A multi-rate adaptive compatible coherent demodulation method and device

Abstract

The application provides a kind of multi-rate adaptive compatible coherent demodulation method and device, belong to coherent demodulation technical field;The method comprises: the pre-treatment of coherent light signal;Adopt the double-stage mechanism of coarse identification and fine identification combination to automatically identify the rate and modulation format of signal, wherein fine identification utilizes lightweight CNN classifier;According to the recognition result, adaptively configure the core demodulation parameters such as sampling rate, equalizer order, operation bit width and carrier recovery algorithm;Through the unified configurable modular architecture, complete synchronization, equalization, carrier recovery and other demodulation processing;Soft decision is made to demodulated signal and real-time monitoring performance index;According to the monitoring result, closed-loop dynamic fine tuning is carried out on demodulation parameter;The application realizes the automatic identification and compatible demodulation of 100G / 200G / 400G and PM-QPSK / PM-16QAM signal, and is suitable for general coherent acquisition platform, intelligent optical network receiver and multi-rate optical communication tester.

Description

Technical Field

[0001] This invention relates to the field of coherent demodulation technology, and in particular to a multi-rate adaptive compatible coherent demodulation method and apparatus. Background Technology

[0002] High-speed coherent optical communication is a core technology of modern optical transmission networks, widely used in backbone networks, data center interconnection, and 5G bearer networks. With the diversification of service requirements, optical networks need to support multiple rate levels such as 100G, 200G, and 400G simultaneously, and be compatible with multiple modulation formats such as PM-QPSK and PM-16QAM.

[0003] However, signals with different rates and modulation formats place different demands on the sampling rate, equalizer complexity, and carrier recovery algorithm of the demodulation system. How to achieve adaptive and compatible demodulation of multiple rates and multiple modulation formats in a single receiving platform has become a technical problem that urgently needs to be solved in this field.

[0004] Currently, existing high-speed coherent demodulation systems mainly adopt a single-rate dedicated design scheme.

[0005] For example, 100G systems are typically based on PM-QPSK modulation, employing a fixed processing flow of 80GSps sampling, constant modulus equalization (CMA), and fourth-order frequency offset estimation combined with VV phase recovery; 200G systems are based on PM-16QAM modulation, employing 160GSps sampling, improved CMA equalization, and high-order carrier recovery algorithms; 400G systems use higher sampling rates (such as 320GSps), multi-tap equalizers, and nonlinear compensation modules. These solutions are optimized for their respective rates, but their architectures are incompatible, requiring independently designed hardware modules, resulting in significant waste of hardware resources, high equipment costs, and poor system flexibility.

[0006] To meet the requirements of multi-rate compatibility, some studies in recent years have attempted to introduce adaptive mechanisms.

[0007] For example, some solutions employ modulation format recognition methods based on constellation diagram features, performing single-stage classification decisions by extracting features such as constellation point distribution and mean square error. However, the accuracy of such methods drops significantly in low optical signal-to-noise ratio (OSNR, e.g., below 12dB), making it difficult to meet the requirements of engineering applications. Other solutions introduce deep learning models (such as convolutional neural networks) for modulation format recognition, but traditional deep learning networks have a large number of parameters, high computational complexity, and inference latency typically reaching millisecond levels, making them unsuitable for deployment on real-time hardware platforms such as FPGAs, and unable to simultaneously perform joint recognition of rate and modulation format. Furthermore, while some solutions achieve semi-automatic parameter configuration, they employ fixed parameter adjustment strategies and lack a closed-loop optimization mechanism based on performance feedback, resulting in large fluctuations in demodulation performance and long convergence times after rate switching, failing to meet the needs of rapid service scheduling in dynamic optical networks.

[0008] In summary, the existing technologies have the following main defects: (1) different demodulation architectures are incompatible, hardware reuse rate is low, and cost is high; (2) there is a lack of efficient, real-time and high-precision automatic identification mechanism for rate and modulation format, which relies on manual configuration and cannot adapt to dynamic rate switching scenarios; (3) the core demodulation parameters (sampling rate, equalizer order, carrier recovery algorithm, etc.) are fixed or only statically loaded, which cannot be adaptively adjusted according to the input signal state, resulting in significant demodulation performance loss; (4) there is a lack of closed-loop optimization during multi-rate switching, resulting in large performance fluctuations, slow convergence, and inability to meet real-time service requirements.

[0009] Therefore, there is an urgent need to provide a coherent demodulation method and apparatus that can achieve multi-rate adaptive compatibility to solve the above problems and meet the practical engineering needs of general coherent acquisition platforms, intelligent optical network receivers and multi-rate optical communication testers. Summary of the Invention

[0010] The purpose of this invention is to overcome the shortcomings of the prior art and provide a complete multi-rate adaptive coherent demodulation closed-loop system that can be used for real-time processing of FPGAs. It aims to achieve functions such as automatic identification, adaptive parameter configuration, unified modular demodulation, and closed-loop performance optimization, and is applicable to engineering scenarios such as general coherent acquisition platforms, intelligent optical network receivers, and multi-rate optical communication testers.

[0011] To achieve the above objectives, this application proposes a multi-rate adaptive compatible coherent demodulation method for adaptive compatible coherent demodulation of coherent optical signals, comprising the following steps: The received coherent optical signal is preprocessed to obtain the preprocessed baseband signal; The preprocessed baseband signal is automatically identified in terms of rate and modulation format to obtain the identification result; wherein the automatic identification adopts a two-stage mechanism combining a coarse identification stage and a fine identification stage. Based on the identification results, adaptive parameter configuration is performed, and demodulation parameters matching the rate and modulation format are loaded; Based on the identification results and the configured demodulation parameters, the preprocessed baseband signal is subjected to multi-rate compatible demodulation through a unified and configurable modular architecture to obtain the demodulated standard constellation signal. The multi-rate compatible demodulation processing includes adaptive synchronization processing, adaptive sampling rate conversion processing, adaptive equalization processing, and multi-rate carrier recovery processing. The demodulated standard constellation signal is subjected to decision output and performance monitoring to obtain demodulation result output, performance index data output and performance monitoring feedback data. Based on the performance monitoring feedback data, closed-loop optimization feedback is performed to dynamically fine-tune the demodulation parameters, resulting in optimized parameter feedback, which is then applied to adaptive parameter configuration, adaptive equalization processing, and multi-rate carrier recovery processing. The coherent demodulation method supports at least 100G, 200G, and 400G rates, as well as PM-QPSK and PM-16QAM modulation formats for coherent optical signal compatible demodulation.

[0012] As a further solution, the two-stage mechanism specifically includes: The constellation diagram features, symbol rate, spectral bandwidth, and frame structure features of the preprocessed baseband signal are extracted, and a multi-dimensional feature vector is constructed. Coarse identification stage: Based on the symbol rate and spectral bandwidth of the preprocessed baseband signal, rapid classification is performed to classify the preprocessed baseband signal into the corresponding rate range; Precise recognition stage: Within the corresponding rate range, the multi-dimensional feature vector is input into the trained lightweight CNN classifier, which outputs accurate recognition results of rate and modulation format.

[0013] As a further solution, the adaptive parameter configuration step further includes: Obtain accurate rate and modulation format recognition results from a lightweight CNN classifier; Parameter combinations are dynamically loaded based on the built-in parameter mapping table; Verify and output the parameter combination corresponding to the recognition result; The specific demodulation parameters corresponding to the parameter combinations are configured based on the end-to-end demodulation parameter configuration package. The parameter mapping table supports online updates to expand new rates and modulation formats; the parameter combinations include: sampling rate, equalizer type, equalizer order, carrier recovery algorithm type, and equalizer operation bit width.

[0014] As a further solution, the parameter adaptive adjustment strategy of the parameter mapping table is further configured to: adaptively adjust the number of taps and the operation bit width of the equalizer according to the identified signal rate; wherein... For 100G signals, the configuration is 32 taps, 16-bit operation width, and the equalizer type adopts the basic CMA algorithm. For 200G signals, the configuration is 64 taps and 18-bit operation width. The equalizer type adopts a joint algorithm of improved CMA and Least Mean Square (LMS). For 400G signals, the configuration is 128 taps, 20-bit operation width, and the equalizer type adopts a joint algorithm of improved CMA and Least Mean Square (LMS).

[0015] As a further solution, the adaptive synchronization processing is performed using a variable-length synchronization word detection algorithm; wherein... The variable-length synchronization word detection algorithm supports adaptive matching of frame synchronization words at rates of 100G / 200G / 400G, and the synchronization word length is automatically adjusted to 64bit / 128bit / 256bit according to the rate, with a synchronization time ≤10μs. After automatic adjustment, an adaptive clock recovery algorithm is executed to ensure clock synchronization accuracy at different rates.

[0016] As a further solution, the adaptive sampling rate conversion process adaptively switches between 80GSps / 160GSps / 320GSps sampling rates according to the sampling rate requirements of different baseband signals; wherein, by employing a fractional sampling rate conversion algorithm, signal integrity is ensured during different rate conversions.

[0017] As a further solution, the adaptive equalization processing is performed using a multi-tap equalizer with a configurable number of taps. The number of taps is adaptively adjusted to 32 / 64 / 128 taps according to the rate, and the complexity of the multi-tap equalizer is automatically adjusted to meet the dispersion compensation requirements of different rates.

[0018] As a further solution, the multi-rate carrier recovery process adaptively adjusts the carrier recovery algorithm based on the identified modulation format; wherein... Frequency offset estimation: An adaptive order frequency offset estimation algorithm is adopted; for 100G signals, a fourth power algorithm is used, and for 200G / 400G signals, a higher order fourth power algorithm is used. Phase recovery: Different phase recovery algorithms are used for different modulation formats: for PM-QPSK signals, the VV phase recovery algorithm is used; for PM-16QAM signals, the partitioned VV phase recovery algorithm is used. Nonlinear compensation: For 400G signals, the nonlinear compensation coefficient is dynamically adjusted according to the signal power to compensate for phase distortion caused by fiber nonlinearity.

[0019] As a further solution, the closed-loop optimization feedback is specifically configured as follows: Real-time monitoring of error vector amplitude (EVM) and bit error rate (BER) performance metrics with a granularity of 10 μs; When judging BER>1×10 -9 If EVM > 8%, perform at least one of the following optimization operations: First, adjust the step size of the adaptive equalizer; If the performance still does not meet the threshold, increase the number of iterations of the carrier recovery algorithm; For 400G signals, the nonlinear compensation coefficient is adjusted synchronously.

[0020] On the other hand, the present invention also provides a multi-rate adaptive compatible coherent demodulation apparatus for implementing the multi-rate adaptive compatible coherent demodulation method described in any one of the above descriptions, characterized in that it includes: Signal input and preprocessing unit: preprocesses the received coherent optical signal to obtain the preprocessed baseband signal; Rate and modulation format identification unit: automatically identifies the rate and modulation format of the preprocessed baseband signal to obtain the identification result; wherein, the automatic identification adopts a two-stage mechanism combining a coarse identification stage and a fine identification stage; Adaptive parameter configuration unit: performs adaptive parameter configuration based on the recognition result, and loads demodulation parameters that match the rate and modulation format; Multi-rate compatible demodulation unit: Based on the identification results and the configured demodulation parameters, the preprocessed baseband signal is subjected to multi-rate compatible demodulation through a unified and configurable modular architecture to obtain the demodulated standard constellation signal; The multi-rate compatible demodulation processing includes adaptive synchronization processing, adaptive sampling rate conversion processing, adaptive equalization processing, and multi-rate carrier recovery processing. Decision output and performance monitoring unit: performs decision output and performance monitoring on the demodulated standard constellation signal to obtain demodulation result output, performance index data output and performance monitoring feedback data; Closed-loop optimization feedback unit: Based on the performance monitoring feedback data, it performs closed-loop optimization feedback, dynamically fine-tunes the demodulation parameters, obtains optimized parameter feedback, and applies it to adaptive parameter configuration, adaptive equalization processing, and multi-rate carrier recovery processing.

[0021] Compared with related technologies, the multi-rate adaptive compatible coherent demodulation method and apparatus provided by the present invention have the following advantages: 1. To address the contradiction between accuracy and real-time performance in traditional recognition schemes, this invention proposes a two-stage lightweight CNN recognition and adaptive configuration mechanism of "coarse recognition (based on symbol rate / bandwidth) + fine recognition (lightweight convolutional CNN)". This mechanism achieves microsecond-level real-time recognition while ensuring high accuracy and can be directly deployed on real-time hardware platforms such as FPGA.

[0022] 2. The unified and configurable modular demodulation architecture and mixed-precision resource allocation of this invention break away from the traditional idea that multi-rate systems require multiple independent hardware sets. All core modules (synchronization, equalization, carrier recovery) adopt unified hardware logic and adapt to different rates only through parameter configuration (number of taps, bit width, algorithm type); in particular, the mixed-precision bit width optimization significantly reduces resource consumption.

[0023] 3. This invention focuses on specific algorithms and fast closed-loop optimization for high-order modulation, including a partitioned VV phase recovery algorithm for PM-16QAM to improve linewidth tolerance, and a microsecond-level closed-loop parameter fine-tuning mechanism based on performance monitoring to solve the performance fluctuation problem after switching. Attached Figure Description

[0024] The accompanying drawings, which are incorporated in and form part of this specification, illustrate embodiments consistent with this application and, together with the description, serve to explain the principles of this application.

[0025] To more clearly illustrate the technical solutions in the embodiments of this application or related technologies, the accompanying drawings used in the description of the embodiments or related technologies will be briefly introduced below. Obviously, those skilled in the art can obtain other drawings based on these drawings without creative effort.

[0026] Figure 1 A schematic diagram illustrating the steps of a multi-rate adaptive compatible coherent demodulation method provided by the present invention; The purpose, features, and advantages of this application will be further explained in conjunction with the embodiments and with reference to the accompanying drawings. Detailed Implementation

[0027] To make the objectives, technical solutions, and advantages of the embodiments of the present invention clearer, the technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. The components of the embodiments of the present invention described and shown in the accompanying drawings can generally be arranged and designed in various different configurations.

[0028] Example 1 Please see Figure 1This embodiment provides a multi-rate adaptive compatible coherent demodulation method for adaptively compatible coherent demodulation of coherent optical signals, including the following steps: The received coherent optical signal is preprocessed to obtain the preprocessed baseband signal; The preprocessed baseband signal is automatically identified in terms of rate and modulation format to obtain the identification result; wherein the automatic identification adopts a two-stage mechanism combining a coarse identification stage and a fine identification stage. Based on the identification results, adaptive parameter configuration is performed, and demodulation parameters matching the rate and modulation format are loaded; Based on the identification results and the configured demodulation parameters, the preprocessed baseband signal is subjected to multi-rate compatible demodulation through a unified and configurable modular architecture to obtain the demodulated standard constellation signal. The multi-rate compatible demodulation processing includes adaptive synchronization processing, adaptive sampling rate conversion processing, adaptive equalization processing, and multi-rate carrier recovery processing. The demodulated standard constellation signal is subjected to decision output and performance monitoring to obtain demodulation result output, performance index data output and performance monitoring feedback data. Based on the performance monitoring feedback data, closed-loop optimization feedback is performed to dynamically fine-tune the demodulation parameters, resulting in optimized parameter feedback, which is then applied to adaptive parameter configuration, adaptive equalization processing, and multi-rate carrier recovery processing. The coherent demodulation method supports at least 100G, 200G, and 400G rates, as well as PM-QPSK and PM-16QAM modulation formats for coherent optical signal compatible demodulation.

[0029] It should be noted that the core of this embodiment lies in constructing a complete closed-loop processing flow of "identification—configuration—demodulation—monitoring—optimization". First, the input coherent optical signal is preprocessed (such as gain adjustment and filtering) to provide a high-quality signal for subsequent identification. Then, a two-stage mechanism of coarse identification + fine identification is adopted to automatically determine the signal rate (100G / 200G / 400G) and modulation format (PM-QPSK / PM-16QAM), which solves the problems of low identification accuracy or inability to deploy in real time in the prior art.

[0030] Based on this, matching demodulation parameters (such as sampling rate, equalizer order, carrier recovery algorithm, etc.) are adaptively loaded according to the identification results, and demodulation processing such as synchronization, equalization, and carrier recovery is completed using a unified and configurable modular architecture. Finally, the results are output through soft decision and performance indicators are monitored in real time. Based on the monitoring results, the demodulation parameters are dynamically fine-tuned in a closed loop to ensure that the demodulation performance converges quickly and remains stable. The entire method supports compatible demodulation of multiple rates and modulation formats, requires no manual intervention, and can dynamically respond to changes in signal rate.

[0031] Compared with existing technologies that use multiple independent hardware sets, fixed parameters, and lack automatic recognition, this embodiment achieves system automation and versatility through two-stage recognition and adaptive parameter configuration; it achieves high reuse of hardware resources through a unified modular architecture (reuse rate increased by more than 70%); and it solves the pain points of large performance fluctuations and slow convergence after rate switching through a closed-loop optimization mechanism (switching latency ≤100μs, performance fluctuation ≤0.5dB).

[0032] Therefore, the method covered in this embodiment is not only applicable to the real-time demodulation of 100G / 200G / 400G PM-QPSK / PM-16QAM signals, but can also be extended to other rates and modulation formats, providing an efficient, flexible and low-cost solution for general coherent acquisition platforms, intelligent optical network receivers and other devices.

[0033] Furthermore, the two-stage mechanism specifically includes: The constellation diagram features, symbol rate, spectral bandwidth, and frame structure features of the preprocessed baseband signal are extracted, and a multi-dimensional feature vector is constructed. Coarse identification stage: Based on the symbol rate and spectral bandwidth of the preprocessed baseband signal, rapid classification is performed to classify the preprocessed baseband signal into the corresponding rate range; Precise recognition stage: Within the corresponding rate range, the multi-dimensional feature vector is input into the trained lightweight CNN classifier, which outputs accurate recognition results of rate and modulation format.

[0034] Specifically, to address the problems of low recognition accuracy and high complexity in existing technologies, this embodiment adopts a mechanism of "feature extraction + two-stage pattern matching + lightweight CNN classification": (1) Feature extraction: Extract six key features of the signal, including constellation diagram features (number of constellation points, distribution density, mean square error), symbol rate, spectral bandwidth, and frame structure features (synchronization word, frame length), and construct a multi-dimensional feature vector to solve the problem of poor robustness of single feature recognition. (2) Two-stage identification: Coarse identification stage: First, rapid classification is performed based on symbol rate and spectral bandwidth. The symbol rate of 100G signal is 28GBaud and the spectral bandwidth is 50GHz, 200G is 56GBaud and 100GHz, and 400G is 112GBaud and 200GHz. This step can quickly classify the signal into the corresponding rate range, with a time of ≤0.1μs, thus narrowing the identification range. Fine recognition stage: For the intervals after coarse recognition, extract constellation diagrams and frame structure features, and input them into a trained lightweight CNN classifier for accurate classification; (3) Lightweight CNN design: In view of the problem of high complexity of existing deep learning models, the CNN of this invention adopts a lightweight structure of 1×1 convolutional layer + fully connected layer, with a total of only 256 parameters, less than 1000 floating-point operations, and inference time ≤0.8μs. It can be directly deployed on the FPGA platform in real time, solving the problem that traditional deep learning cannot run in real time. (4) Recognition effect: In scenarios where OSNR is as low as 12dB, the recognition accuracy is still ≥99.9%, which is much higher than the 95% accuracy of existing single-stage recognition schemes, and the overall recognition delay is ≤1μs, which meets the real-time requirements.

[0035] Furthermore, the adaptive parameter configuration step further includes: Obtain accurate rate and modulation format recognition results from a lightweight CNN classifier; Parameter combinations are dynamically loaded based on the built-in parameter mapping table; Verify and output the parameter combination corresponding to the recognition result; The specific demodulation parameters corresponding to the parameter combinations are configured based on the end-to-end demodulation parameter configuration package. The parameter mapping table supports online updates to expand new rates and modulation formats; the parameter combinations include: sampling rate, equalizer type, equalizer order, carrier recovery algorithm type, and equalizer operation bit width.

[0036] First, the system obtains the accurate recognition result (i.e., the signal rate and modulation format) output by the lightweight CNN classifier. Then, based on a built-in parameter mapping table, it dynamically loads the parameter combination corresponding to this result. This mapping table pre-stores the optimal demodulation parameters for different rates and modulation formats. To ensure the correctness and completeness of the parameters, the system also verifies the loaded parameter combination and outputs it only after confirming its accuracy. Finally, through a full-link demodulation parameter configuration package, the parameter combinations are distributed and configured to each demodulation module (such as the equalizer and carrier recovery unit), achieving a complete adaptive conversion from recognition to demodulation parameters.

[0037] Secondly, this parameter combination covers the core configuration items required for multi-rate demodulation, specifically including: sampling rate (supporting adaptive switching of 80GSps / 160GSps / 320GSps), equalizer type (such as basic CMA or improved CMA+LMS joint algorithm), equalizer order (32 / 64 / 128 tap adaptive adjustment), carrier recovery algorithm type (such as fourth power frequency offset estimation, partitioned VV phase recovery, etc.), and equalizer operation bit width (16 / 18 / 20 bit mixed precision).

[0038] The multi-dimensional parameter combination design in this embodiment enables the system to accurately configure the most suitable demodulation algorithm and hardware resources according to the accuracy and complexity requirements of signals of different rates. This avoids the performance waste or resource redundancy caused by using fixed parameters in traditional solutions, and achieves the best balance between performance and resource utilization.

[0039] In addition, a key innovation of this technology is that the parameter mapping table supports online updates. When the system needs to expand its support for new rate levels (such as 800G) or new modulation formats (such as PM-64QAM), there is no need to modify the underlying hardware logic or restart the device. The upgrade can be completed simply by updating the parameter mapping table remotely or locally.

[0040] This design significantly enhances the system's scalability and future adaptability, while reducing subsequent maintenance and upgrade costs. Simultaneously, the end-to-end demodulation parameter configuration package ensures the consistency and completeness of parameter distribution, preventing demodulation anomalies caused by parameter omissions or mismatches, and providing strong support for the stable and reliable operation of multi-rate adaptive demodulation.

[0041] Furthermore, the parameter adaptive adjustment strategy of the parameter mapping table is further configured as follows: adaptively adjusting the number of taps and the operation bit width of the equalizer according to the identified signal rate; wherein, For 100G signals, the configuration is 32 taps, 16-bit operation width, and the equalizer type adopts the basic CMA algorithm. For 200G signals, the configuration is 64 taps and 18-bit operation width. The equalizer type adopts a joint algorithm of improved CMA and Least Mean Square (LMS). For 400G signals, the configuration is 128 taps, 20-bit operation width, and the equalizer type adopts a joint algorithm of improved CMA and Least Mean Square (LMS).

[0042] Specifically, this embodiment adaptively adjusts the core demodulation parameters based on the rate identification results to achieve multi-rate compatibility; the parameter adaptive adjustment strategy is shown in Table 1: Table 1. Parameter Adaptive Adjustment Comparison Table In addition, this embodiment has a built-in updatable multi-rate parameter mapping table. The identification result triggers the loading of the corresponding parameter combination, with a configuration delay of ≤10μs and support for dynamic rate switching. The mapping table supports remote online updates and can be subsequently expanded to support new rates such as 800G without modifying the hardware logic, thus exhibiting strong scalability.

[0043] Furthermore, the adaptive synchronization processing is performed using a variable-length synchronization word detection algorithm; wherein, The variable-length synchronization word detection algorithm supports adaptive matching of frame synchronization words at rates of 100G / 200G / 400G, and the synchronization word length is automatically adjusted to 64bit / 128bit / 256bit according to the rate, with a synchronization time ≤10μs. After automatic adjustment, an adaptive clock recovery algorithm is executed to ensure clock synchronization accuracy at different rates.

[0044] Furthermore, the adaptive sampling rate conversion process adaptively switches between 80GSps / 160GSps / 320GSps sampling rates according to the sampling rate requirements of different baseband signals; wherein, by employing a fractional sampling rate conversion algorithm, signal integrity is ensured during conversion at different rates.

[0045] Furthermore, the adaptive equalization process is performed using a multi-tap equalizer with a configurable number of taps. The number of taps is adaptively adjusted to 32 / 64 / 128 taps according to the speed, and the complexity of the multi-tap equalizer is automatically adjusted to meet the dispersion compensation requirements of different speeds.

[0046] Furthermore, the multi-rate carrier recovery processing adaptively adjusts the carrier recovery algorithm based on the identified modulation format; wherein, Frequency offset estimation: An adaptive order frequency offset estimation algorithm is adopted; for 100G signals, a fourth power algorithm is used, and for 200G / 400G signals, a higher order fourth power algorithm is used. Phase recovery: Different phase recovery algorithms are used for different modulation formats: for PM-QPSK signals, the VV phase recovery algorithm is used; for PM-16QAM signals, the partitioned VV phase recovery algorithm is used. Nonlinear compensation: For 400G signals, the nonlinear compensation coefficient is dynamically adjusted according to the signal power to compensate for phase distortion caused by fiber nonlinearity.

[0047] As a further solution, the closed-loop optimization feedback is specifically configured as follows: Real-time monitoring of error vector amplitude (EVM) and bit error rate (BER) performance metrics with a granularity of 10 μs; When judging BER>1×10 -9If EVM > 8%, perform at least one of the following optimization operations: First, adjust the step size of the adaptive equalizer; If the performance still does not meet the threshold, increase the number of iterations of the carrier recovery algorithm; For 400G signals, the nonlinear compensation coefficient is adjusted synchronously.

[0048] Example 2 This embodiment also provides a multi-rate adaptive compatible coherent demodulation device for implementing a multi-rate adaptive compatible coherent demodulation method as described in any one of Embodiments 1, comprising: Signal input and preprocessing unit: preprocesses the received coherent optical signal to obtain the preprocessed baseband signal; Rate and modulation format identification unit: automatically identifies the rate and modulation format of the preprocessed baseband signal to obtain the identification result; wherein, the automatic identification adopts a two-stage mechanism combining a coarse identification stage and a fine identification stage; Adaptive parameter configuration unit: performs adaptive parameter configuration based on the recognition result, and loads demodulation parameters that match the rate and modulation format; Multi-rate compatible demodulation unit: Based on the identification results and the configured demodulation parameters, the preprocessed baseband signal is subjected to multi-rate compatible demodulation through a unified and configurable modular architecture to obtain the demodulated standard constellation signal; The multi-rate compatible demodulation processing includes adaptive synchronization processing, adaptive sampling rate conversion processing, adaptive equalization processing, and multi-rate carrier recovery processing. Decision output and performance monitoring unit: performs decision output and performance monitoring on the demodulated standard constellation signal to obtain demodulation result output, performance index data output and performance monitoring feedback data; Closed-loop optimization feedback unit: Based on the performance monitoring feedback data, it performs closed-loop optimization feedback, dynamically fine-tunes the demodulation parameters, obtains optimized parameter feedback, and applies it to adaptive parameter configuration, adaptive equalization processing, and multi-rate carrier recovery processing.

[0049] It should be noted that: the multi-rate compatible demodulation unit in this embodiment adopts a unified modular design, and each core module supports adaptive parameter adjustment to achieve compatible demodulation of signals at different rates. The hardware logic of all modules is completely reused, with only the parameters differing, achieving 100% reuse of hardware resources; wherein, the multi-rate compatible demodulation unit includes: (1) Adaptive synchronization processing module Frame synchronization: It adopts a variable length synchronization word detection algorithm, supports adaptive matching of frame synchronization words for 100G / 200G / 400G signals, and the synchronization word length is automatically adjusted to 64bit / 128bit / 256bit according to the rate, with a synchronization time ≤10μs; Clock recovery: Adaptive clock recovery based on Gardner algorithm, sampling rate dynamically switches according to recognition results, clock recovery range ±200ppm, ensuring clock synchronization accuracy at different rates, timing error ≤0.1%.

[0050] (2) Adaptive Equalization Module The core architecture uses a multi-tap equalizer with a configurable number of taps. The number of taps is adaptively adjusted according to the speed (32 / 64 / 128 taps), and the complexity of the equalizer is automatically adjusted to meet the dispersion compensation requirements of different speeds. Hybrid precision bit width optimization: Adaptively adjusts the operational bit width of the equalizer taps to meet the precision requirements of different speeds. 100G signals use a 16-bit width. 200G uses 18bit, The 400G uses 20-bit, which reduces hardware resource consumption while ensuring demodulation accuracy. Compared with the fixed 20-bit solution, resource consumption is reduced by 40%. Algorithm switching: The 100G signal uses the basic CMA algorithm, which has a fast convergence speed and meets the requirements of low complexity. The 200G / 400G signal adopts an improved CMA+LMS joint algorithm, which balances convergence speed and compensation accuracy, and solves the problem of slow convergence of CMA algorithm under high-order modulation. Polarization demultiplexing: Supports adaptive demultiplexing of dual polarization signals, adjusts the equalizer weight update strategy according to the modulation format, adapts to polarization state changes at different rates, and has a polarization mode dispersion tolerance of ≥40ps.

[0051] (3) Multi-rate carrier recovery module Frequency offset estimation: An adaptive order frequency offset estimation algorithm is adopted, and the fourth power algorithm is used for 100G signals, which has low computational complexity. The 200G / 400G signal uses a high-order fourth power algorithm with an estimation range of ±200MHz and an accuracy of <1kHz, which meets the high-precision requirements of high-order modulation. Phase recovery: Different phase recovery algorithms are used for different modulation formats. PM-QPSK signals use the traditional VV phase recovery algorithm, which has low complexity and meets real-time requirements; The PM-16QAM signal uses an improved partitioned VV phase recovery algorithm, which divides the 16QAM constellation diagram into 4 regions, and performs phase estimation independently in each region. At the same time, a phase noise suppression module is added. Compared with the traditional VV algorithm, the phase noise tolerance is improved by 2 times, and it can support laser linewidths of up to 500kHz, solving the problem of phase noise sensitivity under high-order modulation. Nonlinear compensation: An adaptive nonlinear compensation submodule is embedded in the 400G signal to dynamically adjust the nonlinear compensation coefficient according to the signal power, thereby compensating for phase distortion caused by fiber nonlinearity and increasing the transmission distance by 20%.

[0052] (4) Adaptive sampling rate conversion module It adapts to the sampling rate requirements of different input signals and supports adaptive switching of sampling rates of 80GSps / 160GSps / 320GSps; it adopts a fractional sampling rate conversion algorithm to ensure signal integrity during different rate conversions, with a sampling rate conversion error of ≤0.1% and no signal distortion.

[0053] Furthermore, the decision output and performance monitoring unit specifically includes: Soft decision: The decision threshold is adaptively adjusted according to the modulation format. The PM-QPSK signal adopts 4-point decision and the PM-16QAM signal adopts 16-point decision. The soft decision LLR information is output, which supports soft decision FEC decoding and improves the transmission distance of the system. Performance monitoring: Real-time calculation of EVM, MER, SNR, BER and other indicators of signals at different rates, with a monitoring granularity of 10μs. The monitoring results are fed back to the closed-loop optimization module in real time for fine-tuning of demodulation parameters.

[0054] Furthermore, the closed-loop optimization feedback unit fine-tunes the tuning parameters based on performance monitoring results to optimize demodulation performance and address the issues of large performance fluctuations and slow convergence after rate switching. The optimization strategy is as follows: (1) If BER>1×10-9 or EVM>8% is detected, the equalizer step size is first adaptively adjusted. The step size starts from the initial value of 0.001 and is adjusted by 0.0002 each time until the performance index meets the threshold. (2) If the performance is still not satisfactory after adjusting the step size, adjust the number of iterations of the carrier recovery algorithm, gradually increasing it from the default 5 times to 10 times to improve the accuracy of phase recovery; (3) For 400G signals, the coefficients of nonlinear compensation are adjusted synchronously to optimize the effect of nonlinear compensation; Dynamic adjustment: After rate switching, parameters are quickly adjusted through the closed-loop optimization module to stabilize the demodulation performance in the optimal state, with optimization latency ≤50μs and performance fluctuation ≤0.5dB, which is far superior to the 2dB fluctuation and millisecond-level convergence time of existing technologies.

[0055] In practical use and verification, the system provided in this embodiment has the following technical advantages: 1. This embodiment adopts a two-stage process of coarse recognition + fine recognition, combined with a lightweight CNN classifier with only 256 parameters, to achieve a recognition accuracy of ≥99.9% under low OSNR and a recognition latency of ≤1μs, solving the problem that traditional deep learning cannot be deployed in real time; 2. This embodiment designs a hybrid precision adaptive parameter configuration strategy: it automatically adjusts core parameters such as sampling rate, equalizer order, and operation bit width according to the recognition results. The parameter mapping table can be updated online and supports new rate expansion. At the same time, through hybrid precision bit width optimization, the hardware resource consumption is reduced by 40%. 3. This embodiment adopts a unified configurable modular demodulation architecture: the hardware logic of all demodulation modules is fully reused, and different rates are adapted only by parameter configuration, thereby improving the hardware resource reuse rate by more than 70% and solving the problem of resource waste in existing multi-module solutions; 4. This embodiment innovates a partitioned multi-rate carrier recovery algorithm: a partitioned VV phase recovery algorithm is proposed for PM-16QAM signals, which improves the phase noise tolerance by 2 times, supports lasers with a linewidth of 500kHz, and improves the transmission performance of 400G signals by combining adaptive nonlinear compensation. 5. This embodiment constructs a microsecond-level fast closed-loop optimization mechanism: based on performance monitoring with a 10μs granularity, it realizes dynamic fine-tuning of parameters within 50μs, and the performance fluctuation after rate switching is ≤0.5dB, which solves the problems of large performance fluctuation and slow convergence after switching in the existing scheme; 6. This embodiment is compatible with general coherent acquisition platforms, reduces the hardware cost and resource occupancy of multi-rate demodulation systems, supports dynamic rate switching scenarios, and can be directly applied to real-time hardware platforms such as FPGA.

[0056] In summary, this embodiment employs a two-stage mechanism of coarse recognition + fine recognition, combined with a lightweight CNN classifier, to automatically identify the input signal's rate (100G / 200G / 400G) and modulation format (PM-QPSK / PM-16QAM), maintaining a recognition accuracy of ≥99.9% even with low OSNR. Based on the recognition results, it adaptively adjusts core parameters such as sampling rate, equalizer order, computational bit width, and carrier recovery algorithm, using hybrid precision optimization to reduce hardware resource consumption. Furthermore, this embodiment adopts a unified modular demodulation architecture, combined with partitioned carrier recovery and fast closed-loop optimization, to achieve high-precision demodulation of signals with different rates, with a rate switching delay ≤100μs and performance fluctuation ≤0.5dB.

[0057] Therefore, this invention has strong multi-rate compatibility, excellent automatic identification and configuration capabilities, high hardware resource reuse rate, and supports dynamic rate switching. It is suitable for general coherent acquisition platforms, intelligent optical network receivers, and multi-rate optical communication testers, and has extremely high engineering application value.

[0058] The above are only some embodiments of this application and do not limit the patent scope of this application. All equivalent structural transformations made under the technical concept of this application and using the contents of the specification and drawings of this application, or direct / indirect applications in other related technical fields, are included in the patent protection scope of this application.

Claims

1. A multi-rate adaptive compatible coherent demodulation method, applied to adaptive compatible coherent demodulation of coherent optical signals, characterized in that, Includes the following steps: The received coherent optical signal is preprocessed to obtain the preprocessed baseband signal; The preprocessed baseband signal is automatically identified in terms of rate and modulation format to obtain the identification result; wherein the automatic identification adopts a two-stage mechanism combining a coarse identification stage and a fine identification stage. Based on the identification results, adaptive parameter configuration is performed, and demodulation parameters matching the rate and modulation format are loaded; Based on the identification results and the configured demodulation parameters, the preprocessed baseband signal is subjected to multi-rate compatible demodulation through a unified and configurable modular architecture to obtain the demodulated standard constellation signal. The multi-rate compatible demodulation processing includes adaptive synchronization processing, adaptive sampling rate conversion processing, adaptive equalization processing, and multi-rate carrier recovery processing. The demodulated standard constellation signal is subjected to decision output and performance monitoring to obtain demodulation result output, performance index data output and performance monitoring feedback data. Based on the performance monitoring feedback data, closed-loop optimization feedback is performed to dynamically fine-tune the demodulation parameters, resulting in optimized parameter feedback, which is then applied to adaptive parameter configuration, adaptive equalization processing, and multi-rate carrier recovery processing. The coherent demodulation method supports at least 100G, 200G, and 400G rates, as well as PM-QPSK and PM-16QAM modulation formats for coherent optical signal compatible demodulation. The two-stage mechanism specifically includes: The constellation diagram features, symbol rate, spectral bandwidth, and frame structure features of the preprocessed baseband signal are extracted, and a multi-dimensional feature vector is constructed. Coarse identification stage: Based on the symbol rate and spectral bandwidth of the preprocessed baseband signal, rapid classification is performed to classify the preprocessed baseband signal into the corresponding rate range; Precise recognition stage: Within the corresponding rate range, the multi-dimensional feature vector is input into the trained lightweight CNN classifier, which outputs accurate recognition results of rate and modulation format. The adaptive parameter configuration step further includes: Obtain accurate rate and modulation format recognition results from a lightweight CNN classifier; Parameter combinations are dynamically loaded based on the built-in parameter mapping table; Verify and output the parameter combination corresponding to the recognition result; The specific demodulation parameters corresponding to the parameter combinations are configured based on the end-to-end demodulation parameter configuration package. The parameter mapping table supports online updates to accommodate new rates and modulation formats. The parameter combinations include: sampling rate, equalizer type, equalizer order, carrier recovery algorithm type, and equalizer operation bit width. The multi-rate carrier recovery processing adaptively adjusts the carrier recovery algorithm based on the identified modulation format. Frequency offset estimation: An adaptive order frequency offset estimation algorithm is adopted; for 100G signals, a fourth power algorithm is used, and for 200G / 400G signals, a higher order fourth power algorithm is used. Phase recovery: Different phase recovery algorithms are used for different modulation formats: for PM-QPSK signals, the VV phase recovery algorithm is used; for PM-16QAM signals, the partitioned VV phase recovery algorithm is used. Nonlinear compensation: For 400G signals, the nonlinear compensation coefficient is dynamically adjusted according to the signal power to compensate for phase distortion caused by fiber nonlinearity.

2. The multi-rate adaptive compatible coherent demodulation method according to claim 1, characterized in that, The parameter adaptive adjustment strategy of the parameter mapping table is further configured to: adaptively adjust the number of taps and the operation bit width of the equalizer according to the identified signal rate; wherein... For 100G signals, the configuration is 32 taps, 16-bit operation width, and the equalizer type adopts the basic CMA algorithm. For 200G signals, the configuration is 64 taps and 18-bit operation width. The equalizer type adopts a joint algorithm of improved CMA and Least Mean Square (LMS). For 400G signals, the configuration is 128 taps, 20-bit operation width, and the equalizer type adopts a joint algorithm of improved CMA and Least Mean Square (LMS).

3. The multi-rate adaptive compatible coherent demodulation method according to claim 1, characterized in that, The adaptive synchronization processing is performed using a variable-length synchronization word detection algorithm; wherein... The variable-length synchronization word detection algorithm supports adaptive matching of frame synchronization words at rates of 100G / 200G / 400G, and the synchronization word length is automatically adjusted to 64bit / 128bit / 256bit according to the rate, with a synchronization time ≤10μs. After automatic adjustment, an adaptive clock recovery algorithm is executed to ensure clock synchronization accuracy at different rates.

4. The multi-rate adaptive compatible coherent demodulation method according to claim 1, characterized in that, The adaptive sampling rate conversion process adaptively switches between 80GSps / 160GSps / 320GSps sampling rates according to the sampling rate requirements of different baseband signals; wherein, by employing a fractional sampling rate conversion algorithm, signal integrity is ensured during conversion at different rates.

5. The multi-rate adaptive compatible coherent demodulation method according to claim 1, characterized in that, The adaptive equalization process is performed using a multi-tap equalizer with a configurable number of taps. The number of taps is adaptively adjusted to 32 / 64 / 128 taps according to the speed, and the complexity of the multi-tap equalizer is automatically adjusted to meet the dispersion compensation requirements of different speeds.

6. The multi-rate adaptive compatible coherent demodulation method according to claim 1, characterized in that, The specific configuration of the closed-loop optimization feedback is as follows: Real-time monitoring of error vector amplitude (EVM) and bit error rate (BER) performance metrics with a granularity of 10 μs; When judging BER>1×10 -9 If EVM > 8%, perform at least one of the following optimization operations: First, adjust the step size of the adaptive equalizer; If the performance still does not meet the threshold, increase the number of iterations of the carrier recovery algorithm; For 400G signals, the nonlinear compensation coefficient is adjusted synchronously.

7. A multi-rate adaptive compatible coherent demodulation apparatus, used to implement the multi-rate adaptive compatible coherent demodulation method as described in any one of claims 1 to 6, characterized in that, include: Signal input and preprocessing unit: preprocesses the received coherent optical signal to obtain the preprocessed baseband signal; Rate and modulation format identification unit: automatically identifies the rate and modulation format of the preprocessed baseband signal to obtain the identification result; wherein, the automatic identification adopts a two-stage mechanism combining a coarse identification stage and a fine identification stage; Adaptive parameter configuration unit: performs adaptive parameter configuration based on the recognition result, and loads demodulation parameters that match the rate and modulation format; Multi-rate compatible demodulation unit: Based on the identification results and the configured demodulation parameters, the preprocessed baseband signal is subjected to multi-rate compatible demodulation through a unified and configurable modular architecture to obtain the demodulated standard constellation signal; The multi-rate compatible demodulation processing includes adaptive synchronization processing, adaptive sampling rate conversion processing, adaptive equalization processing, and multi-rate carrier recovery processing. Decision output and performance monitoring unit: performs decision output and performance monitoring on the demodulated standard constellation signal to obtain demodulation result output, performance index data output and performance monitoring feedback data; Closed-loop optimization feedback unit: Based on the performance monitoring feedback data, it performs closed-loop optimization feedback, dynamically fine-tunes the demodulation parameters, obtains optimized parameter feedback, and applies it to adaptive parameter configuration, adaptive equalization processing, and multi-rate carrier recovery processing.

Images