A method for extracting parameters of a GaN HEMT nonlinear current model

Abstract

This invention discloses a method for extracting parameters from a GaN HEMT nonlinear current model, belonging to the field of semiconductor device modeling technology. First, parasitic parameters are removed from the measured current data of the GaN HEMT device to obtain intrinsic current data. Then, based on the intrinsic current data under a first bias condition, a composite fitting strategy of initial value enhancement and cold / hot start is used to extract parameters in the nonlinear current model that are unrelated to self-heating and trapping effects. Further, based on the intrinsic current data under a second bias condition, and combined with the extracted irrelevant parameters, a multi-round fitting and R² evaluation strategy is used to extract parameters related to self-heating and trapping effects from the model. This invention significantly reduces the dependence on initial parameter values ​​through a hybrid fitting strategy, improving the robustness, automation, and overall accuracy of parameter extraction, making it suitable for high-precision GaN HEMT device modeling and circuit design.

Description

Technical Field

[0001] This invention relates to the field of semiconductor device modeling technology, and more specifically to a method for extracting parameters of a GaN HEMT nonlinear current model. Background Technology

[0002] Gallium nitride (GaN) high electron mobility transistors (HEMTs) have become key devices in modern radio frequency and power electronic systems due to their advantages such as high frequency, high power density, and high efficiency. Accurate device models are fundamental to circuit design, and parameter extraction from nonlinear current models is the core and most challenging aspect of modeling. The Angelov model is widely used due to its continuously differentiable equations and good convergence, but it contains a large number of parameters to characterize complex effects such as self-heating and trapping, making parameter extraction extremely difficult.

[0003] Currently, most mainstream parameter extraction methods rely on single optimization algorithms such as least squares. These methods are extremely sensitive to the initial parameter values. Due to the complex characteristics of GaN HEMTs and the presence of noise in experimental data, unreasonable initial values ​​can easily lead to the fitting getting stuck in local optima or failing to converge, making it difficult to obtain a high-precision model in one go. Modeling engineers often need to rely on experience to perform multiple manual adjustments, which seriously affects the automation and efficiency of modeling and increases development costs.

[0004] Therefore, how to propose a robust and highly automated parameter extraction method for GaN HEMT nonlinear current models is a problem that urgently needs to be solved by those skilled in the art. Summary of the Invention

[0005] In view of this, the present invention provides a method for extracting parameters of GaN HEMT nonlinear current model, which solves the problems of unsatisfactory parameter extraction accuracy and low degree of automation in existing algorithms, and guides the modeling of high-precision GaN HEMT devices and the design of high-frequency / high-power circuits.

[0006] To achieve the above objectives, the present invention adopts the following technical solution:

[0007] This invention discloses a method for extracting parameters of a GaN HEMT nonlinear current model, comprising:

[0008] S1. Based on the intrinsic current data under the first bias condition, the self-heating and trap-independent parameters in the GaN HEMT nonlinear current model are extracted using a composite fitting strategy of initial value enhancement for cold and hot start-up.

[0009] S2. Based on the intrinsic current data under the second bias condition, combined with the self-heating and trap-independent parameters, multi-round fitting and R... 2 The evaluation strategy extracts self-heating and trap-related parameters from the GaN HEMT nonlinear current model;

[0010] Specifically, parasitic parameters are removed from the measured current data of the GaN device to obtain the intrinsic current data.

[0011] Preferably, the GaN HEMT nonlinear current model is the Angelov current model.

[0012] Preferably, the measured current data under the first bias condition is as follows: at the static bias point V gsq =0 and V dsq Pulse IV characteristic data obtained by measurement at =0;

[0013] The intrinsic current data under the first bias condition are obtained by removing parasitic parameters from the measured current data under the first bias condition.

[0014] Preferably, S1 includes:

[0015] S11. Regarding the current drain-source voltage V ds At the bias point, perform cold start fitting. If the cold start fitting is successful, select the parameter combination with the best fitting accuracy as the fitting result for the current bias point; otherwise, execute S12.

[0016] S12. Regarding the current drain-source voltage V ds For the bias point, perform a warm-start fitting. If the warm-start fitting is successful, select the parameter combination with the best fitting accuracy index as the fitting result of the current bias point; otherwise, use the preset conservative parameter values ​​as the output of the current bias point.

[0017] S13. Traverse all target leakage power supply bias points, repeat S11 to S12, obtain discrete values ​​of each irrelevant parameter under different biases, process the discrete values ​​to obtain self-heating and trap irrelevant parameters.

[0018] Preferably, S11 includes:

[0019] Regarding the current drain-source voltage V ds Within the preset constraints of each model parameter, multiple sets of initial parameter combinations are randomly generated and used as the starting point of the least squares method to fit the intrinsic current data. The fitting accuracy index corresponding to each set of parameters is calculated. If the fitting accuracy index corresponding to at least one set of initial parameter combinations reaches or exceeds the first preset threshold, the one with the best fitting accuracy index is selected as the fitting result of the current bias point; otherwise, it is determined that the cold start fitting has failed.

[0020] Preferably, S12 includes:

[0021] Regarding the current drain-source voltage V ds The bias point is based on the successfully fitted parameter vector of the previous drain-source voltage bias point. A new initial parameter combination library is generated by adding random noise, scaling or amplifying at least one of these methods. Each set of parameters is then refitted and the fitting accuracy index is calculated. If the fitting accuracy index of at least one set of parameters reaches or exceeds the second preset threshold, the one with the best fitting accuracy index is selected as the fitting result of the current bias point. Otherwise, the hot start fitting is judged as a failure.

[0022] Preferably, the measured current data under the second bias condition is static bias. V gsq Less than the pinch-off voltage V dsq The measured data are pulse current and DCIV data under high bias voltage.

[0023] The intrinsic current data under the second bias condition is obtained by de-embedding the measured current data under the second bias condition.

[0024] Preferably, S2 includes:

[0025] S21. Obtain initial estimates of relevant parameters based on intrinsic current data under the second bias condition;

[0026] S22. Perform multi-algorithm, multi-round fitting and R²-based review to obtain the R² value of each fitting result;

[0027] S23. Select the parameter group whose R² value exceeds the preset R² value from all fitting results, and use it as the extracted value of the relevant parameter;

[0028] S24. Using the extracted values ​​of relevant parameters and irrelevant parameters as the initial set, and combining all intrinsic current data under the first and second bias conditions, the constrained optimization algorithm is used to iteratively optimize all model parameters, and the final parameter set is output.

[0029] Preferably, S22 includes:

[0030] Starting with the initial estimates of relevant parameters and the extracted irrelevant parameters, an initial parameter library is constructed. For each set of initial values ​​in the library, the least squares method, differential evolution algorithm, and simulated annealing algorithm are used sequentially for fitting. The coefficient of determination R² value of each fitting result is calculated and recorded.

[0031] As can be seen from the above technical solutions, compared with the prior art, this invention discloses a method for extracting parameters of a GaN HEMT nonlinear current model, proposing a "cold start-hot start" initial value enhancement mechanism. By setting the parameter constraint range once, the fitting process does not require manual modification of the initial value settings, realizing a one-stop extraction of nonlinear current model parameters from measured parameter input, simulation condition setting to fitting parameter output. In complex multi-parameter model fitting scenarios, this invention uses least squares method, quasi-Newton algorithm and other fitting algorithms to form an algorithm library, and combines R... 2 The dominant accuracy review strategy enables high-precision modeling of nonlinear current models. Attached Figure Description

[0032] To more clearly illustrate the technical solutions in the embodiments of the present invention or the prior art, the drawings used in the description of the embodiments or the prior art will be briefly introduced below. Obviously, the drawings described below are only embodiments of the present invention. For those skilled in the art, other drawings can be obtained based on the provided drawings without creative effort.

[0033] Figure 1 The flowchart shows the model parameter extraction process based on the initial value enhancement cold and hot start fitting strategy.

[0034] Figure 2 For multi-round fitting and R 2 Evaluation strategy flowchart;

[0035] Figure 3 This is a schematic diagram showing the comparison and verification results of the measured and simulated current model. Detailed Implementation

[0036] The technical solutions of the embodiments of the present invention will be clearly and completely described below with reference to the accompanying drawings. Obviously, the described embodiments are only some embodiments of the present invention, and not all embodiments. 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.

[0037] The measured current data of GaN devices includes the influence of parasitic parameters, while the extraction of device current model parameters targets the intrinsic nonlinear current source model of the device. Therefore, in this embodiment, before extracting the device current model parameters, it is necessary to remove parasitic parameters, especially parasitic resistance. To this end, this embodiment uses the parasitic resistance (including gate parasitic resistance Rg, drain parasitic resistance Rd, and source parasitic resistance Rs) removal method from patent ZL202510946845.8 to remove the influence of parasitic parameters from the measured current data and obtain the intrinsic current data.

[0038] In this embodiment, the nonlinear current model used is the Angelov current model, and the model equations are shown in (1)-(6). The model equations contain 22 model parameters to be extracted, which can be divided into parameters that are independent of device self-heating and trap effects and parameters that are related to device self-heating and trap effects according to their properties.

[0039] (1)

[0040] (2)

[0041] (3)

[0042] (4)

[0043] (5)

[0044] (6)

[0045] In the formula, I ds It is the drain-source current; M ipkb The coefficients of the hyperbolic tangent function of the gate-voltage polynomial M ipk Amplitude correction parameters; I pkth It is the current when the transconductance of the device is at its maximum; It is a gate-voltage polynomial; These are coefficients describing the channel length modulation effect; V ds and V gs These are the drain-source and gate-source intrinsic port voltages of the device, respectively. I pkT0 It is a parameter I pkth The value at 25°C K 0 is the temperature coefficient, and ΔT represents the real-time change in channel temperature relative to the extraction temperature of 25°C. V gseff It is the equivalent gate voltage used to describe the trap effect; P 1 、P 2 、P 3 These are the gate voltage polynomial coefficients; V pk1 、V pk2 、V pk3 and V gsmThese are gate voltage parameters related to the transconductance peak value; P n0 , P n1 , alphaP n These are the fitting parameters for the gate voltage polynomial coefficients; M pkb Used to control M ipk The upper bound; Q m These are the coefficients used to calculate the transconductance modulation term; V GSQ It is a static gate voltage bias; V DSQ It is a static leakage voltage bias; a 1 and V gspinch Parameters related to surface traps; a 2 and V dssubs0 Parameters related to the body trap.

[0046] This invention discloses a method for extracting parameters of a GaN HEMT nonlinear current model, comprising the following steps:

[0047] S1. Based on the intrinsic current data under the first bias condition, the self-heating and trap-independent parameters in the GaN HEMT nonlinear current model are extracted using a composite fitting strategy of initial value enhancement for cold and hot start-up.

[0048] To minimize the impact of GaN device self-heating and trapping effects on current characteristics, the measured current data under the first bias condition in this embodiment are selected. V gsq =0, V dsq The pulse IV characteristic data under static bias =0 is combined with the parasitic parameter de-embedding method to remove parasitic resistance and obtain the intrinsic current data under the first bias condition. Then, parameters independent of self-heating and trapping effects are extracted.

[0049] According to formula (2) I pkT0 These are model parameters that are irrelevant and can be directly related to current characteristics, obtained by calculating the drain-source current. I ds The peak transconductance can be used to extract the corresponding current value as... I pkT0 The initial value. For other irrelevant parameters, i.e. M ipkb , Q m , V pk1 ,V pk2 , V pk3 , P 1 (V ds ) , P 2 (V ds ) , P 3 (V ds ) , It is necessary to combine the measured current data of the device with different V ds The transfer characteristic curve below, using the following method Figure 1 The initial value enhancement cold and hot start fitting strategy shown includes:

[0050] S11. Regarding the current drain-source voltage V ds At the bias point, perform a cold start fitting. If the cold start fitting is successful, select the parameter combination with the best fitting accuracy as the fitting result for the current bias point; otherwise, execute S12.

[0051] Specifically, regarding the current drain-source voltage V ds Within the preset constraints of each model parameter, multiple sets of initial parameter combinations are randomly generated and used as the starting point for the least squares method to fit the intrinsic current data. The fitting accuracy index corresponding to each set of parameters is calculated. If the fitting accuracy index corresponding to at least one set of initial parameter combinations reaches or exceeds the first preset threshold, the one with the best fitting accuracy index is selected as the fitting result of the current bias point; otherwise, it is determined that the cold start fitting has failed.

[0052] Before parameter extraction, a constraint range is set for each parameter to be extracted. Based on the constraint range, multiple sets of initial parameter combinations containing small, large, and conservative initial values ​​are randomly generated. Specifically, large initial values ​​correspond to values ​​close to the upper limit within the constraint range, small initial values ​​correspond to values ​​close to the lower limit, and conservative initial values ​​correspond to values ​​near the midpoint of the constraint range. Each initial parameter combination is used as the starting point for the least squares method, and the static bias is adjusted accordingly. V gsq =0, V dsq Fit the measured pulse current data at =0.

[0053] S12. Regarding the current drain-source voltage V dsFor the bias point, perform a warm-start fitting. If the warm-start fitting is successful, select the parameter combination with the best fitting accuracy as the fitting result for the current bias point; otherwise, use the preset conservative parameter values ​​as the output for the current bias point.

[0054] Hot start fitting includes: with the current V ds The previous bias point V ds Based on the successfully fitted parameter vector of the bias point, a new initial parameter combination library is generated by adding random noise, scaling or amplifying at least one of the following methods (see Table 1 for specific calculation methods).

[0055] Table 1. Method for constructing the initial parameter library for hot start ( vec (For the original parameter set)

[0056]

[0057] Based on this, starting from each initial parameter combination in the initial value library, the least squares method is used to apply the static bias as... V gsq =0, V dsq Fit the measured pulse current data at =0.

[0058] If at least one set of parameters in the warm-start process has a fitting accuracy index that reaches or exceeds the second preset threshold, then the one with the best fitting accuracy index is selected as the fitting result of the current bias point; otherwise, the warm-start fitting is judged as a failure, and the preset conservative parameter value is used as the output of the current bias point.

[0059] When both "cold start" and "warm start" fail, it means that... V ds The measured data quality is poor under the bias, making parameter acquisition difficult. Considering that the failure of parameter extraction under individual biases has a limited impact on the overall process, the fallback mechanism will ultimately select a conservative initial value that meets the constraints as the final fitting result under that bias condition.

[0060] S13. Traverse all target drain-source voltage bias points, repeat S11 to S12, obtain discrete values ​​of each irrelevant parameter under different biases, process the discrete values ​​to obtain self-heating and trap irrelevant parameters.

[0061] In this embodiment, for M ipkb , Q m , V pk1 , V pk2 , Vpk3 , The parameter values ​​are determined by averaging; for parameters that include a drain bias-related sub-model... P 1 (V ds ) , P 2 (V ds ) , P 3 (V ds ) The least squares method was used to extract the bias correlation fitting.

[0062] S2. Based on the intrinsic current data under the second bias condition, combined with self-heating and trap-independent parameters, multi-round fitting and R... 2 The evaluation strategy extracts self-heating and trapping-related parameters from the GaN HEMT nonlinear current model.

[0063] For parameters related to self-heating and trapping effects, this embodiment first adopts the extraction method in patent ZL201410800156.8, combined with measured current data under the second bias condition: static bias. V gsq Less than the pinch-off voltage V dsq Using measured pulse current data at high bias voltage and measured DCIV data, intrinsic current data under the second bias condition were obtained after removing parasitic resistance using the parasitic parameter de-embedding method. This allowed for the acquisition of parameters related to self-heating and trapping effects in the current model. K 0 、a 1 、a 2 、 V gspinch 、V dssubs0 The initial value of .

[0064] After obtaining initial values ​​for parameters related to self-heating and trapping effects, this embodiment uses R-based... 2 The comprehensive fitting strategy for evaluation completes the extraction of parameters related to self-heating and trapping effects, such as... Figure 2 As shown, it includes the following steps:

[0065] S21. Based on the intrinsic current data under the second bias condition, obtain the initial estimates of the relevant parameters, i.e. K 0 、a 1、 a 2 、V gspinch 、V dssubs0 The initial value of .

[0066] S22. Perform multi-algorithm, multi-round fitting and R²-driven review, including: using the initial estimates of relevant parameters and the irrelevant parameters extracted in S1 as the starting point to construct an initial parameter library; for each set of initial values ​​in the library, use the least squares method, differential evolution algorithm and simulated annealing algorithm to fit in sequence, and calculate and record the coefficient of determination R² value of each fitting result.

[0067] The initial estimates of relevant parameters and the self-heating and trap-independent parameters extracted from S1 are used as initial values ​​for optimization. Within the preset parameter range, an initial value library is constructed by adding random noise as shown in Table 1. A multi-round fitting iteration method is adopted, that is, each round of fitting process starts from an initial value in the initial value library, traverses all fitting algorithms specified in the algorithm library, and records the fitting result and coefficient of determination (R²) of each algorithm. 2 The value is calculated using formula (7). The algorithm library includes least squares, differential evolution, and simulated annealing algorithms.

[0068] (7)

[0069] Where RSS represents the residual sum of squares and TSS represents the total sum of squares.

[0070] After the fitting process for a single initial value is completed, the R-squared of the fitting round result is calculated. 2 .

[0071] S23. Select the parameter group whose R² value exceeds the preset R² value from all fitting results, and use it as the extracted value of the relevant parameter.

[0072] After multiple rounds of comprehensive fitting, R0 was selected from the fitting results. 2 The result closest to 1 is used as the final set of model parameters.

[0073] S24. Using the irrelevant parameters and the relevant parameters extracted in step S23 as the initial set, and combining all intrinsic current data under the first and second bias conditions, the constrained optimization algorithm is used to iteratively optimize all model parameters, and the final parameter set is output.

[0074] After completing the above extraction steps, the extracted parameters independent of self-heating and trapping effects, as well as the initial values ​​of parameters related to self-heating and trapping effects, are used as the initial parameter set for global optimization, combined with the static bias. V gsq =0,V dsq =0、 V gsq Less than the pinch-off voltage, V dsq Using measured pulse current data with high bias voltage and measured DCIV data as input, global optimization is performed on all current model parameters.

[0075] In the global optimization, the iteration rounds and accuracy requirements for the comprehensive optimization are first set, and the optimization range of each model parameter is the same as the parameter constraint range used in S1 and S2. Simultaneously, R is updated. 2 The evaluation uses a comprehensive fitting strategy algorithm library, setting least squares, constrained least squares, and constrained finite-memory quasi-Newton algorithm as the algorithms used in each iteration. In each iteration, starting from the initial value, each fitting algorithm in the library is traversed to obtain the highest R-value. 2 The results corresponding to the values ​​are output as the final fitting parameters.

[0076] This embodiment verifies the current model parameter extraction algorithm proposed in this invention based on measured current data of a 0.1μm GaN HEMT device with a size of 4×30μm.

[0077] The measured gate voltage bias range of the device is -6.0V to 0.0V with a step size of 0.2V, and the drain bias voltage range is 0.0V to 20.0V with a step size of 0.5V. After obtaining the current model parameters using the model parameter extraction algorithm proposed in this invention, the parameters are substituted into the Angelov model of equations (1) to (6) for simulation and measurement comparison verification. In addition, in order to evaluate the advantages of this algorithm and the traditional current model parameter extraction algorithm, this embodiment uses the current model parameter extraction algorithm in patent ZL201410800156.8 as the traditional algorithm and performs corresponding current model parameter extraction for comparison. The simulation and measurement comparison verification results of the DCIV of the model are as follows. Figure 3 As shown in the figure, the red solid line represents the verification results corresponding to the model parameters obtained by the parameter extraction algorithm proposed in this invention, the orange dashed line represents the verification results corresponding to the model parameters obtained based on the traditional algorithm, and the blue circle represents the measured data.

[0078] like Figure 3 As shown, compared to traditional algorithms, the proposed method integrates initial value enhancement cold-hot start and R... 2 The current model extracted by the algorithm for evaluating the fitting strategy can accurately predict the current characteristics under different biases, especially the current characteristics in the linear and saturation regions.

[0079] To evaluate the accuracy improvement brought by the present invention compared with the traditional algorithm, this embodiment uses the accuracy calculation method based on the root mean square error of RMSE using formula (8), and maps the error reflected by RMSE to a percentage system. The results show that the accuracy of the parameter extraction algorithm of the present invention reaches 97.95%, while the accuracy of the traditional algorithm is only 77.30%, an improvement of 20.65%, which further verifies the effectiveness of the current model parameter extraction algorithm in this embodiment.

[0080] (8)

[0081] Where score is the percentage accuracy obtained using the root mean square error method, and y_meas is the measured data for comparison.

[0082] The various embodiments in this specification are described in a progressive manner, with each embodiment focusing on its differences from other embodiments. Similar or identical parts between embodiments can be referred to interchangeably. For the apparatus disclosed in the embodiments, since they correspond to the methods disclosed in the embodiments, the description is relatively simple; relevant parts can be referred to the method section.

[0083] The above description of the disclosed embodiments enables those skilled in the art to make or use the invention. Various modifications to these embodiments will be readily apparent to those skilled in the art, and the general principles defined herein may be implemented in other embodiments without departing from the spirit or scope of the invention. Therefore, the invention is not to be limited to the embodiments shown herein, but is to be accorded the widest scope consistent with the principles and novel features disclosed herein.

Claims

1. A method for extracting parameters of a GaN HEMT nonlinear current model, characterized in that, include: S1. Based on the intrinsic current data under the first bias condition, the self-heating and trap-independent parameters in the GaN HEMT nonlinear current model are extracted using a composite fitting strategy of initial value enhancement for cold and hot start-up, including: S11. Regarding the current drain-source voltage V ds At the bias point, perform cold start fitting. If the cold start fitting is successful, select the parameter combination with the best fitting accuracy as the fitting result for the current bias point; otherwise, execute S12. Among them, regarding the current drain-source voltage V ds Within the preset constraints of each model parameter, multiple sets of initial parameter combinations are randomly generated and used as the starting point of the least squares method to fit the intrinsic current data. The fitting accuracy index corresponding to each set of parameters is calculated. If the fitting accuracy index corresponding to at least one set of initial parameter combinations reaches or exceeds the first preset threshold, the one with the best fitting accuracy index is selected as the fitting result of the current bias point; otherwise, it is determined that the cold start fitting has failed. S12. Regarding the current drain-source voltage V ds For the bias point, perform a warm-start fitting. If the warm-start fitting is successful, select the parameter combination with the best fitting accuracy index as the fitting result of the current bias point; otherwise, use the preset conservative parameter values ​​as the output of the current bias point. Among them, regarding the current drain-source voltage V ds The bias point is based on the successfully fitted parameter vector of the previous drain-source voltage bias point. A new initial parameter combination library is generated by adding random noise, scaling or amplifying at least one of these methods. Each set of parameters is then refitted and the fitting accuracy index is calculated. If the fitting accuracy index of at least one set of parameters reaches or exceeds the second preset threshold, the one with the best fitting accuracy index is selected as the fitting result of the current bias point. Otherwise, the hot start fitting is judged as a failure. S13. Traverse all target leakage voltage power supply bias points, repeat S11 to S12, obtain discrete values ​​of each irrelevant parameter under different biases, process the discrete values ​​to obtain self-heating and trap irrelevant parameters. S2. Based on the intrinsic current data under the second bias condition, combined with the self-heating and trap-independent parameters, multi-round fitting and R... 2 The evaluation strategy extracts self-heating and trap-related parameters from the GaN HEMT nonlinear current model; Specifically, parasitic parameters are removed from the measured current data of the GaN device to obtain the intrinsic current data.

2. The method for extracting parameters of a GaN HEMT nonlinear current model according to claim 1, characterized in that, The GaN HEMT nonlinear current model is the Angelov current model.

3. The method for extracting parameters of a GaN HEMT nonlinear current model according to claim 1, characterized in that, The measured current data under the first bias condition are: at the static bias point V gsq =0 and V dsq Pulse IV characteristic data obtained by measurement at =0; The intrinsic current data under the first bias condition are obtained by removing parasitic parameters from the measured current data under the first bias condition.

4. The method for extracting parameters of a GaN HEMT nonlinear current model according to claim 1, characterized in that, The measured current data under the second bias condition is for static bias. V gsq Less than the pinch-off voltage V dsq The measured data are pulse current and DCIV data under high bias voltage. The intrinsic current data under the second bias condition is obtained by de-embedding the measured current data under the second bias condition.

5. The method for extracting parameters of a GaN HEMT nonlinear current model according to claim 4, characterized in that, S2 include: S21. Obtain initial estimates of relevant parameters based on intrinsic current data under the second bias condition; S22. Perform multi-algorithm, multi-round fitting and R²-based review to obtain the R² value of each fitting result; S23. Select the parameter group whose R² value exceeds the preset R² value from all fitting results, and use it as the extracted value of the relevant parameter; S24. Using the extracted values ​​of relevant parameters and irrelevant parameters as the initial set, and combining all intrinsic current data under the first and second bias conditions, a constrained optimization algorithm is used to iteratively optimize all model parameters, and the final parameter set is output.

6. The method for extracting parameters of a GaN HEMT nonlinear current model according to claim 5, characterized in that, S22 includes: Starting with the initial estimates of relevant parameters and the extracted irrelevant parameters, an initial parameter library is constructed. For each set of initial values ​​in the library, the least squares method, differential evolution algorithm, and simulated annealing algorithm are used sequentially for fitting. The coefficient of determination R² value of each fitting result is calculated and recorded.

Images