Adaptive dynamic performance optimization method and system for source measurement unit

Abstract

The invention relates to a self-adaptive dynamic performance optimization method and system for a source measurement unit, and the method specifically comprises the following steps: Step 1: signal collection and performance index extraction: controlling the source measurement unit to output a step signal, and collecting and outputting response waveform data; calculating rise time, stable time and overshoot of the current output waveform according to the output response waveform data; step 2, intelligent parameter recommendation: inputting the rise time, the stabilization time and the overshoot obtained in the step 1 into a pre-constructed intelligent recommendation algorithm model in real time to obtain an adjusted PID control parameter; 3, parameter boundary constraint and safety protection are carried out; step 4, updating the parameters of the controller and carrying out online iteration; according to the invention, multi-dimensional intelligent balance and optimization of dynamic performance are realized; and the method has high adaptivity and intelligent level.

Description

Technical Field

[0001] This invention relates to the fields of electronic measurement technology, automatic control technology and power electronics technology, and in particular to a method and system for intelligent optimization of the dynamic performance of a high-precision source measurement unit (SMU). Background Technology

[0002] A Source Measure Unit (SMU) is a type of test instrument that integrates high-precision source output and precision measurement. It is widely used in semiconductor device characteristic analysis, sensor testing, electrochemical research, and various high-precision power supply and load simulation scenarios. As the operating frequency and performance of electronic devices continue to increase, testing tasks place higher demands on the dynamic performance of SMUs, especially in scenarios involving high-speed step excitation, rapid source adaptation, and transient response. Traditional control strategies often struggle to balance fast response and output stability.

[0003] Existing SMUs mostly employ fixed-parameter proportional-integral-derivative (PID) control or simple gain adjustment methods to achieve closed-loop control. While these methods are structurally simple, they are prone to the following shortcomings when facing different output levels, different load conditions, or rapid dynamic changes: 1. Difficulty in balancing response speed and stability: Fixed parameters cannot adapt to the inertia and damping characteristics of different operating points, resulting in slow response under some test conditions and large overshoot under others. 2. Lack of adaptive capability: When the output level, load characteristics, or operating mode changes, traditional controllers require manual readjustment of PID parameters, failing to achieve real-time optimization and increasing debugging and maintenance costs. 3. Prominent overshoot and oscillation issues: In high-gain or high-speed control scenarios, the system often exhibits transient overshoot, oscillation, and even stability degradation, affecting measurement accuracy and test repeatability. 4. Limited testing efficiency: Due to the lack of automated parameter tuning mechanisms, parameter optimization relies on manual experience, leading to long test preparation times and difficulty in ensuring consistency between different batches or different operators.

[0004] With the development of instrument intelligence and algorithm technology, researchers have attempted to introduce methods such as fuzzy control, expert systems, and machine learning into SMU control to achieve dynamic optimization under multiple operating conditions. However, existing solutions are mostly limited to offline modeling or single-index optimization, failing to dynamically balance multi-dimensional performance indicators such as response time, settling time, and overshoot. Furthermore, they lack sufficient safety boundary constraints, making it difficult to ensure stable operation under high voltage and high current conditions. Therefore, there is an urgent need for a method and system capable of real-time acquisition of output response characteristics, intelligent analysis of performance indicators, and adaptive adjustment of control parameters within safety boundaries. This would simultaneously improve response speed and output stability under multiple load conditions, significantly reduce undesirable dynamic behaviors such as overshoot and oscillation, and thus improve the overall testing efficiency and adaptability of the SMU. Summary of the Invention

[0005] To address the problems of slow response, difficulty in balancing overshoot and oscillation, and the need for manual parameter tuning in existing technologies during multi-condition dynamic output testing, this invention proposes an adaptive dynamic performance optimization method and system for source measurement units (SMUs). This method and system can optimize control parameters in real time within safety boundaries, significantly improving dynamic performance. The invention achieves this by online acquisition and analysis of the SMU's transient response characteristics and constructing an intelligent recommendation algorithm based on expert rules, thereby realizing the proportional gain K... p Integral gain K i Differential gain K d The iterative adaptive adjustment ensures high accuracy while shortening response time and suppressing overshoot and oscillation, thereby improving overall testing efficiency and stability.

[0006] Terminology Explanation:

[0007] ADC: Analog-to-Digital Converter, is an electronic device that converts continuous-time analog signals into discrete-time digital signals.

[0008] To achieve the above objectives, the present invention provides the following technical solution:

[0009] An adaptive dynamic performance optimization method for a source measurement unit specifically includes the following steps:

[0010] Step 1: Signal Acquisition and Performance Index Extraction

[0011] The control source measurement unit outputs a step signal and acquires the output response waveform data; based on the output response waveform data, it calculates the rise time, settling time, and overshoot of the current output waveform.

[0012] Rise time is defined as the time required for the output to rise from 10% to 90% of the final steady-state value, or the time to first reach 90% of the target value at the sampling point.

[0013] Settling time is defined as the time from receiving a step input signal to the output voltage or current signal entering and remaining within the steady-state error band; the steady-state error band refers to the time from when the output voltage or current signal enters the range of the target value y. set The steady-state error band centered at y and with a relative error of δ, i.e. set Within the range of ±δ, where δ is set according to the test accuracy requirements;

[0014] Overshoot is defined as the relative error between the maximum response value and the setpoint.

[0015] Step 2: Intelligent Parameter Recommendation: The rise time, settling time, and overshoot obtained in Step 1 are input in real time into a pre-built intelligent recommendation algorithm model. This model incorporates multiple rules based on experience and theoretical analysis. By matching the response patterns under different operating conditions, it determines the adjustment strategy that the controller should adopt, obtaining the adjusted PID control parameters, including the proportional gain K. p Integral gain K i and differential gain K d ;

[0016] Step 3: Parameter boundary constraints and safety protection: Limit the adjusted PID control parameters to ensure they are within the range allowed by the hardware and safe to operate.

[0017] Step 4: Controller parameter update and online iteration:

[0018] The constrained PID control parameters are written into the internal controller of the source measurement unit to achieve real-time optimization of the next output; through continuous iteration from Step 1 to Step 4, it gradually converges to the optimal state: the response time is shortened, the settling time is reduced, and overshoot and oscillation are significantly suppressed; the iteration stops automatically according to the termination condition, or can be manually terminated by the user to adapt to different test tasks.

[0019] According to a preferred embodiment of the present invention, in Step 1, after the control source measurement unit outputs a step signal, the transient response waveform data of the output voltage or current is acquired by an external oscilloscope and an internal ADC, and the rise time T of the current output waveform is calculated. r Settling time T s and overshoot σ;

[0020] Ascent time T r The discrete form is represented by the first condition that satisfies y. n ≥0.9y set Sampling time, rise time T r Including the time of these sampling points, where y n It is the voltage value at a certain moment, y set The target value is set;

[0021] Overshoot σ: The calculation formula is as follows:

[0022]

[0023] Among them, y max To the maximum value of the response waveform, y set The target value is set;

[0024] According to a preferred embodiment of the present invention, the intelligent recommendation algorithm model includes the following core rules:

[0025] First rule: Overshoot suppression;

[0026] When σ>σ th When an overshoot is detected, σ th It is a positive threshold close to zero;

[0027] If T r <T ref Then adjust K according to the first formula group. p K i and K d ;T ref For reference adaptive response time; T ref The settings are configured according to different modes; the first formula group includes formulas (2), (3), and (4), as shown below:

[0028] K p_new =K p_old *[0.8+0.2*(1-min(σ,1.0))] (2);

[0029] K i_new =K i_old *[1.0-1.13*(min(1.0,(T ref / T r -1.0))] (3);

[0030] K d_new =K d_old *[μ+υ*(min(σ,1.0)] (4);

[0031] Among them, K p_new It is the adjusted proportional gain, K p_old This is the proportional gain before adjustment, K. i_new It is the adjusted integral gain, K i_old K is the integral gain before adjustment. d_new It is the adjusted differential gain, K d_old is the differential gain before adjustment, μ is a basic adjustment coefficient, υ is an overshoot adjustment weight coefficient, and σ is the overshoot.

[0032] If T r ≥T ref If the response speed is normal or slow, adjust K according to the second formula group. p K i and K d ;

[0033] The second set of formulas includes equations (5), (6), and (7), as shown below:

[0034] K p_new =K p_old*[δ*(T ref / T r )](5);

[0035] K i_new =K i_old *[η+θ*(1-min(σ,1.0))](6);

[0036] K d_new =K d_old *[ι+κ*(T r / T ref )](7);

[0037] Where δ, η, θ, ι, κ are preset empirical coefficients, δ is the basic coefficient for proportional gain adjustment, η is the basic coefficient for integral gain adjustment, θ is the overshoot compensation weight coefficient for integral gain, σ is the overshoot amount, ι is the basic coefficient for differential gain adjustment, κ is the response speed weight coefficient for differential gain, and 0 < δ < 1, η ≈ 1.

[0038] Second rule: Accelerate response;

[0039] When σ≤σ th And T r >T ref When the value is +ΔT, it is determined that the system has no overshoot but a slow response, indicating excessive system damping; ΔT is the buffer value. Therefore, K is adjusted according to the third formula group. p K i and K d And decrease K d ;

[0040] The third set of formulas includes equations (8), (9), and (10), as shown below:

[0041] K p_new =K p_old *[1.0+ε*(T r / T ref -1.0)] (8);

[0042] K i_new =K i_old *[1.0+ζ*(T r / T ref -1.0)](9);

[0043] K d_new =K d_old *[ρ-τ*(T r / T ref -1.0)](10);

[0044] Where ε is the response speed deviation weighting coefficient of the proportional gain, ζ is the response speed deviation weighting coefficient of the integral gain, and τ is the response speed deviation weighting coefficient of the differential gain.

[0045] Third rule: Oscillation suppression and termination conditions;

[0046] When multiple iterations satisfy T r ≤T ref , |σ|≤σ th And K i *λ>K p Stop adjusting parameters when the time is right.

[0047] If the waveform is detected to continuously and alternately exceed the target value, oscillation is determined, and K is reduced. p or K i To increase damping, the following suppression strategies are adopted, including equations (11) and (12), as shown below:

[0048] K p_new =K p_old *λ p (11);

[0049] K i_new =K i_old *λ i (12);

[0050] Where, λ p , λ i ∈(0,1) is the damping coefficient (taken as 0.7 to 0.9).

[0051] According to a preferred embodiment of the present invention, in Step 3, the adjusted PID control parameters are subjected to amplitude limiting processing using equations (13), (14), and (15); as shown below:

[0052] K p =max(K) p_min ,min(K p ,K p_max ))(13);

[0053] K i =max(K) i_min ,min(K i ,K i_max ))(14);

[0054] K d =max(K) d_min ,min(K d ,K d_max ))(15);

[0055] Among them, K p_min Kp_max K i_min K i_max K d_min K d_max These are the upper and lower limits of the parameters set based on the hardware characteristics of the source measurement unit.

[0056] A computer device includes a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement steps of an adaptive dynamic performance optimization method for a source measurement unit.

[0057] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of an adaptive dynamic performance optimization method for a source measurement unit.

[0058] An adaptive dynamic performance optimization system for a source measurement unit, comprising:

[0059] The signal acquisition and performance index extraction module is configured as follows:

[0060] The control source measurement unit outputs a step signal and acquires the output response waveform data; based on the output response waveform data, it calculates the rise time, settling time, and overshoot of the current output waveform.

[0061] Rise time is defined as the time required for the output to rise from 10% to 90% of the final steady-state value, or the time to first reach 90% of the target value at the sampling point.

[0062] Settling time is defined as the time from receiving a step input signal to the output voltage or current signal entering and remaining within the steady-state error band; the steady-state error band refers to the time from when the output voltage or current signal enters the range of the target value y. set The steady-state error band centered at y and with a relative error of δ, i.e. set Within the range of ±δ, where δ is set according to the test accuracy requirements;

[0063] Overshoot is defined as the relative error between the maximum response value and the setpoint.

[0064] The intelligent parameter recommendation module is configured as follows:

[0065] The obtained rise time, settling time, and overshoot are input in real time into a pre-built intelligent recommendation algorithm model. This model incorporates multiple rules based on experience and theoretical analysis. By matching the response patterns under different operating conditions, it determines the adjustment strategy that the controller should adopt, thus obtaining the adjusted PID control parameters, including the proportional gain K. p Integral gain K i and differential gain K d ;

[0066] The parameter boundary constraint and safety protection module is configured to: limit the adjusted PID control parameters to ensure that they are within the range allowed by the hardware and safe to operate.

[0067] The controller parameter update and online iteration module is configured to: write the constrained PID control parameters into the internal controller of the source measurement unit to achieve real-time optimization of the next output; continuously iterate and gradually converge to the optimal state: the response time is shortened, the settling time is reduced, and overshoot and oscillation are significantly suppressed; the iteration stops automatically according to the termination condition, or is terminated manually by the user, to adapt to different test tasks.

[0068] Compared with the prior art, the adaptive dynamic performance optimization method and system for source measurement units provided by the present invention have the following significant advantages and beneficial effects:

[0069] 1. Achieves multi-dimensional intelligent balancing and optimization of dynamic performance: By real-time acquisition and analysis of multiple key performance indicators such as rise time Tr, settling time Ts, and overshoot σ of the SMU output waveform, and intelligently adjusting PID control parameters based on an expert rule system, it can automatically balance response speed and stability under different loads and operating conditions. This method effectively solves the inherent contradiction between speed and stability in traditional fixed-parameter PID controllers, significantly improving the overall dynamic testing performance of the SMU.

[0070] 2. High level of adaptability and intelligence: The system's built-in intelligent recommendation algorithm model integrates multiple sets of adaptive adjustment rules and formulas, which can dynamically adjust K based on real-time performance indicators. p K i K d The parameters can be set without manual intervention or pre-modeling. The algorithm supports adaptive parameter settings for multiple levels (such as voltage levels ±600mV, ±6V, ±60V and current levels ±1μA to ±10A), and can dynamically adjust empirical coefficients and thresholds according to hardware characteristics and output modes, demonstrating strong environmental adaptability and intelligence.

[0071] 3. Significantly suppresses overshoot and oscillation, improving test accuracy and repeatability: By introducing overshoot suppression rules and oscillation detection mechanisms (such as damping coefficient λ) p , λ i The system can quickly identify and suppress undesirable dynamic phenomena such as overshoot and oscillation, avoiding system stability degradation. This not only improves the quality of the output waveform but also ensures high accuracy and repeatability of the measurement results, making it particularly suitable for applications such as semiconductor characteristic analysis and sensor testing where extremely high test stability is required.

[0072] 4. Built-in safety boundary constraint mechanism to ensure stable system operation: All parameter adjustments are made within preset safety boundaries (e.g., K). p ∈[0.01,120.0], K i ∈[0.001,50.0], K d The upper and lower limits are ∈[0.001,10.0]), and can be dynamically configured according to the voltage and current levels. This mechanism effectively prevents problems such as controller instability and hardware overload caused by parameter mutations or algorithm misjudgments, ensuring the safe and reliable operation of the system under high power and high voltage conditions.

[0073] 5. Improved testing efficiency and automation, reducing labor costs: This invention achieves a fully automated closed-loop optimization process from signal acquisition, performance analysis, parameter recommendation to controller updates, eliminating the need for manual parameter tuning, greatly shortening test preparation time and improving testing efficiency. Simultaneously, the system supports log recording and feedback functions, facilitating subsequent data analysis and process optimization, significantly reducing long-term operation and maintenance costs.

[0074] 6. Excellent system integration and scalability: The system can be seamlessly integrated into existing SMU control software, supports PXIe bus communication, multi-module synchronous control, and external script calls, facilitating the construction of multi-channel parallel testing and large-scale automated testing systems. The algorithm supports user-defined rules and coefficients, possessing excellent flexibility and scalability, and can adapt to more complex application needs in the future. Attached Figure Description

[0075] Figure 1 This is a schematic diagram of the overall architecture of the adaptive dynamic performance optimization system for the source measurement unit of the present invention;

[0076] Figure 2 This is a flowchart illustrating the implementation of intelligent parameter recommendation in the adaptive dynamic performance optimization method for source measurement units of the present invention.

[0077] Figure 3 This is a schematic diagram of the hardware and software architecture of the adaptive dynamic performance optimization system for the source measurement unit of the present invention;

[0078] Figure 4 This is a schematic diagram of the adaptive dynamic performance index test (current mode, load connection) of the source measurement unit of the present invention;

[0079] Figure 5 This is a schematic diagram of the test of the measurement accuracy index of the source measurement unit described in this invention (voltage mode, no load connection);

[0080] Figure 6 This is a schematic diagram illustrating the overshoot before enabling the adaptive dynamic performance optimization method for the source measurement unit of the present invention;

[0081] Figure 7 This is a schematic diagram of overshoot after enabling the adaptive dynamic performance optimization method for the source measurement unit of the present invention;

[0082] Figure 8 This is a schematic diagram of the response time before activation of the adaptive dynamic performance optimization method for the source measurement unit of the present invention;

[0083] Figure 9 This is a schematic diagram showing the response time after the adaptive dynamic performance optimization method for the source measurement unit of the present invention is enabled;

[0084] Figure 10 This is a schematic diagram illustrating the oscillation suppression before activation of the adaptive dynamic performance optimization method for the source measurement unit of the present invention.

[0085] Figure 11 This is a schematic diagram illustrating the oscillation suppression after the adaptive dynamic performance optimization method for the source measurement unit of the present invention is enabled. Detailed Implementation

[0086] The present invention will be further defined below with reference to the accompanying drawings and embodiments, but is not limited thereto.

[0087] Example 1

[0088] This invention designs an adaptive dynamic performance optimization method and system for a source measurement unit (SMU). Its core lies in automatically calculating the source adaptation adaptive response time T by real-time acquisition of the SMU's output response signal. r Settling time T s And key dynamic performance indicators such as overshoot σ, and based on a built-in expert rule system and adaptive adjustment algorithm, iteratively optimize the PID parameters (K) of the SMU internal controller. p K i K d This allows for an optimal balance between response speed and accuracy while ensuring system stability. All parameter adjustments are performed within preset safety boundaries, ensuring operational safety and reliability.

[0089] An adaptive dynamic performance optimization method for source measurement units is proposed, the core of which lies in establishing an intelligent recommendation system that takes the system's dynamic performance indicators as input and PID parameter adjustments as output. For example... Figure 1 As shown, the specific steps include:

[0090] Step 1: Signal Acquisition and Performance Index Extraction

[0091] The control source measurement unit outputs a step signal and acquires the output response waveform data; based on the output response waveform data, it calculates the rise time, settling time, and overshoot of the current output waveform.

[0092] Rise time is defined as the time required for the output to rise from 10% to 90% of the final steady-state value, or the time to first reach 90% of the target value at the sampling point.

[0093] Settling time is defined as the time from receiving a step input signal (such as a step change in voltage) to the output voltage or current signal entering and remaining within the steady-state error band; the steady-state error band refers to the time from when the output voltage or current signal enters the range of the target value y. set The steady-state error band centered at y and with a relative error of δ, i.e. set Within the range of ±δ, where δ is set according to the test accuracy requirements; common values ​​are 0.001 (i.e. ±0.1%) or less.

[0094] Overshoot is defined as the relative error between the maximum response value and the setpoint.

[0095] Step 2: Intelligent Parameter Recommendation: The rise time, settling time, and overshoot obtained in Step 1 are input in real-time into a pre-built intelligent recommendation algorithm model. This model incorporates multiple rules based on experience and theoretical analysis. By matching the response patterns under different operating conditions, it determines the adjustment strategy that the controller should adopt. The model considers both speed and stability, as well as hardware constraints. The adjusted PID control parameters, including the proportional gain K, are obtained. p Integral gain K i and differential gain K d ;

[0096] Step 3: Parameter Boundary Constraints and Safety Protection: All parameter adjustments are performed within preset safety boundaries to ensure stable system operation under any conditions and prevent oscillations or hardware damage caused by sudden parameter changes. To prevent control parameters from exceeding the hardware safety range, the adjusted PID control parameters need to be limited to ensure they remain within the hardware's allowable and safe operating range.

[0097] Step 4: Controller parameter update and online iteration:

[0098] The constrained PID control parameters are written into the internal controller of the source measurement unit to achieve real-time optimization of the next output. Then, a step signal is re-output and the response waveform is acquired, entering the next round of the "measurement-analysis-recommendation-constraint-update" iterative process. Through continuous iteration from Step 1 to Step 4, it gradually converges to the optimal state: response time is shortened, settling time is reduced, and overshoot and oscillation are significantly suppressed. The iteration stops automatically according to the termination condition or can be manually terminated by the user to adapt to different test tasks.

[0099] Among them, the shortened response time refers to the system's rise time T.r Reduce, T r Defined as the time required for the output signal to rise from 10% to 90% of its steady-state value; a decrease in steady-state time refers to: steady-state time T s Defined as the time required from the start of a step signal output until the output voltage or current signal first enters and remains within the steady-state error band; significant suppression of overshoot and oscillation refers to: the maximum overshoot of the response waveform is reduced, and oscillation suppression refers to: the amplitude and number of oscillations of the response waveform before and after reaching the steady-state value are reduced.

[0100] Example 2

[0101] The adaptive dynamic performance optimization method for a source measurement unit described in Example 1 differs in that:

[0102] In Step 1, after the control source measurement unit outputs a step signal, it acquires the transient response waveform data of the output voltage or current through an external oscilloscope and an internal ADC, and calculates the rise time T of the current output waveform. r Settling time T s and overshoot σ;

[0103] Ascent time T r The discrete form is represented by the first condition that satisfies y. n ≥0.9y set Sampling time, rise time T r Including the time of these sampling points, where y n It is the voltage value at a certain moment, y set The target value is set;

[0104] Overshoot σ: The calculation formula is as follows:

[0105]

[0106] Among them, y max To the maximum value of the response waveform, y set This step sets a target value; it ensures that a multidimensional index that fully characterizes the dynamic performance is obtained, providing accurate input for subsequent adaptive algorithms.

[0107] The intelligent recommendation algorithm model includes the following core rules:

[0108] First rule: Overshoot suppression;

[0109] When σ>σ th When an overshoot is detected, σ th For example, σ is a positive threshold close to zero. th =0.001; This indicates that the system has insufficient damping and poor stability.

[0110] If Tr <T ref Then adjust K according to the first formula group. p K i and K d ;T ref For reference adaptive response time; T ref The settings vary depending on the mode; this indicates that the system response is too fast and underdamped. Overshoot is suppressed and system stability is maintained by reducing the proportional gain, slightly reducing the integral gain, and enhancing the derivative action. Specifically, this is manifested by significantly reducing K. p (Multiplied by a coefficient less than 1) to suppress overshoot, slightly reducing K i To prevent integral saturation and significantly enhance K d To improve the system's damping capability, oscillations and overshoot can be more effectively suppressed by predicting the trend of error changes. The first set of formulas includes equations (2), (3), and (4), as shown below:

[0111] K p_new =K p_old *[0.8+0.2*(1-min(σ,1.0))] (2);

[0112] K i_new =K i_old *[1.0-1.13*(min(1.0,(T ref / T r -1.0))] (3);

[0113] K d_new =K d_old *[μ+υ*(min(σ,1.0)] (4);

[0114] Among them, K p_new It is the adjusted proportional gain, K p_old This is the proportional gain before adjustment, K. i_new It is the adjusted integral gain, K i_old K is the integral gain before adjustment. d_new It is the adjusted differential gain, K d_old is the differential gain before adjustment, μ is a basic adjustment coefficient, υ is an overshoot adjustment weight coefficient, and σ is the overshoot.

[0115] If T r ≥T ref If the response speed is normal or slow, adjust K according to the second formula group. p K i and K dThis situation indicates that the system exhibits both an oscillating tendency (requiring a reduced response speed to decrease overshoot) and inherent inertia (needing an enhanced response capability to increase speed). Therefore, the adjustment strategy of the second set of formulas needs to strike a balance between suppressing overshoot and accelerating the response, specifically by moderately reducing the proportional gain and appropriately adjusting the weights of the integral and derivative. The oscillating tendency of the system is suppressed by significantly reducing the proportional gain. This is achieved by reducing K... p After mitigating the overall response, the integral action is appropriately enhanced to maintain a certain response speed. Furthermore, considering the inherent inertia of the system, the derivative action is enhanced to effectively compensate for system inertia, thereby accelerating the response speed.

[0116] The second set of formulas includes equations (5), (6), and (7), as shown below:

[0117] K p_new =K p_old *[δ*(T ref / T r )](5);

[0118] K i_new =K i_old *[η+θ*(1-min(σ,1.0))](6);

[0119] K d_new =K d_old *[ι+κ*(T r / T ref )](7);

[0120] Where δ, η, θ, ι, κ are preset empirical coefficients, δ is the basic coefficient for proportional gain adjustment, η is the basic coefficient for integral gain adjustment, θ is the overshoot compensation weight coefficient for integral gain, σ is the overshoot amount, ι is the basic coefficient for differential gain adjustment, κ is the response speed weight coefficient for differential gain, and 0 < δ < 1, η ≈ 1.

[0121] Second rule: Accelerate response;

[0122] When σ≤σ th And T r >T ref When the value is +ΔT, it is determined that the system has no overshoot but a slow response, indicating excessive system damping; ΔT is the buffer value. Therefore, K is adjusted according to the third formula group. p K i and K d At the same time, increase K. p and K i To improve response speed and reduce K d To reduce the damping effect of the differential term and avoid it excessively inhibiting the system's response speed.

[0123] The third set of formulas includes equations (8), (9), and (10), as shown below:

[0124] K p_new =K p_old *[1.0+ε*(T r / T ref -1.0)] (8);

[0125] K i_new =K i_old *[1.0+ζ*(T r / T ref -1.0)] (9);

[0126] K d_new =K d_old *[ρ-τ*(T r / T ref -1.0)] (10);

[0127] Where ε is the response speed deviation weighting coefficient of the proportional gain, ζ is the response speed deviation weighting coefficient of the integral gain, and τ is the response speed deviation weighting coefficient of the differential gain.

[0128] Third rule: Oscillation suppression and termination conditions;

[0129] When multiple iterations satisfy T r ≤T ref , |σ|≤σ th And K i *λ>K p (λ is a proportionality constant, for example, satisfying K) i *10.5≥K p When optimal performance is achieved, parameter adjustments are stopped to avoid unnecessary oscillations. The intelligent recommendation algorithm model supports multi-level adaptive adjustment, including voltage levels of ±600mV, ±6V, and ±60V, and current levels from ±1μA to ±10A, with different T values ​​set for different levels. ref and σ th Threshold.

[0130] If a waveform is detected to continuously and alternately exceed the target value, this indicates a typical phenomenon of continuous system oscillation. This means the system's output response cannot stabilize at the setpoint, but rather fluctuates repeatedly around it. Specifically, it refers to the system output value fluctuating around the setpoint y. set The waveform oscillates back and forth between the upper (positive overshoot) and lower (undershoot) ranges, forming a periodic waveform with very slow decay. This is identified as oscillation, and K is then reduced. p or K i To increase damping, the following suppression strategies are adopted, including equations (11) and (12), as shown below:

[0131] Kp_new =K p_old *λ p (11);

[0132] K i_new =K i_old *λ i (12);

[0133] Where, λ p , λ i ∈(0,1) is the damping coefficient (taken as 0.7 to 0.9).

[0134] This step uses systematic rules to accurately identify different dynamic states and recommend corresponding strategies. The expert rule system supports user-defined thresholds (such as σ). th T ref It can add rules and coefficients (such as α, β, γ, etc.) and can be extended to adapt to special application scenarios, with good flexibility and scalability.

[0135] In Step 3, the adjusted PID control parameters are limited using equations (13), (14), and (15), as shown below:

[0136] K p =max(K) p_min ,min(K p ,K p_max )) (13);

[0137] K i =max(K) i_min ,min(K i ,K i_max (14);

[0138] K d =max(K) d_min ,min(K d ,K d_max )) (15);

[0139] Among them, K p_min K p_max K i_min K i_max K d_min K d_max These are the upper and lower limits of the parameters set according to the hardware characteristics of the source measurement unit. The upper and lower limits are set separately for different output levels and modes.

[0140] The upper and lower limits of the parameters are dynamically configured based on the voltage range (e.g., ±600mV, ±6V, ±60V) and current range (±1μA to ±10A) of the SMU. This constraint ensures that controller instability or hardware overload will not occur in high-power or high-voltage modes, providing safety protection for global optimization.

[0141] The intelligent recommendation algorithm proposed in this invention has high adaptability and real-time performance. It can dynamically adjust PID parameters according to the actual output response of the SMU without relying on pre-modeling or manual intervention, which significantly improves the system's adaptability under different loads and different levels.

[0142] See Figure 2 This embodiment combines Figure 2 This paper describes the workflow of the intelligent recommendation algorithm of this invention. The algorithm automatically starts after each acquisition of the SMU's step response waveform, using the real-time calculated rise time, overshoot, and settling time as input. The algorithm first determines whether overshoot exists: if overshoot is present, different parameter adjustment strategies are selected based on the current response speed. For fast responses, the proportional gain is mainly reduced and the integral gain is fine-tuned to suppress overshoot; for slower responses, overshoot suppression is combined with response speed optimization; if there is no overshoot but the response is slow, both the proportional and integral gains are increased to accelerate the response. After each parameter adjustment, the algorithm checks whether a preset termination condition has been met. If the condition is met, optimization stops; otherwise, the next round of iterative adjustment begins. This cyclical optimization mechanism continuously brings the system's dynamic performance closer to its optimal state. Through intelligent condition judgment and parameter adjustment strategies, this algorithm achieves automatic optimization of the SMU's dynamic performance, achieving a balance between fast response and stable output without manual intervention.

[0143] Figure 4 This is a schematic diagram of the adaptive dynamic performance index test (current mode, load connection) of the source measurement unit of the present invention; Figure 5 This is a schematic diagram of the test of the measurement accuracy index of the source measurement unit described in this invention (voltage mode, no load connection); Figure 6 This is a schematic diagram illustrating the overshoot before enabling the adaptive dynamic performance optimization method for the source measurement unit of the present invention; Figure 7 This is a schematic diagram of overshoot after enabling the adaptive dynamic performance optimization method for the source measurement unit of the present invention; Figure 8 This is a schematic diagram of the response time before activation of the adaptive dynamic performance optimization method for the source measurement unit of the present invention; Figure 9 This is a schematic diagram showing the response time after the adaptive dynamic performance optimization method for the source measurement unit of the present invention is enabled; Figure 10 This is a schematic diagram illustrating the oscillation suppression before activation of the adaptive dynamic performance optimization method for the source measurement unit of the present invention. Figure 11This is a schematic diagram illustrating the oscillation suppression after the adaptive dynamic performance optimization method for the source measurement unit of the present invention is enabled.

[0144] This demonstrates intuitively the significant effect of the intelligent recommendation algorithm of this invention in improving the dynamic performance of SMU.

[0145] Taking the ±6V voltage range as an example, the SMU outputs a +5V step signal, and the load condition is a transient change from 10% to 90% of the range. The response waveform is acquired using an oscilloscope, and the waveform characteristics before and after the intelligent recommendation algorithm is enabled are recorded.

[0146] Performance Comparison Analysis: Overshoot Suppression Effect (Corresponding) Figure 6 ): Figure 6 Before the algorithm was enabled, the response waveform showed obvious overshoot, with an overshoot of 5.2%. Figure 7 After the algorithm was enabled, the waveform rose smoothly, and the overshoot decreased to 0.1%, significantly improving output stability.

[0147] Response time optimization (corresponding) Figure 8 ): Figure 8 Before the algorithm was enabled, the rise time was 90μs and the settling time was 130μs. Figure 9 After enabling the algorithm, the rise time was shortened to 39μs and the settling time was reduced to 65μs, resulting in a significant improvement in response speed.

[0148] Oscillation suppression effect (corresponding) Figure 10 ): Figure 10 Before the display algorithm was enabled, the waveform oscillated multiple times near the stable value for a relatively long period of time. Figure 11 After the algorithm was enabled, the oscillation phenomenon was completely eliminated, and the waveform quickly and smoothly converged to the target value.

[0149] Example 3

[0150] A computer device includes a memory and a processor, the memory storing a computer program, the processor executing the computer program to implement the steps of an adaptive dynamic performance optimization method for a source measurement unit as described in Embodiment 1 or 2.

[0151] Example 4

[0152] A computer-readable storage medium having a computer program stored thereon, which, when executed by a processor, implements the steps of an adaptive dynamic performance optimization method for a source measurement unit as described in Embodiment 1 or 2.

[0153] Example 5

[0154] An adaptive dynamic performance optimization system for a source measurement unit, comprising:

[0155] The signal acquisition and performance index extraction module is configured as follows:

[0156] The control source measurement unit outputs a step signal and acquires the output response waveform data; based on the output response waveform data, it calculates the rise time, settling time, and overshoot of the current output waveform.

[0157] Rise time is defined as the time required for the output to rise from 10% to 90% of the final steady-state value, or the time to first reach 90% of the target value at the sampling point.

[0158] Settling time is defined as the time from receiving a step input signal (such as a step change in voltage) to the output voltage or current signal entering and remaining within the steady-state error band; the steady-state error band refers to the time from when the output voltage or current signal enters the range of the target value y. set The steady-state error band centered at y and with a relative error of δ, i.e. set Within the range of ±δ, where δ is set according to the test accuracy requirements; common values ​​are 0.001 (i.e. ±0.1%) or less.

[0159] Overshoot is defined as the relative error between the maximum response value and the setpoint.

[0160] The intelligent parameter recommendation module is configured as follows:

[0161] The obtained rise time, settling time, and overshoot are input in real time into a pre-built intelligent recommendation algorithm model. This model incorporates multiple rules based on empirical and theoretical analysis. By matching the response patterns under different operating conditions, it determines the adjustment strategy that the controller should adopt. The model considers both speed and stability, as well as hardware constraints. The adjusted PID control parameters, including the proportional gain K, are then obtained. p Integral gain K i and differential gain K d ;

[0162] The parameter boundary constraint and safety protection module is configured such that all parameter adjustments are made within preset safety boundaries, ensuring stable system operation under any conditions and preventing oscillations or hardware damage caused by sudden parameter changes. To prevent control parameters from exceeding the hardware safety range, the adjusted PID control parameters need to be limited to ensure they remain within the hardware's allowable and safe operating range.

[0163] The controller parameter update and online iteration module is configured to: write the constrained PID control parameters into the internal controller of the source measurement unit to achieve real-time optimization of the next output; then re-output the step signal and acquire the response waveform, entering the next round of the "measurement-analysis-recommendation-constraint-update" iterative process. Continuous iteration gradually converges to the optimal state: response time is shortened, settling time is reduced, and overshoot and oscillation are significantly suppressed; iteration automatically stops according to termination conditions, or can be manually terminated by the user, to adapt to different test tasks.

[0164] Among them, the shortened response time refers to the system's rise time T. r Reduce, T r Defined as the time required for the output signal to rise from 10% to 90% of its steady-state value; a decrease in steady-state time refers to: steady-state time T s Defined as the time required from the start of a step signal output until the output voltage or current signal first enters and remains within the steady-state error band; significant suppression of overshoot and oscillation refers to: a reduction in the maximum overshoot of the response waveform; and oscillation suppression refers to a reduction in the amplitude and frequency of oscillations in the response waveform before and after reaching the steady-state value. The adaptive optimization algorithm proposed in this invention can be embedded in SMU control software to form a complete closed-loop control system. The system achieves continuous optimization through the organic synergy of hardware acquisition, data analysis, algorithm decision-making, and control updates. The specific system architecture is as follows:

[0165] At the output end, the response is acquired and digitized in real time by a high-precision ADC and oscilloscope to ensure the complete fidelity of transient characteristics;

[0166] The performance analysis logic calculates rise time, settling time, and overshoot of the sampled data and compares them against thresholds.

[0167] The intelligent recommendation algorithm performs inference and parameter generation based on real-time metrics and execution rules;

[0168] The controller updates the PID parameters in real time based on the algorithm recommendation and drives the SMU output to form the next round of response;

[0169] The system synchronously records iterative data, performance changes, and parameter tuning trends for subsequent performance tracking and batch optimization.

[0170] Through the PXIe bus and multi-module synchronization mechanism, the system supports multi-channel parallel testing and external script calls, facilitating batch testing and long-term operational monitoring, thereby significantly improving testing efficiency and scalability. In summary, this invention effectively solves the dynamic performance optimization problem of SMU under multiple operating conditions through an intelligent real-time parameter adjustment mechanism, achieving an intelligent balance between high precision and high speed, and possesses strong engineering practical value and promising application prospects. This system can be integrated into existing SMU control platforms, supports multiple communication protocols (such as SCPI and PXIe), and has good compatibility and portability.

[0171] like Figure 1 As shown, the core of the adaptive dynamic performance optimization system constructed in this invention lies in forming a closed-loop control system. This system mainly includes four functional modules: a data acquisition module, a performance analysis module, an intelligent recommendation module, and a control execution module. These modules are integrated into the SMU's host computer control software, and interact with the SMU hardware modules via the PXIe bus for high-speed data exchange and command transmission. The data acquisition module is the system's perception layer. It consists of a high-precision ADC inside the SMU and an external oscilloscope (such as MSO4054), responsible for synchronously acquiring the actual output voltage or current response waveform of the SMU after the control execution module sends a signal. Its high sampling rate (2.5GS / s) and digitizer function ensure the complete and accurate capture of transient response characteristics. The performance analysis module is the system's evaluation layer. It receives raw waveform data from the data acquisition module and calculates three core dynamic performance indicators in real time: source-adaptive adaptive response time T0. r Settling time T s The calculated index value, along with the overshoot σ, is converted into a standardized data packet and sent to the intelligent recommendation module, which embeds an intelligent recommendation algorithm model. This module receives the T signal from the performance analysis module. r T s The system collects σ data and compares it with the preset threshold value for the current gear. Based on multiple built-in rules (rules one, two, and three) and corresponding formula groups (formula groups one, two, and three), it performs logical judgments and calculations, outputting a set of optimized PID parameter suggestions. The control execution module receives the parameter suggestions from the intelligent recommendation module and performs safety boundary constraint processing before sending them to the embedded controller of the SMU to ensure that all parameters are within the stable range allowed by the hardware. Subsequently, this module updates the new set of safe parameters to the controller via the PXIe bus, completing the control preparation for the next output. Figure 1 The arrows clearly illustrate the system's workflow, forming a complete closed loop of "measurement → analysis → decision-making → execution".

[0172] See Figure 3 This embodiment combines Figure 3Briefly describe the hardware and software architecture of the system of this invention. For example... Figure 3 As shown, the system adopts a layered architecture, consisting of a hardware layer, a driver layer, and a software application layer. The hardware layer includes high-resolution power measurement modules, a PXIe chassis, an oscilloscope, electronic loads, and other test equipment, connected via PXIe bus and GPIB / USB interfaces. The driver layer provides the underlying drivers and control interfaces for the hardware devices, ensuring communication and synchronization between them. The software application layer, as the core of the system, integrates four major functional modules: data acquisition, performance analysis, intelligent recommendation, and control execution. It achieves fully automated operations such as real-time waveform acquisition, performance index calculation, intelligent PID parameter recommendation, and controller parameter updates. The system also provides a human-machine interface, supporting functions such as parameter setting, process monitoring, and test data recording.

Claims

1. An adaptive dynamic performance optimization method for a source measurement unit, characterized in that, Specifically, the following steps are included: Step 1: Signal Acquisition and Performance Index Extraction The control source measurement unit outputs a step signal and acquires the output response waveform data; based on the output response waveform data, it calculates the rise time, settling time, and overshoot of the current output waveform. Rise time is defined as the time required for the output to rise from 10% to 90% of the final steady-state value, or the time to first reach 90% of the target value at the sampling point. Settling time is defined as the time from receiving a step input signal to the output voltage or current signal entering and remaining within the steady-state error band; the steady-state error band refers to the time from when the output voltage or current signal enters the range of the target value y. set The steady-state error band centered at y and with a relative error of δ, i.e. set Within the range of ±δ, where δ is set according to the test accuracy requirements; Overshoot is defined as the relative error between the maximum response value and the setpoint. Step 2: Intelligent Parameter Recommendation: The rise time, settling time, and overshoot obtained in Step 1 are input in real time into a pre-built intelligent recommendation algorithm model. This model incorporates multiple rules based on experience and theoretical analysis. By matching the response patterns under different operating conditions, it determines the adjustment strategy that the controller should adopt, obtaining the adjusted PID control parameters, including the proportional gain K. p Integral gain K i and differential gain K d ; Step 3: Parameter boundary constraints and safety protection: Limit the adjusted PID control parameters to ensure they are within the range allowed by the hardware and safe to operate. Step 4: Controller parameter update and online iteration: The constrained PID control parameters are written into the internal controller of the source measurement unit to achieve real-time optimization of the next output; Through continuous iteration from Step 1 to Step 4, the system gradually converges to the optimal state: response time is shortened, settling time is reduced, and overshoot and oscillation are significantly suppressed; the iteration stops automatically according to the termination condition, or can be manually terminated by the user, to adapt to different test tasks.

2. The adaptive dynamic performance optimization method for a source measurement unit according to claim 1, characterized in that, In Step 1, after the control source measurement unit outputs a step signal, it acquires the transient response waveform data of the output voltage or current through an external oscilloscope and an internal ADC, and calculates the rise time T of the current output waveform. r Settling time T s and overshoot σ; Ascent time T r The discrete form is represented by the first condition that satisfies y. n ≥0.9y set Sampling time, rise time T r Including the time of these sampling points, where y n It is the voltage value at a certain moment, y set The target value is set; Overshoot σ: The calculation formula is as follows: Among them, y max To respond to the maximum value of the waveform, y set The target value is set.

3. The adaptive dynamic performance optimization method for a source measurement unit according to claim 1, characterized in that, The intelligent recommendation algorithm model includes the following core rules: First rule: Overshoot suppression; When σ>σ th When an overshoot is detected, σ th It is a positive threshold close to zero; If T r <T ref Then adjust K according to the first formula group. p K i and K d ;T ref For reference adaptive response time; T ref The settings are configured according to different modes; the first formula group includes formulas (2), (3), and (4), as shown below: K p_new =K p_old *[0.8+0.2*(1-min(σ,1.0))] (2); K i_new =K i_old *[1.0-1.13*(min(1.0,(T ref / T r -1.0))] (3); K d_new =K d_old *[μ+υ*(min(σ,1.0)] (4); Among them, K p_new It is the adjusted proportional gain, K p_old This is the proportional gain before adjustment, K. i_new It is the adjusted integral gain, K i_old K is the integral gain before adjustment. d_new It is the adjusted differential gain, K d_old is the differential gain before adjustment, μ is a basic adjustment coefficient, υ is an overshoot adjustment weight coefficient, and σ is the overshoot. If T r ≥T ref If the response speed is normal or slow, adjust K according to the second formula group. p K i and K d ; The second set of formulas includes equations (5), (6), and (7), as shown below: K p_new =K p_old *[δ*(T ref / T r )] (5); K i_new =K i_old *[η+θ*(1-min(σ,1.0))] (6); K d_new =K d_old *[i+k*(T r / T ref )] (7); Where δ, η, θ, ι, κ are preset empirical coefficients, δ is the basic coefficient for proportional gain adjustment, η is the basic coefficient for integral gain adjustment, θ is the overshoot compensation weight coefficient for integral gain, σ is the overshoot amount, ι is the basic coefficient for differential gain adjustment, and κ is the response speed weight coefficient for differential gain, and 0<δ<1, η≈1. Second rule: Accelerate response; When σ≤σ th And T r >T ref When the value is +ΔT, it is determined that the system has no overshoot but a slow response, indicating excessive system damping; ΔT is the buffer value. Therefore, K is adjusted according to the third formula group. p K i and K d ; The third set of formulas includes equations (8), (9), and (10), as shown below: K p_new =K p_old *[1.0+ε*(T r / T ref -1.0)] (8); K i_new =K i_old *[1.0+ζ*(T r / T ref -1.0)] (9); K d_new =K d_old *[p-t*(T r / T ref -1.0)] (10); Where ε is the response speed deviation weighting coefficient of the proportional gain, ζ is the response speed deviation weighting coefficient of the integral gain, and τ is the response speed deviation weighting coefficient of the differential gain. Third rule: Oscillation suppression and termination conditions; When multiple iterations satisfy T r ≤T ref , |σ|≤σ th And K i *λ>K p Stop adjusting parameters when the time is right. If the waveform is detected to continuously and alternately exceed the target value, oscillation is determined, and K is reduced. p or K i To increase damping, the following suppression strategies are adopted, including equations (11) and (12), as shown below: K p_new =K p_old *l p (11); K i_new =K i_old *l i (12); Where, λ p , λ i ∈(0,1) is the damping coefficient.

4. The adaptive dynamic performance optimization method for a source measurement unit according to claim 3, characterized in that, l i ∈(0.7, 0.9).

5. The adaptive dynamic performance optimization method for a source measurement unit according to claim 1, characterized in that, In Step 3, the adjusted PID control parameters are limited using equations (13), (14), and (15), as shown below: K p =max(K p_min ,min(K p ,K p_max )) (13); K i =max(K i_min ,min(K i ,K i_max )) (14); K d =max(K d_min ,min(K d ,K d_max )) (15); Among them, K p_min K p_max K i_min K i_max K d_min K d_max These are the upper and lower limits of the parameters set based on the hardware characteristics of the source measurement unit.

6. A computer device comprising a memory and a processor, wherein the memory stores a computer program, characterized in that, When the processor executes the computer program, it implements the steps of the adaptive dynamic performance optimization method for a source measurement unit as described in any one of claims 1-5.

7. A computer-readable storage medium having a computer program stored thereon, characterized in that, When the computer program is executed by the processor, it implements the steps of the adaptive dynamic performance optimization method for a source measurement unit as described in any one of claims 1-5.

8. An adaptive dynamic performance optimization system for a source measurement unit, characterized in that, include: The signal acquisition and performance index extraction module is configured as follows: The control source measurement unit outputs a step signal and acquires the output response waveform data; Based on the output response waveform data, calculate the rise time, settling time, and overshoot of the current output waveform; Rise time is defined as the time required for the output to rise from 10% to 90% of the final steady-state value, or the time to first reach 90% of the target value at the sampling point. Settling time is defined as the time from receiving a step input signal to the output voltage or current signal entering and remaining within the steady-state error band; the steady-state error band refers to the time from when the output voltage or current signal enters the range of the target value y. set The steady-state error band centered at y and with a relative error of δ, i.e. set Within the range of ±δ, where δ is set according to the test accuracy requirements; Overshoot is defined as the relative error between the maximum response value and the setpoint. The intelligent parameter recommendation module is configured as follows: The obtained rise time, settling time, and overshoot are input in real time into a pre-built intelligent recommendation algorithm model. This model incorporates multiple rules based on experience and theoretical analysis. By matching the response patterns under different operating conditions, it determines the adjustment strategy that the controller should adopt, thus obtaining the adjusted PID control parameters, including the proportional gain K. p Integral gain K i and differential gain K d ; The parameter boundary constraint and safety protection module is configured to: limit the adjusted PID control parameters to ensure that they are within the range allowed by the hardware and safe to operate. The controller parameter update and online iteration module is configured to: write the constrained PID control parameters into the internal controller of the source measurement unit to achieve real-time optimization of the next output; continuously iterate and gradually converge to the optimal state: the response time is shortened, the settling time is reduced, and overshoot and oscillation are significantly suppressed; the iteration stops automatically according to the termination condition, or is terminated manually by the user, to adapt to different test tasks.

Images