A dual-pzt-driven high-speed sweep laser and a three-stage synchronous non-jump-mode sweep method thereof

Abstract

The application provides a dual-PZT driven high-speed sweep frequency laser and a three-stage synchronous non-jump mode sweep frequency method thereof, and belongs to the technical field of high-speed sweep frequency lasers. The application solves the technical problems that the existing sweep frequency laser has a sharp deterioration in synchronization accuracy and a frequent loss of mode lock due to nonlinear errors such as piezoelectric hysteresis, creep, temperature drift and high-frequency phase lag when the sweep frequency laser is tuned at a speed of kHz or above, and lacks environmental disturbance suppression capability and multi-scene adaptability. The dual-PZT driven high-speed sweep frequency laser comprises an optical resonant mode selection system, a dual-PZT tuning driving system, a differential ring capacitive displacement sensing system, an R-SOA internal source modulation frequency locking system, an FPGA intelligent three-stage synchronous control system and a multi-field coupled temperature control and environment monitoring system. An LSTM hybrid prediction model is embedded in the FPGA, and errors can be suppressed stage by stage. The application can realize continuous tuning without mode jump, has a sub-MHz level frequency synchronization accuracy, can be stably operated for ten thousand hours without maintenance, and is compatible with multiple scenes with one machine.

Description

Technical Field

[0001] This invention belongs to the field of swept-frequency laser technology, and relates to a high-speed swept-frequency laser driven by dual PZT, particularly a three-stage synchronous mode-skipping-free swept-frequency method for a high-speed swept-frequency laser driven by dual PZT. Background Technology

[0002] As a core light source in fields such as frequency-modulated continuous-wave lidar, optical coherence tomography, and spectral analysis, the performance of swept-frequency lasers directly determines the measurement accuracy and imaging resolution of the entire system. With the increasing demands for scanning speed, tuning range, and spectral purity in various applications, high-speed, mode-hopping-free swept-frequency lasers have become a key bottleneck restricting technological development. During laser frequency tuning, if the longitudinal modes of the resonant cavity and the passband of the filter lose synchronization, it will trigger longitudinal mode switching, resulting in mode hopping. This causes abrupt changes in the output light frequency and phase, severely degrading the signal-to-noise ratio and ranging accuracy of coherent detection.

[0003] To achieve mode-hopping-free frequency sweeping, existing technologies mainly employ two approaches. One approach relies on the mechanical rotation of gratings or mirrors, synchronously changing the external cavity length and the filter center wavelength to ensure the laser always operates in a single longitudinal mode; for example, the Littman / Littrow external cavity laser uses flexible hinges or star structures to adjust the grating angle, combined with synchronous tuning of the piezoelectric ceramic and the injected current, achieving continuous mode-hopping-free tuning up to approximately 105 GHz. The other approach uses dual PZTs to drive the etalon and external cavity mirror separately, achieving coarse synchronization through a fixed ratio of open-loop voltage. However, both approaches have inherent limitations: the mechanical tuning structure is complex, the motion control precision requirements are extremely high, and the scanning speed is slow (typically only tens of Hz); traditional dual PZT open-loop driving cannot effectively compensate for nonlinear errors such as hysteresis, creep, temperature drift, and high-frequency phase lag of the piezoelectric ceramic, resulting in a sharp deterioration in synchronization accuracy and frequent mode-hopping loss at high-speed frequency sweeps above kHz. In addition, existing systems lack the ability to actively suppress disturbances such as ambient temperature and vibration, resulting in poor long-term operational stability and difficulty in meeting the needs of multiple scenarios such as high-speed scanning and high-precision modes.

[0004] Based on this, we propose a high-speed sweeping laser driven by dual PZT and its three-stage synchronous mode-skipping sweeping method. Summary of the Invention

[0005] The purpose of this invention is to address the aforementioned problems in existing technologies by proposing a high-speed swept-frequency laser driven by dual PZT and its three-stage synchronous mode-hopping-free swept-frequency method. The technical problem to be solved by this invention is: how to achieve continuous tuning without mode hopping and sub-MHz frequency synchronization under high-speed swept-frequency, while taking into account adaptability to multiple scenarios and long-term stable operation at the tens of thousands of hours level.

[0006] The objective of this invention can be achieved through the following technical solutions:

[0007] A high-speed swept-frequency laser driven by dual PZTs comprises three main parts: a laser hardware system, a three-level synchronous mode-hopping-free swept-frequency control method, and a mechanism-LSTM hybrid intelligent compensation algorithm. These include an optical resonant mode selection system, a dual PZT tuning drive system, a differential ring capacitor displacement sensing system, an R-SOA intrinsic modulation frequency locking system, an FPGA intelligent three-level synchronous control system, and a multi-field coupled temperature control and environmental monitoring system.

[0008] The optical resonant mode selection system performs gain amplification, coarse mode selection, fine mode selection, and beam splitting detection, and includes the following units:

[0009] The R-SOA active gain unit employs an InP-based reflective semiconductor optical amplifier chip, with a gain bandwidth covering the C-band or extended C+L band. The R-SOA chip has an anti-reflection coating on its front end and a high-reflection coating on its rear end, forming a resonant cavity with the external cavity mirror. It features a DC bias current of 150mA, an RF modulation bandwidth of 100MHz, a modulation peak-to-peak value of 10mA, and a carrier differential gain of [missing information]. The relaxation oscillation frequency is 700MHz. The chip is mounted on an independent TEC temperature-controlled stage with a temperature stability of ±0.01℃.

[0010] The 6T-FP standard etalon coarse selection unit utilizes a 25μm air-gap fused silica parallel substrate, with multilayer dielectric films fabricated on both sides via ion beam assisted deposition. The film system is designed as follows: Structure. Basic fineness F=60, FSR=6THz, passband full width at half maximum (FWHM)=FSR / F=100GHz. The etalon is driven by the first PZT to change the air gap length, achieving 6THz coarse filtering.

[0011] The fine mode selection unit of the external cavity mirror uses a planar dielectric film mirror with a reflectivity of 80%±1% and a substrate material of fused silica, with a surface accuracy better than λ / 10. The mirror and the far end face of the R-SOA form an external cavity resonant cavity with an external cavity FSR > 10 GHz and a corresponding external cavity physical length < 15 mm. The mirror is driven by a second PZT, and fine tuning of the longitudinal mode frequency is achieved by changing the external cavity length.

[0012] The beam splitting and coupling detection unit sets a 1% beam splitter in the standard etalon transmission optical path. After passing through the focusing lens, the beam is received by a high-speed InGaAs PIN photodiode with a bandwidth of 125MHz and a responsivity of 0.9A / W@1550nm. The output optical beat frequency signal is sent to the frequency-locked demodulation circuit.

[0013] The main laser output coupling unit uses an aspherical lens to collimate the diverging laser output from the R-SOA. The lens has a numerical aperture (NA) of 0.5~0.8 and a focal length of 1.5~3.0 mm. The laser then enters a single-mode fiber via a precision fiber coupler and outputs through an FC / APC standard interface, with a return loss >60 dB.

[0014] The electrically controlled dual-etalon switching unit incorporates two 6T etalons: an F=60 high-speed etalon and an F=200 high-precision etalon. An electrically controlled translation stage completes the optical path switching. The hybrid modeling and computation unit pre-stores two Airy transmission mechanism parameter libraries, automatically updating the passband FWHM, loss coefficient, and synchronization error threshold after etalon switching. The F=60 etalon has a single-pass 10GHz frequency offset loss of 0.17dB and a wide synchronization tolerance, suitable for high-speed radar scenarios; the F=200 etalon has a narrower passband and stronger side-mode suppression, suitable for high-precision OCT scenarios. One device is compatible with both applications without requiring replacement of the entire unit.

[0015] A dual-axis tilt closed-loop correction unit for the reflector, equipped with a miniature shear piezoelectric trimmer, extracts the DC loss component from the 50MHz modulation error signal as the tilt error source, which is then superimposed onto the optical precision closed-loop PID output to dynamically correct the mirror's pitch / yaw tilt. This keeps the tilt error stably controlled within 0.1μrad over the long term, reducing the SMSR fluctuation from ±2dB to ±0.5dB.

[0016] The beam splitter adaptive gain adjustment unit features a transimpedance amplifier with multiple digitally controllable gain settings, and the FPGA automatically switches gain levels based on the main output power monitoring signal. Hybrid modeling adds an optical power loss correction term to compensate for the transmission loss fluctuations of the beam splitter and etalon with wavelength / temperature, thus extending the linear range of the error signal.

[0017] Synchronous optical constraint formula:

[0018]

[0019] ,

[0020] Where K is the synchronization ratio. For the standard to have a free spectral range, For the free spectral range of the external cavity, The optical length of the external cavity. Let c be the standard air gap thickness, c be the speed of light in vacuum, and λ be the center wavelength of the laser. The displacement required for synchronous tuning of the external cavity PZT to one FSR.

[0021] The dual PZT tuning drive system is either a dual shear type PZT tuning drive system or a dual telescopic PZT tuning drive system.

[0022] The dual-shear type PZT tuning drive system realizes piezoelectric tuning with standard air gap and outer cavity length, and includes the following units:

[0023] The standard PZT shearing actuator uses a multilayer piezoelectric ceramic shear plate operating in shear mode. It measures 5mm × 5mm × 2mm, with a piezoelectric shear coefficient d15 ≥ 800pC / N, a load resonant frequency ≥ 20kHz, a stroke ±1μm, closed-loop resolution ≤ 0.1nm, hysteresis < 1%, and tilt error < 0.1μrad. The single FSR tuning displacement is λ / 2 ≈ 775nm. The PZT shear plate has gold electrodes on both sides and is equipped with a flexible hinge anti-tilt limiting structure, with a lateral offset < 0.5mrad.

[0024] The external cavity shearing PZT actuator uses a multilayer piezoelectric ceramic shear plate operating in shear mode. Its dimensions are 10mm × 10mm × 5mm, with a piezoelectric shear coefficient d15 ≥ 800pC / N, a load resonant frequency ≥ 15kHz, a stroke ± 500μm, a closed-loop resolution ≤ 0.5nm, hysteresis < 5%, and tilt error < 0.5μrad. The single FSR synchronous tuning displacement is 600 × 775nm = 465μm.

[0025] Two PZTs are driven independently with a fixed synchronization ratio of 600:1, satisfying the synchronization displacement constraint. ,in, The displacement required for synchronous tuning of the external cavity PZT to one FSR. The displacement required to tune an FSR for a standard etalon PZT.

[0026] The high-voltage amplification and drive unit employs a high-voltage operational amplifier to amplify the 0~5V low-voltage control signal output from the FPGA to a 0~150V piezoelectric drive voltage, with a slew rate ≥50V / μs, output current ≥100mA, and bandwidth ≥100kHz. It incorporates built-in overvoltage, overcurrent, and short-circuit hardware protection circuits, as well as EMI shielding and filtering. An LC low-pass filter is configured at the output to suppress high-frequency noise.

[0027] The PZT status acquisition and protection unit installs a temperature thermistor and a vibration acceleration sensor near the PZT shear plate to collect the PZT ceramic temperature and structural vibration amplitude in real time. When the threshold is exceeded, the high-voltage drive is cut off and the data is uploaded to the hybrid modeling unit for error compensation.

[0028] The layered composite phase feedforward unit consists of three levels of feedforward superposition: PZT second-order dynamics-based phase compensation; additional compensation term for scanning acceleration; and independent phase coefficient tables for 1kHz, 2kHz, 5kHz, 10kHz, and 30kHz ranges. At 10kHz scanning, the PZT phase lag is reduced from 170° to less than 20°, and further compressed to less than 8° at the 30kHz extreme scanning, resulting in a 40% reduction in synchronization error.

[0029] The creep online identification and correction unit continuously acquires hourly creep displacement data under closed-loop steady-state conditions, and incrementally updates the hybrid modeling four-dimensional residual database in real time. The hysteresis voltage range is fitted using 20 polynomial segments, replacing single cubic interpolation. The creep synchronization error during 8 hours of continuous scanning was reduced from 1.5MHz to 0.3MHz.

[0030] An adaptive bandwidth current limiting protection unit is used, featuring a digitally controllable RC bandwidth network. The FPGA automatically switches the drive bandwidth based on the real-time scanning frequency. The upper limit of the output current is dynamically adjusted based on the temperature collected by the PZT thermistor, reducing power consumption at high temperatures to slow piezoelectric aging and suppressing high-speed resonance noise.

[0031] The dual-telescopic PZT tuning drive system realizes piezoelectric tuning with standard air gap and outer cavity length, and includes the following units:

[0032] The standard etalon telescopic PZT actuator uses a stacked or tubular telescopic piezoelectric ceramic actuator. It drives the 6T-FP standard etalon air gap thickness change through longitudinal telescopic movement to achieve coarse mode selection tuning in the 6THz range. The single FSR tuning displacement is λ / 2≈775nm, and the stroke meets the full range scanning requirements of the standard etalon.

[0033] The external cavity telescopic PZT actuator unit adopts a large-stroke stacked piezoelectric ceramic actuator. It drives the displacement of the external cavity mirror through longitudinal telescopic movement, changing the optical length of the external cavity to achieve fine tuning of the longitudinal mode frequency. The single FSR synchronous tuning displacement is 600×775nm=465μm, the stroke is ±500μm, and it maintains a fixed synchronization ratio of 600:1 with the standard PZT.

[0034] The high-voltage amplification drive unit amplifies the low-voltage control signal output by the FPGA to a 0~150V high-voltage drive signal. It has high-voltage slew rate and wide bandwidth characteristics, drives dual-channel telescopic PZT actuators to respond quickly, and has built-in overvoltage, overcurrent, short-circuit hardware protection and EMI filtering.

[0035] The PZT status acquisition and protection unit installs a temperature thermistor and a vibration acceleration sensor near the telescopic PZT actuator to collect the PZT ceramic temperature and structural vibration amplitude in real time. When the threshold is exceeded, the high voltage drive is cut off and the data is uploaded to the hybrid modeling unit for error compensation.

[0036] The layered composite phase feedforward unit integrates the second-order dynamics basic phase compensation of PZT, the additional compensation of scanning acceleration, and the three-level superposition of graded frequency coefficients to compensate for the phase lag of the stretchable PZT under high-speed scanning. The phase lag is reduced from 170° to less than 20° under 10kHz scanning.

[0037] The creep online identification and correction unit continuously collects hourly creep displacement under closed-loop steady state, updates the hybrid modeling four-dimensional residual database in real time incrementally, and fits piecewise polynomials in the hysteresis voltage range. The creep synchronization error is reduced from 1.5MHz to 0.3MHz after 8 hours of continuous scanning.

[0038] The adaptive bandwidth current limiting protection unit sets up a digitally controllable RC bandwidth network, and the FPGA automatically switches the drive bandwidth according to the real-time scanning frequency; it dynamically adjusts the upper limit of the output current based on the temperature collected by the PZT thermistor, reduces power at high temperatures to slow down piezoelectric aging, and suppresses high-speed resonance noise.

[0039] The differential ring capacitive displacement sensing system is integrated into two PZT shear plates, realizing coarse positioning based on self-capacitance plus high-precision detection based on differential capacitance, and includes the following units:

[0040] A magnetron sputtering ring electrode sensing unit is constructed by fabricating a ring-shaped differential capacitance sensor electrode on the surface of a PZT shear sheet using magnetron sputtering. The ring electrode has an inner diameter of 1.5 mm, an outer diameter of 2.5 mm, and a thickness of 100 nm. The electrode material is platinum, and the adhesion layer is titanium. Magnetron sputtering process parameters: background vacuum... The sputtering pressure is 0.5~1.0 Pa, the sputtering power is 100~200 W, and the deposition temperature is room temperature to 150℃. The inner and outer rings form differential feedback capacitors Cin+ and Cin-. The self-capacitance channel realizes coarse positioning of PZT, and the differential capacitance detects nanometer-level displacement with a resolution ≤0.08 nm.

[0041] The capacitor signal modulation and demodulation unit uses the differential capacitance detection principle to eliminate common-mode interference. It includes a high-frequency AC excitation source, an instrument differential amplifier, an active bandpass filter, and a 16-bit ADC sampling circuit with a noise floor of <0.01nm. It converts minute differential capacitance changes into a standard 0~3.3V analog voltage signal.

[0042] The displacement synchronization transmission unit uses a FIFO high-speed cache and an LVDS differential transmission chip to synchronously transmit dual-channel displacement data to the FPGA displacement closed-loop calculation unit without transmission delay.

[0043] The electrode temperature drift common-mode decoupling unit incorporates a hybrid modeling approach, adding an electrode capacitance temperature coupling mechanism equation to separate two types of errors: PZT thermal expansion and electrode dielectric constant temperature drift. The demodulation circuit also incorporates common-mode negative feedback amplification. The zero-point drift due to ±10℃ temperature changes is reduced from 0.8nm to 0.15nm.

[0044] The dual-channel bit synchronization timing alignment unit adds a hardware synchronization latch signal to the LVDS transmission, which is from the same source as the FPGA's 100kHz master control clock; the hybrid modeling adds a sampling time difference compensation term to eliminate the dynamic synchronization error caused by the microsecond-level sampling offset of the two ADCs.

[0045] The adaptive vibration filtering unit analyzes vibration amplitude in real time by inputting data from PZT vibration sensors and dynamically adjusts the filter window length—lengthening the filter window during severe vibrations and shortening it during static measurements. The frequency error caused by environmental vibrations in the 1~10kHz range is reduced from 1.3MHz to 0.2MHz.

[0046] The R-SOA internal modulation frequency locking system generates a double-sideband error signal based on R-SOA current modulation, replacing the traditional PDH external electro-optic modulation scheme, and includes the following units:

[0047] The 50MHz DDS generator unit uses the AD9833 DDS chip and a low-temperature drift crystal oscillator. It outputs a 50MHz sine wave through a phase-locked loop and a low-noise operational amplifier AD8065, with a frequency accuracy of 0.1Hz and a phase noise of <-120dBc / Hz@10kHz. It is synchronous with the demodulation local oscillator.

[0048] The R-SOA bias superposition unit consists of a 150mA constant current source, a broadband current mirror, a DC blocking high-voltage capacitor, and an RF choke. It uses a broadband bias three-way circuit to superimpose a 50MHz sinusoidal modulation signal onto the R-SOA DC bias. The operating frequency range is 10kHz~100MHz, with insertion loss <0.5dB and 50Ω impedance matching. The peak-to-peak modulation current is 10mA, and the bandwidth is 100MHz. Modulation expression: ,in, The DC bias current is set to 150mA. The peak-to-peak value of the modulation current / modulation amplitude is taken as 5~15mA. The frequency is 50MHz, the modulation depth is 0.1~0.5rad, and t is time. The instantaneous phase of the modulated signal, This is the modulation angular frequency.

[0049] The photoelectric transimpedance amplifier unit uses a Thorlabs PDA10CF high-speed PIN photodiode in conjunction with an AD8015 transimpedance amplifier, achieving a transimpedance gain of [missing information]. The bandwidth is 125MHz, and the noise level is <1nV / √Hz. The output signal is then fed into a mixer after passing through a 10MHz high-pass filter to remove low-frequency noise.

[0050] The dual-balanced mixer demodulation unit uses a Mini-Circuits ADE-1 mixer with 50Ω impedance matching, 0dBm local oscillator, and 6dB conversion loss. The photoelectric signal is mixed with the 50MHz local oscillator signal and then passed through a fourth-order Butterworth low-pass filter to output a low-frequency error signal. Error signal formula: ,in, Frequency detuning / frequency deviation For the standard etalon transmission function, For optical carrier frequency, For radio frequency modulation frequency, For the upper sideband frequency, This refers to the lower sideband frequency;

[0051] When Δv=0, the error signal returns to zero, corresponding to the complete alignment of the transmission peak of the external cavity longitudinal mode with the standard etalon; when there is a deviation, the amplitude and polarity of the error signal reflect the magnitude and direction of the deviation.

[0052] The error signal conditioning unit uses the zero-drift operational amplifier AD8628, equipped with a DC offset adjustment potentiometer and a limiting protection circuit, to amplify the demodulated error signal to a standard level of 0~±5V and send it to the FPGA optical precision closed-loop operation unit.

[0053] The modulation amplitude adaptive unit dynamically adjusts the modulation current in the range of 5~15mA. The FPGA adjusts the DDS output amplitude in a closed loop according to the output amplitude of the photodetector to maintain the modulation depth stable at the optimal operating point of 0.3rad, and the signal-to-noise ratio of the error signal is uniform and stable across the entire band.

[0054] The demodulation zero-point self-calibration unit automatically shuts off 50MHz modulation during the lockout gap, stores the inherent DC offset of the acquisition and demodulation circuit in the compensation register, and superimposes real-time offset corrections onto the optical PID input. The long-term zero-point drift is reduced from 0.08MHz to 0.01MHz.

[0055] The filter bandwidth dynamic matching unit enables high-precision switching between 1~2kHz operating conditions and 5kHz low-pass bandwidth to improve lock-in response; it maintains 3kHz low-pass to suppress harmonics in 10~30kHz high-speed operating conditions; the high-pass filter adaptively adjusts the cutoff frequency according to optical power to reduce weak light signal loss.

[0056] The FPGA intelligent three-level synchronous control system is the main computing unit, using a Xilinx Zynq / Artix7 FPGA as the main control chip and a 100kHz unified sampling clock. It includes the following units:

[0057] The mechanism-LSTM hybrid modeling feedforward unit incorporates a PZT second-order underdamped dynamics model, an FP etalon transmission Airy model, a laser rate equation, and a 600:1 synchronous displacement constraint, among other underlying physical mechanism layers. It is complemented by a four-dimensional measured residual database correction layer, which calculates the compensation voltage in real time through three-dimensional cubic interpolation. The LSTM neural network uses 6-dimensional features from a 50-step historical window—driving voltage, actual displacement, temperature, voltage derivative, displacement derivative, and scan frequency—as input to predict the physical model error compensation amount. Hybrid model formula:

[0058] ,

[0059] ,

[0060] in, To ultimately predict the displacement, To predict displacement based on physical mechanisms, The amount of compensation for LSTM time series prediction; This is the final PZT drive voltage. For physical mechanism theory voltage, For the target displacement of the standard etalon, This is the final PZT drive voltage. The current temperature. For the rate of temperature change, For scanning acceleration, Correcting voltages for the four-dimensional residual database. For vibration compensation voltage, This is the tilt compensation voltage.

[0061] The first-level open-loop coarse synchronization unit relies solely on hybrid modeling feedforward voltage to drive the dual PZTs, with no displacement feedback and a synchronization error of 50~100nm. It is used for rapid debugging of prototypes and provides an initial operating point for subsequent closed-loop circuits.

[0062] The unit employs a two-stage dual-channel displacement PID controller with two independent PID cores, a bandwidth of 10kHz, and an anti-integral saturation module. The basic PID parameter values ​​are: Kp=2.5, Ki=0.8, Kd=0.1. It inputs the measured displacement from the dual-channel differential capacitance sensing unit and outputs a corrected external cavity PZT drive voltage, compressing the synchronization error to ±1nm.

[0063] A three-level optical precision PID unit with a low-speed PID core and a bandwidth of 1~5kHz. Basic PID parameter values: Kp_opt=0.5, Ki_opt=0.05, Kd_opt=0.01. Lock-in threshold |e(Δv)|<0.1MHz. It receives and demodulates optical error signals, superimposes them onto the displacement closed-loop output, achieving MHz-level frequency precision alignment, SMSR stability >45dB, and automatic compensation for temperature drift, PZT aging, and optical component deformation.

[0064] The timing unit is mode-hopping-free, featuring a high-speed state machine equipped with a 100μs high-precision timer and a phase-locked loop (PLL) capture scan generator. Triggering occurs when the etalon PZT displacement reaches +775nm: etalon PZT returns to zero → external cavity PZT jumps to -465μm → 100Hz small-amplitude scan capture → re-locking. Timing specifications: PZT step < 100μs, loop capture < 1ms, total mode-hopping time < 2ms, power fluctuation ≤ 0.3dB.

[0065] The recapture unit for lost-lock operation includes an error amplitude threshold comparator and a 3-recapture count register. Upon loss of lock, it immediately switches to a secondary displacement closed loop, with the external cavity PZT scanning at 200Hz and ±5~20nm to capture the zero point; successful capture restores the tertiary closed loop; three consecutive failed captures trigger a full-range reset and update the residual database.

[0066] The DAC output unit consists of two 16-bit AD5761 DACs with a 1MHz update rate, a synchronization timing synchronizer, and digital limiting protection. It converts the final control voltage of the three-stage synchronization into an analog signal and outputs it to the high-voltage PZT amplifier unit.

[0067] The four-dimensional database incremental self-updating unit automatically collects four-dimensional data of voltage, scanning frequency, temperature, and displacement error during mode skipping reset and lockout recapture intervals. It uses local cubic interpolation to incrementally update the database without overwriting the factory calibration data. The entire unit can operate for 8000 hours without needing to be returned to the factory for recalibration, and the synchronization accuracy decay is <0.05MHz.

[0068] The three-level adaptive PID self-tuning unit performs a 100Hz ±10nm small-amplitude self-tuning scan after each power-on and mode jump reset. It automatically retrieves three sets of PID parameters (high precision, standard, and extreme high speed) based on the current scan frequency, temperature, and etalon precision, reducing closed-loop overshoot and lockout probability by 60%.

[0069] The optical field dynamics hard-constraint unit embeds a 443 ps photon lifetime and a 700 MHz R-SOA relaxation oscillation frequency as hard constraints into the mechanism layer. When the scan rate approaches the relaxation oscillation limit, it automatically reduces the scan acceleration and increases the feedforward compensation weight. The physical basis for this constraint is that the shortest optical field setup time is approximately 2 ns, and the relaxation oscillation frequency corresponds to the maximum response rate of the gain medium. At the 30 kHz limit scan, the power attenuation is reduced from 15% to 7%, and the dynamic linewidth is compressed by 30%.

[0070] The hierarchical fault diagnosis unit can separate four types of error sources in real time: PZT phase lag, temperature drift, environmental vibration, and optical tilt. It outputs standardized fault codes and accurately locates the corresponding secondary functional unit, improving fault diagnosis efficiency by 80%.

[0071] The automatic operating condition identification unit reads the sweep frequency command from the host computer and automatically identifies the sweep frequency. It allows for one-click switching of feedforward coefficient, closed-loop bandwidth, and etalon mechanism parameters. It also features built-in hard protection for sweep rate, PZT acceleration, and optical error threshold, with automatic speed reduction when exceeding the boundary to avoid device damage and SMSR drop.

[0072] The fully automatic calibration unit enables one-button start-up for full-dimensional automated calibration. It automatically traverses voltage and temperature, scanning frequency segment tests, collects displacement, optical error, and power data, automatically generates and stores a four-dimensional residual database, and simultaneously outputs a factory performance grading report. The calibration time for a single machine is reduced from 4 hours to 40 minutes.

[0073] The multi-field coupled temperature control and environmental monitoring system suppresses temperature drift errors and includes the following units:

[0074] The standard etalon temperature control unit features a miniature TEC cooling chip with a high-precision NTC thermistor and an aluminum thermally conductive base, achieving a temperature control accuracy of ±0.01℃ and real-time acquisition of the standard etalon air gap substrate temperature.

[0075] The R-SOA temperature control unit features a TEC chip platform with thermal resistance pads and a heat dissipation metal housing, providing a temperature control accuracy of ±0.01℃, stabilizing the gain of the chip's bandgap wavelength, and suppressing carrier refractive index temperature drift.

[0076] External endoscope temperature control unit, small TEC with heat insulation bracket and temperature acquisition circuit, temperature control accuracy ±0.01℃.

[0077] Temperature refractive index correction formula: ,in, ΔT represents the refractive index at temperature T, 0.000273 represents the thermo-optic coefficient of fused silica, and ΔT represents the temperature change. The three temperature data are uploaded to the hybrid modeling and computing unit every 100ms, and the synchronization ratio is corrected in real time.

[0078] A three-channel temperature control cross-decoupling unit is used to establish a three-channel temperature coupling mechanism matrix, and hybrid modeling is used to decouple heat conduction cross-interference in real time. When a single channel experiences a temperature change, the synchronization ratio offset of the other two channels is pre-compensated. The synchronization error for a 1℃ step temperature change is reduced from 1.5MHz to 0.2MHz.

[0079] A temperature change rate feedforward pre-compensation unit is added, and the temperature change rate dT / dt is acquired by thermistor differentiation. A dynamic temperature drift compensation term is added to the hybrid model. The stabilization time for sudden ambient temperature changes is reduced from 30s to 3s.

[0080] The dynamic linewidth monitoring unit uses the delayed self-heterodyne interferometry method to analyze the dynamic linewidth in real time through short-time Fourier transform; when the linewidth expansion exceeds the threshold, it automatically adjusts the R-SOA bias current or adjusts the PZT drive waveform to achieve active compression.

[0081] The frequency modulation linearity monitoring and predistortion unit uses wavelength monitoring data to calculate the deviation between the actual frequency-time curve and the ideal linear curve, and uses an iterative predistortion algorithm to shape the swept frequency waveform, reducing the frequency sweep nonlinearity by more than an order of magnitude.

[0082] The PZT aging condition monitoring and life prediction unit periodically records the PZT's displacement-voltage curve, resonant frequency, and capacitance sensor baseline, extracts aging characteristics, predicts the remaining life through an LSTM model, and adaptively adjusts the controller parameters to extend the effective service life.

[0083] Globally embedded long-term reliability optimization logic:

[0084] (1) Vibration active collaborative compensation logic: PZT vibration sensor data is connected to the hybrid modeling and computing unit in real time to calculate vibration displacement disturbance, and superimposed reverse compensation voltage to offset 1~10kHz environmental vibration.

[0085] (2) PZT aging long-term slow compensation logic: The whole machine automatically performs micro static calibration every 10 hours of continuous operation, updates the piezoelectric coefficient theoretical model, compensates for long-term piezoelectric coefficient drift, and extends the stable service life to more than 10,000 hours.

[0086] A three-level synchronous mode-hopping-free frequency sweep control method for a high-speed frequency sweep laser driven by dual PZTs is proposed, employing a three-level hierarchical error suppression architecture of "open-loop coarse synchronization + displacement closed-loop intermediate synchronization + optical error fine closed-loop". The overall process is divided into five execution stages: fully automatic mechanism-data hybrid modeling offline calibration stage, three-level hierarchical adaptive synchronous real-time control stage, adaptive smooth mode-hopping-free reset continuous frequency sweep stage, hierarchical diagnostic loss-of-lock detection and recapture stage, and background long-term adaptive compensation stage.

[0087] Phase 1: Fully Automated Mechanism-Data Hybrid Modeling and Offline Calibration

[0088] Performed by the FPGA fully automatic calibration unit:

[0089] S1.1. The multi-zone temperature control unit automatically traverses the temperature range of -10℃ to 60℃ in segments;

[0090] S1.2. The FPGA open-loop unit automatically outputs 0~10V drive voltage in segments, and the displacement sensing unit synchronously collects nanometer-level displacement data;

[0091] S1.3. Automatically traverse multiple scanning frequencies of 1kHz, 2kHz, 5kHz, 10kHz, and 30kHz to collect the original errors of phase hysteresis, delay, and creep.

[0092] S1.4. Automatically construct a four-dimensional residual database, complete the solidification of the cubic interpolation function, and simultaneously fit the second-order dynamics of PZT, the transmission Airy function of the FP etalon, the laser rate equation, and the complete set of mechanism parameters of the optical field dynamics.

[0093] S1.5. Train the LSTM timing compensation network and burn the network weights and database to the FPGA on-chip Flash;

[0094] S1.6. Automatically complete equipment classification determination, store the corresponding operating condition parameter library, and output a complete factory calibration report.

[0095] Phase Two: Three-Level Hierarchical Adaptive Synchronous Real-Time Control

[0096] S2.1. Open-loop coarse synchronization. After the host computer issues a frequency sweep command, the automatic operating condition identification unit determines the operating level and retrieves the corresponding layered composite phase feedforward coefficients and etalon fineness mechanism parameters. The hybrid modeling and computing unit superimposes multi-dimensional compensation quantities such as creep, temperature change rate, vibration, and optical power loss, and outputs a 600:1 ratio feedforward drive voltage to coarsely align the etalon transmission peak with the external cavity longitudinal mode. The open-loop synchronization error is 50~100nm, providing an initial operating point for subsequent closed-loop operation.

[0097] S2.2. Synchronization in Displacement Closed-Loop. After common-mode temperature drift decoupling and adaptive vibration filtering preprocessing, the differential ring capacitive sensing unit synchronously latches dual-channel displacement data and sends it to a dual-channel adaptive PID controller. The PID self-tuning unit matches the optimal control parameters in real time based on the current operating conditions, compares the target displacement output with the hybrid model, and the dual-channel 10kHz bandwidth PID controller corrects the drive voltage in real time, eliminating static hysteresis, creep, and sampling time difference errors. The synchronization error is compressed to ±1nm.

[0098] S2.3. Precise Closed-Loop Optical Error Correction. A 50MHz DDS unit adaptively adjusts the modulation current to maintain an optimal modulation depth of 0.3rad, and the filter bandwidth automatically switches according to operating conditions; the demodulation circuit superimposes the zero-point self-calibration offset in real time; the tilt correction unit finely adjusts the dual-axis piezoelectricity of the reflector based on loss errors. The error signal, after mixing and demodulation, is input to the PID controller of the external cavity PZT for fine adjustment, with the alignment accuracy of the longitudinal mode and the etalon passband as the final control target. Synchronization accuracy ≤ ±0.1MHz, SMSR stability > 45dB, and automatic compensation for temperature drift, PZT aging, optical element deformation, and beam tilt.

[0099] The synchronously operating optical field dynamics constraint unit automatically reduces the scanning rate and increases the feedforward compensation weight when the scanning rate approaches the relaxation oscillation limit, thus suppressing power attenuation and dynamic linewidth broadening. The layered error diagnosis unit outputs the amplitude values ​​of various error sources and fault codes in real time.

[0100] Phase 3: Adaptive Smooth Mode-Jump-Free Reset Continuous Frequency Sweep

[0101] S3.1. The displacement sensing unit monitors the displacement of the etalon PZT in real time, and triggers the timing unit when the displacement reaches +775nm.

[0102] S3.2. The timing unit executes a three-stage smooth step transition: the etalon PZT returns to zero in three steps, and the external cavity PZT synchronously transitions to -465μm in three steps, with a total step response time of <100μs.

[0103] S3.3. Adaptively configure the capture scan amplitude according to the current synchronization error amplitude, and perform a 100Hz small scan to search for the optical error zero point;

[0104] S3.4. Upon detecting zero error, it automatically switches to a three-stage optical precision closed loop, with a loop acquisition time of <1ms, a total mode hopping time of <2ms, and output optical power fluctuation ≤0.3dB. During mode hopping intervals, it synchronously performs incremental self-updates of the residual database to compensate for short-term creep and temperature drift. Cyclic execution achieves continuous frequency sweeping with no mode hopping throughout the C-band 4.4THz range.

[0105] Phase Four: Hierarchical Diagnostic Loss-Lock Detection and Recapture

[0106] S4.1. The optical error conditioning unit outputs the error amplitude in real time. After the unlocking logic unit determines that the lock has been lost, it immediately switches to the second-level displacement closed loop.

[0107] S4.2. Adaptive capture scanning amplitude performs a 200Hz scan to search for the optical error zero point. Successful capture restores the three-level closed loop.

[0108] S4.3. If three consecutive capture failures occur, a full-range reset is performed, and the temperature, vibration, and scanning frequency data of this loss of lock are recorded simultaneously, and the hybrid modeling residual database is incrementally updated.

[0109] S4.4. The hierarchical error diagnosis unit outputs standardized fault codes and uploads them to the host computer to complete fault tracing.

[0110] Phase 5: Long-term adaptive compensation in the background

[0111] The machine automatically initiates a micro static calibration every 10 hours of continuous operation to update the PZT piezoelectric coefficient and compensate for device aging drift.

[0112] Vibration sensor data is input to the hybrid modeling unit in real time, and a reverse compensation voltage is superimposed to offset environmental vibrations of 1~10kHz.

[0113] Temperature control coupling decoupling and temperature change rate pre-compensation operate continuously in real time to dynamically offset the synchronization offset caused by changes in ambient temperature.

[0114] The complete algorithm for mechanism-LSTM hybrid modeling is as follows:

[0115] Layered mathematical mechanism formula:

[0116] (1) The underlying mechanism of second-order dynamics of PZT: in, Let be the transfer function, and s be the Laplace operator. The undamped natural angular frequency, The damping ratio is 0.05~0.15, and fr is the resonant frequency of the PZT under load.

[0117] Phase lag: Where f is the driving frequency, arctan(·) is the arctangent function, and fn is the resonant frequency. When the driving frequency f is much lower than the resonant frequency fn, the PZT response is flat. When f approaches fn, the response amplitude increases sharply and the phase lag increases rapidly to 90°. When f exceeds fn, the phase lag approaches 180° and the response amplitude decays rapidly.

[0118] (2) Theoretical synchronous displacement constraints: ,in, t represents the theoretical displacement of the external cavity at time t, and 600 is the synchronization ratio. Let t be the air gap displacement of the etalon at time t.

[0119] (3) Improved Preisach hysteresis model: Where x(t) is the PZT displacement, It is a hysteretic nonlinear term. Let be the linear piezoelectric term, N be the number of hysteresis operators, u(t) be the driving voltage, γ be the hysteresis kernel function, and μi be the weighting coefficient of the i-th hysteresis operator. The rising edge flip threshold for the i-th hysteresis operator. η is the rising edge toggling threshold of the i-th hysteresis operator, and η is the linear piezoelectric coefficient.

[0120] (4) Multi-source coupling error decomposition model:

[0121] in, This is the final PZT drive voltage. The hysteresis error component is represented by V, where V is the driving voltage. Let t be the creep error component, and t be time. This is the phase lag error component. For the sweep frequency, For scanning acceleration, This is the temperature drift error component. For temperature deviation, For the rate of temperature change, For vibration error components, This is the optical loss error component. To output optical power.

[0122] (5) Mechanism-data hybrid correction output model:

[0123] ,

[0124] in, This is the final PZT drive voltage. For physical mechanism theory voltage, For the target displacement of the standard etalon, For the sweep frequency, The current temperature. For the rate of temperature change, For scanning acceleration, Correcting voltages for the four-dimensional residual database. For vibration compensation voltage, This is the tilt compensation voltage.

[0125] (6) Feedforward phase lead compensation:

[0126] in, This is the feedforward compensation control signal, where FF stands for Feed-Forward and t is time. The original driving signal is τ, where τ is the phase compensation time constant. The first derivative of the driving signal. Let fn be the second derivative of the driving signal, and fn be the natural frequency of the PZT. These are the coefficients of the second-order Taylor expansion.

[0127] (7) Hard constraint criterion for optical field dynamics: in, For frequency scanning rate, For frequency infinitesimal elements, Let be the time infinitesimal element, and 2 be the constant coefficient. This is the resonant frequency of the PZT under load. This is the resonant angular frequency of the PZT. The standard has a full width at half maximum (FWHM) of the transmission peak. It serves as the denominator for the criterion.

[0128] (8) Optical locking criterion: in, For the standard etalon transmittance function, For optical carrier frequency, The optical upper sideband frequency, The optical lower sideband frequency, The transmittance of the upper sideband. The transmittance is the lower sideband. This is the sideband transmittance differential error signal. To lock the judgment frequency threshold.

[0129] Compared with existing technologies, this dual-PZT driven high-speed swept-frequency laser and its three-stage synchronous mode-skipping-free swept-frequency method have the following advantages:

[0130] This invention is based on a three-level synchronization architecture and a mechanism-LSTM hybrid intelligent algorithm. It suppresses errors step by step through open-loop coarse synchronization, displacement closed-loop intermediate synchronization, and optical fine closed-loop, achieving a synchronization error of <1.8 GHz and SMSR >45 dB at a 10 kHz sweep frequency, while maintaining stability even at the 30 kHz extreme sweep rate. The internal modulation frequency locking eliminates the need for external electro-optical components, and multi-field coordinated temperature control and vibration compensation provide strong environmental adaptability. Fully automatic calibration, incremental self-update, and aging compensation enable maintenance-free operation for tens of thousands of hours. It is compatible with dual-dot switching and various scenarios such as radar and OCT. Attached Figure Description

[0131] Figure 1 This is a layered overall block diagram of the entire system of the present invention;

[0132] Figure 2 A schematic diagram of a stretchable PZT structure for the optical path of an optical resonant mode selection system;

[0133] Figure 3 A schematic diagram of the PZT shear plate structure for the optical path of an optical resonant mode selection system;

[0134] Figure 4 Cross-sectional view of an integrated ring capacitive sensor for a dual-shear PZT tuned drive system;

[0135] Figure 5 This is a top view of the electrode structure of a ring-shaped differential capacitive displacement sensor.

[0136] Figure 6 Schematic diagram of the complete circuit for R-SOA internal 50MHz modulation and demodulation;

[0137] Figure 7 A block diagram of the parallel computing data flow for an FPGA intelligent three-level synchronous control system;

[0138] Figure 8 This is a three-level synchronous control timing logic diagram;

[0139] Figure 9 Comparison of adaptive segmented smooth mode-hopping reset waveforms;

[0140] Figure 10 Complete flowchart of the mechanism-LSTM hybrid modeling algorithm;

[0141] Figure 11 This is a comparison curve of the full performance indicators of the present invention with four types of existing light sources;

[0142] Figure 12 Comparison waveforms of synchronization error with and without LSTM compensation for 10kHz frequency sweep. Detailed Implementation

[0143] The following are specific embodiments of the present invention, which are described in conjunction with the accompanying drawings. However, the present invention is not limited to these embodiments.

[0144] like Figures 1-12 As shown, a high-speed sweeping laser driven by dual PZT consists of three main parts: laser hardware system, three-level synchronous mode-skipping sweeping control method, and mechanism-LSTM hybrid intelligent compensation algorithm. It includes an optical resonant mode selection system, a dual PZT tuning drive system, a differential ring capacitor displacement sensing system, an R-SOA intrinsic modulation frequency locking system, an FPGA intelligent three-level synchronous control system, and a multi-field coupling temperature control and environmental monitoring system.

[0145] The optical resonant mode selection system performs gain amplification, coarse mode selection, fine mode selection, and beam splitting detection, and includes the following units:

[0146] The R-SOA active gain unit employs an InP-based reflective semiconductor optical amplifier chip, with a gain bandwidth covering the C-band (1530~1565nm) or the extended C+L band (1525~1610nm). The R-SOA chip's front end is coated with an anti-reflection film (reflectivity <0.1%), and the rear end is coated with a high-reflectivity film (reflectivity >90%), forming a resonant cavity with the external cavity mirror. The DC bias current is 150mA, the RF modulation bandwidth is 100MHz, the modulation peak-to-peak value is 10mA, and the carrier differential gain is [not specified]. The relaxation oscillation frequency is 700MHz. The chip is mounted on an independent TEC temperature-controlled stage with a temperature stability of ±0.01℃.

[0147] The 6T-FP etalon (6-layer dielectric film structure FP etalon) coarse selection module uses a 25μm air-gap fused silica parallel substrate, and multilayer dielectric films are fabricated on both sides using ion beam assisted deposition (IBAD) process. The film system is designed as follows: Structure (H is) n≈2.1; L is (n≈1.45). The basic fineness F=60 (corresponding to reflectivity 94.9%), FSR=6THz, and the passband full width at half maximum (FWHM)=FSR / F=100GHz. The etalon is driven by the first PZT to change the air gap length, achieving 6THz coarse filtering.

[0148] The fine mode selection unit of the external cavity mirror uses a planar dielectric film mirror with a reflectivity of 80%±1% and a substrate material of fused silica, with a surface accuracy better than λ / 10. The mirror and the far end face of the R-SOA form an external cavity resonant cavity with an external cavity FSR > 10 GHz and a corresponding external cavity physical length < 15 mm. The mirror is driven by a second PZT, and fine tuning of the longitudinal mode frequency is achieved by changing the external cavity length.

[0149] The beam splitting and coupling detection unit sets a 1% beam splitter in the standard etalon transmission optical path. After passing through the focusing lens, the beam is received by a high-speed InGaAs PIN photodiode with a bandwidth of 125MHz and a responsivity of 0.9A / W@1550nm. The output optical beat frequency signal is sent to the frequency-locked demodulation circuit.

[0150] The main laser output coupling unit uses an aspherical lens to collimate the diverging laser output from the R-SOA. The lens has a numerical aperture (NA) of 0.5~0.8 and a focal length of 1.5~3.0 mm. The laser then enters a single-mode fiber via a precision fiber coupler (coupling efficiency >60%) and is output through an FC / APC standard interface with a return loss >60 dB.

[0151] The electrically controlled dual-etalon switching unit incorporates two 6T etalons: an F=60 high-speed etalon and an F=200 high-precision etalon. An electrically controlled translation stage completes the optical path switching. The hybrid modeling and computation unit pre-stores two Airy transmission mechanism parameter libraries, automatically updating the passband FWHM, loss coefficient, and synchronization error threshold after etalon switching. The F=60 etalon has a single-pass 10GHz frequency offset loss of 0.17dB and a wide synchronization tolerance, suitable for high-speed radar scenarios; the F=200 etalon has a narrower passband and stronger side-mode suppression, suitable for high-precision OCT scenarios. One device is compatible with both applications without requiring replacement of the entire unit.

[0152] A dual-axis tilt closed-loop correction unit for the reflector, equipped with a miniature shear piezoelectric trimmer, extracts the DC loss component from the 50MHz modulation error signal as the tilt error source, which is then superimposed onto the optical precision closed-loop PID output to dynamically correct the mirror's pitch / yaw tilt. This keeps the tilt error stably controlled within 0.1μrad over the long term, reducing the SMSR fluctuation from ±2dB to ±0.5dB.

[0153] The beam splitter adaptive gain adjustment unit features a transimpedance amplifier with multiple digitally controllable gain settings, and the FPGA automatically switches gain levels based on the main output power monitoring signal. Hybrid modeling adds an optical power loss correction term to compensate for the transmission loss fluctuations of the beam splitter and etalon with wavelength / temperature, thus extending the linear range of the error signal.

[0154] Synchronous optical constraint formula:

[0155]

[0156] ,

[0157] Where K is the synchronization ratio. For the standard to have a free spectral range, For the free spectral range of the external cavity, The optical length of the external cavity. The standard equipotential bonding element has an air gap thickness of 25 μm, where c is the speed of light in a vacuum and λ is the center wavelength of the laser. The displacement required for synchronous tuning of the external cavity PZT to one FSR.

[0158] The dual PZT tuning drive system is either a dual shear type PZT tuning drive system or a dual telescopic PZT tuning drive system.

[0159] The dual-shear PZT tuned drive system achieves synchronous tuning of the etalon air gap thickness and the external cavity optical length through the piezoelectric effect. Its core actuator can adopt two configurations: such as... Figure 3 The shear-type PZT (shear mode) shown utilizes the shear deformation of piezoelectric ceramics to drive the mirror or etalon to produce lateral displacement, thereby changing the optical path; alternatively, methods such as... Figure 2 The dual-telescopic PZT tuning drive system (stacked or tubular structure) shown achieves displacement output through longitudinal expansion and contraction. The two configurations differ in displacement stroke, resonant frequency, and hysteresis characteristics, but both meet the synchronous tuning requirements of this invention. The dual-PZT tuning drive system, with its higher stiffness, faster response speed, and smaller tilt error, is more suitable for high-speed frequency sweep scenarios; meanwhile, the telescopic PZT, as an alternative, can also be adapted to the control system and synchronization algorithm of this invention.

[0160] The dual-shear type PZT tuned drive system includes the following units:

[0161] The standard etalon shearing PZT actuator uses a multilayer piezoelectric ceramic shear plate operating in shear mode. It measures 5mm × 5mm × 2mm, with a piezoelectric shear coefficient d15 ≥ 800pC / N, a load resonant frequency ≥ 20kHz, a stroke ± 1μm, closed-loop resolution ≤ 0.1nm, hysteresis < 1%, and tilt error < 0.1μrad. The single FSR tuning displacement is λ / 2 ≈ 775nm (@1550nm). The PZT shear plate has gold electrodes on both sides and is equipped with a flexible hinge anti-tilt limiting structure, with a lateral offset < 0.5mrad.

[0162] The external cavity shearing PZT actuator uses a multilayer piezoelectric ceramic shear plate operating in shear mode. Its dimensions are 10mm × 10mm × 5mm, with a piezoelectric shear coefficient d15 ≥ 800pC / N, a load resonant frequency ≥ 15kHz, a stroke ± 500μm, a closed-loop resolution ≤ 0.5nm, hysteresis < 5%, and tilt error < 0.5μrad. The single FSR synchronous tuning displacement is 600 × 775nm = 465μm.

[0163] Two PZTs are driven independently with a fixed synchronization ratio of 600:1, satisfying the synchronization displacement constraint. ,

[0164] in, The displacement required for synchronous tuning of the external cavity PZT to one FSR. The displacement required to tune an FSR for a standard etalon PZT.

[0165] The high-voltage amplification and drive unit employs a high-voltage operational amplifier (OPA454) to amplify the 0~5V low-voltage control signal output from the FPGA to a 0~150V piezoelectric drive voltage, with a slew rate ≥50V / μs, output current ≥100mA, and bandwidth ≥100kHz. It incorporates built-in overvoltage, overcurrent, and short-circuit hardware protection circuits, as well as EMI shielding and filtering. An LC low-pass filter (cutoff frequency ≥100kHz) is configured at the output to suppress high-frequency noise.

[0166] The PZT status acquisition and protection unit installs a temperature thermistor (PT1000, accuracy ±0.01℃) and a vibration acceleration sensor (bandwidth 1kHz) near the PZT shear plate to acquire the PZT ceramic temperature and structural vibration amplitude in real time. When the threshold is exceeded, the high-voltage drive is cut off and the data is uploaded to the hybrid modeling unit for error compensation.

[0167] The layered composite phase feedforward unit consists of three levels of feedforward superposition: ① PZT second-order dynamics-based phase compensation; ② additional compensation term for scanning acceleration (a_scan×0.0001); ③ independent phase coefficient tables for 1kHz, 2kHz, 5kHz, 10kHz, and 30kHz. At 10kHz scanning, the PZT phase lag is reduced from 170° to less than 20°, and further compressed to less than 8° at the 30kHz extreme scanning, resulting in a 40% reduction in synchronization error.

[0168] The creep online identification and correction unit continuously acquires hourly creep displacement data under closed-loop steady-state conditions, and incrementally updates the hybrid modeling four-dimensional residual database (driving voltage-scanning frequency-temperature-displacement error) in real time. The hysteresis voltage range is fitted with a 20-segment polynomial, replacing single cubic interpolation. The creep synchronization error during 8 hours of continuous scanning was reduced from 1.5MHz to 0.3MHz.

[0169] An adaptive bandwidth current limiting protection unit is implemented, featuring a digitally controllable RC bandwidth network. The FPGA automatically switches the drive bandwidth based on the real-time scanning frequency (expanding bandwidth to improve response during high-speed operation and narrowing bandwidth to suppress noise during low-speed operation). The upper limit of the output current is dynamically adjusted based on the temperature collected by the PZT thermistor, reducing power at high temperatures to slow piezoelectric aging and suppressing high-speed resonance noise.

[0170] The dual-telescopic PZT tuning drive system realizes standard piezoelectric tuning with air gap and outer cavity length, and includes the following units:

[0171] The standard etalon telescopic PZT actuator uses a stacked or tubular telescopic piezoelectric ceramic actuator. It drives the 6T-FP standard etalon air gap thickness change through longitudinal telescopic movement to achieve coarse mode tuning in the 6THz range. The single FSR tuning displacement is λ / 2≈775nm, and the stroke meets the full range scanning requirements of the standard etalon.

[0172] The external cavity telescopic PZT actuator unit adopts a large-stroke stacked piezoelectric ceramic actuator. It drives the displacement of the external cavity mirror through longitudinal telescopic movement, changing the optical length of the external cavity to achieve fine tuning of the longitudinal mode frequency. The single FSR synchronous tuning displacement is 600×775nm=465μm, the stroke is ±500μm, and it maintains a fixed synchronization ratio of 600:1 with the standard PZT.

[0173] The high-voltage amplification drive unit amplifies the low-voltage control signal output by the FPGA to a 0~150V high-voltage drive signal. It has the characteristics of high voltage slew rate (≥50V / μs) and wide bandwidth (≥100kHz), driving dual-channel telescopic PZT actuators to respond quickly. It has built-in overvoltage, overcurrent, short circuit hardware protection and EMI filtering.

[0174] The PZT status acquisition and protection unit installs a temperature thermistor and a vibration acceleration sensor near the telescopic PZT actuator to collect the PZT ceramic temperature and structural vibration amplitude in real time. When the threshold is exceeded, the high voltage drive is cut off and the data is uploaded to the hybrid modeling unit for error compensation.

[0175] The layered composite phase feedforward unit integrates the second-order dynamics basic phase compensation of PZT, the additional compensation of scanning acceleration, and the three-level superposition of graded frequency coefficients to compensate for the phase lag of the stretchable PZT under high-speed scanning. The phase lag is reduced from 170° to less than 20° under 10kHz scanning.

[0176] The creep online identification and correction unit continuously collects hourly creep displacement under closed-loop steady state, updates the hybrid modeling four-dimensional residual database in real time incrementally, and fits piecewise polynomials in the hysteresis voltage range. The creep synchronization error is reduced from 1.5MHz to 0.3MHz after 8 hours of continuous scanning.

[0177] The adaptive bandwidth current limiting protection unit sets up a digitally controllable RC bandwidth network, and the FPGA automatically switches the drive bandwidth according to the real-time scanning frequency; it dynamically adjusts the upper limit of the output current based on the temperature collected by the PZT thermistor, reduces power at high temperatures to slow down piezoelectric aging, and suppresses high-speed resonance noise.

[0178] The differential ring capacitance displacement sensing system is integrated into two PZT shear plates, achieving coarse positioning based on self-capacitance and high-precision detection based on differential capacitance. It includes the following units:

[0179] A magnetron sputtering ring electrode sensing unit is constructed by fabricating a ring-shaped differential capacitance sensor electrode on the surface of a PZT shear sheet using magnetron sputtering. The ring electrode has an inner diameter of 1.5 mm, an outer diameter of 2.5 mm, and a thickness of 100 nm. The electrode material is platinum (Pt), and the adhesion layer is titanium (Ti). Magnetron sputtering process parameters: background vacuum... The sputtering pressure is 0.5~1.0 Pa, the sputtering power is 100~200 W, and the deposition temperature is room temperature to 150℃. The inner and outer rings form differential feedback capacitors Cin+ and Cin-. The self-capacitance channel realizes coarse positioning of PZT, and the differential capacitance detects nanometer-level displacement with a resolution ≤0.08 nm.

[0180] The capacitor signal modulation and demodulation unit uses the differential capacitance detection principle to eliminate common-mode interference. It includes a high-frequency AC excitation source (1MHz, resolution better than 4aF, corresponding displacement resolution better than 0.1nm), an instrument differential amplifier (common-mode rejection ratio >80dB), an active bandpass filter (cutoff frequency 100kHz), and a 16-bit ADC sampling circuit with a noise floor <0.01nm. It converts minute differential capacitance changes into a standard 0~3.3V analog voltage signal.

[0181] The displacement synchronization transmission unit adopts FIFO high-speed cache and LVDS differential transmission chip to synchronously transmit dual-channel displacement data (standard etalon PZT and external cavity PZT) to the FPGA displacement closed-loop calculation unit without transmission delay.

[0182] The electrode temperature drift common-mode decoupling unit incorporates a hybrid modeling approach, adding an electrode capacitance temperature coupling mechanism equation to separate two types of errors: PZT thermal expansion and electrode dielectric constant temperature drift. The demodulation circuit also incorporates common-mode negative feedback amplification. The zero-point drift due to ±10℃ temperature changes is reduced from 0.8nm to 0.15nm.

[0183] The dual-channel bit synchronization timing alignment unit adds a hardware synchronization latch signal to the LVDS transmission, which is from the same source as the FPGA's 100kHz master control clock; the hybrid modeling adds a sampling time difference compensation term to eliminate the dynamic synchronization error caused by the microsecond-level sampling offset of the two ADCs.

[0184] The adaptive vibration filtering unit analyzes vibration amplitude in real time by inputting data from PZT vibration sensors and dynamically adjusts the filter window length—lengthening the filter window during intense vibrations (enhancing noise suppression) and shortening it during static measurements (improving response speed). The frequency error caused by environmental vibrations in the 1~10kHz range is reduced from 1.3MHz to 0.2MHz.

[0185] The R-SOA internal modulation frequency locking system generates a double-sideband error signal based on R-SOA current modulation, replacing the traditional PDH external electro-optic modulation scheme. It includes the following units:

[0186] The 50MHz DDS generator unit uses the AD9833 DDS chip and a low-temperature drift crystal oscillator. It outputs a 50MHz sine wave through a phase-locked loop and a low-noise operational amplifier AD8065, with a frequency accuracy of 0.1Hz and a phase noise of <-120dBc / Hz@10kHz. It is synchronous with the demodulation local oscillator.

[0187] The R-SOA bias superposition unit consists of a 150mA constant current source, a broadband current mirror, a DC blocking high-voltage capacitor, and an RF choke. It uses a broadband bias three-way (Bias-T) circuit to superimpose a 50MHz sinusoidal modulation signal onto the R-SOA DC bias. The operating frequency range is 10kHz~100MHz, with insertion loss <0.5dB and 50Ω impedance matching (VSWR <1.2:1). The peak-to-peak modulation current is 10mA (dynamically adjustable range 5~15mA), and the bandwidth is 100MHz. Modulation expression: ,in, The DC bias current is set to 150mA. The peak-to-peak value of the modulation current / modulation amplitude is set to 5~15mA (adaptive adjustment). The frequency is 50MHz, the modulation depth is 0.1~0.5rad, and t is time. The instantaneous phase of the modulated signal (unit: rad). The modulation angular frequency (unit: rad / s).

[0188] The photoelectric transimpedance amplifier unit uses a Thorlabs PDA10CF high-speed PIN photodiode in conjunction with an AD8015 transimpedance amplifier, achieving a transimpedance gain of [missing information]. V / A, bandwidth 125MHz, noise <1nV / √Hz. The output signal is fed into the mixer after being filtered by a 10MHz high-pass filter to remove low-frequency noise.

[0189] The dual-balanced mixer demodulation unit uses a Mini-Circuits ADE-1 mixer with 50Ω impedance matching, 0dBm local oscillator, and 6dB conversion loss. The photoelectric signal is mixed with the 50MHz local oscillator signal and then passed through a fourth-order Butterworth low-pass filter (with a dynamically switching cutoff frequency of 3kHz or 5kHz) to output a low-frequency error signal. The error signal formula is: ,in, Frequency detuning / frequency deviation The standard etalon transmission function (Airy function). For optical carrier frequency, For radio frequency modulation frequency, For the upper sideband frequency, This refers to the lower sideband frequency;

[0190] When Δv=0, the error signal returns to zero, corresponding to the complete alignment of the transmission peak of the external cavity longitudinal mode with the standard etalon; when there is a deviation, the amplitude and polarity of the error signal reflect the magnitude and direction of the deviation.

[0191] The error signal conditioning unit uses the zero-drift operational amplifier AD8628, equipped with a DC offset adjustment potentiometer and a limiting protection circuit, to amplify the demodulated error signal to a standard level of 0~±5V and send it to the FPGA optical precision closed-loop operation unit.

[0192] The modulation amplitude adaptive unit dynamically adjusts the modulation current in the range of 5~15mA. The FPGA adjusts the DDS output amplitude in a closed loop according to the output amplitude of the photodetector to maintain the modulation depth stable at the optimal operating point of 0.3rad, and the signal-to-noise ratio of the error signal is uniform and stable across the entire band.

[0193] The demodulation zero-point self-calibration unit automatically shuts off 50MHz modulation during the lockout gap, stores the inherent DC offset of the acquisition and demodulation circuit in the compensation register, and superimposes real-time offset corrections onto the optical PID input. The long-term zero-point drift is reduced from 0.08MHz to 0.01MHz.

[0194] The filter bandwidth dynamic matching unit enables high-precision switching between 1~2kHz operating conditions and 5kHz low-pass bandwidth to improve lock-in response; it maintains 3kHz low-pass to suppress harmonics in 10~30kHz high-speed operating conditions; the high-pass filter adaptively adjusts the cutoff frequency according to optical power to reduce weak light signal loss.

[0195] The FPGA intelligent three-level synchronous control system is the main computing unit, using a Xilinx Zynq / Artix7 FPGA as the main control chip and a 100kHz unified sampling clock. It includes the following units:

[0196] The mechanism-LSTM hybrid modeling feedforward unit incorporates a PZT second-order underdamped dynamics model, an FP etalon transmission Airy model, a laser rate equation, and a 600:1 synchronous displacement constraint, among other underlying physical mechanism layers. It is complemented by a four-dimensional measured residual database (driving voltage, scanning frequency, temperature, and displacement error) correction layer, which calculates the compensation voltage in real time through three-dimensional cubic interpolation. The LSTM neural network uses 6-dimensional features (driving voltage, actual displacement, temperature, voltage derivative, displacement derivative, and scanning frequency) from a 50-step historical window as input to predict the amount of physical model error compensation. Hybrid model formula:

[0197] ,

[0198] ,

[0199] in, To ultimately predict the displacement, To predict displacement based on physical mechanisms, The amount of compensation for LSTM time series prediction; This is the final PZT drive voltage. For physical mechanism theory voltage, For the target displacement of the standard etalon, This is the final PZT drive voltage. The current temperature. For the rate of temperature change, For scanning acceleration, Correcting voltages for the four-dimensional residual database. For vibration compensation voltage, This is the tilt compensation voltage.

[0200] The first-level open-loop coarse synchronization unit relies solely on hybrid modeling feedforward voltage to drive the dual PZTs, with no displacement feedback and a synchronization error of 50~100nm. It is used for rapid debugging of prototypes and provides an initial operating point for subsequent closed-loop circuits.

[0201] The unit employs a two-stage dual-channel displacement PID controller with two independent PID cores, a bandwidth of 10kHz, and an anti-integral saturation module. Basic PID parameter values ​​are: Kp=2.5, Ki=0.8, Kd=0.1 (the adaptive PID unit dynamically optimizes based on operating conditions). It inputs the measured displacement from the dual-channel differential capacitance sensing unit and outputs a corrected external cavity PZT drive voltage, compressing the synchronization error to ±1nm (corresponding to a frequency error ≤±600kHz).

[0202] A three-level optical precision PID unit with a low-speed PID core and a bandwidth of 1~5kHz (dynamically switching). Basic PID parameter values: Kp_opt=0.5, Ki_opt=0.05, Kd_opt=0.01 (adaptive adjustment). Lock-in threshold |e(Δv)|<0.1MHz. Receives and demodulates optical error signals, superimposing them onto the displacement closed-loop output to achieve MHz-level frequency precision alignment, SMSR stability >45dB, and automatic compensation for temperature drift, PZT aging, and optical component deformation.

[0203] The timing unit is mode-hopping-free, featuring a high-speed state machine equipped with a 100μs high-precision timer and a phase-locked loop (PLL) capture scan generator. Triggering occurs when the etalon PZT displacement reaches +775nm: etalon PZT returns to zero → external cavity PZT jumps to -465μm → 100Hz small-amplitude scan capture → re-locking. Timing specifications: PZT step < 100μs, loop capture < 1ms, total mode-hopping time < 2ms, power fluctuation ≤ 0.3dB.

[0204] The recapture unit for lost-lock operation includes an error amplitude threshold comparator and a 3-recapture count register. Upon loss of lock, it immediately switches to a secondary displacement closed loop, with the external cavity PZT scanning at 200Hz and ±5~20nm (adaptive amplitude) to capture the zero point; upon successful capture, the tertiary closed loop is restored; after three consecutive failed captures, a full-range reset is performed and the residual database is updated.

[0205] The DAC output unit consists of two 16-bit AD5761 DACs with a 1MHz update rate, a synchronization timing synchronizer, and digital limiting protection. It converts the final control voltage of the three-stage synchronization into an analog signal and outputs it to the high-voltage PZT amplifier unit.

[0206] The four-dimensional database incremental self-updating unit automatically collects four-dimensional data of voltage, scanning frequency, temperature, and displacement error during mode skipping reset and lockout recapture intervals. It uses local cubic interpolation to incrementally update the database without overwriting the factory calibration data. The entire unit can operate for 8000 hours without needing to be returned to the factory for recalibration, and the synchronization accuracy decay is <0.05MHz.

[0207] The three-level adaptive PID self-tuning unit performs a 100Hz, ±10nm small-amplitude self-tuning scan after each power-on and mode jump reset. It automatically selects three sets of PID parameters based on the current scan frequency, temperature, and etalon precision: high precision (1~2kHz), standard (5~10kHz), and extreme high speed (20~30kHz). This reduces the probability of closed-loop overshoot and lockout by 60%.

[0208] The optical field dynamics hard-constraint unit embeds a 443 ps photon lifetime and a 700 MHz R-SOA relaxation oscillation frequency as hard constraints into the mechanism layer. When the scan rate approaches the relaxation oscillation limit, it automatically reduces the scan acceleration and increases the feedforward compensation weight. The physical basis for this constraint is that the shortest optical field setup time is approximately 2 ns (determined by the photon lifetime), and the relaxation oscillation frequency corresponds to the maximum response rate of the gain medium. At the 30 kHz limit scan, the power attenuation is reduced from 15% to 7%, and the dynamic linewidth is compressed by 30%.

[0209] The hierarchical fault diagnosis unit can separate four types of error sources in real time: PZT phase lag, temperature drift, environmental vibration, and optical tilt. It outputs standardized fault codes (such as E001 - phase lag exceeds the limit, E002 - temperature drift exceeds the limit, etc.) to accurately locate the corresponding secondary functional unit, improving fault diagnosis efficiency by 80%.

[0210] The automatic operating condition identification unit reads the sweep frequency command from the host computer and automatically identifies the sweep frequency (1~30kHz). It allows one-click switching of feedforward coefficient, closed-loop bandwidth, and etalon mechanism parameters (F=60 or F=200). It also features built-in hard protection for sweep rate, PZT acceleration, and optical error threshold, with automatic speed reduction when exceeding the boundary to avoid device damage and SMSR drop.

[0211] The fully automatic calibration unit enables one-button start-up for full-dimensional automated calibration. It automatically traverses voltage (0~10V segments), temperature (-10℃~60℃), and scanning frequency (1kHz, 2kHz, 5kHz, 10kHz, 30kHz) segments for testing, collecting displacement, optical error, and power data. It automatically generates and stores a four-dimensional residual database and simultaneously outputs a factory performance grading report. The calibration time for a single machine is reduced from 4 hours to 40 minutes.

[0212] LSTM neural network architecture:

[0213] Input layer: History window length T=50, feature dimension 6

[0214] LSTM layer 1: Hidden units 128, returning sequence

[0215] LSTM layer 2: Hidden unit 64, returning sequence

[0216] Dropout layer: Drop rate 0.2

[0217] Fully connected layer: 32 units, ReLU activated

[0218] Output layer: 1 unit (displacement compensation)

[0219] A multi-field coupled temperature control and environmental monitoring system suppresses temperature drift errors and includes the following components:

[0220] The standard etalon temperature control unit features a miniature TEC cooling chip with a high-precision NTC thermistor (accuracy ±0.01℃) and an aluminum thermally conductive base, achieving a temperature control accuracy of ±0.01℃ and real-time acquisition of the standard etalon air gap substrate temperature.

[0221] The R-SOA temperature control unit features a TEC chip platform with thermal resistance pads and a heat dissipation metal housing, providing a temperature control accuracy of ±0.01℃, stabilizing the gain of the chip's bandgap wavelength, and suppressing carrier refractive index temperature drift.

[0222] External endoscope temperature control unit, small TEC with heat insulation bracket and temperature acquisition circuit, temperature control accuracy ±0.01℃.

[0223] Temperature refractive index correction formula: ,in, ΔT represents the refractive index at temperature T, 0.000273 represents the thermo-optic coefficient of fused silica, and ΔT represents the temperature change. The three temperature data are uploaded to the hybrid modeling and computing unit every 100ms, and the synchronization ratio is corrected in real time.

[0224] A three-channel temperature control cross-decoupling unit is used to establish a three-channel temperature coupling mechanism matrix, and hybrid modeling is used to decouple heat conduction cross-interference in real time. When a single channel experiences a temperature change, the synchronization ratio offset of the other two channels is pre-compensated. The synchronization error for a 1℃ step temperature change is reduced from 1.5MHz to 0.2MHz.

[0225] A temperature change rate feedforward pre-compensation unit is added, and the temperature change rate dT / dt is acquired by thermistor differentiation. A dynamic temperature drift compensation term is added to the hybrid model. The stabilization time for sudden ambient temperature changes is reduced from 30s to 3s.

[0226] The dynamic linewidth monitoring unit uses delayed self-heterodyne interferometry to analyze the dynamic linewidth in real time through short-time Fourier transform (STFT). When the linewidth expansion exceeds the threshold, it automatically adjusts the R-SOA bias current or adjusts the PZT drive waveform to achieve active compression.

[0227] The frequency modulation linearity monitoring and predistortion unit uses wavelength monitoring data to calculate the deviation between the actual frequency-time curve and the ideal linear curve, and uses an iterative predistortion algorithm to shape the swept frequency waveform, reducing the frequency sweep nonlinearity by more than an order of magnitude.

[0228] The PZT aging condition monitoring and life prediction unit periodically records the PZT's displacement-voltage curve, resonant frequency, and capacitance sensor baseline, extracts aging characteristics, predicts the remaining life through an LSTM model, and adaptively adjusts the controller parameters to extend the effective service life.

[0229] Globally embedded long-term reliability optimization logic:

[0230] (1) Vibration active collaborative compensation logic: PZT vibration sensor data is connected to the hybrid modeling and computing unit in real time to calculate vibration displacement disturbance, and superimposed reverse compensation voltage to offset 1~10kHz environmental vibration.

[0231] (2) PZT aging long-term slow compensation logic: The whole machine automatically performs micro static calibration every 10 hours of continuous operation, updates the piezoelectric coefficient theoretical model, compensates for long-term piezoelectric coefficient drift, and extends the stable service life to more than 10,000 hours.

[0232] This three-level synchronous mode-skipping frequency sweep control method adopts a three-level hierarchical error suppression architecture of "open-loop coarse synchronization + displacement closed-loop intermediate synchronization + optical error fine closed-loop". The overall process is divided into five execution stages: fully automatic mechanism-data hybrid modeling offline calibration stage, three-level hierarchical adaptive synchronous real-time control stage, adaptive smooth mode-skipping reset continuous frequency sweep stage, hierarchical diagnostic loss-of-lock detection and recapture stage, and background long-term adaptive compensation stage.

[0233] Phase 1: Fully Automated Mechanism-Data Hybrid Modeling and Offline Calibration

[0234] Performed by the FPGA fully automatic calibration unit:

[0235] S1.1. The multi-zone temperature control unit automatically traverses the temperature range of -10℃ to 60℃ in segments;

[0236] S1.2. The FPGA open-loop unit automatically outputs 0~10V drive voltage in segments, and the displacement sensing unit synchronously collects nanometer-level displacement data;

[0237] S1.3. Automatically traverse multiple scanning frequencies of 1kHz, 2kHz, 5kHz, 10kHz, and 30kHz to collect the original errors of phase hysteresis, delay, and creep.

[0238] S1.4. Automatically construct a four-dimensional residual database (driving voltage-scanning frequency-temperature-displacement error), complete the solidification of the cubic interpolation function, and simultaneously fit the second-order dynamics of PZT, the transmission Airy function of the FP etalon, the laser rate equation, and the complete set of mechanism parameters of the optical field dynamics.

[0239] S1.5. Train the LSTM timing compensation network and burn the network weights and database to the FPGA on-chip Flash;

[0240] S1.6. Automatically complete equipment classification (high precision / standard / extreme high speed three levels), store the corresponding operating condition parameter library, and output a complete factory calibration report.

[0241] Phase Two: Three-Level Hierarchical Adaptive Synchronous Real-Time Control

[0242] S2.1. Open-loop coarse synchronization. After the host computer issues a frequency sweep command, the automatic operating condition identification unit determines the operating range (1~2kHz / 5~10kHz / 20~30kHz) and retrieves the corresponding layered composite phase feedforward coefficients and etalon fineness mechanism parameters (F=60 or F=200). The hybrid modeling and computing unit superimposes multi-dimensional compensation quantities such as creep, temperature change rate, vibration, and optical power loss, and outputs a 600:1 ratio feedforward drive voltage to coarsely align the etalon transmission peak with the external cavity longitudinal mode. The open-loop synchronization error is 50~100nm, providing an initial operating point for subsequent closed-loop operation.

[0243] S2.2. Synchronization in Displacement Closed-Loop. After common-mode temperature drift decoupling and adaptive vibration filtering preprocessing, the differential ring capacitive sensing unit synchronously latches dual-channel displacement data and sends it to the dual-channel adaptive PID controller. The PID self-tuning unit matches the optimal control parameters (three independent parameter libraries) in real time according to the current operating conditions, compares the target displacement output by the hybrid model, and the dual-channel 10kHz bandwidth PID controller corrects the drive voltage in real time, eliminating static hysteresis, creep, and sampling time difference errors. The synchronization error is compressed to ±1nm (corresponding to a frequency error ≤ ±600kHz).

[0244] S2.3. Precise Closed-Loop Optical Error Control. A 50MHz DDS unit adaptively adjusts the modulation current (5~15mA range) to maintain an optimal modulation depth of 0.3rad; the filter bandwidth automatically switches according to operating conditions (5kHz for high-precision operation, 3kHz for high-speed operation); the demodulation circuit superimposes the zero-point self-calibration offset in real time; the tilt correction unit finely adjusts the dual-axis piezoelectricity of the reflector based on loss errors. The error signal, after mixing and demodulation, is input to the PID controller of the external cavity PZT (bandwidth dynamically switching from 1~5kHz) for fine adjustment, with the alignment accuracy of the longitudinal mode and the etalon passband as the final control target. Synchronization accuracy ≤ ±0.1MHz, SMSR stability > 45dB, automatic compensation for temperature drift, PZT aging, optical element deformation, and beam tilt.

[0245] The synchronously operating optical field dynamics constraint unit automatically reduces the scanning rate and increases the feedforward compensation weight when the scanning rate approaches the relaxation oscillation limit, thus suppressing power attenuation and dynamic linewidth broadening. The layered error diagnosis unit outputs the amplitude values ​​of various error sources and fault codes in real time.

[0246] Phase 3: Adaptive Smooth Mode-Jump-Free Reset Continuous Frequency Sweep

[0247] S3.1. The displacement sensing unit monitors the displacement of the etalon PZT in real time, and triggers the timing unit when the displacement reaches +775nm (completes one FSR tuning);

[0248] S3.2. The timing unit performs a three-stage smooth step transition: the standard etalon PZT returns to zero in three steps (reducing to 30% within 30μs, then to 10% within another 30μs, and finally to zero within the last 40μs), and the external cavity PZT synchronously transitions to -465μm in three steps (transitioning to 30% within 30μs, then to 70% within another 30μs, and finally to 100% within the last 40μs), with a total step response time <100μs;

[0249] S3.3. Adaptively configure the capture scan amplitude according to the current synchronization error amplitude (for errors > 5 GHz, use ±20 nm large-amplitude scan for fast capture; for errors < 1 GHz, use ±5 nm small-amplitude scan for fine locking), and perform a 100 Hz small-amplitude scan to search for the optical error zero point;

[0250] S3.4. Upon detecting zero error, it automatically switches to a three-stage optical precision closed loop, with a loop acquisition time of <1ms, a total mode hopping time of <2ms, and output optical power fluctuation ≤0.3dB. During mode hopping intervals, it synchronously performs incremental self-updates of the residual database to compensate for short-term creep and temperature drift. Cyclic execution achieves continuous frequency sweeping with no mode hopping throughout the C-band 4.4THz range.

[0251] Phase Four: Hierarchical Diagnostic Loss-Lock Detection and Recapture

[0252] S4.1. The optical error conditioning unit outputs the error amplitude in real time. After the unlocking logic unit determines that the lock has been lost, it immediately switches to the second-level displacement closed loop.

[0253] S4.2. Adaptive capture scanning amplitude performs 200Hz scanning to search for the optical error zero point (expanding the scanning range when the error amplitude is large and narrowing it when the amplitude is small), and recovers the three-level closed loop upon successful capture;

[0254] S4.3. If three consecutive capture failures occur, a full-range reset is performed, and the temperature, vibration, and scanning frequency data of this loss of lock are recorded simultaneously, and the hybrid modeling residual database is incrementally updated.

[0255] S4.4. The hierarchical error diagnosis unit outputs standardized fault codes (such as E001, E002, etc.) and uploads them to the host computer to complete fault tracing.

[0256] Phase 5: Long-term adaptive compensation in the background (continuous parallel operation of the entire machine)

[0257] The machine automatically initiates a micro static calibration every 10 hours of continuous operation to update the PZT piezoelectric coefficient and compensate for device aging drift.

[0258] Vibration sensor data is input to the hybrid modeling unit in real time, and a reverse compensation voltage is superimposed to offset environmental vibrations of 1~10kHz.

[0259] Temperature control coupling decoupling and temperature change rate pre-compensation operate continuously in real time to dynamically offset the synchronization offset caused by changes in ambient temperature.

[0260] Mechanism-LSTM Hybrid Modeling Complete Algorithm

[0261] Layered Mathematical Mechanism Formula

[0262] (1) The underlying mechanism of second-order dynamics of PZT: in, Let be the transfer function (in the Laplace domain), and s be the Laplace operator. The undamped natural angular frequency, The damping ratio is 0.05~0.15, and fr is the resonant frequency of the PZT under load.

[0263] Phase lag: Where f is the driving frequency, arctan(·) is the arctangent function, and fn is the resonant frequency. When the driving frequency f is much lower than the resonant frequency fn, the PZT response is flat (no attenuation, no phase lag). When f approaches fn, the response amplitude increases sharply (resonance), and the phase lag increases rapidly to 90°. When f exceeds fn, the phase lag approaches 180°, and the response amplitude decays rapidly.

[0264] (2) Theoretical synchronous displacement constraints: ,in, t represents the theoretical displacement of the external cavity at time t, and 600 is the synchronization ratio. Let t be the air gap displacement of the etalon at time t.

[0265] (3) Improved Preisach hysteresis model: Where x(t) is the PZT displacement, It is a hysteretic nonlinear term. Let be the linear piezoelectric term, N be the number of hysteresis operators, u(t) be the driving voltage, γ be the hysteresis kernel function, and μi be the weighting coefficient of the i-th hysteresis operator. The rising edge flip threshold for the i-th hysteresis operator. η is the rising edge toggling threshold of the i-th hysteresis operator, and η is the linear piezoelectric coefficient.

[0266] (4) Multi-source coupling error decomposition model:

[0267] in, This is the final PZT drive voltage. The hysteresis error component is represented by V, where V is the driving voltage. Let t be the creep error component, and t be time. This is the phase lag error component. For the sweep frequency, For scanning acceleration, This is the temperature drift error component. For temperature deviation, For the rate of temperature change, For vibration error components, This is the optical loss error component. To output optical power.

[0268] (5) Mechanism-data hybrid correction output model:

[0269] ,in, This is the final PZT drive voltage. For physical mechanism theory voltage, For the target displacement of the standard etalon, For the sweep frequency, The current temperature. For the rate of temperature change, For scanning acceleration, Correcting voltages for the four-dimensional residual database. For vibration compensation voltage, This is the tilt compensation voltage.

[0270] (6) Feedforward phase lead compensation: in, FF stands for Feed-Forward, and t is time. The original driving signal (uncompensated) is given by τ, which is the phase compensation time constant. It is the first derivative (rate of change) of the driving signal. fn is the second derivative of the driving signal (changing acceleration), and fn is the natural frequency of PZT (load resonant frequency). These are the coefficients of the second-order Taylor expansion.

[0271] (7) Hard constraint criterion for optical field dynamics: in, The frequency scan rate (left side of the scan rate limit criterion) is the scanning rate. For frequency infinitesimal elements, 1 represents the time infinitesimal element, and 2 represents the constant coefficient (numerator). This is the resonant frequency of the PZT under load. This is the resonant angular frequency of the PZT. The standard is the full width at half maximum (FWHM) of the transmission peak. The denominator (product term) is used as the criterion.

[0272] (8) Optical locking criterion: in, It is the standard etalon transmittance function (Airy function). For optical carrier frequency, The optical upper sideband frequency, The optical lower sideband frequency, The transmittance of the upper sideband. The transmittance is the lower sideband. This is the sideband transmittance differential error signal. To lock the frequency threshold (locking precision).

[0273] Example 1: Standard industrial-grade 10kHz FMCW lidar light source

[0274] Hardware configuration: Optical path switched to F=60 standard etalon unit; ChipMing P11.Z500 shearing PZT execution unit; OPA454 high voltage amplification unit; magnetron sputtering ring electrode capacitive sensing unit (resolution 0.08nm); Zynq FPGA main controller; 10kHz displacement closed loop, 3kHz optical closed loop; all adaptive compensation units enabled.

[0275] Calibration process: A one-click fully automated calibration pipeline completes the solidification of the four-dimensional residual database and the training and deployment of the LSTM network in 40 minutes.

[0276] Actual performance: Maximum scan 10kHz, scan rate 44THz / ms, steady-state synchronization error <1.8GHz, long-term 8-hour creep + temperature drift total error <0.3MHz, SMSR stability 41~43dB, mode jump reset 1.2ms, power fluctuation 0.3dB, SMSR fluctuation under vibration disturbance ±0.4dB.

[0277] Example 2: High-precision research-grade 2kHz OCT light source

[0278] Hardware configuration: Electronic control switched to F=200 high-precision etalon unit; PZT replaced with PI P-843.10 high-precision low-hysteresis actuator (hysteresis <1%, tilt error <0.1μrad); optical closed-loop bandwidth adaptively switched to 5kHz; PID self-tuning loaded with high-precision parameter group; LSTM database encrypted calibration.

[0279] Actual performance: 2kHz scan, synchronization error <0.5GHz, SMSR stability >44.5dB, dynamic linewidth <100kHz, temperature change lock-up recovery time <3s.

[0280] Example 3: Extremely High-Speed ​​30kHz Spectral Analysis Light Source

[0281] Hardware configuration: Customized 15kHz high-resonance PZT execution unit, multiple PZTs in parallel to improve stiffness; R-SOA replaced with a high differential gain chip (relaxation oscillation frequency increased to 1GHz); hybrid model optical field dynamic constraint automatic speed limiting protection; layered composite feedforward to compress phase lag to within 8°; adaptive filtering bandwidth matched to high-speed operating conditions; vibration active compensation fully enabled.

[0282] Actual performance: Maximum scan frequency 30kHz, scan speed 132THz / ms, synchronization error <8GHz, SMSR ≥35dB, power attenuation of only 7% at 30kHz scan, dynamic linewidth compression 30%, mode skipping time 1.8ms.

[0283] Comparative Example 1: Traditional Dual PZT Open-Loop Solution

[0284] The modular architecture is without hierarchical units, differential capacitive sensing closed loop, optical precision closed loop unit, and mechanism-LSTM hybrid modeling and computing unit; it only has a single-stage open-loop proportional drive, with a maximum scan of 3kHz, a synchronization error of 60~90nm (corresponding to tens of GHz frequency error), SMSR < 38dB, and complete loss of lock at 10kHz scan.

[0285] Comparative Example 2: First-generation unoptimized three-level synchronization scheme (no adaptive unit)

[0286] Synchronization error 1.8GHz, 8-hour temperature drift creep error 1.5MHz, mode skipping power fluctuation 0.6~1dB, 10kHz scan SMSR fluctuation ±2dB, calibration time 4 hours, no fault diagnosis capability, and incompatible with multiple scenarios.

[0287] The above comparative examples demonstrate that the present invention achieves its effect through the synergistic effect of modular hierarchical hardware architecture, three-level closed-loop error suppression, mechanism-LSTM hybrid intelligent algorithm, and end-to-end adaptive optimization unit.

[0288] In summary, this invention employs a three-tiered, progressive error suppression architecture—"open-loop coarse synchronization—displacement-closed-loop intermediate synchronization—optical error fine-loop"—replacing the traditional single-stage open-loop drive scheme. The first-stage open-loop, based on a mechanism-LSTM hybrid model, outputs a 600:1 ratio feedforward voltage to achieve MHz-level coarse alignment. The second-stage displacement-closed-loop compresses the synchronization error to ±1nm through differential capacitance sensing. The third-stage optical fine-loop, relying on R-SOA intrinsic modulation frequency-locking technology, improves the frequency alignment accuracy to ≤±0.1MHz. This three-stage synergy results in a synchronization error <1.8GHz and SMSR stability >45dB under standard 10kHz operating conditions, while maintaining a side-mode rejection ratio ≥35dB under extreme 30kHz operating conditions, thus solving the technical challenge of mode hopping and loss of lock-up in high-speed swept-frequency lasers over a wide bandwidth.

[0289] This invention deeply integrates four physical mechanisms—PZT second-order dynamics, FP etalon Airy transmission function, laser rate equation, and 600:1 synchronous displacement constraint—with an LSTM temporal neural network. The physical mechanism layer provides rigid constraint boundaries to prevent error explosion during high-frequency extrapolation. The LSTM network predicts the physical model residuals using six-dimensional features (voltage, displacement, temperature, voltage derivative, displacement derivative, and scanning frequency) from the past 50 steps, compensating for nonlinear errors such as hysteresis, creep, and temperature drift in real time. At 10kHz scanning, the PZT phase hysteresis is reduced from 170° to less than 20°, and at the 30kHz extreme scanning, it is compressed to less than 8°, reducing the synchronization error by more than 40%. This achieves a wide-condition adaptive accuracy that is unattainable by traditional pure physical models or pure data-driven methods.

[0290] This invention utilizes the current modulation characteristics of the R-SOA gain chip to directly generate a 50MHz double-sideband modulation signal for optical frequency locking, replacing the expensive external electro-optic modulator, isolator, and complex optical path in traditional PDH solutions. The modulation amplitude adaptive closed-loop control dynamically adjusts within the 5~15mA range, maintaining an optimal modulation depth of 0.3rad; the demodulation zero-point online self-calibration unit compensates for inherent circuit offsets in real time, with long-term zero-point drift ≤0.01MHz; the filter bandwidth automatically switches between high-precision (5kHz) and high-speed (3kHz) operating conditions. This solution simplifies the system structure, reduces costs, and improves long-term operational stability, allowing for 8000 hours of continuous operation without requiring factory recalibration.

[0291] This invention constructs a multi-field coupled collaborative control network encompassing temperature, vibration, and optical tilt. The three independent TEC temperature control channels achieve an accuracy of ±0.01℃. Through a cross-coupling decoupling matrix, pre-compensation is provided for the other two channels in the event of a single channel temperature change. The synchronization error for a 1℃ step temperature change is reduced from 1.5MHz to 0.2MHz, and the temperature change lock-in recovery time is shortened from 30s to 3s. The vibration active compensation logic provides real-time inverse counteraction to environmental vibrations from 1 to 10kHz, reducing the frequency error from 1.3MHz to 0.2MHz. The dual-axis tilt closed-loop correction of the reflector stably controls the tilt error within 0.1μrad, narrowing the SMSR fluctuation from ±2dB to ±0.5dB. The synergy of these three elements makes it possible to move the system from the laboratory to the industrial field.

[0292] This invention deeply integrates intelligent operation and maintenance functions such as a fully automated calibration pipeline, incremental self-updating of a four-dimensional residual database, hierarchical fault diagnosis coding, PZT aging life prediction, and micro self-calibration. One-click calibration reduces the calibration time for a single machine from 4 hours to 40 minutes; incremental self-updating ensures that the synchronization accuracy decay is <0.05MHz during 8000 hours of continuous operation, eliminating the need for recalibration at the factory; hierarchical fault coding separates four types of error sources—phase lag, temperature drift, vibration, and tilt—in real time, accurately locating them to the corresponding functional units, improving fault diagnosis efficiency by 80%; micro static calibration every 10 hours compensates for PZT aging drift, extending the stable working life of the entire machine to over 10,000 hours and reducing the total life cycle cost.

[0293] This invention achieves comprehensive adaptation of a single hardware platform to diverse application scenarios through an electronically controlled dual-etalon switching unit (F=60 high-speed mode and F=200 high-precision mode), a three-level PID parameter self-tuning library (three independent sets of parameters: high precision / standard / extreme high speed), layered phase feedforward grading coefficients (1kHz, 2kHz, 5kHz, 10kHz, and 30kHz independent configurations), and optical field dynamics hard constraint speed limiting protection. The standard industrial-grade 10kHz FMCW radar light source has a synchronization error of <1.8GHz and an SMSR ≥40dB; the high-precision 2kHz OCT light source has a synchronization error of <0.5GHz, a dynamic linewidth of <100kHz, and an SMSR >45dB; and the extreme 30kHz spectral analysis scan rate reaches up to 132THz / ms with a power attenuation of only 7%. This multi-functional design reduces user equipment investment costs.

[0294] This invention achieves fully integrated, end-to-end hardware functionality encompassing optical resonant mode selection, dual-shear PZT execution, differential ring capacitor displacement sensing, R-SOA intrinsic modulation frequency locking, FPGA intelligent control, and multi-field coupled temperature control. The magnetron sputtering ring electrode displacement resolution is ≤0.08nm, the PZT travel reaches ±500μm while meeting a fixed synchronization ratio of 600:1, the high-voltage amplification unit has a slew rate ≥50V / μs and an output of 0~150V, and a 24-bit Σ-Δ ADC and a 16-bit 1MHz DAC ensure control accuracy. Through a three-stage closed-loop system and a hybrid intelligent algorithm, the overall system achieves a frequency synchronization error <1.8GHz at a 10kHz sweep rate, and an SMSR ≥35dB at a 30kHz extreme sweep rate. The dynamic linewidth is reduced by 30% compared to traditional solutions, the mode-hopping time is <2ms, and the power fluctuation is ≤0.3dB.

[0295] The specific embodiments described herein are merely illustrative of the spirit of the invention. Those skilled in the art to which this invention pertains may make various modifications or additions to the described specific embodiments or use similar methods to substitute them, without departing from the spirit of the invention or exceeding the scope defined by the appended claims.

Claims

1. A high-speed swept-frequency laser driven by dual PZTs, comprising an optical resonant mode selection system, a dual PZT tuning drive system, a differential ring capacitor displacement sensing system, an R-SOA intrinsic modulation frequency locking system, an FPGA intelligent three-level synchronous control system, and a multi-field coupling temperature control and environmental monitoring system, characterized in that, The optical resonant mode selection system includes an R-SOA active gain unit, a 6T-FP etalon coarse mode selection unit, an external cavity mirror fine mode selection unit, a beam splitting coupling detection unit, and a main laser output coupling unit, along with an electronically controlled dual etalon switching unit, a mirror dual-axis tilt closed-loop correction unit, and a beam splitting detection adaptive gain adjustment unit. The R-SOA is biased at 150mA and superimposed with 50MHz RF modulation, with a modulation peak-to-peak value of 10mA. The etalon FSR is 6THz, the air gap thickness is 25μm, and it supports adaptive switching between F=60 high-speed mode and F=200 high-precision mode. The external cavity FSR is >10GHz, and the dual PZT synchronization ratio is fixed. Where K is the synchronization ratio. For the standard to have a free spectral range, The external cavity has a free spectral range; the dual PZT tuning drive system is either a dual shear-type PZT tuning drive system or a dual telescoping type PZT tuning drive system. The dual shear-type PZT tuning drive system includes a standard etalon shear PZT execution unit, an external cavity shear PZT execution unit, a high-voltage amplification drive unit, and a PZT status acquisition and protection unit, and is equipped with a layered composite phase feedforward unit, a creep online identification and correction unit, and an adaptive bandwidth current limiting protection unit; the differential ring capacitor displacement sensing system includes a magnetron sputtering ring electrode sensing unit, a capacitor signal modulation and demodulation unit, and a displacement synchronization transmission unit, and is equipped with an electrode temperature drift common-mode decoupling unit, a dual-path position synchronization timing alignment unit, and an adaptive vibration filtering unit; the R-SOA internal modulation frequency locking system includes a 50MHz DDS generation unit, an R-SOA bias superposition unit, an optoelectronic transimpedance amplification unit, a dual-balanced mixer demodulation unit, and an error signal conditioning unit, and is equipped with a modulation amplitude adaptive unit, a demodulation zero-point self-calibration unit, and a filter bandwidth dynamic matching unit; no external electro-optic modulator is required, and the demodulation error signal is... ,in, Frequency detuning / frequency deviation For the standard etalon transmission function, For optical carrier frequency, For radio frequency modulation frequency, For the upper sideband frequency, The lower sideband frequency is used. When Δv=0, the error signal returns to zero, corresponding to the complete alignment of the external cavity longitudinal mode with the etalon transmission peak. The FPGA intelligent three-level synchronous control system includes a mechanism-LSTM hybrid modeling feedforward unit, a first-level open-loop coarse synchronization unit, a second-level dual-channel displacement PID unit, a third-level optical fine PID unit, a mode-skipping-free timing unit, a lock-out recapture unit, and a DAC output unit. It also embeds a four-dimensional database incremental self-updating unit, a third-level adaptive PID self-tuning unit, an optical field dynamics hard constraint unit, a hierarchical fault diagnosis unit, an automatic operating condition identification unit, and a fully automatic calibration unit. The built-in mechanism-LSTM hybrid prediction model uses PZT second-order dynamics, FP etalon transmission Airy function, and laser rate... The equation serves as the physical constraint layer, using six-dimensional features (driving voltage, actual displacement, temperature, voltage derivative, displacement derivative, and scanning frequency) over a 50-step historical window as LSTM inputs, and outputting displacement compensation. A three-segment smooth step reset is employed for mode-skipping timing. The multi-field coupled temperature control and environmental monitoring system includes an R-SOA temperature control unit, a standard etalon temperature control unit, and an external cavity mirror temperature control unit, along with a three-way temperature control cross-decoupling unit, a temperature change rate feedforward pre-compensation unit, a dynamic linewidth monitoring unit, a frequency modulation linearity monitoring and pre-distortion unit, and a PZT aging state monitoring and lifespan prediction unit. The three independent TEC temperature control accuracy is ±0.01℃, and temperature, vibration, and air pressure data are simultaneously collected and fed into the hybrid model for thermo-mechanical-optical coupling compensation.

2. A high-speed swept-frequency laser driven by dual PZTs according to claim 1, characterized in that, The layered composite phase feedforward unit integrates PZT second-order dynamics-based phase compensation, scanning acceleration additional compensation, and graded frequency coefficients in a three-level superposition. The LSTM neural network has an input window length of T=50, and the input features are six dimensions: driving voltage, actual displacement, temperature, voltage derivative, displacement derivative, and scanning frequency. The dual-layer LSTM has 128 and 64 hidden layers respectively, a dropout rate of 0.2, and a 32-unit ReLU activated fully connected layer, outputting displacement compensation. The modulation amplitude adaptive unit dynamically adjusts the modulation current range of 5~15mA to maintain an optimal modulation depth of 0.3rad. The demodulation zero-point online self-calibration unit shuts down the inherent offset compensation of the 50MHz modulation acquisition circuit during the unlock gap.

3. A high-speed swept-frequency laser driven by dual PZTs according to claim 2, characterized in that, The four-dimensional database incremental self-updating unit automatically collects four-dimensional data of voltage, scanning frequency, temperature, and displacement error during mode hopping reset and lockout recapture intervals, and incrementally updates the database using local cubic interpolation without overwriting the factory basic calibration data, with synchronization accuracy attenuation <0.05MHz; the optical field dynamics hard constraint unit uses a photon lifetime of 443ps and a relaxation oscillation frequency of 700MHz R-SOA as physical limit thresholds, and automatically reduces scanning acceleration and increases feedforward compensation weight when the scanning rate approaches the threshold; the hierarchical fault diagnosis coding unit separates four types of error sources in real time: PZT phase error, temperature drift, environmental vibration, and optical tilt, outputs standardized fault codes, and accurately locates the corresponding secondary functional unit.

4. A high-speed swept-frequency laser driven by dual PZTs according to claim 3, characterized in that, The fully automated calibration pipeline unit traverses the entire temperature, voltage, and scanning frequency range with a single click, shortening the calibration time per machine and automatically classifying high-precision / standard / extreme high-speed equipment levels. The multi-field temperature control cross-coupling decoupling unit establishes a three-way temperature coupling mechanism matrix, pre-compensating for the synchronous proportional offset of the other two channels when a single channel experiences a temperature change. The temperature change rate feedforward pre-compensation unit collects the temperature differential dT / dt and adds a dynamic compensation term to the hybrid model, shortening the stabilization time in scenarios with sudden changes in ambient temperature. The adaptive smooth mode-hopping timing logic drives the PZT in three steps, with a total step time of <100μs. The capture scanning amplitude automatically switches between ±5nm / ±20nm levels based on the real-time frequency offset, with an average loop capture time of 0.3ms.

5. A high-speed swept-frequency laser driven by dual PZTs according to claim 4, characterized in that, The high-speed swept-frequency laser features continuous vibration compensation and PZT micro self-calibration logic every 10 hours, ensuring stable operation and extended service life. The high-voltage amplification unit uses an OPA454 high-voltage operational amplifier with an output of 0~150V, a slew rate ≥50V / μs, and an output current ≥100mA, coupled with overvoltage, overcurrent, and short-circuit hardware protection. Displacement acquisition employs a 24-bit Σ-Δ ADC with a sampling rate ≥200kHz and an effective bit depth ≥20bit. The DAC uses a 16-bit AD5761 with an update rate ≥1MHz. The improved Preisach hysteresis model, PZT second-order dynamic transfer function, and 600:1 FSR synchronization ratio constraint constitute a complete physical mechanism layer, serving as the constraint boundary for the LSTM network and preventing high-frequency extrapolation error explosion.

6. A high-speed swept-frequency laser driven by dual PZTs according to claim 5, characterized in that, The dual-telescopic PZT tuning drive system includes a standard telescopic PZT execution unit, an external cavity telescopic PZT execution unit, a high voltage amplification drive unit, and a PZT status acquisition and protection unit, and is equipped with a hierarchical composite phase feedforward unit, a creep online identification and correction unit, and an adaptive bandwidth current limiting protection unit. Standard etalon telescopic PZT actuator: Employs a stacked or tubular telescopic piezoelectric ceramic actuator, driving the 6T-FP standard etalon air gap thickness change through longitudinal telescopic movement to achieve coarse mode tuning in the 6THz range; the single FSR tuning displacement is λ / 2≈775nm, and the stroke meets the full-range scanning requirements of the standard etalon. External cavity telescopic PZT actuator: Employs a large-stroke stacked piezoelectric ceramic actuator, driving the external cavity mirror displacement through longitudinal telescopic movement to change the external cavity optical length and achieve fine longitudinal mode frequency tuning; the single FSR synchronous tuning displacement is 600×775nm=465μm, and the stroke is ±500μm. m, maintaining a fixed synchronization ratio of 600:1 with the standard PZT; High-voltage amplification drive unit: amplifies the low-voltage control signal output by the FPGA to a 0~150V high-voltage drive signal, with high-voltage slew rate and wide bandwidth characteristics, driving the dual-channel telescopic PZT actuator to respond quickly, with built-in overvoltage, overcurrent, short-circuit hardware protection and EMI filtering; PZT status acquisition and protection unit: installs a temperature thermistor and vibration acceleration sensor near the telescopic PZT actuator to acquire the PZT ceramic temperature and structural vibration amplitude in real time, and cuts off the high-voltage drive when the threshold is exceeded and uploads the data to the hybrid modeling unit for error compensation; Layered composite phase feedforward unit: Integrating PZT second-order dynamics-based phase compensation, scanning acceleration additional compensation, and graded frequency coefficients in a three-level superposition, it compensates for the phase lag of the stretchable PZT under high-speed scanning, reducing the phase lag from 170° to less than 20° at 10kHz scanning; Creep online identification and correction unit: Continuously acquiring hourly creep displacement under displacement closed-loop steady state, incrementally updating the hybrid modeling four-dimensional residual database in real time, and performing piecewise polynomial fitting of the hysteresis voltage range; Adaptive bandwidth current limiting protection unit: Setting up a digitally controllable RC bandwidth network, the FPGA automatically switches the drive bandwidth according to the real-time scanning frequency; dynamically adjusting the upper limit of the output current based on the temperature collected by the PZT thermistor, reducing power at high temperatures to slow piezoelectric aging and suppressing high-speed resonance noise.

7. A three-stage synchronous mode-skipping-free frequency sweeping method for a high-speed frequency sweeping laser driven by dual PZTs as described in claim 6, characterized in that, It includes the following execution phases: Phase 1: Fully Automated Mechanism-Data Hybrid Modeling and Offline Calibration FPGA fully automatic calibration unit execution: S1.

1. The multi-zone temperature control unit automatically traverses the temperature range of -10℃ to 60℃ in segments; S1.

2. The open-loop unit automatically outputs a 0~10V drive voltage in segments, and the displacement sensing unit synchronously collects nanometer-level displacement data; S1.

3. It automatically traverses multiple scanning frequencies of 1kHz, 2kHz, 5kHz, 10kHz, and 30kHz, and collects the original errors of phase hysteresis, creep, and hysteresis; S1.

4. It automatically constructs a four-dimensional residual database, solidifies the cubic interpolation function, and synchronously fits the second-order dynamics of PZT, the transmission Airy function of the FP etalon, the laser rate equation, and the complete set of mechanism parameters of the optical field dynamics; S1.

5. It trains the LSTM timing compensation network and solidifies it to the FPGA; S1.

6. Automatically complete equipment classification determination, store the corresponding operating condition parameter library, and output a complete factory calibration report; Phase Two, Three-Level Hierarchical Adaptive Synchronous Real-Time Control: S2.

1. First-level open-loop coarse synchronization, the working condition automatic identification unit determines the working level, retrieves the corresponding layered composite phase feedforward coefficient and etalon fineness mechanism parameters, the mechanism-LSTM hybrid model outputs a 600:1 ratio feedforward voltage, and coarsely aligns the etalon transmission peak with the external cavity longitudinal mode. S2.

2. Secondary displacement closed-loop synchronization: The differential capacitance sensing unit outputs dual displacement after common-mode temperature drift decoupling and adaptive vibration filtering preprocessing. The adaptive PID calculation unit matches the best control parameters according to the current working conditions to correct the drive voltage, and the synchronization error is compressed to ±1nm. S2.

3. Three-level optical error fine closed loop: the internal modulation frequency locking system adaptively adjusts the modulation amplitude to maintain optimal modulation depth, dynamically switches the filter bandwidth, and calibrates the demodulation zero point online; the mirror tilt correction unit dynamically corrects beam tilt; optical PID finely adjusts the external cavity PZT drive voltage, with synchronization accuracy ≤ ±0.1MHz and SMSR stability > 45dB; synchronous operation optical field dynamics hard constraint unit automatically reduces speed and increases feedforward compensation weight when the scan rate approaches the relaxation oscillation limit; layered error diagnosis unit outputs fault codes in real time. Phase 3, Adaptive Smooth Mode-Jump-Free Reset Continuous Frequency Sweep: S3.

1. The displacement sensing unit monitors the displacement of the etalon PZT in real time. When the displacement reaches +775nm, the timing unit is triggered. S3.

2. The timing unit performs a three-stage smooth step transition. The etalon PZT returns to zero in three steps, and the external cavity PZT synchronously transitions to -465μm in three steps. S3.

3. The adaptive capture scanning amplitude logic dynamically configures the scanning range according to the real-time error amplitude and performs a 100Hz small-amplitude scan to search for the optical error zero point. S3.

4. After detecting the zero point, it automatically switches to a three-level optical fine closed loop. The loop capture time is <1ms, the total mode-hopping time is <2ms, and the output optical power fluctuation is ≤0.3dB. During mode skipping intervals, incremental self-updates of the four-dimensional residual database are performed synchronously to continuously complete uninterrupted frequency sweeping of the C-band at 4.4THz; Phase Four: Hierarchical Diagnostic Loss Detection and Recapture S4.

1. The optical error conditioning unit outputs the error amplitude in real time. After the unlocking logic unit determines that the lock has been lost, it immediately switches to the second-level displacement closed loop. S4.

2. Adaptive amplitude scanning to search for the optical error zero point; S4.

3. If three consecutive capture failures occur, a full-range reset is performed, and the temperature, vibration, and scanning frequency data of this loss of lock are recorded and incrementally updated in the hybrid modeling residual database. S4.

4. The hierarchical error diagnosis unit outputs standardized fault codes and uploads them to the host computer to complete fault tracing; Phase 5, Long-term Adaptive Compensation in the Background: The entire machine operates in parallel with vibration countermeasures, temperature change rate pre-compensation, and PZT micro-calibration every 10 hours, employing a triple background logic to suppress creep, aging, and environmental disturbance errors over the long term.

8. The three-stage synchronous mode-skipping-free frequency sweeping method for a high-speed frequency sweeping laser driven by dual PZTs according to claim 7, characterized in that, The mechanism-LSTM hybrid model integrates four physical mechanisms as constraints: PZT second-order dynamics, improved Preisach hysteresis, FP etalon transmission Airy function, and laser rate equation. The LSTM time-series network predicts the physical model residuals using six-dimensional features from 50 historical steps. The standard operating condition is a 10kHz scan rate of 44THz / ms, with a synchronization error <1.8GHz and SMSR ≥40dB; the extreme operating condition is a 30kHz scan rate of 132THz / ms, with a synchronization error <8GHz and SMSR ≥30dB; and the high-precision operating condition is a 2kHz scan rate of <0.5GHz, with SMSR ≥45dB and dynamic linewidth <100kHz.

9. A three-stage synchronous mode-skipping-free frequency sweeping method for a high-speed frequency sweeping laser driven by dual PZTs according to claim 8, characterized in that, The multi-field coupled temperature control system reduces the stabilization time for sudden temperature changes from 30 seconds to 3 seconds by using cross-coupling decoupling and feedforward pre-compensation for temperature change rate.

10. A three-stage synchronous mode-skipping-free frequency sweeping method for a high-speed frequency sweeping laser driven by dual PZTs according to claim 9, characterized in that, The optical field dynamics hard constraint unit has a built-in hard constraint of 443ps photon lifetime and 700MHz relaxation oscillation frequency; the adaptive vibration time domain filtering and vibration active collaborative compensation logic jointly suppress 1~10kHz environmental vibration; the high-speed sweep frequency laser automatically starts micro static calibration every 10 hours of continuous operation to update the piezoelectric coefficient theoretical model to compensate for PZT long-term aging drift.

Images