Double-feedback control debugging method of ultra-precision servo system

Abstract

The invention discloses a double-feedback control debugging method of an ultra-precision servo system, and relates to the technical field of semiconductor manufacturing equipment. In order to solve the defects that in the prior art, coupling interference is serious in the debugging process, frequency domain performance optimization is insufficient, filter configuration does not have adaptivity, and an efficient anti-interference debugging mechanism under multi-source disturbance is lacked, the technical scheme provided by the invention comprises the steps that a double-feedback servo system comprising a speed ring and a position ring is constructed; debugging a speed feedback loop, optimizing controller gain parameters based on frequency domain analysis, and outputting speed loop response characteristics; a position feedback loop is debugged, and position loop controller parameters are configured in combination with speed loop response and acceleration sensor signals; the notch filter is automatically adjusted according to the dynamic vibration characteristics; and frequency domain self-adaptive debugging of the double-feedback controller is completed. The method is suitable for servo control debugging work of high-precision motion systems such as a wafer carrying table in semiconductor manufacturing equipment, and is especially suitable for nanoscale positioning scenes such as a photoetching machine and measuring equipment.

Description

Technical Field

[0001] This invention relates to the field of semiconductor manufacturing equipment technology, specifically to a debugging method for a dual closed-loop servo system used for precision motion control of wafers. Background Technology

[0002] With the continuous development of semiconductor manufacturing technology, high-end equipment such as lithography machines and metrology equipment are placing more stringent demands on the control precision and response speed of motion platforms. This is especially true in nanometer-level positioning and high-frequency dynamic response scenarios, where the performance requirements for servo systems (such as bandwidth, noise immunity, and stability margin) are significantly increased. For high-precision motion mechanisms such as wafer stages, the industry typically employs a dual-feedback closed-loop control structure. This involves using different types of sensors for feedback in the velocity and position loops: the velocity loop uses a grating ruler-reading head system to achieve high-resolution velocity measurement, while the position loop obtains nanometer-level position feedback through a mirror-laser interferometer system to enhance the overall positioning performance of the system.

[0003] In current industrial practice, the debugging methods for dual-feedback servo control systems mostly employ empirical rules or time-domain response analysis (such as the step response method) to perform PI parameter tuning and filter design. For example, some manufacturers repeatedly adjust PID parameters by preset position / speed commands and observing the time-domain error curve; existing technologies have proposed adding notch filters to the speed or position loops to suppress certain mechanical resonance frequencies, but these methods still have significant limitations in terms of parameter tuning efficiency, frequency domain performance optimization, and disturbance adaptability.

[0004] On the one hand, there is coupling interference between the velocity loop and the position loop, and traditional methods cannot achieve independent tuning of the two, resulting in low system debugging efficiency and limited performance optimization effects. On the other hand, time-domain methods cannot directly reflect the frequency domain characteristics of the system, such as bandwidth and phase margin, often resulting in insufficient system bandwidth or insufficient robustness. At the same time, in actual operation, the servo system also faces multi-source disturbances, including load changes, friction fluctuations, structural resonance, and environmental vibration. If these disturbances are not suppressed in time, they will seriously affect the stability and accuracy of the motion platform.

[0005] Furthermore, existing technologies generally lack system identification mechanisms for the system's natural frequencies and real-time analysis methods for residual vibrations, and have failed to establish a direct coupling relationship between accelerometers and frequency responses, making it difficult to achieve adaptive adjustment of filter parameters based on dynamic vibration characteristics.

[0006] In summary, existing technologies suffer from drawbacks such as severe coupling interference during the debugging process, insufficient frequency domain performance optimization, lack of adaptive filter configuration, and absence of an efficient anti-interference debugging mechanism under multi-source disturbances. Summary of the Invention

[0007] To address the shortcomings of existing technologies, such as severe coupling interference during the debugging process, insufficient frequency domain performance optimization, lack of adaptive filter configuration, and absence of an efficient anti-interference debugging mechanism under multi-source disturbances, the technical solution provided by this invention is as follows: A dual-feedback control debugging method for an ultra-precision servo system includes: Steps for constructing a dual-feedback servo system that includes a velocity loop and a position loop; The steps for debugging the speed feedback loop, optimizing the controller gain parameters based on frequency domain analysis, and outputting the speed loop response characteristics; The steps for debugging the position feedback loop and configuring the position loop controller parameters by combining the velocity loop response and acceleration sensor signals; The steps involve generating filter parameters and automatically adjusting the notch filter based on dynamic vibration characteristics. Steps to complete the frequency domain adaptive tuning of the dual feedback controller.

[0008] Furthermore, in a preferred embodiment, the step of debugging the speed feedback loop includes: obtaining the platform's frequency domain response through frequency sweep signal excitation, and dynamically configuring the speed loop controller's bandwidth and phase margin based on the response data.

[0009] Furthermore, in a preferred embodiment, the step of debugging the position feedback loop includes: after the velocity feedback loop is debugged, using the frequency domain characteristics of the velocity loop as a constraint, fusing the position sensor output and residual vibration index to achieve position controller gain self-tuning.

[0010] Furthermore, in a preferred embodiment, the accelerometer is used to collect the micro-vibration response signal of the target platform during operation, extract the main mode frequency and its energy characteristics, and use them as the basis for subsequent filter parameter adjustment.

[0011] Furthermore, in a preferred embodiment, the parameter generation of the notch filter includes: adaptively setting the filter center frequency and damping ratio based on the main vibration components identified in the acceleration signal spectrum.

[0012] Furthermore, in a preferred embodiment, the debugging processes of the speed feedback loop and the position feedback loop are independent of each other but have coupled feedback logic, which is transmitted through intermediate variables to achieve frequency domain collaborative optimization of the dual closed-loop controller.

[0013] A dual feedback control debugging device for an ultra-precision servo system includes: A module for constructing a dual-feedback servo system that includes a velocity loop and a position loop; A module for debugging the speed feedback loop, optimizing controller gain parameters based on frequency domain analysis, and outputting the speed loop response characteristics; A module for debugging the position feedback loop and configuring the position loop controller parameters by combining the velocity loop response and acceleration sensor signals; A module that generates filter parameters and automatically adjusts the notch filter based on dynamic vibration characteristics; A module for completing the frequency domain adaptive tuning of the dual feedback controller.

[0014] The system also provides a computer storage medium for storing a computer program, which, when read by the computer, executes the method.

[0015] A computer is also provided, including a processor and a storage medium, wherein the computer executes the method when the processor reads a computer program stored in the storage medium.

[0016] A computer program product is also provided, which, when executed, implements the method.

[0017] Compared with the prior art, the advantages of the technical solution provided by the present invention are as follows: This scheme effectively avoids coupling interference between the dual feedback loops by introducing independent frequency domain tuning methods for the velocity and position loops. This transforms the debugging process from traditional trial-and-error tuning to a structured, repeatable frequency domain optimization process. Compared to existing technologies where the velocity and position loops interfere with each other and require multiple rounds of testing to achieve optimal parameter combinations, this scheme, by disconnecting the feedback links and tuning separately, combined with frequency domain excitation signal injection and Bode plot analysis, can quickly obtain high phase margin and bandwidth indicators, achieving a dual improvement in control accuracy and stability.

[0018] This scheme introduces a swept-frequency excitation and frequency response function (FRF) modeling method during the speed loop commissioning phase, combined with static natural frequency identification, to form a complete spatial resonance distribution information set. By analyzing the Bode plot characteristics between excitation-response pairs, the resonant frequencies that have the greatest impact on system performance are effectively identified and separated. Compared to the problem of traditional commissioning methods being unable to quantitatively analyze the system's resonant structure, this approach achieves precise positioning of the target modal frequencies, providing a clear basis for filter configuration.

[0019] This solution utilizes an accelerometer for dynamic residual acceleration analysis, enabling real-time monitoring and feedback of resonant energy during actual operation. When the vibration amplitude at a specific frequency continuously exceeds a threshold, the center frequency of the notch filter is automatically adjusted to precisely suppress interference at that frequency, thus constructing a closed-loop optimization mechanism of "detection-response-compensation." Compared to existing technologies that require static filter parameter configuration and exhibit slow response to changes in operating conditions, this method significantly improves the system's adaptability to dynamic disturbances and its control robustness.

[0020] This scheme employs a multi-filter collaborative configuration mechanism. A low-pass filter is introduced into the velocity loop to reduce high-frequency noise interference to the system; a notch filter suppresses specific mechanical structure resonant frequencies; and lead-lag compensation is used to improve dynamic tracking capability and low-speed steady-state error control, respectively. Unlike existing technologies where multiple filter parameters require empirical adjustment and it is difficult to ensure a balance between bandwidth and phase margin, this scheme guides filter combination optimization through frequency domain indicators, enabling the system to possess both fast response capability and good stability margin.

[0021] This scheme incorporates sweep speed commands and sensitivity function analysis during position loop tuning. By controlling the slope of the logarithmic amplitude-frequency curve to be close to -20 dB / dec, precise proportional gain setting is achieved. Furthermore, by combining structural mode information fused in the position feedback, the design of the lead-lag filter is guided, enabling the system to maintain sub-nanometer positioning accuracy during low-speed micro-motion stages while maintaining high dynamic tracking capability at high speeds. Compared to existing tuning schemes, this approach can better accommodate controller performance under different operating conditions and adapt to complex multi-tasking applications.

[0022] Through a dual verification mechanism of static and dynamic performance testing, this solution can continuously monitor servo performance under different operating conditions, detect and quickly suppress residual vibration components, and improve the wafer platform's disturbance recovery time and steady-state error performance. Unlike traditional methods that require frequent shutdowns for troubleshooting, this testing mechanism is tightly integrated with the debugging method, significantly improving the overall reliability and engineering maintainability of the servo system.

[0023] It is suitable for servo control debugging of high-precision motion systems such as wafer stages in semiconductor manufacturing equipment, and is especially suitable for nanometer-level positioning scenarios such as lithography machines and measurement equipment. Attached Figure Description

[0024] Figure 1 Flowchart for Static Natural Frequency Identification Figure 2 Flowchart for dynamic acceleration residual analysis; Figure 3 A flowchart illustrating a dual feedback control debugging method for an ultra-precision servo system; Figure 4 Bode plot for the location loop of traditional debugging methods; Figure 5 A speed loop Bode plot is used for debugging a dual feedback control system for an ultra-precision servo system. Figure 6 A position loop Bode plot is used to demonstrate a dual feedback control debugging method for an ultra-precision servo system. Figure 7 Time-domain step plot for debugging traditional methods; Figure 8This is a time-domain step diagram for a dual-feedback control debugging method of an ultra-precision servo system. Detailed Implementation

[0025] To make the advantages and benefits of the technical solution provided by the present invention clearer, the technical solution provided by the present invention will now be described in further detail with reference to the accompanying drawings, specifically: Implementation Method 1: This implementation method provides a dual feedback control debugging method for an ultra-precision servo system, including: Steps for constructing a dual-feedback servo system that includes a velocity loop and a position loop; The steps for debugging the speed feedback loop, optimizing the controller gain parameters based on frequency domain analysis, and outputting the speed loop response characteristics; The steps for debugging the position feedback loop and configuring the position loop controller parameters by combining the velocity loop response and acceleration sensor signals; The steps involve generating filter parameters and automatically adjusting the notch filter based on dynamic vibration characteristics. Steps to complete the frequency domain adaptive tuning of the dual feedback controller.

[0026] The steps for debugging the speed feedback loop include: obtaining the platform's frequency domain response through frequency sweep signal excitation, and dynamically configuring the bandwidth and phase margin of the speed loop controller based on the response data.

[0027] The steps for debugging the position feedback loop include: after the velocity feedback loop is debugged, the position controller gain is self-tuned by fusing the position sensor output and residual vibration index with the velocity loop frequency domain characteristics as constraints.

[0028] Accelerometers are used to collect micro-vibration response signals of the target platform during operation, extract the main mode frequency and its energy characteristics, and use them as the basis for subsequent filter parameter adjustment.

[0029] The parameters of the notch filter are generated by adaptively setting the filter center frequency and damping ratio based on the main vibration components identified in the acceleration signal spectrum.

[0030] The debugging processes of the speed feedback loop and the position feedback loop are independent of each other but have coupled feedback logic. The logic is passed through intermediate variables to achieve frequency domain collaborative optimization of the dual closed-loop controller.

[0031] Implementation Method 2, see below Figure 1-3 This embodiment further describes the technical solution provided in Embodiment 1 in more detail, specifically: Based on a dual-closed-loop servo architecture, a step-by-step decoupling and frequency domain optimization strategy is adopted, specifically including the following steps: Step 1: Construct a dual feedback control architecture In the target ultra-precision servo system, a dual closed-loop control structure comprising a speed feedback loop and a position feedback loop is constructed. The speed feedback loop is based on current control, superimposed with a speed feedback controller; the position feedback loop constructs a position feedback path on the outer layer and integrates a position controller. The two feedback loops can be modeled separately in terms of physical structure and control model, enabling decoupled analysis and independent debugging.

[0032] Step 2: Establish debugging interface and sensor acquisition mechanism After the controller structure is configured, an interface module for frequency domain debugging is configured, and a high-precision position encoder and a MEMS accelerometer are integrated into the platform motion unit. The position encoder is used to acquire nanometer-level displacement changes, and the accelerometer is used to acquire the micro-vibration response of the structure. During servo system operation, a test signal injection channel is set up to apply frequency sweep excitation to the system.

[0033] Step 3: Perform frequency domain identification and modeling Under test excitation, system response data is collected and frequency domain transformation is performed to obtain amplitude and phase frequency response curves. For the velocity feedback loop, the dynamic characteristics of the system are extracted by analyzing the amplitude decay trend, phase lag, and resonance abrupt change point. For acceleration data, fast Fourier transform is performed to identify frequency energy distribution, principal mode frequency position, and harmonic response behavior.

[0034] The outputs of this stage include: velocity loop frequency response model, residual vibration spectrum, system bandwidth and margin parameters.

[0035] Step 4: Optimize the speed feedback controller Based on the frequency domain model of the velocity loop, adjust the feedforward gain, scaling factor, and low-pass compensation coefficient of its controller to ensure the system has the target bandwidth, desired phase margin, and suppresses mid-frequency disturbance response. During commissioning, the controller response strength can be gradually increased, and the system should be monitored for any new structural resonance issues.

[0036] After optimization, the system obtains a basic and stable dynamic response characteristic, and its frequency domain data is output for subsequent filtering and position controller configuration.

[0037] Step 5: Perform residual vibration frequency band analysis After the velocity loop configuration is stabilized, acceleration signals are collected during system operation. Modal energy analysis is used to extract characteristic indicators such as the main residual frequencies, peak energy, and energy sustaining time. This analysis is used to determine the structural coupling dynamic behavior of the platform under unsteady-state response, such as platform body structural resonance and slide rail vibration feedback.

[0038] The identification results are output as multiple high-energy main frequency points and their bandwidth ranges.

[0039] Step Six: Adaptively Configure Notch Filter Parameters Based on the identified dominant mode frequency, multiple notch filters are constructed, and parameters such as center frequency, bandwidth, and quality factor are set for each frequency point to automatically generate the filter transfer function. The filters are embedded in the velocity feedback path to suppress the disturbance effect of the resonant frequency band in real time, and the parameters are dynamically refreshed according to the residual vibration changes during the operation of the servo system, forming a "spectrum-control" closed-loop vibration suppression system.

[0040] This step significantly reduces human intervention and improves the adaptive capability of filter configuration.

[0041] Step 7: Debug the position feedback controller After the velocity loop frequency domain optimization and filter configuration are completed, the debugging phase of the position control loop begins. Combining the velocity loop frequency response characteristics and the filtered feedback signal, the proportional, integral, and feedforward parameters of the position controller are configured to ensure that the system maintains a fast response while suppressing tracking errors and overshoot. A step-by-step iterative approach can be used to improve the system's stability margin, and the position tracking accuracy under various operating conditions can be tested.

[0042] The final control parameters should meet the following requirements: fast tracking, zero steady-state error, and no closed-loop resonance peak.

[0043] Step 8: Complete the system debugging parameter fixing and deployment The control parameters and filter configuration parameters of the speed loop and position loop obtained from the above debugging are solidified and uploaded to the controller firmware configuration area. Subsequently, during the mass production or operation and maintenance phase of the platform, this parameter group can be quickly reused to achieve batch debugging and deployment. If the platform structure is slightly different, this method can be reused for rapid parameter tuning, which has good system adaptability.

[0044] The ultra-precision servo system debugged using the method provided in this embodiment has the following performance advantages: Frequency domain decoupling optimization of the position control loop and speed control loop was achieved, avoiding debugging coupling interference; Automatically identify the resonant frequency band and configure a notch filter to effectively suppress residual vibration during platform operation; This improves the bandwidth and phase margin of the servo system, enhancing dynamic response performance. It reduces reliance on the experience of debugging personnel and improves debugging efficiency and batch consistency; It is suitable for various precision platform controller architectures and has cross-platform capabilities.

[0045] Implementation Method 3: Combination Figure 1-8 This embodiment describes the technical solution provided above in further detail through specific examples. Specifically: A dual-feedback independent tuning method based on frequency domain analysis is proposed. This method combines the vibration spectrum feature learning function measured by an external independent accelerometer. By adjusting the PID parameters of the servo loop and various filters, and integrating static natural frequency identification and dynamic acceleration residual analysis, frequency response analysis and optimization are performed sequentially and distributed from the velocity loop to the position loop. This achieves efficient decoupling and tuning of the position loop (laser ruler) and the velocity loop (grating ruler), improves system bandwidth and anti-interference capability, and solves the problem of multi-source disturbance suppression that is difficult to overcome by traditional methods.

[0046] 1. The system architecture is as follows: 1) Feedback module: Speed ​​loop: Reading head – grating ruler system, provides commutation for the servo system, increasing the stability of the entire system.

[0047] Positioning ring: The laser ruler-reflector system can more accurately reflect the positional changes of the load end (wafer), improving the positioning accuracy and repeatability of the motion system.

[0048] Accelerometer sensor: Five accelerometers (range: ±0.055g, sensitivity: 40V / g).

[0049] 2) Control module: Dual-core Arm processor, capable of independently running the speed loop control algorithm and the position loop control algorithm; It supports multiple excitation signal injections in the frequency domain and Bode plot analysis.

[0050] 2. Independent tuning of the speed loop 1) System Modeling and Open-Loop Transfer Function The feedback signal of the reading head-grating ruler system serves as the input to the speed loop, and the motor torque command serves as the output. Its open-loop transfer function can be expressed as follows:

[0051] in: , For the velocity loop PI parameters; For the load's rotational inertia, It is a viscous damping system; This refers to system delay (including signal processing and actuator delay).

[0052] 2) Incentive Injection and Data Acquisition First, disconnect the position loop feedback to avoid double-coupling interference and ensure the independence of the speed loop tuning. Then, inject a sweep frequency current signal with gradually varying amplitude. The frequency range is 1-3000 Hz, and the amplitude A is dynamically adjusted according to the motor's rated current (3% of the rated value is recommended). The excitation system has a full-frequency response. The speed signal output from the reading head – grating ruler system is recorded synchronously. The open-loop transfer function is calculated using the FRF (Frequency Response Function):

[0053] Simultaneously, signals are acquired from five accelerometers distributed at the four corners and center of the motion platform (sampling rate ≥ 10 kHz). Vibration data collection is required under two working conditions: static motion platform (static natural frequency identification); and actual working condition motion (dynamic residual feedback mechanism of vibration).

[0054] 3) Bode plot analysis and parameter optimization First, perform initial parameter tuning, estimating the PI parameter using empirical formulas: ,

[0055] in The target bandwidth (e.g., 100 Hz). = 0.707 is the damping ratio. , This represents the load inertia.

[0056] With the motion platform stationary, a spatial resonance distribution dataset is constructed by combining vibration data of various parts of the platform measured by five acceleration sensors. The third-order key resonant frequencies and their spatial distribution weight coefficients are then extracted using modal analysis algorithms.

[0057] When the motion platform is running along the target trajectory under actual working conditions, the acceleration sensor signal is collected in real time, and the residual vibration energy distribution is extracted after the signal-to-noise ratio is enhanced.

[0058] Then, various filters are applied for optimization: Low-pass filters effectively suppress high-frequency noise, such as switching noise from the motor driver (e.g., PWM carrier frequency) or high-frequency interference introduced by high-frequency mechanical vibrations (e.g., bearing friction noise) in the speed feedback signal. Low-pass filters remove these unwanted signals through a cutoff frequency, ensuring the stability of speed control. The cutoff frequency is set to three times the target bandwidth. Furthermore, the rapid response of the speed loop can cause phase lag in the high-frequency band; low-pass filters reduce high-frequency gain, preventing self-oscillation of the system under high-frequency disturbances.

[0059]

[0060] Notch filters effectively suppress mechanical resonance in ultra-precision servo systems. The mechanical transmission structure between the motor and the load (such as couplings and lead screws) may have resonance points at specific frequencies. Notch filters deeply attenuate these frequencies, preventing the accumulation of resonance energy that could lead to servo system malfunction. Furthermore, if periodic interference exists in the speed feedback signal (such as periodic errors caused by encoder misalignment), notch filters can precisely filter out these fixed-frequency interference components.

[0061] By employing an improved balanced reduction method, the first three main modes are retained during the static natural frequency identification process to generate coarse notch filter parameters. During dynamic vibration residual feedback analysis, when the frequency energy exceeds the threshold three times consecutively, the center frequency of the notch filter is adaptively adjusted to obtain accurate notch filter parameters.

[0062]

[0063] in, , 1, 2 * 2000.

[0064] Lead-lag compensation: In the speed loop of an ultra-precision servo system, lead-lag compensation can improve dynamic response, which is the effect of lead compensation; it can also enhance low-frequency stability, an effect contributed by lag compensation. During high-speed scanning, lead compensation provides phase lead in the mid-to-high frequency range (e.g., 500–1000 Hz), accelerating the system's tracking speed command and reducing dynamic lag error. During low-speed precision adjustment, lag compensation increases low-frequency gain, suppresses low-frequency oscillations caused by integral saturation, and ensures steady-state speed accuracy. Overall, it involves adjusting the phase margin of the servo loop; the compensation formula is:

[0065] in Used for phase lead, Used for phase lag.

[0066] 4) Optimize objectives and validation The frequency domain tuning specifications are: gain margin ≥ 10 dB, phase margin ≥ 40°.

[0067] First of all Keep 0 unchanged, then gradually increase the value starting from 0. Simultaneously observe the changes in the amplitude curve within the amplitude-frequency curve, and stop adjusting when the slope of the amplitude curve near the 0 dB crossover frequency approaches -20 dB / dec. Start adjusting To improve low-frequency gain and suppress steady-state error, the three types of filters mentioned above are added during the adjustment process to reduce resonant peaks, avoid phase lag, and maximize system bandwidth. Note that the notch filter should be placed as far away as possible from the crossover frequency to avoid reducing system bandwidth and thus decreasing system margin.

[0068] 3. Independent tuning of position loop 1) Frequency response analysis First, a frequency sweep speed command is injected, with a frequency range of 1-3000 Hz. Unlike the current command injected via the speed loop, the frequency sweep speed command is closely related to the system's maximum speed, so the amplitude of the frequency sweep command must be dynamically adjusted according to the maximum speed. Then, the deviation between the actual position and the commanded position is obtained using a laser ruler, and the sensitivity function is calculated.

[0069] 2) Bode plot analysis and parameter tuning Proportional gain optimization: Adjust position loop proportional gain To avoid the occurrence of resonance peaks, the slope of the logarithmic amplitude-frequency curve in the logarithmic frequency response curve should be close to -20 dB / dec at the 0 dB crossover frequency.

[0070] Phase compensation design: The following filters can be added to the position loop: Notch filters primarily eliminate external interference. In semiconductor cleanrooms, vibrations from the external environment, such as the periodic vibrations of air conditioners or vacuum pumps, can be transmitted to the wafer stage through the internal structure of the equipment. Notch filters can effectively isolate such environmental interference. Furthermore, while vibrations caused by the inherent frequency of the wafer stage structure can directly affect the positioning control of the servo system through position feedback, notch filters have limited effectiveness in suppressing this vibration and require combination with other methods.

[0071] Lead-lag filter: Lead filters can optimize phase margin in the position loop. Near the crossover frequency of the position loop, lead compensation can improve phase margin, avoid resonance peaks caused by phase lag, and enhance system stability. Meanwhile, hysteresis filters can effectively suppress low-frequency errors. For steady-state errors in position control, such as minute positional deviations caused by guide rail friction, hysteresis compensation can improve low-frequency gain, reduce static errors, and ensure nanometer-level positioning accuracy.

[0072] The spatial resonance distribution dataset obtained through static natural frequency identification is fused with the position feedback signal to form a composite position feedback signal. This composite position feedback can be used to predict position change trends or as a feedforward signal to generate parameters such as natural frequency and damping ratio in a lead-lag filter.

[0073] 4. Performance Testing 1) Static performance testing Keep the motion stage stationary and close its position loop. Set the sampling frequency to 20 kHz and the sampling duration to 10 s to sample the position error of the motion stage. Observe the time-domain and frequency-domain representations of the position error, and record the peak-to-peak signal amplitude in the time domain and the frequency and amplitude of the most prominent vibration peaks in the frequency domain. If vibrations other than the natural frequency of the wafer stage appear in the frequency domain amplitude curve, the environmental vibration needs to be re-investigated or the filter needs to be modified to ensure a clean vibration spectrum. Then, record the peak-to-peak signal amplitude in the time domain again until the system specifications are met.

[0074] 2) Dynamic performance testing Based on the actual dynamic performance indicators of the system (speed fluctuations or rapid tuning capability during uniform motion), the motion table is tested under specified working conditions. The servo loop of the system can be tuned in a targeted manner based on the test results.

[0075] Compared with traditional methods, this approach has the following advantages: 1. Improved tuning efficiency Traditional debugging methods rely on trial and error based on experience, resulting in long debugging times and poor parameter stability. Even slight changes in load can lead to control instability and poor noise immunity. This solution, however, decouples the servo loop, allowing for separate tuning of the position and speed loops. This frequency-domain decoupling debugging method significantly shortens debugging time and ensures stable servo performance even when the load varies within a certain range, demonstrating strong noise immunity.

[0076] 2. Improved servo performance By using new debugging methods, servo loop bandwidth can be increased by 40%, wafer stage static jitter can be reduced by 50%, and disturbance rejection recovery time can be reduced by 50%.

[0077] In the above embodiments, the following tuning strategies can also be added to optimize the tuning method: 1. Dynamic parameter switching strategy The control parameter set is dynamically adjusted based on the motion speed (high-speed scanning / low-speed positioning). Acceleration feedforward is enabled in high-speed mode, and the position loop stiffness is enhanced in low-speed mode. The control mode is switched according to the wafer motion speed threshold, increasing the speed loop bandwidth in high-speed mode and enhancing the position loop proportional gain in low-speed mode.

[0078] 2. Hybrid Disturbance Observation and Compensation Techniques Luenberger-Kalman Hybrid Observer: A perturbation estimation method that combines model-driven and data-driven approaches.

[0079] Multi-source disturbance classification compensation: Generate targeted feedforward signals for disturbances such as friction, thermal drift, and mechanical resonance.

[0080] Integrating a Luenberger observer with a Kalman filter, it estimates unmodeled disturbances in real time and injects feedforward compensation.

[0081] During the trial phase, Figure 4 The Bode plot results obtained using traditional tuning methods suffer from drawbacks. In these methods, the velocity loop and position loop tuning are coupled, and there is no multi-mode filter coordinating architecture, resulting in a lengthy tuning time, typically around 4 hours. The tuning results are also unsatisfactory, with a bandwidth of 37.5 Hz, a gain margin of 2.8 dB, and a phase margin of 37°. Not only is the bandwidth low, but the frequency domain specifications are also not met.

[0082] Figure 5 , 6 These are the velocity loop Bode plot and the position loop Bode plot obtained by applying this invention. Due to the application of a dual-feedback independent frequency domain tuning scheme, the servo control loop is decoupled, thus generating two Bode plots. Furthermore, the use of a multi-mode filter collaborative architecture shortened the debugging time to one hour, and the debugging results met the frequency domain specifications: bandwidth of 60Hz, gain margin of 11 dB, and phase margin of 40.5°.

[0083] Figure 7 , Figure 8 These are step graphs before and after applying the tuning method of this invention. A step excitation is applied to the servo system to simulate its response capability under the most demanding operating conditions. The graphs show that after applying the tuning method of this invention, the overshoot decreases from 10% to 3%, the rise time decreases from 20ms to 10ms, and the settling time decreases from 50ms to 20ms. This demonstrates that the response speed and tuning performance of the servo system are significantly improved after applying the tuning method of this invention.

[0084] The above description of several specific embodiments further details the technical solution provided by the present invention in order to highlight the advantages and benefits of the technical solution provided by the present invention. However, the above-described specific embodiments are not intended to limit the present invention. Any reasonable modifications and improvements to the present invention, combinations of embodiments, and equivalent substitutions based on the spirit and principles of the present invention should be included within the protection scope of the present invention.

Claims

1. A dual-feedback control debugging method for an ultra-precision servo system, characterized in that, include: Steps for constructing a dual-feedback servo system that includes a velocity loop and a position loop; The steps for debugging the speed feedback loop, optimizing the controller gain parameters based on frequency domain analysis, and outputting the speed loop response characteristics; The steps for debugging the position feedback loop and configuring the position loop controller parameters by combining the velocity loop response and acceleration sensor signals; The steps involve generating filter parameters and automatically adjusting the notch filter based on dynamic vibration characteristics. Steps to complete the frequency domain adaptive tuning of the dual feedback controller.

2. The dual feedback control debugging method for an ultra-precision servo system according to claim 1, characterized in that, The steps for debugging the speed feedback loop include: obtaining the platform's frequency domain response through frequency sweep signal excitation, and dynamically configuring the bandwidth and phase margin of the speed loop controller based on the response data.

3. The dual feedback control debugging method for an ultra-precision servo system according to claim 1, characterized in that, The steps for debugging the position feedback loop include: after the velocity feedback loop is debugged, the position controller gain is self-tuned by fusing the position sensor output and residual vibration index with the velocity loop frequency domain characteristics as constraints.

4. The dual feedback control debugging method for an ultra-precision servo system according to claim 1, characterized in that, Accelerometers are used to collect micro-vibration response signals of the target platform during operation, extract the main mode frequency and its energy characteristics, and use them as the basis for subsequent filter parameter adjustment.

5. The dual feedback control debugging method for an ultra-precision servo system according to claim 1, characterized in that, The parameters of the notch filter are generated by adaptively setting the filter center frequency and damping ratio based on the main vibration components identified in the acceleration signal spectrum.

6. The dual feedback control debugging method for an ultra-precision servo system according to claim 1, characterized in that, The debugging processes of the speed feedback loop and the position feedback loop are independent of each other but have coupled feedback logic. The logic is passed through intermediate variables to achieve frequency domain collaborative optimization of the dual closed-loop controller.

7. A dual feedback control debugging device for an ultra-precision servo system, characterized in that, include: A module for constructing a dual-feedback servo system that includes a velocity loop and a position loop; A module for debugging the speed feedback loop, optimizing controller gain parameters based on frequency domain analysis, and outputting the speed loop response characteristics; A module for debugging the position feedback loop and configuring the position loop controller parameters by combining the velocity loop response and acceleration sensor signals; A module that generates filter parameters and automatically adjusts the notch filter based on dynamic vibration characteristics; A module for completing the frequency domain adaptive tuning of the dual feedback controller.

8. A computer storage medium for storing computer programs, characterized in that, When the computer program is read by the computer, the computer executes the method of claim 1.

9. A computer, comprising a processor and a storage medium, characterized in that, When the processor reads the computer program stored in the storage medium, the computer executes the method of claim 1.

10. A computer program product, as a computer program, is characterized by: When the computer program is executed, it implements the method of claim 1.

Images