Optical chip data carrying method

Abstract

The invention discloses an optical chip data carrying method, which comprises the following steps of: S1, generating a multi-wavelength optical signal, and separating the multi-wavelength optical signal by using a wavelength division demultiplexer; s2, the FPGA or the single-chip microcomputer generates a control signal, and the control signal is subjected to AD9764 digital-to-analog conversion; s3, the analog signal and the light beam are coupled, and a Mach-Zehnder interferometer modulates the analog signal and the light beam; s4, combining the light beams by a wavelength division multiplexer to form a multi-wavelength modulation signal; s5, the modulated optical signal is input into an optical chip, and optical calculation is carried out; s6, transmitting the signal to a receiving end through a low-loss optical fiber to ensure that the signal is stable; s7, the wavelength division demultiplexer at the receiving end decomposes the signal, and the Mach-Zehnder interferometer recovers; and S8, converting into an electric signal, completing data demodulation, and realizing stable transmission. According to the invention, wavelength division multiplexing, optical chip parallel computing and Mach-Zehnder interferometer modulation technologies are combined, and optical signal data carrying and optical computing processing are realized.

Description

Technical Field

[0001] The present invention relates to the technical fields of optical communication and optical computing, and particularly to a method for carrying data on an optical chip. Background Art

[0002] With the rapid development of big data, cloud computing, and artificial intelligence technologies, data centers, high-performance computing (HPC), and intelligent network devices have put forward higher requirements for the data transmission rate, bandwidth, and stability. Optical communication technology occupies an important position in the modern communication field due to its high bandwidth, low latency, and electromagnetic interference resistance. Especially in the application scenarios of large-scale data transmission and ultra-high-speed computing, optical chips, as the core components of optical communication systems, play a key role in the data modulation, transmission, and demodulation processes. However, there are still multiple technical bottlenecks in the existing methods for carrying data on optical chips, resulting in limited system performance and difficulty in meeting the growing data transmission requirements.

[0003] The current optical chip data transmission system usually adopts the optoelectronic conversion (O / E / O) method, that is, the electrical signal is converted into an optical signal for transmission through an optoelectronic conversion module, and the optical signal is converted back into an electrical signal at the receiving end for data parsing. This method can meet the basic requirements in the medium and low-speed data transmission environment, but with the increase in data volume and transmission rate, the limitations of this method gradually emerge. First, the bandwidth of the optoelectronic conversion module is limited, which restricts the data throughput capacity of the overall system and cannot fully utilize the high-speed advantage of optical communication. Second, frequent optoelectronic conversion leads to an increase in system power consumption. Especially in large-scale data centers and cloud computing scenarios, the energy consumption of the traditional optoelectronic conversion architecture has become an important optimization target. In addition, the incompatibility of the interface protocols between optical chips and peripheral devices (such as FPGAs, MCUs, etc.) increases the complexity of system integration and affects the adaptability and flexibility of the devices.

[0004] In order to improve data transmission efficiency, wavelength division multiplexing (WDM) technology has been widely used in recent years. WDM improves the parallelism of data transmission by simultaneously transmitting multiple optical signals with different wavelengths in a single optical fiber. However, the existing WDM systems mainly rely on external digital signal processing (DSP) units for signal equalization, error correction, etc., which not only increases the hardware cost but also raises the system power consumption. In addition, multi-wavelength parallel transmission may cause crosstalk and dispersion effects between signals, which in turn affect the stability of data transmission. Although fiber dispersion compensation technologies (such as dispersion compensation fiber DCF and digital dispersion equalization) can improve this problem to a certain extent, the existing solutions still rely on external compensation modules and fail to achieve efficient optimization inside the optical chip, resulting in insufficient system integration.

[0005] In terms of data modulation, the existing technology usually uses a Mach-Zehnder interferometer (MZI) for phase modulation of optical signals. The MZI modulator is based on the principle of optical interference and controls the phase change through an electrical signal to achieve optical carrier modulation of data. This method has improved the signal modulation rate to a certain extent, but there are still many problems. First, the stability of the MZI modulator is limited and it is easily affected by temperature drift and environmental noise, resulting in signal distortion. Second, it is difficult for the MZI modulator to be efficiently integrated with the wavelength division multiplexing (WDM) technology, leading to the inability to fully utilize the overall computing power of the optical chip. In addition, the optical signal after MZI modulation still needs to be processed by an external DSP or FPGA, further increasing the system complexity and power consumption.

[0006] At the data receiving end, the existing methods for loading data on optical chips also face challenges. Due to the dispersion effect and non-linear effects (such as self-phase modulation SPM, cross-phase modulation XPM, four-wave mixing FWM) during fiber optic transmission, the optical signal will be distorted after long-distance transmission, resulting in an increase in the bit error rate. Traditional compensation methods include dispersion compensation fiber (DCF), digital signal processing (DSP), and coherent detection technology, but these solutions usually rely on external hardware units and do not provide effective optimization at the optical chip level. In addition, the data demodulation process of existing optical chips still relies on optoelectronic conversion and fails to directly achieve data recovery in the optical domain, further limiting the real-time performance and energy efficiency of the system.

[0007] The defects of the existing optical chip data loading technology are mainly reflected in the following aspects: First, the bandwidth utilization rate is low. Traditional optoelectronic conversion methods cannot fully utilize the high bandwidth characteristics of optical fibers, resulting in limited data transmission rates. Second, the system power consumption is relatively high. Due to frequent optoelectronic conversion and multi-stage signal processing, the overall energy consumption is large, especially in long-distance transmission or high-throughput application scenarios. In addition, the interface compatibility is poor. There is a lack of a unified interface standard between the optical chip and peripheral devices (such as FPGA, MCU), resulting in high integration difficulty and increased development costs. Fourth, the transmission stability is insufficient. During long-distance transmission, the dispersion and non-linear effects cause the signal quality to decline, affecting data integrity and reliability. Finally, the system architecture is complex. Existing optical communication systems rely on multiple independent signal processing units, resulting in a complex overall design and difficulty in implementing a low-power, high-integration optimization scheme.

[0008] The present invention proposes a method for data loading based on an optical chip. By integrating technologies such as FPGA for optical signal processing, Mach-Zehnder interferometer (MZI) modulation, wavelength division multiplexing (WDM), and parallel computing on the optical chip, efficient data modulation, transmission, and demodulation are achieved. This method integrates a data processing unit inside the optical chip, reducing the dependence on external optoelectronic conversion modules, and improving the integration and energy efficiency of the system. In addition, the data modulation process is optimized through the parallel computing ability of the optical chip, improving the signal quality and transmission stability. The implementation of the present invention will effectively solve problems such as bandwidth bottlenecks, high power consumption, and poor interface compatibility in existing optical chip data loading technologies, providing a better solution for future high-speed optical communication, data centers, and high-performance computing.

[0009] Therefore, how to provide a method for predicting big data device failures based on machine learning is an urgent problem that needs to be solved by those skilled in the art. Summary of the Invention

[0010] An object of the present invention is to propose a method for data loading based on an optical chip. This method combines advanced technologies such as wavelength division multiplexing (WDM), Mach-Zehnder interferometer (MZI) modulation, FPGA parallel processing, and optical computing. By integrating a data processing unit inside the optical chip, efficient modulation, transmission, and demodulation of multi-wavelength optical signals are achieved, optimizing the performance of the optical communication system. The system uses an FPGA to generate high-speed digital signals and perform digital-to-analog conversion, conducts coherent modulation through an MZI, and optimizes the data modulation process by combining the parallel computing ability of the optical chip to achieve seamless data loading and low-power consumption transmission. At the same time, this method adopts optical domain signal processing technology, reducing the dependence on external DSP and optoelectronic conversion modules, improving the system integration, reducing energy consumption, and enhancing the stability and compatibility of the system. Compared with traditional optical communication systems, the present invention has significant advantages in high-bandwidth data transmission, low-power consumption computing, and multi-wavelength signal processing, can meet the needs of data centers, high-performance computing, and intelligent network devices, and provides important technical support for the efficient development of future optical communication systems.

[0011] A method for loading data on an optical chip according to an embodiment of the present invention includes the following steps:

[0012] S1. Generate multiple optical signals with different wavelengths through a multi-wavelength light source, and use a wavelength demultiplexer to separate the generated multi-wavelength optical signals into multiple independent wavelength light beams;

[0013] S2. Generate a control signal and digital data by an FPGA or a single-chip microcomputer, and convert the digital data into an analog signal through an AD9764 digital-to-analog converter;

[0014] S3. Couple the analog signal with each independent wavelength beam, and use a Mach-Zehnder interferometer to modulate each beam to achieve the loading of the electrical signal onto the optical signal;

[0015] S4. Use a wavelength division multiplexer to recombine the modulated independent wavelength beams to form a multi-wavelength modulated optical signal;

[0016] S5. Input the multi-wavelength modulated optical signal into the optical chip, and perform data loading and calculation on the optical signal through an optical logic and parallel processing module;

[0017] S6. Transmit the multi-wavelength modulated optical signal processed by the optical chip to the remote receiving unit through a low-loss optical fiber;

[0018] S7. At the receiving end, use a wavelength demultiplexer to decompose the wavelength of the transmitted optical signal, separate it into multiple individual wavelength optical signals, and use a Mach-Zehnder interferometer to modulate and recover the signal for each optical signal;

[0019] S8. Convert the optical signal processed by the receiving end into an electrical signal, and complete the restoration of the data through a data demodulation and information recovery module, thereby realizing data loading and transmission.

[0020] Optionally, the specific steps of S1 include:

[0021] S11. Provide a multi-wavelength light source that can generate optical signals of different wavelengths. The wavelength range of the optical signals is λ1, λ2, …, λ n , where n is the number of channels of the optical signal, and λ i , i = 1, 2, …, n. Each wavelength corresponds to an independent optical signal channel;

[0022] S12. Perform preliminary power equalization on the optical signal output by the light source so that the power P(λ i ) of each wavelength optical signal satisfies:

[0023]

[0024] where ΔP is the set power equalization threshold;

[0025] S13. Input the optical signal into a wavelength demultiplexer, and separate the input multi-wavelength optical signal into individual wavelength beams through optical filtering and wavelength division technology. Each beam corresponds to a single wavelength λ i , and the output optical signal power satisfies:

[0026] P out (λ i ) = T DeMux (λ i ) · P in(λ i );

[0027] wherein, T DeMux (λ i ) is the transmission coefficient of the wavelength by the wavelength demultiplexer, P in (λ i ) is the input power, and P out (λ i ) is the output power;

[0028] S14. Perform optical signal shaping on the n separated independent wavelength beams, and control the optical power level of each beam to satisfy the set optical power range:

[0029] P min ≤ P(λ i ) ≤ P max ;

[0030] wherein, P min and P max are the set minimum and maximum power thresholds respectively;

[0031] S15. Output the shaped optical signal for signal modulation, data loading, and transmission processing.

[0032] Optionally, the S2 specifically includes:

[0033] S21. Generate a digital control signal D(t) by an FPGA or a single-chip microcomputer. The signal D(t) represents the data to be loaded onto the optical signal, and the signal adopts an N-bit parallel data format, where D(t) = {d1, d2,..., d N}, d i ∈ {0, 1};

[0034] S22. Sample the digital control signal D(t) and perform serial-to-parallel conversion at the set signal rate f s to obtain the corresponding high-speed digital signal D s (t):

[0035]

[0036] N represents the number of bits of the digital signal, and D s (t) is used for subsequent digital-to-analog conversion processing;

[0037] S23. Input the high-speed digital signal D s (t) into the AD9764 digital-to-analog converter to convert and obtain the corresponding analog signal S analog (t):

[0038]

[0039] Among them, V ref is the reference voltage of AD9764;

[0040] S24. Filter the analog signal S analog (t) to remove high-frequency noise using a low-pass filter to obtain a smoothed signal S f (t):

[0041] S f (t) = S analog (t) * h LPF (t);

[0042] Among them, h LPF (t) represents the impulse response of the low-pass filter, and * represents the convolution operation;

[0043] S25. Adjust the power of the filtered analog signal S f (t) to ensure that the signal power P S satisfies:

[0044] P min ≤ P S ≤ P max ;

[0045] Among them, P min and P max are the set minimum and maximum power thresholds respectively;

[0046] S26. Send the adjusted analog signal S f (t) to the input end of the Mach-Zehnder interferometer.

[0047] Optionally, the specific content of S3 includes:

[0048] S31. Input the independent wavelength light beam λ i , i = 1, 2,..., n into the Mach-Zehnder interferometer, and the light beam power satisfies:

[0049] P min ≤ P(λ i ) ≤ P max ;

[0050] Among them, P min and P max are the set minimum and maximum optical power thresholds;

[0051] S32. Input the analog signal S f (t) into the modulation electrode of the Mach-Zehnder interferometer for phase modulation of the optical signal;

[0052] S33. Inside the Mach-Zehnder interferometer, two optical couplers are used to split and combine the incident light respectively, and the modulation process satisfies:

[0053] E out (t) = E in (t)·e j*(t) ;

[0054] where E in (t) is the electric field strength of the incident optical signal, E out (t) is the electric field strength of the output optical signal, and φ(t) is the phase change determined by the analog signal S f (t). The relationship of φ(t) is as follows:

[0055]

[0056] where V π is the half-wave voltage of the Mach-Zehnder interferometer;

[0057] S34. The phase-modulated optical signal is recombined at the second optical coupler to form the optical signal E mod (t) loaded with data:

[0058]

[0059] S35. Adjust the power of the modulated optical signal E mod (t) to ensure that the output optical power satisfies:

[0060] P min ≤ P mod ≤ P max ;

[0061] where P mod is the power of the modulated optical signal;

[0062] S36. Output the modulated optical signal E mod (t) with adjusted power to the wavelength division multiplexer for optical signal multiplexing and transmission.

[0063] Optionally, the S4 specifically includes:

[0064] S41. Receive the modulated optical signal E mod,i (t), where i = 1, 2,..., n, and each optical signal corresponds to an independent wavelength λ i , and ensure that its power satisfies:

[0065] P min ≤ P mod,i ≤ P max ;

[0066] Among them, P mod,i is the power of the modulated optical signal corresponding to the wavelength λ i , P min and P max are respectively the set minimum and maximum power thresholds;

[0067] S42. Input the modulated optical signal E mod,i (t) into the wavelength division multiplexer, and use optical multiplexing technology to combine the modulated optical signals of multiple wavelengths λ1, λ2, …, λ n to form a multi-wavelength multiplexed optical signal E MUX (t), and its mathematical expression is:

[0068]

[0069] Among them, E MUX (t) is the multiplexed optical signal, which contains the modulated optical signals of n independent wavelengths;

[0070] S43. Perform optical power equalization on the multi-wavelength multiplexed optical signal E MUX (t) to ensure that the optical power of all channels is maintained within the target range and satisfies:

[0071]

[0072] Among them, ΔP is the set optical power equalization threshold;

[0073] S44. Perform spectral optimization processing on the multi-wavelength multiplexed optical signal E MUX (t), and use the spectral equalization filter H eq (λ) to compensate the optical power of different wavelengths to ensure the uniformity of the multiplexed signal and satisfy:

[0074] E MUX,eq (t) = E MUX (t) * H eq (λ);

[0075] Among them, E MUX,eq (t) is the multiplexed optical signal after spectral optimization;

[0076] S45. Perform power adjustment on the multi-wavelength multiplexed optical signal E MUX,eq (t) so that the output optical power satisfies:

[0077] P MUX,min ≤P MUX ≤P MUX,max ;

[0078] Among them, P MUX,min and P MUX,maxThey are the minimum and maximum power thresholds of the set multiplexed optical signal respectively;

[0079] S46. Output the adjusted multi-wavelength multiplexed optical signal E MUX,eq (t) to the input end of the optical chip for optical data processing and parallel computing.

[0080] Optionally, the S5 specifically includes:

[0081] S51. Receive the multi-wavelength multiplexed optical signal E MUX,eq (t). The optical signal E MUX,eq (t) includes n different wavelengths λ i , i = 1, 2,..., n, and satisfies the power range:

[0082] P MUX,min ≤P MUX ≤P MUX,max ;

[0083] Among them, P MUX is the power of the multiplexed optical signal, and P MUX,min and P MUX,max are the set minimum and maximum power thresholds respectively;

[0084] S52. Input the optical signal E MUX,eq (t) into the optical chip, and use the optical computing unit of the optical chip to perform parallel processing on the optical signal E MUX,eq (t), so that the optical signals E MUX,eq (t) with different wavelengths are data-carried and calculated on multiple optical channels in the optical chip. The optical computing model satisfies:

[0085] E chip (t) = f chip (E MUX,eq (t), W, B);

[0086] Among them, E chip (t) is the output optical signal after the optical chip calculation, f chip (·) is the internal calculation function of the optical chip, and W and B respectively represent the optical computing weight matrix and the bias term;

[0087] S53. Perform coherent detection on the optical signal E chip (t) after being processed by the optical chip, extract the optical intensity information using a photodetector, calculate the output optical power P chip , and ensure that it satisfies:

[0088] P chip,min ≤P chip ≤P chip,max ;

[0089] Among them, P chip,minand P chip,max are respectively the minimum power and the maximum power of the optical signal output by the optical chip set;

[0090] S54. Perform wavelength routing on the optical signal calculated inside the optical chip to ensure that optical signals of different wavelengths are allocated to corresponding optical channels according to the set path. The wavelength routing function satisfies:

[0091] λ out,i = R(λ in,i );

[0092] where λ out,i is the output wavelength after calculation by the optical chip, λ in,i is the input wavelength, and R(·) is the wavelength routing function of the optical chip;

[0093] S55. Perform power normalization on the optical signal calculated by the optical chip to ensure that all output optical signals are stable. The normalization calculation satisfies:

[0094]

[0095] where P norm,i is the optical power after normalization, is the total optical power of all channels;

[0096] S56. Output the optically-signal E chip,norm (t) with normalized power to the output port of the optical chip for optical signal transmission and remote reception processing.

[0097] Optionally, the specific content of S6 includes:

[0098] S61. Receive the optical signal E chip,norm (t). The optical signal E chip,norm (t) contains n wavelengths λ i , i = 1, 2,..., n, and its power satisfies:

[0099] P chip,min ≤ P chip,norm,i ≤ P chip,max ;

[0100] where P chip,norm,i is the optical signal power corresponding to the wavelength λ i , P chip,min and P chip,max are respectively the set minimum and maximum power thresholds;

[0101] S62. Input the optical signal E chip,norm (t) into the optical fiber transmission channel and use a low-loss optical fiber to achieve long-distance transmission. The total power loss of the optical fiber transmission system satisfies:

[0102] P loss = P chip - P fiber,out ;

[0103] where P loss is the loss during optical fiber transmission, and P fiber,out is the optical power at the output end of the optical fiber;

[0104] S63. Perform dispersion compensation on the optical signal E chip,norm (t) in the optical fiber, and use a dispersion compensation module to correct the broadening of the optical pulse. The optical signal E disp (t) after dispersion compensation satisfies:

[0105] E disp (t) = E chip,norm (t) * h disp (t);

[0106] where h disp (t) is the impulse response of the dispersion compensation filter, and * represents the convolution operation;

[0107] S64. Perform non - linear compensation on the optical signal E disp (t), and use optical non - linear equalization technology to correct the non - linear effects during transmission. After the optical signal is compensated, it satisfies:

[0108]

[0109] where, and respectively represent the Fourier transform and the inverse Fourier transform, L is the total length of the optical fiber, ω is the angular frequency of the optical signal, E disp (t) is the optical signal after dispersion compensation, ω m is the weight parameter of the neural network, · represents scalar multiplication, b is the bias term, and M is the length of the filtering window;

[0110] S65. Amplify the optical signal E NL (t) during transmission, and use an erbium - doped fiber amplifier to perform gain compensation on the optical signal. The power of the amplified signal satisfies:

[0111] P EDFA,i = G EDFA · P fiber,out,i

[0112] where P EDFA,i is the optical power after amplification, and G EDFA is the gain of the optical amplifier;

[0113] S66. Output the optical signal E NL (t) after gain compensation to the receiving - end optical fiber port for optical signal demultiplexing and data recovery processing.

[0114] Optionally, the S7 specifically includes:

[0115] S71. Receive the optical signal E NL (t), where the optical signal E NL (t) contains n wavelengths λ i , i = 1, 2,..., n, and ensure that the signal power satisfies:

[0116] P EDFA,min ≤ P EDFA,i ≤ P EDFA,max ;

[0117] where P EDFA,i is the optical power corresponding to the wavelength λ i , P EDFA,min and P EDFA,max are the set minimum and maximum power thresholds, respectively;

[0118] S72. Use a wavelength division multiplexer to decompose the wavelength of the optical signal E NL (t), and split the multi-wavelength multiplexed signal into n independent single-wavelength signals E DeMux,i (t). The decomposed signal satisfies:

[0119] E DeMux,i (t) = H DeMux (λ i ) · E NL (t);

[0120] where H DeMux (λ i ) is the transfer function of the wavelength division multiplexer;

[0121] S73. Perform coherent detection on each separated optical signal E DeMux,i (t), and use a coherent receiver to extract the amplitude and phase information of the optical signal. The coherent detection signals I cho,i (t) and Q coh,i (t) satisfy:

[0122] I cho,i (t) = |E DeMux,i (t)| cos(θ i (t));

[0123] Q coh,i (t) = |E DeMux,i (t)| sin(θ i (t));

[0124] where θ i (t) is the phase information of the optical signal;

[0125] S74. Input the coherent detection signals I cho,i (t) and Q coh,i (t) into a Mach-Zehnder interferometer. Convert the optical signal into an electrical signal through an optoelectronic conversion unit. The converted signal S PD,i (t) satisfies:

[0126] S PD,i (t) = R PD ·(I cho,i (t) 2 +Q cho,i (t) 2 );

[0127] Where R PD is the responsivity of the photodetector;

[0128] S75. Perform digital processing on the signal S PD,i (t) after optoelectronic conversion. Sample the signal using an analog-to-digital converter. The discretized digital signal D ADC,i [k]:

[0129]

[0130] Where T s is the sampling period, V ADC is the reference voltage of the ADC, and N is the number of quantization bits;

[0131] S76. Input the digital signal D ADC,i [k] into a data recovery module. Use digital signal processing algorithms for signal equalization, symbol decision, and bit error rate correction, and finally recover the transmitted data D rec,i [k]:

[0132] D rec,i [k] = f DPS (D ADC,i [k]);

[0133] Where f DPS (·) is a digital signal processing function, including equalization filtering H EQ (k), symbol decision f dec (k), and bit error rate compensation f FEC (k). The specific calculations are as follows:

[0134] D EQ,i [k] = H EQ (k) * D ADC,i [k];

[0135] D dec,i [k] = f dec (D EQ,i [k]);

[0136] D rec,i [k]=f FEC (D dec,i [k]);

[0137] Among them, for f dec (k) symbol decision, f FEC (k) is used for forward error correction;

[0138] S77. Transmit the demodulated data D recover,i (t) to the data storage and processing unit for data storage, analysis and application.

[0139] Optionally, the S8 specifically includes:

[0140] S81. Receive the demodulated data D recover,i (t), where i = 1, 2,..., n, and the data D recover,i (t) is obtained by converting the optically coherent demodulated and error-compensated signal:

[0141] D recover,i (t)=f demod (E DSP,i (t));

[0142] Among them, f demod (·) is the data demodulation function:

[0143] f demod (E DSP,i (t)) = Threshold(E DSP,i (t));

[0144] Among them, Threshold(·) represents the threshold decision function;

[0145] S82. Perform error detection on the recovered data D recover,i (t), and use cyclic redundancy check or hash check code to calculate the bit error rate. The bit error rate calculation formula is as follows:

[0146]

[0147] Among them, N eror is the number of detected error bits, N total is the total number of bits, and the error detection function f BER (·):

[0148]

[0149] Among them, D ref,i (k) is the reference data bit stream, and ⊕ represents the exclusive OR operation;

[0150] S83. Perform error correction on the error - code data. Forward error - correction codes are used for error correction, and the corrected data D corr,i (t) is calculated as follows:

[0151] D corr,i (t) = f FEC (D recover,i (t));

[0152] where f FEC (·) is the forward error - correction function:

[0153] f FEC (D recover,i (t)) = D recover,i (t) ⊕ Syndrome(D recover,i (t));

[0154] where Syndrome(·) is the parity - check synthesis calculation function;

[0155] S84. Classify the corrected data D corr,i (t). According to the allocation rule of the wavelength λ i , map the data to different storage queues The storage - queue rules are as follows:

[0156]

[0157] where the data - classification function f classify (·) is expressed as:

[0158] f classify (D corr,i (t)) = Hash(λ i ) mod M;

[0159] where Hash(λ i ) is the wavelength hash function and M is the total number of storage queues;

[0160] S85. Perform format conversion on the data in the storage queue . Use a digital - signal processing module to perform signal - coding conversion so that the data meets the protocol requirements of different application scenarios. The conversion formula is as follows:

[0161] D final,i (t) = f encode (D corr,i (t));

[0162] where the data - encoding function f encode (·) is expressed as:

[0163] f encode (Dcorr,i (t)) = LDPC_Encode(D corr,i (t));

[0164] Among them, LDPC_Encode(·) represents the low-density parity-check encoding function;

[0165] S86. Output the data D final,i (t) after format conversion to the target storage device or computing system for further processing or analysis, and the stored data format satisfies:

[0166]

[0167] Among them, S is the finally stored data set, which contains data of all wavelength channels.

[0168] The beneficial effects of the present invention are as follows:

[0169] First of all, by integrating a data processing unit inside the optical chip, the present invention realizes the efficient modulation, transmission, and demodulation of multi-wavelength optical signals, breaking through the dependence on external optoelectronic conversion modules in traditional optical communication systems. A high-speed digital signal is generated by the FPGA, and combined with the AD9764 digital-to-analog converter to convert the digital signal into an analog signal, ensuring the efficient loading of data in the optical domain. Using the Mach-Zehnder interferometer (MZI) for coherent modulation, the system can independently modulate optical signals of different wavelengths, and at the same time, the wavelength-division multiplexing (WDM) technology is used to combine the modulated signals to achieve more efficient multi-channel parallel data transmission. Compared with the existing optical communication systems, the present invention directly completes data modulation and signal multiplexing in the optical domain, avoiding the frequent optoelectronic conversion process, significantly reducing power consumption, and at the same time improving the transmission bandwidth and data throughput capacity of the system.

[0170] Secondly, by utilizing the parallel computing ability of the optical chip, the present invention optimizes the data modulation and transmission processes, effectively reducing the complexity and latency problems of the system. Traditional optical communication systems usually rely on external digital signal processing (DSP) units for equalization and error correction during data modulation and demodulation, while the present invention integrates a computing unit inside the optical chip and uses optical matrix operations to modulate and normalize the input optical signals, reducing the need for external processing units. Combining power equalization and optical signal optimization algorithms, the system can dynamically adjust the power levels of different wavelength signals to ensure the stability of signal transmission quality. Compared with the traditional scheme, the present invention realizes the optimization of data modulation and transmission at the chip level, greatly improving the integration degree of the optical communication system and reducing the energy consumption of signal processing.

[0171] Furthermore, through low-loss optical fiber transmission and optimization of optical signal stability, the present invention solves the stability problem of the existing optical chip data loading method in long-distance transmission. During signal transmission, the system adopts a dispersion compensation algorithm and an adaptive nonlinear equalization mechanism, effectively reducing the dispersion effect and nonlinear distortion, and improving the stability of optical signals in long-distance transmission. In addition, the system utilizes optical domain power control technology to dynamically adjust the power level of optical signals during transmission, ensuring the consistency of different wavelength signals at the receiving end and avoiding signal attenuation and crosstalk problems. Compared with the prior art, the present invention can maintain a low bit error rate during long-distance optical signal transmission, improving the reliability and efficiency of data transmission.

[0172] In addition, the present invention adopts wavelength division demultiplexing (DeMUX) technology at the receiving end to separate multi-wavelength modulated optical signals into independent single-wavelength signals, and performs demodulation and coherent detection through a Mach-Zehnder interferometer (MZI) to ensure the accurate recovery of optical signals. Combining the optoelectronic conversion and data decoding optimization strategies, the system can quickly and accurately convert optical signals into electrical signals, and through adaptive error correction and data recovery algorithms, improve the accuracy of data demodulation. Compared with the traditional optoelectronic conversion demodulation method, the present invention reduces the dependence on external DSP processing units during the data recovery process, while reducing the system complexity, improving the efficiency and stability of data recovery.

[0173] Finally, the present invention realizes an efficient, low-power, and stable optical chip data loading method. Through optical chip-level data modulation, wavelength division multiplexing and demultiplexing, optical signal optimization, and improvement of long-distance transmission stability, it effectively solves the bottlenecks of existing optical chip data transmission schemes in terms of bandwidth, power consumption, compatibility, and transmission stability. This method not only improves the application ability of optical chips in high-performance computing, data centers, cloud computing, and other fields, but also has broad application prospects in future optical communication systems. Description of the Drawings

[0174] The drawings are used to provide a further understanding of the present invention, and constitute a part of the specification. Together with the embodiments of the present invention, they are used to explain the present invention, and do not constitute a limitation to the present invention. In the drawings:

[0175] Figure 1 is the overall flowchart of a method for loading data on an optical chip proposed by the present invention;

[0176] Figure 2 is the framework diagram of the data loading and transmission system based on an optical chip in the present invention. Detailed Embodiments

[0177] The present invention will now be described in further detail with reference to the accompanying drawings. These drawings are all simplified schematic diagrams, only illustrating the basic structure of the present invention in a schematic manner, and therefore only showing the components related to the present invention.

[0178] Reference Figure 1 and Figure 2 , a method for carrying data on an optical chip, comprising the following steps:

[0179] S1. Generate multiple optical signals with different wavelengths through a multi-wavelength light source, and use a wavelength division multiplexer to separate the generated multi-wavelength optical signals into multiple independent wavelength light beams;

[0180] S2. Generate a control signal and digital data by an FPGA or a single-chip microcomputer, and convert the digital data into an analog signal through an AD9764 digital-to-analog converter;

[0181] S3. Couple the analog signal with each independent wavelength light beam, and use a Mach-Zehnder interferometer to modulate each light beam to achieve the loading of the electrical signal onto the optical signal;

[0182] S4. Use a wavelength division multiplexer to recombine the modulated independent wavelength light beams to form a multi-wavelength modulated optical signal;

[0183] S5. Input the multi-wavelength modulated optical signal into the optical chip, and perform data loading and calculation on the optical signal through an optical logic and parallel processing module;

[0184] S6. Transmit the multi-wavelength modulated optical signal processed by the optical chip to a remote receiving unit through a low-loss optical fiber;

[0185] S7. At the receiving end, use a wavelength division multiplexer to decompose the wavelength of the transmitted optical signal, separate it into multiple individual wavelength optical signals, and use a Mach-Zehnder interferometer to modulate and signal recovery process each optical signal;

[0186] S8. Convert the optical signal processed by the receiving end into an electrical signal, and complete the restoration of the data through a data demodulation and information recovery module, thereby realizing data loading and transmission.

[0187] In this embodiment, the specific content of S1 includes:

[0188] S11. Provide a multi-wavelength light source, which can generate optical signals with different wavelengths. The wavelength range of the optical signals is λ1, λ2, …, λ n , where n is the number of channels of the optical signal, and λ i , i = 1, 2, …, n, and each wavelength corresponds to an independent optical signal channel;

[0189] S12. Perform preliminary power equalization on the optical signal output by the light source so that the power P(λ i ) of each wavelength optical signal satisfies:

[0190]

[0191] where ΔP is the set power equalization threshold;

[0192] S13. Input the optical signal into an optical demultiplexer, and separate the input multi-wavelength optical signal into n independent wavelength light beams through optical filtering and wavelength division technology. Each light beam corresponds to a single wavelength λ i , and the power of the output optical signal satisfies:

[0193] P out (λ i ) = T DeMux (λ i ) · P in (λ i );

[0194] where T DeMux (λ i ) is the transmission coefficient of the optical demultiplexer for the wavelength, P in (λ i ) is the input power, and P out (λ i ) is the output power;

[0195] S14. Perform optical signal shaping on the n separated independent wavelength light beams, and control the optical power level of each light beam so that it satisfies the set optical power range:

[0196] P min ≤ P(λ i ) ≤ P max ;

[0197] where P min and P max are the set minimum and maximum power thresholds respectively;

[0198] S15. Output the shaped optical signal for signal modulation, data loading, and transmission processing.

[0199] In this embodiment, the specific steps of S2 include:

[0200] S21. Generate a digital control signal D(t) by an FPGA or a single-chip microcomputer. The signal D(t) represents the data to be loaded onto the optical signal. The signal adopts an N-bit parallel data format, where D(t) = {d1, d2,..., d N}, and d i ∈ {0, 1};

[0201] S22. Sample the digital control signal D(t) and perform serial-to-parallel conversion at the set signal rate f s to obtain the corresponding high-speed digital signal D s (t):

[0202]

[0203] N represents the number of bits of the digital signal, and D s (t) is used for subsequent digital-to-analog conversion processing;

[0204] S23. Input the high-speed digital signal D s (t) into the AD9764 digital-to-analog converter to convert and obtain the corresponding analog signal S analog (t):

[0205]

[0206] where V ref is the reference voltage of the AD9764;

[0207] S24. Filter the analog signal S analog (t), and use a low-pass filter to remove high-frequency noise to obtain a smoothed signal S f (t):

[0208] S f (t) = S analog (t) * h LPF (t);

[0209] where h LPF (t) represents the impulse response of the low-pass filter, and * represents the convolution operation;

[0210] S25. Adjust the power of the filtered analog signal S f (t) to ensure that the signal power P S satisfies:

[0211] P min ≤ P S ≤ P max ;

[0212] where P min and P max are the set minimum and maximum power thresholds respectively;

[0213] S26. Send the adjusted analog signal S f (t) to the input end of the Mach-Zehnder interferometer.

[0214] In this embodiment, the specific content of S3 includes:

[0215] S31. Input the independent wavelength beam λ i , i = 1, 2, …, n into the Mach-Zehnder interferometer, and the beam power satisfies:

[0216] P min ≤ P(λ i ) ≤ P max ;

[0217] where P min and P max are the set minimum and maximum optical power thresholds;

[0218] S32. Input the analog signal S f (t) into the modulation electrode of the Mach-Zehnder interferometer for phase modulation of the optical signal;

[0219] S33. Inside the Mach-Zehnder interferometer, use two optical couplers to split and combine the incident light respectively, and the modulation process satisfies:

[0220] E out (t) = E in (t) · e jφ(t) ;

[0221] where E in (t) is the electric field strength of the incident optical signal, E out (t) is the electric field strength of the output optical signal, φ(t) is the phase change determined by the analog signal S f (t), and the relationship of φ(t) is as follows:

[0222]

[0223] where V π is the half-wave voltage of the Mach-Zehnder interferometer;

[0224] S34. The phase-modulated optical signal recombines at the second optical coupler to form the optical signal E mod (t) loaded with data:

[0225]

[0226] S35. Adjust the power of the modulated optical signal E mod (t) to ensure that the output optical power satisfies:

[0227] P min ≤ P mod ≤ P max ;

[0228] where P modis the power of the modulated optical signal;

[0229] S36. Output the modulated optical signal E mod (t) to the wavelength division multiplexer for optical signal multiplexing and transmission.

[0230] In this embodiment, the said S4 specifically includes:

[0231] S41. Receive the modulated optical signal E mod,i (t), where i = 1, 2,..., n, and each optical signal corresponds to an independent wavelength λ i , and ensure that its power satisfies:

[0232] P min ≤P mod,i ≤P max ;

[0233] where, P mod,i is the power of the modulated optical signal corresponding to the wavelength λ i , P min and P max are respectively the set minimum and maximum power thresholds;

[0234] S42. Input the modulated optical signal E mod,i (t) to the wavelength division multiplexer, and use optical multiplexing technology to combine the modulated optical signals of multiple wavelengths λ1, λ2,..., λ n to form a multi-wavelength multiplexed optical signal E MUX (t), and its mathematical expression is:

[0235]

[0236] where, E MUX (t) is the multiplexed optical signal, including the modulated optical signals of n independent wavelengths;

[0237] S43. Perform optical power equalization on the multi-wavelength multiplexed optical signal E MUX (t) to ensure that the optical power of all channels remains within the target range and satisfies:

[0238]

[0239] where, ΔP is the set optical power equalization threshold;

[0240] S44. Perform spectral optimization processing on the multi-wavelength multiplexed optical signal E MUX (t), and use the spectral equalization filter H eq (λ) to compensate the optical power of different wavelengths to ensure the uniformity of the multiplexed signal and satisfy:

[0241] EMUX,eq E(t) = MUX E(t) * H(λ); eq

[0242] Wherein, E MUX,eq (t) is the multiplexed optical signal after spectral optimization;

[0243] S45. Adjust the power of the multi-wavelength multiplexed optical signal E MUX,eq (t) so that the output optical power satisfies:

[0244] P MUX,min ≤ P MUX ≤ P MUX,max ;

[0245] Wherein, P MUX,min and P MUX,max are respectively the set minimum and maximum power thresholds of the multiplexed optical signal;

[0246] S46. Output the adjusted multi-wavelength multiplexed optical signal E MUX,eq (t) to the input end of the optical chip for optical data processing and parallel computing.

[0247] In this embodiment, the specific steps of S5 include:

[0248] S51. Receive the multi-wavelength multiplexed optical signal E MUX,eq (t). The optical signal E MUX,eq (t) contains n different wavelengths λ i , i = 1, 2,..., n, and satisfies the power range:

[0249] P MUX,min ≤ P MUX ≤ P MUX,max ;

[0250] Wherein, P MUX is the power of the multiplexed optical signal, and P MUX,min and P MUX,max are respectively the set minimum and maximum power thresholds;

[0251] S52. Input the optical signal E MUX,eq (t) into the optical chip, and use the optical computing unit of the optical chip to perform parallel processing on the optical signal E MUX,eq (t), so that the optical signals E MUX,eq (t) with different wavelengths perform data loading and computing on multiple optical channels in the optical chip. The optical computing model satisfies:

[0252] E chip (t) = f chip (E MUX,eq (t), W, B); ​

[0253] Among them, E chip (t) is the output optical signal after optical chip calculation, and f chip (·) is the internal calculation function of the optical chip. W and B respectively represent the optical calculation weight matrix and the bias term;

[0254] S53. Perform coherent detection on the optical signal E chip (t) after being processed by the optical chip, extract the optical intensity information using a photodetector, and calculate the output optical power P chip , and ensure that it satisfies:

[0255] P chip,min ≤P chip ≤P chip,max ;

[0256] Among them, P chip,min and P chip,max are respectively the minimum power and the maximum power of the output optical signal of the set optical chip;

[0257] S54. Perform wavelength routing on the optical signal obtained by internal calculation of the optical chip to ensure that different wavelength signals are allocated to the corresponding optical channels according to the set path. The wavelength routing function satisfies:

[0258] λ out,i =R(λ in,i );

[0259] Among them, λ out,i is the output wavelength after optical chip calculation, λ in,i is the input wavelength, and R(·) is the optical chip wavelength routing function;

[0260] S55. Perform power normalization on the optical signal after optical chip calculation to ensure that all output optical signals are stable. The normalization calculation satisfies:

[0261]

[0262] Among them, P norm,i is the normalized optical power, is the sum of the optical powers of all channels;

[0263] S56. Output the optically-signal E chip,norm (t) after power normalization to the output port of the optical chip for optical signal transmission and remote reception processing.

[0264] In this embodiment, the S6 specifically includes:

[0265] S61. Receive the optical signal E chip,norm (t), and the optical signal E chip,norm (t) contains n wavelengths λ i, where \(i = 1, 2, \ldots, n\), and their powers satisfy:

[0266] P chip,min \(\leq P\) chip,norm,i \(\leq P\) chip,max ;

[0267] Among them, \(P\) chip,norm,i is the optical signal power corresponding to the wavelength \(\lambda\) i , \(P\) chip,min and \(P\) chip,max are respectively the set minimum and maximum power thresholds;

[0268] S62. Input the optical signal \(E\) chip,norm (t) into the optical fiber transmission channel, and use low-loss optical fiber to achieve long-distance transmission. The total power loss of the optical fiber transmission system satisfies:

[0269] P loss \(= P\) chip - P fiber,out ;

[0270] Among them, \(P\) loss is the loss during optical fiber transmission, and \(P\) fiber,out is the optical power at the output end of the optical fiber;

[0271] S63. Perform dispersion compensation on the optical signal \(E\) chip,norm (t) in the optical fiber. Use a dispersion compensation module to correct the broadening of optical pulses. The optical signal \(E\) disp (t) after dispersion compensation satisfies:

[0272] E disp (t) \(= E\) chip,norm (t) * h disp (t);

[0273] Among them, \(h\) disp (t) is the impulse response of the dispersion compensation filter, and * represents the convolution operation;

[0274] S64. Perform non-linear compensation on the optical signal \(E\) disp (t). Use optical non-linear equalization technology to correct the non-linear effects during transmission. After the optical signal is compensated, it satisfies:

[0275]

[0276] Among them, and respectively represent the Fourier transform and the inverse Fourier transform, \(L\) is the total length of the optical fiber, \(\omega\) is the angular frequency of the optical signal, \(E\) disp (t) is the optical signal after dispersion compensation, \(\omega\) m weight parameters of the neural network, · represents scalar multiplication, \(b\) is the bias term, and \(M\) is the length of the filtering window;

[0277] S65. Amplify the optical signal E NL (t) during transmission, and use an erbium-doped fiber amplifier to perform gain compensation on the optical signal. The power of the amplified signal satisfies:

[0278] P EDFA,i = G EDFA ·P fiber,out,i

[0279] where P EDFA,i is the optical power after amplification, and G EDFA is the gain of the optical amplifier;

[0280] S66. Output the gain-compensated optical signal E NL (t) to the receiving end fiber port for optical signal demultiplexing and data recovery processing.

[0281] In this embodiment, the S7 specifically includes:

[0282] S71. Receive the optical signal E NL (t). The optical signal E NL (t) contains n wavelengths λ i , i = 1, 2,..., n, and ensure that the signal power satisfies:

[0283] P EDFA,min ≤ P EDFA,i ≤ P EDFA,max ;

[0284] where P EDFA,i is the optical power corresponding to the wavelength λ i , P EDFA,min and P EDFA,max are respectively the set minimum and maximum power thresholds;

[0285] S72. Use a wavelength demultiplexer to perform wavelength decomposition on the optical signal E NL (t), and split the multi-wavelength multiplexed signal into n independent single-wavelength signals E DeMux,i (t). The decomposed signal satisfies:

[0286] E DeMux,i (t) = H DeMux (λ i )·E NL (t);

[0287] where H DeMux (λ i ) is the transfer function of the wavelength demultiplexer;

[0288] S73. For each separated optical signal E DeMux,i(t) Perform coherent detection, and use a coherent receiver to extract the amplitude and phase information of the optical signal. The coherent detection signal I cho,i (t) and Q coh,i (t) satisfy:

[0289] I cho,i (t) = |E DeMux,i (t)| cos(θ i (t));

[0290] Q coh,i (t) = |E DeMux,i (t)| sin(θ i (t));

[0291] where θ i (t) is the phase information of the optical signal;

[0292] S74. Input the coherent detection signals I cho,i (t) and Q coh,i (t) into a Mach-Zehnder interferometer, convert the optical signal into an electrical signal through a photoelectric conversion unit, and the converted signal S PD,i (t) satisfies:

[0293] S PD,i (t) = R PD · (I cho,i (t) 2 + Q cho,i (t) 2 );

[0294] where R PD is the responsivity of the photodetector;

[0295] S75. Perform digital processing on the signal S PD,i (t) after photoelectric conversion. Sample the signal using an analog-to-digital converter, and the discretized digital signal D ADC,i [k]:

[0296]

[0297] where T s is the sampling period, V ADC is the reference voltage of the ADC, and N is the number of quantization bits;

[0298] S76. Input the digital signal D ADC,i [k] into a data recovery module, perform signal equalization, symbol decision, and bit error rate correction using digital signal processing algorithms, and finally recover the transmitted data D rec,i [k]:

[0299] D rec,i[k]=f DPS (D ADC,i [k]);

[0300] where f DPS (·) is a digital signal processing function, including equalization filtering H EQ (k), symbol decision f dec (k), and bit error rate compensation f FEC (k), and the specific calculation is as follows:

[0301] D EQ,i [k]=H EQ (k)*D ADC,i [k];

[0302] D dec,i [k]=f dec (D EQ,i [k]);

[0303] D rec,i [k]=f FEC (D dec,i [k]);

[0304] where f dec (k) is for symbol decision, and f FEC (k) is for forward error correction;

[0305] S77. Transmit the demodulated data D recover,i (t) to the data storage and processing unit for data storage, analysis, and application.

[0306] In this embodiment, the S8 specifically includes:

[0307] S81. Receive the demodulated data D recover,i (t), where i = 1, 2,..., n, and the data D recover,i (t) is obtained by converting the optically coherent demodulated and error-compensated signal:

[0308] D recover,i (t)=f demod (E DSP,i (t));

[0309] where f demod (·) is a data demodulation function:

[0310] f demod (E DSP,i (t))=Threshold(E DSP,i (t));

[0311] where Threshold(·) represents a threshold decision function;

[0312] S82. Perform error detection on the restored data D recover,i (t), and calculate the bit error rate using cyclic redundancy check or hash check code. The formula for calculating the bit error rate is as follows:

[0313]

[0314] where N eror is the number of detected error bits, and N total is the total number of bits. The error detection function f BER (·):

[0315]

[0316] where D ref,i (k) is the reference data bit stream, and ⊕ represents the exclusive OR operation;

[0317] S83. Perform error correction on the error data. Use forward error correction code for error correction. The formula for the corrected data D corr,i (t) is as follows:

[0318] D corr,i (t) = f FEC (D recover,i (t));

[0319] where f FEC (·) is the forward error correction function:

[0320] f FEC (D recover,i (t)) = D recover,i (t) ⊕ Syndrome(D recover,i (t));

[0321] where Syndrome(·) is the parity check comprehensive calculation function;

[0322] S84. Classify the corrected data D corr,i (t). According to the allocation rule of the wavelength λ i , map the data to different storage queues The storage queue rules are as follows:

[0323]

[0324] where the data classification function f classify (·) is expressed as:

[0325] f classify (D corr,i (t)) = Hash(λ i ) mod M;

[0326] Among them, Hash(λ i ) is a wavelength hash function, and M is the total number of storage queues;

[0327] S85. Perform format conversion on the data in the storage queue , and use a digital signal processing module to perform signal encoding conversion to make the data meet the protocol requirements of different application scenarios. The conversion formula is as follows:

[0328] D final,i (t) = f encode (D corr,i (t));

[0329] Among them, the data encoding function f encode (·) is expressed as:

[0330] f encode (D corr,i (t)) = LDPC_Encode(D corr,i (t));

[0331] Among them, LDPC_Encode(·) represents a low-density parity-check encoding function;

[0332] S86. Output the data D final,i (t) after format conversion to a target storage device or computing system for further processing or analysis. The storage data format satisfies:

[0333]

[0334] Among them, S is the finally stored data set, which contains data of all wavelength channels.

[0335] Example 1:

[0336] To verify the feasibility of the present invention in practical applications, the present invention is applied to the optical communication system of a large data center. This data center mainly serves high-performance computing (HPC) and artificial intelligence (AI) training tasks, and has extremely high requirements for the bandwidth, power consumption, and stability of data transmission. The traditional method of loading data on optical chips has obvious performance bottlenecks in this environment, mainly manifested in low utilization rate of data transmission bandwidth, high power consumption in the optoelectronic conversion process, increased bit error rate due to the influence of dispersion and nonlinear effects on optical signals during transmission, and interface compatibility issues between optical chips and peripheral devices (such as FPGA, MCU). Therefore, in the high-speed optical communication system of this data center, the method of loading data based on optical chips of the present invention is adopted to optimize the data modulation, transmission, and demodulation processes, improve data transmission efficiency, and reduce power consumption.

[0337] During the application process, an optical chip is used as the core signal processing unit inside the data center. The FPGA generates high-speed digital signals, and combined with the AD9764 digital-to-analog converter, the digital signals are converted into analog signals to ensure the stable loading of data. Subsequently, the optical signal is coherently modulated through a Mach-Zehnder interferometer (MZI) to achieve the optical carrier transmission of data. After the optical signal modulation is completed, the wavelength-division multiplexing (WDM) technology is used to combine multi-wavelength signals to improve the fiber transmission efficiency, and the data is transmitted through a low-loss optical fiber. At the receiving end, a wavelength demultiplexer (DeMUX) is used to analyze the multi-wavelength signals, and the MZI is used for demodulation and coherent detection to ensure the accurate recovery of the optical signal. In addition, combined with the adaptive error correction and data recovery algorithms, the demodulation process of the optical signal is optimized to improve the accuracy of the data.

[0338] To verify the advantages of the present invention, the performance of the optical communication system in the data center was tested under the traditional optical chip data loading method and the method of the present invention respectively, and the key performance indicators were recorded, including bandwidth utilization, system power consumption, bit error rate, signal delay, transmission stability, and device compatibility. The test period was from January 2024 to June 2024, and the test location was a large data center in a first-tier city in China.

[0339] In terms of bandwidth utilization, the average bandwidth utilization of the traditional optical chip data loading method was about 62.3%, while the average bandwidth utilization of the method of the present invention reached 91.7% in the same environment, an increase of 47.2%. This shows that the method of the present invention can utilize the transmission capacity of the optical fiber more efficiently and improve the data throughput. In terms of system power consumption, the average power consumption of the traditional method was 9.5W / channel, while the average power consumption of the method of the present invention dropped to 6.1W / channel, with obvious energy-saving effects, a overall reduction of 35.8%.

[0340] In the bit error rate test, after the traditional method was used for long-distance (500km) optical fiber transmission, the average bit error rate was 7.2×10 -6 , while the bit error rate of the method of the present invention was only 2.1×10 -6 , a reduction of 70.8%, indicating that the present invention is more stable in signal modulation and demodulation and can reduce signal loss. In terms of signal delay, the average end-to-end delay of the traditional method was 12.8ms, while the average delay of the method of the present invention dropped to 8.5ms, a reduction of 33.6%, effectively improving the real-time communication ability of the data center. In addition, in the compatibility test, the method of the present invention was compatible with 97.6% of the FPGA devices, while the compatibility of the traditional method was only 81.2%, indicating that the method of the present invention has better adaptability under different hardware platforms.

[0341] In the transmission stability test, by monitoring the system uptime over a long period (continuous operation for 6 months), the system uptime ratio of the traditional method was 99.21%, while that of the method of the present invention reached 99.93%. The system reliability was greatly improved, and the task failure rate caused by data transmission errors was significantly reduced. The application of the present invention effectively improved the stability of the data center in the ultra-large-scale data transmission environment, significantly reduced the data loss caused by signal distortion, and improved the reliability and operation efficiency of the overall computing tasks.

[0342] From the above test results, it can be seen that the data loading method based on optical chips of the present invention is superior to the traditional method in multiple key indicators such as bandwidth utilization, power consumption, bit error rate, signal delay, transmission stability, and device compatibility, and can provide a more efficient and stable optical communication solution in the fields of high-performance computing, data centers, and cloud computing.

[0343] Table 1 Experimental data of performance comparison of the optical chip data loading method

[0344]

[0345] The above data prove that the data loading method based on optical chips of the present invention is significantly superior to the traditional method in multiple key performance indicators, providing reliable technical support for the optimization and upgrade of future high-performance optical communication systems.

Claims

1. A method for carrying data on an optical chip, characterized in that It includes the following steps: S1. Generate optical signals of multiple different wavelengths through a multi-wavelength light source, and use a wavelength division multiplexer to separate the generated multi-wavelength optical signals into multiple independent wavelength light beams; S2. Generate a control signal and digital data by an FPGA or a single-chip microcomputer, and convert the digital data into an analog signal through an AD9764 digital-to-analog converter; S3. Couple the analog signal with each independent wavelength light beam, and use a Mach-Zehnder interferometer to modulate each light beam to realize the loading of an electrical signal onto an optical signal; S4. Use a wavelength division multiplexer to recombine the modulated independent wavelength light beams to form a multi-wavelength modulated optical signal; S5. Input the multi-wavelength modulated optical signal into the optical chip, and perform data loading and calculation on the optical signal through an optical logic and parallel processing module; S6. Transmit the multi-wavelength modulated optical signal processed by the optical chip to a remote receiving unit through a low-loss optical fiber; S7. At the receiving end, use a wavelength division multiplexer to decompose the transmitted optical signal by wavelength, separate it into multiple individual wavelength optical signals, and use a Mach-Zehnder interferometer to modulate and perform signal recovery processing on each optical signal; S8. Convert the optical signal processed at the receiving end into an electrical signal, and complete the restoration of the data through a data demodulation and information recovery module, so as to realize data loading and transmission.

2. The method for carrying optical chip data according to claim 1, wherein The specific content of S1 includes: S11. Provide a multi-wavelength light source. The light source can generate optical signals of different wavelengths, and the wavelength range of the optical signals is λ1, λ2, …, λ n , where n is the number of channels of the optical signals, and λ i , i = 1, 2, …, n. Each wavelength corresponds to an independent optical signal channel; S12. Perform preliminary power equalization on the optical signals output by the light source so that the power P(λ i ) of each wavelength optical signal satisfies: where ΔP is a set power equalization threshold; S13. Input the optical signal into an optical demultiplexer, and separate the input multi-wavelength optical signal into individual wavelength beams through optical filtering and wavelength division techniques. Each beam corresponds to a single wavelength λ i , and the output optical signal power satisfies: P out (λ i ) = T DeMux (λ i )·P in (λ i ); Among them, T DeMux (λ i ) is the transmission coefficient of the wavelength by the wavelength demultiplexer, P in (λ i ) is the input power, P out (λ i ) is the output power; S14. Perform optical signal shaping on the separated n independent wavelength light beams, and control the optical power level of each light beam to make it meet the set optical power range: P min ≤P(λ i )≤P max ; where P min and P max are the set minimum and maximum power thresholds, respectively; S15. Output the shaped optical signal for signal modulation, data loading, and transmission processing.

3. The method for carrying optical chip data according to claim 1, characterized in that The specific content of S2 includes: S21. Generate a digital control signal D(t) by an FPGA or a microcontroller. The signal D(t) represents the data to be loaded onto the optical signal, and the signal adopts an N-bit parallel data format, where D(t) = {d1, d2, …, d N}, d i ∈ {0, 1}; S22. Sample the digital control signal D(t) and perform serial-to-parallel conversion at the set signal rate f s to obtain the corresponding high-speed digital signal D s (t): where N represents the number of bits of the digital signal, and D s (t) is used for subsequent digital-to-analog conversion processing; S23. Input the high-speed digital signal D s (t) into the AD9764 digital-to-analog converter to obtain the corresponding analog signal S analog (t): Among them, V ref is the reference voltage of AD9764; S24. Filter the analog signal S analog (t), and use a low-pass filter to remove high-frequency noise to obtain a smoothed signal S f (t): S f S(t) = analog S(t) * h LPF (t); where h LPF (t) represents the impulse response of the low-pass filter, and * represents the convolution operation; S25. Perform power adjustment on the filtered analog signal S f (t) to ensure that the signal power P S satisfies: P min ≤P S ≤P max ; Among them, P min and P max are respectively the set minimum and maximum power thresholds; S26. Send the adjusted analog signal S f (t) to the input end of the Mach-Zehnder interferometer.

4. A method for carrying optical chip data according to claim 1, characterized in that The specific content of S3 includes: S31. Input the independent wavelength beam λ i , where i = 1, 2, …, n, into the Mach-Zehnder interferometer, and the beam power satisfies: i , i = 1, 2, …, n into the Mach-Zehnder interferometer, and the beam power satisfies: P min ≤P(λ i )≤P max ; where P min and P max are the set minimum and maximum optical power thresholds; S32. Input the analog signal S f (t) into the modulation electrode of the Mach-Zehnder interferometer for phase modulation of the optical signal; S33. Inside the Mach-Zehnder interferometer, use two optical couplers to split and combine the incident light respectively, and the modulation process satisfies: E out f(t) = E in f(t) · e jφ(t) ; Among them, E in (t) is the electric field strength of the incident optical signal, and E out (t) is the electric field strength of the output optical signal. φ(t) is the phase change determined by the analog signal S f (t). The relationship of φ(t) is as follows: where V π is the half-wave voltage of the Mach-Zehnder interferometer; S34. The phase-modulated optical signal is recombined at the second optical coupler to form the data-loaded optical signal E mod (t): S35. Adjust the power of the modulated optical signal E mod (t) to ensure that the output optical power meets the requirement: P min ≤P mod ≤P max ; where P mod is the power of the modulated optical signal; S36. Output the modulated optical signal E mod (t) after power adjustment to the wavelength division multiplexer for optical signal multiplexing and transmission.

5. A method for loading optical chip data according to claim 1, characterized in that The specific content of S4 includes: S41. Receive the modulated optical signal E mod,i (t), where i = 1, 2, …, n, and each optical signal corresponds to an independent wavelength λ i , and ensure that its power satisfies: P min ≤P mod,i ≤P max ; Among them, P mod,i is the power of the modulated optical signal corresponding to the wavelength λ i , P min and P max are respectively the set minimum and maximum power thresholds; S42. Input the modulated optical signal E mod,i (t) into a wavelength division multiplexer, and use optical multiplexing technology to combine the modulated optical signals of multiple wavelengths λ1, λ2, …, λ n to form a multi-wavelength multiplexed optical signal E MUX (t), and its mathematical expression is: Among them, E MUX (t) is the multiplexed optical signal, which contains n modulated optical signals with independent wavelengths; S43. Perform optical power equalization on the multi-wavelength multiplexed optical signal E MUX (t) to ensure that the optical powers of all channels are maintained within the target range, satisfying: where ΔP is a set optical power equalization threshold; S44. Perform spectral optimization processing on the multi-wavelength multiplexed optical signal E MUX (t), and use the spectral equalization filter H eq (λ) to compensate the optical powers of different wavelengths to ensure the uniformity of the multiplexed signal, satisfying: E MUX,eq E(t) = MUX E(t)*H eq (λ); Among them, E MUX,eq (t) is the multiplexed optical signal after spectral optimization; S45. Adjust the power of the multi-wavelength multiplexed optical signal E MUX,eq (t) so that the output optical power meets the following condition: P MUX,min ≤P MUX ≤P MUX,max ; Among them, P MUX,min and P MUX,max are respectively the minimum and maximum power thresholds of the multiplexed optical signal set; S46. Output the adjusted multi-wavelength multiplexed optical signal E MUX,eq (t) to the input end of the optical chip for optical data processing and parallel computing.

6. A method for carrying optical chip data according to claim 1, characterized in that, The specific content of S5 includes: S51. Receive a multi-wavelength multiplexed optical signal E MUX,eq (t), where the optical signal E MUX,eq (t) includes n different wavelengths λ i , i = 1, 2, …, n, and satisfies the power range: P MUX,min ≤P MUX ≤P MUX,max ; Among them, P MUX is the power of the multiplexed optical signal, P MUX,min and P MUX,max are respectively the set minimum and maximum power thresholds; S52. Input the optical signal E MUX,eq (t) into the optical chip, and use the optical computing unit of the optical chip to perform parallel processing on the optical signal E MUX,eq (t), so that optical signals E MUX,eq (t) with different wavelengths perform data loading and calculation on multiple optical channels in the optical chip. The optical computing model satisfies: E chip (t) = f chip (E MUX,eq (t), W, B); Among them, E chip (t) is the output optical signal after optical chip calculation, and f chip (·) is the internal calculation function of the optical chip. W and B represent the optical calculation weight matrix and the bias term respectively; S53. Coherently detect the optical signal E chip (t) after being processed by the optical chip, extract the optical intensity information using a photodetector, and calculate the output optical power P chip , and ensure that the following is satisfied: P chip,min ≤P chip ≤P chip,max ; Among them, P chip,min and P chip,max are respectively the minimum power and the maximum power of the optical signal output by the set optical chip; S54. Perform wavelength routing on the optical signal calculated inside the optical chip to ensure that different wavelength signals are allocated to the corresponding optical channels according to the set path, and the wavelength routing function satisfies: λ out,i = R(λ in,i ); Among them, λ out,i is the output wavelength after optical chip calculation, and λ in,i is the input wavelength. R(·) is the optical chip wavelength routing function; S55. Perform power normalization on the optical signal calculated by the optical chip to ensure that all output optical signals are stable, and the normalization calculation satisfies: Among them, P norm,i is the normalized optical power, and is the sum of the optical powers of all channels; The optical signal E after power normalization chip,norm (t) is output to the output port of the optical chip for optical signal transmission and remote reception processing.

7. A method for carrying optical chip data according to claim 1, characterized in that The specific content of S6 includes: S61. Receive the optical signal E chip,norm (t), where the optical signal E chip,norm (t) contains n wavelengths λ i , i = 1, 2, …, n, and its power satisfies: P chip,min ≤P chip,norm,i ≤P chip,max ; Among them, P chip,norm,i is the optical signal power corresponding to the wavelength λ i , P chip,min and P chip,max are respectively the set minimum and maximum power thresholds; S62. Input the optical signal E chip,norm (t) into the optical fiber transmission channel, and use low-loss optical fiber to achieve long-distance transmission. The total power loss of the optical fiber transmission system satisfies: P loss = P chip -P fiber,out ; Among them, P loss is the loss during optical fiber transmission, and P fiber,out is the optical power at the output end of the optical fiber; S63. Disperse compensation is performed on the optical signal E chip,norm in the optical fiber. A dispersion compensation module is used to correct the broadening of the optical pulse. The optical signal E disp (t) after dispersion compensation satisfies: E disp I(t) = E chip,norm (t) * h disp (t); where h disp (t) is the impulse response of the dispersion compensation filter, and * represents the convolution operation; S64. Nonlinearly compensate the optical signal E disp (t), and use optical nonlinear equalization technology to correct the nonlinear effects during transmission. After the optical signal compensation, it satisfies: Among them, and represent the Fourier transform and the inverse Fourier transform respectively, L is the total length of the optical fiber, ω is the angular frequency of the optical signal, E disp (t) is the optical signal after dispersion compensation, ω m is the weight parameter of the neural network, · represents scalar multiplication, b is the bias term, and M is the length of the filtering window; S65. Amplify the optical signal E NL (t) during the transmission process, and use an erbium-doped fiber amplifier to perform gain compensation on the optical signal. The power of the amplified signal satisfies: P EDFA,i = G EDFA ·P fiber,out,i Among them, P EDFA,i is the amplified optical power, and G EDFA is the optical amplifier gain; S66. Output the optical signal E NL (t) after gain compensation to the receiving end optical fiber port for optical signal demultiplexing and data recovery processing.

8. A method for loading optical chip data according to claim 1, characterized in that, The specific content of S7 includes: S71. Receive the optical signal E NL (t), where the optical signal E NL (t) contains n wavelengths λ i , i = 1, 2, …, n, and ensure that the signal power satisfies: P EDFA,min ≤P EDFA,i ≤P EDFA,max ; wherein, P EDFA,i is the optical power corresponding to the wavelength λ i , P EDFA,min and P EDFA,max are respectively the set minimum and maximum power thresholds; S72. Using a wavelength demultiplexer to perform wavelength decomposition on the optical signal E NL (t), splitting the multi-wavelength multiplexed signal into n independent single-wavelength signals E DeMux,i (t), and the decomposed signal satisfies: E DeMux,i E(t) = H DeMux (λ i )·E NL (t); Among them, H DeMux (λ i ) is the transfer function of the wavelength demultiplexer; S73. For each separated optical signal E DeMux,i (t), perform coherent detection, and use a coherent receiver to extract the amplitude and phase information of the optical signal. The coherent detection signals I cho,i (t) and Q coh,i (t) satisfy: I cho,i (t) = |E DeMux,i (t)| cos(θ i (t)); Q coh,i E(t) = |E DeMux,i (t)| sin(θ i (t)); where θ i (t) is the phase information of the optical signal; S74. Input the coherent detection signals I cho,i (t) and Q coh,i (t) into a Mach-Zehnder interferometer, convert the optical signal into an electrical signal through an optoelectronic conversion unit, and the converted signal S PD,i (t) satisfies: S PD,i S(t) = R PD ·(I cho,i (t) 2 +Q cho,i (t) 2 ); Among them, R PD Responsivity of the photodetector; S75. For the signal S PD,i (t) after photoelectric conversion, perform digital processing, and sample the signal using an analog-to-digital converter. The discretized digital signal D ADC,i [k]: Among them, T s is the sampling period, V ADC is the reference voltage of the ADC, and N is the number of quantization bits; S76. Input the digital signal D ADC,i [k] into the data recovery module, and perform signal equalization, symbol decision, and bit error rate correction using digital signal processing algorithms, and finally recover the transmitted data D rec,i [k]: D rec,i [k] = f DPS (D ADC,i [k]); Among them, f DPS (·) is a digital signal processing function, including equalization filtering H EQ (k), symbol decision f dec (k) and bit error rate compensation f FEC (k), and the specific calculation is as follows: D EQ,i [k] = H EQ (k) * D ADC,i [k]; D dec,i [k] = f dec (D EQ,i [k]); D rec,i [k] = f FEC (D dec,i [k]); Among them, for f dec (k) symbol decision, f FEC (k) for forward error correction; S77. Send the demodulated data D recover,i (t) to the data storage and processing unit for data storage, analysis, and application.

9. A method for loading optical chip data according to claim 1, characterized in that The specific content of S8 includes: S81. Receive the demodulated data D recover,i (t), where i = 1, 2, …, n, and the data D recover,i (t) is obtained by converting the optical signal after coherent demodulation and error compensation: D recover,i (t) = f demod (E DSP,i (t)); where f demod (·) is the data demodulation function: f demod (E DSP,i (t)) = Threshold(E DSP,i (t)); where Threshold(·) represents a threshold decision function; S82. Perform error detection on the restored data D recover,i (t), and use cyclic redundancy check or hash check code to calculate the bit error rate. The bit error rate calculation formula is as follows: Among them, N eror is the number of error bits detected, and N total is the total number of bits. The error detection function f BER (·): Among them, D ref,i (k) is the reference data bitstream, represents an exclusive OR operation; S83. Error correction processing is performed on the error data, and forward error correction code is used for error correction. The corrected data D corr,i (t) is calculated as follows: D corr,i ψ(t) = f FEC (D recover,i (t)); where f FEC (·) is a forward error correction function: where Syndrome(·) is a parity check comprehensive calculation function; S84. Classify the corrected data D corr,i (t), and map the data to different storage queues according to the wavelength λ i allocation rule The storage queue rules are as follows: Among them, the data classification function f classify (·) is expressed as: f classify (D corr,i (t)) = Hash(λ i ) mod M; Among them, Hash(λ i ) is a wavelength hash function, and M is the total number of storage queues; S85. Perform format conversion on the data in the storage queue Q λi and perform signal encoding conversion using a digital signal processing module to make the data meet the protocol requirements of different application scenarios. The conversion formula is as follows: D final,i (t) = f encode (D corr,i (t)); Among them, the data encoding function f encode (·) is expressed as: f encode (D corr,i (t)) = LDPC_Encode(D corr,i (t)); where LDPC_Encode(·) represents a low-density parity-check coding function; S86. Output the data D final,i (t) after format conversion to the target storage device or computing system for further processing or analysis, and the stored data format satisfies: where S is the finally stored data set, which contains the data of all wavelength channels.

Images