CN-122021548-A - GaN HEMT nonlinear current model parameter extraction method

CN122021548ACN 122021548 ACN122021548 ACN 122021548ACN-122021548-A

Abstract

The invention discloses a method for extracting parameters of a GaN HEMT nonlinear current model, and belongs to the technical field of semiconductor device modeling. The method comprises the steps of firstly, carrying out parasitic parameter de-embedding on actual measurement current data of a GaN HEMT device to obtain intrinsic current data, then extracting parameters irrelevant to self-heating and trap effects in a nonlinear current model by adopting a composite fitting strategy of initial value enhanced cold and hot starting based on the intrinsic current data under a first bias condition, and extracting parameters relevant to the self-heating and trap effects in the model by adopting a multi-round fitting and R2 evaluation strategy based on the intrinsic current data under a second bias condition and combining the extracted irrelevant parameters. The method obviously reduces the dependence on the initial value of the parameter through the hybrid fitting strategy, improves the robustness, the degree of automation and the overall precision of parameter extraction, and is suitable for modeling and circuit design of a high-precision GaN HEMT device.

Inventors

MAO SHUMAN
YAO XIANGYU
XU YUEHANG

Assignees

电子科技大学长三角研究院(湖州)

Dates

Publication Date: 20260512
Application Date: 20260415

Claims (9)

1. The method for extracting the nonlinear current model parameters of the GaN HEMT is characterized by comprising the following steps of: s1, extracting self-heating and trap independent parameters in a GaN HEMT nonlinear current model by utilizing a composite fitting strategy of initial value enhanced cold and hot starting based on intrinsic current data under a first bias condition; S2, based on intrinsic current data under a second bias condition, combining the self-heating and trap independent parameters, and extracting self-heating and trap related parameters in a GaN HEMT nonlinear current model by utilizing a multi-round fitting and R 2 evaluation strategy; and performing parasitic parameter de-embedding on the actually measured current data of the GaN device to acquire the intrinsic current data.
2. The method for extracting parameters of a nonlinear current model of a GaN HEMT according to claim 1, wherein the nonlinear current model of the GaN HEMT is Angelov.
3. The method for extracting the nonlinear current model parameters of the GaN HEMT according to claim 1 is characterized in that the measured current data under the first bias condition is pulse I-V characteristic data obtained by measuring under the static bias point V gsq = 0 and V dsq = 0; and carrying out parasitic parameter de-embedding on the actually measured current data under the first bias condition to obtain intrinsic current data under the first bias condition.
4. A method for extracting nonlinear current model parameters of a GaN HEMT according to claim 3, wherein S1 comprises: S11, aiming at the bias point of the current drain-source voltage V ds , performing cold start fitting, if the cold start fitting is successful, selecting a parameter combination with the optimal fitting precision index as a fitting result of the current bias point, otherwise, performing S12; S12, performing hot start fitting aiming at the bias point of the current drain-source voltage V ds , if the hot start fitting is successful, selecting a parameter combination with the optimal fitting precision index as a fitting result of the current bias point, otherwise, adopting a preset conservative parameter value as output of the current bias point; S13, traversing all target leakage voltage source bias points, repeatedly executing S11 to S12 to obtain discrete values of all independent parameters under different biases, and processing the discrete values to obtain self-heating and trap independent parameters.
5. The method for extracting nonlinear current model parameters of a GaN HEMT according to claim 4, wherein S11 comprises: and randomly generating a plurality of groups of initial parameter combinations in a preset constraint range of each model parameter aiming at the bias point of the current drain-source voltage V ds , respectively fitting the intrinsic current data as a starting point of a least square method, and calculating a fitting precision index corresponding to each group of parameters, wherein if the fitting precision index corresponding to at least one group of initial parameter combinations reaches or exceeds a first preset threshold, the optimal fitting precision index is selected as a fitting result of the current bias point, otherwise, judging that cold start fitting fails.
6. The method for extracting nonlinear current model parameters of a GaN HEMT according to claim 4, wherein S12 comprises: Aiming at the current drain-source voltage V ds bias point, taking a successful fitting parameter vector of the last drain-source voltage bias point of the current drain-source voltage bias point as a reference, generating a new initial parameter combination library by adding random noise, scaling or amplifying at least one mode, respectively carrying out fitting again on each group of parameters and calculating a fitting precision index, if the fitting precision index corresponding to at least one group of parameters reaches or exceeds a second preset threshold, selecting the optimal fitting precision index from the fitting precision indexes as a fitting result of the current bias point, otherwise, judging that the hot start fitting fails.
7. The method for extracting parameters of a nonlinear current model of a GaN HEMT according to claim 1, wherein the measured current data under the second bias condition is pulse current measured data with static bias V gsq smaller than pinch-off voltage and V dsq high bias voltage and DCIV measured data; and de-embedding the actually measured current data under the second bias condition to obtain intrinsic current data under the second bias condition.
8. The method for extracting nonlinear current model parameters of a GaN HEMT according to claim 7, wherein S2 comprises: S21, acquiring initial estimated values of related parameters based on intrinsic current data under a second bias condition; S22, performing multi-algorithm multi-round fitting and R2 leading examination to obtain a decision coefficient R2 value of each fitting result; s23, selecting parameter groups with R2 values exceeding a preset R2 value in all fitting results as extraction values of related parameters; S24, taking the extracted values of the related parameters and the irrelevant parameters as an initial set, carrying out iterative optimization on all model parameters by adopting a constraint optimization algorithm in combination with all intrinsic current data under the first bias condition and the second bias condition, and outputting a final parameter set.
9. The method for extracting nonlinear current model parameters of a GaN HEMT according to claim 8, wherein S22 comprises: And for each group of initial values in the library, fitting by adopting a least square method, a differential evolution algorithm and a simulated annealing algorithm in sequence, calculating and recording the value of a decision coefficient R2 for obtaining the fitting result each time.

Description

GaN HEMT nonlinear current model parameter extraction method Technical Field The invention relates to the technical field of semiconductor device modeling, in particular to a GaN HEMT nonlinear current model parameter extraction method. Background Gallium nitride (GaN) High Electron Mobility Transistors (HEMTs) have become key devices for modern radio frequency and power electronic systems due to their high frequency, high power density, and high efficiency. The accurate device model is the basis of circuit design, and the parameter extraction of the nonlinear current model is the core and difficulty of modeling. The Angelov model is widely adopted because of its continuous equation and good convergence, but it contains a large number of parameters to characterize the complex effects of self-heating, trapping, etc., making parameter extraction extremely difficult. The current mainstream parameter extraction method mostly depends on a single optimization algorithm such as a least square method. Such methods are extremely sensitive to the initial value settings of the parameters. Because GaN HEMT characteristic is complicated, actual measurement data have noise, unreasonable initial values are very easy to cause fitting to be trapped into local optimum or can not converge, and a high-precision model is difficult to obtain at one time. Often, modeling engineers need to manually debug for many times by experience, which seriously affects the automation degree and efficiency of modeling and increases the research and development cost. Therefore, how to provide a parameter extraction method with strong robustness and high automation degree for a nonlinear current model of a GaN HEMT is a problem to be solved by those skilled in the art. Disclosure of Invention In view of the above, the invention provides a method for extracting parameters of a GaN HEMT nonlinear current model, which is used for solving the problems of non-ideal parameter extraction precision, low automation degree and the like of the existing algorithm and guiding high-precision GaN HEMT device modeling and high-frequency/high-power circuit design. In order to achieve the above purpose, the present invention adopts the following technical scheme: The invention discloses a method for extracting parameters of a GaN HEMT nonlinear current model, which comprises the following steps: s1, extracting self-heating and trap independent parameters in a GaN HEMT nonlinear current model by utilizing a composite fitting strategy of initial value enhanced cold and hot starting based on intrinsic current data under a first bias condition; S2, based on intrinsic current data under a second bias condition, combining the self-heating and trap independent parameters, and extracting self-heating and trap related parameters in a GaN HEMT nonlinear current model by utilizing a multi-round fitting and R 2 evaluation strategy; and performing parasitic parameter de-embedding on the actually measured current data of the GaN device to acquire the intrinsic current data. Preferably, the GaN HEMT nonlinear current model is Angelov current model. Preferably, the measured current data under the first bias condition is pulse I-V characteristic data obtained by measuring at a static bias point V gsq =0 and V dsq =0; and carrying out parasitic parameter de-embedding on the actually measured current data under the first bias condition to obtain intrinsic current data under the first bias condition. Preferably, S1 comprises: S11, aiming at the bias point of the current drain-source voltage V ds, performing cold start fitting, if the cold start fitting is successful, selecting a parameter combination with the optimal fitting precision index as a fitting result of the current bias point, otherwise, performing S12; S12, performing hot start fitting aiming at the bias point of the current drain-source voltage V ds, if the hot start fitting is successful, selecting a parameter combination with the optimal fitting precision index as a fitting result of the current bias point, otherwise, adopting a preset conservative parameter value as output of the current bias point; S13, traversing all target leakage voltage source bias points, repeatedly executing S11 to S12 to obtain discrete values of all independent parameters under different biases, and processing the discrete values to obtain self-heating and trap independent parameters. Preferably, S11 includes: and randomly generating a plurality of groups of initial parameter combinations in a preset constraint range of each model parameter aiming at the bias point of the current drain-source voltage V ds, respectively fitting the intrinsic current data as a starting point of a least square method, and calculating a fitting precision index corresponding to each group of parameters, wherein if the fitting precision index corresponding to at least one group of initial parameter combinations reaches or exceeds a first preset threshold, the optim