Optical processing module and optical processing method
By setting waveguide paths of different lengths in the optical domain to compensate for optical signals, the high power consumption and delay problems of optical reservoir computing networks are solved, and efficient and low power consumption optical signal processing is achieved.
Patent Information
- Application Number
- PCT/CN2026/075969
- Authority / Receiving Office
- WO · WO
- Patent Type
- Applications
- Current Assignee / Owner
- Priority Date
- 2025-02-27
- Filing Date
- 2026-01-30
- Publication Date
- 2026-11-12
AI Technical Summary
Existing optical reservoir computing networks suffer from problems such as high demand for storing large matrix weight coefficients, slow computing speed, and high power consumption when handling complex tasks. Furthermore, existing technologies require converting optical signals to the electrical domain for compensation, which leads to increased power consumption and latency.
An optical processing module is used to compensate for optical signals by setting waveguide paths of different lengths in the optical domain. The signal is then transmitted using a photonic computing device, which reduces linear and nonlinear distortion and decreases the workload of digital signal processing.
Direct signal compensation in the optical domain reduces power consumption, complexity, and system latency, while improving the reliability and efficiency of signal processing.
Smart Images

Figure CN2026075969_12112026_PF_FP_ABST
Abstract
Description
A light processing module and a light processing method
[0001] This application claims priority to Chinese Patent Application No. 202510229580.X, filed on February 27, 2025, entitled "A Light Processing Module and a Light Processing Method", the entire contents of which are incorporated herein by reference. Technical Field
[0002] This application relates to the field of artificial intelligence (AI) technology, and more particularly to a light processing module and a light processing method. Background Technology
[0003] Artificial intelligence (AI) is the theory, methods, technology, and application systems that use digital computers or machines controlled by digital computers to simulate, extend, and expand human intelligence, perceive the environment, acquire knowledge, and use that knowledge to achieve optimal results. In other words, AI is a branch of computer science that attempts to understand the essence of intelligence and produce a new kind of intelligent machine that can react in a way similar to human intelligence. AI studies the design principles and implementation methods of various intelligent machines, enabling them to possess the functions of perception, reasoning, and decision-making.
[0004] In recent years, artificial neural networks have made significant progress, and they are increasingly serving the real world. Neural network structures are generally divided into feedforward networks and recurrent networks. Feedforward neural networks are mainly used for static (non-temporal) data processing because individual input data, even when given sequentially, are processed independently. Recurrent neural networks are suitable for dynamic (temporal) data processing because they can embed the temporal dependencies of the input into their dynamic behavior. Many real-world scenarios are temporal, such as prediction, adaptive filtering, computing device control or recognition, noise reduction, vision, and speech. With the advent of the big data era, dynamic signal processing will greatly reduce the overhead of storing and transmitting large datasets and the need for post-processing. Reservoir computing (RC) networks in recurrent neural networks are suitable for a wide variety of classification and prediction tasks.
[0005] The main characteristic of RC networks is that the input weights and reservoir weights are fixed; only the output weights need to be trained, which greatly shortens the training time and reduces the training difficulty, and allows for the simultaneous execution of multiple tasks. For complex tasks, optical RC neural networks require storing large matrices of weight coefficients, resulting in high memory requirements, slow computation speed, and high power consumption. In contrast, in optical neural networks, the weight coefficients are an inherent part of the photonic computing device, and information is transmitted at the speed of light, resulting in high speed, large bandwidth, and low power consumption.
[0006] In order to better realize optical reservoir calculation, there is an urgent need for an optical computing device with a simple computing device structure, a large number of nodes, a wide range of adjustable topology, and a small number of active devices. Summary of the Invention
[0007] In a first aspect, this application provides an optical processing module, comprising: a first unit; the first unit includes a first device, a first waveguide path, a second waveguide path, a third waveguide path, and a fourth waveguide path; the first waveguide path is used to transmit a first optical signal; the first device is used to divide the first optical signal based on signal strength to obtain a first sub-signal and a second sub-signal; the second waveguide path is used to transmit the first sub-signal; the third waveguide path is used to transmit the second sub-signal; the lengths of the second waveguide path and the third waveguide path are different; the fourth waveguide path is used to transmit a second optical signal, the second optical signal being an optical signal obtained by converging the first sub-signal transmitted through the second waveguide path and the second sub-signal transmitted through the third waveguide path.
[0008] The approach of this application is to compensate for optical signals by setting two optical waveguide paths (a second optical waveguide path and a third optical waveguide path) of different lengths within each cascaded unit (e.g., the first unit). By transmitting sub-signals of different intensities in the optical signal through the two waveguide paths respectively, a phase difference is generated, thereby changing the content of the input signal (i.e., the resulting second optical signal), thus compensating for the optical signal and reducing its linear and nonlinear distortion. This architecture compensates for optical signals in the optical domain without requiring prior photoelectric conversion, thereby reducing the workload of digital signal processing (DSP), reducing power consumption, complexity, and system latency, and ultimately improving reliability.
[0009] In one possible implementation, the optical processing module further includes: a first input interface; the first input interface is used to acquire first information, the first information indicating the division ratio of the optical signal; the first device is specifically used to divide the first optical signal based on the signal strength according to the first information.
[0010] The specific compensation method can be controlled by the division ratio of the first optical signal. For different system deployments, different optical signal modulation formats, different transmission rates, and different wavelengths, the optimal (or better) division ratio may be different. The division ratio can be determined by model training. Therefore, in this embodiment, the division ratio is set as an adjustable parameter. The user can input the division ratio through the first input interface. For example, the division ratio is determined by model training. The user can input the division ratio determined after training through the first input interface (for example, indicated by the first information).
[0011] In one possible implementation, the first device is a tunable coupler.
[0012] In one possible implementation, the first unit further includes: a second device and a fifth waveguide path; the fifth waveguide path is used to transmit a third sub-signal obtained by dividing the second optical signal based on signal strength; the second device is used to perform a second adjustment on the signal strength or phase of the third sub-signal.
[0013] The signal strength or phase of the third sub-signal transmitted on the fifth waveguide path is adjusted by the second device, thereby increasing the degree of freedom of compensation.
[0014] In one possible implementation, the optical processing module further includes: a second input interface; the second input interface is used to acquire second information, the second information indicating the parameters of the first adjustment; the first device is specifically used to perform a second adjustment on the signal strength or phase of the third sub-signal according to the second information.
[0015] In one possible implementation, the optical processing module further includes: a second unit and a sixth waveguide path; the second unit includes a third device and a seventh waveguide path; the sixth waveguide path is used to transmit a fourth sub-signal obtained by dividing the second optical signal based on signal strength; the third device is used to perform a second adjustment on the signal strength or phase of the fourth sub-signal transmitted through the fourth waveguide path; the seventh waveguide path is used to transmit a third optical signal, which is an optical signal obtained by converging the third sub-signal after the first adjustment and the fourth sub-signal after the second adjustment.
[0016] In one possible implementation, the optical processing module includes multiple cascaded units, including the first unit, an eighth waveguide path, a ninth waveguide path, and a tenth waveguide path; the eighth waveguide path and the ninth waveguide path have different lengths; the eighth waveguide path is used to transmit the converged signal of the multiple cascaded units; the ninth waveguide path is used to transmit the converged signal of the multiple cascaded units; and the tenth waveguide path is used to transmit a fourth optical signal, which is the optical signal obtained by converging the converged signal transmitted through the eighth waveguide path and the converged signal transmitted through the ninth waveguide path.
[0017] For example, the outputs (i.e., converged signals) of multiple cascaded units can be multiplexed, and each multiplexed signal can be transmitted and converged through waveguide paths of different lengths, thereby achieving more flexible phase compensation.
[0018] In one possible implementation, the optical processing module further includes a third device for making a third adjustment to the signal strength or phase of the converged signal transmitted on the eighth waveguide path.
[0019] In one possible implementation, the optical processing module further includes: a third input interface; the third input interface is used to acquire third information, the third information indicating the parameters of the third adjustment; the third device is specifically used to perform a third adjustment on the signal strength or phase of the converged signal transmitted on the eighth waveguide path according to the third information.
[0020] In one possible implementation, the optical processing module further includes a fourth device for making a fourth adjustment to the signal strength or phase of the converged signal transmitted on the ninth waveguide path.
[0021] In one possible implementation, the optical processing module further includes: a fourth input interface; the fourth input interface is used to acquire fourth information, the fourth information indicating the parameters of the fourth adjustment; the fourth device is specifically used to perform a fourth adjustment on the signal strength or phase of the converged signal transmitted on the ninth waveguide path according to the fourth information.
[0022] In one possible implementation, the second device, the third device, or the fourth device includes a Mach-Zehnder interferometer and a phase shifter.
[0023] Secondly, this application also provides an optical processing method applied to an optical processing apparatus, the optical processing apparatus comprising: a first unit; the first unit comprising a first device, a first waveguide path, a second waveguide path, a third waveguide path, and a fourth waveguide path; the method comprising: the first waveguide path transmitting a first optical signal; the first device dividing the first optical signal based on signal strength to obtain a first sub-signal and a second sub-signal; the second waveguide path transmitting the first sub-signal; the third waveguide path transmitting the second sub-signal; the lengths of the second waveguide path and the third waveguide path being different;
[0024] The fourth waveguide path transmits a second optical signal, which is the converged optical signal resulting from the convergence of the first sub-signal transmitted through the second waveguide path and the second sub-signal transmitted through the third waveguide path.
[0025] In one possible implementation, the optical processing module further includes: a first input interface; the method further includes: the first input interface acquiring first information, the first information indicating the division ratio of the optical signal; the first device specifically dividing the first optical signal based on the first information and signal strength.
[0026] In one possible implementation, the first device is a tunable coupler.
[0027] In one possible implementation, the first unit further includes: a second device and a fifth waveguide path; the method further includes:
[0028] The fifth waveguide path transmits a third sub-signal obtained by dividing the second optical signal based on signal strength;
[0029] The second device makes a second adjustment to the signal strength or phase of the third sub-signal.
[0030] In one possible implementation, the light processing module further includes: a second input interface; the method further includes:
[0031] The second input interface acquires second information, which indicates the parameters adjusted by the first input interface.
[0032] The first device specifically adjusts the signal strength or phase of the third sub-signal based on the second information.
[0033] In one possible implementation, the optical processing module further includes: a second unit and a sixth waveguide path; the second unit includes a third device and a seventh waveguide path; the method further includes:
[0034] The sixth waveguide path transmits the fourth sub-signal obtained by dividing the second optical signal based on signal strength;
[0035] The third device performs a second adjustment on the signal strength or phase of the fourth sub-signal transmitted through the fourth waveguide path;
[0036] The seventh waveguide path transmits a third optical signal, which is the optical signal resulting from the convergence of the third sub-signal after the first adjustment and the fourth sub-signal after the second adjustment.
[0037] In one possible implementation, the optical processing module includes multiple cascaded units, including the first unit, an eighth waveguide path, a ninth waveguide path, and a tenth waveguide path; the eighth waveguide path and the ninth waveguide path have different lengths; the method further includes:
[0038] The eighth waveguide path transmits the converged signals of the multiple cascaded units;
[0039] The ninth waveguide path transmits the converged signal of the multiple cascaded units;
[0040] The tenth waveguide path transmits a fourth optical signal, which is the optical signal resulting from the convergence of the convergence signal transmitted through the eighth waveguide path and the convergence signal transmitted through the ninth waveguide path.
[0041] In one possible implementation, the optical processing module further includes a third device, and the method further includes: the third device performing a third adjustment on the signal strength or phase of the converged signal transmitted on the eighth waveguide path.
[0042] In one possible implementation, the light processing module further includes: a third input interface; the method further includes:
[0043] The third input interface acquires third information, which indicates the parameters to be adjusted in the third manner;
[0044] The third device adjusts the signal strength or phase of the converged signal transmitted on the eighth waveguide path according to the third information.
[0045] In one possible implementation, the optical processing module further includes a fourth device, and the method further includes: the fourth device performing a fourth adjustment on the signal strength or phase of the converged signal transmitted on the ninth waveguide path.
[0046] In one possible implementation, the light processing module further includes: a fourth input interface; the method further includes:
[0047] The fourth input interface acquires fourth information, which indicates the fourth adjusted parameter;
[0048] The fourth device specifically adjusts the signal strength or phase of the converged signal transmitted on the ninth waveguide path according to the fourth information.
[0049] In one possible implementation, the second device, the third device, or the fourth device includes a Mach-Zehnder interferometer and a phase shifter.
[0050] For further details on optical computing methods, please refer to any of the implementation schemes in the first aspect, which will not be elaborated here.
[0051] Thirdly, this application also provides a processor chip, including a driving circuit and an optical computing device according to any embodiment of the first aspect. The processor chip here may also be referred to as an optical computing engine, a chip system, etc., and this application does not specifically limit it. The aforementioned driving circuit is used to drive the optical computing device. For example, the driving circuit can provide the necessary driving signals to the active devices in the optical computing device, enabling the active devices to operate normally in the optical computing device.
[0052] Fourthly, this application also provides a computing device, including the processor chip and memory described in the third aspect. The memory can be used to store data to be computed and / or the computation results of the data to be computed by the optical computing device in the processor chip. Attached Figure Description
[0053] To more clearly illustrate the technical methods of the embodiments of this application, the accompanying drawings used in the embodiments will be briefly described below.
[0054] Figure 1 is a schematic diagram of an application architecture provided in an embodiment of this application;
[0055] Figures 2 to 5 are schematic diagrams of an application architecture provided in an embodiment of this application;
[0056] Figures 6 to 9 are schematic diagrams illustrating the effects of embodiments of this application. Detailed Implementation
[0057] The embodiments of this application are described below with reference to the accompanying drawings. The terminology used in the implementation section of this application is for explaining specific embodiments only and is not intended to limit the scope of this application.
[0058] The embodiments of this application will now be described with reference to the accompanying drawings. Those skilled in the art will recognize that, with technological advancements and the emergence of new scenarios, the technical solutions provided in the embodiments of this application are equally applicable to similar technical problems.
[0059] The terms "first," "second," etc., used in the specification, claims, and accompanying drawings of this application are used to distinguish similar objects and are not necessarily used to describe a specific order or sequence. It should be understood that such terms are interchangeable where appropriate; this is merely a way of distinguishing objects with the same attributes in the embodiments of this application. Furthermore, the terms "comprising" and "having," and any variations thereof, are intended to cover non-exclusive inclusion, so that a process, method, system, product, or apparatus that comprises a series of elements is not necessarily limited to those elements, but may include other elements not explicitly listed or inherent to those processes, methods, products, or apparatuses.
[0060] First, the overall workflow of an artificial intelligence system is described, as shown in Figure 1. Figure 1 is a structural diagram of the main framework of artificial intelligence. The framework is then elaborated on from two dimensions: the "Intelligent Information Chain" (horizontal axis) and the "IT Value Chain" (vertical axis). The "Intelligent Information Chain" reflects a series of processes from data acquisition to processing. For example, it could be the general process of intelligent information perception, intelligent information representation and formation, intelligent reasoning, intelligent decision-making, and intelligent execution and output. In this process, data undergoes a condensation process of "data—information—knowledge—wisdom." The "IT Value Chain" reflects the value that artificial intelligence brings to the information technology industry, from the underlying infrastructure of human intelligence and information (provided and processed by technology) to the industrial ecosystem of the system.
[0061] (1) Infrastructure
[0062] Infrastructure provides computing power to support artificial intelligence systems, enabling communication with the external world and providing support through a basic platform. This communication occurs through sensors; computing power is provided by intelligent chips (hardware acceleration chips such as CPUs, NPUs, GPUs, ASICs, and FPGAs); and the basic platform includes distributed computing frameworks and related platform guarantees and support, which may include cloud storage and computing, interconnected networks, etc. For example, sensors communicate with the outside world to acquire data, and this data is provided to intelligent chips in the distributed computing system provided by the basic platform for computation.
[0063] (2) Data
[0064] The data at the next layer of infrastructure is used to represent the data sources in the field of artificial intelligence. The data involves graphics, images, voice, text, and IoT data from traditional devices, including business data from existing systems and sensor data such as force, displacement, liquid level, temperature, and humidity.
[0065] (3) Data processing
[0066] Data processing typically includes methods such as data training, machine learning, deep learning, search, reasoning, and decision-making.
[0067] Among them, machine learning and deep learning can perform intelligent information modeling, extraction, preprocessing, and training of data by symbolizing and formalizing it.
[0068] Reasoning refers to the process in which, in a computer or intelligent system, the machine thinks and solves problems by simulating human intelligent reasoning, based on reasoning control strategies and using formalized information. Typical functions include search and matching.
[0069] Decision-making refers to the process of making decisions based on intelligent information after reasoning, and it typically provides functions such as classification, sorting, and prediction.
[0070] (4) General ability
[0071] After the data processing mentioned above, the results of the data processing can be used to form some general capabilities, such as algorithms or a general system, for example, translation, text analysis, computer vision processing, speech recognition, image recognition, etc.
[0072] (5) Smart Products and Industry Applications
[0073] Intelligent products and industry applications refer to products and applications of artificial intelligence systems in various fields. They are the encapsulation of overall artificial intelligence solutions, productizing intelligent information decision-making and realizing practical applications. Their application areas mainly include: intelligent terminals, intelligent transportation, intelligent healthcare, autonomous driving, smart cities, etc.
[0074] 1. Optical computing:
[0075] Optical computing, also known as optical computing, utilizes the optical properties of optical devices to perform signal processing in the optical domain, thereby enabling corresponding mathematical calculations.
[0076] 2. All-optical nonlinearity:
[0077] The process where, thanks to the nonlinear coefficients of optical devices, the output signal light is not a linear function of the amplitude of the input signal light under the action of pump light is called an all-optical nonlinear process. The pump light, also known as the pump source in a laser, provides energy to excite lower-energy-level particles to higher-energy-level ones. In this process, both the pump light and the input signal light are optical signals, hence the term "all-optical effect." This process can include a nonlinear response between the amplitude / intensity of the output signal light and the amplitude / intensity of the input signal light, as well as changes in the frequency and phase of the output signal light compared to the input signal light. The pump light and the input signal light in this process can be the same beam or two different optical signals.
[0078] 3. Reservoir compute (RC):
[0079] Reservoir computing, also known as reservoir computing or pool computing, is a simplified RNN network architecture. An RC network consists of an input layer, a reservoir layer, and an output layer. The reservoir layer further includes a feedforward component and a feedback loop. On one hand, RC networks do not require training of the input and reservoir layers during training, making them feasible for hardware implementation. On the other hand, the presence of the feedback loop gives RC networks a certain short-term memory capability, making them suitable for scenarios such as time-series data classification and prediction.
[0080] With the rapid maturation of generative AI, the computing power required for large-scale models will be centralized in the cloud, giving rise to new models of intelligent computing cloud services. This includes the generation and transfer of computing power, and the centralization of computing service providers. To achieve more intelligent judgment and more real-time response, the training of large-scale models requires the continuous exchange of new data, which may span hundreds or even thousands of kilometers, necessitating efficient transmission between data centers and between data centers and enterprise users. This transmission process heavily relies on the development of fiber optic communication. Low-latency, low-power silicon-based photonic signal processing chips integrate multiple functions into a single chip, reducing coupling losses between devices and improving system reliability and stability.
[0081] Currently, various neural network architectures have been implemented on silicon photonic chips using micro-ring arrays and Mark-Zehnder interferometer network structures. Reservoir computing neural networks, as a variant of recurrent neural networks (RNNs), possess both linear memory and nonlinear processing capabilities, and require only the output layer for training, resulting in low complexity. Therefore, they can be well implemented on silicon-based photonic chip platforms.
[0082] However, after receiving optical signals transmitted through optical fibers, silicon photonics chips experience some distortion due to the transmission of signals within the fiber. Therefore, an all-optical signal compensator can be used to compensate for this distortion. Existing technologies convert the optical signal to the electrical domain and compensate for the electrical signal, which results in higher chip power consumption and latency.
[0083] The optical computing device involved in the embodiments of this application is described below.
[0084] Referring to FIG2, the optical computing device 200 provided in the embodiments of this application may include a first unit 201; wherein, the first unit 201 includes a first device, a first waveguide path, a second waveguide path, a third waveguide path, and a fourth waveguide path.
[0085] The optical computing device can be a silicon photonic chip, which can be used to implement reservoir computing neural networks. Specifically, the optical computing device can include multiple cascaded units, corresponding to multiple cascaded units in the reservoir pool layer of the reservoir computing neural network, and each cascaded unit can perform nonlinear transformation on the input signal.
[0086] In one example, taking multiple cascaded units including a first unit 201, the first unit 201 can be one of the multiple cascaded units. In this case, the first unit 201 can receive the optical signal input to the optical computing device (e.g., the first optical signal in the embodiment of this application) and transmit the optical signal (a portion of it) after being transmitted through the optical waveguide of the first unit 201 to the next adjacent cascaded unit.
[0087] The first unit 201 can also be the middle unit among multiple cascaded units. In this case, the first unit 201 can receive the optical signal transmitted by the adjacent previous cascaded unit (e.g., the first optical signal in the embodiment of this application) and transmit the optical signal (a portion of it) after being transmitted through the optical waveguide of the first unit 201 to the adjacent next cascaded unit (e.g., the second unit 202 in the embodiment of this application).
[0088] Furthermore, the output optical signal of each cascade unit (except for the last cascade unit, which is the cascade unit furthest from the input) can consist of two parts (divided according to light intensity). One part can be input to the next adjacent cascade unit, and the other part of the output optical signal of each cascade unit can be converged. The converged signal can be used as the output of multiple cascade units.
[0089] A detailed illustration of multiple cascaded units can be found in Figure 3.
[0090] The reservoir computational neural network can include a reservoir layer, which consists of one or more interconnected nonlinear nodes. This reservoir layer performs nonlinear transformations on the input signal, processing it nonlinearly in a higher dimension (in other words, it takes a one-dimensional input, multiplies it by weights, and projects the signal into a higher dimension). Cascading reservoirs enhances the network architecture's memory length by linearly combining the outputs of each reservoir, enabling the processing of higher-speed and longer-dispersion optical communication signals. Furthermore, cascading on a silicon-based platform requires only simple waveguide connections, making it highly feasible.
[0091] The aforementioned cascaded units can correspond to cascaded reservoirs, and the convergence operation of the outputs of multiple cascaded units can be achieved through waveguide connections.
[0092] In this embodiment of the application, the first unit 201 may include the first waveguide path 2011, which is used to transmit a first optical signal.
[0093] Optionally, the first optical signal can be an optical signal containing multiple wavelengths. The signal transmitted in the optical fiber can be a wavelength division multiplexing (WDM) signal containing multiple wavelengths obtained through WDM processing. The WDM signal can be compensated by an optical processing device without demultiplexing it. In other words, the embodiments of this application can support synchronous compensation of multiple wavelength signals, greatly improving the ability to process multiple concurrent signals.
[0094] Optionally, the first optical signal is a single-wavelength optical signal obtained by demultiplexing a signal containing multiple wavelengths. The signal transmitted in the optical fiber can be a wavelength division multiplexing signal containing multiple wavelengths obtained by wavelength division multiplexing processing. When the wavelength division multiplexing signal is demultiplexed, the single-wavelength signal can be compensated by an optical processing device.
[0095] In this embodiment of the application, the first unit 201 may include a first device 2012, which is used to divide the first optical signal based on signal intensity to obtain a first sub-signal and a second sub-signal.
[0096] In one possible implementation, the first device 2012 is a tunable coupler.
[0097] In one possible implementation, the optical processing module further includes: a first input interface; the first input interface is used to acquire first information, the first information indicating the division ratio of the optical signal; the first device 2012 is specifically used to divide the first optical signal based on the signal strength according to the first information.
[0098] For example, the first optical signal can be divided into a 6:4 ratio based on signal strength. Then the strength of the first optical signal is A, the signal strength of the first sub-signal can be 0.6*A, and the signal strength of the second sub-signal can be 0.4*A.
[0099] The approach of this application is to compensate for optical signals by setting two optical waveguide paths (a second optical waveguide path and a third optical waveguide path) of different lengths within each cascaded unit (e.g., the first unit 201). By transmitting sub-signals of different intensities in the optical signal through the two waveguide paths respectively, a phase difference is generated, thereby changing the content of the input signal (i.e., the resulting second optical signal), thus compensating for the optical signal and reducing its linear and nonlinear distortion. This architecture compensates for optical signals in the optical domain without requiring prior photoelectric conversion, thereby reducing the workload of digital signal processing (DSP), reducing power consumption, complexity, and system latency, and ultimately improving reliability.
[0100] The specific compensation method can be controlled by the division ratio of the first optical signal. For different system deployments, different optical signal modulation formats, different transmission rates, and different wavelengths, the optimal (or better) division ratio may be different. The division ratio can be determined by model training. Therefore, in this embodiment, the division ratio is set as an adjustable parameter. The user can input the division ratio through the first input interface. For example, the division ratio is determined by model training. The user can input the division ratio determined after training through the first input interface (for example, indicated by the first information).
[0101] For example, the system can first compare the reservoir output signal with the initial reference signal to determine the optimal weight parameters through training. This training process combines the input and target signals and uses a specific algorithm (such as linear regression) to optimize the weight parameters, ensuring that the output signal is as close as possible to the original undistorted signal in the transmission link. After training, these weight parameters are configured through precise hardware modeling to guarantee optimal performance in actual hardware operation. Once these parameters are successfully configured and deployed, the system can achieve automated signal compensation and recovery without additional adjustments.
[0102] The first device 2012 in this embodiment, and the second and third devices described later, can be programmable units, that is, devices with controllable parameters. Through the programmable unit system, real-time adjustment and optimization can be performed for different optical signal modulation formats, transmission rates, and wavelengths. The parameters of the programmable units can be determined through model training and input through the input interface.
[0103] In this embodiment of the application, referring to FIG3, FIG3 is a structural schematic diagram of the first unit 201, wherein the first unit 201 may include the second waveguide path 2013, the second waveguide path 2013 is used to transmit the first sub-signal; the third waveguide path is used to transmit the second sub-signal; the length of the second waveguide path 2013 and the length of the third waveguide path are different.
[0104] In this embodiment of the application, the first unit 201 may include the fourth waveguide path 2014 for transmitting a second optical signal. The second optical signal is an optical signal obtained by converging the first sub-signal transmitted through the second waveguide path 2013 and the second sub-signal transmitted through the third waveguide path.
[0105] In one possible implementation, the first unit 201 further includes: a second device and a fifth waveguide path; the fifth waveguide path is used to transmit a third sub-signal obtained by dividing the second optical signal based on signal strength; the second device is used to make a second adjustment to the signal strength or phase of the third sub-signal.
[0106] The second optical signal can be divided into two parts based on intensity. One part serves as the input to the next cascaded unit (e.g., the second unit 202) (e.g., the fourth sub-signal in a subsequent embodiment), while the other part can be transmitted through the waveguide structure and then used as the output of the cascaded unit, converging with the outputs of other cascaded units.
[0107] In this embodiment of the application, taking the fifth waveguide path in the first unit 201 as an example, the signal strength or phase of the third sub-signal transmitted on the fifth waveguide path can be adjusted by the second device, thereby increasing the degree of freedom of compensation.
[0108] The second device can be a Mach-Zehnder interferometer (MZI) or a phase shifter. The MZI can adjust the signal intensity of the optical signal, while the phase shifter can adjust the phase of the optical signal.
[0109] In one possible implementation, the optical processing module further includes: a second input interface; the second input interface is used to acquire second information, the second information indicating the parameters of the first adjustment; the first device 2012 is specifically used to perform a second adjustment on the signal strength or phase of the third sub-signal according to the second information.
[0110] The amplitude and phase of the optical signal can be adjusted in real time via programming, as described above. This flexibility not only ensures accurate compensation under different signal conditions but also allows for dynamic response to changes in the network environment. Through optical control, the system can effectively suppress thermal crosstalk and phase errors, significantly improving signal quality and reducing noise introduced during transmission.
[0111] In one possible implementation, the optical processing module further includes: a second unit 202 and a sixth waveguide path; the second unit 202 includes a third device and a seventh waveguide path; the sixth waveguide path is used to transmit a fourth sub-signal obtained by dividing the second optical signal based on signal strength; the third device is used to perform a second adjustment on the signal strength or phase of the fourth sub-signal transmitted through the fourth waveguide path; the seventh waveguide path is used to transmit a third optical signal, which is an optical signal obtained by converging the third sub-signal after the first adjustment and the fourth sub-signal after the second adjustment.
[0112] In this embodiment, the optical signal distorted during transmission enters a cascaded reservoir computational neural network architecture. This architecture consists of multiple cascaded reservoir units, with the optical signal processed layer by layer as it passes through these units. Within each reservoir unit, adjustments to the feedback strength and length enable each reservoir to effectively handle linear and nonlinear distortions generated during transmission and enhance the system's ability to process time-dependent signals. The output signals of each reservoir are output in parallel and weighted by a programmable unit combining MZI and a phase shifter. l e iΦl l = 1, 2, 3. These programmable units are highly flexible and can precisely adjust the output weights of each reservoir, thereby optimizing the overall output signal of the cascaded reservoir architecture.
[0113] In one possible implementation, the optical processing module includes multiple cascaded units, including the first unit 201, an eighth waveguide path, a ninth waveguide path, and a tenth waveguide path; the eighth waveguide path and the ninth waveguide path have different lengths; the eighth waveguide path is used to transmit the converged signal of the multiple cascaded units; the ninth waveguide path is used to transmit the converged signal of the multiple cascaded units; and the tenth waveguide path is used to transmit a fourth optical signal, which is the optical signal obtained by converging the converged signal transmitted through the eighth waveguide path and the converged signal transmitted through the ninth waveguide path.
[0114] For example, the outputs (i.e., converged signals) of multiple cascaded units can be multiplexed, and each multiplexed signal can be transmitted and converged through waveguide paths of different lengths, thereby achieving more flexible phase compensation.
[0115] In one possible implementation, the optical processing module further includes a third device for adjusting the signal strength or phase of the converged signal transmitted on the eighth waveguide path. By adjusting the signal strength or phase of optical signals on waveguide paths of different lengths using the third device, more flexible signal compensation can be achieved.
[0116] In one possible implementation, the optical processing module further includes: a third input interface; the third input interface is used to acquire third information, the third information indicating the parameters of the third adjustment; the third device is specifically used to perform a third adjustment on the signal strength or phase of the converged signal transmitted on the eighth waveguide path according to the third information.
[0117] In one possible implementation, the optical processing module further includes a fourth device for making a fourth adjustment to the signal strength or phase of the converged signal transmitted on the ninth waveguide path.
[0118] In one possible implementation, the optical processing module further includes: a fourth input interface; the fourth input interface is used to acquire fourth information, the fourth information indicating the parameters of the fourth adjustment; the fourth device is specifically used to perform a fourth adjustment on the signal strength or phase of the converged signal transmitted on the ninth waveguide path according to the fourth information.
[0119] In one possible implementation, the second device, the third device, or the fourth device includes a Mach-Zehnder interferometer and a phase shifter.
[0120] Figure 4 illustrates the architecture of an embodiment of this application. The entire system consists of two core modules: an all-optical cascaded reservoir computing neural network architecture and a reservoir computing all-optical readout layer architecture. These modules are connected sequentially and process optical signals from left to right. The output signal (i.e., the converged signal) processed by the cascaded reservoir architecture can be transmitted to the reservoir computing all-optical readout layer for further processing.
[0121] Referring to Figure 4, this architecture consists of an all-optical cascaded reservoir computation neural network architecture and an all-optical readout layer architecture. The parallel output of the cascaded reservoir architecture enters the reservoir readout architecture, with signals input and output via waveguides. The all-optical cascaded reservoir computation neural network architecture and the all-optical readout layer architecture are interconnected via waveguides, forming a serial processing unit. This design is equivalent to a complete architecture that directly implements reservoir computation in the optical domain, enabling efficient completion of reservoir and readout layer computational tasks within the optical domain. This all-optical computation method eliminates the need for electrical domain processing, significantly improving system response speed and reducing latency, making it particularly suitable for signal processing in high-speed optical communication scenarios.
[0122] The device module provided in this application embodiment can be applied to the receiving end of a general optical module unit in optical communication, as shown in Figure 5. Figure 5(a) shows the architecture of the prior art. In this module, the distorted optical signal is transmitted to the receiving end through optical fiber and enters the silicon-based photonic chip for compensation and recovery processing. This processing is completed directly in the optical domain, avoiding the DSP steps in the electrical domain. It is worth noting that two deployment schemes are provided here: (1) single-chip compensation for single-wavelength (Figure 5(b)); (2) single-chip compensation for multiple-wavelength (Figure 5(c)).
[0123] This application's embodiments enable seamless integration with existing optical modules, integrating the entire structure and photodetector onto a single silicon photonic chip. Before the optical signal undergoes photoelectric conversion, signal compensation and recovery are performed, thus eliminating the need for traditional DSP algorithms. This design effectively reduces the power consumption and computational resource overhead of the optical module, while simplifying the signal processing flow and significantly improving the efficiency and performance of the optical communication system.
[0124] To ensure the effectiveness of this architecture in practical applications, the embodiments of this application have undergone comprehensive modeling and optimization. During the modeling process, the relationship between the distorted signal and the original signal in the optical fiber communication link is analyzed, and machine learning or other training algorithms are used to optimize various parameters in the architecture. The optimized parameters accurately reflect reservoir feedback, readout layer weights, MZI configuration, and phase adjustment of the phase shifter. Through precise hardware modeling, it is ensured that the compensation effect reaches an ideal level under various signal conditions in practical applications.
[0125] In practical applications, micro-heaters based on silicon-based photonic chips are key components. They are used to apply specific voltage values to adjust the configuration of the MZI and phase shifter. By regulating the voltage of the micro-heaters, the system can achieve precise control of the optical signal. The specific configuration relationship is shown in Figure 6, which details the adjustment relationship between voltage value, temperature change, and the MZI and phase shifter, thereby ensuring that the system can achieve optimal signal compensation and recovery under different signal transmission conditions.
[0126] Figure 6 shows the configuration relationship between the MZI and the phase shifter voltages. The combination of the MZI and the phase shifter programming unit can realize the weight configuration in the range of [-1,1]. The configuration order is to first use the MZI to configure the weight value in the range of [0,1], and then use the phase shifter to control the positive or negative sign of the weight value.
[0127] In this embodiment, deployment scheme 1 (single-chip compensation for single-wavelength) is used to conduct a real-world experimental verification of high-speed optical communication signals. The data format and communication parameters are shown in Table 1 below:
[0128] Table 1
[0129] This study targets direct modulation and detection applications in data centers, focusing on both 800GbE and 1.6TbE communication scenarios. The specific experimental setup includes 8 wavelength division multiplexing (WDM) channels with a channel spacing of 3.2nm (400GHz), a communication rate of 100Gbaud, and a distance of 10km. OOK and PAM4 modulation formats are used, with each symbol (baud) representing one and two bits respectively. The input power is set to 0dBm to simultaneously account for the effects of linear dispersion and nonlinear distortion. During on-chip training, 10,000 symbols are used to optimize the system parameters. Once training is complete, another 10,000 symbols are used for verification to obtain the BER (Browser Error Rate).
[0130] In this embodiment, compensation and restoration are first performed for 800GbE (8x100 Gbps OK) and 1.6TbE (8x200 Gbps) in the C-band scenario, resulting in the following two technical effects:
[0131] By optimizing the programmable unit, in the single-chip single-wavelength compensation scheme, the embodiments of this application can realize all-optical wavelength for 100Gbaud signals with different wavelengths and modulation formats, and meet the FEC requirements, as shown in Figure 7. Figure 7 is a comparison of BER for all-optical compensation of channel 100Gbps OOK and 100Gbaud PAM4 signals.
[0132] The solution in this application completely eliminates the use of DSP compensation, and only requires about 100mW of power consumption to achieve all-optical compensation of a single-wavelength 100Gbaud PAM4 signal. Compared with DSP compensation, the power consumption is reduced by more than 200 times, as shown in Figure 8, which is a comparison of power consumption for single-channel PAM4 signal compensation.
[0133] The use of all-optical compensation in the embodiments of this application will not generate additional power consumption for higher speed signals, while the DSP will increase power consumption as the speed increases.
[0134] Furthermore, the numerical simulation in this application mainly focuses on C-band optical communication signals, and the data format and communication parameters are shown in Table 2 below:
[0135] Table 2
[0136] For deployment scheme 2 (i.e., single-chip multi-wavelength compensation) in this application embodiment, the communication rate and distance are set to 56 Gbaud and 5 km, respectively, with 16 WDM channels and a channel spacing of approximately 1.6 nm (i.e., 200 GHz), and PAM4 modulation format is adopted. To avoid the influence of nonlinear effects, the input power is set to 14 dBm, thus simultaneously considering the effects of linear dispersion and nonlinear distortion. During the training process, 10,000 symbols are used to optimize the parameters in the system. Once training is complete, another 10,000 symbols are used for verification to obtain the BER value.
[0137] In this embodiment, the present application focuses on compensation and recovery for a 16-channel 56Gbaud PAM4 signal transmission distance of 5km in a C-band scenario. As shown in Figure 9, Figure 9 compares the BER performance of a single-chip compensated multi-wavelength C-band communication system.
[0138] For the single-chip multi-wavelength compensation scheme, the all-optical processor in this application embodiment can eliminate the need for some DSP compensation while meeting the requirements of HD-FEC. Since the optical fiber has different degrees of dispersion distortion for different wavelengths, the scheme in this application can compensate for some optical fiber distortion and reduce the complexity of the DSP. The all-optical processor can reduce the overall power consumption (48.1W → 9.88W) by reducing the complexity of the DSP. Deployment scheme 2 adopts the form of single-chip multi-wavelength compensation, which can reduce the hardware chip cost compared with scheme 1.
[0139] Because this optical computing device can provide high-performance computing, it can meet the computing needs of various new vertical application scenarios such as smart cities, smart transportation, smart manufacturing and smart homes for high bandwidth and low energy consumption.
[0140] It should also be noted that the device embodiments described above are merely illustrative. The units described as separate components may or may not be physically separate, and the components shown as units may or may not be physical units; that is, they may be located in one place or distributed across multiple network units. Some or all of the modules can be selected to achieve the purpose of this embodiment according to actual needs. In addition, in the device embodiment drawings provided in this application, the connection relationship between modules indicates that they have a communication connection, which can be implemented as one or more communication buses or signal lines.
[0141] Through the above description of the embodiments, those skilled in the art can clearly understand that this application can be implemented by means of software plus necessary general-purpose hardware, or it can be implemented by special-purpose hardware including application-specific integrated circuits, special-purpose CPUs, special-purpose memory, special-purpose components, etc. Generally, any function performed by a computer program can be easily implemented by corresponding hardware, and the specific hardware structure used to implement the same function can also be diverse, such as analog circuits, digital circuits, or special-purpose circuits. However, for this application, software program implementation is more often the preferred implementation method. Based on this understanding, the technical solution of this application, in essence, or the part that contributes to the prior art, can be embodied in the form of a software product. This computer software product is stored in a readable storage medium, such as a computer floppy disk, USB flash drive, mobile hard disk, ROM, RAM, magnetic disk, or optical disk, etc., and includes several instructions to cause a computer device (which may be a personal computer, server, or network device, etc.) to execute the methods described in the various embodiments of this application.
[0142] In the above embodiments, implementation can be achieved, in whole or in part, through software, hardware, firmware, or any combination thereof. When implemented in software, it can be implemented, in whole or in part, as a computer program product.
[0143] The computer program product includes one or more computer instructions. When the computer program instructions are loaded and executed on a computer, all or part of the processes or functions described in the embodiments of this application are generated. The computer may be a general-purpose computer, a special-purpose computer, a computer network, or other programmable device. The computer instructions may be stored in a computer-readable storage medium or transmitted from one computer-readable storage medium to another. For example, the computer instructions may be transmitted from one website, computer, server, or data center to another website, computer, server, or data center via wired (e.g., coaxial cable, fiber optic, digital subscriber line (DSL)) or wireless (e.g., infrared, wireless, microwave, etc.) means. The computer-readable storage medium may be any available medium that a computer can store or a data storage device such as a server or data center that integrates one or more available media. The available medium may be a magnetic medium (e.g., floppy disk, hard disk, magnetic tape), an optical medium (e.g., DVD), or a semiconductor medium (e.g., solid-state disk (SSD)).
Claims
1. A light processing module, characterized in that, include: The first unit includes a first device, a first waveguide path, a second waveguide path, a third waveguide path, and a fourth waveguide path. The first waveguide path is used to transmit the first optical signal; The first device is used to divide the first optical signal based on signal intensity to obtain a first sub-signal and a second sub-signal; The second waveguide path is used to transmit the first sub-signal; The third waveguide path is used to transmit the second sub-signal; the length of the second waveguide path and the length of the third waveguide path are different. The fourth waveguide path is used to transmit a second optical signal, which is the optical signal obtained by converging the first sub-signal transmitted through the second waveguide path and the second sub-signal transmitted through the third waveguide path through the first device.
2. The optical processing module according to claim 1, characterized in that, The first optical signal is an optical signal containing multiple wavelengths; or, the first optical signal is an optical signal of a single wavelength obtained by demultiplexing a signal containing multiple wavelengths.
3. The light processing module according to claim 1 or 2, characterized in that, The optical processing module further includes: a first input interface; The first input interface is used to acquire first information, which indicates the division ratio of the optical signal; The first device is specifically used to divide the first optical signal based on signal strength according to the first information.
4. The optical processing module according to any one of claims 1 to 3, characterized in that, The first device is a tunable coupler.
5. The light processing module according to any one of claims 1 to 4, characterized in that, The first unit further includes: a second device and a fifth waveguide path; The fifth waveguide path is used to transmit a third sub-signal obtained by dividing the second optical signal based on signal strength; The second device is used to make a second adjustment to the signal strength or phase of the third sub-signal.
6. The light processing module according to claim 5, characterized in that, The optical processing module further includes: a second input interface; The second input interface is used to obtain second information, which indicates the first adjusted parameter; The first device is specifically used to make a second adjustment to the signal strength or phase of the third sub-signal based on the second information.
7. The optical processing module according to any one of claims 1 to 6, characterized in that, The optical processing module further includes: a second unit and a sixth waveguide path; the second unit includes a third device and a seventh waveguide path; The sixth waveguide path is used to transmit the fourth sub-signal obtained by dividing the second optical signal based on signal strength; The third device is used to make a second adjustment to the signal strength or phase of the fourth sub-signal transmitted through the fourth waveguide path; The seventh waveguide path is used to transmit a third optical signal, which is the optical signal obtained by converging the third sub-signal after the first adjustment and the fourth sub-signal after the second adjustment.
8. The light processing module according to any one of claims 1 to 7, characterized in that, The optical processing module includes multiple cascaded units, including the first unit, an eighth waveguide path, a ninth waveguide path, and a tenth waveguide path; the eighth waveguide path and the ninth waveguide path have different lengths. The eighth waveguide path is used to transmit the converged signal of the multiple cascaded units; The ninth waveguide path is used to transmit the converged signal of the multiple cascaded units; The tenth waveguide path is used to transmit a fourth optical signal, which is the optical signal obtained by converging the converged signal transmitted through the eighth waveguide path and the converged signal transmitted through the ninth waveguide path.
9. The optical processing module according to claim 8, characterized in that, The optical processing module further includes a third device, which is used to make a third adjustment to the signal strength or phase of the converged signal transmitted on the eighth waveguide path.
10. The optical processing module according to claim 9, characterized in that, The optical processing module further includes: a third input interface; The third input interface is used to obtain third information, which indicates the parameters of the third adjustment; The third device is specifically used to make a third adjustment to the signal strength and phase of the converged signal transmitted on the eighth waveguide path based on the third information.
11. The light processing module according to any one of claims 8 to 10, characterized in that, The optical processing module further includes a fourth device, which is used to make a fourth adjustment to the signal strength or phase of the converged signal transmitted on the ninth waveguide path.
12. The optical processing module according to claim 11, characterized in that, The optical processing module further includes: a fourth input interface; The fourth input interface is used to obtain fourth information, which indicates the fourth adjustment parameter; The fourth device is specifically used to make a fourth adjustment to the signal strength and phase of the converged signal transmitted on the ninth waveguide path based on the fourth information.
13. The optical processing module according to any one of claims 4 to 12, characterized in that, The second device, the third device, or the fourth device includes: a Mach-Zehnder interferometer and a phase shifter.
14. A light processing method, characterized in that, The method is applied to an optical processing device, the optical processing device comprising: a first unit; the first unit comprising a first device, a first waveguide path, a second waveguide path, a third waveguide path, and a fourth waveguide path; the method comprising: The first waveguide path transmits the first optical signal; The first device divides the first optical signal based on signal intensity to obtain a first sub-signal and a second sub-signal; The second waveguide path transmits the first sub-signal; The third waveguide path transmits the second sub-signal; the length of the second waveguide path and the length of the third waveguide path are different. The fourth waveguide path transmits a second optical signal, which is the converged optical signal resulting from the convergence of the first sub-signal transmitted through the second waveguide path and the second sub-signal transmitted through the third waveguide path.
15. The method according to claim 14, characterized in that, The optical processing module further includes: a first input interface; the method further includes: The first input interface acquires first information, which indicates the division ratio of the optical signal; Specifically, the first device divides the first optical signal based on signal strength according to the first information.
16. The method according to claim 14 or 15, characterized in that, The first unit further includes: a second device and a fifth waveguide path; the method further includes: The fifth waveguide path transmits a third sub-signal obtained by dividing the second optical signal based on signal strength; The second device performs a second adjustment on the signal strength and phase of the third sub-signal.
17. A processor chip, characterized in that, The device includes a driving circuit and a light processing module as described in any one of claims 1 to 13, wherein the driving circuit is used to drive the light processing module.
18. A computing device, characterized in that, Includes the processor chip and memory as described in claim 17.
