Automatic frequency stabilization system based on dual-core correlation matching intelligent peak searching

Abstract

The invention discloses an automatic frequency stabilization system based on dual-core correlation matching intelligent peak searching. The automatic frequency stabilization system comprises the following steps that S1, a saturated absorption spectrum light path and a circuit are built, debugged and optimized; s2, constructing a double-matching template; s3, triangular wave voltage scanning and clock timing are started, and signals are collected and transmitted; s4, starting a KCF algorithm, applying a template to track model matching, judging a result obtained by matching, finding a peak point and solving a corresponding scanning voltage; and S5, recording a peak point scanning voltage, closing a scanning signal, starting sinusoidal signal modulation, and giving a triangular wave scanning voltage corresponding to a peak value to the sinusoidal signal as an offset. The spectrum line peak value suspected area can be effectively positioned, the double-branch and correlation filtering tracking method can perform precise secondary searching in the suspected area, the locking precision is greatly improved, the long-term stability of 10-12 orders of magnitude is ensured, and meanwhile the good vibration interference resistance can be achieved when a vibration signal occurs.

Description

Technical Field

[0001] The present invention relates to the technical field of automatic frequency stabilization systems, and particularly to an automatic frequency stabilization system based on dual-core correlation matching and intelligent peak searching. Background Art

[0002] Gravimetry intersects and integrates with many disciplines such as seismology, geology, fundamental physics, geodynamics, hydrology, oceanography, and astronomy, and has developed vigorously. It is a key discipline for serving strategic needs such as the national surveying and mapping benchmark, resource exploration, military security, and disaster monitoring, as well as for researching basic geoscience issues. The gravitational field can effectively characterize the distribution of the Earth's material density. Precise gravitational field information can provide important physical field information for geophysics, resource exploration, detection of basic physical indicators, and navigation, and provide data support for the construction of background information. At the same time, it also plays a core role in the application of target objects such as oil and gas resource exploration, tidal model construction, submarine gravitational field construction, gravitational background field construction, and passive navigation. Therefore, the accurate measurement of gravitational indicators is particularly important for the extension of basic discipline theories and practical applications. As a new type of high-precision absolute gravity measurement inertial sensor, the cold atom gravimeter has significant performance advantages in gravity measurement. The cold atom gravimeter uses a laser with strong coherence, good monochromaticity, and stable frequency as the coherent light source, and jointly constructs a magneto-optical trap to cool alkali metal atoms with an external magnetic field. By adjusting the phase of the laser, beam splitting, inversion, and beam combination of Raman interference are realized, and the laser frequency is further adjusted to scan the interference fringes to calculate the local gravity value. With the development of time, the cold atom interferometer has gradually moved from the static laboratory state to the field, and cold atom gravimeters with a certain degree of miniaturization and intelligence have been developed. As the core of the cold atom gravimeter, the optical path system in the entire system of the mobile cold atom gravimeter is also vulnerable to environmental factors, resulting in problems such as loss and drift of frequency points. A stable laser frequency will have a greater impact on links such as atomic cooling, Raman interference, fluorescence detection, and the accuracy of gravity dynamic detection, severely restricting the accuracy of gravity measurement. In order to meet the need for mobility, miniaturization is an important strategy for the laser system, and the compactness of the laser system is a research direction. The birth of the laser has promoted the progress of a series of science and technologies. In order to obtain a laser that meets the requirements, that is, a laser with a narrow linewidth and high frequency stability, researchers have studied the laser closed-loop frequency stabilization system. A typical laser closed-loop frequency stabilization system consists of three parts: a reference frequency, a frequency discriminator, and a feedback controller. Widely used reference frequencies include the transition frequencies of atoms or molecules and the characteristic frequencies of optical resonators. The main difference lies in the different frequency discrimination methods. When using the transition frequencies of atoms and molecules as the reference frequency, the frequency discrimination signal is often obtained from the absorption spectrum or dispersion spectrum of atoms and molecules. Researchers have designed many frequency discrimination methods and their frequency stabilization technologies, such as modulation spectroscopy frequency stabilization technology, modulation transfer spectroscopy frequency stabilization technology, bias spectroscopy frequency stabilization technology, magneto-dichroism frequency stabilization technology, etc. The transmission and reflection characteristics of a high-Q F-P cavity have characteristics similar to the atomic absorption spectral line and have an extremely narrow linewidth. The typical frequency stabilization technology based on the F-P cavity is the pdh frequency stabilization technology.

[0003] After years of development, the closed-loop frequency stabilization technology of lasers has achieved very prominent results and has been widely applied in many research fields. However, it has very obvious deficiencies, mainly manifested as the instability of the laser frequency stabilization system, that is, the unlocking phenomenon is likely to occur. This is because the effective frequency locking range of the above-mentioned laser closed-loop frequency stabilization technology is usually narrow. When the laser system suffers inevitable interference, the laser frequency jumps out of the effective frequency locking range and the unlocking occurs. Such phenomena seriously limit the application scenarios of the laser closed-loop frequency stabilization system, such as long-term, complex field environment, unattended, spaceborne and other experiments. Therefore, a miniaturized automatic frequency stabilization system that can be applied in multiple scenarios without human intervention, quickly recover and maintain long-term stability under unlocking conditions such as loss or drift of laser frequency points is particularly important. However, many devices are locked based on the peaks of individual typical Lamb dips without considering the frequency locking method of all saturation absorption peaks; the mode of calculating the differential signal of a specific transition frequency peak as the frequency locking point is extremely susceptible to environmental factor changes in the dynamic measurement of cold atom gravimeters, resulting in strong fluctuations and drifts in the differential signal, and thus unable to correctly find the locking point. The problem with such methods is that the locking point cannot be found for a long time, resulting in the system repeatedly unlocking and sweeping the frequency to find the locking point, leading to an overly long period, and inevitably introducing an overly long "dead time", resulting in abnormal gravity measurements. The traditional method of finding the locking point by differentiation has good results in static experiments, but has fatal defects in dynamic measurement environments. In recent years, artificial intelligence technology has made great progress, and the computing power and perception ability of AI large models have been greatly improved. Applying the method of artificial intelligence algorithms will effectively provide new ideas for frequency stabilization in complex field environments. The target tracking algorithm can quickly find the target area and lock it in a complex strong interference background during the real-time movement of the target, and can be excellently applied to the identification and tracking of spectral lines. The target tracking algorithm based on kernel correlation filtering can distinguish the foreground from the background to the greatest extent by constructing a kernel correlation filter, and the constructed ridge regression loss function can effectively separate the foreground from the background. Therefore, this algorithm is introduced into peak locking. However, the kernel correlation filtering algorithm only constructs a set of cyclic transfer matrices to find candidate targets in the real-time moving area and cannot accurately locate. Therefore, we propose an automatic frequency stabilization system based on dual-core correlation matching intelligent peak searching to solve the above problems. Summary of the Invention

[0004] The purpose of the present invention is to solve the deficiencies existing in the prior art, and to propose an automatic frequency stabilization system based on dual-core correlation matching intelligent peak searching.

[0005] In order to achieve the above purpose, the present invention adopts the following technical solutions:

[0006] An automatic frequency stabilization system based on dual-core correlation matching intelligent peak searching, comprising the following steps:

[0007] S1. Build and debug and optimize the saturated absorption spectrum optical path and circuit;

[0008] S2. Construct a double matching template;

[0009] S3. Start triangular wave voltage scanning and clock timing, collect signals and transmit them;

[0010] S4. Start the KCF algorithm, apply the template tracking model for matching, discriminate the obtained matching results, find the peak point and obtain the corresponding scanning voltage;

[0011] S5. Record the scanning voltage at the peak point and turn off the scanning signal, and at the same time turn on the sine signal modulation. At the same time, give the triangular wave scanning voltage corresponding to the peak to the sine signal as an offset, and set the frequency and amplitude of the sine signal;

[0012] S6. The photodetector collects the optical signal. The sine signal with the same frequency and the optical signal collected by the photodetector are sent to the multiplier through the phase shifter to obtain the pre-differential error signal;

[0013] S7. Obtain the DC component of the error signal through a low-pass filter with a set cut-off frequency;

[0014] S8. Construct a PID feedback, send the DC component as the differential error signal into the PID feedback to form a closed-loop circuit to achieve locking of the required transition frequency point;

[0015] S9. Determine whether the error signal is greater than the set threshold to determine whether to unlock, including: the detection module monitors whether to unlock. Once the error signal is greater than 1×10 -3 Or the PD detection signal is less than 0.1, indicating that the laser is unlocked. At this time, restart the triangular wave scanning of the laser, the KCF intelligently determines the frequency locking point, modulates and demodulates, and the PID feedback control stabilizes the frequency locking point operation to complete automatic frequency relocking.

[0016] The present invention can be used in the optical path module and modulation transfer module related to miniaturization. The laser driver circuit provides stable current control and an intelligent PID temperature control module (current control accuracy: 0.01 mA, temperature control accuracy: 0.01 °C). It simplifies the miniaturized optical path construction and starts triangular wave scanning to cooperate with the laser driver circuit to adjust and optimize the spectral line. First, a standard curve template containing 6 saturated absorption peaks and a curve region of a specified peak are constructed to form a double template. Triangular wave scanning is started to obtain the real-time spectral line. The analog-to-digital conversion module collects the spectral line voltage and triangular wave voltage, and the dual-branch parallel kernel correlation filtering algorithm is run to find the peak value. The parallel operation of the dual-branch kernel correlation filtering algorithm can not only effectively locate the suspected peak region of the spectral line, but also the dual-branch and correlation filtering tracking methods can perform precise secondary searching in the suspected region, greatly improving the locking accuracy. When both branches meet the set value requirements, the triangular wave scanning is turned off and the scanning voltage at the locking point is recorded. The modulation transfer module is started, the spectral line information signal is collected into the modulation transfer module, the sine wave modulation is sent into the laser driver circuit to modulate the laser tube, the same-frequency sine signal is phase-shifted by the phase shifter and then demodulated, and the spectral line signal obtained from the photodetector is sent into the multiplier. The DC component is filtered out by the low-pass filter and sent into the subsequent PID module as an error signal for feedback fine-tuning of the scanning voltage at the locking point to ensure the stability of the output frequency point, ensuring a long-term stability of the order of 10 -12 orders of magnitude.

[0017] The frequency locking monitoring module of the present invention continuously monitors the error signal and the spectral line voltage signal. When the measured signal fails to meet the set requirements, it indicates that the laser frequency point drifts or the frequency is abnormal and unlocked. At this time, the modulation transfer (modulation and demodulation) module is turned off, the triangular wave scanning module is started to scan and obtain the scanning spectral line, and then the dual-branch KCF algorithm is started for peak searching again. When the peak searching again meets the set point requirements, it indicates re-locking. The triangular wave scanning module is turned off again and the scanning voltage at the peak point is recorded. The modulation transfer module is synchronously started, the error signal is obtained, and the scanning voltage at the peak point is feedback fine-tuned by PID to maintain the frequency point locked state. At the same time, the unlocking monitoring module is started to monitor whether it is unlocked in real time. The intelligent re-locking module can quickly find and complete the locking after unlocking, can complete the re-locking operation within seconds, and can remain unlocked for several months after re-locking, and ensure a long-term stability of the order of 10 -12 orders of magnitude. When vibration occurs, due to the stable sample anti-drift interference ability of the intelligent KCF peak searching algorithm, when the vibration signal is superimposed on the optical path spectral line signal, it can still meet the double set requirements. Therefore, when the vibration signal occurs, it can have good anti-vibration interference ability. Description of the Drawings

[0018] Figure 1 It is a diagram showing the detection of the laser wavefront by the Thorlabs wavefront detector;

[0019] Figure 2 It is a diagram of the laser temperature control and current drive interface;

[0020] Figure 3 It is the hyperfine energy level structure diagram of Rb atoms;

[0021] Figure 4 It is the automatic frequency stabilization system diagram of saturated absorption spectrum;

[0022] Figure 5 It is the physical optical path construction diagram;

[0023] Figure 6 It is the triangular wave scanning spectral line diagram;

[0024] Figure 7 It is the double-template construction diagram of saturated absorption peak;

[0025] Figure 8 It is the scanning spectral line diagram of signal generator;

[0026] Figure 9 It is the structure diagram of double-branch KCF algorithm;

[0027] Figure 10 It is the structure diagram of double-template kernel correlation intelligent peak-seeking automatic frequency stabilization system;

[0028] Figure 11 It is the overall system flow chart;

[0029] Figure 12 It is the frequency stabilization structure diagram;

[0030] Figure 13 It is the frequency stabilization spectral line and error signal diagrams before and after filtering;

[0031] Figure 14 It is the stability test diagram of wavelength meter;

[0032] Figure 15 It is the 100-hour stability test diagram;

[0033] Figure 16 It is the re-locking test diagram;

[0034] Figure 17 It is the Allan variance diagram. Specific implementation manners

[0035] Next, the technical solutions in the embodiments of the invention will be clearly and completely described in conjunction with the accompanying drawings in the embodiments of the invention. Obviously, the described embodiments are only a part of the embodiments of the invention, rather than all of the embodiments.

[0036] Refer to Figure 1-17 , an automatic frequency stabilization system based on dual-kernel correlation matching intelligent peak-seeking, includes the following steps:

[0037] S1. Build, debug and optimize the saturated absorption spectrum optical path and circuit;

[0038] S2. Construct a double matching template;

[0039] S3. Start the triangular wave voltage scan and clock timing, collect signals and transmit them;

[0040] S4. Start the KCF algorithm, apply the template tracking model for matching, discriminate the obtained matching results, find the peak point and obtain the corresponding scanning voltage;

[0041] S5. Record the scanning voltage at the peak point and turn off the scanning signal, and at the same time turn on the sine signal modulation. At the same time, give the triangular wave scanning voltage corresponding to the peak as the offset to the sine signal, and set the frequency and amplitude of the sine signal;

[0042] S6. The photodetector collects the optical signal. The optical signal collected by the photodetector and the sine signal of the same frequency after passing through the phase shifter are sent to the multiplier to obtain the pre-differential error signal;

[0043] S7. Obtain the DC component of the error signal through a low-pass filter with a set cut-off frequency;

[0044] S8. Construct a PID feedback. Use the DC component as the differential error signal to send into the PID feedback to form a closed-loop circuit to achieve the locking of the required transition frequency point;

[0045] S9. Determine whether the error signal is greater than the set threshold to determine whether to unlock, including: the detection module monitors whether to unlock. Once the error signal is greater than 1×10 -3 Or the PD detection signal is less than 0.1, indicating that the laser is unlocked. At this time, restart the triangular wave scan of the laser, the KCF intelligent determines the frequency locking point, modulates and demodulates, and the PID feedback control stabilizes the frequency locking point operation to complete the automatic frequency relocking.

[0046] Among them, in step S1: The optical path construction includes:

[0047] Set the driving current of the laser, turn on the temperature PID, drive the laser tube to emit light by adjusting the driver current, set the current value to 110 mA, and turn on the temperature PID. At this time, the laser tube emits light normally, and the spot range is good. The effect is as Figure 1 shown, Figure 1 The Solibro wavefront detector detects the laser wavefront condition and finds that the light energy emitted by the laser can well meet the conditions of the high-speed beam.

[0048] After multiple tunings of the driving current of the laser, it can emit light better in the range of 110 mA and meet the subsequent spectral line requirements. The laser adjusts the temperature control PID to set the automatic parameter tuning function to meet the requirements of temperature adjustment in the dynamic environment scenario. Figure 2 The laser adjustment interface is given. Figure 2The working current of the middle drive interface is set to 110 mA, the maximum output current is set to 140 mA, the maximum output current of the laser tube is 180 mA, a 40 mA margin is required, and the output power of the laser is 47 mw, meeting the required current and power for the experiment. The temperature PID drives the PID parameters according to the required ambient set temperature and turns on the temperature control module. The temperature control accuracy is 0.001 °C, and it basically meets the set temperature required by the laser tube during the experiment.

[0049] Calculation of the saturated absorption peak frequency:

[0050] The coupling of the electron spin angular momentum L and the orbital angular momentum S (S = 1 / 2) (LS coupling) generates a fine energy level structure. The value range of the total angular momentum J of the coupled electrons is |L - S| ≤ J ≤ |L + S|, that is, J = |L ± 1 / 2|. Figure 3 This is the hyperfine energy level structure of the Rb atom.

[0051] For the atom in 5 2 S 1 / 2 → 5 2 P 1 / 2 and 5 2 S 1 / 2 → 5 2 P 3 / 2 The transitions are respectively called the D1 line and the D2 line, and the value (unit: MHz) between the energy levels is the energy level spacing. The Rb element contains 85 Rb and 87 Rb, two isotopes, with natural abundances of 72.15% and 27.85% respectively. The fine energy levels of the ground state and the excited state of the atom are usually denoted as n 2S+1 S J and n 2S+1 P J Among them, the principal quantum number n determines the layer where the outermost valence electrons are located. For the Rb atom, n = 5. When in the ground state, L = 0, and J g = 1 / 2, corresponding to the fine energy level of 5 2 S 1 / 2 ; when in the lowest excited state, L = 1, and J e1 = 1 / 2 and J e2 = 3 / 2. Among them, J e1 = 1 / 2 corresponds to the fine energy level of 5 2 P 1 / 2 , J e2 = 3 / 2 corresponds to the fine energy level of 5 2 P 3 / 2 .

[0052] The coupling of the total electron angular momentum J and the total nuclear angular momentum M (JM coupling) generates a hyperfine energy level structure. The value range of the total angular momentum F of the coupled atom is |J - M| ≤ F ≤ |J + M|. In the ground state, J = 1 / 2 for both. 85For Rb with M = 5 / 2, the ground-state fine energy level 5 2 S 1 / 2 splits into two hyperfine energy levels of F g = 2 and F g = 3; 87 For Rb with M = 3 / 2, the ground-state fine energy level 5 2 S 1 / 2 splits into two hyperfine energy levels of F g = 1 and F g = 2. Similarly, the hyperfine energy levels of the excited state F e1 and F e2 can also be obtained.

[0053] The main reasons for the broadening of atomic spectral lines are: the natural broadening of atoms without external influence; the Doppler broadening caused by random thermal motion; and the pressure broadening due to collisions between absorbing atoms. Among them, the Lorentzian linear natural broadening is on the order of MHz, and the pressure broadening in a dilute atomic gas cell without buffer gas is so small that it can be ignored, while the Gaussian linear Doppler broadening can reach the order of hundreds of MHz. Assume that a two-level atom with velocity v moves to the left and is irradiated by a probe light with frequency ω L , propagating to the right.

[0054] Due to the Doppler effect, the frequency felt by the atom dop When ω 0 is equal to the atomic resonance frequency ω pro , the velocity v of the atomic group that absorbs the probe light is: In the formula, Δ, L are the detuning amount and wave vector of the laser, respectively. It can be seen that due to the Doppler effect, the laser around the resonance frequency will also be absorbed by the atoms, making the absorption spectral line of the atoms broaden. If a counter-propagating pump light with the same frequency ω but higher power is introduced into the rubidium cell. Similarly, from formula it can be seen that when Δ≠0, the pump light will be absorbed by the atomic group with velocity Figure 5 . At this time, the two beams of light are absorbed by atomic groups with different velocities and do not affect each other; when Δ = 0, both beams of light are absorbed by the atomic group with zero velocity. The absorption of the high-power pump light by the atoms becomes saturated, so the absorption of the probe light is weak, resulting in strong transmission of the probe light. A raised peak, namely the saturated absorption peak, is observed in the absorption spectrum, and the corresponding spectral line is called the saturated absorption spectrum, as

[0055] shown. The Rb atom is a multi-level model. Here, a three-level model is taken as an example for illustration, as Figure 6 shown. The resonance frequencies of the three-level atom are ω 1 and ω 2 respectively, then there are two atomic groups with different velocities that will absorb photons.

[0056] When ω L =(ω 1 +ω 2 ) / 2, from Equation and Equation it is obtained that v 1 =v 2 . It can be seen that when the laser scanning frequency is equal to the intermediate frequency of two hyperfine energy levels, a saturation absorption peak will also be generated when the atom interacts with it, that is, a cross saturation absorption peak.

[0057] Figure 6 is the saturation absorption spectrum of the D2 line transition of Rb atoms. The saturation absorption peaks are respectively 87 Rb's F g =2→F e2 =(1,3) cross line (C 13 ), F g =2→F e2 =(2,3) cross line (C 23 ) and F g =2→F e2 =3 transition line (T 3 ) and 85 Rb's F g =3→F e2 =(2,4) cross line (C 24 ), F g =3→F e2 =(3,4) cross line (C 34 ) and F g =3→F e2 =4 transition line (T 4 ).

[0058] The wavelengths and frequencies of the saturation absorption peaks are shown in Table 1.

[0059] Table 1 Frequencies and Wavelengths of Rb Atom Saturation Absorption Peaks

[0060]

[0061]

[0062] Frequency stabilization system design;

[0063] The principle of laser frequency stabilization is as follows: Select a stable frequency as the reference frequency. When the laser frequency deviates from the reference frequency, an error signal that can characterize this deviation is generated by identifying the deviation, and the error signal is fed back to the laser. The reference frequency mostly adopts the central frequency of the transition spectrum line of atoms or molecules or the resonance frequency of the optical resonator. The automatic frequency stabilization system uses the central frequency of the atomic saturated absorption spectrum as the frequency reference, and modulates and demodulates the laser by using the method of wavelength modulation spectroscopy, so as to obtain the frequency discrimination signal required for frequency stabilization. The main components of the automatic frequency stabilization system are as Figure 4 shown, and it is mainly composed of six parts: an optical part, a single-chip microcomputer control circuit, a modulation and demodulation circuit, a proportional-integral-derivative (PID) feedback circuit, a host computer module, and a laser controller.

[0064] The physical diagram of the optical path construction is as Figure 5 shown, which mainly involves a laser driver circuit, a temperature control module, an optical isolator, a collimating lens, a focusing lens, a half-wave plate, a quarter-wave plate, a rubidium atomic gas cell, a reflector, a photodetector, and a host computer for displaying the spectral line shape and subsequent frequency stabilization operations, etc.

[0065] Figure 5 This is a physical display of the optical path construction. The optical path is collimated by a yellow single-mode optical fiber and a small part of the light is separated by a polarization beam splitter prism as the pump light. The pump light passes through the rubidium cell and the quarter-wave plate and is transmitted to the 0-degree reflector. After reflection, it passes through the rubidium cell again. The power of the pump light is about 0.45 mw, and the power of the probe light is about 46 μw, and the spectral line information can be effectively observed. The power ratio of the pump light to the probe light is about 10:1. The probe light is transmitted through the polarization beam splitter prism and focused by a 40-mm focusing lens onto the photodetector for photoelectric conversion. The data acquisition card collects and transmits it to the host computer to display the spectral line information and subsequent frequency stabilization processing. After the spectral line information is optimized by the triangular wave amplitude and frequency, it is as Figure 6 shown.

[0066] Among them, step S2 of constructing the double matching template includes:

[0067] The spectral line information obtained by the photodetector is optimized to obtain the real-time spectral line. Multiple periods are intercepted through the LabView program and saved as an array. The required template is obtained by offline simulation of the array. Since the rubidium cell is natural abundance rubidium, the spectral line information converted by the photodetector here contains 85 Rb and 87 all the spectral line information of 87 Rb. Here, the g Rb spectral line is used as the reference spectral line, which contains 6 basic saturated absorption peaks. Intercept the F e2 =2→F g =(2,3) transition frequency as the reference frequency locking point, corresponding to the second saturated absorption peak of the spectral line. Therefore, 6 basic saturated absorption peaks are intercepted through the array as template 1, and F g =2→Fe2 =(2,3) corresponds to the transition frequency of the second template, where F g =2→F e2 =(2,3) corresponds to the second saturation absorption peak of Template 1. The production of the double template is as Figure 7 shown.

[0068] Figure 7 The production process of the template is given. By comparing the standard curve and optimizing the optical path, the spectral line information containing 6 saturation absorption peaks is obtained. Multiple groups of multi-cycle spectral line data are saved through data acquisition from the real-time spectral line. The spectral line profile that can effectively characterize the current environment and is relatively appropriate is extracted through offline simulation. A set of data is selected from it, and the spectral line segment containing 6 saturation absorption peaks is intercepted as the standard template of the first branch, as shown by Template 1 in the figure. In order to more effectively locate the position of the saturation absorption peak, any one of the saturation absorption peaks is selected as Template 2. Considering that the red-detuned cooling light of the frequency points used in the cold atom gravimeter is at F g =2→F e2 =(2,3) transition frequency vicinity, so the second saturation absorption peak of Template 1 is selected as Template 2, as Figure 8 Template 2 in the figure is F g =2→F e2 =(2,3) corresponding spectral line saturation absorption peak.

[0069] Among them, in step S3: Turn on the triangular wave voltage scan and clock timing, and the signal acquisition and transmission include;

[0070] The triangular wave scanning module is built. The triangular wave scan is constructed through a signal generator, and the wavelength meter is used to view the change range of the scanning frequency, and real-time simulation experiments are carried out. The signal generator is set with a frequency of 1 Hz and a voltage amplitude of 1 V. At this time, the spectral line information is good, and 6 saturation absorption peaks can be completely displayed. The saturation absorption peak segment and the signal generator setting interface are as Figure 8 shown.

[0071] Figure 8 The frequency is set to 1 Hz, the voltage amplitude is 0.75 V, the offset and phase are set to 0 in the figure. Since the impedance required by the driver is 50 Ω, the impedance is adjusted to a low impedance. The spectral line effect in the experiment is good, and the peaks are relatively sharp, which can provide good spectral line information for subsequent experiments.

[0072] To achieve miniaturization, the present invention realizes triangular wave scanning through a program and optimizes parameters. The program is written using LabView, which is convenient for later application of DSP or FPGA for programming. To collect signals, the triangular wave scanning signal needs to be sent to the SMA interface of the driver first. The triangular wave signal is sent out through the AO interface of the data acquisition card, and the spectral line signal of the PD detector is collected through the AI interface. Set clock synchronization, and the program sequentially executes the sending and collection operations. The LabView display control can view the spectral line information in real time.

[0073] S4. Start the KCF algorithm, apply the template tracking model for matching, and discriminate the obtained matching results to find the peak point and obtain the corresponding scanning voltage, including:

[0074] Based on the kernel correlation filtering tracking algorithm, as Figure 9 shown.

[0075] Figure 9 Show the structure diagram of the dual-branch KCF algorithm. Template 1 and the spectral line information collected by the acquisition card in the current period are used for correlation calculation to update the correlation filter, obtaining the response point area. After FFT operation, it is used for correlation calculation with the spectral line information in the (t + 1)-th period. After inverse FFT operation, the corresponding point position is obtained, and the maximum value of the response area is obtained through classification by the classifier. Branch 2 maps the spectral line area of the correlation operation of Branch 1, and the mapped spectral line area is used for correlation calculation with Template 2 to update the correlation filter 2. After FFT operation, it is used for correlation calculation with the result obtained by Branch 1 at time (t + 1) after mapping. After iFFT operation, the response point position is obtained and sent to the classifier for discrimination operation. When the response values obtained from the discrimination calculations of both branches are greater than the preset value, it indicates that the locking point is found. At this time, record the spectral line voltage position of the locking point and calculate the corresponding scanning voltage value of the scanning voltage.

[0076] S5. Record the scanning voltage at the peak point and turn off the scanning signal, and simultaneously turn on the sine signal modulation. At the same time, load the triangular wave scanning voltage corresponding to the peak value onto the sine signal as an offset, and set the frequency of the sine signal to 10 kHz and the amplitude to 0.01 V.

[0077] S6. The photodetector collects the optical signal. The same-frequency sine signal and the optical signal collected by the photodetector are sent to the multiplier through the phase shifter to obtain the pre-differential error signal, including:

[0078] The same-frequency signal modulation and demodulation. The atomic resonance frequency can be used as the reference frequency source. The feedback system needs to receive an error signal and adjusts according to the feedback of the error signal to keep the laser stable near the resonance frequency point. The modulation and demodulation method can obtain the differential signal. The error signal fluctuates slightly around 0 and is sent to the PID feedback control as the first-order differential error signal.

[0079] The modulation method is to apply a modulation to the laser frequency ω 0Add a small sine modulation signal \(A\sin(\Omega t)\), where \(\Omega\) is the modulation frequency and \(A\) is the modulation amplitude, as small as possible. Load it on the laser driver as the modulation signal to modulate the laser.

[0080] To obtain the differential signal, the modulation signal is loaded onto the laser for continuous sine modulation. At this time, demodulation is required to obtain the differential signal.

[0081] S7. Obtain the DC component from the error signal through a low-pass filter with a set cut-off frequency, including:

[0082] During the demodulation process, a sine signal with the same frequency is phase-shifted by a phase shifter and used as the reference signal to multiply with the spectral line signal of the PD detector as the mixing signal. Then, the DC component is extracted through a low-pass filter. At this time, the DC component is exactly the differential signal, fluctuating around 0.

[0083] S8. Construct a PID feedback. Feed the DC component as the differential error signal into the PID feedback to form a closed-loop circuit to achieve locking at the desired transition frequency point.

[0084] S9. Determine whether the error signal is greater than the set threshold to determine whether to unlock, including: The detection module monitors whether to unlock. Once the error signal is greater than \(1\times10\) -3 or the PD detection signal is less than 0.1, it indicates that the laser is unlocked. At this time, restart the triangular wave scanning of the laser, and the KCF intelligently determines the frequency locking point, performs modulation and demodulation, and the PID feedback control stabilizes the frequency locking point operation to complete automatic frequency relocking.

[0085] In order to overcome the frequency locking method based on the peak locking of individual typical Lamb dips without considering all saturated absorption peaks; by calculating the differential signal of a specific transition frequency peak as the frequency locking point mode, in the dynamic measurement of cold atom gravimeters, it is extremely vulnerable to environmental factor changes, resulting in strong fluctuations and drifts in the differential signal, and it is impossible to correctly find the locking point, resulting in repeated unlocking and frequency scanning for a long time to find the locking point, introducing an overly long "dead time", which seriously affects the accuracy of gravity dynamic measurement. The present invention proposes an automatic frequency stabilization system based on dual-core correlation matching intelligent peak searching. Figure 10 and Figure 11 Give the specific operation method. This method first builds an optical path that is easy to miniaturize, sets Figure 2 the constant current in it to 110 mA, the maximum current to 140 mA, sets the temperature to 21 °C, automatically tunes the current environmental temperature control PID parameters, builds the optical path to align the pump light and the probe light to coincide and pass through the rubidium cell, adjusts the power ratio of the pump light and the probe light until spectral line information is observed, and optimizes each link to achieve the best effect. Figure 4 and Figure 5Give the specific operations for setting up the optical path. The host computer sets up a triangular wave scanning module, adjusts the triangular wave frequency to 1 Hz and the amplitude to 750 mV, turns on the triangular wave scanning, and transmits it to the SMA driver through a usb-6211 data acquisition card. The data acquisition card collects the signals of the PD detector. The host computer displays the spectral line information and finely tunes the phase and offset of the triangular wave scanning to the optimal values. Figure 9 Give the process of finding the scanning voltage at specific frequency points. The collected spectral line information is sent to the dual-template matching kernel correlation algorithm for intelligent peak searching, and the scanning voltage at the frequency points is searched for and recorded. Figure 12 Give the automatic frequency stabilization method after peak searching. After finding the frequency-locking peak, turn off the triangular wave scanning and simultaneously turn on the sine modulation module. Set the sine signal frequency to 10 kHz and the amplitude to 0.01. Send it to the SMA driver through the same path of the acquisition card. The PD detector collects the spectral line voltage and the sine signal of the same frequency. After the phase shifter shifts the phase, they are sent to the multiplier to extract the error signal. The error signal passes through a low-pass filter to filter out the DC component, and is identified by the unlocking identification module to determine whether the differential signal is higher than the preset 10 -3 and whether the PD detection voltage is less than 0.1 mV to determine whether unlocking occurs. If there is no unlocking, it is sent to the PI module to feedback and finely adjust the scanning voltage at the frequency point. If unlocking occurs, turn on the triangular wave scanning module again. The dual-template kernel correlation intelligent peak searching module searches for peaks and determines the locking point scanning. After the modulation and demodulation module, it is determined again whether unlocking occurs. If no unlocking is seen, the PI feedback adjusts the current at the frequency point. The system repeatedly executes this loop process to ensure automatic frequency stability. The long-term frequency stability can reach 10 -12 orders of magnitude. The unlocking and locking time is about 4 seconds. For the long-term locking test, no unlocking was seen in 108 hours.

[0086] The above is only a preferred specific implementation manner of the present invention, but the protection scope of the present invention is not limited thereto. Any person skilled in the art within the technical scope disclosed by the present invention, according to the technical solution and inventive concept of the present invention, makes equivalent replacements or changes, and all should be covered within the protection scope of the present invention.

Claims

1. An automatic frequency stabilization system based on dual-core correlation matching and intelligent peak search, characterized in that: The following steps are involved: S1. Build, debug and optimize the saturated absorption spectrum optical path and circuit; S2, constructing a double matching template; S3, start triangle wave voltage scanning and clock timing, collect signals and transmit; S4, start the KCF algorithm, apply the template tracking model matching, judge the matching results, find the peak point and obtain the corresponding scanning voltage; S5, record the peak scanning voltage and turn off the scanning signal, and turn on the sine signal modulation at the same time, and give the triangular wave scanning voltage corresponding to the peak value to the sine signal as an offset, and set the sine signal frequency and amplitude; S6, the photoelectric detector collects the photoelectric signal, and the photoelectric signal obtained by using the same frequency sinusoidal signal through the phase shifter and the photoelectric detector is sent to the multiplier to obtain the pre-differential error signal; S7, obtaining a DC component by passing the error signal through a low-pass filter with a set cutoff frequency; S8, construct PID feedback, use the DC component as the differential error signal and send it into the PID feedback to form a closed loop to achieve the desired transition frequency point locking; S9, determining whether the error signal is greater than a set threshold value and determining whether the lock is unlocked include: the detection module monitors whether the lock is unlocked, once the error signal is greater than 1×10 -3 Or if the PD detection signal is less than 0.1, it means the laser is unlocked. At this time, restart the laser triangle wave scanning, KCF intelligently determines the frequency locking point, modulates and demodulates, and PID feedback controls the stable frequency locking point operation to complete the automatic frequency relocking.

2. The automatic frequency stabilization system based on dual-core correlation matching intelligent peak search according to claim 1, characterized in that: According to step S1, the optical path construction includes: laser drive current setting, temperature PID opening, driving the laser tube to emit light by adjusting the driver current, setting the current value to 110mA, and temperature PID opening. At this time, the laser tube emits light normally, the spot range is good, and the laser drive current is tuned multiple times. It can emit light well in the range of 110mA and meet the subsequent spectral line requirements. The laser adjusts the temperature control PID to set the automatic parameter setting function to meet the temperature regulation requirements in dynamic environment scenarios.

3. The automatic frequency stabilization system based on dual-core correlation matching intelligent peak search according to claim 1, characterized in that: According to step S1, the saturation absorption peak frequency is calculated as follows: the electron spin angular momentum L is coupled with the orbital angular momentum S (S = 1 / 2) to produce a fine energy level structure, and the total angular momentum J of the coupled electrons is in the range of |LS|≤J≤|L+S|, that is, J = |L±1 / 2|. 2 S 1 / 2 →5 2 P 1 / 2 and 5 2 S 1 / 2 →5 2 P 3 / 2 The transitions are called D1 line and D2 line respectively, and the value between energy levels is the energy level spacing. The Rb element contains 85 Rb and 87 The natural abundances of two isotopes of Rb are 72.15% and 27.85%, respectively. The fine energy levels of the ground and excited states of atoms are usually denoted by n 2S+1 S J and n 2S+1 P J , where the principal quantum number n determines the number of layers of the valence electrons outside the nucleus. For Rb atoms, n = 5, and when L = 0 in the ground state, we get J g =1 / 2, corresponding to a fine energy level of 5 2 S 1 / 2 , when the lowest excited state is L=1, we get J e1 =1 / 2 and J e2 =3 / 2, where J e1 =1 / 2 corresponds to a fine energy level of 5 2 P 1 / 2 , J e2 =3 / 2 corresponds to a fine energy level of 5 2 P 3 / 2 The total angular momentum of the electron J is coupled with the total angular momentum of the nucleus M to produce a hyperfine energy level structure. The total angular momentum of the atom F after coupling has a value range of |JM|≤F≤|J+M|. In the ground state, J=1 / 2. 85 Rb's M = 5 / 2, so the ground state fine level is 5 2 S 1 / 2 Split into F g =2 and F g =3 two hyperfine levels, 87 Rb's M = 3 / 2, so the ground state fine level is 5 2 S 1 / 2 Split into F g =1 and F g =2 two hyperfine energy levels, and similarly, the excited state F e1 and F e2 of the superfine levels.

4. The automatic frequency stabilization system based on dual-core correlation matching intelligent peak search according to claim 1, characterized in that: According to step S1, the principle of laser frequency stabilization is as follows: a stable frequency is selected as the reference frequency. When the laser frequency deviates from the reference frequency, an error signal that can characterize the deviation is generated by identifying the deviation, and the error signal is fed back to the laser. The reference frequency is mostly the center frequency of the transition spectrum of the atomic molecule or the resonant frequency of the optical resonant cavity. The automatic frequency stabilization system uses the center frequency of the atomic saturated absorption spectrum as the frequency reference, and uses the amplitude-frequency hybrid modulation spectrum method to modulate and demodulate the laser, thereby obtaining the frequency discrimination signal required for frequency stabilization.

5. The automatic frequency stabilization system based on dual-core correlation matching intelligent peak search according to claim 1, characterized in that: According to step S2, constructing a double matching template includes: the spectral line information obtained by the photodetector is tuned to obtain a real-time spectral line, multiple cycles are intercepted by the LabView program, saved as an array, and intercepted by array offline simulation to obtain the required template. Since the rubidium bubble is naturally abundant rubidium, the spectral line information converted by the photodetector includes 85 Rb and 87 All spectral line information of Rb is 87 The Rb spectrum is the reference spectrum, which contains 6 basic saturation absorption peaks. g =2→F e2 =(2,3) transition frequency is used as the reference locking frequency point, corresponding to the second saturation absorption peak of the spectrum line. Therefore, 6 basic saturation absorption peaks are intercepted through the array as template 1, and F is intercepted. g =2→F e2 =(2,3) corresponds to the transition frequency of template 2, which corresponds to the second saturation absorption peak of template 1.

6. The automatic frequency stabilization system based on dual-core correlation matching intelligent peak search according to claim 1, characterized in that: According to step S3, a triangle wave scanning module is built, a triangle wave scan is constructed through a signal generator, and a wavelength meter is used to check the scanning frequency variation range, and a real-time simulation test is performed. The signal generator is set to a frequency of 1 Hz and a voltage amplitude of 1 V. At this time, the spectral line information is good and can fully display the 6 saturation absorption peaks.

7. The automatic frequency stabilization system based on dual-core correlation matching intelligent peak search according to claim 1, characterized in that: According to step S5, the frequency of the sinusoidal signal is set to 10 kHz and the amplitude is set to 0.01 V.

8. The automatic frequency stabilization system based on dual-core correlation matching intelligent peak search according to claim 1, characterized in that: According to step S6, the same frequency signal is modulated and demodulated, and the atomic resonance frequency can be used as a reference frequency source. The feedback system needs to receive an error signal, and the laser is stably maintained near the resonance frequency point based on the feedback adjustment of the error signal. The modulation and demodulation method can remove the differential signal, and the error signal fluctuates slightly around 0, and is sent to the PID feedback control as a first-order differential error signal. The modulation method adds a small sinusoidal modulation signal A sin(Ωt) to the laser frequency w0, where Ω is the modulation frequency and A is the modulation amplitude, which is as small as possible, and is loaded on the laser drive as a modulation signal to modulate the laser. In order to obtain the differential signal, the modulation signal is loaded on the laser for uninterrupted sinusoidal modulation, and demodulated to obtain the differential signal.

9. The automatic frequency stabilization system based on dual-core correlation matching intelligent peak search according to claim 1, characterized in that: According to step S7, during the demodulation process, a sinusoidal signal of the same frequency is used as a reference signal after being phase-shifted by a phase shifter and multiplied with the spectrum line signal of the PD detector as a mixing signal. Then, a DC component is extracted through a low-pass filter. The DC component is a differential signal and fluctuates around 0.

Images