Frequency spectrum drift feedback method for improving real-time phase detection precision

Abstract

The invention discloses a frequency spectrum drift feedback method applied to plasma density diagnosis for improving real-time phase detection precision. According to the method, an FPGA is used as a core platform, and reference and measurement signals containing three frequency components are collected through an ADC; after cross-clock domain processing and ping-pong storage, windowing FFT is carried out to obtain frequency spectrum information; extracting frequency positions and dynamic tracking change values of three peak values in the two paths of signals through a multi-spectral peak identification and tracking algorithm, and transmitting the frequency positions and the dynamic tracking change values to an upper computer to draw a trend curve; and comparing the actual frequency with the theoretical window width, feeding back a frequency spectrum drift early warning if the actual frequency exceeds the range, otherwise, adaptively calculating the window width for each frequency point to carry out dynamic filtering, and finally outputting three paths of phase differences through IFFT (Inverse Fast Fourier Transform), complex conjugate multiplication and arc tangent operation. Real-time high-precision calculation of the phase difference of the multi-frequency signals is achieved, spectrum drift can be processed in a self-adaptive mode, spectrum peak overlapping is effectively prevented, and the real-time performance, precision and reliability of plasma density diagnosis are remarkably improved.

Description

Technical Field

[0001] This invention relates to the field of plasma density diagnostic technology, and specifically discloses a feedback method for improving the accuracy of real-time phase detection in response to spectral drift. Background Technology

[0002] Energy is the cornerstone of human development, and controlled nuclear fusion, due to its clean and efficient characteristics, is considered the core of future energy. Achieving fusion requires precise control of plasma density, temperature, and confinement time, with real-time monitoring of plasma density parameters being particularly crucial. Therefore, accurate diagnosis and monitoring of plasma electron and current densities are essential prerequisites for ensuring the progress of fusion research and the safe operation of reactors.

[0003] Far-infrared laser diagnostic technology is widely used in the diagnosis of high-temperature plasmas due to its diverse measurement methods and lack of unnecessary disturbance and contamination to the plasma. In nuclear fusion reactors, measuring plasma electron density and current density using far-infrared laser polarization interferometry and providing real-time feedback control has become an important method. Among these methods, the three-wave polarization interferometry is a common approach. It uses a laser to output three beams with slightly different frequencies. Through optical beam combining and splitting, polarization and interference signals can be obtained without external grating modulation, avoiding errors introduced by modulation and improving measurement stability and accuracy. Its working principle is as follows: one beam serves as a reference beam, directly coupled to the third intrinsic beam and entering the reference detector; the other beam serves as the probe beam, first passing through the plasma, undergoing a phase change due to its refractive index, and then coupling with the intrinsic beam into the probe detector. The two detectors output reference and measurement signals containing three intermediate frequencies, respectively. By comparing the phase difference between these two signals, the Faraday rotation angle and plasma electron density can be calculated, and thus the poloidal magnetic field distribution and plasma current density can be inferred. The schematic diagram is shown below. Figure 1 As shown.

[0004] In a three-wave polarization interferometry system, signal frequency stability is a crucial prerequisite for ensuring the accuracy of phase measurements. In such a system, the three beat frequency signals generated by the three lasers correspond to the beat frequencies of the local oscillator and the left- and right-hand circularly polarized light, respectively. Maintaining a fixed spectral order is essential for accurate physical quantity calculations. However, due to laser output frequency drift, even if the spectral order remains unchanged, the beat frequency peaks may still alternate, overlap, or become excessively close. Specifically, for example… Figure 2As shown, this can lead to severe crosstalk during data processing, introducing significant noise into the system and potentially causing errors in phase measurement results. Furthermore, long-term slow drift can accumulate substantial systematic errors, ultimately making it impossible to meet the requirements for phase detection. Therefore, real-time monitoring of laser signal drift and timely feedback are crucial for maintaining signal stability and ensuring the accuracy of system measurement results.

[0005] Traditional phase detection techniques rely primarily on software calculations, making it difficult to provide real-time feedback on key plasma parameters. This results in poor real-time monitoring of critical physical quantities such as electron density and current density. Furthermore, the increasingly complex structure and function of future nuclear fusion reactors necessitate the detection of a growing number of potentially unstable signal frequencies. This can easily lead to peak overlap during spectral analysis, severely impacting the accuracy of data processing and phase detection. Ultimately, this results in significant errors in phase calculations, hindering the precise and safe control of controlled nuclear fusion devices. Additionally, existing technologies largely depend on manual assessment of signal frequency drift to maintain intermediate frequency signal stability, significantly limiting their flexibility and scalability.

[0006] Therefore, how to achieve real-time monitoring and timely feedback of the stability of mixed signals containing multiple frequencies, so as to reduce the phase calculation error caused by spectral peak overlap, ensure the real-time detection accuracy of plasma electron and current density, and improve the flexibility and scalability requirements, is an important technical problem that urgently needs to be solved. Summary of the Invention

[0007] The purpose of this invention is to provide a spectral drift feedback method for improving real-time phase detection. Based on FPGA technology, this method preprocesses two signals containing three frequencies, performs spectral analysis, and obtains six frequency points from the two signals through peak finding. These points are then compared with theoretical values ​​set by the host computer. If the signals are within a specified range, signal stability is fed back to the host computer; if they exceed the range, spectral drift is promptly reported to the host computer, allowing operators to adjust the signal frequency in a timely manner. This method is adaptable to the analysis of signals containing multiple different frequencies and provides real-time feedback of relevant information. It can effectively reduce spectral peak overlap caused by spectral drift, thereby reducing phase calculation errors and ensuring the real-time performance and high accuracy of plasma electron and current density detection. This meets the stringent requirements of controlled nuclear fusion devices for laser signal frequency stability.

[0008] To achieve the above objectives, the present invention adopts the following technical solution: A spectral drift feedback method for improving real-time phase detection is proposed. This method uses a programmable gate array (FPGA) as the processing core and specifically includes the following steps:

[0009] After the system starts the phase measurement work after receiving the host computer instruction, it continuously collects external signals through an ADC chip. The external signals include a reference signal and a measurement signal, and both signals contain three different frequency components.

[0010] Perform cross-clock domain processing and ping-pong storage operations on the collected reference signal and measurement signal respectively, realize data segmentation, and continuously input to subsequent operations to support pipelined processing.

[0011] Perform windowing processing on the data segments after the above operations and perform fast Fourier transform (FFT) to obtain the frequency domain information of each of the two signals.

[0012] Analyze the frequency domain information, perform multi-spectrum peak identification and tracking operations on the frequency domain signals. Three actual frequency points and their corresponding dynamic tracking change values can be obtained for each signal, and they are uploaded to the host computer in real time to respectively draw the corresponding frequency dynamic change trend curves. Compare the window width calculation ranges of the three actual frequency points with their respective theoretical frequency points. If it is not within the theoretical range, there is a spectrum drift phenomenon, and an instruction indicating the spectrum drift phenomenon and the current actual frequency point position are promptly fed back to the host computer; if it is within the theoretical range, continue with the subsequent work.

[0013] If normal work can be carried out after theoretical comparison, calculate the adaptive window width range according to the actual frequency points where the three maxima are located in the spectrum, retain all the spectrum information within the window width range, set the remaining spectrum data to zero, and feed back the actual window width data corresponding to the three calculated frequency points to the host computer.

[0014] Perform inverse fast Fourier transform (IFFT) on the processed spectrum data, and then obtain the calculation results of three phase differences in the two signals through arctangent operation and send them to the host computer.

[0015] The multi-spectrum peak identification and tracking method is as follows: Dynamically detect and record the three largest peaks and their corresponding frequency point positions based on the traversal method, and compare and calculate with the saved frequency point positions of the previous segment. The core idea is to sequentially compare and update conditions. Each time a new data is input, it is successively judged whether it meets the conditions for updating the largest peak, the second peak or the third peak, and the peak list is recorded in real time. Let the three peak points be d0, d1, d2 (d0 < d1 < d2), the corresponding frequency point variables be p0, p1, p2, which are used to record the frequency point positions corresponding to d0, d1, d2 respectively. The three frequency point variables saved in the previous segment of data are p0', p1', p2', the dynamic tracking change values are Δ0, Δ1, Δ2, and the input new data value is d i , and its corresponding frequency point position is p i , and their initial values are all 0.

[0016] Step 1: For the maximum peak d0, if d i > d0, then pass the original value of d0 and its frequency point position p0 to d1 and its frequency point position p1 in sequence, pass the original value of d1 and its frequency point position p1 to d2 and its frequency point position p2, let d0 = d i , and record its frequency position p0 = p i ; Otherwise, continue to Step 2.

[0017] Step 2: For the second peak d1, if d1 < d i < d0, then pass the original value of d1 and its frequency point position p1 to d2 and its frequency point position p2, keep the value of d0 and its frequency point position p0 unchanged, let d1 = d i , and record its frequency point position p1 = p i ; Otherwise, continue to Step 3.

[0018] Step 3: For the third peak d2, if d2 < d i < d1, then let d2 = d i , and record its frequency point position p2 = p i , keep the values of d0, d1 and their frequency point positions p0, p1 unchanged; Otherwise, d0, d1, and d2 are not updated.

[0019] After traversing all the input data, finally obtain the three largest peak points d0, d1, d2 and their corresponding frequency point positions p0, p1, p2. Compare and calculate the three frequency point positions with the saved p0', p1', p2' respectively, and perform the following operations: Δ0 = p0 - p0'; Δ1 = p1 - p1'; Δ2 = p2 - p2'.

[0020] After execution, save the values of p0, p1, p2 to p0', p1', p2' for the next tracking operation.

[0021] The specific calculation method of the adaptive window width range is as follows: Let the current frequency peak frequency point be n i (i = 0, 1).

[0022] If the frequency point n i+1 - n i > 6 and the distances of the frequency point from the zero point and the spectrum center point are both greater than 3, then set the filtering window width as [n i - 3, n i + 3], and each frequency point range contains 7 - point spectrum information; Otherwise, it is necessary to calculate the dynamic filtering window width, that is, starting from the actual peak frequency point in the signal spectrum, search in both the left and right directions respectively. If encountering a spectrum zero point, spectrum center point or other peak frequency points, calculate the dynamic range based on this, and obtain an asymmetric adaptive filtering window width (taking the integer part) for each frequency point.

[0023] For frequency point n0, if a zero point of the spectrum is encountered to the left, the width of the left window is [n0 / 2, n0], otherwise it is [n0-3, n0]. If a frequency point n1 is encountered to the right and n1-n0≤6, the width of the right window is [n0, n0+(n1-n0) / 2], otherwise it is [n0, n0+3].

[0024] For frequency point n1, if frequency point n0 is encountered to the left and n1-n0≤6, then the width of the left window is [n0+(n1-n0) / 2+1, n1], otherwise it is [n1-3, n1]; if frequency point n2 is encountered to the right and n2-n1≤6, then the width of the right window is [n1, n1+(n2-n1) / 2], otherwise it is [n1, n1+3].

[0025] For frequency point n2, if it encounters frequency point n1 to the left and n2-n1≤6, then the left window width is [n1+(n2-n1) / 2+1, n2], otherwise it is [n2-3, n2]; if it encounters the spectral center point n to the right... k And n k If -n²≤6, then the width of the right window is [n², n²+(n... k [-n2) / 2], otherwise [n2, n2+3].

[0026] The real-time phase detection spectral drift feedback method of this invention provides a system with FPGA as the processing core, realizing high real-time and high-precision detection and spectral drift monitoring of phase differences in signals containing three frequency components. Through all-hardware FFT / IFFT, traversal peak finding and tracking, and adaptive dynamic window width calculation, it can accurately extract the actual positions, dynamic trends, and phase information of three frequency points in a single workflow, and compare them with the theoretical range in real time. Once spectral drift is detected, it immediately feeds back to the host computer, effectively avoiding phase calculation errors caused by overlapping spectral peaks. This method significantly improves the real-time performance and accuracy of plasma electron density detection. Its adaptive processing mechanism has both strong robustness and versatility, fully meeting the stringent requirements of controlled nuclear fusion devices for laser signal frequency stability. Its core advantages are reflected in the following aspects.

[0027] The system developed in this invention significantly improves the processing speed and real-time performance of relevant data in fusion experiments. Based on an FPGA platform, this method integrates a series of complex operations—signal acquisition, cross-clock domain processing, ping-pong buffering, FFT / IFFT transformation, multi-peak identification and tracking algorithms, adaptive dynamic window width calculation, and phase calculation—into the hardware logic through a hardware pipeline design. This approach eliminates the latency and uncertainty caused by operating system scheduling and software iteration, resulting in extremely low and strictly controllable latency throughout the entire process from external signal input to phase results and drift state feedback. It enables real-time monitoring of the plasma experiment's state, meeting the stringent real-time requirements of the plasma diagnostics field.

[0028] The adaptive spectrum processing method proposed in this invention significantly improves the accuracy of phase difference calculation results and the versatility of the system. Based on a fixed window width, an adaptive dynamic window width calculation method is proposed. This method can calculate the optimal left and right window widths for each peak point in real time and independently, according to the actual distribution of peak points in the spectrum. It retains the effective spectral range containing the dominant frequency energy while minimizing harmonic and noise interference, thus solving the problem of spectral peak overlap caused by spectrum drift and providing a high-quality data stream for subsequent high-precision phase calculations.

[0029] The real-time monitoring and feedback early warning mechanism for spectral drift constructed in this invention enhances the stability and reliability of the system. This mechanism compares the actual frequency points obtained from multi-peak identification and tracking with the theoretical range in real time. If the frequency is within a safe range, the filter window width is determined based on the actual frequency point and fed back to the host computer. Otherwise, if the spectral drift is severe, an early warning is immediately sent to the host computer. It can also monitor the laser's frequency state in real time using the frequency dynamic change trend curve plotted based on dynamically tracked numerical values, predicting frequency drift trends. This provides strong support for staff to monitor the plasma experiment status in real time and adjust parameters promptly, and also ensures the system outputs stable and reliable phase difference measurement results. Attached Figure Description

[0030] Figure 1 This is a schematic diagram of the three-wave polarization interferometry principle of the present invention;

[0031] Figure 2 This is a diagram showing the phenomenon of spectral peaks approaching (or even overlapping) due to the spectral drift phenomenon of this invention;

[0032] Figure 3 This is a block diagram of the phase measurement system design of the present invention;

[0033] Figure 4 This is a flowchart of the multi-peak identification and tracking method of the present invention;

[0034] Figure 5 This is a flowchart of the adaptive window width range calculation method of the present invention. Detailed Implementation

[0035] The following detailed description, in conjunction with the accompanying drawings and embodiments, provides an improved spectral drift feedback method for real-time phase detection according to the present invention. The following embodiments will help those skilled in the art to further understand the present invention, but do not limit the invention in any way. It should be noted that those skilled in the art can make several modifications and improvements without departing from the concept of the present invention. These all fall within the scope of protection of the present invention.

[0036] Figure 3 This is a block diagram of the phase measurement system design of the present invention, as shown in the figure, which includes the following steps.

[0037] Step 1: After receiving the phase measurement command from the host computer, the system starts the measurement process; the ADC chip continuously and synchronously acquires the externally input reference signal and measurement signal; both the reference signal and the measurement signal contain three signal components of different frequencies.

[0038] Step 2: Perform cross-clock domain processing on the two acquired digital signals to ensure stable synchronization of the data in subsequent processing modules; then perform ping-pong storage operation on the two signals to segment the data, achieving seamless buffering and pipelined continuous processing of the data.

[0039] Step 3: Window the two preprocessed signal data segments to reduce spectral leakage; perform Fast Fourier Transform (FFT) on the two windowed signals respectively to obtain the frequency domain information of the reference signal and the measurement signal.

[0040] Step 4: Analyze the frequency domain information, perform multi-peak identification and tracking on the spectrum of each signal, locate the actual frequency points corresponding to the three amplitude maxima and their corresponding dynamic tracking changes, and upload them to the host computer in real time to draw the corresponding frequency dynamic change trend curves; compare each actual frequency point with the preset theoretical frequency point window width calculation range: if any actual frequency point exceeds the corresponding theoretical window width range, it is determined that a spectrum drift has occurred, the system immediately sends a spectrum drift alarm command to the host computer, and suspends the subsequent process; if all actual frequency points are within the corresponding theoretical window width range, proceed to step 5.

[0041] Step 5: Based on the three actual frequency points identified in Step 4, calculate the corresponding adaptive window width range for each; for the spectrum of each signal, only retain the spectrum data that falls within the three adaptive window width ranges, and set the spectrum data outside the window width ranges to zero; at the same time, upload the calculated window width data corresponding to the three actual frequency points to the host computer.

[0042] Step 6: Perform the inverse fast Fourier transform (IFFT) on the two filtered spectrum data in Step 5 respectively to recover six groups of complex time-domain signals; calculate the phase differences corresponding to the three frequency components respectively through complex conjugate multiplication and arctangent operations on the corresponding signals; finally, send the three calculated phase difference results to the host computer to complete one phase measurement work.

[0043] Figure 4 is the flow chart of the multi-peak recognition and tracking method of the present invention, and the Figure 1 multi-peak recognition and tracking method in Step 4 includes the following steps.

[0044] Step 1: Set three peak variables d0, d1, and d2, which are used to store the currently recognized maximum peak, sub-peak, and third peak respectively; three corresponding frequency point variables p0, p1, and p2, which are used to record the frequency point positions corresponding to d0, d1, and d2 respectively; the three frequency point variables p0', p1', and p2' saved in the previous data; the dynamically tracked change values are Δ0, Δ1, and Δ2, and the initial values of all variables are set to 0.

[0045] Step 2: The frequency domain data stream obtained after the signal data collected by the system is preprocessed and FFT is continuously input into the traversal process, and each newly input data is set as d i , and its corresponding frequency point position is p i .

[0046] Step 3: If d i > d0, then first assign the current value of d0 and its frequency point position p0 to d1 and its corresponding frequency point position p1 respectively; secondly, assign the current value of d1 and its frequency point position p1 to d2 and its corresponding frequency point position p2 respectively; finally, let d0 = d i , and record its corresponding frequency point position p0 = p i .

[0047] Step 4: If the condition in Step 3 is not satisfied, then judge whether to update the sub-peak d1. If d1 <d i <d0, then first assign the current value of d1 and its frequency point position p1 to d2 and its corresponding frequency point position p2 respectively; then let d1 = d i , and record its corresponding frequency point position p1 = p i ; d0 and its frequency point position p0 remain unchanged.

[0048] Step 5: If the conditions in Steps 3 and 4 are not satisfied, then judge whether to update the third peak d2: If d2 <d i <d1, then let d2 = d i , and record its corresponding frequency point position p2 = d i, d0 and d1 and their frequency point positions p0 and p1 remain unchanged.

[0049] Step 6: If the current data does not meet any of the conditions in Steps 3, 4, and 5 above, the records of d0, d1, d2 and their corresponding frequency point positions p0, p1, p2 remain unchanged.

[0050] Step 7: After traversing all the input data, the finally obtained d0, d1, d2 are the three largest peaks identified in the spectrum, and their corresponding frequency point positions p0, p1, p2 are the frequency point positions where these three peaks are located.

[0051] Step 8: Compare and calculate the three current frequency point positions p0, p1, p2 with the saved p0', p1', p2' respectively: Δ0 = p0 - p0'; Δ1 = p1 - p1'; Δ2 = p2 - p2'. After the calculation, save the values of p0, p1, p2 to p0', p1', p2' for the next tracking operation.

[0052] Figure 5 is the flow chart of the adaptive window width calculation method of the present invention, and the Figure 1 adaptive window width calculation method in Step 5 includes the following steps.

[0053] Step 1: Set the three peak frequency points obtained by multi-spectrum peak identification, which are n0, n1, n2 in ascending order of frequency, where n0 < n1 < n2. Define the zero point position and the spectrum center point position n of the spectrum k .

[0054] Step 2: For adjacent peak frequency points, determine whether the condition is satisfied: If for all i (i = 0, 1), it is satisfied that n i+1 -n i > 6, and the minimum distance of each peak frequency point n i from the spectrum zero point and the spectrum center point is greater than 3, then execute Step 3 and adopt the standard fixed window width; otherwise, execute Step 4 and adopt the dynamic adaptive window width.

[0055] Step 3: If the conditions set in Step 2 are satisfied, assign the same fixed window width range [n i -3, n i +3] to each peak frequency point n_i.

[0056] Step 4: If the conditions set in Step 2 are not satisfied, calculate the dynamic window widths with asymmetric left and right boundaries for n0, n1, and n2 respectively, corresponding to Steps 5, 6, and 7.

[0057] Step 5: For the lowest frequency peak point n0, set its dynamic window width range to [L0, R0] (rounding down to the integer part). Calculate the left boundary L0: if the distance between n0 and the zero point of the spectrum is ≤3, then L0 = n0 / 2, otherwise L0 = n0 - 3; calculate the right boundary R0: if n1 - n0 ≤ 6, then R0 = n0 + (n1 - n0) / 2, otherwise R0 = n0 + 3.

[0058] Step 6: For the intermediate frequency peak point n1, set its dynamic window width range to [L1, R1] (rounding down to the integer part). Calculate the left boundary L1: if n1-n0≤6, then L1=n0+(n1-n0) / 2+1, otherwise L1=n1-3; calculate the right boundary R1: if n2-n1≤6, then R1=n1+(n2-n1) / 2, otherwise R1=n1+3.

[0059] Step 7: For the highest frequency peak point n2, set its dynamic window width range to [L2, R2] (rounding down to the integer part). Calculate the left boundary L2: if n2 - n1 ≤ 6, then L2 = n1 + (n2 - n1) / 2 + 1, otherwise L2 = n2 - 3; Calculate the right boundary R2: if the spectral center point position n k -n²≤6, then R²=n²+(n k -n2) / 2, otherwise R2=n2+3.

Claims

1. A spectral drift feedback method for improving real-time phase detection accuracy, characterized in that, include: S1: After receiving the instruction from the host computer, the system synchronously acquires the continuously input reference signal and measurement signal through the ADC chip. The reference signal and measurement signal each contain three signal components of different frequencies. S2: Perform cross-clock domain processing on the acquired reference signal and measurement signal respectively, and perform ping-pong storage operation on the processed signal to segment the data for subsequent pipelined processing; S3: Window the two signal data segments processed in step S2, and perform Fast Fourier Transform (FFT) on them respectively to obtain the frequency domain information of the reference signal and the measurement signal. S4: Analyze frequency domain information, perform multi-peak identification and tracking operations on the frequency domain information of each signal, identify three amplitude maxima and their corresponding actual frequency points, and obtain three dynamic tracking change values ​​by comparing the current data segment with the three frequency points saved after the previous data segment processing, so as to track the actual frequency point position in real time; upload these three values ​​and the three actual frequency point data to the host computer to draw the corresponding frequency dynamic change trend curves; compare each actual frequency point with the preset theoretical frequency window width range. If any actual frequency point exceeds its corresponding theoretical frequency window width range, it is determined that a spectrum drift has occurred, and a spectrum drift warning command is sent to the host computer; If all actual frequency points are within their corresponding theoretical frequency window widths, then proceed to step S5. S5: Based on the three actual frequency points identified in step S4, calculate the adaptive window width range for each frequency point. For the spectrum data of each signal, only the data within the adaptive window width range is retained, and the data outside the range is set to zero. At the same time, the adaptive window width data corresponding to the three actual frequency points is fed back to the host computer. S6: Perform inverse fast Fourier transform (IFFT) on the two spectrum data processed in step S5, then calculate the phase difference corresponding to the three frequency components through complex conjugate multiplication and arctangent operation, and send the three sets of phase difference results to the host computer.

2. The real-time phase detection spectral drift feedback method according to claim 1, characterized in that, The multi-peak identification and tracking operation described in step S4 specifically includes: Let the current data segment have three peak variables d0, d1, d2 and their corresponding frequency point position variables p0, p1, p2, and the three frequency point variables p0', p1', p2' stored in the previous data segment, dynamically tracking the changing values ​​Δ0, Δ1, Δ2, all initially set to 0. Traverse the input spectral amplitude data stream, for each current data segment d... i The corresponding frequency point is p. i Perform the following operations: If d i If d0 > 0, then assign the current value of d0 and its frequency position p0 to d1 and its corresponding frequency position p1, and assign the current value of d1 and its frequency position p1 to d2 and its corresponding frequency position p2. Let d0 = d i And record its frequency point position as p0=p i ; If the above conditions are not satisfied and d1 < d i < d0, then assign the current value of d1 and its frequency point position p1 to d2 and its corresponding frequency point position p2, and set d1 = d i and record its frequency point position as p1 = p i , and keep d0 and its corresponding frequency point position p0 unchanged; If the above two conditions are not satisfied and d2 < d i < d1, then set d2 = d i and record its frequency point position as p2 = p i , and keep d0 and d1 and their corresponding frequency point positions p0 and p1 unchanged; If the current data does not meet any of the above conditions, all variables remain unchanged; After the traversal is complete, d0, d1, and d2 are the three identified maximum peak values, and their corresponding frequency point position variables p0, p1, and p2 are the three actual frequency points. These are then compared and calculated with the saved p0', p1', and p2', respectively, and the following operations are performed: Δ0=p0-p0'; Δ1=p1-p1'; Δ2=p2-p2'. After execution, the values ​​of p0, p1, and p2 are saved to p0', p1', and p2 for use in the next tracking operation.

3. The real-time phase detection spectral drift feedback method according to claim 1, characterized in that, The calculation of the adaptive window width range in step S5 specifically includes: Let the three peak frequency points be n0, n1, and n2 in ascending order of frequency; If the following condition is met: for all i (i=0,1), n i+1 -n i >6, and each n i Distance from the spectral zero and spectral center point n k If the minimum distance is greater than 3, then for each n i Allocate fixed window width range [n i -3, n i +3]; If the above conditions are not met, then the following rules apply to each frequency point n. i Calculate the dynamic window width [L] i , R i (Take the integer part): For n0: its left boundary L0: if the distance between n0 and the zero point of the spectrum is ≤3, then L0=n0 / 2, otherwise L0=n0-3; its right boundary R0: if n1-n0≤6, then R0=n0+(n1-n0) / 2, otherwise R0=n0+3; For n1: its left boundary L1: if n1-n0≤6, then L1=n0+(n1-n0) / 2+1, otherwise L1=n1-3; its right boundary R1: if n2-n1≤6, then R1=n1+(n2-n1) / 2, otherwise R1=n1+3; For n2: its left boundary L2: if n2-n1≤6, then L2=n1+(n2-n1) / 2+1, otherwise L2=n2-3; its right boundary R2: if n k -n²≤6, then R²=n²+(n k -n2) / 2, otherwise R2=n2+3.

4. A phase measurement system incorporating a spectral drift feedback mechanism, characterized in that, The system is used to implement the method according to any one of claims 1 to 3, the system comprising: Signal acquisition module: used to continuously and synchronously acquire reference signals and measurement signals containing three different frequency components through an ADC chip; Data preprocessing module: used to process and store the two acquired signals across clock domains, and to segment the data to support pipelined processing by subsequent modules; Spectrum analysis module: used to perform windowing and Fourier transform (FFT) on the two preprocessed signals to obtain their respective frequency domain information; Peak identification, tracking and drift judgment module: used to execute multi-peak identification and tracking algorithm to obtain three actual frequency points and their corresponding dynamic tracking change values ​​and upload them to the host computer in real time. The three actual frequency points are compared with the corresponding theoretical range to determine whether a spectrum drift phenomenon has occurred and generate corresponding feedback instructions. Adaptive window width calculation and filtering module: used to calculate the adaptive window width range of each of the three actual frequency points and filter the corresponding spectrum data; Phase calculation module: used to perform IFFT transformation on the filtered spectrum to obtain six sets of complex time-domain signals, and perform pairwise complex conjugate multiplication and arctangent operation on them to obtain three sets of phase difference results; Communication interface module: used for receiving instructions and providing data feedback to the host computer.

5. The real-time phase detection spectral drift feedback method according to claim 1, characterized in that, The method significantly improves the real-time performance and accuracy of plasma electron density detection. Its adaptive processing mechanism has excellent versatility, which not only fully meets the stringent requirements of fusion experiments for the stability of laser signal frequency, but also provides a reliable basis for experimenters to monitor the plasma state in real time and adjust parameters in a timely manner, thereby comprehensively enhancing the stability and operational reliability of the system.

6. The method for multi-spectral peak identification and tracking and adaptive window width range calculation according to claims 2 and 3, characterized in that, The method can calculate appropriate left and right window widths for each peak based on the distribution of actual peak points in each data segment's spectrum, and compare them with the peak points of the previous data segment to determine the trend of change. While preserving the effective spectrum range containing the main frequency energy to a maximum extent, it filters out harmonic and noise interference to a certain extent, thereby effectively solving the problem of spectral peak overlap that may be caused by spectrum drift, and providing reliable data for subsequent high-precision phase calculations.

7. The phase measurement system including a spectral drift feedback mechanism according to claim 4, characterized in that, The system greatly improves the processing speed and real-time performance of relevant data in fusion experiments, achieving extremely low and strictly controllable latency throughout the entire process from external signal acquisition to phase results and drift state feedback. It can continuously and stably monitor the state of plasma experiments in real time, fully meeting the stringent requirements of high real-time performance and high precision in the field of plasma diagnostics.

Images