Composite control system and method for suppressing timing jitter of passive Q-switched laser

Abstract

The invention provides a compound control system and method for suppressing timing jitter of a passively Q-switched laser, and the system obtains the time sequence and energy information of laser output in real time through a photoelectric sampling circuit, constructs a high-precision closed-loop feedback mechanism, achieves the subnanosecond measurement of a laser pulse interval, and provides a data basis for the precise sensing of the time sequence jitter. A neural network model is introduced into a closed-loop control circuit, nonlinear dynamic characteristics of an output pulse sequence of a laser can be effectively modeled, a long-time dependency relationship is captured, and prospective prediction of light-emitting delay deviation and energy deviation of a next pulse is realized. A PID compensator is combined to carry out fusion adjustment on a prediction error and a real-time feedback signal, a control instruction with a dynamic response capability is generated, and the adaptability of the system to factors such as environment disturbance, device aging and temperature drift is remarkably improved.

Description

Technical Field

[0001] This application relates to the field of laser control technology, and in particular to a composite control system and method for suppressing timing jitter in a passively Q-switched laser. Background Technology

[0002] In the research and application of solid-state lasers, Q-switching technology, as a key means to achieve high-energy, short-pulse output, has long received widespread attention. The core of Q-switching technology is to dynamically control the loss within the laser resonant cavity, suppress laser oscillation in the early stages of pumping, and accumulate a large number of inverted particles in the gain medium. Subsequently, the cavity loss is rapidly reduced, causing the stored energy to be released in a concentrated manner in the form of giant pulses, thereby obtaining a peak power much higher than that of continuous output. It can be divided into two types: active and passive Q-switching.

[0003] Active Q-switching relies on external electrical or acoustic signals to drive electro-optic or acousto-optic modulators, offering advantages such as high repetition rate, good pulse stability, and ease of synchronization. However, active Q-switching requires complex high-voltage control circuitry, limiting its application in miniaturized and low-cost scenarios. In contrast, passive Q-switching uses a saturable absorber as its core component, leveraging its nonlinear absorption characteristics to achieve automatic Q-value switching. At low light intensities, the saturable absorber strongly absorbs the laser, maintaining a high-loss state within the cavity. When the intracavity light intensity reaches its saturation threshold, the absorber is "bleached," the transmittance rapidly increases, and the cavity Q-value surges, triggering the formation of a giant pulse. Due to their simple structure and small size, passively Q-switched lasers are widely used in photoacoustic imaging, lidar, and other fields. However, the recovery time and saturation threshold of the saturable absorber, as well as the energy storage process of the gain medium, are affected by various factors such as ambient temperature, pump power fluctuations, and device aging. This results in randomness in the timing of each pulse generation and easy drift in the repetition frequency, especially under low repetition frequency or variable power operation conditions. This severely restricts the application of passive Q-switching under high precision requirements.

[0004] To improve the stability of Q-switched systems, technologies such as material innovation, waveguide acousto-optic all-fiber integration, and synergistic nonlinear effects optimize carrier dynamics, reduce losses, and increase damage thresholds, driving the development of lasers towards higher energy, narrower pulse widths, and higher stability. However, these methods suffer from problems such as complex material fabrication processes, high costs, and insufficient system flexibility. Furthermore, existing passive Q-switching systems lack real-time feedback mechanisms, making it difficult to adapt to dynamic load changes. Summary of the Invention

[0005] In view of the above practical problems and the shortcomings of the existing technology, the main technical problem to be solved by the present invention is to provide a composite control system and method for suppressing timing jitter in passively Q-switched lasers.

[0006] To address the aforementioned technical problems, this application provides a composite control system for suppressing timing jitter in passively Q-switched lasers, employing the following technical solution: including:

[0007] A laser head, comprising a pump source, a laser gain medium, a passively Q-switched crystal, and an optical resonant cavity, for generating passively Q-switched laser pulses;

[0008] The photoelectric sampling circuit is used to collect the pulse timing signal and energy signal of the laser output by the laser head in real time, and convert the pulse timing signal into a digital timestamp with a resolution of no more than 22ps, and convert the interval between adjacent laser pulses into a high-precision time difference with a resolution of no more than 45ps.

[0009] The closed-loop control circuit includes a neural network model and a PID compensator. The neural network model receives historical pulse timing data collected by the photoelectric acquisition circuit and predicts the delay deviation and energy deviation of the next pulse. The PID compensator integrates the predicted value of the neural network model with the real-time feedback signal and generates pulse interval adjustment command and voltage amplitude adjustment command through a proportional-integral-derivative algorithm.

[0010] The pulse drive circuit is used to receive the instructions from the PID compensator and output a pump drive signal with a nanosecond-level adjustable pulse interval and a dynamically adjustable voltage amplitude to the pump source, thereby realizing dual-parameter closed-loop feedforward control of time delay and energy.

[0011] In a preferred embodiment, the photoelectric acquisition circuit includes a beam splitter, a high-speed photodetector, a signal conditioning circuit, and a time-to-digital converter;

[0012] The beam splitter is used to split the laser output from the optical resonant cavity into a main output beam and a feedback beam, and to connect the feedback beam to the high-speed photodetector.

[0013] The high-speed photodetector is used to convert the feedback optical signal into an electrical signal and input the electrical signal to the signal conditioning circuit;

[0014] The signal conditioning circuit includes a high-speed comparator and a monostable multivibrator. The high-speed comparator shapes the electrical signal, and the shaped signal is input to the monostable multivibrator to obtain an extended pulse width output signal. The response time of the high-speed comparator is less than 1 ns, and the adjustable pulse width of the monostable multivibrator output is 0.1-100 ns. ;

[0015] The time-to-digital converter is used to generate a reference clock and capture the interval between adjacent laser pulses, outputting high-precision time difference data.

[0016] In a preferred embodiment, the neural network model is a temporal convolutional neural network, which uses causal convolution, dilated convolution and residual connection structures to extract temporal features, and the output layer is mapped to the joint predicted value of time delay bias and energy bias through a fully connected layer;

[0017] The input to the temporal convolutional neural network is historical pulse sequence data, which includes the time delay deviation of the pulse interval and the driving pulse energy deviation of N consecutive pulses, where N is not less than 10; the time deviation of the pulse interval is calculated from the adjacent pulse interval output by the time-to-digital converter, and the driving pulse energy deviation is obtained by the photoelectric sampling circuit after measurement and digital processing.

[0018] The temporal convolutional neural network outputs a predicted delay deviation and an predicted energy deviation for the next pulse. The predicted delay deviation is used to generate the pulse interval adjustment command, and the predicted energy deviation is used to generate the voltage amplitude adjustment command.

[0019] In a preferred embodiment, the temporal convolutional neural network comprises no more than 3 causal convolutional layers, each convolutional kernel width is no greater than 5, the number of channels is no greater than 64, residual connections are used between layers, and ReLU is used as a non-linear activation function.

[0020] In a preferred embodiment, the pulse driving circuit supports dynamic adjustment of the interval between adjacent pulses in steps no greater than 10 ns; dynamic adjustment of the output voltage amplitude in steps no greater than 0.1 V; and the voltage amplitude adjustment range is 0–5.2 V.

[0021] The pulse drive circuit uses a multi-parameter linkage control algorithm to synchronously adjust the pulse interval and voltage amplitude of the drive circuit, controlling the time delay deviation within ±3ns and the energy fluctuation within ±5%.

[0022] In a preferred embodiment, a threshold T1 is set based on the repetition frequency stability requirement, and a threshold T2 is set based on the pump source safety operation requirement; the logic of the multi-parameter linkage control includes:

[0023] If the time delay deviation is less than the threshold T1, the pulse interval of the drive circuit is adjusted first to keep the pulse voltage amplitude of the drive circuit constant.

[0024] If the time delay deviation is greater than the threshold T1 and the pulse voltage amplitude of the driving circuit is less than the threshold T2, then the pulse voltage amplitude of the driving circuit should be adjusted first to keep the pulse interval of the driving circuit unchanged.

[0025] If the time delay deviation is greater than the threshold T1 and the pulse voltage amplitude of the drive circuit is greater than the threshold T2, then the pulse voltage amplitude of the drive circuit is set to T2 to adjust the pulse interval of the drive circuit.

[0026] In a preferred embodiment, the closed-loop control circuit is deployed on an embedded MCU platform, and the neural network model achieves real-time inference after model compression and fixed-point quantization, with a single prediction latency of less than 1. .

[0027] In a preferred embodiment, the laser gain medium is selected from one of Nd:YAG, Yb:YAG, or YVO4 crystals; the passively Q-switched crystal uses Cr 4+ YAG, one side of which is directly bonded to the laser gain medium; the pump source is a pulse-driven high-brightness laser diode, the operating wavelength of which matches the absorption peak of the laser gain medium.

[0028] To address the aforementioned technical problems, this application provides a composite control method for suppressing timing jitter in passively Q-switched lasers, employing the following technical solution: The control method includes:

[0029] The pulse timing signal and energy signal output by the laser head are acquired in real time by the photoelectric sampling circuit. The acquired data is input into the time convolutional neural network model to predict the time delay deviation and energy deviation of the next pulse. The predicted values ​​and real-time feedback signals are input into the PID compensator, and the pulse interval adjustment and voltage amplitude adjustment are calculated by the proportional-integral-derivative algorithm. The pulse drive circuit outputs the corresponding pump drive signal to the pump source according to the adjustment, realizing the dual-parameter closed-loop feedforward control of time delay and energy.

[0030] In a preferred embodiment, the input samples of the temporal convolutional neural network model are generated using a sliding window method, with a window size of not less than 15. Each sample contains the pulse interval and energy value features of N consecutive historical pulses. After standardization, the samples are input into the causal convolutional layer, and the network output layer generates a joint prediction value for the delay deviation and energy deviation of the next pulse.

[0031] In summary, this application has the following beneficial effects:

[0032] This invention proposes a composite control system for suppressing timing jitter in passively Q-switched lasers. It acquires real-time timing and energy information of the laser output through a photoelectric sampling circuit, constructing a high-precision closed-loop feedback mechanism to achieve sub-nanosecond measurement of laser pulse intervals, providing a data foundation for accurate timing jitter sensing. A neural network model is introduced into the closed-loop control circuit, effectively modeling the nonlinear dynamic characteristics of the laser output pulse sequence, capturing long-term dependencies, and enabling look-ahead prediction of the next pulse's output delay and energy deviation. A PID compensator is combined to fuse and adjust the prediction error with the real-time feedback signal, generating control commands with dynamic response capabilities, significantly improving the system's adaptability to environmental disturbances, device aging, and temperature drift. Attached Figure Description

[0033] Figure 1 A schematic diagram of a composite control system for suppressing timing jitter in a passively Q-switched laser provided by the present invention;

[0034] Figure 2 The optical path diagram of the passive Q-switched laser head provided by this invention;

[0035] Figure 3 This is a schematic diagram of the composite control process of the TCN neural network layer and the PID compensation layer provided by the present invention;

[0036] Figure 4 This is a schematic diagram of the pulse drive circuit provided by the present invention. Detailed Implementation

[0037] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. All other embodiments obtained by those skilled in the art based on the embodiments of the present invention without creative effort are within the scope of protection of the present invention.

[0038] In the description of this invention, it should be noted that the terms "upper," "lower," "inner," "outer," "top / bottom," etc., indicate the orientation or positional relationship based on the orientation or positional relationship shown in the accompanying drawings. They are used only for the convenience of describing the invention and for simplifying the description, and do not indicate or imply that the device or element referred to must have a specific orientation, or be constructed and operated in a specific orientation. Therefore, they should not be construed as limitations on the invention. Furthermore, the terms "first" and "second" are used for descriptive purposes only and should not be construed as indicating or implying relative importance.

[0039] In the description of this invention, it should be noted that, unless otherwise explicitly specified and limited, the terms "installed", "equipped", "sleeved / connected", "connected", etc., should be interpreted broadly. For example, "connection" can be a wall-mounted connection, a detachable connection, or an integral connection; it can be a mechanical connection or an electrical connection; it can be a direct connection or an indirect connection through an intermediate medium; it can be a connection within two components. For those skilled in the art, the specific meaning of the above terms in this invention can be understood according to the specific circumstances.

[0040] The following is in conjunction with the appendix Figures 1-4 This application will be described in further detail.

[0041] This embodiment provides a composite control system for suppressing timing jitter in a passively Q-switched laser. It achieves a stable repetition rate for the passively Q-switched laser through a fusion control strategy combining a neural network and a PID compensator, making it suitable for high-precision applications such as photoacoustic imaging and precision measurement. The system includes a laser head, a photoelectric sampling circuit, a closed-loop control circuit, and a pulse drive circuit. Wherein:

[0042] The laser head receives electrical energy and then emits pump light to excite the laser crystal. The gain of the laser crystal is proportional to the accumulated pump light energy it receives. When the gain is greater than the loss, a laser pulse is formed and output through the laser resonant cavity. The laser head includes, from left to right, a pump source, coupling optical elements, a laser gain medium, and a passively Q-switched crystal. The laser gain medium and the passively Q-switched crystal are directly bonded to form a bonded crystal. The bonded crystal and the output coupling mirror form a laser resonant cavity, which is used to absorb and amplify the energy output by the pump source, generating and outputting the optical pulse of the passively Q-switched laser.

[0043] In this embodiment, the laser resonator is a straight cavity with a short length. The pump source is a pulse-driven high-brightness laser diode (LD) with an operating wavelength matching the absorption peak of the laser gain medium, preferably 808nm (for Nd:YAG), 940nm (for Yb:YAG), or 808 / 880nm dual-wavelength pumping (for YVO4) to achieve efficient energy conversion. The coupling optical element can use a modular focusing lens group with a focal length of 11mm to collimate and focus the output light from the fiber; it can also use thermally bonded Nd:YAG / Cr... 4 ⁺:YAG bonded crystal, wherein the Nd:YAG layer thickness is 1 mm, the doping concentration is 1.0 ± 0.1 at.%, and Cr 4 ⁺: The YAG layer thickness is 1.5mm, with an initial transmittance of 85%. The two ends of the crystal are coated with a high-reflection coating (HR@1064nm) and an anti-reflection coating (AR@808nm), resulting in an output mirror transmittance of 30%.

[0044] In this embodiment, the laser gain medium is selected from one of Nd:YAG, Yb:YAG, or YVO4 crystals, and the passively Q-switched crystal uses Cr. 4+ YAG, one side of which is directly bonded to the laser gain medium, reduces the number of optical interfaces in the cavity, reduces losses and thermal lensing effects, and improves beam quality and long-term operational stability.

[0045] The photoelectric acquisition circuit is used to acquire the pulse timing signal and energy signal of the output laser in real time, and convert the pulse timing signal into a digital timestamp with a resolution of ≤22ps. The photoelectric acquisition circuit includes a beam splitter, a high-speed photodetector, a signal conditioning circuit, and a time-to-digital converter (TDC).

[0046] The beam splitter divides the laser output from the laser resonant cavity into two parts: output light and feedback light. The feedback light is connected to the high-speed photodetector. The high-speed photodetector converts the feedback light signal into an electrical signal and inputs the electrical signal to the signal conditioning circuit. The signal conditioning circuit is used for pulse shaping. The signal conditioning circuit uses a high-speed comparator to connect the shaped signal to the input of a monostable multivibrator and adjusts the output to obtain a signal with extended pulse width. The time-to-digital converter (TDC) generates a reference clock through a ring oscillator, uses the START / STOP signal to capture the interval between adjacent laser pulses, and combines the calibration data of the delay chain and the ring oscillator to output a high-precision time difference between adjacent laser pulses (resolution 45 ps).

[0047] like Figure 1 As shown, the specific connection method of the photoelectric sampling circuit is as follows: The beam splitter divides the optical pulse output from the laser resonant cavity into two paths. One path serves as the main optical path output, and the other path converts the optical signal into an electrical signal through a high-speed photodetector. The electrical signal output from the high-speed photodetector is input to the signal conditioning circuit. The signal conditioning circuit can be composed of a high-speed comparator and a monostable multivibrator. The high-speed comparator shapes the obtained electrical pulse and inputs it to the monostable multivibrator. After receiving a brief trigger signal, the monostable multivibrator outputs a pulse signal with a stable amplitude whose width is determined by the values ​​of externally connected resistors and capacitors. The adjustable pulse width range is 0.1-100. The electrical signal is sent to a time-to-digital converter (TDC), which then converts the time interval of the electrical signal into a digital signal.

[0048] In this embodiment, the high-speed comparator in the signal conditioning circuit has a response time of less than 1 ns, and the adjustable pulse width of the monostable multivibrator output is 0.1-100 ns. This is used to match the input signal width requirements of the time-to-digital converter (TDC), ensuring the reliability and anti-interference capability of time measurement. A fiber optic beam splitter with a splitting ratio of 4:1 is employed. The high-speed photodetector is an avalanche photodiode (APD) with a response speed ≥10GHz and an operating wavelength of 1064nm. The high-speed comparator is an LMH7322, and the monostable multivibrator is an AD8310. External connections include a 1kΩ resistor and a 10nF capacitor. The pulse width output range is adjustable from 0.1 to 100. The trigger outputs a stable square wave signal to the time-to-digital converter TDC-GP22.

[0049] The closed-loop control circuit adopts a fusion control strategy of time convolutional neural network and PID compensator. By combining the timing prediction capability of neural network and the dynamic response capability of PID compensator, pulse interval adjustment command and voltage amplitude adjustment command are generated to achieve stable control of repetition frequency.

[0050] The main control unit of the closed-loop control circuit collects continuous pulse sequence data as input to the time convolutional neural network. The data includes the time difference sequence between two adjacent pulses sampled by the time-to-digital converter (TDC) and the historical drive signal energy sequence of the pulse drive circuit calculated from historical data. The historical pulse timing data includes the time delay deviation and drive pulse energy deviation over N consecutive pulse intervals, where N ≥ 10, and the time delay deviation Δt... i The energy deviation ΔE is calculated from the standard deviation of the interval between adjacent pulses output by the time-to-digital converter (TDC) module. i It is obtained by measuring and digitizing using a photodetector and signal conditioning circuit.

[0051] In this embodiment, the sliding window method is used to generate input samples for the neural network model (TCN). The window size is not less than 15, and each sample contains the pulse interval and energy value features of 15 consecutive historical pulses. The sample shape is (15, 2). The time delay deviation Δt of each sample for the next pulse is also specified. i and energy deviation ΔE i The sample shape is (2,). After standardizing these two feature values, they are input into the causal convolutional layer.

[0052] The TCN neural network model is a regression predictor that takes a multivariate time series segment of shape (15,2) as input and outputs a bi-objective prediction for the next time step. Each prediction requires the simultaneous output of two independent but related physical quantities, hence the shape (2,). In Python, (2,) represents a one-dimensional vector containing two scalar elements.

[0053] The closed-loop control circuit includes a neural network model (TCN) and a PID compensator. The closed-loop control circuit can use a ZynqAXU3EGB as the main control unit to deploy a lightweight TCN time convolutional neural network model and complete PID compensation and control command output. The input to the neural network model is the timing deviation signal output from the time-to-digital converter (TDC) and historical pulse sequence data (including pulse interval and drive pulse energy). The output of the neural network model is the predicted delay deviation (Δt) and energy deviation (ΔE) for the next pulse. The predicted delay deviation (Δt) is used to generate pulse interval adjustment commands, and the predicted energy deviation (ΔE) is used to generate voltage amplitude adjustment commands.

[0054] Temporal features are extracted using a multi-layer causal convolutional and residual connection structure. The output layer is mapped to a joint predicted value of time delay deviation and energy deviation through a fully connected layer, serving as the feedforward input for the fusion control strategy. The PID compensator integrates the output value of the neural network model with the real-time feedback signal to generate pulse interval adjustment commands and voltage amplitude adjustment commands.

[0055] In this embodiment, the closed-loop control circuit is deployed on an embedded MCU platform. The neural network model (TCN) is processed by model compression and fixed-point quantization to achieve real-time inference, with a single prediction latency of less than 1. This meets the closed-loop control bandwidth requirements of high-repetition-rate lasers (≥10kHz). For example... Figure 3 The neural network model (TCN) analyzes the time characteristics of the digital signal to generate compensation parameters, which are then passed to a PID compensator. The PID compensator dynamically adjusts the output of the pulse drive circuit based on the compensation parameters, thereby regulating the drive current of the pump source in real time and achieving closed-loop control of the laser pulse width and energy.

[0056] The neural network model (TCN) has a network structure consisting of ≤3 causal convolutional layers, with each convolutional kernel width ≤5 and the number of channels ≤64. Gradient propagation capability is enhanced through residual connections. The causal convolutional design ensures that the output depends only on historical time-series data, avoiding future information contamination. The dilated convolution expands the receptive field exponentially, improving the ability to model long-term dependencies.

[0057] In this embodiment, the lightweight neural network model (TCN) achieves learning and prediction through dilated convolutions and residual connections in three causal convolutional layers. The network consists of three cascaded causal convolutional layers: the first neural network layer is a causal convolutional layer with a dilation coefficient d=1, performing one-dimensional convolution along the time axis, a kernel size of 5×1, 2 input channels corresponding to the pulse interval and pulse energy, and 64 output channels. A left padding strategy is used with a padding length of 4 to maintain the consistency of the time series length. The second neural network layer is a causal convolutional layer with a dilation coefficient d=2, a kernel size of 5×1, 64 input and 64 output channels, and a left padding length of 8 to ensure that the receptive field expands with depth while maintaining sequence alignment. The third neural network layer is a causal convolutional layer with a dilation coefficient d=4, with other parameters consistent with the second layer, and a left padding length of 16 to capture longer-distance temporal dependencies. Each convolutional layer uses ReLU as a non-linear activation function to enhance the model's expressive power.

[0058] like Figure 3As shown, the neural network model (TCN) introduces a 1×1 convolutional cross-layer residual connection between the first and third layers. This does not change the spatial dimension of the feature map, but only performs information fusion between channels, which is used to alleviate the gradient vanishing problem in deep network training and improve feature propagation efficiency. The network ends at the output layer, which flattens the output tensor of the third convolution into a one-dimensional vector, and performs dimension mapping through a fully connected layer to generate a joint prediction value for the pulse delay deviation Δt and energy deviation ΔE in the next cycle.

[0059] like Figure 3 As shown, the PID compensation layer of the PID compensator receives the real-time feedback signal output from the neural network layer, including the pulse interval and the pulse energy of the drive circuit. Based on the difference between the real-time feedback signal and the predicted value, the time error is calculated. ) and energy error ( The proportional, integral, and derivative terms are weighted and fused to generate the initial control command.

[0060] The pulse drive circuit receives instructions from the PID compensator, dynamically adjusts the pump drive parameters, and outputs a pump drive signal with an adjustable pulse interval in the nanosecond (ns) range and a dynamically adjustable voltage amplitude (range 0–5.2V) to the pump source. This achieves dual-parameter closed-loop feedforward control of time delay and energy, which not only effectively suppresses the inherent repetition frequency jitter problem of passively Q-switched lasers, but also controls the output stability within ±5%, which is far superior to traditional open-loop or single feedback control methods.

[0061] The pulse drive circuit receives an external low-phase-noise reference clock, multiplies it using an internal phase-locked loop and a voltage-controlled oscillator to generate a high-frequency time base, and uses a programmable delay unit (DLL) to achieve independent phase offset adjustment of the output channel, with a delay resolution of up to picosecond level. The output channel is then frequency-divided and phase-configured before its amplitude is adjusted to generate the high-precision closed-loop feedforward control signal.

[0062] like Figure 4 As shown, the pulse drive circuit receives an external low phase noise reference clock and generates a high-precision closed-loop feedforward control signal with dual degrees of freedom in time delay and energy through phase-locked loop frequency multiplication, programmable delay unit phase adjustment, single-channel frequency division and configuration, and amplitude adjustment.

[0063] In this embodiment, a PGA and a high-speed DAC unit can be used. The pulse drive circuit receives a 4MHz external low-phase-noise reference clock signal to provide a reference for the subsequent generation of the high-frequency time base. The reference clock signal is fed into a phase-locked loop (PLL), where it undergoes frequency multiplication by a phase detector, loop filter, and voltage-controlled oscillator (VCO) to generate a 20GHz high-frequency clock signal. The multiplied high-frequency clock signal then enters a programmable delay unit for phase offset adjustment, achieving picosecond-level delay resolution. The adjusted high-frequency clock signal is then frequency-divided according to the control commands of the closed-loop control circuit and undergoes amplitude adjustment to finally output the voltage amplitude at any given time. The rising and falling edges are precisely controlled by the drive pulse signal output by the closed-loop control circuit.

[0064] The pulse drive circuit supports dynamic adjustment of the interval between adjacent pulses in steps ≤10ns; dynamic adjustment of the output voltage amplitude in steps ≤0.1V; and synchronous adjustment of the pulse interval and voltage amplitude of the drive circuit through a multi-parameter linkage control algorithm, so that the time delay deviation Δt ≤±3ns and the energy fluctuation ≤±5%.

[0065] The logic of the multi-parameter linkage control includes:

[0066] If the time delay deviation is less than the threshold T1, the pulse interval of the drive circuit is adjusted first to keep the pulse voltage amplitude of the drive circuit constant.

[0067] If the time delay deviation is greater than the threshold T1 and the pulse voltage amplitude of the driving circuit is less than the threshold T2, then the pulse voltage amplitude of the driving circuit should be adjusted first to keep the pulse interval of the driving circuit unchanged.

[0068] If the time delay deviation is greater than the threshold T1 and the pulse voltage amplitude of the drive circuit is greater than the threshold T2, then the pulse voltage amplitude of the drive circuit is set to T2 to adjust the pulse interval of the drive circuit.

[0069] When the continuous pulse timeout or energy exceeds the limit, the pump source is triggered to shut down or the pulse parameters are reset to the preset safe value.

[0070] Includes a fault protection module: control parameters are dynamically updated based on the joint prediction output of a neural network and a PID compensator, achieving closed-loop stability of the repetition frequency. For example... Figure 3 As shown, the output control layer of the pulse drive circuit converts the initial control command generated by the PID compensation layer into the actual drive signal, and determines which control strategy or fault protection to adopt based on whether the pulse interval deviation and the pulse voltage amplitude of the drive circuit exceed the preset threshold.

[0071] This embodiment also provides a composite control method for suppressing timing jitter in a passively Q-switched laser. This control method is executed by the closed-loop control circuit of the composite control system, and includes:

[0072] The pulse timing signal and energy signal output by the laser head are acquired in real time by the photoelectric sampling circuit. The acquired data is input into the time convolutional neural network model to predict the time delay deviation and energy deviation of the next pulse. The predicted values ​​and real-time feedback signals are input into the PID compensator, and the pulse interval adjustment and voltage amplitude adjustment are calculated by the proportional-integral-derivative algorithm. The pulse drive circuit outputs the corresponding pump drive signal to the pump source according to the adjustment, realizing the dual-parameter closed-loop feedforward control of time delay and energy.

[0073] The above description is merely a preferred embodiment of the present invention, but the design concept of the present invention is not limited thereto. Any non-substantial modifications made to the present invention by those skilled in the art within the scope of the technology disclosed in the present invention using this concept shall be deemed as an infringement of the protection scope of the present invention.

Claims

1. A composite control system for suppressing timing jitter in a passively Q-switched laser, characterized in that: include: A laser head, comprising a pump source, a laser gain medium, a passively Q-switched crystal, and an optical resonant cavity, for generating passively Q-switched laser pulses; The photoelectric sampling circuit is used to collect the pulse timing signal and energy signal of the laser output by the laser head in real time, and convert the pulse timing signal into a digital timestamp with a resolution of no more than 22ps, and convert the interval between adjacent laser pulses into a high-precision time difference with a resolution of no more than 45ps. The closed-loop control circuit includes a neural network model and a PID compensator. The neural network model receives historical pulse timing data collected by the photoelectric acquisition circuit and predicts the delay deviation and energy deviation of the next pulse. The PID compensator integrates the predicted value of the neural network model with the real-time feedback signal and generates pulse interval adjustment command and voltage amplitude adjustment command through a proportional-integral-derivative algorithm. The pulse drive circuit is used to receive the instructions from the PID compensator and output a pump drive signal with a nanosecond-level adjustable pulse interval and a dynamically adjustable voltage amplitude to the pump source, thereby realizing dual-parameter closed-loop feedforward control of time delay and energy.

2. The composite control system for suppressing timing jitter in a passively Q-switched laser according to claim 1, characterized in that: The photoelectric acquisition circuit includes a beam splitter, a high-speed photodetector, a signal conditioning circuit, and a time-to-digital converter; The beam splitter is used to split the laser output from the optical resonant cavity into a main output beam and a feedback beam, and to connect the feedback beam to the high-speed photodetector. The high-speed photodetector is used to convert the feedback optical signal into an electrical signal and input the electrical signal to the signal conditioning circuit; The signal conditioning circuit includes a high-speed comparator and a monostable multivibrator. The high-speed comparator shapes the electrical signal, and the shaped signal is input to the monostable multivibrator to obtain an extended pulse width output signal. The response time of the high-speed comparator is less than 1 ns, and the adjustable pulse width of the monostable multivibrator output is 0.1-100 ns. ; The time-to-digital converter is used to generate a reference clock and capture the interval between adjacent laser pulses, outputting high-precision time difference data.

3. The composite control system for suppressing timing jitter in a passively Q-switched laser according to claim 1, characterized in that: The neural network model is a temporal convolutional neural network, which uses causal convolution, dilated convolution and residual connection structure to extract temporal features. The output layer is mapped to the joint predicted value of time delay bias and energy bias through a fully connected layer. The input to the temporal convolutional neural network is historical pulse sequence data, which includes the time delay deviation of the pulse interval and the driving pulse energy deviation of N consecutive pulses, where N is not less than 10; the time deviation of the pulse interval is calculated from the adjacent pulse interval output by the time-to-digital converter, and the driving pulse energy deviation is obtained by the photoelectric sampling circuit after measurement and digital processing. The temporal convolutional neural network outputs a predicted delay deviation and an predicted energy deviation for the next pulse. The predicted delay deviation is used to generate the pulse interval adjustment command, and the predicted energy deviation is used to generate the voltage amplitude adjustment command.

4. The composite control system for suppressing timing jitter in a passively Q-switched laser according to claim 3, characterized in that: The temporal convolutional neural network contains no more than 3 causal convolutional layers, each with a kernel width of no more than 5 and a channel number of no more than 64. Residual connections are used between the layers, and ReLU is used as the non-linear activation function.

5. The composite control system for suppressing timing jitter in a passively Q-switched laser according to claim 1, characterized in that: The pulse driving circuit supports dynamic adjustment of the interval between adjacent pulses in steps no greater than 10ns; the output voltage amplitude is dynamically adjusted in steps no greater than 0.1V, and the voltage amplitude adjustment range is 0–5.2V. The pulse drive circuit uses a multi-parameter linkage control algorithm to synchronously adjust the pulse interval and voltage amplitude of the drive circuit, controlling the time delay deviation within ±3ns and the energy fluctuation within ±5%.

6. A composite control system for suppressing timing jitter in a passively Q-switched laser according to claim 5, characterized in that: The threshold T1 is set according to the repetition frequency stability requirement, and the threshold T2 is set according to the pump source safety operation requirement. The logic of the multi-parameter linkage control includes: If the time delay deviation is less than the threshold T1, the pulse interval of the drive circuit is adjusted first to keep the pulse voltage amplitude of the drive circuit constant. If the time delay deviation is greater than the threshold T1 and the pulse voltage amplitude of the driving circuit is less than the threshold T2, then the pulse voltage amplitude of the driving circuit should be adjusted first to keep the pulse interval of the driving circuit unchanged. If the time delay deviation is greater than the threshold T1 and the pulse voltage amplitude of the drive circuit is greater than the threshold T2, then the pulse voltage amplitude of the drive circuit is set to T2 to adjust the pulse interval of the drive circuit.

7. A composite control system for suppressing timing jitter in a passively Q-switched laser according to claim 1, characterized in that: The closed-loop control circuit is deployed on an embedded MCU platform. The neural network model, after model compression and fixed-point quantization, achieves real-time inference with a single prediction latency of less than 1 second. .

8. The composite control system for suppressing timing jitter in a passively Q-switched laser according to claim 1, characterized in that: The laser gain medium is selected from one of Nd:YAG, Yb:YAG, or YVO4 crystals; the passively Q-switched crystal uses Cr... 4+ YAG, one side of which is directly bonded to the laser gain medium; the pump source is a pulse-driven high-brightness laser diode whose operating wavelength matches the absorption peak of the laser gain medium.

9. A composite control method for suppressing timing jitter in a passively Q-switched laser, applied to the system described in any one of claims 1-8, characterized in that: The control method includes: The pulse timing signal and energy signal output by the laser head are acquired in real time by the photoelectric sampling circuit. The acquired data is input into the time convolutional neural network model to predict the time delay deviation and energy deviation of the next pulse. The predicted values ​​and real-time feedback signals are input into the PID compensator, and the pulse interval adjustment and voltage amplitude adjustment are calculated by the proportional-integral-derivative algorithm. The pulse drive circuit outputs the corresponding pump drive signal to the pump source according to the adjustment, realizing the dual-parameter closed-loop feedforward control of time delay and energy.

10. A composite control method for suppressing timing jitter in a passively Q-switched laser according to claim 9, characterized in that: The input samples of the temporal convolutional neural network model are generated using the sliding window method, with a window size of not less than 15. Each sample contains the pulse interval and energy value features of N consecutive historical pulses. After standardization, the samples are input into the causal convolutional layer, and the network output layer generates a joint prediction value for the delay deviation and energy deviation of the next pulse.

Images