VGA-based 5G miniature optical transceiver module self-calibration method

Abstract

The invention discloses a VGA (Video Graphics Array)-based 5G miniature optical transceiver module self-calibration method, which belongs to the technical field of optical transceiver module self-calibration, and comprises the following steps: S1, establishing a self-calibrated benchmark reference value through an initialization module and collecting environment parameters and benchmark optical parameters; s2, dynamically collecting and receiving optical power and environment change data, and judging whether deviation exceeds a threshold value or not to trigger model retraining by collecting actual operation data of environment parameters and driving parameters and combining a neural network model prediction target value; integration of a new data set and re-fitting of a non-linear relation enable the model to continuously adapt to the influence of long-term factors; key parameters such as a gain coefficient or a temperature compensation curve of a proportional-integral-differential control algorithm are dynamically corrected, long-term precision and reliability of a self-calibration system are guaranteed, and the problem of calibration failure caused by failure of a static model is avoided.

Description

Technical Field

[0001] This invention belongs to the field of optical transceiver module self-calibration technology, specifically referring to a self-calibration method for a 5G micro optical transceiver module based on VGA. Background Technology

[0002] With the increasing demand for high speed and low latency in 5G communication, the VGA-based miniature optical transceiver module integrates environmental parameter monitoring, neural network modeling and dynamic gain adjustment technologies to achieve adaptive calibration of transmitted and received optical power, ensuring signal stability and energy efficiency optimization under complex operating conditions.

[0003] However, existing VGA-based self-calibration methods for 5G miniature optical transceivers still have certain shortcomings. These methods suffer from poor adaptability of static models, lagging control algorithms, inaccurate temperature compensation, and insufficient long-term stability. Traditional methods rely heavily on linear models, which struggle to cope with complex environmental changes. They lack dynamic gain adjustment and closed-loop feedback mechanisms, leading to signal fluctuations and increased bit error rates. Temperature compensation, lacking real-time prediction and closed-loop adjustment, is susceptible to device aging and lacks an iterative model update mechanism, resulting in continuous degradation of calibration accuracy over long-term operation. Summary of the Invention

[0004] The purpose of this invention is to provide a self-calibration method for a VGA-based 5G micro optical transceiver module, in order to solve the problems mentioned in the background art.

[0005] To achieve the above objectives, the present invention provides the following technical solution: a self-calibration method for a 5G micro optical transceiver module based on VGA, comprising the following steps:

[0006] S1. By initializing the module and collecting environmental parameters and reference light parameters, a self-calibrated reference value is established;

[0007] S2. Dynamically acquire and receive optical power and environmental change data;

[0008] S3. By inputting the baseline parameters and real-time data into the neural network model, a nonlinear mapping relationship between environmental parameters and light parameters is established.

[0009] S4. Optimize the receiver signal strength by calculating the ideal gain value based on the model output and adjusting the VGA gain to stabilize the received optical power.

[0010] S5. By combining temperature monitoring and model prediction, the laser driving parameters are adjusted to compensate for the temperature drift deviation of the emitted light power.

[0011] S6. By periodically comparing the calibrated parameters with the ideal values ​​and triggering model updates, a self-calibration strategy is implemented to adapt to long-term operational changes.

[0012] S7 optimizes overall energy consumption while ensuring performance by monitoring power consumption and dynamically adjusting VGA gain accuracy.

[0013] Preferably, in step S1, when the optical module is started, the control unit first sends an initialization command to each functional module to activate the built-in environmental sensor and power supply voltage monitoring module, and collects the current ambient temperature, humidity and power supply voltage values ​​in real time; then, the laser drive circuit is controlled to output the default operating current, and the initial emitted optical power of the transmitter and the initial received optical power of the receiver are detected by the photodetector, and the laser bias current and modulation depth parameters are read simultaneously; finally, the collected environmental parameters and optical parameters are stored in non-volatile memory as reference values ​​for subsequent calibration processes.

[0014] Preferably, in step S2, during the operation of the optical module, the data acquisition unit is activated to adjust the gain coefficient of the receiving optical signal in real time via VGA, enabling the photodetector to dynamically capture changes in the current received optical power. Simultaneously, temperature, humidity, and power supply voltage fluctuation data output by environmental sensors are collected, and the analog optical signal adjusted by VGA and the environmental parameters are converted into digital signals via an analog-to-digital converter. The control unit stores the processed digital data in a buffer group by timestamp and periodically triggers the data preprocessing process, including outlier filtering, signal smoothing, and multi-dimensional parameter alignment, ultimately forming a structured dataset containing received optical power, environmental parameters, and VGA gain status, providing real-time monitoring basis for subsequent neural network model input.

[0015] Preferably, in step S3, the baseline environmental parameters and optical parameters stored in S1 are used as input samples, combined with the real-time environmental parameters dynamically acquired in S2 and the received optical power data after VGA gain adjustment, to construct a multi-dimensional training dataset; the original data is normalized by the data preprocessing module, the temporal features of the environmental parameters and optical parameters are extracted, and the training set and validation set are divided; the preprocessed data is input into a neural network model based on a long short-term memory network, with the environmental parameters as the input layer and the optical parameters as the output layer, and the model weights are iteratively optimized through the backpropagation algorithm.

[0016] Preferably, in step S3, during model training, the VGA gain sampling frequency and data update cycle are adjusted synchronously to ensure the real-time performance and consistency of the input data, ultimately generating a neural network model that reflects the nonlinear relationship between environmental changes and optical parameter response, as shown in the formula:

[0017]

[0018] In the formula, min wThis means minimizing the loss function L by optimizing the weights w of the neural network; L represents the loss function. The standardization factor is represented by N; the total number of training samples is represented by N. This indicates time series modeling; T represents the number of time steps or the number of samples; y t This represents the actual optical parameter value at time step t; x t This represents the environmental parameters and VGA gain value at time step t; f(x) t ;W) indicates that when inputting x t The predicted optical parameter values ​​under weight W; W represents the model weights.

[0019] Preferably, in step S4, during the operation of the optical module, the control unit calls the neural network model trained in step S3, takes the currently acquired environmental parameters and the current VGA gain value as input, and the model outputs the corresponding ideal received optical power target value; subsequently, the actual optical power value measured in real time by the photodetector is compared with the target value predicted by the model, and the deviation between the two is calculated, as shown in the formula:

[0020] Where e(t) = y pred (t)-y actual (t);

[0021] In the formula, ΔG represents the VGA gain adjustment value calculated based on the deviation between the model prediction and the actual received optical power; K p The proportionality coefficient represents the weight of the direct impact of the current deviation e(t) on the gain adjustment ΔG; K i The integral coefficient represents the weight of the cumulative effect of historical deviation e(τ) on ΔG; K d The differential coefficient represents the rate of change of the deviation. Weight of the impact on ΔG; This represents the cumulative value of the deviation e(τ) from the initial time to the current time t; This represents the rate of change of the deviation e(t) over time; e(t) = y pred (t)-y actual (t) represents the ideal received optical power y predicted by the model. pred (t) and the actual measured value y actual The difference in (t).

[0022] Preferably, in step S4, based on a preset gain adjustment algorithm, combined with deviation values ​​and historical adjustment data, a new VGA gain adjustment command is dynamically generated; the control unit sends the adjustment command to the VGA driver module to adjust its gain coefficient to compensate for optical power fluctuations caused by environmental changes, and at the same time feeds back the latest VGA gain value and environmental parameters to the model input terminal; by periodically triggering model prediction and gain adjustment operations, the signal strength at the receiving end is always maintained within a preset stable range.

[0023] Preferably, in step S5, during laser operation, the temperature monitoring module collects the laser cavity and ambient temperature in real time and transmits the temperature data as input to the pre-trained neural network model. The model outputs the target value of the ideal emitted light power corresponding to the current temperature. The control unit compares the measured value of the ideal emitted light power actually emitted by the laser with the target value predicted by the model and calculates the deviation value. Based on the preset driving parameter adjustment algorithm, combined with the deviation value and historical adjustment data, the control unit dynamically generates adjustment instructions for the laser driving parameters. The driving circuit updates the bias current and modulation current values ​​of the laser according to the adjustment instructions to compensate for the drift of the ideal emitted light power caused by temperature changes. At the same time, the latest temperature data and the adjusted driving parameters are fed back to the model input.

[0024] Preferably, in step S6, during long-term system operation, the control unit periodically collects current environmental parameters and calibrated VGA gain values, laser drive parameters, and actual operating data, and calculates ideal parameter target values ​​based on the neural network model trained in step S3; by comparing the deviation between the actual values ​​and the target values, it determines whether the deviation exceeds a preset threshold range; if the deviation exceeds the threshold, the model update process is triggered, integrating the latest collected data with historical calibration data and inputting it as a new dataset into the model training module to refit the nonlinear relationship between environmental parameters and optical power; simultaneously, based on the model update results, the key parameters in the self-calibration strategy are dynamically corrected to adapt to changes in system characteristics caused by long-term factors such as device aging and environmental drift; the updated model and strategy parameters are synchronized to the control unit, replacing the original version, forming a closed-loop iterative mechanism.

[0025] Preferably, in step S7, during the operation of the optical module, the control unit collects system power consumption data in real time through a current sensor or power meter, and determines the current operating status based on a preset power consumption-performance mapping relationship; dynamically adjusts the adjustment accuracy of the VGA gain according to the power consumption status, reducing unnecessary high-frequency gain adjustment operations while ensuring the stability of the received optical power; simultaneously, the adjusted VGA gain setting is sent to the driver module, and the actual change trend of power consumption and received optical power is continuously monitored; if the performance index is detected to be close to the critical threshold, the high-precision adjustment mode of the VGA gain is restored first to ensure communication quality; by periodically evaluating the balance between power consumption and performance, the precision control parameters in the gain adjustment strategy are dynamically updated to achieve a dynamic trade-off between power consumption optimization and signal quality maintenance.

[0026] Compared with the prior art, the beneficial effects of the present invention are:

[0027] 1. This invention significantly improves the model's adaptability to complex environmental changes by constructing a multi-dimensional training dataset of baseline parameters and real-time data, and by using a long short-term memory network to model the nonlinear relationship between environmental parameters and optical parameters. Data normalization and temporal feature extraction ensure the standardization of model input, while the division of training and validation sets enhances the model's generalization performance. The generated neural network model can accurately capture the dynamic correlation between environmental parameters and optical power response, providing a high-precision prediction basis for subsequent ideal gain calculation, transmitted optical power compensation, and self-calibration strategy optimization, thereby achieving stable operation of the optical module under complex working conditions.

[0028] 2. This invention effectively stabilizes the signal strength at the receiving end by dynamically adjusting the VGA gain using an ideal target value of received optical power output based on a neural network model and combined with a proportional-integral-derivative control algorithm. By comparing the model prediction value with the actual measured optical power value in real time, the deviation is calculated and a gain adjustment command is generated, achieving the triple control objectives of rapid response to current errors, elimination of historical accumulated deviations, and suppression of dynamic fluctuations. The periodically triggered model prediction and gain adjustment operations form a closed-loop feedback mechanism to ensure that the received optical power is always maintained within the preset range, thereby improving the robustness of communication quality and reducing the risk of signal distortion or increased bit error rate caused by environmental disturbances.

[0029] 3. This invention combines temperature monitoring and model prediction to dynamically adjust laser driving parameters, which can accurately compensate for temperature drift deviations in emitted optical power. By collecting real-time data on the laser cavity and ambient temperature, the pre-trained neural network model is used to predict the ideal target value of emitted optical power, and the deviation signal is generated by comparing it with the actual measured value. The driving parameter adjustment algorithm dynamically optimizes the bias current and modulation current based on the deviation and historical data, effectively addressing optical power fluctuations caused by temperature changes. A closed-loop feedback mechanism ensures the stability of the transmitter, extends device lifespan, and reduces the risk of performance degradation due to temperature drift.

[0030] 4. This invention ensures that the self-calibration strategy adapts to changes in system characteristics over the long term by periodically comparing calibrated parameters with ideal values ​​and triggering model updates; by collecting actual operating data of environmental and driving parameters, and combining the target value predicted by the neural network model, it determines whether the deviation exceeds the threshold to trigger model retraining; the integration of new datasets and the refitting of nonlinear relationships enable the model to continuously adapt to the effects of long-term factors such as component aging and environmental drift; dynamic correction of key parameters such as the gain coefficient of the proportional-integral-derivative control algorithm or the temperature compensation curve forms a closed-loop iterative mechanism, ensuring the long-term accuracy and reliability of the self-calibration system and avoiding calibration failure problems caused by static model failure. Attached Figure Description

[0031] Figure 1 The following is the operational flow of the self-calibration method for a VGA-based 5G micro optical transceiver module according to the present invention. Figure 1 ;

[0032] Figure 2 The following is the operational flow of the self-calibration method for a VGA-based 5G micro optical transceiver module according to the present invention. Figure 2 ;

[0033] Figure 3 The following is the operational flow of the self-calibration method for a VGA-based 5G micro optical transceiver module according to the present invention. Figure 3 ;

[0034] Figure 4 The following is the operational flow of the self-calibration method for a VGA-based 5G micro optical transceiver module according to the present invention. Figure 4 . Detailed Implementation

[0035] 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. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0036] Example

[0037] Please see Figure 1-4 As shown, the present invention provides a technical solution comprising the following steps:

[0038] S1. By initializing the module and collecting environmental parameters and reference light parameters, a self-calibrated reference value is established;

[0039] S2. Dynamically acquire and receive optical power and environmental change data;

[0040] S3. By inputting the baseline parameters and real-time data into the neural network model, a nonlinear mapping relationship between environmental parameters and light parameters is established.

[0041] S4. Optimize the receiver signal strength by calculating the ideal gain value based on the model output and adjusting the VGA gain to stabilize the received optical power.

[0042] S5. By combining temperature monitoring and model prediction, the laser driving parameters are adjusted to compensate for the temperature drift deviation of the emitted light power.

[0043] S6. By periodically comparing the calibrated parameters with the ideal values ​​and triggering model updates, a self-calibration strategy is implemented to adapt to long-term operational changes.

[0044] S7 optimizes overall energy consumption while ensuring performance by monitoring power consumption and dynamically adjusting VGA gain accuracy.

[0045] Preferably, in step S1, when the optical module is started, the control unit first sends an initialization command to each functional module to activate the built-in environmental sensor and power supply voltage monitoring module, and collects the current ambient temperature, humidity and power supply voltage values ​​in real time; then, the laser drive circuit is controlled to output the default operating current, and the initial emitted optical power of the transmitter and the initial received optical power of the receiver are detected by the photodetector, and the laser bias current and modulation depth parameters are read simultaneously; finally, the collected environmental parameters and optical parameters are stored in non-volatile memory as reference values ​​for subsequent calibration processes.

[0046] Preferably, in step S2, during the operation of the optical module, the data acquisition unit is activated to adjust the gain coefficient of the receiving optical signal in real time via VGA, enabling the photodetector to dynamically capture changes in the current received optical power. Simultaneously, temperature, humidity, and power supply voltage fluctuation data output by environmental sensors are collected, and the analog optical signal adjusted by VGA and the environmental parameters are converted into digital signals via an analog-to-digital converter. The control unit stores the processed digital data in a buffer group by timestamp and periodically triggers the data preprocessing process, including outlier filtering, signal smoothing, and multi-dimensional parameter alignment, ultimately forming a structured dataset containing received optical power, environmental parameters, and VGA gain status, providing real-time monitoring basis for subsequent neural network model input.

[0047] Preferably, in step S3, the baseline environmental parameters and optical parameters stored in S1 are used as input samples, combined with the real-time environmental parameters dynamically acquired in S2 and the received optical power data after VGA gain adjustment, to construct a multi-dimensional training dataset; the original data is normalized by the data preprocessing module, the temporal features of the environmental parameters and optical parameters are extracted, and the training set and validation set are divided; the preprocessed data is input into a neural network model based on a long short-term memory network, with the environmental parameters as the input layer and the optical parameters as the output layer, and the model weights are iteratively optimized through the backpropagation algorithm.

[0048] Preferably, in step S3, during model training, the VGA gain sampling frequency and data update cycle are adjusted synchronously to ensure the real-time performance and consistency of the input data, ultimately generating a neural network model that reflects the nonlinear relationship between environmental changes and optical parameter response, as shown in the formula:

[0049]

[0050] In the formula, min w This means minimizing the loss function L by optimizing the weights w of the neural network; L represents the loss function. The standardization factor is represented by N; the total number of training samples is represented by N. This indicates time series modeling; T represents the number of time steps or the number of samples; y t This represents the actual optical parameter value at time step t; x t This represents the environmental parameters and VGA gain value at time step t; f(x) t ;W) indicates that when inputting x t The predicted optical parameter values ​​under weight W; W represents the model weights.

[0051] Preferably, in step S4, during the operation of the optical module, the control unit calls the neural network model trained in step S3, takes the currently acquired environmental parameters and the current VGA gain value as input, and the model outputs the corresponding ideal received optical power target value; subsequently, the actual optical power value measured in real time by the photodetector is compared with the target value predicted by the model, and the deviation between the two is calculated, as shown in the formula:

[0052] Where e(t) = y pred (t)-y actual (t);

[0053] In the formula, ΔG represents the VGA gain adjustment value calculated based on the deviation between the model prediction and the actual received optical power; K p The proportionality coefficient represents the weight of the direct impact of the current deviation e(t) on the gain adjustment ΔG; K i The integral coefficient represents the weight of the cumulative effect of historical deviation e(τ) on ΔG; K d The differential coefficient represents the rate of change of the deviation. Weight of the impact on ΔG; This represents the cumulative value of the deviation e(τ) from the initial time to the current time t; This represents the rate of change of the deviation e(t) over time; e(t) = y pred (t)-y actual (t) represents the ideal received optical power y predicted by the model. pred (t) and the actual measured value y actual The difference in (t).

[0054] Preferably, in step S4, based on a preset gain adjustment algorithm, combined with deviation values ​​and historical adjustment data, a new VGA gain adjustment command is dynamically generated; the control unit sends the adjustment command to the VGA driver module to adjust its gain coefficient to compensate for optical power fluctuations caused by environmental changes, and at the same time feeds back the latest VGA gain value and environmental parameters to the model input terminal; by periodically triggering model prediction and gain adjustment operations, the signal strength at the receiving end is always maintained within a preset stable range.

[0055] Preferably, in step S5, during laser operation, the temperature monitoring module collects the laser cavity and ambient temperature in real time and transmits the temperature data as input to the pre-trained neural network model. The model outputs the target value of the ideal emitted light power corresponding to the current temperature. The control unit compares the measured value of the ideal emitted light power actually emitted by the laser with the target value predicted by the model and calculates the deviation value. Based on the preset driving parameter adjustment algorithm, combined with the deviation value and historical adjustment data, the control unit dynamically generates adjustment instructions for the laser driving parameters. The driving circuit updates the bias current and modulation current values ​​of the laser according to the adjustment instructions to compensate for the drift of the ideal emitted light power caused by temperature changes. At the same time, the latest temperature data and the adjusted driving parameters are fed back to the model input.

[0056] Preferably, in step S6, during long-term system operation, the control unit periodically collects current environmental parameters and calibrated VGA gain values, laser drive parameters, and actual operating data, and calculates ideal parameter target values ​​based on the neural network model trained in step S3; by comparing the deviation between the actual values ​​and the target values, it determines whether the deviation exceeds a preset threshold range; if the deviation exceeds the threshold, the model update process is triggered, integrating the latest collected data with historical calibration data and inputting it as a new dataset into the model training module to refit the nonlinear relationship between environmental parameters and optical power; simultaneously, based on the model update results, the key parameters in the self-calibration strategy are dynamically corrected to adapt to changes in system characteristics caused by long-term factors such as device aging and environmental drift; the updated model and strategy parameters are synchronized to the control unit, replacing the original version, forming a closed-loop iterative mechanism.

[0057] Preferably, in step S7, during the operation of the optical module, the control unit collects system power consumption data in real time through a current sensor or power meter, and determines the current operating status based on a preset power consumption-performance mapping relationship; dynamically adjusts the adjustment accuracy of the VGA gain according to the power consumption status, reducing unnecessary high-frequency gain adjustment operations while ensuring the stability of the received optical power; simultaneously, the adjusted VGA gain setting is sent to the driver module, and the actual change trend of power consumption and received optical power is continuously monitored; if the performance index is detected to be close to the critical threshold, the high-precision adjustment mode of the VGA gain is restored first to ensure communication quality; by periodically evaluating the balance between power consumption and performance, the precision control parameters in the gain adjustment strategy are dynamically updated to achieve a dynamic trade-off between power consumption optimization and signal quality maintenance.

[0058] Working Principle: First, upon startup, the module is initialized and environmental and reference optical parameters are collected to establish a calibration benchmark. During operation, the received optical power, environmental changes, and power supply status are dynamically monitored, and the VGA gain is adjusted in real time to stabilize the signal strength. A neural network model is trained to establish a nonlinear mapping relationship between environmental and optical parameters, and a PID algorithm is used to dynamically correct the VGA gain to compensate for optical power deviations caused by environmental fluctuations. Simultaneously, the temperature monitoring module adjusts the laser drive parameters based on model predictions to compensate for transmitter temperature drift. The system periodically triggers model updates, integrates new data to refit the relationship to adapt to long-term changes, and dynamically adjusts the VGA gain accuracy through power consumption monitoring. This optimizes energy consumption while ensuring performance, ultimately forming a closed-loop iterative mechanism to ensure the long-term stable operation of the optical module in complex environments.

[0059] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their likenesses.

[0060] The present invention and its embodiments have been described above. This description is not restrictive, and the accompanying drawings are only one embodiment of the present invention; the actual structure is not limited thereto. In conclusion, if those skilled in the art are inspired by this description and design similar structures and embodiments without departing from the spirit of the invention, such designs should fall within the protection scope of the present invention.

Claims

1. A self-calibration method for a 5G micro optical transceiver module based on VGA, characterized in that, Includes the following steps: S1. By initializing the module and collecting environmental parameters and reference light parameters, a self-calibrated reference value is established; S2. Dynamically acquire and receive optical power and environmental change data; S3. By inputting the baseline parameters and real-time data into the neural network model, a nonlinear mapping relationship between environmental parameters and light parameters is established. S4. Optimize the receiver signal strength by calculating the ideal gain value based on the model output and adjusting the VGA gain to stabilize the received optical power. S5. By combining temperature monitoring and model prediction, the laser driving parameters are adjusted to compensate for the temperature drift deviation of the emitted light power. S6. By periodically comparing the calibrated parameters with the ideal values ​​and triggering model updates, a self-calibration strategy is implemented to adapt to long-term operational changes. S7 optimizes overall energy consumption while ensuring performance by monitoring power consumption and dynamically adjusting VGA gain accuracy.

2. The self-calibration method for a VGA-based 5G micro optical transceiver module according to claim 1, characterized in that: When the optical module is started, S1 first sends an initialization command to each functional module through the control unit to activate the built-in environmental sensor and power supply voltage monitoring module, and collects the current ambient temperature, humidity and power supply voltage values ​​in real time. Subsequently, the laser drive circuit is controlled to output the default operating current, and the initial emitted optical power at the transmitter and the initial received optical power at the receiver are detected by a photodetector. The laser bias current and modulation depth parameters are read simultaneously. Finally, the collected environmental and optical parameters are stored in a non-volatile memory as reference values ​​for subsequent calibration processes.

3. The self-calibration method for a VGA-based 5G micro optical transceiver module according to claim 1, characterized in that: In step S2, during the operation of the optical module, the data acquisition unit is activated to adjust the gain coefficient of the receiving optical signal in real time through VGA, enabling the photodetector to dynamically capture changes in the current received optical power. Simultaneously, temperature, humidity, and power supply voltage fluctuation data output by the environmental sensor are collected, and the analog optical signal adjusted by VGA and the environmental parameters are converted into digital signals through the analog-to-digital conversion module. The control unit stores the processed digital data in a buffer according to timestamps and periodically triggers the data preprocessing process to ultimately form a structured dataset containing received optical power, environmental parameters, and VGA gain status.

4. The self-calibration method for a VGA-based 5G micro optical transceiver module according to claim 1, characterized in that: In step S3, the reference environmental parameters and optical parameters stored in S1 are used as input samples, and combined with the real-time environmental parameters dynamically acquired in S2 and the received optical power data after VGA gain adjustment, a multi-dimensional training dataset is constructed. The data preprocessing module normalizes the raw data, extracts the temporal features of environmental and optical parameters, and divides the data into training and validation sets. The preprocessed data is input into a neural network model based on a long short-term memory network, with environmental parameters as the input layer and light parameters as the output layer. The model weights are iteratively optimized through the backpropagation algorithm.

5. The self-calibration method for a VGA-based 5G micro optical transceiver module according to claim 4, characterized in that: In step S3, during model training, the gain sampling frequency and data update cycle of the VGA are adjusted synchronously to ensure the real-time performance and consistency of the input data. This ultimately generates a neural network model that reflects the nonlinear relationship between environmental changes and optical parameter responses, as shown in the formula: In the formula, min w This means minimizing the loss function L by optimizing the weights w of the neural network; L represents the loss function. The standardization factor is represented by N; the total number of training samples is represented by N. This indicates time series modeling; T represents the number of time steps or the number of samples; y t This represents the actual optical parameter value at time step t; x t This represents the environmental parameters and VGA gain value at time step t; f(x) t ;W) indicates that when inputting x t The predicted optical parameter values ​​under weight W; W represents the model weights.

6. The self-calibration method for a VGA-based 5G micro optical transceiver module according to claim 1, characterized in that: In step S4, during the operation of the optical module, the control unit calls the neural network model trained in step S3, taking the currently acquired environmental parameters and the current VGA gain value as inputs, and the model outputs the corresponding ideal received optical power target value; subsequently, the actual optical power value measured in real time by the photodetector is compared with the target value predicted by the model, and the deviation between the two is calculated using the following formula: where e(t) = y pred (t) - y actual (t); In the formula, ΔG represents the VGA gain adjustment value calculated based on the deviation between the model prediction and the actual received optical power; K p The proportionality coefficient represents the weight of the direct impact of the current deviation e(t) on the gain adjustment ΔG; K i The integral coefficient represents the weight of the cumulative effect of historical deviation e(τ) on ΔG; K d The differential coefficient represents the rate of change of the deviation. Weight of the impact on ΔG; This represents the cumulative value of the deviation e(τ) from the initial time to the current time t; This represents the rate of change of the deviation e(t) over time; e(t) = y pred (t)-y actual (t) represents the ideal received optical power y predicted by the model. pred (t) and the actual measured value y actual The difference in (t).

7. The self-calibration method for a VGA-based 5G micro optical transceiver module according to claim 6, characterized in that: S4, based on a preset gain adjustment algorithm, combines deviation values ​​and historical adjustment data to dynamically generate new VGA gain adjustment commands; the control unit sends the adjustment commands to the VGA driver module to adjust its gain coefficient to compensate for optical power fluctuations caused by environmental changes, and at the same time feeds back the latest VGA gain value and environmental parameters to the model input terminal; by periodically triggering model prediction and gain adjustment operations, the signal strength at the receiving end is always maintained within a preset stable range.

8. The self-calibration method for a VGA-based 5G micro optical transceiver module according to claim 1, characterized in that: In step S5, during laser operation, the temperature monitoring module collects the laser cavity and ambient temperature in real time and transmits the temperature data as input to the pre-trained neural network model. The model outputs the target value of the ideal emitted light power corresponding to the current temperature. The control unit compares the measured value of the ideal emitted light power actually emitted by the laser with the target value predicted by the model and calculates the deviation value. Based on the preset driving parameter adjustment algorithm, combined with the deviation value and historical adjustment data, the control unit dynamically generates adjustment instructions for the laser driving parameters. The driving circuit updates the bias current and modulation current values ​​of the laser according to the adjustment command to compensate for the drift of the ideal emitted light power caused by temperature changes; at the same time, it feeds back the latest temperature data and the adjusted driving parameters to the model input.

9. The self-calibration method for a VGA-based 5G micro optical transceiver module according to claim 1, characterized in that: In the long-term operation of the system, the control unit periodically collects the current environmental parameters, the calibrated VGA gain value, and the actual operating data of the laser drive parameters, and calculates the target value of the ideal parameters based on the neural network model trained in S3. By comparing the deviation between the actual value and the target value, it is determined whether the value exceeds the preset threshold range. If the deviation exceeds the threshold, the model update process is triggered, integrating the latest collected data with the historical calibration data and inputting it into the model training module as a new dataset to refit the nonlinear relationship between environmental parameters and optical power. Meanwhile, based on the model update results, key parameters in the self-calibration strategy are dynamically corrected to adapt to changes in system characteristics caused by long-term factors; the updated model and strategy parameters are synchronized to the control unit to replace the original version, forming a closed-loop iterative mechanism.

10. The self-calibration method for a VGA-based 5G micro optical transceiver module according to claim 1, characterized in that: In step S7, during the operation of the optical module, the control unit collects system power consumption data in real time through a current sensor or power meter, and determines the current operating status based on a preset power consumption-performance mapping relationship; dynamically adjusts the adjustment accuracy of VGA gain according to the power consumption status, reducing unnecessary high-frequency gain adjustment operations while ensuring the stability of received optical power; simultaneously, the adjusted VGA gain setting is sent to the driver module, and the actual change trend of power consumption and received optical power is continuously monitored; if the performance index is detected to be close to the critical threshold, the high-precision adjustment mode of VGA gain is restored first to ensure communication quality; by periodically evaluating the balance between power consumption and performance, the precision control parameters in the gain adjustment strategy are dynamically updated to achieve a dynamic trade-off between power consumption optimization and signal quality maintenance.

Images