High stability optical parametric oscillator assisted by intelligent feedback algorithm

Abstract

The application discloses a kind of high stability optical parametric oscillator assisted by intelligent feedback algorithm, and relates to the technical field of optical parametric oscillator.The oscillator includes 515 nm pump laser source, half-wave plate, polarization beam splitter, plane mirror, dichroic mirror, concave focusing cavity mirror, nonlinear crystal, output coupling mirror, four-quadrant photodetector based on STM32, DAC high-voltage drive board and piezoelectric deflection mirror.The application utilizes four-quadrant photodetector to monitor laser output power in real time, and collects data through the combination of ADC and direct memory access (DMA), realizes double buffering operation by means of full transmission and half transmission mechanism, to efficiently obtain the performance indicators required by SPGD algorithm.The algorithm continuously iterates and optimizes, controls piezoelectric deflection mirror in real time, precisely adjusts cavity length and beam pointing, and effectively stabilizes laser output power.

Description

Technical Field

[0001] This invention relates to the field of optical parametric oscillator technology, and in particular to a highly stable optical parametric oscillator based on the SPGD algorithm. Background Technology

[0002] Optical parametric oscillators (OPOs) are crucial laser frequency conversion devices, and the stability of their output power directly impacts their performance in applications such as precision measurement, spectral analysis, and quantum optics. In practical applications, the output power of OPOs is susceptible to interference from factors such as pump source fluctuations, temperature variations, and mechanical vibrations, leading to energy fluctuations and reduced system reliability and repeatability. Therefore, developing efficient and rapid power stabilization control methods is of significant practical value.

[0003] Traditional OPO power stabilization systems often employ proportional-integral-derivative (PID) control algorithms, adjusting pump power or cavity length via feedback to achieve stable output. However, PID control suffers from limitations such as limited response speed, complex parameter tuning, and poor adaptability to multiple disturbance coupling mechanisms. Especially in wide-tunable or ultrafast laser OPO systems, it is difficult to achieve high-bandwidth dynamic suppression at the millisecond or even microsecond level, limiting its application in practical scenarios. Summary of the Invention

[0004] The technical problem to be solved by the present invention is to provide a highly stable optical parametric oscillator based on an intelligent feedback algorithm to improve response speed and control accuracy.

[0005] To solve the above-mentioned technical problems, the technical solution adopted by the present invention is: a highly stable optical parametric oscillator assisted by an intelligent feedback algorithm, comprising a pump laser source, wherein the green light emitted from the pump laser source is first split by a half-wave plate and a polarizing beam splitter, then reflected by a plane mirror, transmitted through a first dichroic mirror and enters the resonant cavity of the optical parametric oscillator, the green light is reflected by a first concave focusing cavity mirror and focused into a nonlinear crystal to generate signal light and idler light; the signal light and idler light are reflected and collimated at a second concave focusing cavity mirror, the idler light is transmitted out of the resonant cavity of the optical parametric oscillator after passing through a second dichroic mirror, the signal light is reflected and incident on an output coupling mirror, the output signal light is incident on a beam splitter, and the remaining... Part of the energy is reflected and continues to oscillate within the resonant cavity. The reflected signal light returns to the resonant cavity, is reflected by the first dichroic mirror, and then enters the plane mirror, returning to the resonant cavity to meet the newly incident pump light synchronously. It then enters the nonlinear crystal again and oscillates repeatedly within the cavity. After the transmitted signal light passes through the beam splitter, part of its energy is collected by the four-quadrant photodetector and conditioned by the preamplifier before being connected to the microcontroller. The microcontroller processes the collected digital signal and then transmits it to the DAC high-voltage driver board through the SPI communication interface to generate a high-voltage analog drive signal. Finally, the signal signal acts on the piezoelectric deflector in a closed-loop feedback manner to achieve stable control of the laser output power.

[0006] The beneficial effects of the above technical solution are as follows: the S5980 chip in the four-quadrant detector of the oscillator can convert optical signals into current in the 320-1100 nm spectral range, and then internally amplify and adjust it into a 0-3.3V voltage signal; the STM32 microcontroller is configured with multi-channel ADC differential and single-ended input modes, and implements high-frequency iteration and signal processing of the SPGD algorithm in direct memory access (DMA) interrupts; the system outputs a voltage from 0 to +20V through the DAC81416, which is amplified by 6 times gain and then drives the piezoelectric deflector to achieve a maximum angle adjustment of 3 mrad, thereby quickly correcting the cavity length and beam pointing and stabilizing the output power in real time. This system significantly improves the response speed of output power jitter monitoring of the optical parametric oscillator and the control accuracy of the resonant cavity, and can complete a large range of power locking in the microsecond range, which is especially suitable for applications with high requirements for output power stability, such as bioimaging, biophotonics, and micro / nano fabrication. Attached Figure Description

[0007] The present invention will now be described in further detail with reference to the accompanying drawings and specific embodiments.

[0008] Figure 1 This is a schematic block diagram of the oscillator described in an embodiment of the present invention;

[0009] Figure 2 This is a control flowchart of the oscillator described in an embodiment of the present invention;

[0010] The components are: 1. Pump laser source; 2. Half-wave plate; 3. Polarizing beam splitter; 4. Plane mirror; 5. First dichroic mirror; 6. First concave focusing cavity mirror; 7. Nonlinear crystal; 8. Second concave focusing cavity mirror; 9. Second dichroic mirror; 10. Output coupling mirror; 11. Plane mirror; 12. Beam splitter; 13. Four-quadrant photodetector; 14. Microcontroller; 15. DAC high-voltage driver board; 16. Piezoelectric deflector; 17. Green light; 18. Signal light; 19. Idle light. Detailed Implementation

[0011] 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 a part of the embodiments of the present invention, and not all of them. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.

[0012] Many specific details are set forth in the following description in order to provide a full understanding of the invention. However, the invention may also be practiced in other ways different from those described herein, and those skilled in the art can make similar extensions without departing from the spirit of the invention. Therefore, the invention is not limited to the specific embodiments disclosed below.

[0013] like Figure 1 As shown in the figure, this invention discloses a highly stable optical parametric oscillator assisted by an intelligent feedback algorithm, including a pump laser source 1. The green light 17 emitted from the pump laser source 1 is first split by a half-wave plate 2 and a polarization beam splitter 3, and then reflected by a plane mirror 4. The pump laser source 1 described in this application can be an ultrashort pulse laser source with a center wavelength of 515 nm. The 515 nm pump laser source emits 515 nm green light. It should be noted that the oscillator is also applicable to optical parametric oscillator systems with output wavelengths of other wavelengths, and its stabilization method is not limited to the specific wavelength example described. After passing through the first dichroic mirror 5, the light enters the resonant cavity of the optical parametric oscillator. The 515 nm green light 17 is reflected and focused by the first concave focusing cavity mirror 6 and enters the nonlinear crystal 7 to generate signal light 18 and idler light 19. The signal light 18 and idler light 19 are reflected and collimated at the second concave focusing cavity mirror 8. The idler light 19 is transmitted away from the optical parametric oscillator system after passing through the second dichroic mirror 9. The signal light 18 is reflected and enters the output coupling mirror 10. The output signal light enters the beam splitter 12. The remaining energy is reflected and continues to oscillate in the resonant cavity.

[0014] The reflected signal light 18 returns to the resonant cavity, is reflected by the first dichroic mirror 5, and then enters the plane mirror 11 before returning to the cavity and synchronously meeting the newly incident pump light 17. It then re-enters the nonlinear crystal 7 and oscillates repeatedly within the cavity. The transmitted signal light, after passing through the beam splitter 12, continues to transmit most of its energy, while a very small portion is collected by the four-quadrant photodetector 13 and conditioned by a preamplifier before being input to the STM32 microcontroller 14. The collected digital signal undergoes iterative processing using the Stochastic Parallel Gradient Descent (SPGD) algorithm in the interrupt service routine. It should be noted that although this embodiment uses the SPGD algorithm as an example, those skilled in the art should understand that any intelligent auxiliary control algorithm capable of real-time optimization based on system performance indicators falls within the scope of this invention. The intelligent auxiliary control system algorithm includes a series of intelligent control algorithms capable of autonomously optimizing system performance, such as particle swarm optimization, model predictive control, and reinforcement learning algorithms. The signal is then transmitted to the DAC high-voltage driver board 15 via the SPI communication interface to generate a high-voltage analog drive signal. Finally, it acts on the piezoelectric deflector 16 in a closed-loop feedback manner to achieve stable control of the laser output power.

[0015] Inside the STM32 microcontroller 14, the ADC digitizes the analog signal at a high sampling rate and stores the conversion result in a designated memory area in real time via DMA1. The processing of the digital signal is completed in a specially configured interrupt service routine: the system uses the half-transfer and full-transfer interrupts of DMA1 to trigger the control flow, in which the stochastic parallel gradient descent (SPGD) algorithm is executed. This algorithm is used to calculate the power deviation in real time and generate a set of optimized control variables in parallel iterations.

[0016] The iteratively obtained digital control signal is transmitted to the SPI buffer via DMA2, and the high-speed SPI communication interface transmits the data to the DAC high-voltage driver board 15. This board integrates a multi-channel DAC chip (DAC81416) and a high-voltage operational amplifier, converting the digital signal into an analog voltage and amplifying it to a 100-volt high-voltage output to drive the piezoelectric deflector. The piezoelectric deflector makes fine angle adjustments based on the applied high-voltage signal, thereby adjusting the cavity length and beam direction in real time. This ultimately forms a closed-loop feedback control system with high response speed and strong anti-interference capability, achieving precise and stable control of the laser output power.

[0017] In this application, preferably, the photoelectric sensor chip used in the four-quadrant photodetector 13 is an S5980, with a spectral response range of 320 nm to 1100 nm and a photosensitive surface divided into four quadrants: A, B, C, and D. The chip converts the received optical power into a current signal, which is then conditioned by the integrated amplifier circuit inside the four-quadrant photodetector 13 and converted into a 0-3.3 V voltage signal before being output to the STM32 microcontroller.

[0018] The STM32 microcontroller's input I / O ports are configured as differential input mode ports for ADC1 and ADC2, and a single-ended input mode port for ADC3. The execution of the SPGD algorithm is placed within the DMA transfer completion interrupt and half-transfer interrupt service functions inside the STM32 microcontroller 14. The advantage of this double-buffering mechanism is that it enables parallel execution of data acquisition and algorithm processing, greatly reducing data processing latency, avoiding memory access conflicts, and ensuring that the system still possesses superior real-time response capabilities under high-speed sampling conditions.

[0019] Furthermore, in the STM32 microcontroller, ADC1 and ADC2 operate in differential input mode and sample synchronously to acquire the conditioned beam position signal (X and Y directions); ADC3 operates in single-ended input mode to acquire the total power signal. The beam position information is used to determine the beam alignment status. When the beam is within the detector's effective area, the SPGD algorithm is activated, using the total power sampled by ADC3 as feedback to control the output voltage.

[0020] The photoelectric sensor chip converts the collected optical power into four current signals. These four signals are then converted into voltage signals by an inverting transimpedance amplifier composed of four operational amplifiers. The voltage signals are then processed in two paths: one path groups the four voltage signals in pairs, combines them using two inverting adder amplifiers, and inputs them to the differential inputs of STM32's ADC1 and ADC2 to calculate the positional offset of the light spot in the X and Y directions; the other path combines the four voltage signals using an inverting adder amplifier, and its output is connected to the single-ended input port of STM32's ADC3 to acquire and monitor the total output power of the laser.

[0021] ADC sampling data is transferred via DMA1, where channel 0 is used to transfer 16-bit data from ADC3, and channel 1 is used to transfer 32-bit data in ADC1 and ADC2 synchronous mode (the high 16 bits are ADC2 data, and the low 16 bits are ADC1 data). Data is written to a 32-bit memory array with 64 elements, although 63 units are actually used. This design aims to meet the DMA transfer requirement that the buffer size be a power of 2, thereby improving access efficiency and system stability.

[0022] The DAC high-voltage driver board 15 includes a DAC81416 digital-to-analog converter chip and subsequent voltage amplification circuitry. The DAC81416 has 16 output channels with an output voltage range of 0 to +20 V. The amplification circuitry is set with a voltage gain of 6, and can output a maximum voltage of ±120 V. The DAC high-voltage driver board adopts a three-channel output architecture: one 120 V high-voltage bias reference signal to establish a linear operating range for the piezoelectric ceramic actuator; and two independent 0-120 V differential control signals as common-mode voltage references for the drive bridge, corresponding to two orthogonal deflection axes respectively. By precisely modulating the midpoint potential of the piezoelectric ceramic pair, a push-pull drive is achieved for the piezoelectric ceramic actuator, causing one unit to extend while the other unit retracts, thereby precisely controlling the mirror deflection.

[0023] In this application, the piezoelectric deflector 16 includes four piezoelectric ceramics. The driving voltage range of the piezoelectric ceramics is 0-120 V. Two-dimensional θx and θy deflection is achieved by differentially controlling the three voltages. The maximum deflection range is 3 mrad.

[0024] Working principle:

[0025] The photoelectric sensor chip converts the received optical power signal into four current signals, which are then converted into voltage signals by an inverting transimpedance amplifier composed of four precision operational amplifiers. These voltage signals are then processed in two paths. One path groups the four signals into pairs, combines them through two inverting adder amplifiers, and inputs them to the ADC1 and ADC2 ports of the STM32 microcontroller. These two ADCs are configured in differential input mode and operate in synchronous sampling mode to calculate the position offset of the light spot in the X and Y directions in real time. This position information is used to monitor the beam alignment status and trigger the SPGD algorithm when the light spot is within the effective area of ​​the detector. The other path combines the four voltage signals through an inverting adder amplifier and connects them to the single-ended input port of ADC3 to acquire the total laser power signal, which serves as the feedback control quantity for the SPGD algorithm.

[0026] The positions x and y of the light spot can be obtained from the output signal sampled by the four-quadrant photodetector using the following two formulas:

[0027]

[0028]

[0029] ADC sampling data is transferred via DMA1, where channel 0 transmits 16-bit data from ADC3, and channel 1 transmits 32-bit combined data obtained from ADC1 and ADC2 in synchronous mode (the high 16 bits are ADC2 data, and the low 16 bits are ADC1 data). The data is written to a 32-bit memory array with a capacity of 64. The DMA transfer employs a double-buffered mechanism of full transfer and half transfer. During the transfer interruption phase, the power value used for calculation is the average of the 21 power samples that have already been transferred.

[0030] The iterative process of the SPGD algorithm is as follows: First, during the first half-transmission interruption, a set of random unidirectional perturbations is simultaneously injected into both control channels of the piezoelectric deflector. Then, a full transmission interruption is used to collect and update the system's response data. Next, during the second half-transmission interruption, a perturbation in the opposite direction to the first is applied, and the data is updated again using a full transmission interruption. Based on the results of these two samplings, during the third half-transmission interruption, the control voltage of each channel is calculated and updated according to the SPGD control law. Finally, in the subsequent full transmission interruption, by comparing the current optical power with the initial value, it is determined whether the iteration has achieved a performance improvement, thus completing one complete optimization cycle. The algorithm's iteration frequency is approximately 6.17 kHz.

[0031] The Stochastic Parallel Gradient Descent (SPGD) algorithm used in the feedback system is an intelligent feedback iterative algorithm. One of its advantages is that it is independent of the system model, so it can be used without knowing the explicit mapping between the system's inputs and outputs. Figure 2 The specific implementation framework of the SPGD algorithm is shown. Here, its iterative control flow is briefly introduced using a single control variable. The algorithm uses a two-way perturbation approach to iteratively optimize the voltage control vector u based on the system evaluation function power P. The update formula for the control vector u is:

[0032]

[0033] Where δu is the perturbation vector and μ is the learning rate, the system evaluation function P is optimized by stochastic gradient descent, thereby improving the system power stability and anti-interference capability.

[0034] The DAC high-voltage driver board integrates a USB-to-serial communication module based on the FT232RL chip. This module establishes a high-efficiency data uplink channel, enabling critical data processed internally by the STM32 microcontroller (including real-time input signals from each channel and control outputs after algorithmic calculations) to be uploaded to the host computer via a serial port tool in digital stream form. Operators can observe, record, and analyze this data in real time, providing intuitive data for dynamic debugging of the system's closed-loop control.

[0035] The optical parametric oscillator described in this invention overcomes the bottleneck of insufficient power stability of existing near-infrared optical parametric oscillators during long-term operation, and provides a new light source solution with higher output power and better stability for fields such as spectral analysis, micro-nano fabrication, and optical frequency combs that rely on broadband tunable light sources.

Claims

1. A highly stable optical parametric oscillator assisted by an intelligent feedback algorithm, characterized in that: The system includes a pump laser source (1). The green light (17) emitted from the pump laser source is first split by a half-wave plate (2) and a polarization beam splitter (3), then reflected by a plane mirror (4), transmitted through a first dichroic mirror (5), and enters the resonant cavity of an optical parametric oscillator. The green light (17) is reflected and focused by a first concave focusing cavity mirror (6) and injected into a nonlinear crystal (7) to generate signal light (18) and idler light (19). The signal light (18) and idler light (19) are reflected and collimated at a second concave focusing cavity mirror (8). The idler light (19) is transmitted away from the resonant cavity of the optical parametric oscillator after passing through a second dichroic mirror (9). The signal light (18) is reflected and incident on an output coupling mirror (10). Part of the signal light is output and transmitted to a beam splitter (12). The remaining part of the signal light is reflected and then... The signal light (18) continues to oscillate within the resonant cavity. After returning to the resonant cavity, the reflected signal light (18) is reflected by the first dichroic mirror (5) and incident on the plane mirror (11). It then returns to the resonant cavity and synchronously meets the newly incident pump light (17), and is incident on the nonlinear crystal (7) again, thus oscillating repeatedly within the cavity. After the transmitted signal light passes through the beam splitter (12), part of its energy is collected by the four-quadrant photodetector (13), and after being conditioned by the preamplifier, it is connected to the microcontroller (14). The microcontroller (14) processes the collected digital signal and then transmits it to the DAC high-voltage driver board (15) through the SPI communication interface to generate a high-voltage analog drive signal. Finally, it acts on the piezoelectric deflector (16) in the form of closed-loop feedback to achieve stable control of the output power of the signal light.

2. The high-stability optical parametric oscillator assisted by the intelligent feedback algorithm as described in claim 1, characterized in that: The pump laser source is a 515 nm pump laser source (1), and the emitted green light is 515 nm green light (17).

3. The high-stability optical parametric oscillator assisted by the intelligent feedback algorithm as described in claim 1, characterized in that: The four-quadrant photodetector (13) includes an S5980 photoelectric sensor chip with a spectral response range of 320 nm to 1100 nm and a photosensitive surface divided into four quadrants: A, B, C, and D. The chip converts the received light power into a current signal, which is then conditioned by the integrated amplifier circuit inside the four-quadrant photodetector (13) and converted into a 0-3.3 V voltage signal before being output to the STM32 microcontroller (14).

4. The highly stable optical parametric oscillator assisted by the intelligent feedback algorithm as described in claim 3, characterized in that: The microcontroller uses an STM32 microcontroller (14), whose input I / O ports are configured as differential input mode ports for ADC1 and ADC2, and single-ended input mode ports for ADC3.

5. The high-stability optical parametric oscillator assisted by the intelligent feedback algorithm as described in claim 1, characterized in that: The microcontroller (14) performs stochastic parallel gradient descent (SPGD) algorithm iterative processing on the acquired digital signals in the interrupt service routine to obtain control signals.

6. The high-stability optical parametric oscillator assisted by the intelligent feedback algorithm as described in claim 4, characterized in that: The photoelectric sensor chip converts the received optical power signal into four current signals for output. These current signals are then converted into voltage signals by an inverting transimpedance amplifier composed of four precision operational amplifiers. The voltage signals are then processed in two paths. One path groups the four signals into pairs, combines them through two inverting adder amplifiers, and inputs them to the ADC1 and ADC2 ports of the STM32 microcontroller. These two ADCs are configured in differential input mode and operate in synchronous sampling mode to calculate the position offset of the light spot in the X and Y directions in real time. The position offset information is used to monitor the beam alignment status and trigger the SPGD algorithm when the light spot is within the effective area of ​​the detector. The other path combines the four voltage signals through an inverting adder amplifier and connects them to the single-ended input port of ADC3 to acquire the total laser power signal, which serves as the feedback control quantity for the SPGD algorithm.

7. The high-stability optical parametric oscillator assisted by the intelligent feedback algorithm as described in claim 5, characterized in that: The SPGD algorithm is placed in the DMA transfer completion interrupt and half-transfer interrupt service functions inside the STM32 microcontroller (14).

8. The high-stability optical parametric oscillator assisted by the intelligent feedback algorithm as described in claim 5, characterized in that: The SPGD algorithm uses a two-way perturbation approach to iteratively optimize the voltage control vector u based on the system evaluation function power P. The update formula for the control vector u is: ； Where δu is the perturbation vector, μ is the learning rate, and P is the power of the system evaluation function.

9. The high-stability optical parametric oscillator assisted by the intelligent feedback algorithm as described in claim 1, characterized in that: The DAC high-voltage driver board (15) includes a DAC81416 digital-to-analog converter chip and a voltage amplifier circuit; the DAC81416 digital-to-analog converter chip has 16 output channels with an output voltage range of 0 to +20 V; the voltage amplifier circuit is set with a voltage gain of 6 times and can output a maximum voltage of ±120 V.

10. The high-stability optical parametric oscillator assisted by the intelligent feedback algorithm as described in claim 1, characterized in that: The piezoelectric deflector (16) contains four piezoelectric ceramics. The driving voltage range of the piezoelectric ceramics is 0-120 V. Two-dimensional θx and θy deflection is achieved by differentially controlling the three voltages. The maximum deflection range is 3 mrad.

Images