A laser automatic frequency stabilization system and method based on real-time automatic peak seeking technology

Abstract

The application discloses a kind of automatic frequency stabilization system and method of laser based on real-time automatic peak seeking technology, system includes the frequency automatic tuning module based on frequency-current-temperature model;Frequency reference automatic identification module based on adaptive threshold method;Frequency automatic closed-loop control module;Frequency lockout and stability periodic monitoring lock state monitoring and judging module.The application solves the problem of traditional frequency tuning method to oscilloscope and artificial dependence.Based on the peak seeking method of adaptive threshold, frequency reference identification positioning is realized without human intervention;Combined with the closed-loop control system of periodic checking lock frequency system stability, not only monitor laser lockout state, but also real-time monitor laser locking precision, and adjust laser closed-loop module according to monitoring state.The application has the advantages of high peak identification precision, strong noise immunity, high locking precision and good long-term stability.

Description

TECHNICAL FIELD

[0001] The present application relates to the technical field of semiconductor laser frequency automatic closed-loop locking, in particular to a laser closed-loop control system using a saturated absorption peak or using an FP cavity transmission peak and reflection peak as a frequency reference, and specifically to a laser automatic frequency stabilization system and method based on real-time automatic peak searching technology. BACKGROUND

[0002] Laser frequency stabilization technology is widely used in laser coherent measurement, spaceborne laser communication, cold atom physics and other fields. In particular, in the field of atomic spin inertia / magnetic field measurement, the laser frequency is difficult to lock and easy to lose, which leads to poor long-term stability of the laser output frequency, and cannot guarantee the long-term stability of the output of the spin light pumping system and detection system of the atomic sensor. The main difficulty of laser frequency locking lies in that the traditional analog closed-loop frequency locking system relies on operating experience and cannot accurately lock the laser frequency at the absorption peak or resonance peak, and cannot automatically search for the peak and re-lock after losing the lock, which leads to the inability to re-close the laser frequency after it drifts out of the closed-loop point. Another difficulty in maintaining long-term stability of the laser frequency locking system is that the stability of the traditional laser analog frequency locking system depends on the wavelength meter acquisition and judgment, and the judgment of whether the locking state meets the requirements needs to be processed manually after a period of work, which leads to a long judgment period and cannot adjust the locking state in real time.

[0003] In the prior art, a laser automatic frequency stabilization system based on saturated absorption spectrum intelligent recognition technology is mentioned (it is mentioned that the mode recognition technology is used to automatically tune the laser output mode and the locking point in the saturated absorption spectrum. Technology [2] (A self-analyzing double-loop digital controller in laser frequency stabilization for inter-satellite laser ranging) mentions the application of the PDH frequency stabilization method to automatically tune the laser PZT and temperature to realize the automatic closed-loop locking of the laser frequency. Both methods have certain deficiencies, the former relies on offline training of spectral samples to construct a spectral feature vector, and when the optical path is fine-tuned, the spectrum is weak or the acquisition circuit performance is poor, there are many noises, the offline spectral features and real-time spectral features are different, and it is easy to misjudge; in addition, the method of searching for the target peak value is not explained in the patent, and it is more biased towards the partial feature matching of the spectrum and the template; and the second laser double closed-loop frequency locking method, the article describes the automatic closed-loop control method of the laser using the PDH frequency stabilization method, and does not describe the automatic frequency reference method. SUMMARY

[0004] The present application solves the problem: overcome the deficiency of the existing frequency stabilization system which needs human eye observation of the oscilloscope and manual operation of commercial lock-in, controller and other instruments to realize the laser frequency reference recognition closed loop, provide a kind of laser automatic frequency stabilization system and method based on real-time automatic peak searching technology, with the advantages of high peak recognition precision, strong noise resistance, high locking precision and good long-term stability.

[0005] In the first aspect, the present application provides a kind of laser automatic frequency stabilization system based on real-time automatic peak searching technology, comprising: frequency automatic tuning module A based on frequency-current-temperature model, frequency reference automatic identification module B based on adaptive threshold method, frequency automatic closed loop control module C and locking state monitoring and judgment module D of frequency loss and stability periodic monitoring;

[0006] Frequency automatic tuning module A based on frequency-current-temperature model, according to the laser frequency-current-temperature model that has been established, the working point of laser current and temperature is automatically tuned to frequency reference range, and triangle wave sweep signal is started, saturated absorption peak or resonance peak spectrum signal appears;

[0007] Frequency reference automatic identification module B based on adaptive threshold method, the spectrum data in single sweep period is triggered and saved to the start edge of sweep period;Each data in spectrum data is compared with its adjacent data in turn, all local peaks in spectrum data are judged and identified, the baseline of each local peak is judged, and the height and width of peak to baseline are calculated, according to the height and width characteristics of saturated absorption peak or FP cavity resonance peak in spectrum, height threshold and width threshold conditions are set, noise and false peak interference produced by Doppler background are screened out, and the peak meeting threshold condition is saturated absorption peak or FP cavity resonance peak as frequency reference, the scanning voltage corresponding to peak value is stored and output to laser current source, and the frequency reference recognition of laser frequency stabilization system is completed;

[0008] Automatic closed loop control module C, modulation and demodulation algorithm based on Cordic algorithm is solved, and the error signal of laser frequency is obtained, according to the error signal obtained by different frequency reference, set P, I, D parameter initial value, use discrete PID control algorithm to calculate the error signal deviation, and stabilize the laser frequency at the set frequency reference point;

[0009] The locking state judgment module D, which periodically monitors frequency unlocking and stability, periodically stores the spectral signal after the laser frequency is locked and calculates the power spectral density of the error signal within that period. First, based on the characteristics of the spectral signal within one period after locking and unlocking, a laser unlocking threshold condition is set. Then, based on the frequency stability requirements after locking, a threshold condition is set at the corresponding angular frequency of the error signal power spectral density to determine if the laser is unlocked. If unlocked, it re-enters the frequency reference automatic identification module B based on the adaptive threshold method, activates a triangular wave to re-find the frequency reference point, and re-locks. If unlocked, it determines whether the error signal power spectral density meets the stability threshold condition. If it meets the stability threshold condition, it enters the next judgment period. If it does not meet the stability threshold condition, it re-enters the automatic closed-loop control module C, adjusts the PID control parameters, re-closes the loop, and re-enters the locking state judgment module D after one period to determine the laser's locking state and stability, ensuring long-term frequency stability after the laser is powered on.

[0010] Furthermore, the frequency automatic tuning module A based on the frequency-current-temperature model is implemented as follows:

[0011] (1) Calibrate the current-frequency coefficient k of the semiconductor laser I-υ Temperature-frequency coefficient k T-υ constant bias k bias A frequency-current-temperature model was established for the semiconductor laser used. laser =k I-υ I set +k T-υ T set +k bias ;

[0012] (2) Calibrate the current-power coefficient k of a semiconductor laser I-P The laser output power P required by the atomic gyroscope or atomic magnetometer laser Calculate the laser operating current

[0013] (3) Based on the frequency-current-temperature model obtained in step 1, and the target frequency range v laser and the calculated laser operating current I set Calculate the operating temperature of the laser.

[0014] (4) The tuning operating point is I set ,T set The triangular wave output signal is activated to sweep the frequency of the laser.

[0015] Furthermore, the frequency reference automatic identification module B based on the adaptive threshold method includes five steps:

[0016] (1) Monitor the start edge and end edge in the triangle wave frequency sweep cycle, and save the start edge and end edge spectrum data containing frequency reference information to array X;

[0017] (2) Compare the array data in sequence. If the array index is i, the data X i satisfies X i-1 X i X i+1 , then record and save X i and its index i as a local maximum value. Identify and store all peak values in the saved group of spectra;

[0018] (3) Based on the different background baselines of different spectra and the different trends of peak values with baseline changes, an adaptive peak baseline determination method is used to find the baseline of each local peak value starting from the first peak. The peak value X i of array index i is iterated to the left to the start data X0 of array index 0 to find the minimum value X i between X0 and X i-left . The peak value of array index i is iterated to the right to X max , and max is the length of the data stored in a frequency sweep cycle to find the minimum value X max between X0 and X i-right ; if X i-left X i-right , then the point with coordinates (i-right, X i-right ) is the base point, and a horizontal line is drawn through this point to be the baseline of the peak value X i . Otherwise, a horizontal line is drawn through (i-left, X i-left ) to be the peak baseline. After determining the peak baseline, the data is iterated to the left and right from X i until the first boundary point with a vertical coordinate equal to the base point value is found, which are respectively recorded as the left boundary point and the right boundary point of the peak value X i , with coordinates (i_left_boun, X i-left-boun ) and (i_right_boun, X i-right-boun ). Therefore, the amplitude of the peak value X i is |X i-right-boun -X i |, and the width of the peak value X i is |i_right_boun-i_left_boun|. The baseline of each peak value is found in sequence to obtain the amplitude and width of each local peak value relative to its baseline;

[0019] (4) Based on the peak amplitude generated by the noise fluctuation is much smaller than the amplitude of the frequency reference peak, the width of the peak formed by the Doppler background is much larger than the width of the frequency reference peak, an adaptive calculation amplitude threshold and width calculation can be used as the basis for screening the frequency reference amplitude threshold Th height and width threshold Th width, according to the range of the applied sweep voltage, determine the number of absorption peaks or resonant peaks N existing in the single cycle spectrum, use the heap sorting algorithm to sort the amplitude of each local peak calculated in step (3) from small to large, select the amplitude of the N+1th peak as the amplitude threshold Th height; Use the heap sorting method to sort the width of all peaks calculated in step three from large to small, select the average width of the top N peaks as the width threshold Th width;

[0020] (5) From the first local peak stored in step (2), traverse all local peaks, according to the peak amplitude threshold judgment condition (X i-height <Th_height) The peak value introduced by noise fluctuation is screened out, and then according to the peak width judgment condition (X i-width >Th_height) The peak value introduced by the Doppler background is screened out, and finally N peaks satisfying the condition are obtained as saturated absorption peaks or FP cavity resonance peaks. Store the triangular wave sweep voltage corresponding to the peak value label position and output to the laser current source, complete the frequency reference identification of the laser frequency stabilization system.

[0021] Further, the frequency automatic closed loop control module C includes three steps:

[0022] (1) After the spectrum signal collected is selected by a digital band-pass filter to obtain the odd harmonic component of the modulated spectrum, the Cordic quadrature demodulation algorithm is used to calculate the error signal of the laser frequency, and the noise resistance of the system is improved;

[0023] (2) Store the sweep voltage corresponding to the peak value position identified by the frequency reference automatic identification module B, and output the voltage value at this point to the laser current source through the digital-to-analog conversion circuit, and close the triangular wave sweep signal;

[0024] (3) Set the P, I and D control parameter initial values for the error signal of different frequency references, use the discrete PID control algorithm to obtain the correction control amount for the error signal calculated by this module, and real-time correct the laser frequency to the set frequency reference point.

[0025] Further, the lock state monitoring and judgment module D of the frequency loss and stability periodic monitoring includes three steps:

[0026] (1) Interval T sample Collect and save the error signal obtained in the frequency automatic closed loop control module C, and calculate Tsample Power spectral density of error signal U err_psd , save the power spectral density value U at the angular frequency reflecting the frequency stability of the laser err_psd_10Hz ; interval T sample Collect and save the spectrum signal, calculate the spectrum signal amplitude at t0+T sample The difference between the spectrum signal amplitude at t0+T and the spectrum signal amplitude collected at t0 U spectrum_interval ;

[0027] (2) Set the error signal power spectral density threshold Th_psd and the spectrum amplitude difference threshold Th_spectrum; Th_psd and Th_spectrum are set according to the frequency locking stability, and the value is obtained by experimental measurement; according to the threshold condition Judge the laser locking state, when the condition is not met, the laser is judged to be out of lock; according to the threshold condition U err_psd_10Hz <Th_psd judge the frequency locking stability of the laser, when the condition is not met, it is judged that the frequency locking accuracy of the laser does not meet the demand; when both conditions are met, it is judged that the laser remains locked and meets the demand accuracy, and the next T sample judgment period is continued to wait;

[0028] (3) If out of lock, re-enter the frequency reference automatic identification module B based on the adaptive threshold method, start the triangular wave to find the frequency reference point again, and re-lock; if not out of lock, judge whether the error signal power spectral density meets the stability threshold condition; if the stability threshold condition is met, enter the next judgment period; if the stability threshold condition is not met, re-enter the automatic closed-loop control module C, adjust the PID control parameters, and re-close the loop, and after an interval period, re-enter the locking state judgment module D to judge the laser locking state and stability, and ensure the long-term frequency stability of the laser after starting.

[0029] In the second aspect, the application provides a laser automatic frequency stabilization method based on real-time automatic peak finding technology, which realizes the following:

[0030] (1) Realize the automatic tuning of the laser frequency based on the frequency-current-temperature model, including the following steps: calibrate the laser power-frequency coefficient, the frequency-current coefficient, the frequency-temperature coefficient, and the constant bias coefficient; establish the laser frequency-current-temperature model; determine the laser output power and the laser output frequency range according to the experimental requirements of the atomic gyroscope or the atomic sensor; calculate the laser current and temperature working point, and start the triangular wave sweep signal;

[0031] (2) The adaptive threshold method is used to realize automatic identification of frequency reference, including the following steps: monitoring the starting edge and ending edge in the triangular wave frequency sweep cycle, and saving the spectrum signal of a cycle; judging and saving all peak values in the spectrum signal; adaptively judging the baseline of each peak, and calculating the amplitude and width of each peak. The specific method is as follows: starting from the first peak, traversing to the left to the starting point, finding the minimum value X i-left between these data, traversing to the right to the ending point, finding the minimum value X i-right between these data; if X i-left X i-right , then draw a horizontal line through the point (i-right, X i-right ) to mark the baseline of the peak value X i , otherwise draw a horizontal line through the point (i-left, X i-left ) to mark the baseline of the peak value; after determining the baseline, traverse the data to the left and right with X i , until the first boundary point with a function value equal to the baseline value is found, which is recorded as the left boundary point (i_left_boun, X i-left-boun ) and the right boundary point (i_right_boun, X i-right-boun ) of the peak value X i , respectively; therefore, the amplitude of the peak value X i-right-boun is |X i -X i |, and the width of the peak value X i-height is |i_right_boun-i_left_boun|; repeat the above calculation for each peak value to obtain the amplitude and width of each local peak value.

[0032] (3) The sorting method is used to adaptively calculate the amplitude threshold and width threshold. The specific method is as follows: using the heap sorting algorithm to sort all peak amplitudes from small to large, and selecting the amplitude of the N+1th ranked peak as the amplitude threshold Th_height; using the heap sorting method to sort all peak widths from large to small, and selecting the average width of the top N peaks as the width threshold Th_width; screening out the peak values introduced by noise fluctuations according to the peak amplitude threshold judgment condition (X i-width <Th_height), and then screening out the peak values introduced by Doppler background according to the peak width judgment condition (X sample >Th_height), to obtain the peak values that can be used as frequency reference, and saving the sweep voltage corresponding to the peak value center.

[0033] (4) The Cordic quadrature demodulation algorithm is used to obtain the error signal of the frequency reference. The discrete PID control algorithm is used to calculate the frequency correction control quantity of the error signal, and the correction control quantity is used to real-time stabilize the frequency of the laser at the set frequency reference point.

[0034] (5) Periodically monitor the output state of the laser spectrum signal and the error signal, judge whether the laser frequency is out of lock, and whether the stability meets the condition, the judgment method is interval T sample Collect and save the spectrum signal (including saturated absorption spectrum or FP cavity resonance spectrum), calculate t0+T sample The amplitude difference of the spectrum signal at time t0 and the spectrum signal amplitude collected at time t0 U spectrum_interval If the amplitude difference meets Then judge that the laser has lost lock, restart the frequency sweep and perform frequency reference re-identification and locking; If it does not meet, the laser remains locked, interval T sample Collect and save the error signal, and calculate T sample The power spectral density of the error signal in T err_psd Save the power spectral density value U err_psd_10Hz At 10Hz frequency; According to the threshold condition U err_psd_10Hz Th_psd, judge the stability state of the laser frequency locking, when the threshold condition is met, it is judged that the laser frequency locking precision does not meet the demand, the PID parameters are adjusted again and the offset amount is calculated again for closed-loop locking, if the threshold condition is not met, it is judged that the laser frequency locking precision meets the demand, and the next judgment period is waited.

[0035] Compared with the prior art, the present application has the following advantages:

[0036] (1) It is applied to a saturated absorption frequency stabilization system taking atomic absorption spectrum as a frequency reference, and can also be applied to a frequency stabilization system taking an FP cavity as a frequency reference. The system comprises four key modules: a frequency automatic tuning module based on a frequency-current-temperature model; a frequency reference automatic identification module based on an adaptive threshold method; a frequency automatic closed-loop control module; and a locking state monitoring and judgment module for frequency out-of-lock and stability periodic monitoring. The present application first establishes a laser frequency-current-temperature model, and automatically tunes the laser operating point to the target frequency range according to the model, solving the problem of dependence on an oscilloscope and manual operation in the traditional tuning method of the laser frequency. The peak seeking method based on the adaptive threshold is used to accurately identify the frequency reference, so as to automatically tune the laser frequency to the saturated absorption peak frequency reference or the FP cavity resonance peak frequency reference without manual intervention, and the error signal is obtained by using a modulation and demodulation algorithm, and the automatic closed-loop control of the frequency is realized; finally, the method of periodically checking the spectrum voltage after frequency locking and the error signal not only monitors the out-of-lock state of the laser, but also monitors the locking precision of the laser in real time, and adjusts the laser closed-loop module according to the monitoring state. The present application has the advantages of high peak identification precision, strong noise resistance, high locking precision and good long-term stability.

[0037] (2) The real-time automatic peak searching module in the present application can be used not only for saturated absorption spectrum frequency reference locking, but also for FP cavity resonance peak locking, and a series of frequency locking reference distinguishing frequency stabilization methods such as DAVLL frequency locking and MTS frequency locking. The automatic peak searching technology of the present application can realize high-precision identification of the frequency reference without eliminating the Doppler background light path (different from the Doppler elimination light path peak searching technology of Toptica Company) and special noise reduction algorithm. The automatic closed-loop control module and the laser frequency locking state judgment module in the present application can not only realize one-key automatic locking of the laser, but also maintain long-term high-stability locking of the laser frequency. After the laser frequency is unlocked, the frequency reference automatic identification and locking module can be jumped back to for simple and efficient frequency locking process.

[0038] (3) The automatic tuning method of the present application realizes automatic frequency tuning without an oscilloscope by combining the tuning characteristics of the laser frequency-power and current-temperature. The automatic frequency reference identification module based on the adaptive threshold method in the present application can quickly and accurately identify various frequency references such as saturated absorption peaks and FP cavity resonance peaks without spectrum noise reduction and additional Doppler background light path elimination module. When applied to saturated absorption peaks of different atoms, the peak threshold value does not need to be manually adjusted, and the method has the advantages of self-adaptation and high identification accuracy. The present application extracts the error signal in a modulation-demodulation mode, further improves the anti-interference ability of the frequency locking system to noise, and further improves the stability of the frequency stabilization system. Another advantage of the present application is that the frequency locking system sets double judgment conditions of error signal power spectral density and spectrum signal difference, which can accurately and quickly judge whether the laser is unlocked and whether the frequency stability after locking meets the set requirements. Therefore, this advantage can guarantee the long-term stability of the laser after frequency stabilization and the stability of the frequency stabilization accuracy to meet the requirements. The present application is executed in an embedded microprocessor, which can not only reduce the size of the frequency locking system, but also realize one-key frequency locking, and at the same time guarantee the small size and high performance of the system. BRIEF DESCRIPTION OF DRAWINGS

[0039] Figure 1 The flowchart of the laser automatic frequency stabilization system and method based on the real-time automatic peak searching technology of the present application;

[0040] Figure 2 The experimental diagram and the spectrum diagram containing saturated absorption peaks of the specific implementation of the present application, the right spectrum diagram includes the peak value baseline and the peak value amplitude, width schematic diagram in the peak searching method;

[0041] Figure 3 The program flowchart of the specific implementation of the present application;

[0042] Figure 4The following figures show the experimental test results and model fitting results of the laser output frequency-current-temperature model in the implementation of this invention: (a) is a comparison of the wavelength-temperature test results obtained by tuning the laser temperature after fixing the current and the model fitting results; (b) is a comparison of the wavelength-current test results obtained by tuning the laser current after fixing the temperature and the model fitting results.

[0043] Figure 5 The following are experimental results of judging the laser lock-up state in the implementation of the present invention: (a) shows that the amplitude difference of the spectral signal after locking does not meet the threshold condition, and the laser is unlocked; (b) shows the experimental results of judging the power spectral density amplitude of the error signal and the threshold condition after locking. Detailed Implementation

[0044] To make the disclosed objectives, technical solutions, and advantages clearer, the invention will be further described in detail below with reference to specific embodiments and the accompanying drawings. It should be noted that any content not described in the drawings or specification, as well as some abbreviations, are well-known to those skilled in the art. Some specific parameters given in this embodiment are merely illustrative; these values ​​can be changed to appropriate values ​​in different implementations.

[0045] like Figure 1 As shown, the system of the present invention includes an automatic frequency tuning module A; an automatic frequency reference identification module B; an automatic frequency closed-loop control module C; and a frequency lock-in status monitoring and judgment module D.

[0046] (1) Based on the calibrated semiconductor laser current-power coefficient, temperature-frequency coefficient, current-frequency coefficient, and constant bias, establish the laser's frequency-current-temperature output model. According to the system's requirements for laser output power and frequency, calculate the laser's operating current and operating temperature sequentially, and tune the laser to this operating point. Enable the triangular wave output signal to achieve laser frequency sweep, obtaining a spectral signal containing atomic saturation absorption peaks or FP cavity resonance peaks.

[0047] (2) The edges of the triangular wave sweep signal are processed to obtain the starting edge and the ending edge of the triangular wave sweep, and the spectral data in the range are stored. All local peaks in the stored spectral data are identified. The baseline of each peak is calculated, and the amplitude and width of the peak are calculated. The amplitude and width are defined as the height of the peak to the baseline and the width between the two endpoints of the intersection of the baseline and the peak, respectively. The amplitudes of all peaks are sorted in ascending order by the heap sorting algorithm, and the height of the N+1 peak is sorted as the amplitude threshold of the N frequency reference peaks. The widths of all peaks are sorted in ascending order by the heap sorting algorithm, and the average of the widths of the first N+1 peaks is sorted as the width threshold of the N frequency reference peaks. Each peak amplitude threshold and width threshold are judged to exclude peak interference caused by noise interference and Doppler background (or other background baseline interference), and the N absorption peaks or resonance peaks that meet the threshold conditions are identified as frequency references. The corresponding sweep voltage of the N frequency reference peaks is marked.

[0048] (3) The low-frequency and high-frequency noise of the spectral signal is filtered out using a digital band-pass filter to obtain the odd harmonic component of the spectrum, and then the Cordic quadrature demodulation algorithm is applied to demodulate the error signal of the frequency reference. The triangular wave voltage value corresponding to the center of the target peak is output to the laser current source through the digital-to-analog conversion circuit, and the triangular wave sweep signal is turned off. The initial value of the PID control parameter of the laser frequency-locked loop is set, and the discrete PID control algorithm is applied to realize frequency closed-loop locking.

[0049] (4) The power spectral density of the error signal is calculated periodically, and the amplitude difference of the locked spectral signal is calculated periodically. The power spectral density value of the error signal at 10 Hz obtained by calculation is compared with the threshold value of the frequency stabilization accuracy set, and the amplitude difference of the spectral signal obtained by calculation is compared with the spectral difference of the frequency-locked system in the locked state. When the amplitude difference of the spectral signal at the beginning and end of the sampling period meets the threshold condition, i.e., the amplitude difference meets the threshold condition, it is determined that the system is in a lost state. The system re-starts the sweep and enters the frequency reference identification module. When the power spectral density value of the error signal at 10 Hz in the sampling period meets the threshold condition, i.e., the power spectral density value meets err_psd_10Hz U

[0050] In this embodiment, the laser frequency stabilization optical path is a saturated absorption frequency stabilization optical path, and the specific experimental environment is as shown in Figure 2 .

[0051] The laser is a 795 nm DFB semiconductor laser, and the laser output beam is divided into two beams by a 45° beam splitter after passing through an isolator, wherein the transmitted light is used as the application light, and the reflected light is used as the pumping light to pass through the Rb atomic cell. The light beam passes through a plane mirror and is used as the detection light, and then passes through the Rb atomic cell again and is transmitted through the 45° beam splitter to be received by a photodetector. The photodetector outputs a saturated absorption spectrum signal, which includes six saturated absorption peaks.

[0052] The method in the application is implemented by FPGA+DSP hardware circuit and related language. The specific flow chart is shown in Figure 3 . First, the expression of the current and temperature of the laser on the frequency is calculated by the method of linear regression calibration: λ 780 = 0.04848T+0.001355I+779.1849. The results obtained by experiment fitting and model are shown in Figure 4 , Figure 4 (a) in the figure is the wavelength-temperature test result obtained by tuning the temperature of the laser with the current fixed, and the comparison figure of the model fitting result; Figure 4 (b) in the figure is the wavelength-current test result obtained by tuning the current of the laser with the temperature fixed, and the comparison figure of the model fitting result. It can be seen from the figure that the model fitting result is accurate and effective. In the example, the working point of the laser is 64 mA and the temperature is 20℃, which is obtained from the current-power relationship and the current-temperature-wavelength relationship. The triangular wave is started to begin the program, and the output 6V triangular wave sweep voltage obtains the spectrum signal shown in Figure 2 , and the frequency reference automatic tuning is completed.

[0053] The starting point and the ending point of the triangular wave sweep program are detected, and the trigger edges of the light storage and the stop storage are generated. Considering that the spectrum sampling rate is high, the spectrum is sampled and saved. In the example, the spectrum is stored in array 1 in the triangular wave sweep period, and according to the sampling frequency, m points are stored in array 1. The data in array 1 is traversed, and all the local peak values in array 1 are judged according to X i-1 <X i <X i+1 (1<i<m) and the local peak values are stored in array 2.

[0054] The peak baseline of each peak value in array 2 is found, and the height from the peak value to the baseline and the width of the peak value are calculated. For example, Figure 2 , in the right spectrum of the peak value A, if the data label of the peak value A is a in array 1, the left traversal from X a to X a in array 1 is performed to find the minimum value X b between X b , such as point B (b, X a ) shown in Figure 2 . In array 1, from Xa Traverse to the right until X m and search for X a up to X m to find the minimum value of X between them c , such as Figure 2 the point C(c, X c ). If X b < X c , then draw a horizontal line through (c, X c ) and denote it as the baseline of peak A. Otherwise, draw a horizontal line through (b, X b ) and denote it as the peak value baseline. As shown in Figure 2 , the dashed line is the baseline of peak A. Calculate the absolute value of the difference in the ordinate between peak A and point C, |X a - X c |, which is the height from peak A to the baseline and is denoted as the amplitude of peak A. In array 1, traverse from X a to the left until point D(d, X d ), where it satisfies In array 1, traverse from X a to the left until point E(e, X e ), where it satisfies Calculate the absolute value of the difference in the abscissa between D and E, |X e - X d |, which is the width of peak A.

[0055] As shown in Figure 2 , the spectrum contains a total of six absorption peaks. Sort the heights of each peak in array 2. Use the heap sort algorithm to sort the amplitudes of all peaks from largest to smallest. Select the amplitude of the 7th ranked peak as the amplitude threshold Th_height (based on the fact that the peak amplitudes introduced by noise fluctuations are much smaller than those of absorption peaks or resonance peaks); use the heap sort method to sort the widths of all peaks calculated in step three from largest to smallest. Select the average value of the widths of the top 6 ranked peaks as the width threshold Th_width (based on the fact that the peak widths introduced by the Doppler background are much larger than those of absorption peaks or resonance peaks); start traversing all local peaks from the first local peak stored in array 2. Screen out the peaks introduced by noise fluctuations according to the peak amplitude threshold judgment condition (X i-height < Th_height), and then screen out the peaks introduced by the Doppler background according to the peak width judgment condition (X i-width > Th_height). Finally, obtain 6 peaks that meet the conditions and record the triangular wave sweep voltage at the positions corresponding to the peak labels. Complete the frequency reference identification. As shown in Figure 3 in the process, the frequency reference to be locked corresponds to the triangular wave voltage, which is output to the laser current source through DA. Turn off the triangular wave sweep signal and simultaneously generate a trigger signal for the frequency closed-loop control module to start.

[0056] After the spectrum signal is collected, the first harmonic error signal of the absorption peak is obtained through a digital band-pass filter, and the error signal is used to obtain a low-noise peak error signal by using a Cordic quadrature demodulation algorithm. After receiving the PID start trigger signal generated by the frequency reference identification module, the error signal is obtained by a discrete PID control algorithm, and the PID correction amount U t is output to the laser current source through DA. The frequency automatic closed-loop control is completed. Wherein U t The P, I and D parameters K p , K i , and K d are expressed as follows.

[0057]

[0058] In the formula, K p is a proportional coefficient, K i is an integral coefficient, and K d is a differential coefficient, e t represents the error signal data collected at time t, e n represents the integral amount of the error signal within time t, and U t is the correction amount calculated by using the discrete PID control algorithm on the error signal. The amount is input to the current source through the voltage-to-current circuit inside the current source, and is converted into the tuning amount of the laser frequency.

[0059] In order to ensure the long-time high-stability locking of the laser frequency stabilization system, the spectrum signal error signal after locking is periodically collected to check whether the laser is locked and whether the stability after locking meets the requirements. As shown in Figure 3 After the PID correction amount is output to the laser current source, the spectrum signal and the error signal within time T (T=30min) are collected and saved at a sampling rate of 200Hz, and a total of N T sampling points are saved. The amplitude difference of the spectrum signal within time T is calculated The power spectral density of the error signal within time T is calculated, and the error signal power spectral density value at 10Hz is saved. When the amplitude difference of the spectrum signal is greater than the locking threshold Th_lock, it is determined that the frequency locking system is unlocked, as shown in Figure 5As shown in (a), after the 3-4 hour lock-in, the spectrum signal voltage jumps and never recovers to the locked state, which proves that the laser has been unlocked. The triangular wave sweep is restarted, the stored spectrum is entered and the peak identification is turned on, and the laser is relocked to the target frequency reference. When the spectrum signal amplitude difference is less than Th_lock, it is preliminarily determined that the frequency locking system is still locked. It is continuously determined whether the 10 Hz error signal power spectral density is less than the frequency locking accuracy threshold Th_accuracy. If the accuracy threshold is not met, the program jumps to the PID parameter adjustment module, and is relocked in a closed loop. Figure 5 As shown in (b), when the 10 Hz power spectral density amplitude is far lower than the threshold condition, it is determined that the accuracy does not meet the locking demand accuracy, and the PID parameters can be adjusted to relock. If the accuracy threshold is met, the register is emptied, the next T period error signal and spectrum signal are saved, and the locking state and locking accuracy are determined.

[0060] The above includes the description of the preferred embodiments of the present application, which is intended to specifically describe the technical features of the present application, and is not intended to limit the content of the application to the specific forms described in the embodiments. Other modifications and variations made in accordance with the main idea of the application are also protected by the present patent. The main idea of the application is defined by the claims, not by the specific description of the embodiments.

Claims

1. A laser automatic frequency stabilization system based on real-time automatic peak finding technology, characterized in that, It has the advantages of high peak recognition accuracy, strong anti-noise capability, high locking accuracy, and good long-term stability, including: frequency automatic tuning module A based on frequency-current-temperature model, frequency reference automatic identification module B based on adaptive threshold method, frequency automatic closed-loop control module C, and locking status monitoring and judgment module D for frequency unlocking and periodic monitoring of stability. The frequency auto-tuning module A, based on the frequency-current-temperature model, automatically tunes the laser current and temperature to the operating point within the frequency reference range according to the established laser frequency-current-temperature model, and activates the triangular wave sweep signal to produce spectral signals with saturation absorption peaks or resonance peaks. The frequency reference automatic identification module B based on the adaptive threshold method is triggered at the start edge of the sweep cycle and saves the spectral data within a single sweep cycle. It compares each data point in the spectral data with its preceding and following adjacent data to identify all local peaks in the spectral data. For each local peak, it determines its baseline and calculates the height and width from the peak to the baseline. Based on the height and width characteristics of the saturated absorption peaks or FP cavity resonance peaks in the spectrum, it sets height and width threshold conditions to filter out false peak interference caused by noise and Doppler background. Peaks that meet the threshold conditions are the saturated absorption peaks or FP cavity resonance peaks used as the frequency reference. The scanning voltage corresponding to the peak is stored and output to the laser current source to complete the frequency reference identification of the laser frequency stabilization system. Automatic closed-loop control module C uses a modulation and demodulation algorithm based on the Cordic algorithm to obtain the error signal of the laser frequency. According to the different error signals obtained from different frequency references, the initial values ​​of P, I, and D parameters are set, and the error signal is calculated by using a discrete PID control algorithm to stabilize the laser frequency at the set frequency reference point. The locking state judgment module D, which periodically monitors frequency unlocking and stability, periodically stores the spectral signal after the laser frequency is locked and calculates the power spectral density of the error signal within that period. First, based on the characteristics of the spectral signal within one period after locking and unlocking, a laser unlocking threshold condition is set. Then, based on the frequency stability requirements after locking, a threshold condition is set at the corresponding angular frequency of the error signal power spectral density to determine if the laser is unlocked. If unlocked, it re-enters the frequency reference automatic identification module B based on the adaptive threshold method, activates a triangular wave to re-find the frequency reference point, and re-locks. If unlocked, it determines whether the error signal power spectral density meets the stability threshold condition. If it meets the stability threshold condition, it enters the next judgment period. If it does not meet the stability threshold condition, it re-enters the automatic closed-loop control module C, adjusts the PID control parameters, re-closes the loop, and re-enters the locking state judgment module D after one period to determine the laser's locking state and stability, ensuring long-term frequency stability after the laser is powered on.

2. The laser automatic frequency stabilization system with real-time automatic peak finding technology according to claim 1, characterized in that: The automatic frequency tuning module A based on the frequency-current-temperature model is implemented as follows: (1) Calibrate the current-frequency coefficient of the semiconductor laser Temperature-frequency coefficient constant bias A frequency-current-temperature model was established for the semiconductor laser used. ; (2) Calibrate the current-power coefficient of the semiconductor laser The laser output power required by the atomic gyroscope or atomic magnetometer Calculate the laser operating current = ; (3) Based on the frequency-current-temperature model obtained in step 1, and the target frequency range and the calculated laser operating current Calculate the operating temperature of the laser. ; (4) The tuning operating point is , The triangular wave output signal is activated to sweep the frequency of the laser.

3. The laser automatic frequency stabilization system with real-time automatic peak finding technology according to claim 1, characterized in that: The frequency reference automatic identification module B based on the adaptive threshold method includes five steps: (1) Monitor the start and end edges of the triangular wave sweep cycle, and save the spectral data containing frequency reference information at the start and end edges to an array. ; (2) Compare the array data sequentially. If the data with array index i is... satisfy Then it is saved as a local maximum record. and its label It identifies and stores all peaks in a set of spectra. (3) Based on the different background baselines of different spectra and the different trends of peak height with the baseline, an adaptive peak baseline judgment method is adopted. Starting from the first peak, the baseline of each local peak is found sequentially, and the peak with array number i is the peak. Traverse left to the starting data with array index 0. Search to Minimum value between The peak value with array index i is traversed to the right until... Search to Minimum value between ;if < The coordinates are ( The point is taken as the base point, and a horizontal line is drawn through that point as the peak value. The baseline, and conversely, the baseline is taken from ( Draw a horizontal line and denote it as the peak baseline; After determining the peak baseline, re-apply using Traverse the data left and right until the first boundary point whose ordinate equals the base point value is encountered, and record each boundary point as a peak value. The coordinates of the left and right boundary points are ( )and( Therefore, peak value The amplitude is denoted as | |, peak The width is denoted as | |;Sequentially find the baseline for each peak to obtain the amplitude and width of each local peak relative to its baseline; (4) Since the peak amplitude of noise fluctuations is much smaller than that of the frequency reference peak, and the width of the peak formed by the Doppler background is much larger than that of the frequency reference peak, an adaptive amplitude threshold and width calculation method can be used to select the amplitude threshold as the basis for frequency reference. and width threshold The number of absorption peaks or resonances N in a single-cycle spectrum is determined based on the range of applied sweep voltage. The amplitudes of each local peak calculated in step (3) are sorted from smallest to largest using a heap sort algorithm, and the amplitude of the N+1th ranked peak is selected as the amplitude threshold. ; Sort all peak widths calculated in step three from largest to smallest using heap sort, and select the average width of the top N peaks as the width threshold. ; (5) Starting from the first local peak stored in step (2), traverse all local peaks and determine the peak amplitude threshold condition ( Peaks introduced by noise fluctuations are filtered out, and then the peak width is determined based on the criteria ( The peaks introduced by the Doppler background are screened out, and N peaks that meet the conditions are finally obtained as saturated absorption peaks or FP cavity resonant peaks. The triangular wave sweep voltage corresponding to the peak position is stored and output to the laser current source to complete the frequency reference identification of the laser frequency stabilization system.

4. The laser automatic frequency stabilization system with real-time automatic peak finding technology according to claim 1, characterized in that: The automatic closed-loop frequency control module C includes three steps: (1) After the acquired spectral signal is selected by a digital bandpass filter to obtain the odd harmonic components of the modulated spectrum, the Cordic orthogonal demodulation algorithm is used to solve the error signal of the laser frequency to improve the noise immunity of the system. (2) Store the sweep voltage at the position corresponding to the peak value identified by the frequency reference automatic identification module B, and output the voltage value at this point to the laser current source through the digital-to-analog converter circuit to turn off the triangular wave sweep signal; (3) Set initial values ​​for P, I, and D control parameters for error signals of different frequency references, use discrete PID control algorithm to obtain the correction control quantity for the error signal calculated by the module, and use the correction control quantity to stabilize the laser frequency at the set frequency reference point in real time.

5. The laser automatic frequency stabilization system with real-time automatic peak finding technology according to claim 1, characterized in that: The frequency unlocking and stability periodic monitoring locking state monitoring and judgment module D includes three steps: (1) Interval The error signal obtained from the frequency automatic closed-loop control module C is acquired and saved, and calculations are performed. Power spectral density of time error signal Store the power spectral density value at the angular frequency that reflects the frequency stability of the laser. ;interval Acquire and save spectral signals, calculate The amplitude of the spectral signal at time and Amplitude difference of spectral signals collected at different times ; (2) Set the power spectral density threshold of the error signal Spectral amplitude difference threshold Stability setting based on frequency lock This value was obtained through experimental measurement; based on the threshold condition. The laser's locked state is determined; if this condition is not met, the laser is considered unlocked; based on a threshold condition... The laser frequency locking stability status is assessed. If this condition is not met, the laser frequency locking accuracy is determined to be insufficient. If both conditions are met, the laser is determined to be locked and has achieved the required accuracy, and the process continues to wait for the next condition. Determine the cycle; (3) If the lock is lost, the frequency reference automatic identification module B based on the adaptive threshold method will be re-entered, the triangular wave will be activated to find the frequency reference point again, and the lock will be re-locked; If the laser is not lost, the system checks whether the power spectral density of the error signal meets the stability threshold condition. If it does, the system proceeds to the next judgment cycle. If it does not meet the stability threshold condition, the system re-enters the automatic closed-loop control module C, adjusts the PID control parameters, re-closes the loop, and re-enters the lockout judgment module D after one cycle to judge the laser's lockout status and stability, ensuring the long-term frequency stability of the laser after it is powered on.

6. A laser automatic frequency stabilization method based on real-time automatic peak finding technology of the system described in claim 1, characterized in that, The implementation is as follows: (1) Automatic tuning of laser frequency based on frequency-current-temperature model includes the following steps: calibrating laser power-frequency coefficient, frequency-current coefficient, frequency-temperature coefficient, and constant bias coefficient. Establish a laser frequency-current-temperature model; determine the laser output power and laser output frequency range according to the experimental requirements of atomic gyroscopes or atomic sensors; calculate the laser current and temperature operating point, and enable the triangular wave sweep signal; (2) Automatic identification of frequency reference based on adaptive threshold method includes the following steps: monitoring the start and end edges of the triangular wave sweep cycle and saving the spectral signal of one cycle; judging and saving all peaks in the spectral signal; adaptively judging the baseline of each peak and calculating the amplitude and width of each peak. The specific method is to traverse from the first peak to the starting point to the left and find the minimum value between these data. Traverse to the right until Minimum value between ;if < Then the coordinates are ( The point is marked with a horizontal line and recorded as the peak value. The baseline, and conversely, the coordinates are taken as ( Draw a horizontal line from the point ( ) as the peak baseline; after determining the baseline, re-establish the baseline using... Traverse the data left and right until the first boundary point where the function value equals the base point value is encountered, and record these as peak values. left boundary point ( ) and right boundary point ( Therefore, peak value The amplitude is denoted as | |, peak The width is denoted as | |; Repeat the above calculations for each peak to obtain the amplitude and width of each local peak; (3) The amplitude threshold and width threshold are adaptively calculated based on the sorting method. Specifically, the heap sort algorithm is used to sort all the peak amplitudes from smallest to largest, and the amplitude of the N+1th ranked peak is selected as the amplitude threshold. ; Sort all peaks by width from largest to smallest using heap sort, and select the average width of the top N peaks as the width threshold. Based on the peak amplitude threshold determination criteria ( Peaks introduced by noise fluctuations are filtered out, and then the peak width is determined based on the criteria ( The peaks introduced by the Doppler background are filtered out to obtain the peak values ​​that can be used as a frequency reference, and the sweep voltage corresponding to the peak center is saved. (4) Obtain the error signal of the frequency reference based on the Cordic quadrature demodulation algorithm, use the discrete PID control algorithm to calculate the frequency correction control quantity of the error signal, and use the correction control quantity to stabilize the laser frequency at the set frequency reference point in real time. (5) Periodically monitor the output status of the laser's spectral signal and error signal to determine whether the laser frequency has lost lock and whether the stability meets the conditions. The determination method is to periodically monitor the output status of the laser's spectral signal and error signal. Acquire and save spectral signals, calculate The amplitude of the spectral signal at time and Amplitude difference of spectral signals collected at different times If the amplitude difference satisfies If the condition is met, the laser is determined to have lost lock, and the frequency sweep is restarted to re-identify and lock the frequency reference; otherwise, the laser remains locked, and the interval is... Collect and save error signals, and calculate... Power spectral density of time error signal Save the power spectral density value at 10 Hz. Based on threshold conditions The laser frequency locking stability status is judged. If the threshold condition is met, it is determined that the laser frequency locking accuracy has not met the requirements. The PID parameters are readjusted and the correction amount is recalculated for closed-loop locking. If the threshold condition is not met, it is determined that the laser frequency locking accuracy meets the requirements, and the system waits for the next judgment cycle.

Images