High-speed silicon light engine method and device for silicon light module

Abstract

The invention relates to the related technical field of silicon photon chips, in particular to a high-speed silicon optical engine method and device for a silicon optical module, and the method comprises the steps: obtaining basic technical parameters of an optical fiber connection product with the silicon optical module; the refractive index distribution and the waveguide section size of the silicon optical waveguide are configured, and a multi-mode waveguide array is constructed; an electro-optical modulator array is integrated in a multi-mode waveguide array, and the coupling efficiency of waveguides and optical fibers is adjusted in real time through dynamic impedance matching; meanwhile, the high-speed electric drive circuit is used for driving the electro-optical modulator array and optimizing rising edge characteristics / falling edge characteristics of signals, and power consumption adjustment is carried out on the high-speed electric drive circuit. The technical problem that the optical signal quality is reduced due to the fact that the silicon optical waveguide design mostly adopts fixed parameters and mode crosstalk is prone to occurring in a high-speed transmission scene is solved, and the technical effects that the waveguide refractive index is dynamically configured, the multi-mode waveguide array is constructed, the mode crosstalk is effectively reduced, the transmission loss is reduced, and the optical signal transmission quality is improved are achieved.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of silicon photonics chip, and particularly relates to a high-speed silicon light engine method and device for a silicon optical module. BACKGROUND

[0002] As a core component of the silicon optical module, the performance of the silicon light engine directly determines the modulation, transmission efficiency and system stability of the optical signal. However, in the current silicon light engine technology, the silicon optical waveguide design mostly adopts fixed parameters, which is difficult to adapt to the optical transmission demand in different scenarios, and has the problems of large mode crosstalk and high transmission loss. In the high-speed transmission above 400 Gbps, the optical signal distortion is serious, which leads to a significant increase in the bit error rate. In addition, the high-frequency signal attenuation and impedance mismatch of the high-speed electrical driving circuit make the chip power consumption too high and the working stability poor, which seriously restricts the reliability of the silicon optical module in the high-speed scenario. SUMMARY

[0003] The present application provides a high-speed silicon light engine method and device for a silicon optical module, aiming to solve the technical problem that the silicon optical waveguide design in the prior art mostly adopts fixed parameters, which easily causes mode crosstalk in the high-speed transmission scenario, resulting in a decrease in the quality of the optical signal.

[0004] In view of the above problems, the technical scheme of the present application is as follows:

[0005] In a first aspect, the present application provides a high-speed silicon light engine method for a silicon optical module, wherein the method comprises: acquiring basic technical parameters of a fiber connection product with a silicon optical module; configuring the refractive index distribution and the waveguide cross-sectional size of the silicon optical waveguide based on the basic technical parameters, constructing a multimode waveguide array; integrating an electro-optic modulator array in the multimode waveguide array, adjusting the coupling efficiency of the waveguide and the optical fiber in real time through dynamic impedance matching; simultaneously, driving the electro-optic modulator array with a high-speed electrical driving circuit and optimizing the signal rising edge characteristic and the falling edge characteristic, constructing a thermal-electric-optical coupling model to adjust the power consumption of the high-speed electrical driving circuit.

[0006] Preferably, a common mode choke is integrated at the output end of the high-speed electrical driving circuit to compensate for high-frequency signal attenuation through dynamic pre-emphasis.

[0007] Preferably, a MEMS adjustable optical grating is integrated at the output end of the waveguide to monitor the reflected light power spectrum in real time; based on the reflected light power spectrum, the reflected phase noise is demodulated to trigger the piezoelectric ceramic micro displacement adjustment.

[0008] Preferably, the phase noise integral value is acquired, and when the phase noise integral value exceeds a preset integral threshold, a piezoelectric ceramic displacement adjustment instruction is triggered; the piezoelectric ceramic displacement adjustment instruction is used to control the waveguide end face curvature radius adjustment accuracy within the tolerance standard.

[0009] Preferably, a graphene-based micro heat pipe array is integrated below the electro-optical modulator, an infrared thermal imaging sensor is deployed, and the chip junction temperature is fed back in real time to dynamic power consumption adjustment; when the chip junction temperature exceeds the preset upper temperature limit, a gradient frequency reduction strategy is triggered, and redundant heat is directed out through the heat pipe.

[0010] Preferably, the heat pipe operating mode is dynamically adjusted according to the chip junction temperature, and when the chip junction temperature is in a temperature range starting from 80% of the preset upper temperature limit, the heat pipe capillary action is strengthened to dissipate heat.

[0011] Preferably, a waveguide-fiber-drive circuit joint calibration module is constructed, a 4D parameter compensation matrix is generated using a wavelength sensitive matrix tester, and based on the 4D parameter compensation matrix, the modulator bias voltage and the fiber input angle are dynamically optimized through federated learning.

[0012] Preferably, local data is uploaded to a cloud platform for global aggregation and update, and after global aggregation and update, the modulator bias voltage and the fiber input angle are output.

[0013] Preferably, based on an edge computing node, a lightweight federated learning agent is deployed, and random noise is injected into local data to formulate an adversarial network.

[0014] In a second aspect, the application provides a high-speed silicon light engine device for a silicon light module, wherein the device comprises: a parameter acquisition module for acquiring basic technical parameters of a fiber connection product with a silicon light module; a multi-modal waveguide array construction module for configuring the refractive index distribution and waveguide cross-sectional size of a silicon light waveguide based on the basic technical parameters, and constructing a multi-modal waveguide array; an impedance matching module for integrating an electro-optical modulator array in the multi-modal waveguide array, and adjusting the coupling efficiency of the waveguide and the fiber in real time through dynamic impedance matching; and a power consumption adjustment module for driving the electro-optical modulator array with a high-speed electrical drive circuit and optimizing the signal rising edge characteristic and falling edge characteristic, and constructing a thermal-electric-optical coupling model to adjust the power consumption of the high-speed electrical drive circuit.

[0015] In summary, one or more technical solutions provided in the application achieve the technical effect of dynamically configuring waveguide refractive index, constructing a multi-modal waveguide array to effectively reduce mode crosstalk, reduce transmission loss, and improve optical signal transmission quality.

[0016] The above description is only a summary of the technical solutions of the application. In order to more clearly understand the technical means of the application, the specific embodiments of the application can be implemented in accordance with the content of the description, and in order to make the above and other purposes, features and advantages of the application more obvious and easy to understand, the following specific embodiments of the application are described. BRIEF DESCRIPTION OF DRAWINGS

[0017] In order to more clearly illustrate the technical solutions in the embodiments of the present application, the drawings needed in the embodiment description will be briefly introduced. Obviously, the drawings in the following description are only some embodiments of the present application, and other drawings can be obtained by those skilled in the art without creative effort.

[0018] Figure 1 A flowchart of a high-speed silicon optical engine method for a silicon optical module in an embodiment.

[0019] Figure 2 A high-speed silicon optical engine device architecture diagram for a silicon optical module in an embodiment.

[0020] Reference signs: parameter acquisition module 11, multimodal waveguide array construction module 12, impedance matching module 13, power consumption adjustment module 14. DETAILED DESCRIPTION

[0021] The embodiments of the present application provide a high-speed silicon optical engine method and device for a silicon optical module, which solves the technical problem that existing silicon optical waveguide designs mostly use fixed parameters, and mode crosstalk easily occurs in high-speed transmission scenarios, resulting in a decline in optical signal quality.

[0022] The technical solutions in the embodiments of the present application will be described clearly and completely in combination with the drawings in the embodiments of the present application. Obviously, the described embodiments are only some of the embodiments of the present application, not all. Based on the embodiments in the present application, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present application.

[0023] It should be noted that the terms "comprising" and "having" and any variations thereof are intended to cover non-exclusive inclusion, for example, a process, method, device, product or server including a series of steps or units does not have to be limited to those clearly listed steps or units, but can include other steps or modules that are not clearly listed or inherent to the process, method, product or device.

[0024] Embodiment one, the present application will be described in detail in combination with the drawings, as shown in Figure 1 The present application provides a high-speed silicon optical engine method for a silicon optical module, which comprises:

[0025] Obtaining the basic technical parameters of a fiber connection product with a silicon optical module; based on the basic technical parameters, configuring the refractive index distribution and waveguide cross-section size of the silicon optical waveguide, and constructing a multimodal waveguide array.

[0026] Specifically, the basic technical parameters of the fiber connection product include the core diameter, the cladding diameter, the numerical aperture, the cutoff wavelength, the transmission loss, etc. of the optical fiber. The refractive index distribution of the silicon optical waveguide refers to the change of the refractive index of the silicon material in the waveguide with the position. The refractive index of silicon is about 3.48, and the refractive index of the silica cladding is about 1.44. The large refractive index difference enables the silicon optical waveguide to effectively confine the optical field, but also increases the complexity of the design. The waveguide cross-sectional size refers to the width and thickness of the waveguide, which has an important influence on the number of modes supported by the waveguide and the mode characteristics. For example, the width of the waveguide is generally between a few hundred nanometers and a few microns. A narrower waveguide generally supports single-mode operation, while a wider waveguide can support multi-mode operation. The multi-mode waveguide array refers to an array composed of multiple waveguides with different mode characteristics. By reasonably designing the parameters of each waveguide, a variety of modes of light can be transmitted in the array, thereby increasing the capacity and flexibility of information transmission.

[0027] In one embodiment, the basic technical parameters of the fiber connection product are obtained, and the refractive index distribution and the waveguide cross-sectional size of the silicon optical waveguide are configured based on the basic technical parameters of the fiber connection product with a silicon optical module. A multi-mode waveguide array is constructed. By precisely controlling the geometry and refractive index distribution of the waveguide, the mode crosstalk can be effectively reduced, and the transmission quality of the optical signal can be improved. For example, when the waveguide width is 1.2 microns, the silicon optical waveguide can support multiple mode transmissions. These modes are completely overlapped in space but have orthogonality, so they can maintain their independence when transmitted in a straight waveguide and avoid mutual coupling crosstalk.

[0028] An electro-optical modulator array is integrated in the multi-mode waveguide array. The coupling efficiency of the waveguide and the optical fiber is adjusted in real time through dynamic impedance matching. At the same time, a high-speed electrical driving circuit is used to drive the electro-optical modulator array and optimize the signal rising edge characteristics / falling edge characteristics. A thermal-electric-optical coupling model is constructed to adjust the power consumption of the high-speed electrical driving circuit.

[0029] Specifically, the electro-optical modulator array refers to an array composed of multiple modulators capable of converting electrical signals into optical signals, each modulator can independently control the amplitude, phase or frequency of the optical signal; the purpose of dynamic impedance matching is to reduce signal reflection, improve transmission efficiency, and ensure good coupling effect between waveguide and optical fiber, the coupling efficiency refers to the efficiency of coupling optical signals from waveguide into optical fiber; the high-speed electrical driving circuit refers to a circuit capable of quickly providing electrical signals to drive the electro-optical modulator array to work, its performance directly affects the modulation speed and quality of optical signals; the signal rising edge characteristic / falling edge characteristic refers to the time required for the signal to jump from low level to high level or from high level to low level, which has important influence on the bandwidth and rate of optical signal transmission; the thermo-electro-optical coupling model is a model that comprehensively considers the interaction of thermal, electrical and optical fields, which is used to analyze and adjust the power consumption of high-speed electrical driving circuit to achieve dynamic balance between power consumption and temperature.

[0030] In one embodiment, the electro-optical modulator array is integrated in the multi-mode waveguide array, so that each waveguide can independently modulate the optical signal, thereby realizing parallel processing of the optical signal and improving the transmission efficiency and capacity of the optical signal. Through the dynamic impedance matching technology, the coupling efficiency of the waveguide and the optical fiber is adjusted in real time, which can ensure that the optical signals of different wavelengths and modes can be efficiently coupled from the waveguide to the optical fiber, effectively improving the performance of the optical link, for example, at a transmission rate of 100Gbps, the dynamic impedance matching technology can increase the coupling efficiency by about 20% and reduce the loss of optical signals by about 1.5dB.

[0031] At the same time, the high-speed electrical driving circuit drives the electro-optical modulator array and optimizes the signal rising edge and falling edge characteristics, specifically, the signal rising time can be shortened and the falling time can be shortened, thereby improving the modulation rate and bandwidth of the optical signal to meet the demand of high-speed optical communication; the thermo-electro-optical coupling model is constructed to adjust the power consumption of the high-speed electrical driving circuit, which can dynamically adjust the working state of the circuit according to the power consumption and heat generated by the electrical driving circuit, and realize the balance between power consumption and temperature, for example, when the power of the electrical driving circuit is increased, the chip temperature is maintained within a certain range through the adjustment of the thermo-electro-optical coupling model, thereby improving the stability and reliability of the circuit, and further improving the performance and efficiency of the entire silicon optical engine.

[0032] Further, the method further comprises:

[0033] A common mode choke coil is integrated at the output end of the high-speed electrical driving circuit to compensate for high-frequency signal attenuation through dynamic pre-emphasis.

[0034] Specifically, the common-mode choke coil is used for suppressing the common-mode noise of high-frequency signals. Its working principle is that the inductance of the coil produces a large inductive reactance to the high-frequency signals, thereby suppressing the transmission of the common-mode noise, while having little effect on the differential-mode signals. The common-mode choke coil is usually composed of two coils wound on a magnetic core. When the common-mode current flows through, the magnetic fields generated by the two coils superimpose each other, forming a larger inductive reactance, thereby suppressing the common-mode current. When the differential-mode current flows through, the magnetic fields generated by the two coils cancel each other out, having little effect on the differential-mode signals. Dynamic pre-emphasis is a signal compensation method that appropriately amplifies or adjusts the high-frequency signals at the signal sending end to compensate for the amplitude drop of the signals caused by high-frequency attenuation during transmission, thereby improving the quality and transmission distance of the signals. High-frequency signal attenuation refers to the phenomenon that the amplitude of the high-frequency component of the signal gradually decreases during high-speed transmission due to factors such as transmission line loss and reflection, which affects the integrity and transmission efficiency of the signal.

[0035] In one embodiment, a common-mode choke coil is integrated at the output end of the high-speed electrical driving circuit, effectively suppressing the common-mode noise of high-frequency signals, improving the anti-interference ability of the signals, and ensuring that the driving signals received by the electro-optical modulator array have high quality and stability. Through dynamic pre-emphasis technology, the amplitude and phase of the signals are adjusted in real time according to the frequency characteristics and transmission characteristics of the signals, compensating for the attenuation of high-frequency signals during transmission, so that the signals maintain good rising and falling edge characteristics when reaching the electro-optical modulator.

[0036] Optionally, in a high-speed optical communication system, when the signal rate is as high as 400 Gbps, the amplitude of the signal at the end of the transmission line without dynamic pre-emphasis technology may decrease by more than 30%, while after using dynamic pre-emphasis technology, the decrease in signal amplitude can be controlled within 10%, thereby significantly improving the quality and transmission efficiency of the signals, and ensuring the high-speed and stable transmission of optical signals.

[0037] Further, the present application provides a method for real-time adjustment of the coupling efficiency of waveguides and optical fibers through dynamic impedance matching, which comprises:

[0038] A MEMS adjustable grating is integrated at the output end of the waveguide to monitor the reflected optical power spectrum in real time. According to the reflected optical power spectrum, the piezoelectric ceramic micro-displacement adjustment is triggered by demodulating the reflected phase noise.

[0039] Specifically, the MEMS tunable grating is a tunable optical element manufactured using micro-electro-mechanical system technology, which can diffract and modulate incident light by changing its period, fill factor or tilt angle to control the propagation direction and intensity of light. For example, by controlling the tilt angle of the grating, the diffraction angle of light can be changed to achieve precise control of the optical signal. The reflected optical power spectrum refers to the power distribution of the reflected optical signal at different wavelengths from the output end of the waveguide. By monitoring the reflected optical power spectrum, important information such as the transmission characteristics, loss conditions and coupling state of the optical fiber and waveguide can be obtained. The reflected phase noise refers to the random fluctuation of the reflected optical signal in phase, which is usually caused by changes in the physical environment of the optical fiber or waveguide, interference during transmission and other factors. Phase noise can affect the quality and coherence of the optical signal. By demodulating the reflected phase noise, the transmission state of the optical signal can be monitored in real time, and the corresponding adjustment mechanism can be triggered to optimize the signal quality.

[0040] In one embodiment, a MEMS tunable grating is integrated at the output end of the waveguide to monitor the reflected optical power spectrum in real time. The monitoring frequency can reach thousands of times per second, covering a wavelength range of typically 1260nm to 1650nm, thereby comprehensively grasping the transmission characteristics of the optical signal in the waveguide. According to the reflected optical power spectrum, the reflected phase noise is demodulated to accurately trigger the piezoelectric ceramic micro-displacement adjustment, which improves the coupling efficiency of the waveguide and optical fiber, reduces the coupling loss, and effectively enhances the transmission quality and stability of the optical signal.

[0041] Further, the present application provides a method for demodulating the reflected phase noise according to the reflected optical power spectrum, which comprises:

[0042] obtaining a phase noise integral value, and triggering a piezoelectric ceramic displacement adjustment instruction when the phase noise integral value exceeds a preset integral threshold; and adjusting the curvature radius of the waveguide end face using the piezoelectric ceramic displacement adjustment instruction to control the adjustment accuracy within the tolerance standard.

[0043] Specifically, the phase noise integral value refers to a value obtained by integrating the reflected phase noise within a certain time interval, which is used to quantify the cumulative effect of phase noise and can reflect the phase stability of the optical signal during transmission; the preset integral threshold is a standard value preset for judging whether the phase noise exceeds the allowed range, and when the phase noise integral value exceeds the threshold, it indicates that the phase stability of the optical signal has been greatly affected and needs to be adjusted; the piezoelectric ceramic displacement adjustment instruction is a control signal used to drive the piezoelectric ceramic element to perform micro-displacement adjustment; piezoelectric ceramic is a material with piezoelectric effect that can produce a small deformation under the action of an electric field, and this deformation can be used to accurately adjust the curvature radius of the waveguide end face; the waveguide end face curvature radius adjustment precision refers to the accuracy of adjusting the curvature radius of the waveguide end face, which directly affects the coupling efficiency between the waveguide and the optical fiber and the quality of the optical signal transmission, to ensure the coupling efficiency and signal quality; the tolerance standard refers to the allowed error range, which is a specific requirement for the adjustment precision, and the adjustment precision of the waveguide end face curvature radius is usually required to be within ±0.5 microns to ensure efficient coupling and transmission of the optical signal.

[0044] In one embodiment, the phase noise integral value is obtained and compared with the preset integral threshold to achieve dynamic adjustment of the coupling efficiency between the waveguide and the optical fiber. Specifically, when the phase noise integral value exceeds the preset integral threshold, it indicates that the phase stability of the optical signal is disturbed, at which time the piezoelectric ceramic displacement adjustment instruction is triggered to control the waveguide end face curvature radius adjustment precision within the tolerance standard through piezoelectric ceramic micro-displacement adjustment, thereby maintaining the coupling efficiency between the waveguide and the optical fiber.

[0045] Optionally, in an optical communication system, when the optical fiber is affected by external temperature changes or mechanical vibrations, causing phase noise accumulation, the coupling efficiency between the waveguide and the optical fiber can be improved from the initial 60% to more than 85% through the above adjustment mechanism, significantly improving the transmission quality of the optical signal. At the same time, accurate curvature radius adjustment can also reduce the reflection loss of the optical signal, thereby further enhancing the overall performance and stability of the system and ensuring efficient and reliable operation of the optical communication system.

[0046] Further, the present application provides a method for constructing a thermal-electric-optical coupling model to adjust the power consumption of the high-speed electric driving circuit, which comprises:

[0047] A graphene-based micro heat pipe array is integrated below the electro-optical modulator, an infrared thermal imaging sensor is deployed, and the chip junction temperature is fed back in real time to dynamic power consumption adjustment; when the chip junction temperature exceeds the preset upper temperature limit, a gradient frequency reduction strategy is triggered and redundant heat is directed out through the heat pipe.

[0048] Specifically, the graphene-based micro heat pipe array, i.e., a heat dissipation element, has an extremely high thermal conductivity, which is usually above 5000 W / (m·K), much higher than that of traditional metal materials, so that the micro heat pipe can efficiently conduct and diffuse heat. The micro heat pipe array is composed of multiple micro heat pipes, each of which contains working liquid inside. When one end is heated, the liquid evaporates to take away heat, and the other end condenses to release heat when it is cooled, thereby achieving rapid heat transfer; the infrared thermal imaging sensor is a sensor that can detect the infrared radiation of an object surface and convert it into temperature information, which can monitor the junction temperature of the chip in real time without contacting the object; dynamic power consumption adjustment refers to adjusting the power consumption of the electric driving circuit according to the real-time situation of the chip temperature, load, etc., to ensure efficient operation of the circuit while avoiding overheating; the gradient frequency reduction strategy refers to gradually reducing the working frequency of the electric driving circuit to reduce power consumption and heat generation when the chip temperature is too high, so as to protect the chip from damage.

[0049] In one embodiment, the graphene-based micro heat pipe array is integrated below the electro-optical modulator, and the infrared thermal imaging sensor is deployed to monitor the chip junction temperature in real time, which can achieve precise control of the chip temperature. For example, when the chip load reaches more than 80% and the environmental temperature exceeds 35℃, the temperature of the electro-optical modulator rapidly rises to more than 90℃. At this time, the temperature information is fed back to the dynamic power consumption adjustment system in real time through the infrared thermal imaging sensor, and the gradient frequency reduction strategy is triggered, which can rapidly reduce the chip temperature to below 75℃ in a short time, while the redundant heat is directed out through the heat pipe, effectively avoiding the distortion of the optical signal and the rise of the bit error rate caused by overheating. Not only does it improve the stability and reliability of the chip, but also ensures the efficient operation of the entire silicon optical engine in the high-speed optical communication scenario. At the same time, by optimizing the working mode of the heat pipe, when the chip junction temperature is in the temperature range starting from 80% of the preset upper limit temperature, the heat dissipation is strengthened by starting the capillary action of the heat pipe, which can further improve the heat dissipation efficiency and make the chip work stably in a wider temperature range.

[0050] Further, the method comprises:

[0051] According to the dynamic adjustment of the chip junction temperature, the working mode of the heat pipe is adjusted. When the chip junction temperature is in the temperature range starting from 80% of the preset upper limit temperature, the heat dissipation is strengthened by starting the capillary action of the heat pipe.

[0052] Specifically, the chip junction temperature refers to the junction temperature of the transistors inside the chip, reflecting the heat accumulation of the chip during operation, and is an important indicator of the thermal state of the chip; the level of chip junction temperature will directly affect the performance, power consumption and life of the chip; the preset upper temperature limit is a temperature threshold set according to the design specifications and safe operation requirements of the chip, and when the chip junction temperature exceeds this preset upper temperature limit, it may cause the chip performance to decline, power consumption to increase, or even damage, for example, the common silicon optical chip preset upper temperature limit is usually between 85℃ to 100℃; the heat pipe capillary action enhanced heat dissipation refers to using the capillary structure inside the heat pipe to enhance the circulation of the working liquid and improve the heat dissipation capacity of the heat pipe. Capillary action is the phenomenon of liquid automatically rising or diffusing in a thin tube or porous material due to surface tension. In the heat pipe, capillary action enables liquid to quickly return from the condensing end to the evaporating end, maintaining the high efficiency of the heat pipe heat dissipation.

[0053] In one embodiment, the heat pipe operating mode is dynamically adjusted according to the chip junction temperature to accurately control the chip temperature and ensure its stable operation, for example, when the chip junction temperature is in a temperature range starting from 80% of the preset upper temperature limit, i.e. if the preset upper temperature limit is 85℃, the heat pipe capillary action enhanced heat dissipation is started when the chip junction temperature reaches 68℃, effectively suppressing temperature overshoot, reducing chip temperature fluctuations, reducing the risk of chip damage caused by temperature fluctuations, and reducing optical signal distortion and bit error rate, ensuring long-term stable operation of the system.

[0054] Further, the method further comprises:

[0055] A waveguide-fiber-drive circuit joint calibration module is constructed, and a 4D parameter compensation matrix is generated using a wavelength sensitive matrix tester; based on the 4D parameter compensation matrix, the modulator bias voltage and the fiber input angle are dynamically optimized through federated learning.

[0056] Specifically, the waveguide-fiber-drive circuit joint calibration module aims to ensure the cooperation and performance optimization between the silicon optical waveguide, optical fiber and electro-optical modulator drive circuit. Through joint calibration, the transmission efficiency and signal quality of the entire optical communication system can be improved. The wavelength sensitive matrix tester is an instrument that can measure the transmission characteristics of optical signals at different wavelengths and generate matrix data. By detecting the wavelength, intensity, phase and other parameters of the optical signal, a 4D parameter compensation matrix is generated to compensate and optimize the transmission characteristics of the optical signal. The 4D parameter compensation matrix is a four-dimensional data matrix that contains compensation parameters for optical signals at different wavelengths, different modes, different polarization states and different transmission paths. It can accurately guide the adjustment of modulator bias voltage and fiber input angle to achieve the best optical signal transmission effect. Federated learning is a distributed machine learning technique that collaboratively trains models on multiple devices or nodes while ensuring data privacy and security, thereby dynamically optimizing modulator bias voltage and fiber input angle and improving system adaptability and performance.

[0057] In one embodiment, the waveguide-fiber-drive circuit joint calibration module is constructed, and the 4D parameter compensation matrix is generated by the wavelength sensitive matrix tester, which can realize fine calibration and optimization of the optical communication system. Based on the 4D parameter compensation matrix, the modulator bias voltage and fiber input angle are dynamically optimized by federated learning, which can enable the system to maintain high performance under different working conditions and environments.

[0058] Optionally, in a 100Gbps optical communication system, the coupling efficiency of the optical fiber and the waveguide is improved by the joint calibration module and the 4D parameter compensation matrix, and the extinction ratio of the modulator is optimized to more than 25dB, which reduces the bit error rate of the optical signal by about 3 orders of magnitude. In an environment with a temperature change range of -5℃ to 75℃, the modulator bias voltage is dynamically adjusted by federated learning, which can ensure that the transmission quality of the optical signal is not affected by temperature, so that the bit error rate of the system is always maintained below 1e-12, while the transmission bandwidth of the optical signal is improved, significantly improving the performance and reliability of the entire optical communication system.

[0059] Further, the present application provides a method for dynamically optimizing the modulator bias voltage and the fiber input angle by federated learning, which further comprises:

[0060] Uploading local data to a cloud platform for global aggregation and updating; after global aggregation and updating, outputting the modulator bias voltage and the fiber input angle.

[0061] Specifically, the local data refers to the parameter data collected and processed on each distributed silicon optical module device about the optical communication performance; uploading to the cloud platform means that these data are transmitted to a centralized remote server system through the Internet for storage and further analysis; the cloud platform has powerful computing and data storage capabilities and can process data from multiple different locations.

[0062] In the federated learning environment, the global aggregated update refers to the aggregation of the model updates or training data uploaded by each local device, the updating of the global model through averaging weighting, etc., and the distribution of the optimized model to each local device for the next round of training, which can utilize a large amount of distributed data to improve the generalization ability of the model while protecting data privacy; the modulator bias voltage refers to the bias voltage applied to the electrode of the electro-optical modulator, which is used to control the working state and modulation efficiency of the modulator; the fiber input angle refers to the coupling angle between the optical fiber and the silicon optical waveguide, which is crucial for efficient coupling of optical signals; after the global model is updated, the optimal modulator bias voltage and fiber input angle are obtained for subsequent optical signal modulation and transmission.

[0063] In one embodiment, in a distributed optical fiber communication network, there are a large number of silicon optical modules, each module collects data about the quality of optical signal transmission, such as signal strength, bit error rate, transmission loss, etc., in different working environments; uploading the local data to the cloud platform can establish a large-scale database with the characteristics of optical communication performance; the cloud platform adopts a federated learning architecture, in which thousands of nodes are deployed, each node is responsible for processing and updating part of the model parameters.

[0064] Optionally, in the aggregated update process, the cloud platform analyzes and aggregates the data uploaded by each silicon optical module, and optimizes the calculation model of the modulator bias voltage and the fiber input angle using machine learning algorithms, for example, the cloud platform can perform statistical analysis on more than a million data points, and if it is found that when the modulator bias voltage is adjusted from 2.5V to 2.8V, the bit error rate can be reduced by 15%; and when the fiber input angle is optimized from 8° to 8.2°, the coupling efficiency can be increased by 10%; after the global aggregated update, the cloud platform outputs the optimized modulator bias voltage and fiber input angle parameters to each silicon optical module device, thereby improving the transmission efficiency and stability of the entire system.

[0065] Further, the application provides that after the global aggregated update, the modulator bias voltage and the fiber input angle are output, and the method further comprises:

[0066] Based on the edge computing node, a lightweight federated learning agent is deployed; in the lightweight federated learning agent, random noise injection is performed on the local data, and an adversarial network is formulated.

[0067] Specifically, edge computing nodes refer to computing devices at the edge of the network (close to the data source or user end) for preliminary processing and analysis of local data, which are small servers, intelligent gateways or dedicated hardware devices deployed in fiber communication networks, characterized by fast local data processing, reduced dependence on central servers and reduced data transmission delay; the lightweight federated learning agent refers to a simplified version of the federated learning client deployed on the edge computing node, which has smaller memory usage, lower computing resource requirements and faster response speed compared to the original federated learning client, and can efficiently run on resource-constrained edge devices while maintaining the basic functions of federated learning such as model updating, data encryption and communication optimization.

[0068] Random noise injection is used for data privacy protection, adding random noise to local data to make it difficult to be stolen or restored to the original information during transmission and processing. Specifically, a certain amplitude of random disturbance is added to the feature dimension of the data, for example, adding random noise in the range of [-0.5, 0.5] to the optical signal intensity data, to ensure the privacy of the data during federated learning; the adversarial network refers to a network structure built during federated learning to improve the robustness and generalization ability of the model; through adversarial training, the model's ability to adapt to different data distributions is enhanced during the learning process by simultaneously resisting random noise and data disturbance. Specifically, the adversarial network includes a generator and a discriminator, the generator tries to generate noise data that can confuse the discriminator, and the discriminator tries to distinguish between real data and generated data. Through this adversarial training, the stability and performance of federated learning are improved.

[0069] In one embodiment, each fiber connection node of the high-speed communication network of the silicon optical module is equipped with an edge computing device, and a lightweight federated learning agent is deployed on the edge computing node; the agent first preprocesses the locally collected optical communication data (such as modulator bias voltage, fiber input angle, optical signal quality, etc.), including data cleaning and feature extraction; before uploading the data to the cloud, the random noise injection module will disturb the data according to the preset noise intensity parameters (such as noise amplitude accounting for 5% of the original data) to ensure data privacy and security.

[0070] The lightweight federated learning agent end uses local computing resources for preliminary model training, and uploads the model update information (rather than the original data) obtained by training to the cloud platform; in the cloud platform, the global model is optimized by aggregating the updates of each edge node, and at the same time, the adversarial network module starts to work, simulates various possible noise interference and data abnormal situations, and performs adversarial training on the global model; after multiple iterations, the robustness of the global model is significantly improved; the optimized global model is distributed back to each edge computing node, and the lightweight federated learning agent end adjusts the local modulator bias voltage and fiber input angle according to the new model parameters, thereby improving the transmission efficiency and stability of the entire silicon optical communication network.

[0071] In summary, the embodiments of the present application have at least the following technical effects:

[0072] Since the basic technical parameters of the optical fiber connection product with the silicon optical module are obtained, the refractive index distribution and the waveguide cross-sectional size of the silicon optical waveguide are configured, the multi-mode waveguide array is constructed, the electro-optical modulator array is integrated in the multi-mode waveguide array, the coupling efficiency of the waveguide and the optical fiber is adjusted in real time through dynamic impedance matching, and at the same time, the high-speed electrical driving circuit is used to drive the electro-optical modulator array and optimize the signal rising edge characteristic / falling edge characteristic, and a thermal-electric-optical coupling model is constructed to adjust the power consumption of the high-speed electrical driving circuit. The present application provides a high-speed silicon optical engine method and device for a silicon optical module. The technical effects of dynamically configuring the waveguide refractive index, constructing the multi-mode waveguide array, effectively reducing the mode crosstalk, reducing the transmission loss, and improving the optical signal transmission quality are achieved.

[0073] Embodiment two, based on the same inventive concept as the high-speed silicon optical engine method for a silicon optical module in the foregoing embodiments, as Figure 2 shown, the present application provides a high-speed silicon optical engine device for a silicon optical module, which comprises:

[0074] The parameter acquisition module 11 is configured to obtain the basic technical parameters of the optical fiber connection product with the silicon optical module.

[0075] The multi-mode waveguide array construction module 12 is configured to configure the refractive index distribution and the waveguide cross-sectional size of the silicon optical waveguide based on the basic technical parameters, and construct a multi-mode waveguide array.

[0076] The impedance matching module 13 is configured to integrate an electro-optical modulator array in the multi-mode waveguide array, and adjust the coupling efficiency of the waveguide and the optical fiber in real time through dynamic impedance matching.

[0077] The power consumption adjustment module 14 is configured to, at the same time, use a high-speed electrical driving circuit to drive the electro-optical modulator array and optimize the signal rising edge characteristic / falling edge characteristic, and construct a thermal-electric-optical coupling model to adjust the power consumption of the high-speed electrical driving circuit.

[0078] Further, the high-speed silicon optical engine device for the silicon optical module is also used to execute the following method:

[0079] The common-mode choke coil is integrated at the output end of the high-speed electrical driving circuit, and the high-frequency signal attenuation is compensated by dynamic pre-emphasis.

[0080] Further, the impedance matching module 13 is also used to execute the following method:

[0081] The MEMS adjustable grating is integrated at the output end of the waveguide, the reflected light power spectrum is monitored in real time, and the piezoelectric ceramic micro-displacement adjustment is triggered according to the reflected phase noise demodulated based on the reflected light power spectrum.

[0082] Further, the impedance matching module 13 is also used to execute the following method:

[0083] The phase noise integral value is obtained, and when the phase noise integral value exceeds a preset integral threshold, a piezoelectric ceramic displacement adjustment instruction is triggered; the waveguide end face curvature radius adjustment precision is controlled within a tolerance standard by using the piezoelectric ceramic displacement adjustment instruction.

[0084] Further, the power consumption adjustment module 14 is also used to execute the following method:

[0085] The graphene-based micro heat pipe array is integrated below the electro-optical modulator, the infrared thermal imaging sensor is deployed, the chip junction temperature is fed back to the dynamic power consumption adjustment in real time, and when the chip junction temperature exceeds a preset upper temperature limit, the gradient frequency reduction strategy is triggered and the redundant heat is directed out through the heat pipe.

[0086] Further, the power consumption adjustment module 14 is also used to execute the following method:

[0087] The heat pipe working mode is adjusted dynamically according to the chip junction temperature, and when the chip junction temperature is in a temperature interval starting from 80% of the preset upper temperature limit, the heat pipe capillary action is started to strengthen heat dissipation.

[0088] Further, the power consumption adjustment module 14 is also used to execute the following method:

[0089] The waveguide-fiber-driving circuit joint calibration module is constructed, the 4D parameter compensation matrix is generated by using the wavelength sensitive matrix tester, and the modulator bias voltage and the fiber input angle are dynamically optimized by federated learning based on the 4D parameter compensation matrix.

[0090] Further, the power consumption adjustment module 14 is also used to execute the following method:

[0091] Upload the local data to the cloud platform for global aggregation update; after global aggregation update, output the modulator bias voltage and the fiber input angle.

[0092] Further, the power consumption adjustment module 14 is also used to execute the following method:

[0093] Based on the edge computing node, a lightweight federated learning agent end is deployed; the local data is subjected to random noise injection at the lightweight federated learning agent end to formulate an adversarial network.

[0094] It should be noted that the above-mentioned sequence of the embodiments of the present application is only for description, and does not represent the advantages and disadvantages of the embodiments. The above-mentioned specific embodiments of the present application are described. The processes depicted in the drawings do not necessarily require the specific order and continuous order shown to achieve the desired results. In some embodiments, multi-task processing and parallel processing are also possible or can be advantageous.

[0095] The above only describes the preferred embodiments of the present application and does not limit the present application. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present application shall be included in the protection scope of the present application.

[0096] The present application is only an exemplary description of the present application, and should be considered to cover any and all modifications, changes, combinations or equivalents within the scope of the present application. Obviously, those skilled in the art can make various modifications and changes to the present application without departing from the scope of the present application. Thus, if these modifications and changes of the present application belong to the scope of the present application and its equivalent technology, the present application intends to include these modifications and changes.

Claims

1. A high-speed silicon photonic engine method for silicon photonic modules, characterized in that, The method includes: Obtain the basic technical parameters of fiber optic connection products equipped with silicon photonics modules; Based on the aforementioned basic technical parameters, the refractive index distribution and waveguide cross-sectional dimensions of the silicon optical waveguide are configured to construct a multimode waveguide array; An electro-optic modulator array is integrated into the multimode waveguide array, and the coupling efficiency between the waveguide and the optical fiber is adjusted in real time through dynamic impedance matching. Simultaneously, a high-speed electric drive circuit is used to drive the electro-optic modulator array and optimize the signal rising / falling edge characteristics. A thermal-electric-optic coupling model is constructed to regulate the power consumption of the high-speed electric drive circuit.

2. The method as described in claim 1, characterized in that, The method includes: A common-mode choke is integrated at the output of the high-speed electric drive circuit to compensate for high-frequency signal attenuation through dynamic pre-emphasis.

3. The method as described in claim 2, characterized in that, The method for adjusting the coupling efficiency between the waveguide and the optical fiber in real time through dynamic impedance matching includes: A MEMS tunable grating is integrated at the waveguide output end to monitor the power spectrum of reflected light in real time. Based on the reflected light power spectrum, the micro-displacement of the piezoelectric ceramic is triggered by demodulating the reflected phase noise.

4. The method as described in claim 3, characterized in that, Based on the reflected light power spectrum, the method includes demodulating the reflected phase noise, comprising: The phase noise integral value is obtained, and when the phase noise integral value exceeds a preset integral threshold, a piezoelectric ceramic displacement adjustment command is triggered. Using the piezoelectric ceramic displacement adjustment command, the waveguide end face curvature radius adjustment accuracy is controlled within the tolerance standard.

5. The method as described in claim 1, characterized in that, The method for adjusting the power consumption of the high-speed electric drive circuit by constructing a thermal-electric-optical coupling model includes: A graphene-based micro heat pipe array is integrated below the electro-optic modulator, and an infrared thermal imaging sensor is deployed to provide real-time feedback on the chip junction temperature for dynamic power consumption adjustment. When the chip junction temperature exceeds the preset upper limit, a gradient frequency reduction strategy is triggered and redundant heat is directed out through a heat pipe.

6. The method as described in claim 5, characterized in that, The method includes: The heat pipe operating mode is dynamically adjusted according to the chip junction temperature. When the chip junction temperature is in the temperature range starting from 80% of the preset upper temperature limit, the heat pipe capillary action is activated to enhance heat dissipation.

7. The method as described in claim 6, characterized in that, The method further includes: A waveguide-fiber-drive circuit joint calibration module was constructed, and a 4D parameter compensation matrix was generated using a wavelength sensitivity matrix tester. Based on the 4D parameter compensation matrix, the modulator bias voltage and fiber input angle are dynamically optimized through federated learning.

8. The method as described in claim 7, characterized in that, The method further includes dynamically optimizing the modulator bias voltage and fiber input angle through federated learning. Upload local data to the cloud platform for global aggregation and updating; After global aggregation update, the modulator bias voltage and fiber input angle are output.

9. The method as described in claim 8, characterized in that, After the global aggregation update, the modulator bias voltage and fiber input angle are output. The method further includes: Deploy a lightweight federated learning agent based on edge computing nodes; At the lightweight federated learning agent, random noise is injected into the local data to formulate an adversarial network.

10. A high-speed silicon photonic engine device for silicon photonic modules, characterized in that, For implementing the high-speed silicon photonics engine method for a silicon photonics module according to any one of claims 1-9, the apparatus comprises: The parameter acquisition module is used to acquire the basic technical parameters of fiber optic connection products equipped with silicon photonics modules; A multimode waveguide array construction module is used to configure the refractive index distribution and waveguide cross-sectional dimensions of silicon optical waveguides based on the aforementioned basic technical parameters, and to construct a multimode waveguide array. An impedance matching module is used to integrate an electro-optic modulator array in the multimode waveguide array, and to adjust the coupling efficiency between the waveguide and the optical fiber in real time through dynamic impedance matching. The power consumption adjustment module is used to simultaneously drive the electro-optic modulator array with a high-speed electric drive circuit and optimize the signal rise / fall characteristics, and construct a thermal-electric-optic coupling model to adjust the power consumption of the high-speed electric drive circuit.

Images