Multivariable decoupling method for temperature control system of test equipment
By employing a generalized internal model control framework and multi-frequency orthogonal disturbance signals in the temperature control system of semiconductor testing equipment, the internal model matrix and decoupling compensator are updated in real time. This solves the problem that fixed decoupling models cannot cope with system nonlinearity and time-varying coupling, and achieves high-precision and efficient temperature control.
Patent Information
- Application Number
- CN202511574093.3
- Authority / Receiving Office
- CN · China
- Patent Type
- Applications(China)
- Current Assignee / Owner
- Filing Date
- 2025-10-31
- Publication Date
- 2025-12-16
AI Technical Summary
In existing temperature control systems for semiconductor testing equipment, fixed decoupling models struggle to handle the nonlinearity and time-varying coupling of the system, resulting in limited control accuracy and efficiency.
By adopting a generalized internal model control framework, multi-frequency orthogonal disturbance signals are injected, and multi-channel synchronous demodulation and model fitting are used to update the internal model matrix and decoupling compensator in real time, thereby realizing dynamic model identification and compensation of coupling paths.
It achieves high-precision decoupling performance under different operating conditions, improves the adaptability and efficiency of the control system, reduces overshoot and oscillation in temperature control, and enhances the accuracy and efficiency of semiconductor testing.
Smart Images
Figure CN121143545A_ABST
Abstract
Description
Technical Field
[0001] This invention relates to the field of semiconductor testing equipment technology, specifically to a multivariable decoupling method for a temperature control system of a testing equipment. Background Technology
[0002] In the research, development, production, and quality inspection of semiconductor devices, different temperature stresses need to be applied to the devices to test their electrical performance and reliability under different temperature conditions. Companies such as Nanjing Hongtai Semiconductor Technology Co., Ltd. specialize in providing such semiconductor testing equipment, such as automated testing equipment, wafer probe stations, or sorting machines. One of their core functions is to provide a precise and controllable testing temperature environment.
[0003] To achieve rapid heating, cooling, and precise temperature control of test heads or wafer chucks, existing temperature control systems typically integrate both heating and cooling units. This constitutes a typical multi-input multi-output (MIMO) control system. However, a significant thermodynamic coupling effect exists between the heating and cooling units: turning on the heater increases the heat load on the cooling system, while fluctuations in the cooling unit can conversely interfere with the stability of heating.
[0004] To address this coupling issue, the industry commonly employs decoupling control strategies based on mathematical models. However, these strategies inherently rely on a pre-established, fixed system model. Semiconductor testing equipment exhibits strong nonlinearity and time-varying characteristics. For example, the heat transfer coefficient, heat capacity, and coupling degree differ significantly between high and low temperature regions, and the loads on different test devices also alter the system's dynamic response. Therefore, using a fixed, offline-identified decoupling model makes it difficult to maintain ideal control performance across the entire operating range of the equipment. When actual operating conditions deviate from the initial conditions established in the model, the decoupling effect is significantly reduced, leading to overshoot, oscillations, or prolonged time to reach steady state in temperature control, directly impacting the accuracy and efficiency of semiconductor testing. Summary of the Invention
[0005] To address the shortcomings of existing technologies, this invention provides a multivariable decoupling method for temperature control systems of testing equipment. This method solves the problem that fixed decoupling models are unable to cope with system nonlinearity and time-varying coupling in temperature control of semiconductor testing equipment, resulting in limited control accuracy and efficiency.
[0006] To achieve the above objectives, the present invention provides a multivariable decoupling method for a temperature control system of a testing device, comprising the following steps: S1: Establish a generalized internal model control framework, which has an internal model matrix containing a master response term and a coupling term; S2: A set of multi-frequency orthogonal disturbance signals consisting of multiple different frequencies is injected into at least two control inputs of the generalized internal model control framework, the multi-frequency orthogonal disturbance signals being used to detect the coupling characteristics between the control inputs; S3: Acquire at least one controlled output signal from the temperature control system of the test equipment; S4: The controlled output signal acquired by S3 is processed by multi-channel synchronous demodulation to extract the response characteristics of each frequency component of the multi-frequency orthogonal disturbance signal injected by S2 on the controlled output signal. S5: Based on the response characteristics extracted in S4, calculate multiple frequency response data points for a coupling path between the control input and the controlled output; S6: Using the multiple frequency response data points calculated in S5, fit a dynamic model of the coupling path; S7: Using the dynamic model fitted by S6, update the coupling terms in the internal model matrix of S1; S8: The generalized internal model control framework adjusts the decoupling compensation based on the internal model matrix updated in S7.
[0007] Preferably, the multi-frequency orthogonal disturbance signal in S2 is specifically: Each of the at least two control inputs is configured with a set of disturbance frequencies consisting of multiple different frequencies; The sets of disturbance frequencies from different control inputs are orthogonal to each other and do not contain overlapping frequencies.
[0008] Preferably, the multi-channel synchronous demodulation in S4 specifically includes: Generate an in-phase reference signal and a quadrature reference signal that are synchronized with each frequency component in the multi-frequency quadrature disturbance signal; The controlled output signal acquired by S3 is orthogonally mixed with the in-phase reference signal and the quadrature reference signal respectively; The mixed signal is low-pass filtered to obtain the response characteristics, which are in-phase and quadrature components.
[0009] Preferably, calculating multiple frequency response data points in step S5 specifically includes: Based on the in-phase component, the quadrature component, and the injection amplitude of the multi-frequency quadrature disturbance signal in S2, the gain response and phase response of the coupling path on each frequency component of the multi-frequency quadrature disturbance signal are calculated, and the gain response and the phase response constitute the plurality of frequency response data points.
[0010] Preferably, the dynamic model in S6 is a first-order plus pure time delay model.
[0011] Preferably, the fitting of the dynamic model in S6 specifically includes: Construct an objective function to minimize the error between the frequency response of the dynamic model at the corresponding frequencies of the plurality of frequency response data points and the plurality of frequency response data points; The objective function is solved by an optimization algorithm to obtain the model parameters of the dynamic model.
[0012] Preferably, the model parameters include the gain, time constant, and pure time delay of the coupling path.
[0013] Preferably, the amplitude of the multi-frequency orthogonal disturbance signal in S2 is set to a preset value, which ensures that the multi-frequency orthogonal disturbance signal does not cause perceptible interference to the main control task of the temperature control system of the test equipment.
[0014] Preferably, the internal model matrix in S1 is a transfer function matrix; The updating of the coupling term in the internal model matrix in S7 specifically involves: The dynamic model fitted by S6 is used as an off-diagonal element of the transfer function matrix for updating.
[0015] Preferably, the generalized internal model control framework in S8 reconstructs a decoupling compensator in real time based on the updated internal model matrix. The decoupling compensator is used to perform feedforward compensation on the control input to counteract the coupling effect.
[0016] This invention provides a multivariable decoupling method for a temperature control system of a testing device. It has the following beneficial effects: 1. This invention injects multi-frequency orthogonal disturbance signals in step S2, and utilizes multi-channel synchronous demodulation in S4 and model fitting in S6 to achieve online identification of the dynamic model of the coupling path. Subsequently, S7 and S8 update the internal model matrix and adjust the decoupling compensation in real time based on the identification results. This method can automatically adapt to the time-varying coupling characteristics caused by changes in the operating point (such as temperature) or equipment aging of the test equipment, ensuring that the control system maintains high-precision decoupling performance under different operating conditions.
[0017] 2. This invention utilizes the multi-frequency orthogonal disturbance signal in S2, allowing for the simultaneous detection of multiple frequency response data points required by S5 with a single injection, significantly improving identification efficiency. Combined with multi-channel synchronous demodulation technology (such as orthogonal mixing and low-pass filtering) in S4, it can accurately extract the weak disturbance response characteristics defined in claim 8 from the controlled output, achieving high signal-to-noise ratio coupling characteristic identification without interfering with the main control task (without generating perceptible interference).
[0018] 3. This invention is based on the Generalized Internal Model Control (GIMC) framework of S1, and explicitly defines the coupling terms in S9 as off-diagonal elements of the internal model matrix. This method uses online identification in S2-S7 to specifically update the coupling terms in S7, while S8 and S10 reconstruct the decoupling compensator in real time based on the updated matrix. This structure clearly separates the main loop control from the decoupling compensation, making the decoupling adjustment logic clear and enabling proactive cancellation of coupling effects through feedforward compensation. Attached Figure Description
[0019] Figure 1 This is a flowchart of the method of the present invention. Detailed Implementation
[0020] The technical solutions in the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. Based on the embodiments of the present invention, all other embodiments obtained by those skilled in the art without creative effort are within the scope of protection of the present invention.
[0021] Please see the appendix Figure 1 This invention provides a multivariable decoupling method for a temperature control system of a test device, comprising the following steps: S1: Establish a generalized internal model control framework. The generalized internal model control framework has an internal model matrix, which contains the main response term and the coupling term. S2: At least two control inputs of the generalized internal model control framework are injected with a set of multi-frequency orthogonal disturbance signals consisting of multiple different frequencies. The multi-frequency orthogonal disturbance signals are used to detect the coupling characteristics between the control inputs. S3: Acquire at least one controlled output signal from the temperature control system of the test equipment; S4: Use multi-channel synchronous demodulation to process the controlled output signal acquired by S3, so as to extract the response characteristics of each frequency component in the multi-frequency orthogonal disturbance signal injected by S2 on the controlled output signal. S5: Based on the response characteristics extracted in S4, calculate multiple frequency response data points of a coupling path between the control input and the controlled output; S6: Using multiple frequency response data points calculated in S5, fit a dynamic model of a coupling path. S7: Update the coupling terms in the internal model matrix of S1 using the dynamic model fitted by S6; S8: The generalized internal model control framework adjusts the decoupling compensation based on the internal model matrix updated in S7.
[0022] S1: Establish a Generalized Internal Model Control (GIMC) framework. This framework runs as the main control algorithm in the controller. The core of the GIMC framework is the use of an internal model matrix. To characterize the dynamic characteristics of the temperature control system of the controlled test equipment.
[0023] The internal model matrix in S1 is a transfer function matrix. The control input vector is described mathematically. (e.g., heating power, cooling power) and the controlled output vector (e.g., the relationship between zone temperature and evaporator temperature).
[0024] Internal model matrix It includes the main response term and coupling terms. Specifically, the main response term is the diagonal element of the transfer function matrix. , indicating the first The control input pair for the first The main influence of the controlled output. Coupling terms are the off-diagonal elements of the transfer function matrix. (in ), indicating the first The control input pair for the first The undesired coupling effects of a controlled output. The purpose of this invention is to accurately identify and compensate for the dynamics represented by coupling terms.
[0025] S2: Inject a set of multi-frequency orthogonal disturbance signals consisting of multiple different frequencies onto at least two control inputs of the generalized internal model control framework. For example, on the heating control input... and refrigeration control input Simultaneously inject disturbances.
[0026] Multi-frequency orthogonal disturbance signal It is a composite sinusoidal signal, and its form is: in Representing the One control input, It is the first The injection amplitude of each frequency component It is the first The angular frequency of each frequency component.
[0027] This multi-frequency quadrature perturbation signal is used to detect the coupling characteristics between control inputs.
[0028] Each of at least two control inputs (e.g.) and ), respectively set a set of perturbation frequencies consisting of multiple different frequencies (e.g. and To ensure that the signal can be separated at the output, the sets of disturbance frequencies of different control inputs are orthogonal to each other, i.e. This ensures that they do not contain overlapping frequencies. Specifically, this corresponds to: for each of at least two control inputs, a set of disturbance frequencies consisting of multiple different frequencies is set; the disturbance frequency sets of different control inputs are orthogonal to each other and do not contain overlapping frequencies.
[0029] The amplitude of the multi-frequency quadrature disturbance signal in S2 It is set to a preset value. This preset value is selected to be small enough to ensure that the output fluctuations caused by its injection are accounted for by the system's thermal inertia. The signal is smoothed out, thus preventing any perceptible interference with the main control task of the temperature control system of the test equipment. Specifically, this corresponds to setting the amplitude of the multi-frequency quadrature disturbance signal in S2 to a preset value, which ensures that the multi-frequency quadrature disturbance signal does not cause any perceptible interference with the main control task of the temperature control system of the test equipment.
[0030] The specific method of disturbance injection is to use the main control signal calculated by the GIMC framework. With disturbance signal The two signals are added together to form the final drive signal applied to the actuator. .
[0031] S3: Acquire at least one controlled output signal from the temperature control system of the test equipment. The controller acquires the controlled output signal in real time through sensors (such as thermocouples or platinum resistance thermometers). (For example, in identification) right When coupling, acquisition ) S4: Utilize multi-channel synchronous demodulation to process the controlled output signal acquired by S3. The purpose of this step is to extract each frequency component (e.g., from the multi-frequency orthogonal perturbation signal injected by S2) from the S2-injected signal. In the controlled output signal The response characteristics on.
[0032] Multi-channel synchronous demodulation is implemented in the controller using a digital phase-locked loop algorithm, specifically including: Generate each frequency component in the multi-frequency orthogonal perturbation signal (e.g.) Synchronous in-phase reference signal and quadrature reference signals .
[0033] The controlled output signal collected by S3 The mixed signal is obtained by performing quadrature mixing (i.e., point-by-point multiplication) with both the in-phase reference signal and the quadrature reference signal. and .
[0034] The mixed signal is low-pass filtered (e.g., using a digital Butterworth filter or a moving average filter) to remove high-frequency components and noise.
[0035] The steady-state output of a filter is its response characteristic, specifically consisting of in-phase and quadrature components.
[0036] The multi-channel synchronous demodulation in S4 specifically includes: generating an in-phase reference signal and a quadrature reference signal that are synchronized with each frequency component in the multi-frequency quadrature disturbance signal; performing quadrature mixing of the controlled output signal acquired in S3 with the in-phase reference signal and the quadrature reference signal respectively; and performing low-pass filtering on the mixed signal to obtain the response characteristics, which are in-phase components and quadrature components.
[0037] S5: Based on the response characteristics extracted in S4 (i.e. and ), calculate the control input ( ) and controlled output ( A coupling path between () (Multiple frequency response data points)
[0038] The specific process for calculating multiple frequency response data points is as follows: based on the in-phase components Orthogonal components And the injection amplitude of multi-frequency orthogonal disturbance signals in S2 The gain response and phase response of the coupling path at each frequency component of the multi-frequency orthogonal disturbance signal are calculated using the following formulas: Gain response: Phase response: Here, atan2 is the four-quadrant arctangent function, used to ensure the phase is within the specified range. Accuracy within the calculated range. The calculated gain and phase responses. The set of these constitutes multiple frequency response data points.
[0039] The calculation of multiple frequency response data points in S5 specifically includes: calculating the gain response and phase response of the coupling path at each frequency component of the multi-frequency quadrature disturbance signal based on the in-phase component, quadrature component, and the injection amplitude of the multi-frequency quadrature disturbance signal in S2. The gain response and phase response constitute multiple frequency response data points.
[0040] S6: Using multiple frequency response data points calculated in S5, fit a dynamic model of the coupling path. .
[0041] Given the characteristics of thermodynamic systems, the preferred dynamic model is a first-order plus pure time delay model (FOPTD), which specifically corresponds to the dynamic model in S6 being a first-order plus pure time delay model.
[0042] Its transfer function structure is as follows: in, These are the model parameters to be identified.
[0043] The model parameters θij specifically include the coupling path gain Kp,ij, the time constant Tij, and the pure time delay τij. This corresponds to the model parameters including the coupling path gain, time constant, and pure time delay. The specific process of fitting the dynamic model includes constructing an objective function J(θij). The objective function is used to minimize the frequency response of the dynamic model at multiple frequency response data points and the error between the multiple frequency response data points. For example, a least-squares objective function in the complex domain can be constructed as follows: Where Wk is an optional weighting coefficient.
[0044] Solve the objective function using optimization algorithms (such as the Levenberg-Marquardt nonlinear optimization algorithm) to find the solution that makes the objective function more efficient and efficient. Minimum optimal parameter set Thus, the model parameters of the dynamic model are obtained. Specifically, this corresponds to: Fitting the dynamic model in S6 includes: constructing an objective function to minimize the frequency response of the dynamic model at the corresponding frequencies of multiple frequency response data points and the error between the multiple frequency response data points; solving the objective function through an optimization algorithm to obtain the model parameters of the dynamic model.
[0045] S7: Dynamic model fitted using S6 Update the coupling terms in the internal model matrix of S1.
[0046] In S7, the coupling terms in the internal model matrix are updated, specifically by adapting the dynamic model fitted in S6. As the transfer function matrix off-diagonal elements Update.
[0047] The internal model matrix in S1 is a transfer function matrix; in S7, the coupling terms in the internal model matrix are updated by using the dynamic model fitted in S6 as an off-diagonal element of the transfer function matrix.
[0048] S8: Generalized internal model control framework based on the internal model matrix updated in S7 Adjust decoupling compensation.
[0049] The adjustment process is as follows: The GIMC framework reconstructs a decoupling compensator in real time based on the updated internal model matrix. This decoupling compensator (e.g., based on...) The calculated feedforward compensation matrix (or inverse decoupling matrix) is used to feedforward compensate the control input to counteract coupling effects. For example, when the controller needs to increase... At that time, the decoupling compensator will adjust according to the updated coupling model. Automatically calculate a compensatory Adjust the amount to offset. The changes affect the refrigeration circuit.
[0050] The generalized internal model control framework in S8 reconstructs a decoupling compensator in real time based on the updated internal model matrix. The decoupling compensator is used to feedforward the control input to counteract the coupling effect.
[0051] The steps S2 to S8 described above are executed cyclically in the controller, causing the internal model matrix to... It can track the coupling characteristics of the test equipment in real time under different operating conditions (such as different temperature points and different loads), thereby achieving continuous and adaptive multivariable decoupling control.
[0052] Although embodiments of the invention have been shown and described, it will be understood by those skilled in the art that various changes, modifications, substitutions and alterations can be made to these embodiments without departing from the principles and spirit of the invention, the scope of which is defined by the appended claims and their equivalents.
Claims
1. A multivariable decoupling method for a temperature control system of a testing device, characterized in that, Includes the following steps: S1: Establish a generalized internal model control framework, which has an internal model matrix containing a master response term and a coupling term; S2: A set of multi-frequency orthogonal disturbance signals consisting of multiple different frequencies is injected into at least two control inputs of the generalized internal model control framework, the multi-frequency orthogonal disturbance signals being used to detect the coupling characteristics between the control inputs; S3: Acquire at least one controlled output signal from the temperature control system of the test equipment; S4: The controlled output signal acquired by S3 is processed by multi-channel synchronous demodulation to extract the response characteristics of each frequency component of the multi-frequency orthogonal disturbance signal injected by S2 on the controlled output signal. S5: Based on the response characteristics extracted in S4, calculate multiple frequency response data points for a coupling path between the control input and the controlled output; S6: Using the multiple frequency response data points calculated in S5, fit a dynamic model of the coupling path; S7: Using the dynamic model fitted by S6, update the coupling terms in the internal model matrix of S1; S8: The generalized internal model control framework adjusts the decoupling compensation based on the internal model matrix updated in S7.
2. The multivariable decoupling method for a temperature control system of a testing device according to claim 1, characterized in that, The multi-frequency orthogonal disturbance signal in S2 is specifically: Each of the at least two control inputs is configured with a set of disturbance frequencies consisting of multiple different frequencies; The sets of disturbance frequencies from different control inputs are orthogonal to each other and do not contain overlapping frequencies.
3. The multivariable decoupling method for a temperature control system of a testing device according to claim 1, characterized in that, The multi-channel synchronous demodulation in S4 specifically includes: Generate an in-phase reference signal and a quadrature reference signal that are synchronized with each frequency component in the multi-frequency quadrature disturbance signal; The controlled output signal acquired by S3 is orthogonally mixed with the in-phase reference signal and the quadrature reference signal respectively; The mixed signal is low-pass filtered to obtain the response characteristics, which are in-phase and quadrature components.
4. The multivariable decoupling method for a temperature control system of a testing device according to claim 3, characterized in that, The calculation of multiple frequency response data points in S5 specifically includes: Based on the in-phase component, the quadrature component, and the injection amplitude of the multi-frequency quadrature disturbance signal in S2, the gain response and phase response of the coupling path on each frequency component of the multi-frequency quadrature disturbance signal are calculated, and the gain response and the phase response constitute the plurality of frequency response data points.
5. The multivariable decoupling method for a temperature control system of a testing device according to claim 1, characterized in that, The dynamic model in S6 is a first-order plus pure time delay model.
6. The multivariable decoupling method for a temperature control system of a testing device according to claim 1, characterized in that, The fitting of the dynamic model in S6 specifically includes: Construct an objective function to minimize the error between the frequency response of the dynamic model at the corresponding frequencies of the plurality of frequency response data points and the plurality of frequency response data points; The objective function is solved by an optimization algorithm to obtain the model parameters of the dynamic model.
7. The multivariable decoupling method for a temperature control system of a testing device according to claim 6, characterized in that, The model parameters include the gain, time constant, and pure time delay of the coupling path.
8. The multivariable decoupling method for a temperature control system of a testing device according to claim 1, characterized in that, The amplitude of the multi-frequency orthogonal disturbance signal in S2 is set to a preset value, which ensures that the multi-frequency orthogonal disturbance signal does not cause perceptible interference to the main control task of the temperature control system of the test equipment.
9. The multivariable decoupling method for a temperature control system of a testing device according to claim 1, characterized in that, The internal model matrix in S1 is a transfer function matrix; The updating of the coupling term in the internal model matrix in S7 specifically involves: The dynamic model fitted by S6 is used as an off-diagonal element of the transfer function matrix for updating.
10. A multivariable decoupling method for a temperature control system of a testing device according to claim 1, characterized in that, The generalized internal model control framework in S8 reconstructs a decoupling compensator in real time based on the updated internal model matrix. The decoupling compensator is used to perform feedforward compensation on the control input to counteract the coupling effect.