A large-scale optical phased array phase mixing calibration method

Abstract

The present application relates to the technical field of integrated photonics, and specifically relates to a large-scale optical phased array phase hybrid calibration method, comprising: S1, establishing a calibration test system comprising a phased array chip to be calibrated, a laser, a polarization controller, a calibration computer, a Fourier imaging system and an infrared camera; S2, setting the iteration number K, the gain factor alpha and the attenuation factor beta; S3, using golden section search to perform single-channel per-element coarse tuning, iteratively locking the quasi-optimal phase of each array element, and obtaining an initial phase vector; S4, based on the SPGD parallel algorithm, generating a Bernoulli distribution disturbance vector with the initial phase vector as the starting point, calculating a dynamic step length in combination with an evaluation function, and synchronously iteratively updating the phases of all array elements; S5, cyclically iterating and judging the convergence condition, and outputting the final calibration phase after the condition is met. The present application effectively improves the calibration efficiency and accuracy through the hybrid calibration architecture combining coarse tuning and fine tuning, has strong robustness, and is suitable for industrial calibration of large-scale optical phased arrays.

Description

Technical Field

[0001] This invention relates to the field of integrated photonics technology, and more specifically to a method for phase mixing calibration of large-scale optical phased arrays. Background Technology

[0002] Optical phased arrays, based on the principle of coherent optical interference, achieve rapid beam scanning and pointing control without mechanical inertia through precise electronic control of the phase and amplitude of array channels. Compared to traditional mechanical and MEMS beam control schemes, they possess core advantages such as being purely solid-state, wear-free, having fast response (nanosecond to microsecond level), high angular resolution, small size, and easy chip-level integration. They can maintain high stability and long lifespan even in harsh environments. With the maturity of integrated optoelectronic technologies such as silicon-based photonics and thin-film lithium niobate, optical phased arrays are rapidly developing towards large-scale arrays, large field of view, low power consumption, and high beam quality. They have irreplaceable technological value and broad application prospects in fields such as lidar, free-space optical communication, and adaptive optics, and have become the mainstream development direction for next-generation high-precision beam control systems.

[0003] Ideally, by precisely applying a phase shift to each channel of an optical phased array that matches the target deflection angle, the light intensity of the far-field target angle light field can be coherently superimposed, the main lobe energy concentrated, and the sidelobe level effectively suppressed. However, in actual fabrication and operation, due to the non-ideal nature of the processing technology, random deviations exist in the waveguide width, thickness, and etching depth, introducing initial phase errors distributed in the range of 0 to 2π in each channel. These random phase errors severely damage the wavefront coherence of the array, directly leading to a decrease in peak power, a deterioration in the peak-to-sidelobe ratio, and a reduction in beam pointing accuracy and system performance. Therefore, phase calibration of the optical phased array is necessary before actual use.

[0004] Existing phase calibration methods for optical phased arrays generally employ neural network calibration: rapid calibration is achieved by constructing a mapping model between far-field speckle and phase error. Chinese invention patent application CN120355605A discloses a phase calibration method, apparatus, electronic device, and storage medium for optical phased arrays. The method includes: inputting a phase image to be calibrated into an optical phased array system to obtain a randomly perturbed phase image corresponding to the phase image; generating the phase image to be calibrated based on the initial emission electric field in the optical phased array; obtaining the randomly perturbed phase image by perturbing the phase image to be calibrated within a preset range; inputting the phase image to be calibrated and the randomly perturbed phase image into a target convolutional neural network model to determine the target phase error of the phase to be calibrated; training the target convolutional neural network model based on a pair of far-field training images; the far-field training image pair includes a far-field training image generated from the optical phased array and a far-field randomly perturbed phase image; and calibrating the initial emission electric field of the phase image to be calibrated based on the target phase error.

[0005] However, the above methods rely on large-scale, high-quality datasets, which are costly to collect and train. Furthermore, the trained models are sensitive to changes in physical parameters such as temperature and optical power, and lack robustness. Summary of the Invention

[0006] Therefore, the technical problem to be solved by the present invention is to overcome the shortcomings of high cost and insufficient robustness of existing optical phased array phase calibration methods, thereby providing a large-scale optical phased array phase hybrid calibration method.

[0007] A large-scale optical phased array phase mixing calibration method includes: Step S1: Establish a calibration test system: Combine the phased array chip to be calibrated, laser, polarization controller, calibration computer, Fourier imaging system and infrared camera to form a calibration test system; Step S2: Set basic parameters: Set the number of single-channel iterations K for the golden section search and the basic gain factor α and decay factor β for adaptive SPGD; Step S3: Golden Section Search Iteration: Starting from the second array element, perform single-channel calibration on the array element to be calibrated, apply phase shifts of 0.5π and 1.5π to the array element to be calibrated, acquire far-field images using an infrared camera, and calculate the corresponding evaluation functions J1 and J2. Based on J1 and J2, filter the search interval and apply the search interval according to the golden ratio. Energy feedback comparisons are performed at selected sampling points. After K iterations, the quasi-optimal initial phase value of the array element is locked. Adjust the corresponding array elements to The remaining array elements are iterated sequentially until all array elements have been iterated and adjusted to obtain the initial phase vector. ; Step S4: Iteration of the adaptive gain SPGD parallel algorithm based on real-time feedback: using the initial phase vector obtained in step S3 As the starting phase vector for parallel optimization in this stage Generate a perturbation vector that conforms to a Bernoulli distribution, apply positive and negative perturbations to it, and measure the change in the evaluation function. The system determines the dynamic step size of the current iteration based on the current evaluation function value, updates the phase values ​​of all array elements synchronously based on the dynamic step size and gradient estimation value, and detects the optimization effect through an infrared camera. It performs multiple rounds of global parallel iterative optimization and dynamically adjusts the gain step size based on the current feedback evaluation function value in each iteration. Step S5: Result Output: When the evaluation function value reaches the preset threshold or the number of iterations s reaches the maximum number of iterations, the search stops and the final calibration voltage value of each channel is output, thereby completing the phase calibration.

[0008] Furthermore: the evaluation function is a function for evaluating the diffraction efficiency of the optical phased array, and the specific evaluation function is as follows:

[0009] Parameter description: For diffraction efficiency; Deflect the target at an angle; This represents the far-field light intensity distribution. This represents the effective field of view of the optical phased array; The full width at half maximum (FWHM) of the diffraction peak is denoted as α.

[0010] Furthermore: the evaluation function is a function for evaluating the peak-to-sidelobe ratio of an optical phased array, and the specific evaluation function is as follows:

[0011] Parameter description: The largest side lobe; The power intensity of the main lobe is denoted by ; PSLR is the peak sidelobe ratio.

[0012] Further: Step S3 includes: S3.1: Pre-sampling logic determination: Starting from the second array element, perform single-channel calibration on the array element to be calibrated: keep the phase of other array elements unchanged, apply phase shifts of 0.5π and 1.5π respectively on this array element, acquire far-field images using an infrared camera, and calculate the corresponding evaluation functions J1(0.5π) and J2(1.5π); S3.2: Filter the search interval and compare the size of J1 and J2; if J1>J2, then set the initial search interval of the array element to [0,π]; if J1≤J2, then set the initial search interval to [π,2π]. S3.3: Golden Section Iteration: Within the selected π length interval, select according to the golden ratio. The sampling points are compared using the evaluation function feedback. After K iterations, the quasi-optimal initial phase value of the array element is locked. ; S3.4: Repeat steps S3.1 to S3.3 for the 3rd to Nth array elements in sequence until the point-by-point initial adjustment of all N channels of the entire array is completed, and combine the quasi-optimal initial phase values ​​locked in each channel into the initial phase vector. .

[0013] Furthermore, step S4 specifically includes the following steps: Step S4.1: Use the initial phase vector obtained in step S3. As the starting phase vector for parallel optimization in this stage Initialize the adaptive SPGD algorithm iteration count s=0 to obtain the initial phase vector. Then calculations were performed to obtain And stipulate ; Step S4.2: Generate a set of random phase perturbation vectors with dimensions consistent with the number of elements to be calibrated and conforming to a Bernoulli distribution. Each perturbation element takes the value of First, apply a positive disturbance to the system. The far-field images were acquired using an infrared camera, and the corresponding evaluation function value J was calculated. + Then apply a negative disturbance to the system. Acquire far-field images and calculate the corresponding evaluation function value J. - ; Calculate the change in the evaluation function ; Step S4.3: Calculate the adaptive gain step size: based on the evaluation function value of the current iteration. Through the formula of negative power function Calculate the dynamic gain step size for the current iteration round. ; Step S4.4: Based on dynamic gain step size Evaluation function change and random perturbation vector Once the gradient estimate is obtained, the phase values ​​of all array elements are updated synchronously according to the phase update formula. When the evaluation function is diffraction efficiency, the phase update formula is: When the evaluation function is the peak-to-sidelobe ratio, the phase update formula is: The iteration count s is incremented by 1, and the updated far-field image is acquired by an infrared camera to detect the beam optimization effect of this iteration. Step S4.5: Iterative Loop: Repeat steps S4.2 to S4.4 above to perform multiple rounds of global parallel iterative optimization. In each round of iteration, the gain step size is dynamically adjusted according to the current feedback evaluation function value until the evaluation function value reaches the preset threshold or the number of iterations s reaches the maximum number of iterations.

[0014] Furthermore, the number of iterations K in step S2 ranges from 4 to 10.

[0015] Further: in step S2, the basic gain factor α is taken as 0.42 to 0.58, and the attenuation factor β is taken as 0.68 to 0.75.

[0016] In the first stage, this invention employs a two-point sampling method. By comparing the evaluation functions of two specific phase points, the half-cycle interval where the peak is located is determined. Then, geometric reduction using the golden ratio is applied to reduce the number of iterations and improve optimization speed. However, due to residual phase errors in the first-stage optimization results, an adaptive gain SPGD parallel algorithm is introduced to further optimize the first-stage results. In the initial optimization phase with low evaluation function values, a large step size is maintained to achieve a high convergence rate. As the target extreme point approaches, the step size decreases with increasing function values, balancing the optimization requirements of convergence speed and stability. This invention achieves rapid calibration of optical phased arrays by using golden section search iteration for initial optimization and adaptive gain SPGD parallel algorithm iteration for further optimization. Compared to existing technologies, this invention offers faster calibration speed and higher accuracy, solving the problems of high cost and insufficient robustness in existing optical phased array phase calibration methods.

[0017] This invention uses diffraction efficiency as an evaluation criterion to judge the calibration status of optical phased arrays. Relatively speaking, the diffraction function is based on the principle of scalar diffraction optics and can completely characterize the light intensity distribution law of the far field of the optical phased array. It can comprehensively cover all energy distribution components such as main lobe, side lobe, grating lobe and stray light. This evaluation method can quantify beam diffraction efficiency from the perspective of energy distribution, objectively reflect the overall energy utilization level of the array, and identify implicit errors such as beam distortion, scanning offset and global energy loss.

[0018] This invention uses the sidelobe peak ratio as an evaluation criterion to judge the calibration status of the optical phased array. Relatively speaking, this evaluation index is simple to calculate, has clear judgment criteria, can quickly determine the convergence effect of phase calibration, is suitable as a quantitative criterion for closed-loop calibration, has strong engineering feasibility, and can intuitively reflect the interference suppression performance of the calibrated beam.

[0019] This invention employs a method of fixing the remaining array elements and calibrating each element independently, minimizing phase coupling interference between elements and improving the accuracy of single-channel phase calculation. By pre-sampling at two feature phase points at 0.5π and 1.5π, the optimal phase interval is quickly determined, compressing the search range to a π-length interval and eliminating the need for blind traversal of the entire interval, significantly reducing the computational load. Simultaneously, the golden ratio sampling iteration within the limited interval enables rapid approximation of the near-optimal phase of the array elements with minimal iterations, ensuring initial adjustment accuracy while shortening the calibration time for each element. After sequential initial adjustment of each element, a high-precision initial phase vector is formed, providing an excellent iterative starting point for subsequent global parallel optimization, reducing the iterative pressure of global optimization, and further accelerating the overall convergence speed.

[0020] This invention uses the phase vector obtained from the golden ratio initial adjustment as the starting value for iteration, avoiding the problems of iterative oscillation and slow convergence caused by random initial values, and improving the search probability of the global optimal solution. It employs a random perturbation vector conforming to the Bernoulli distribution for bidirectional positive and negative perturbation, resulting in strong stability of gradient estimation and adaptability to the optimization requirements of multi-channel synchronous perturbation in large-scale arrays. An adaptive dynamic step-size update mechanism is constructed based on a negative power function, replacing the traditional fixed-gain iteration mode. This achieves fast convergence with large step sizes in the early stages of iteration and fine optimization with small step sizes in the later stages, balancing convergence rate and calibration accuracy. Combined with real-time acquisition of far-field images by an infrared camera to detect the optimization effect, a closed-loop feedback calibration system is formed, which can monitor the beam optimization status in real time, facilitating timely identification of calibration anomalies and improving the stability of the algorithm.

[0021] This invention defines the iteration count range as 4-10. An iteration count K below this range results in too few sampling points, insufficient initial phase adjustment accuracy, and residual initial phase error. Conversely, an excessively high iteration count K leads to a large amount of redundant computation, prolonging the calibration time for large-scale arrays. This numerical range, while ensuring optimal initial phase adjustment accuracy for each array element, compresses computational resource consumption, achieving an optimal balance between calibration accuracy and computational efficiency, making it suitable for industrial calibration scenarios involving large batches and large-scale optical phased arrays.

[0022] This invention limits the range of the fundamental gain function α to 0.42 to 0.58 and the range of the attenuation factor β to 0.68 to 0.75. This parameter range ensures smooth gradient updates during algorithm iteration, avoids phase divergence and beam optimization oscillations, and results in a smooth and stable convergence curve. This effectively improves the robustness of the adaptive SPGD algorithm and further enhances the stability and practicality of this calibration method in engineering measurements. Attached Figure Description

[0023] To more clearly illustrate the specific embodiments of the present invention or the technical solutions in the prior art, the drawings used in the description of the specific embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are some embodiments of the present invention. For those skilled in the art, other drawings can be obtained from these drawings without creative effort.

[0024] Figure 1 This is a flowchart of the large-scale optical phased array phase mixing calibration method of the present invention; Figure 2 This is a flowchart of step S3 in the large-scale optical phased array phase mixing calibration method of the present invention; Figure 3 This is a flowchart of step S4 of the large-scale optical phased array phase mixing calibration method of the present invention. Detailed Implementation

[0025] The technical solution of the present invention will now be clearly and completely described with reference to the accompanying drawings. Obviously, the described embodiments are only some, not all, of the embodiments of the present invention. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0026] Furthermore, the technical features involved in the different embodiments of the present invention described below can be combined with each other as long as they do not conflict with each other.

[0027] A large-scale optical phased array phase mixing calibration method includes: Step S1: Establish a calibration test system: Combine the phased array chip to be calibrated, laser, polarization controller, calibration computer, Fourier imaging system and infrared camera to form a calibration test system; Step S2: Set basic parameters: Set the number of single-channel iterations K for the golden section search and the basic gain factor α and decay factor β for adaptive SPGD; Step S3: Golden Section Search Iteration: Starting from the second array element, perform single-channel calibration on the array element to be calibrated, apply phase shifts of 0.5π and 1.5π to the array element to be calibrated, acquire far-field images using an infrared camera, and calculate the corresponding evaluation functions J1 and J2. Based on J1 and J2, filter the search interval and apply the search interval according to the golden ratio. Energy feedback comparisons are performed at selected sampling points. After K iterations, the quasi-optimal initial phase value of the array element is locked. Adjust the corresponding array elements to The remaining array elements are iterated sequentially until all array elements have been iterated and adjusted to obtain the initial phase vector. ; Step S4: Iteration of the adaptive gain SPGD parallel algorithm based on real-time feedback: using the initial phase vector obtained in step S3 As the starting phase vector for parallel optimization in this stage Generate a perturbation vector that conforms to a Bernoulli distribution, apply positive and negative perturbations to it, and measure the change in the evaluation function. The system determines the dynamic step size of the current iteration based on the current evaluation function value, updates the phase values ​​of all array elements synchronously based on the dynamic step size and gradient estimation value, and detects the optimization effect through an infrared camera. It performs multiple rounds of global parallel iterative optimization and dynamically adjusts the gain step size based on the current feedback evaluation function value in each iteration. Step S5: Result Output: When the evaluation function value reaches the preset threshold or the number of iterations s reaches the maximum number of iterations, the search stops and the final calibration voltage value of each channel is output, thereby completing the phase calibration.

[0028] This invention employs a hybrid calibration architecture combining single-channel element-wise initial calibration with global parallel fine calibration. First, it uses a golden section search to perform initial phase coarse adjustment for each array element, obtaining a high-quality initial phase vector. Then, it utilizes an adaptive SPGD algorithm to achieve parallel iterative optimization across the entire array. This hybrid optimization mode overcomes the drawbacks of traditional single SPGD algorithms, such as reliance on random initial values, slow convergence, and susceptibility to local optima. It also solves the problems of high computational cost and long calibration time associated with full ergonomic search for large-scale arrays. Compared to existing technologies, this invention offers faster calibration speed and higher accuracy, addressing the issues of high cost and insufficient robustness in existing optical phased array phase calibration methods.

[0029] Example 1 (In this example, the diffraction efficiency function is used as the evaluation function) like Figure 1-3 As shown: A large-scale optical phased array phase mixing calibration method, comprising: Step S1: Establish a calibration test system: Combine the phased array chip to be calibrated, laser, polarization controller, calibration computer, Fourier imaging system and infrared camera to form a calibration test system; Step S2: Set basic parameters: Set the single-channel iteration parameter K for the golden section search and the basic gain factor α and attenuation factor β for adaptive SPGD. The iteration number K ranges from 4 to 10. A value below this range will result in too few sampling points, insufficient initial phase adjustment accuracy, and residual initial phase error; a value above this range will generate a large amount of redundant computation, prolonging the calibration time for large-scale arrays. This value range can compress computational resource consumption while ensuring optimal initial phase adjustment accuracy for a single array element, achieving the optimal balance between calibration accuracy and computational efficiency, and is suitable for industrial calibration scenarios involving large batches and large-scale optical phased arrays. In this example, K=5, with five iterations.

[0030] The basic gain function α is set to 0.58, and the attenuation factor β is set to 0.75. This set of parameters is the optimal combination to adapt to the phase modulation characteristics of optical phased arrays. It eliminates the need for repeated adjustments to the algorithm's basic parameters, reducing manual debugging costs and operational difficulty, and improving the algorithm's engineering adaptability. This parameter ratio ensures smooth gradient updates during algorithm iteration, avoiding phase iteration divergence and beam optimization oscillations, resulting in a smooth and stable convergence curve. This effectively improves the robustness of the adaptive SPGD algorithm and further enhances the stability and practicality of this calibration method in engineering measurements. Of course, those skilled in the art can adjust the values ​​of the basic gain function α and the attenuation factor β according to actual conditions (number of iterations and accuracy requirements, etc.).

[0031] Step S3: Includes the following steps: S3.1: Pre-sampling logic determination: Using the first array element as the reference, starting from the second array element, perform single-channel calibration on the array element to be calibrated. Keep the phases of other array elements unchanged, apply phase shifts of 0.5π and 1.5π to this array element respectively, acquire far-field images using an infrared camera, and calculate the corresponding evaluation function. (0.5π) and (1.5π).

[0032] In this embodiment, the evaluation function is the diffraction efficiency function, and its formula is as follows:

[0033] Parameter description: For diffraction efficiency, To the target deflection angle, For far-field light intensity distribution, The effective field of view of the optical phased array. This represents the full width at half maximum (FWHM) of the diffraction peak.

[0034] S3.2: Filter the search range and compare. and The size. If > If so, then the initial search interval for the array element is set to [0,π]; if ≤ Then the initial search interval is set to [π, 2π].

[0035] S3.3: Golden Section Iteration: Within the selected π length interval, select according to the golden ratio. Energy feedback comparison is performed at the sampling points, and after K iterations, the quasi-optimal initial phase value of the array element is locked.

[0036] S3.4: Repeat steps S3.1 to S3.3 for the 3rd to Nth array elements in sequence until the point-by-point initial adjustment of all N channels of the entire array is completed, and combine the quasi-optimal initial phase values ​​locked in each channel into the initial phase vector. .

[0037] Step S4: Includes the following steps: Step S4.1: Use the initial phase vector obtained in step S3. As the starting phase vector for parallel optimization in this stage Initialize the adaptive SPGD algorithm iteration count s=0 to obtain the initial phase vector. Then calculations were performed to obtain And stipulate ; Step S4.2: Generate a set of random phase perturbation vectors with dimensions consistent with the number of elements to be calibrated and conforming to a Bernoulli distribution. Each perturbation element takes the value of ( (Preset small disturbance amplitude); first apply a positive disturbance to the system. The infrared camera is used to acquire far-field images and calculate the corresponding evaluation function values. Then apply a negative disturbance to the system. Acquire far-field images and calculate the corresponding evaluation function values. Calculate the change in the evaluation function. .

[0038] Step S4.3: Calculate the adaptive gain step size: based on the evaluation function value of the current iteration. Through the formula of negative power function Calculate the dynamic gain step size for the current iteration round. ; Step S4.4: Based on dynamic gain step size Evaluation function change and random perturbation vector The gradient estimate is obtained, and the phase update formula is applied. The phase values ​​of all array elements are updated synchronously. The iteration count s is incremented by 1, and the updated far-field image is acquired using an infrared camera to detect the beam optimization effect of this iteration. Step S4.5: Iterative Loop: Repeat steps S4.2 to S4.4 above to perform multiple rounds of global parallel iterative optimization. In each round of iteration, the gain step size is dynamically adjusted according to the current feedback evaluation function value until the evaluation function value reaches the preset threshold or the number of iterations s reaches the maximum number of iterations.

[0039] Step S5: Result Output: When the evaluation function value reaches the preset threshold or the number of iterations s reaches the maximum number of iterations, the search stops and the final calibration voltage value of each channel is output, thereby completing the phase calibration.

[0040] Example 2 (In this example, the evaluation function uses the sidelobe peak ratio function as the evaluation function) like Figure 1-3 As shown: A large-scale optical phased array phase mixing calibration method, comprising: Step S1: Establish a calibration test system: Combine the phased array chip to be calibrated, laser, polarization controller, calibration computer, Fourier imaging system and infrared camera to form a calibration test system; Step S2: Set basic parameters: Set the single-channel iteration parameter K for the golden section search and the basic gain factor α and attenuation factor β for adaptive SPGD. The number of iterations K ranges from 4 to 10. A number of iterations below this range will result in too few sampling points, insufficient initial phase adjustment accuracy, and residual initial phase error; a number of iterations above this range will generate a large amount of redundant computation, prolonging the calibration time for large-scale arrays. This numerical range can compress computational resource consumption while ensuring optimal initial phase adjustment accuracy for a single array element, achieving the optimal balance between calibration accuracy and computational efficiency, and is suitable for industrial calibration scenarios involving large batches and large-scale optical phased arrays. In this example, K=7, and seven iterations are performed.

[0041] The basic gain function α is set to 0.42, and the attenuation factor β is set to 0.68. This set of parameters is the optimal combination to adapt to the phase modulation characteristics of optical phased arrays. It eliminates the need for repeated adjustments to the algorithm's basic parameters, reducing manual debugging costs and operational difficulty, and improving the algorithm's engineering adaptability. This parameter ratio ensures smooth gradient updates during algorithm iteration, avoiding phase iteration divergence and beam optimization oscillations, resulting in a smooth and stable convergence curve. This effectively improves the robustness of the adaptive SPGD algorithm and further enhances the stability and practicality of this calibration method in engineering measurements. Of course, those skilled in the art can adjust the values ​​of the basic gain function α and the attenuation factor β according to actual conditions (number of iterations and accuracy requirements, etc.).

[0042] Step S3: Includes the following steps: S3.1: Pre-sampling logic determination: Using the first array element as the reference, starting from the second array element, single-channel calibration is performed on the array element to be calibrated. Keeping the phases of other array elements unchanged, phase shifts of 0.5π and 1.5π are applied to this array element respectively. Far-field images are acquired using an infrared camera, and the corresponding evaluation function PSLR is calculated. db1 (0.5π) and PSLR db2 (1.5π).

[0043] In this embodiment, the evaluation function is the sidelobe peak ratio function, and its formula is as follows:

[0044] Parameter description: The largest side lobe; The power intensity of the main lobe is denoted by ; PSLR is the peak sidelobe ratio.

[0045] S3.2: Filter the search interval and compare the size of PSLR1 and PSLR2. If PSLR1>PSLR2, the initial search interval of the array element is set to [0,π]; if PSLR1≤PSLR2, the initial search interval is set to [π,2π].

[0046] S3.3: Golden Section Iteration: Within the selected π length interval, select according to the golden ratio. Energy feedback comparison is performed at the sampling points, and after K iterations, the quasi-optimal initial phase value of the array element is locked.

[0047] S3.4: Repeat steps S3.1 to S3.3 for the 3rd to Nth array elements in sequence until the point-by-point initial adjustment of all N channels of the entire array is completed, and combine the quasi-optimal initial phase values ​​locked in each channel into the initial phase vector. .

[0048] Step S4: Includes the following steps: Step S4.1: Use the initial phase vector obtained in step S3. As the starting phase vector for parallel optimization in this stage Initialize the adaptive SPGD algorithm iteration count s=0 to obtain the initial phase vector. Then calculations were performed to obtain And stipulate ; Step S4.2: Generate a set of random phase perturbation vectors with dimensions consistent with the number of elements to be calibrated and conforming to a Bernoulli distribution. Each perturbation element takes the value of ( (Preset small disturbance amplitude). First, apply a positive disturbance to the system. The infrared camera is used to acquire far-field images and calculate the corresponding evaluation function values. Then apply a negative disturbance to the system. Acquire far-field images and calculate the corresponding evaluation function values. Calculate the change in the evaluation function. ; Step S4.3: Calculate the adaptive gain step size: based on the evaluation function value of the current iteration. Through the formula of negative power function Calculate the dynamic gain step size for the current iteration round. ; Step S4.4: Based on dynamic gain step size Evaluation function change and random perturbation vector The gradient estimate is obtained, and the phase update formula is applied. The phase values ​​of all array elements are updated synchronously. The iteration count s is incremented by 1, and the updated far-field image is acquired using an infrared camera to detect the beam optimization effect of this iteration. Step S4.5: Iterative Loop: Repeat steps S4.2 to S4.4 above to perform multiple rounds of global parallel iterative optimization, and dynamically adjust the gain step size according to the current feedback evaluation function value in each iteration until the evaluation function value reaches the preset threshold or the maximum number of iterations is reached.

[0049] Step S5: Result Output: When the evaluation function value reaches the preset threshold or the number of iterations s reaches the maximum number of iterations, the search stops and the final calibration voltage value of each channel is output, thereby completing the phase calibration.

[0050] Obviously, the above embodiments are merely illustrative examples for clear explanation and are not intended to limit the implementation. Those skilled in the art will recognize that other variations or modifications can be made based on the above description. It is neither necessary nor possible to exhaustively list all possible implementations here. However, obvious variations or modifications derived therefrom are still within the scope of protection of this invention.

Claims

1. A method for phase mixing calibration of a large-scale optical phased array, characterized in that, include: Step S1: Establish a calibration test system: Combine the phased array chip to be calibrated, laser, polarization controller, calibration computer, Fourier imaging system and infrared camera to form a calibration test system; Step S2: Set basic parameters: Set the number of single-channel iterations K for the golden section search and the basic gain factor α and decay factor β for adaptive SPGD; Step S3: Golden Section Search Iteration: Starting from the second array element, perform single-channel calibration on the array element to be calibrated, apply phase shifts of 0.5π and 1.5π to the array element to be calibrated, acquire far-field images using an infrared camera, and calculate the corresponding evaluation functions J1 and J2. Based on J1 and J2, filter the search interval and apply the search interval according to the golden ratio. Energy feedback comparisons are performed at selected sampling points. After K iterations, the quasi-optimal initial phase value of the array element is locked. Adjust the corresponding array elements to The remaining array elements are iterated sequentially until all array elements have been iterated and adjusted to obtain the initial phase vector. ; Step S4: Iteration of the adaptive gain SPGD parallel algorithm based on real-time feedback: using the initial phase vector obtained in step S3 As the starting phase vector for parallel optimization in this stage Generate a perturbation vector that conforms to a Bernoulli distribution, apply positive and negative perturbations to it, and measure the change in the evaluation function. The system determines the dynamic step size of the current iteration based on the current evaluation function value, updates the phase values ​​of all array elements synchronously based on the dynamic step size and gradient estimation value, and detects the optimization effect through an infrared camera. It performs multiple rounds of global parallel iterative optimization and dynamically adjusts the gain step size based on the current feedback evaluation function value in each iteration. Step S5: Result Output: When the evaluation function value reaches the preset threshold or the number of iterations s reaches the maximum number of iterations, the search stops and the final calibration voltage value of each channel is output, thereby completing the phase calibration.

2. The large-scale optical phased array phase mixing calibration method according to claim 1, characterized in that: The evaluation function is a function for evaluating the diffraction efficiency of an optical phased array, and the specific evaluation function is as follows: Parameter description: For diffraction efficiency; Deflect the target at an angle; This represents the far-field light intensity distribution. This represents the effective field of view of the optical phased array; The full width at half maximum (FWHM) of the diffraction peak is denoted as α.

3. The large-scale optical phased array phase mixing calibration method according to claim 1, characterized in that: The evaluation function is a function for evaluating the peak-to-sidelobe ratio of an optical phased array. Specifically, the evaluation function is as follows: Parameter description: The largest side lobe; The power intensity of the main lobe is denoted by ; PSLR is the peak sidelobe ratio.

4. The large-scale optical phased array phase mixing calibration method according to any one of claims 1-3, characterized in that: Step S3 includes: S3.1: Pre-sampling logic determination: Starting from the second array element, perform single-channel calibration on the array element to be calibrated: keep the phase of other array elements unchanged, apply phase shifts of 0.5π and 1.5π respectively on this array element, acquire far-field images using an infrared camera, and calculate the corresponding evaluation functions J1(0.5π) and J2(1.5π); S3.2: Filter the search interval and compare the size of J1 and J2; if J1>J2, then set the initial search interval of the array element to [0,π]; if J1≤J2, then set the initial search interval to [π,2π]. S3.3: Golden Section Iteration: Within the selected π length interval, select according to the golden ratio. The sampling points are compared using the evaluation function feedback. After K iterations, the quasi-optimal initial phase value of the array element is locked. ; S3.4: Repeat steps S3.1 to S3.3 for the 3rd to Nth array elements in sequence until the point-by-point initial adjustment of all N channels of the entire array is completed, and combine the quasi-optimal initial phase values ​​locked in each channel into the initial phase vector. .

5. The large-scale optical phased array phase mixing calibration method according to claim 2 or 3, characterized in that: Step S4 specifically includes the following steps: Step S4.1: Use the initial phase vector obtained in step S3. As the starting phase vector for parallel optimization in this stage Initialize the adaptive SPGD algorithm iteration count s=0 to obtain the initial phase vector. Then calculations were performed to obtain And stipulate ; Step S4.2: Generate a set of random phase perturbation vectors with dimensions consistent with the number of elements to be calibrated and conforming to a Bernoulli distribution. Each perturbation element takes the value of First, apply a positive disturbance to the system. The infrared camera is used to acquire far-field images and calculate the corresponding evaluation function values. Then apply a negative disturbance to the system. Acquire far-field images and calculate the corresponding evaluation function values. ; Calculate the change in the evaluation function ; Step S4.3: Calculate the adaptive gain step size: based on the evaluation function value of the current iteration. Through the formula of negative power function Calculate the dynamic gain step size for the current iteration round. ; Step S4.4: Based on dynamic gain step size Evaluation function change and random perturbation vector Once the gradient estimate is obtained, the phase values ​​of all array elements are updated synchronously according to the phase update formula. When the evaluation function is diffraction efficiency, the phase update formula is: When the evaluation function is the peak-to-sidelobe ratio, the phase update formula is: The iteration count s is incremented by 1, and the updated far-field image is acquired by an infrared camera to detect the beam optimization effect of this iteration. Step S4.5: Iterative Loop: Repeat steps S4.2 to S4.4 above to perform multiple rounds of global parallel iterative optimization. In each round of iteration, the gain step size is dynamically adjusted according to the current feedback evaluation function value until the evaluation function value reaches the preset threshold or the number of iterations s reaches the maximum number of iterations.

6. The large-scale optical phased array phase mixing calibration method according to claim 1, characterized in that: The iteration number K in step S2 ranges from 4 to 10.

7. The large-scale optical phased array phase mixing calibration method according to claim 1, characterized in that: In step S2, the basic gain factor α is between 0.42 and 0.58, and the attenuation factor β is between 0.68 and 0.75.

Images