, used to fit low-frequency phase drift and carrier frequency offset. In this embodiment, the multi-granularity phase noise compensation module adopts a fast-slow dual-granularity collaborative architecture. The symbol-level fast phase tracking loop is responsible for short-term rapid phase disturbance correction, while the frame-level slow phase trend estimator is responsible for long-term slow phase trend and frequency offset correction, achieving full coverage compensation for phase noise across the entire linewidth range.

[0078] The loop bandwidth can be the signal response bandwidth of the phase tracking loop. Understandably, a larger loop bandwidth results in stronger tracking capability against high-speed phase jitter. High-speed random phase fluctuations are symbol-by-symbol random phase jitter caused by the laser linewidth, belonging to high-frequency linear phase noise. The instantaneous high-frequency linear phase component can refer to the fast-changing linear phase offset obtained by real-time calculation of a single symbol. The intra-frame pilot symbol can refer to a known fixed reference symbol in the frame structure, used to stably estimate the slowly varying channel parameters of the link. Laser linewidth estimation parameters are physical parameters used to characterize the laser's spectral broadening and determine the low-frequency phase walk intensity. Kalman filter parameters can include filtering coefficients for process noise, observation noise, etc., used to balance estimation smoothness and tracking capability. Low-frequency phase noise and frequency offset can refer to the frame-level slowly varying phase trend and carrier frequency offset caused by temperature and device drift. The slowly varying linear phase component can be the linear phase offset corresponding to the low-frequency phase drift and frequency offset obtained by frame-level smoothing. The linear phase noise estimation result can be the total linear phase offset of the complete laser obtained by fusing the fast and slow phase components. Joint phase correction can be a correction method that unifies the phase rotation cancellation of linear phase noise and nonlinear phase distortion.

[0079] In this embodiment, the symbol-level fast phase tracking loop adaptively configures the optimal loop bandwidth according to the current high-order modulation format. Taking advantage of the smaller constellation spacing and greater sensitivity to phase jitter in high-order QAM, it adaptively increases the tracking bandwidth to accurately capture high-speed random phase fluctuations symbol by symbol, and outputs the high-frequency linear phase component corresponding to each symbol in real time. Furthermore, a frame-level slow phase trend estimator constructs an observation sequence using known pilot symbols within the frame, and dynamically adjusts the noise covariance parameter of the Kalman filter based on the estimated laser linewidth parameter. A larger linewidth enhances the dynamic tracking capability, accurately fitting and offsetting low-frequency phase drift and residual carrier frequency offset in the link, outputting stable frame-level slowly varying linear phase components. The high-frequency fast-changing phase components output at the symbol level are superimposed and fused with the low-frequency slowly varying phase components output at the frame level to obtain a complete linear phase noise estimation result. In this embodiment, the nonlinear phase distortion component introduced by fiber transmission and the linear phase noise component introduced by the laser can be weighted and fused to obtain the total overall phase impairment of the signal. A unified phase reverse rotation correction is performed on the baseband signal to eliminate linear and nonlinear coupled phase impairments in one step, ultimately outputting a high-quality digital signal after full-dimensional joint compensation.

[0080] For example, since phase noise compensation is often strongly coupled with nonlinear damage, this embodiment constructs a multi-granularity joint phase noise compensation (MJPNC) module. Independent tracking loops are set up at the symbol and frame levels. The symbol-level fast phase tracking loop employs a second-order phase-locked loop (PLL) structure, using the decision feedback phase error signal to drive a numerically controlled oscillator (NCO) to track the fast phase random walk caused by the laser linewidth. The loop bandwidth adapts to the current QAM format. The frame-level slow phase trend estimator runs a Kalman filter on the pilot symbols of each frame, jointly estimating the low-frequency components of the laser frequency shift (CFO) and phase noise (PSD). The compensated Kalman state vector includes both phase and frequency shifts, and the process noise covariance matrix Q is dynamically updated based on the real-time estimated laser linewidth Δf.

[0081] In this embodiment, the nonlinear phase estimate φ output by the NN-NLC can be... NL The linear phase noise φ estimated by MJPNC PN Joint addition achieves joint compensation through a unified rotation operation, avoiding the propagation of phase estimation errors introduced by secondary serial processing.

[0082] For example, Figure 4This is a schematic diagram of the structure of a multi-granularity phase noise compensation module (MJPNC) according to Embodiment 2 of the present invention. The MJPNC includes: a nonlinear compensation module phase input interface, a symbol-level fast phase tracking loop, a frame-level Kalman filter, and a joint phase rotator. The symbol-level fast phase tracking loop sequentially includes a decision feedback phase error detector, a second-order phase-locked loop filter, and a numerically controlled oscillator. The frame-level Kalman filter sequentially includes a pilot extraction module, a Kalman prediction step unit, a Kalman update step unit, and a phase / frequency offset state output unit. In this embodiment, the symbol-level fast phase tracking loop outputs instantaneous high-frequency linear phase components, and the frame-level Kalman filter outputs slowly varying linear phase components. The two types of linear phase components are added to obtain a linear phase noise estimation result. The linear phase noise estimation result and the nonlinear phase estimation result are then combined and input into the joint phase rotator, which performs unified joint phase correction and outputs a phase-compensated digital signal.

[0083] In this embodiment, the configuration can simultaneously adapt to high-speed phase jitter and low-frequency phase drift, and achieve joint unified correction of linear phase noise and fiber nonlinear phase. This avoids error accumulation caused by multi-level step-by-step compensation, significantly suppresses phase diffusion of high-order QAM constellations, and significantly improves the phase stability and bit error rate performance of high-speed coherent transmission systems.

[0084] S270. Perform signal fusion processing on the digital signal after dispersion compensation, the digital signal after polarization mode dispersion compensation, the digital signal after nonlinear compensation, and the digital signal after phase noise compensation to obtain the jointly compensated digital signal.

[0085] Among them, signal fusion processing can be performed on multiple sub-signals that have been compensated for a single impairment in the complex domain, and the compensation effects of each channel can be combined to eliminate various transmission impairments such as chromatic dispersion, polarization mode dispersion, optical nonlinearity, and phase noise.

[0086] In this embodiment, after obtaining four parallel sub-signals—the dispersion-compensated digital signal, the polarization mode dispersion-compensated digital signal, the nonlinearity-compensated digital signal, and the phase noise-compensated digital signal—signal fusion processing is performed to integrate the correction effects of each compensation path. This process considers the compensation benefits for various impairments, including chromatic dispersion, polarization mode dispersion, optical nonlinearity, and phase noise, while avoiding residual distortion caused by insufficient compensation from a single compensation module. The final result is a jointly compensated digital signal. It can be understood that in this embodiment, the four compensated digital signals each correspond to the compensation results for a single type of transmission impairment. Based on the impairment prior parameters output by the channel sensing, the fusion weights corresponding to each sub-signal are configured, and the four sub-signals are weighted and fused in the complex domain to obtain the jointly compensated digital signal.

[0087] In this embodiment, the five modules included in the signal joint compensation model are jointly optimized during the training phase using a unified end-to-end loss function, defined as: L total =λ1·L BER +λ2·L EVM +λ3·L NLI _Reg+λ4·L complexity .

[0088] Among them, L BER For soft decision BER estimation (approximated by Q factor); L EVM L is the mean square value of the error vector magnitude; NLI_Reg For physical constraint regularization terms; L Complexity This is a regularization term for network parameters, constraining the total number of parameters to not exceed a preset resource budget. Weight coefficients λ1~λ4 are automatically adjusted according to the modulation format. The trained complete model is exported in TensorFlow Lite quantization format or converted to an RTL description that can run on FPGA / ASIC using High-Level Synthesis (HLS) tools. After deployment, the online adaptive parameter updates of each module (weights of SA-FDE and DU-PMDC) run in parallel with data processing in a pipelined manner, without requiring additional data path latency budget.

[0089] In this embodiment, the end-to-end joint loss function includes four items: format adaptive BER, EVM, physical regularization, and complexity constraints. This ensures the performance balance of each compensation module under different QAM formats and avoids performance bias caused by optimization of a single metric.

[0090] S280. The quality of the jointly compensated digital signal is evaluated using an end-to-end loss function to obtain the quality evaluation result.

[0091] S290. If the quality assessment result does not meet the preset threshold, the internal parameters of the signal joint compensation model are updated through the multi-module joint adaptive mechanism driven by the meta-learning controller, and the signal compensation processing is continued to be performed in parallel by multiple compensation modules contained in the signal joint compensation model until the quality assessment result meets the preset threshold and the target digital signal is output.

[0092] In this embodiment, a multi-module joint adaptive mechanism driven by online meta-learning is used to maintain shared meta-parameters across modulation formats using the MAML framework. This supports burst damage response within 10 symbol periods or less. Furthermore, the neural network-based nonlinear compensation (NN-NLC) scheme reduces computational complexity by several times and significantly lowers hardware power consumption under the condition that Q-factor loss is negligible. In large-scale WDM systems, this can save kilowatt-level system power consumption and has substantial improvement significance for the power usage efficiency (PUE) of data center operators.

[0093] Example 3 Figure 5 This is a schematic diagram of a coherent optical transmission signal impairment compensation device according to Embodiment 3 of the present invention. Figure 5 As shown, the device includes: The signal acquisition module 510 is used to acquire the original damaged digital signal from the coherent receiver.

[0094] The joint compensation module 520 is used to input the original damage digital signal into the signal joint compensation model. The signal joint compensation model performs online adaptive damage compensation processing for chromatic dispersion, polarization mode dispersion, optical nonlinearity and phase noise in parallel through multiple compensation modules contained in the signal joint compensation model to obtain the jointly compensated digital signal. The signal joint compensation model has a built-in meta-learning controller.

[0095] The quality assessment module 530 is used to assess the quality of the jointly compensated digital signal through an end-to-end loss function and obtain the quality assessment result.

[0096] The target digital signal determination module 540 is used to update the internal parameters of the signal joint compensation model through a multi-module joint adaptive mechanism driven by the meta-learning controller if the quality assessment result does not meet the preset threshold, and continue to perform signal compensation processing in parallel through multiple compensation modules included in the signal joint compensation model until the quality assessment result meets the preset threshold and outputs the target digital signal.

[0097] Optionally, the signal joint compensation model includes a channel sensing and modulation identification module, a segmented adaptive dispersion equalization module, a polarization state demultiplexing and polarization mode dispersion joint compensation module, a nonlinear damage compensation module, and a multi-granularity phase noise compensation module. Correspondingly, the joint compensation module 520 includes: The parameter estimation unit is used to input the original damaged digital signal into the signal joint compensation model. Through the channel sensing and modulation identification module, the original damaged digital signal is processed for modulation format identification and online estimation of channel damage parameters to obtain the modulation format and multiple damage prior parameters. The original damaged digital signal, the modulation format, and the multiple damage prior parameters are then input in parallel into the piecewise adaptive dispersion equalization module, the polarization state demultiplexing and polarization mode dispersion joint compensation module, the nonlinear damage compensation module, and the multi-granularity phase noise compensation module, respectively. The chromatic dispersion compensation unit is used to perform chromatic dispersion damage compensation processing on the original damaged digital signal through the piecewise adaptive dispersion equalization module, combined with the modulation format and multiple damage prior parameters, to obtain the dispersion-compensated digital signal. The polarization mode dispersion compensation unit is used to perform polarization mode dispersion damage compensation processing on the original damaged digital signal through polarization state demultiplexing and polarization mode dispersion joint compensation module, so as to obtain the digital signal after polarization mode dispersion compensation. The optical nonlinear compensation unit is used to perform optical nonlinear damage compensation processing on the original damage digital signal through the nonlinear damage compensation module to obtain the nonlinearly compensated digital signal. The phase noise damage compensation unit is used to perform phase noise damage compensation processing on the original damaged digital signal through the multi-granularity phase noise compensation module to obtain the phase noise compensated digital signal. The signal fusion unit is used to perform signal fusion processing on the dispersion-compensated digital signal, the polarization mode dispersion-compensated digital signal, the nonlinear compensation-compensated digital signal, and the phase noise-compensated digital signal to obtain the jointly compensated digital signal.

[0098] Optional, multiple damage prior parameters include dispersion estimation parameters, differential group delay estimation parameters, nonlinear phase noise power estimation parameters, and laser linewidth estimation parameters; The parameter estimation unit is specifically used for: extracting lightweight high-order cumulant features from the original damaged digital signal through the channel sensing and modulation identification module and identifying the current modulation format by combining a multilayer perceptron classifier to obtain the corresponding modulation format and symbol rate; and performing frame-by-frame online estimation processing on the original damaged digital signal based on the modulation format, symbol rate, and pilot-assisted maximum likelihood estimation to obtain dispersion estimation parameters, differential group delay estimation parameters, nonlinear phase noise power estimation parameters, and laser linewidth estimation parameters.

[0099] Optionally, a chromatic dispersion compensation unit is specifically used for: calculating an initial frequency domain equalization filter based on dispersion estimation parameters; performing coarse compensation on the original damaged digital signal based on the initial frequency domain equalization filter to obtain a digital signal with residual dispersion; adaptively determining the filter length of the finite impulse response filter bank according to the modulation format; using an adaptive iterative algorithm to perform fine compensation optimization on the digital signal with residual dispersion; and outputting a dispersion-compensated digital signal.

[0100] Optionally, the polarization state demultiplexing and polarization mode dispersion joint compensation module includes a butterfly-shaped complex finite impulse response expansion layer and a second-order polarization mode dispersion compensation layer. The polarization mode dispersion compensation unit is specifically used to: adaptively determine the number of network layers and filter taps of the butterfly-shaped complex finite impulse response expansion layer based on the modulation format and differential group delay estimation parameters; iteratively update the complex weight matrix and adaptive learning step size of each network layer using multi-loop decision pointing loss as the optimization criterion; complete the polarization demultiplexing of dual-polarization signals and first-order polarization mode dispersion compensation to obtain the compensated intermediate signal; through the second-order polarization mode dispersion compensation layer, combined with the differential group delay estimation parameters at each frequency point, extract the frequency-varying link equivalent Jones matrix of the compensated intermediate signal and construct a frequency differential model to correct the higher-order polarization mode dispersion distortion caused by the variation of differential group delay with optical frequency, and obtain the digital signal after polarization mode dispersion compensation.

[0101] Optionally, an optical nonlinear compensation unit is specifically used for: segmenting the polarization mode dispersion compensated digital signal into segments of fixed length to obtain an equalized complex signal sequence; inputting the equalized complex signal sequence into a nonlinear damage compensation module to predict the corresponding nonlinear phase estimation result and amplitude compression factor; based on the physical constraints of fiber nonlinear transmission and combined with the nonlinear phase noise power estimation parameters, performing regularization correction on the nonlinear phase estimation result and amplitude compression factor to obtain the corrected nonlinear phase estimation result and amplitude compression factor, and performing inverse compensation on the polarization mode dispersion compensated digital signal to obtain the nonlinearly compensated digital signal.

[0102] Optionally, the multi-granularity phase noise compensation module includes a symbol-level fast phase tracking loop and a frame-level slow phase trend estimator; the phase noise impairment compensation unit is specifically used for: adaptively configuring the loop bandwidth according to the modulation format through the symbol-level fast phase tracking loop to compensate for high-speed random phase fluctuations and output instantaneous high-frequency linear phase components; dynamically adjusting the Kalman filter parameters based on intra-frame pilot symbols and combined with laser linewidth estimation parameters output by the frame-level slow phase trend estimator to correct low-frequency phase noise and frequency shifts and output slowly varying linear phase components; determining the linear phase noise estimation result based on the high-frequency linear phase component and the slowly varying linear phase component; fusing the nonlinear phase estimation result and the linear phase noise estimation result and performing joint phase correction to output the jointly compensated digital signal.

[0103] The coherent optical transmission signal impairment compensation device provided in the embodiments of the present invention can execute the coherent optical transmission signal impairment compensation method provided in any embodiment of the present invention, and has the corresponding functional modules and beneficial effects of the method.

[0104] Example 4 Figure 6This is a schematic diagram of an electronic device according to Embodiment 4 of the present invention. The electronic device is intended to represent various forms of digital computers, such as laptop computers, desktop computers, workstations, personal digital assistants, servers, blade servers, mainframe computers, and other suitable computers. The electronic device can also represent various forms of mobile devices, such as personal digital processors, cellular phones, smartphones, wearable devices (such as helmets, glasses, watches, etc.), and other similar computing devices. The components shown herein, their connections and relationships, and their functions are merely illustrative and are not intended to limit the implementation of the invention described and / or claimed herein.

[0105] like Figure 6 As shown, the electronic device 610 includes at least one processor 611 and a memory, such as a read-only memory (ROM) 612 or a random access memory (RAM) 613, communicatively connected to the at least one processor 611. The memory stores computer programs executable by the at least one processor. The processor 611 can perform various appropriate actions and processes based on the computer program stored in the ROM 612 or loaded into the RAM 613 from storage unit 618. The RAM 613 may also store various programs and data required for the operation of the electronic device 610. The processor 611, ROM 612, and RAM 613 are interconnected via a bus 614. An input / output (I / O) interface 615 is also connected to the bus 614.

[0106] Multiple components in electronic device 610 are connected to I / O interface 615, including: input unit 616, such as keyboard, mouse, etc.; output unit 617, such as various types of displays, speakers, etc.; storage unit 618, such as disk, optical disk, etc.; and communication unit 619, such as network card, modem, wireless transceiver, etc. Communication unit 619 allows electronic device 610 to exchange information / data with other devices through computer networks such as the Internet and / or various telecommunications networks.

[0107] Processor 611 can be a variety of general-purpose and / or special-purpose processing components with processing and computing capabilities. Some examples of processor 611 include, but are not limited to, a central processing unit (CPU), a graphics processing unit (GPU), various special-purpose artificial intelligence (AI) computing chips, various processors running machine learning model algorithms, a digital signal processor (DSP), and any suitable processor, controller, microcontroller, etc. Processor 611 performs the various methods and processes described above, such as coherent optical transmission signal impairment compensation methods.

[0108] In some embodiments, the coherent optical transmission signal impairment compensation method may be implemented as a computer program tangibly contained in a computer-readable storage medium, such as storage unit 618. In some embodiments, part or all of the computer program may be loaded and / or installed on electronic device 610 via ROM 612 and / or communication unit 619. When the computer program is loaded into RAM 613 and executed by processor 611, one or more steps of the coherent optical transmission signal impairment compensation method described above may be performed. Alternatively, in other embodiments, processor 611 may be configured to perform the coherent optical transmission signal impairment compensation method by any other suitable means (e.g., by means of firmware).

[0109] Various embodiments of the systems and techniques described above herein can be implemented in digital electronic circuit systems, integrated circuit systems, field-programmable gate arrays (FPGAs), application-specific integrated circuits (ASICs), application-specific standard products (ASSPs), systems-on-a-chip (SoCs), payload-programmable logic devices (CPLDs), computer hardware, firmware, software, and / or combinations thereof. These various embodiments may include implementations in one or more computer programs that can be executed and / or interpreted on a programmable system including at least one programmable processor, which may be a dedicated or general-purpose programmable processor, capable of receiving data and instructions from a storage system, at least one input device, and at least one output device, and transmitting data and instructions to the storage system, the at least one input device, and the at least one output device.

[0110] Computer programs used to implement the methods of the present invention may be written in any combination of one or more programming languages. These computer programs may be provided to a processor of a general-purpose computer, a special-purpose computer, or other programmable data processing device, such that when executed by the processor, the computer programs cause the functions / operations specified in the flowcharts and / or block diagrams to be performed. The computer programs may be executed entirely on a machine, partially on a machine, or as a standalone software package, partially on a machine and partially on a remote machine, or entirely on a remote machine or server.

[0111] In the context of this invention, a computer-readable storage medium can be a tangible medium that may contain or store a computer program for use by or in conjunction with an instruction execution system, apparatus, or device. A computer-readable storage medium may include, but is not limited to, electronic, magnetic, optical, electromagnetic, infrared, or semiconductor systems, apparatus, or devices, or any suitable combination thereof. Alternatively, a computer-readable storage medium may be a machine-readable signal medium. More specific examples of machine-readable storage media include electrical connections based on one or more wires, portable computer disks, hard disks, random access memory (RAM), read-only memory (ROM), erasable programmable read-only memory (EPROM or flash memory), optical fibers, portable compact disk read-only memory (CD-ROM), optical storage devices, magnetic storage devices, or any suitable combination thereof.

[0112] To provide interaction with a user, the systems and techniques described herein can be implemented on an electronic device having: a display device (e.g., a CRT (cathode ray tube) or LCD (liquid crystal display) monitor) for displaying information to the user; and a keyboard and pointing device (e.g., a mouse or trackball) through which the user provides input to the electronic device. Other types of devices can also be used to provide interaction with the user; for example, feedback provided to the user can be any form of sensory feedback (e.g., visual feedback, auditory feedback, or tactile feedback); and input from the user can be received in any form (including sound input, voice input, or tactile input).

[0113] The systems and technologies described herein can be implemented in computing systems that include backend components (e.g., as data servers), or middleware components (e.g., application servers), or frontend components (e.g., user computers with graphical user interfaces or web browsers through which users can interact with implementations of the systems and technologies described herein), or any combination of such backend, middleware, or frontend components. The components of the system can be interconnected via digital data communication of any form or medium (e.g., communication networks). Examples of communication networks include local area networks (LANs), wide area networks (WANs), blockchain networks, and the Internet.

[0114] A computing system can include clients and servers. Clients and servers are generally located far apart and typically interact through communication networks. The client-server relationship is created by computer programs running on the respective computers and having a client-server relationship with each other. The server can be a cloud server, also known as a cloud computing server or cloud host, which is a hosting product within the cloud computing service system to address the shortcomings of traditional physical hosts and VPS services, such as high management difficulty and weak business scalability.

[0115] It should be understood that the various forms of processes shown above can be used, with steps reordered, added, or deleted. For example, the steps described in this invention can be executed in parallel, sequentially, or in different orders, as long as the desired result of the technical solution of this invention can be achieved, and this is not limited herein.

[0116] The specific embodiments described above do not constitute a limitation on the scope of protection of this invention. Those skilled in the art should understand that various modifications, combinations, sub-combinations, and substitutions can be made according to design requirements and other factors. Any modifications, equivalent substitutions, and improvements made within the spirit and principles of this invention should be included within the scope of protection of this invention.

Claims

1. A method for compensating for impairments in coherent optical transmission signals, characterized in that, include: Acquire the raw, impaired digital signal from the coherent receiver; The original damaged digital signal is input into the signal joint compensation model. The multiple compensation modules contained in the signal joint compensation model perform online adaptive damage compensation processing for chromatic dispersion, polarization mode dispersion, optical nonlinearity and phase noise in parallel to obtain the jointly compensated digital signal. The signal joint compensation model has a built-in meta-learning controller. The quality of the jointly compensated digital signal is evaluated using an end-to-end loss function to obtain the quality evaluation result. If the quality assessment result does not meet the preset threshold, the internal parameters of the signal joint compensation model are updated through the multi-module joint adaptive mechanism driven by the meta-learning controller, and the signal compensation processing continues to be performed in parallel by multiple compensation modules included in the signal joint compensation model until the quality assessment result meets the preset threshold and the target digital signal is output.

2. The method according to claim 1, characterized in that, The signal joint compensation model includes a channel sensing and modulation identification module, a segmented adaptive dispersion equalization module, a polarization state demultiplexing and polarization mode dispersion joint compensation module, a nonlinear damage compensation module, and a multi-granularity phase noise compensation module. Accordingly, the original damaged digital signal is input into the signal joint compensation model. Multiple compensation modules included in the signal joint compensation model perform parallel online adaptive damage compensation processing for chromatic dispersion, polarization mode dispersion, optical nonlinearity, and phase noise to obtain the jointly compensated digital signal, including: The original damaged digital signal is input into the signal joint compensation model. The channel sensing and modulation identification module performs modulation format identification processing and online estimation of channel damage parameters on the original damaged digital signal to obtain the modulation format and multiple damage prior parameters. The original damaged digital signal, the modulation format, and the multiple damage prior parameters are then input in parallel into the segmented adaptive dispersion equalization module, the polarization state demultiplexing and polarization mode dispersion joint compensation module, the nonlinear damage compensation module, and the multi-granularity phase noise compensation module. The original damaged digital signal is subjected to chromatic dispersion damage compensation processing by the segmented adaptive dispersion equalization module, combined with the modulation format and the multiple damage prior parameters, to obtain the dispersion-compensated digital signal. The original damaged digital signal is processed by the polarization state demultiplexing and polarization mode dispersion joint compensation module to obtain a digital signal after polarization mode dispersion compensation. The original damage digital signal is subjected to optical nonlinear damage compensation processing by the nonlinear damage compensation module to obtain a nonlinearly compensated digital signal. The original damaged digital signal is subjected to phase noise damage compensation processing by the multi-granularity phase noise compensation module to obtain a phase noise compensated digital signal. The dispersion-compensated digital signal, the polarization mode dispersion-compensated digital signal, the nonlinearity-compensated digital signal, and the phase noise-compensated digital signal are subjected to signal fusion processing to obtain a jointly compensated digital signal.

3. The method according to claim 2, characterized in that, The multiple damage prior parameters include dispersion estimation parameters, differential group delay estimation parameters, nonlinear phase noise power estimation parameters, and laser linewidth estimation parameters; The channel sensing and modulation identification module performs modulation format identification and online channel impairment parameter estimation on the original impaired digital signal to obtain the modulation format and multiple prior impairment parameters, including: The channel sensing and modulation identification module extracts lightweight high-order cumulant features from the original damaged digital signal and combines them with a multilayer perceptron classifier to identify the current modulation format, thereby obtaining the corresponding modulation format and symbol rate. Based on the modulation format, the symbol rate, and the pilot-assisted maximum likelihood estimation, the original damaged digital signal is subjected to frame-by-frame online estimation processing to obtain the dispersion estimation parameters, the differential group delay estimation parameters, the nonlinear phase noise power estimation parameters, and the laser linewidth estimation parameters.

4. The method according to claim 3, characterized in that, The segmented adaptive dispersion equalization module, combined with the modulation format and the multiple prior damage parameters, performs chromatic dispersion damage compensation processing on the original damaged digital signal to obtain a dispersion-compensated digital signal, including: An initial frequency domain equalization filter is calculated based on the dispersion estimation parameters. The original damaged digital signal is coarsely compensated based on the initial frequency domain equalization filter to obtain a digital signal with residual dispersion. The filter length of the finite impulse response filter bank is adaptively determined according to the modulation format, and the digital signal with residual dispersion is finely compensated and optimized using an adaptive iterative algorithm to output the dispersion-compensated digital signal.

5. The method according to claim 3, characterized in that, The polarization state demultiplexing and polarization mode dispersion joint compensation module includes a butterfly-shaped complex finite impulse response expansion layer and a second-order polarization mode dispersion compensation layer. The original damaged digital signal is subjected to polarization mode dispersion damage compensation processing by the polarization state demultiplexing and polarization mode dispersion joint compensation module to obtain a polarization mode dispersion compensated digital signal, including: The number of network layers and the number of filter taps of the butterfly-shaped complex finite impulse response expansion layer are adaptively determined according to the modulation format and the differential group delay estimation parameters. The complex weight matrix and adaptive learning step size of each network layer are iteratively updated using multi-ring decision pointing loss as the optimization criterion to complete the polarization demultiplexing of dual polarization signals and first-order polarization mode dispersion compensation, thereby obtaining the compensated intermediate signal. By using the second-order polarization mode dispersion compensation layer and combining the differential group delay estimation parameters at each frequency point, the frequency-varying link equivalent Jones matrix of the compensated intermediate signal is extracted and a frequency differential model is constructed to correct the high-order polarization mode dispersion distortion caused by the change of differential group delay with optical frequency, thus obtaining the digital signal after polarization mode dispersion compensation.

6. The method according to claim 3, characterized in that, The original damage digital signal is subjected to optical nonlinear damage compensation processing by the nonlinear damage compensation module to obtain a nonlinearly compensated digital signal, including: The digital signal after polarization mode dispersion compensation is segmented into segments of fixed length to obtain an equalized complex signal sequence. The equalized complex signal sequence is input into the nonlinear damage compensation module to predict the corresponding nonlinear phase estimation result and amplitude compression factor. Based on the physical constraints of nonlinear transmission in optical fiber, and combined with the nonlinear phase noise power estimation parameters, the nonlinear phase estimation result and the amplitude compression factor are normalized to obtain the corrected nonlinear phase estimation result and amplitude compression factor. Then, the digital signal after polarization mode dispersion compensation is reversed to obtain the nonlinearly compensated digital signal.

7. The method according to claim 6, characterized in that, The multi-granularity phase noise compensation module includes a symbol-level fast phase tracking loop and a frame-level slow phase trend estimator. The original damaged digital signal is subjected to phase noise damage compensation processing by the multi-granularity phase noise compensation module to obtain a phase noise-compensated digital signal, including: The symbol-level fast phase tracking loop adaptively configures the loop bandwidth according to the modulation format to compensate for high-speed random phase fluctuations and output instantaneous high-frequency linear phase components. The frame-level slow phase trend estimator dynamically adjusts the Kalman filter parameters based on the intra-frame pilot symbols and the laser linewidth estimation parameters output by the channel sense output to correct low-frequency phase noise and frequency offset and output slowly varying linear phase components. The linear phase noise estimation result is determined based on the high-frequency linear phase component and the slowly varying linear phase component; The nonlinear phase estimation result and the linear phase noise estimation result are fused and combined with phase correction to obtain the phase noise compensated digital signal.

8. A coherent optical transmission signal impairment compensation device, characterized in that, include: The signal acquisition module is used to acquire the raw impaired digital signal from the coherent receiver; The joint compensation module is used to input the original damaged digital signal into the signal joint compensation model, and to perform online adaptive damage compensation processing for chromatic dispersion, polarization mode dispersion, optical nonlinearity and phase noise in parallel through multiple compensation modules included in the signal joint compensation model, so as to obtain the jointly compensated digital signal; wherein, the signal joint compensation model has a built-in meta-learning controller; The quality assessment module is used to assess the quality of the jointly compensated digital signal using the end-to-end loss function, and obtain the quality assessment result. The target digital signal determination module is used to update the internal parameters of the signal joint compensation model through a multi-module joint adaptive mechanism driven by the meta-learning controller if the quality assessment result does not meet the preset threshold, and continue to perform signal compensation processing in parallel through multiple compensation modules included in the signal joint compensation model until the quality assessment result meets the preset threshold, and output the target digital signal.

9. An electronic device, characterized in that, The electronic device includes: At least one processor; and A memory communicatively connected to the at least one processor; wherein, The memory stores a computer program that can be executed by the at least one processor, the computer program being executed by the at least one processor to enable the at least one processor to perform the coherent optical transmission signal impairment compensation method according to any one of claims 1-7.

10. A computer-readable storage medium, characterized in that, The computer-readable storage medium stores computer instructions that are used to cause a processor to execute the coherent optical transmission signal impairment compensation method according to any one of claims 1-7.

Images