CN-122001483-A - High-speed data processing method, device and medium based on optical co-encapsulation

CN122001483ACN 122001483 ACN122001483 ACN 122001483ACN-122001483-A

Abstract

The invention discloses a high-speed data processing method, equipment and a medium based on optical co-encapsulation, which relate to the technical field of photoelectric integrated encapsulation and comprise the steps of receiving a high-speed data stream to be transmitted, monitoring a sensor array integrated at key positions of a photon chip and an encapsulation substrate in real time to obtain multidimensional physical quantity sensing data, carrying out fusion and standardization processing on the multidimensional physical quantity sensing data to construct a characteristic vector used for representing a coupling state of a plurality of physical fields in a current encapsulation, inputting the characteristic vector into a digital twin model, describing parameters of nonlinear behaviors of an optical modulator in the current environment, carrying out probability shaping encoding on the high-speed data stream to be transmitted, and obtaining a predistortion lookup table used for compensating nonlinear distortion based on a nonlinear response parameter set and probability shaping parameters used by the probability shaping encoding. The invention realizes the mechanism driving type and high-precision prediction from the environment state to the nonlinear response of the device.

Inventors

WEN WEI
HUANG YONGHAO

Assignees

深圳市立汇通信技术有限公司

Dates

Publication Date: 20260508
Application Date: 20260129

Claims (10)

1. A high-speed data processing method based on optical co-packaging is characterized by comprising the following steps of, Receiving a high-speed data stream to be transmitted, and monitoring a sensor array integrated at key positions of a photon chip and a packaging substrate in real time to obtain multidimensional physical quantity sensing data; fusion and standardization processing are carried out on the multidimensional physical quantity sensing data, and a feature vector for representing the coupling state of the multidimensional physical quantity in the current package is constructed; inputting the feature vector into a digital twin model, and describing parameters of nonlinear behaviors of the optical modulator in the current environment; carrying out probability shaping coding on a high-speed data stream to be transmitted, and acquiring a predistortion lookup table for compensating nonlinear distortion based on a nonlinear response parameter set and probability shaping parameters used by the probability shaping coding; Real-time correction is carried out on the probability shaping symbol sequence according to the predistortion lookup table, and an optical modulator is driven to generate an output optical signal; and receiving and demodulating the output optical signal to obtain a received digital signal, performing performance evaluation on the received digital signal, feeding back an evaluation result to a transmitting end, and dynamically triggering updating of the digital twin model and the predistortion lookup table.
2. The method for processing high-speed data based on optical co-packaging as recited in claim 1, wherein the step of acquiring the multi-dimensional physical quantity sensing data comprises the following steps, Receiving a high-speed data stream to be transmitted, and generating a trigger signal for controlling the synchronization of the sensor array; And sending the trigger signal to a sensor array integrated at the key positions of the photon chip and the packaging substrate, and collecting the multidimensional physical quantity sensing data.
3. The method for high-speed data processing based on optical co-packaging as recited in claim 2, wherein said multi-dimensional physical quantity sensing data includes temperature data, stress strain data and carrier concentration data.
4. The method for processing high-speed data based on optical co-packaging as recited in claim 3, wherein the fusion and standardization process is performed on the multidimensional physical quantity sensing data to construct a feature vector for representing the coupling state of multiple physical fields in the current packaging, specifically comprising the following steps of, Performing space-time alignment and outlier processing on the multidimensional physical quantity sensing data to obtain multidimensional physical quantity sensing data; constructing a heterogeneous graph structure for representing the coupling relation of multiple physical fields in the package based on standardized multidimensional physical quantity sensing data; performing coupling information aggregation on the heterogeneous graph structure to obtain a node state set; And carrying out image pooling and splicing operation on the node state set to generate a feature vector for representing the multi-physical field coupling state in the current package.
5. The method for high-speed data processing based on optical co-packaging as recited in claim 4, wherein the inputting of the feature vector into the digital twin model outputs the nonlinear behavior description parameters of the optical modulator in the current environment comprises the following steps, Mapping the feature vector into an initial state point on the high-dimensional manifold through an embedded transformation layer of the digital twin model; Carrying out nonlinear time sequence evolution on the initial state point, and deducing a dynamic track in a set time window; extracting a coupled oscillator state vector representing the intensity of a nonlinear mechanism in the modulator from the end point of the dynamic track; decoding the coupled oscillator state vector through a parameter regression network of the digital twin model to obtain characteristic parameters of a nonlinear transfer function of the optical modulator; And carrying out dimension reduction and physical meaning mapping on the characteristic parameters to generate nonlinear behavior description parameters of the optical modulator in the current environment.
6. The method for high-speed data processing based on optical co-packaging as recited in claim 5, wherein said obtaining a predistortion look-up table for compensating nonlinear distortion comprises the steps of, Carrying out probability shaping coding on a high-speed data stream to be transmitted, generating a probability shaping symbol sequence, and extracting statistical characteristics of the probability shaping symbol sequence; fusing the nonlinear response parameter set, the probability shaping parameters used by the probability shaping codes and the statistical characteristics to construct a joint condition feature vector; Carrying out iterative optimization on the joint condition feature vector to generate a predistortion mapping relation table; Discretizing and encoding the predistortion mapping relation table to obtain a predistortion lookup table for compensating nonlinear distortion.
7. The method for high-speed data processing based on optical co-packaging as recited in claim 6, wherein said generating the output optical signal comprises the steps of, Extracting a current probability shaping symbol to be processed and a historical probability shaping symbol from the probability shaping symbol sequence to form a context window; Performing predistortion correction on the context window and the predistortion lookup table, and performing filtering compensation on the context window to obtain a driving symbol; The driving symbol is converted into an analog voltage signal, and the optical modulator is driven by the analog voltage signal to generate an output optical signal.
8. The method for processing high-speed data based on optical co-packaging as recited in claim 7, wherein the steps of receiving and demodulating the output optical signal to obtain a received digital signal, performing performance evaluation on the received digital signal, feeding back the evaluation result to the transmitting end, dynamically triggering updating of the digital twin model and the predistortion lookup table are as follows, Receiving and demodulating the output optical signal to produce a sequence of received digital symbols; based on the received digital symbol sequence, performance indexes of multiple dimensions are calculated in parallel, and a multi-dimensional performance characteristic vector is generated; calculating a comprehensive performance degradation index according to the multidimensional performance characteristic vector; comparing the comprehensive performance degradation index with a preset trigger condition to generate an update trigger instruction; And sending an update trigger instruction to the transmitting end, and executing the update of the digital twin model and the predistortion lookup table.
9. A computer device comprises a memory and a processor, wherein the memory stores a computer program, and the computer program is characterized in that the processor realizes the steps of the high-speed data processing method based on optical co-packaging according to any one of claims 1-8 when executing the computer program.
10. A computer-readable storage medium, on which a computer program is stored, characterized in that the computer program, when being executed by a processor, implements the steps of the optical co-encapsulation based high speed data processing method according to any one of claims 1 to 8.

Description

High-speed data processing method, device and medium based on optical co-encapsulation Technical Field The invention relates to the technical field of photoelectric integrated packaging, in particular to a high-speed data processing method, equipment and medium based on optical co-packaging. Background With the rapid increase in data center interconnection and high performance computing demands, transmission rates are advancing toward terabits per second, and conventional pluggable optical modules face severe bottlenecks in bandwidth density and power consumption. The optical co-packaging technology greatly shortens the electric interconnection distance by tightly integrating the photonic chip and the application specific integrated circuit on the substrate level, and becomes a key path for breaking through the bottleneck. Current developments focus on high density integration, high speed modulator design and advanced packaging processes for silicon-based photonic chips. Particularly, micro-ring modulator arrays based on photonic chips are attracting attention due to their compact size and low power consumption characteristics. At the same time, to approach the channel capacity limit, advanced modulation formats such as probability shaping are introduced into the optical interconnect, which improves spectral efficiency by non-uniformly distributing the encoding of signal symbols. The photonic chip is tightly coupled with the electronic chip and the substrate material in the package, forming a complex thermo-force-electricity multi-physical field environment which has dynamic and complex influence on the performance of the photonic device, especially the linearity and wavelength stability of the modulator. The prior art solutions have significant drawbacks in addressing the challenges described above. First, at the signal processing level, most schemes consider probability-shaping coding and nonlinear compensation as independent links. The transfer function of a modulator in a photonic chip, particularly a silicon-based micro-ring modulator, has significant nonlinear characteristics, and the characteristics dynamically change along with the bias point, the temperature and the carrier concentration. The non-uniform distribution characteristic of the probability shaped signal is such that it passes through the non-linear transfer function, a unique distortion spectrum is produced that differs from the uniformly distributed signal. Secondly, in the context of environmental sensing and compensation, existing methods typically integrate limited sensors on photonic chips or package substrates for single point monitoring and employ linear or simple look-up tables for compensation. The method cannot accurately describe the space-time distribution of multidimensional physical quantities such as temperature, stress, carrier concentration and the like in the photonic chip and the nonlinear coupling effect thereof, and cannot establish a high-precision mapping relation between the complex physical field state and the instantaneous nonlinear response parameters of the modulator. Disclosure of Invention The present invention has been made in view of the above-described problems occurring in the prior art. Therefore, the invention provides a high-speed data processing method based on optical co-packaging, which solves the problems of distortion compensation of a probability shaping signal in a dynamic nonlinear photon channel and real-time modeling of the influence of multi-physical field coupling in packaging on modulator performance. In order to solve the technical problems, the invention provides the following technical scheme: The invention provides a high-speed data processing method based on optical co-packaging, which comprises the steps of receiving a high-speed data stream to be transmitted, carrying out real-time monitoring on a sensor array integrated at key positions of a photon chip and a packaging substrate to obtain multidimensional physical quantity sensing data, carrying out fusion and standardization processing on the multidimensional physical quantity sensing data to construct a characteristic vector used for representing the coupling state of multiple physical fields in the current packaging, inputting the characteristic vector into a digital twin model, carrying out probability shaping coding on the high-speed data stream to be transmitted, obtaining a predistortion lookup table used for compensating nonlinear distortion based on a nonlinear response parameter set and probability shaping parameters used by the probability shaping coding, carrying out real-time correction on a probability shaping symbol sequence according to the predistortion lookup table, driving the optical modulator to generate an output optical signal, receiving and demodulating the output optical signal to obtain a received digital signal, carrying out performance evaluation on the received digital signal, feeding back an evaluatio