A parameter extraction method for GaN HEMT based on ASM model

By grouping GaN HEMT devices and using a trust-region algorithm optimizer, the problem of parameter extraction under temperature influence is solved, achieving efficient and accurate parameter extraction and model reuse, thus improving the efficiency and accuracy of parameter extraction for GaN HEMT devices.

CN115204089BActive Publication Date: 2024-11-26CHENGDU HUADA JIUTIAN TECH CO LTD
View PDF 0 Cites 0 Cited by

Patent Information

Application Number
CN202210814388.3
Authority / Receiving Office
CN · China
Patent Type
Patents(China)
Current Assignee / Owner
Filing Date
2022-07-12
Publication Date
2024-11-26
Estimated Expiration
2042-07-12

AI Technical Summary

Technical Problem

Existing GaN HEMT parameter extraction methods do not consider the effect of device temperature, resulting in low extraction accuracy and efficiency, and the model is difficult to reuse for other devices.

Method used

GaN HEMT devices are grouped according to different sizes and temperatures. A trust-region algorithm optimizer is used to extract parameters. First, the capacitance-voltage relationship curve is extracted, and then the current-voltage relationship curve is extracted. The ASM model is used to save and apply the data to devices of the same size at different temperatures.

Benefits of technology

It improves the accuracy and efficiency of parameter extraction, reduces root mean square error, enhances the flexibility and convenience of the model, and saves parameter extraction time.

✦ Generated by Eureka AI based on patent content.

Smart Images

  • Figure CN115204089B_ABST
    Figure CN115204089B_ABST
Patent Text Reader

Abstract

The present application discloses a parameter extraction method, device and computer-readable storage medium of GaN HEMT based on ASM model. The method includes: setting the process parameters of the GaN HEMT; dividing more than one GaN HEMT device into one or more groups based on different sizes and temperatures, wherein the GaN HEMT devices in each group have the same size and different temperatures, and for each group of GaN HEMT devices, first selecting the first GaN HEMT device with the first size S1 and the first temperature T1, adjusting different input parameter values ​​to obtain multiple measurement data output by the first GaN HEMT device, and generating a measurement curve based on the measurement data, wherein the measurement curve includes a capacitance-voltage relationship curve; extracting relevant parameters of the capacitance-voltage relationship curve; saving the ASM model of the first GaN HEMT device; applying the ASM model to GaN HEMT devices with the same size S1 and other temperatures, and continuing to extract relevant parameters of the capacitance-voltage relationship curve.
Need to check novelty before this filing date? Find Prior Art

Description

Technical Field

[0001] The invention relates to the technical field of microelectronic integrated circuits, in particular to a parameter extraction method of a GaN HEMT. Background Art

[0002] Si-based power devices have reached their device limits. GaN materials are used as substrate materials for high electron mobility transistor (HEMT) devices due to their excellent device properties (wide bandgap, high saturation velocity, high thermal stability and high current density). GaN HEMT has demonstrated its excellent properties in the application of high-power microwave integrated circuits, including high breakdown voltage, high operating frequency and high power density. ASM-HEMT has been recognized by the Compact Model Consortium (CMC) as a new standard model for GaN devices. In addition to large signal modeling, ASM-HEMT has achieved accurate modeling of field plate capacitance, trap effect, Kink effect and noise.

[0003] Usually, in the parameter extraction of GaN HEMT, the influence of device temperature is not considered, which reduces the accuracy of parameter extraction. In addition, optimization is to make the objective function converge to the minimum point. The traditional optimization algorithm is to continuously adjust the search direction and then perform a one-dimensional search in the search direction. It has also achieved very practical engineering significance, but there are low efficiency and limited accuracy. In the past, the models after parameter extraction were rarely reused. The extracted models were only applied to the current device and were not very practical for other devices. Therefore, improvements are needed to meet the requirements for better parameter extraction of GaN HEMT. Summary of the invention

[0004] In order to improve the efficiency, convenience and accuracy of parameter extraction of GaN HEMT, the versatility of parameter extraction model of GaN HEMT, and adapt to actual work needs, the present invention provides a parameter extraction method of GaN HEMT based on ASM model, comprising: setting process parameters of the GaN HEMT, wherein the process parameters include one or more of size parameters, functional parameters, and working environment parameters;

[0005] The more than one GaN HEMT devices are divided into one or more groups based on different sizes and temperatures, wherein the GaN HEMT devices in each group have the same size and different temperatures, respectively.

[0006] For each group of GaN HEMT devices, the following parameter extraction is performed:

[0007] Select the first size S 1 , the first temperature T 1a first GaN HEMT device, adjusting different input parameter values ​​to obtain a plurality of measurement data output by the first GaN HEMT device, and generating a measurement curve based on the measurement data, wherein the measurement curve includes a capacitance-voltage relationship curve;

[0008] Extract relevant parameters of capacitance-voltage relationship curve;

[0009] Save the ASM model of the 1st GaN HEMT device;

[0010] Apply the ASM model to the 1 The relevant parameters of the capacitance-voltage relationship curve are extracted from GaN HEMT devices at other temperatures.

[0011] In an implementable manner, the measurement curve further includes a current-voltage relationship curve, and the parameter extraction method further includes extracting relevant parameters of the current-voltage relationship curve.

[0012] In an implementable manner, the process parameters include: device gate length L, device gate width W, operating temperature T, device gate index NF, device gate-source length LSG, and device drain-source length LDG.

[0013] In one implementable manner, extracting parameters related to the capacitance-voltage relationship curve includes: extracting related parameters through a reverse conduction capacitance-drain voltage curve, an output capacitance-drain voltage curve, an input capacitance-drain voltage curve, and a gate capacitance-gate voltage curve.

[0014] In one achievable manner, extracting relevant parameters of the current-voltage relationship curve includes first locally optimizing the parameters and then globally optimizing the parameters, wherein the local optimization includes:

[0015] 1) For the drain current-gate voltage curve id_vg, extract the threshold voltage voff and subthreshold slope nfactor in the low current region of low source-drain voltage Vds;

[0016] 2) For the drain current-gate voltage curve id_vg, extract the low-field mobility parameter u0, mobility attenuation coefficient ua, and mobility second-order attenuation coefficient ub in the low source-drain voltage Vds high current region. ;

[0017] 3) For the drain current-drain voltage curve id_vd, extract the DIBL parameters and subthreshold attenuation parameters in the high source-drain voltage Vds low current region;

[0018] 4) For the leakage current-drain voltage curve id_vd, extract the speed saturation parameter in the high source-drain voltage Vds high current region,

[0019] The parameters obtained through the above local optimization are fine-tuned in the global optimization.

[0020] In one implementable manner, the above local optimization and global optimization are performed based on an optimizer using a trust-region algorithm, and the optimizer directly performs a multi-dimensional search in a region to find the optimal value.

[0021] In one implementation, during the optimization process, when the optimized parameter reaches a boundary value, the optimizer automatically expands the boundary.

[0022] In one implementable manner, the optimization interval is customized using a script, and a function for specifying the range of the optimization interval is called in the script to select an area on the measurement curve graph where parameters need to be extracted.

[0023] In one possible implementation, extracting capacitance-voltage curve parameters includes:

[0024] Extract drain-source capacitance parameters, edge capacitance parameters, bias voltage parameters, drain saturation voltage parameters, and bias voltage parameters from the reverse conduction capacitance-drain voltage (crss-vds) curve;

[0025] Extract the drain-source capacitance parameters and drain edge capacitance parameters from the output capacitance-drain voltage (coss_vds) curve;

[0026] Extract gate-source overlap capacitance parameters from input capacitance-drain voltage (ciss_vds) curve;

[0027] Extract AlGaN layer thickness parameters from gate capacitance_gate voltage (cgg_vgs) curve

[0028] The present invention also provides a GaN HEMT parameter extraction device based on the ASM model, characterized in that it includes a memory and a processor, the memory stores a program running on the processor, and the processor executes the aforementioned GaN HEMT parameter extraction method based on the ASM model when running the program.

[0029] The present invention also provides a computer-readable storage medium having computer instructions stored thereon, wherein the computer instructions, when executed by a processor, execute the aforementioned parameter extraction method of GaN HEMT based on the ASM model.

[0030] Compared with the prior art, the GaN parameter extraction method of the present invention has the following advantages and positive effects:

[0031] 1) Considering the influence of temperature, GaN HEMT devices are grouped according to different sizes and temperatures, and parameter extraction is performed separately, which improves the accuracy of parameter extraction.

[0032] 2) The optimizer "Trust region" defined by the trust region algorithm in the present invention has greatly reduced the root mean square error (RMS) compared with the existing optimizer; it can also automatically expand the boundary and find the optimal value when the region cannot meet the conditions; at the same time, it has a default optimization iteration number with the best effect and the least time consumption. This makes the optimization process fast and efficient, and improves the efficiency and accuracy of parameter extraction.

[0033] 3) By calling functions to customize the optimization interval, it can meet the different needs of users, has strong flexibility and convenience, and has practical engineering significance.

[0034] 4) First extract the capacitance-voltage (cv) curve, then extract the current-voltage (iv) curve. At the same time, the current-voltage (iv) curve adopts the strategy of first local optimization and then global optimization, which makes the extraction more targeted and increases the efficiency and accuracy of the extraction;

[0035] 5) The ASM model after parameter extraction can be saved and applied to devices of the same size but at different temperatures. In this way, only some parameters and curves need to be adjusted for the remaining devices to achieve the effect of parameter extraction for the entire model, which greatly saves the time for parameter extraction and increases the efficiency of parameter extraction, and has strong practical engineering significance. BRIEF DESCRIPTION OF THE DRAWINGS

[0036] Figure 1 A GaN HEMT parameter extraction flow chart based on the ASM model proposed in the present invention;

[0037] Figure 2 It is the simulation and measurement value image after the reverse conduction capacitance-drain voltage (crss_vd) curve parameter extraction is completed in the present invention;

[0038] Figure 3 It is the simulation and measurement value image after the output capacitance-drain voltage (coss_vd) curve parameter extraction is completed in the present invention;

[0039] Figure 4 It is the simulation and measurement value image after the input capacitance-drain voltage (ciss_vd) curve parameter extraction is completed in the present invention;

[0040] Figure 5 It is the simulation and measurement value image after the gate capacitance_gate voltage (cgg_vg) curve parameter extraction is completed in the present invention;

[0041] Figure 6 It is the simulation and measurement value image after the extraction of the drain current-gate voltage (id_vg) curve parameters in the present invention is completed;

[0042] Figure 7 It is the simulation and measurement value image after the extraction of the drain current-gate voltage (id_vg) curve parameters in the present invention is completed;

[0043] Figure 8 It is the simulation and measurement value image after the extraction of the drain current-drain voltage (id_vd) curve parameters in the present invention is completed;

[0044] Fig. 9 It is the simulation and measurement value images after the reverse conduction capacitance-drain voltage (crss_vd) curve parameter extraction is completed in the prior art;

[0045] Fig.10 The simulation and measurement value images after the output capacitance-drain voltage (coss_vd) curve parameters are extracted in the prior art;

[0046] Fig.11 The simulation and measurement value images after the input capacitance-drain voltage (ciss_vd) curve parameters are extracted in the prior art;

[0047] Fig.12 It is the simulation and measurement value image after the gate capacitance_gate voltage (cgg_vg) curve parameter extraction is completed in the prior art;

[0048] Fig.13 It is the simulation and measurement value image after the parameters of drain current-gate voltage (id_vg) curve are extracted in the prior art;

[0049] Fig.14 It is the simulation and measurement value image after the parameters of drain current-gate voltage (id_vg) curve are extracted in the prior art;

[0050] Fig.15 It is the simulation and measurement value images after the parameters of the drain current_drain voltage (id_vd) curve are extracted in the prior art; DETAILED DESCRIPTION

[0051] The following will be combined with the drawings in the embodiments of the present application to clearly describe the technical solutions in the embodiments of the present application. Obviously, the described embodiments are part of the embodiments of the present application, rather than all the embodiments. All other embodiments obtained by ordinary technicians in this field based on the embodiments in the present application belong to the scope of protection of this application.

[0052] The terms "first", "second", etc. in the specification and claims of the present application are used to distinguish similar objects, but not to describe a specific order or sequence. It should be understood that the data used in this way can be interchanged where appropriate, so that the embodiments of the present application can be implemented in an order other than those illustrated or described here, and the objects distinguished by "first", "second", etc. are generally of one type, and the number of objects is not limited. For example, the first object can be one or more. In addition, "and / or" in the specification and claims represents at least one of the connected objects, and the character " / " generally indicates that the objects associated with each other are in an "or" relationship. An accurate large signal model is of great significance to the circuit design of the device. The large signal device model is the basis for the characterization of microwave device characteristics and circuit design optimization. The nonlinear current-voltage (IV) model is the core of the device large signal model, and the extraction of its model parameters is the basis for the device large signal modeling. A fast and accurate parameter extraction method can not only improve the efficiency of modeling, but also shorten the circuit design cycle. The present invention proposes an effective parameter extraction method for GaN HEMT.

[0053] Figure 1 The flowchart of the GaN HEMT parameter extraction method provided by the present invention is as follows: Figure 1 The GaN HEMT parameter extraction method of the present invention is described in detail.

[0054] In step 101, process parameters of the GaN HEMT device are set.

[0055] The process parameters of the GaN HEMT device set in the present invention include: one or more of the device size parameters, functional parameters, and working environment parameters such as device gate length (L), device gate width (W), operating temperature (T), device gate index (NF), device gate-source length (LSG), and device drain-source length (LDG).

[0056] In step 102, more than one GaN HEMT device is divided into one or more groups based on different parameters such as size and temperature.

[0057] Specifically, for example, according to the different sizes S of GaN HEMT devices 1 , S 2 , S 3 …dividing GaN HEMT devices into multiple groups, where GaN HEMT devices of the same size have different temperatures T 1 , T 2 , T 3 …, thus obtaining multiple groups of GaN HEMT devices: G 1 (S 1 , T 1 T2 T 3… ),G 2 (S 2 , T 1 T 2 T 3… ),…The temperatures in different groups may be the same, partially the same, or completely different.

[0058] Refer to Table 1, Group 1 Device G 1 In the figure, the GaN HEMT devices W1000T-40, W1000T150, and W1000T25 have the same size (w = 100u, l = 0.6u), and the temperatures of the three are -40°, 150°, and 25°, respectively. 2 In the figure, GaN HEMT devices W100T-40, W100T150, and W100T25 have the same size (w = 25u, l = 0.6u), and their temperatures are -40°C, 150°C, and 25°C, respectively. 2 In the figure, the GaN HEMT devices W500T-40, W500T150, and W500T25 have the same size (w = 50u, l = 0.6u), and their temperatures are -40°C, 150°C, and 25°C, respectively.

[0059] In this way, each group of GaN HEMT device parameters that need to be extracted are devices of the same size and different temperatures.

[0060]

[0061] Table 1

[0062] In step 103, for each group of GaN HEMT devices G in the plurality of groups, n (S i , T j ), n, i, j = 1, 2, 3 ..., set different input and output parameters, input different parameter values ​​to each group of GaN HEMT devices to obtain the measurement data of the output of each group of GaN HEMT devices, and generate a measurement curve based on the measurement data.

[0063] More specifically, different input and output parameters are set for the GaN HEMT devices in each group, different parameter values ​​are input to each group of GaN HEMT devices to obtain the measurement data of the output of each group of GaN HEMT devices, and a measurement curve is generated based on the measurement data, including one or more of the following methods: For example: for the first device G in the first group of GaN HEMT devices 1 (S 1 ,T 1), set the input parameter as drain voltage (vds), the output parameters as reverse conduction capacitance (crss), output capacitance (coss), and input capacitance (ciss), then change the value of drain voltage (vds) in the range of 0v-12v, and obtain the actual measured values ​​of reverse conduction capacitance (crss), output capacitance (coss), and input capacitance (ciss) of the GaN HEMT device through measuring instruments.

[0064] like Figure 2 As shown, in the measurement curve image obtained based on the above measurement data, the horizontal axis leakage voltage (vds) is the input value, and the vertical axis reverse conduction capacitance (crss) is the output value. Figure 2 The solid points are measured data points, the curve is a simulation diagram, and the RMS represents the average value of the fitting error value, which is 0.57%, and the maximum value RMAX of the error is 1.57%.

[0065] like Figure 3 As shown, in the measurement curve image obtained based on the above measurement data, the horizontal axis leakage voltage (vds) is the input value, and the vertical axis output capacitance (coss) is the output value. Figure 3 The solid points are measured data points, and the curve is a simulation graph. The RMS indicates that the average error value of the fitting is 0.35%, and the maximum error value RMAX is 0.75%.

[0066] like Figure 4 As shown, in the measurement curve image obtained based on the above measurement data, the horizontal axis leakage voltage (vds) is the input value, and the vertical axis input capacitance (ciss) is the output value. Figure 4 The solid points are measured data points, and the curve is a simulation graph. The RMS represents the average error value of the fitting, which is 0.69%, and the maximum error value RMAX is 1.15%.

[0067] Similarly, for the first group G of GaN HEMT devices 1 (S 1 ,T 1 ), set the input parameter to gate voltage (vgs), and give a value ranging from -6v to 6v, and obtain the value of gate capacitance (cgg) of GaN HEMT device through measuring instrument.

[0068] like Figure 5 As shown, in the measurement curve image obtained based on the above measurement data, the horizontal axis gate voltage (vgs) is the input value, and the vertical axis gate capacitance (cgg) is the output value. Figure 5 The solid points are measured data points, and the curve is a simulation graph. The RMS represents the average error value of the fitting, which is 5.53%, and the maximum error value RMAX is 26.5%.

[0069] Similarly, for the first device G in the first group of GaN HEMT devices 1 (S 1 ,T 1 ), set the input parameter as the gate voltage (vgs), and give a value that changes in sequence from 0v-5v, and obtain the value of the leakage current (ids) of the GaN HEMT device through the measuring instrument. At the same time, the source-drain voltage (Vds) is set as the scanning value, which changes in sequence from 0.1v-12.0v.

[0070] like Figure 6 As shown in FIG. 1 , a measurement curve image is obtained based on the above measurement data, wherein the horizontal axis gate voltage (vgs) is the input value, and the vertical axis leakage current (ids) is the output value. Figure 6 The leakage current-gate voltage (id_vg) curve is shown in the following figure under the conditions of gate width w = 100um, gate length l = 0.6um, and temperature t = 25°C. Figure 6 The solid points are measured data points, and the curve is a simulation graph. The RMS represents the average error value of the fitting, which is 7.65%, and the maximum error value RMAX is 9.61%.

[0071] like Figure 7 As shown in FIG. 1 , a measurement curve image is obtained based on the above measurement data, wherein the horizontal axis gate voltage (vgs) is the input value, and the vertical axis leakage current (ids) is the output value. Figure 7 All the leakage current-gate voltage (id_vg) curves under the conditions of gate width w = 100um, gate length l = 0.6um, temperature t = 25℃, Figure 7 The solid points are measured data points, and the curve is a simulation graph. RMS indicates that the average error value of the fitting is 1%, and the maximum error value RMAX is 4.14%.

[0072] Similarly, for the first device G in the first group of GaN HEMT devices 1 (S 1 ,T 1 ), set the input parameter as the drain voltage (vds), and give a value that changes in sequence from 0v-12v, and obtain the value of the drain current (ids) of the GaN HEMT device through the measuring instrument. At the same time, the gate voltage (vgs) is set to a scan value, which changes in sequence from 0.1v-5.0v.

[0073] like Figure 8 As shown in FIG. 1 , a measurement curve image is obtained based on the above measurement data, wherein the horizontal axis leakage voltage (vds) is the input value, and the vertical axis leakage current (ids) is the output value. Figure 8All the leakage current-drain voltage (id_vd) curves under the conditions of gate width w = 100um, gate length l = 0.6um, temperature t = 25℃, Figure 8 The solid points are measured data points, and the curves are simulation graphs. The RMS represents the average error value of the fitting, which is 20.41%, and the maximum error value RMAX is 99.97%.

[0074] In step 104, based on the measurement curve obtained in step 103, multiple groups G of GaN HEMT devices are n (S i , T j ) and extract the parameters of the capacitance-voltage (CV) curve.

[0075] The Advanced Spice Model (ASM) is a physical model based on surface potential. In addition to discussing large signal modeling, the model also accurately models GaN's field plate capacitance, trap effect, Kink effect, and noise characteristics. It is considered to be an industrial compact standard model for power GaN and RF GaN. The ASM model contains process parameters and model parameters, where the process parameters are determined by the device manufacturing process, such as the device's channel length, width, and number of gate fingers.

[0076] In the process of using the optimizer to automatically extract parameters of the ASM model, the essence of parameter extraction is to adjust the model parameters and fit the test curve with the simulation results. Figure 2 The measurement curve image is shown in Figure 1, and RMS represents the fitting error value.

[0077] In the present invention, in the automatic parameter extraction of the ASM model, for example, for a device G in the first group of GaN HEMT devices, 1 (S 1 ,T 1 ) Extract the parameters of the capacitance-voltage (CV) curve, including: 1 GaN HEMT device, fitting at room temperature T 1 Capacitance-voltage (CV) curve at (25°C).

[0078] Among them, (1) extracting the drain-source capacitance parameter (cgdo), drain edge capacitance parameter (cfgd), bias voltage parameter (cgdl), and drain saturation voltage parameter (vdsatcv) from the reverse conduction capacitance-drain voltage (crss-vd) curve;

[0079] (2) Extract the drain-source capacitance parameter (cdso), the drain edge capacitance parameter (cfd), the parameter (aj) that controls the value that is independent of Cds at low source-drain voltage Vds, the capacitance parameter (cj0) of the access region at zero source-drain voltage Vds, the parameter (mz) of Caccd attenuation at high source-drain voltage Vds, and the drain built-in potential parameter (vbi) from the output capacitance-drain voltage (coss_vd) curve;

[0080] (3) Extract the gate-source overlap capacitance parameter (cgso) from the input capacitance-drain voltage (ciss_vd) curve;

[0081] (4) Extract the quantum mechanical effect pre-factor and switch parameter (adosi) in the inversion of the AlGaN layer thickness parameter and other related parameters (tbar), the charge center of mass parameter_the slope parameter of the CV curve under the inversion QME (bdosi), and the charge center of mass parameter_QME inversion starting point parameter (qm0i) from the gate capacitance_gate voltage (cgg_vg) curve. ;

[0082] For size S 1 GaN HEMT device, extract the room temperature T 1 After obtaining the above parameters of the CV curve, save the ASM model of the GaN HEMT device and directly apply the saved ASM model of the GaN HEMT device to the same size S in the first group of GaN HEMT devices. 1 Other temperatures (T 1 , T 2 , T 3 …) and perform parameter extraction based on the ASM model based on the same steps as above.

[0083] In the present invention, the software XModel used has the function of Apply model to other local(s), which can directly apply the adjusted ASM model to the parameter extraction of GaN HEMT devices of the same size and other temperatures. Therefore, when extracting the parameters of GaN HEMT devices of the same size and other temperatures, only fine-tuning is required, which can greatly reduce the workload of parameter extraction.

[0084] In the present invention, first extract the size S 1 , room temperature (T 1 ) under the capacitance-voltage (CV) curve, save the ASM model of the GaN HEMT device, and then apply the ASM model to other temperatures (T 1 , T 2 …), for example, high temperature (150°C) and low temperature (-40°C) conditions are used to extract corresponding parameters, so that only some parameters affected by temperature need to be adjusted in high and low temperatures.

[0085] In step 105, based on the measurement curve obtained in step 103, multiple groups G of GaN HEMT devices are n (S i , T j ) and extract the parameters of the current-voltage (IV) curve.

[0086] The IV parameter extraction process of the present invention is first locally optimized and then globally optimized. The specific process is as follows:

[0087] Among them, local optimization includes:

[0088] Parameters such as cutoff voltage (VOFF) and subthreshold slope (NFACTOR) are extracted from the drain current-gate voltage (id-vg) curve under linear drain current bias conditions. A rough estimate of the cutoff voltage (VOFF) is the linearly proportional gate voltage (vg) value at which the drain current (ID) starts to rise in the drain current-gate voltage (ID-VG) graph. This rough value can be adjusted in the parameter extraction flow to obtain the best results. The subthreshold slope (NFACTOR) controls the subthreshold slope of the device and is extracted by matching the model to the drain current-gate voltage (ID-VG) characteristic under linear drain voltage (VD) conditions on a logarithmic scale.

[0089] The cut-off voltage (VOFF) and subthreshold slope (NFACTOR), carrier low field mobility U0 and mobility vertical field related parameters UA and UB should be extracted from the drain current-gate voltage (ID-VG) characteristic curve under linear drain voltage (VD) conditions. The transconductance GM and its derivatives GM0 and GM00 under linear VD conditions can be accurately simulated using the U0, UA and UB parameters.

[0090] After the linear drain voltage (VD) condition is fitted, parameters should be extracted from the high drain voltage (VD) condition (Vds = 1.2V). As mentioned above, first focus on the drain current-gate voltage (ID-VG) characteristics under high drain voltage (VD) conditions, and extract the cut-off voltage (VOFF) or low current region parameters. Due to the drain induced barrier lowering effect, the cut-off voltage under high drain voltage (VD) conditions is reduced, which can be simulated by extracting DIBL parameters ETA0 and VDSCALE. Under high VD conditions, the subthreshold slope will also decrease, which can be simulated by the CDSCD parameters in the ASM-GaN-HEMT model.

[0091] Next, the self-heating effect is turned on (shmod=1), and the parameters of the drain current-gate voltage (ID-VG) curve in high drain voltage (VD) are extracted. The key parameters extracted under this condition are the velocity saturation parameter (VSAT), the channel length modulation parameter (λ), and the nonlinear series resistance parameters NS0ACCS, S, MEXPACCS, D and U0ACCS. VSAT is best extracted from the intermediate current level, and the transconductance (GM) still increases with the gate voltage (VG) under high VD conditions. The nonlinear series resistance causes the transconductance (GM) to decrease with the gate voltage (VG), and the relevant parameters should be adjusted to the appropriate range. Using the parameters extracted in this step, the transconductance under different drain voltage (VD) conditions can be accurately simulated. After the above local optimization is completed, the overall fit between the measured data and the simulation curve current-voltage (IV) has achieved a good overall fit effect. Next, global optimization is performed to fine-tune the parameters, and the current-voltage (IV) curve can be accurately modeled.

[0092] The above steps complete the DC parameter extraction process of the ASM-GaN-HEMT model at room temperature. 1 , the parameters for extracting the current-voltage (IV) curve are as follows:

[0093] After setting the bias conditions (vth_half = vth / 2, vgg, vth, vdlin) and initial values, local optimization is first performed, including: (1) extracting the threshold voltage (voff) and subthreshold slope (nfactor) of the drain current-gate voltage curve (id_vg) in the low current region of the low source-drain voltage (Vds) (the current region corresponding to before the gate voltage (vgs) reaches the threshold voltage);

[0094] (2) For the drain current-gate voltage curve (id_vg), extract the low-field mobility parameter (u0), mobility attenuation coefficient (ua), and mobility second-order attenuation coefficient (ub) in the low source-drain voltage (Vds) high current region (the current region corresponding to the gate voltage (vgs) reaching the threshold voltage);

[0095] (3) For the drain current-drain voltage curve (id_vd), extract the DIBL parameter and subthreshold attenuation parameters in the high source-drain voltage (Vds) low current region, and extract the drain-induced barrier lowering effect parameter (eta0), the drain voltage-induced subthreshold slope change parameter (cdscd), and the drain-induced barrier lowering effect and source-drain voltage Vds related parameter (vdscale) from the vthgm_vds curve;

[0096] (4) For the drain current-drain voltage curve (id_vd), the speed saturation parameter self-heating and other related parameters are extracted in the high source-drain voltage (Vds) high current region. From the drain current-drain voltage (id_vd) curve and the first-order derivative of the drain current-drain voltage (id′_vd) curve, the parameters drain contact resistance (rdc), source contact resistance (rsc), two-dimensional electron gas density per unit area in the source access region (n0accs), two-dimensional electron gas density per unit area in the drain access region (n0accd), resistance (rth0), mobility temperature-related parameter (ute), access region two-dimensional electron gas density temperature-related parameter (kns0), saturation speed (vsat), access region saturation speed value (vsataccs), and speed saturation parameter (thesat).

[0097] In the embodiment of the present invention, the low source-drain voltage Vds is greater than 0 and less than 1 / 2 of the source-drain voltage Vds; and the high source-drain voltage Vds is greater than 1 / 2 of the source-drain voltage Vds.

[0098] like Figure 6 To locally optimize the image, the horizontal axis gate voltage (vgs) is the input value, and the vertical axis leakage current (ids) is the output value. Figure 6 All the leakage current-gate voltage (id_vg) curves under the conditions of gate width w = 100um, gate length l = 0.6um, temperature t = 25℃, Figure 6 The solid points are measured data points, and the curve is a simulation graph. The RMS represents the average error value of the fitting, which is 7.65%, and the maximum error value RMAX is 9.61%.

[0099] Through the above optimization, the parameters are fine-tuned to obtain better fitting effect.

[0100] After the local optimization is completed, the global optimization is performed

[0101] Figure 7 , Figure 8 The global optimization effect of the present invention is shown.

[0102] like Figure 7 For the global optimization image, the horizontal axis gate voltage (vgs) is the input value, and the vertical axis leakage current (ids) is the output value. Figure 7 All the leakage current-gate voltage (id_vg) curves under the conditions of gate width w = 100um, gate length l = 0.6um, temperature t = 25℃, Figure 7 The solid points are measured data points, and the curve is a simulation graph. RMS indicates that the average error value of the fitting is 1%, and the maximum error value RMAX is 4.14%.

[0103] like Figure 8For the global optimization image, the horizontal axis drain voltage (Vds) is the input value, and the vertical axis drain current (ids) is the output value. Figure 8 All the leakage current-drain voltage (id_vd) curves under the conditions of gate width w = 100um, gate length l = 0.6um, temperature t = 25℃, Figure 8 The solid points are measured data points, and the curve is a simulation graph. The RMS indicates that the average error value of the fitting is 20.41%.

[0104] In the present invention, first extract the size S 1 , room temperature (T 1 ) under the current-voltage (IV) curve, save the ASM model of the GaN HEMT device, and then apply the ASM model to other temperature (T) devices of the same size through the Apply model to other local(s) function of XModel. 1 , T 2 …) of GaN HEMT devices, for example, the corresponding parameters are extracted under high temperature (150°C) and low temperature (-40°C) conditions, so that only the parameters affected by temperature need to be adjusted in high and low temperatures.

[0105] The optimizer used for local optimization and global optimization of the present invention is an optimizer that uses a trust region algorithm. The optimizer is characterized by searching in a region. In a conventional one-dimensional search, the process of moving from point x to the next point can be described as: x+αd, where αd is the displacement in the d direction, which can be recorded as s. The trust region algorithm used by the optimizer of the present invention directly determines the displacement s. At the same time, unlike the one-dimensional search, it does not first determine the search direction d. If the displacement can sufficiently reduce the objective function value, the trust region is expanded; if the objective function value cannot be sufficiently reduced, the trust region is reduced. This iteration continues until convergence. Optimization is to make the objective function converge to the minimum point. The traditional optimization algorithm is to continuously adjust the search direction and then perform a one-dimensional search in the search direction. The newly defined optimizer of the present invention directly searches in multiple dimensions in a region, which can greatly speed up the speed of finding the optimal value. In the application, the optimizer significantly reduces the root mean square error (RMS), and the optimization speed is greatly improved in actual operation; during the optimization process, when the optimized parameter reaches the boundary value, the optimizer can also realize the automatic expansion of the boundary. The principle of automatic expansion is: when the parameter value reaches the set boundary value, add or subtract 0.5 times of its own boundary value on the basis of the current boundary value to set the new boundary value. It supports setting the number of times the soft boundary is automatically expanded (Beyond Boundary Times); at the same time, the default maximum number of iterations (Max Iterations) is also set on the basis of not affecting the accuracy of the optimization results, nor the efficiency of extraction. Therefore, the optimization rate and accuracy are improved.

[0106] The optimizer of the present invention can also customize the optimization interval of parameter extraction based on the actual points of the measurement data using a script, and by calling the function for specifying the optimization interval range in the script, the area where the parameters need to be extracted is framed on the measurement curve graph. The functions called in the script here are all based on the python class script, and the function call is convenient and accurate, which increases the practical engineering application significance. The callable functions are listed as follows (only some are listed):

[0107] double[]getData(String exp): Get all data values ​​in the selected area;

[0108] int nearestIndexOf(double value): returns the index of the value closest to value in the array;

[0109] double abs(double x): returns the absolute value of x;

[0110] double min(): Returns the minimum value in the array;

[0111] double max(): Returns the maximum value in the array;

[0112] double[] derivative(double[] y, double[] x): Differentiates y with respect to x;

[0113] void addRegion(String regionName, double min, double max): Adds a selection range for the region where min < x < max;

[0114] As shown in the Figure 2 and Figure 6 figure, the blue region outlined is defined by the Script. After specifying the maximum and minimum values for the outlined region through functions, the parameter extraction range for steps 103 and 104 above is carried out within this region, enabling the simulation curve to be fitted quickly and directly. Users can achieve more efficient free selection as needed.

[0115] The present invention is illustrated by the following specific embodiments:

[0116] Taking the measurement data of a GaN HEMT with the following process parameters: W = 100um, L = 0.6um, nf = 10, T = 25°C, ldg = 0.7m, lsg = 0.35m as an example, the parameter extraction results obtained after performing DC parameter extraction according to the above steps are as follows: Parameter extraction results for the capacitance-voltage (cv) curve:

[0117]

[0118]

[0119] Parameter extraction results for the current-voltage (iv) curve:

[0120]

[0121]

[0122] Using the parameter values extracted above to plot the corresponding CV curve and IV curve, combined with the measured values, respectively, the curve images of the measured and simulated values of the inverse conductance capacitance-drain voltage (crss_vd) as shown in Figure 2 , the curve images of the measured and simulated values of the output capacitance-drain voltage (coss_vd) as shown in Figure 3 , and as shown in Figure 4The input capacitance-drain voltage (ciss_vd) measured value and simulated value curve image shown in FIG. Figure 5 The curve image of the gate capacitance _ gate voltage (cgg_vg) measured value and simulation value is shown in FIG. Figure 6 The curve image of the drain current _ gate voltage (id_vg) measured value and simulation value obtained as shown in Figure 7 The curve image of the drain current _ gate voltage (id_vg) measured value and simulation value shown in FIG. Figure 8 The graph of the drain current-drain voltage (id_vd) measured and simulated values ​​is shown.

[0123] Attached Figure 2-8 For the optimization simulation results of the present invention, Figure 9-15 The optimization simulation results of the prior art are as follows:

[0124] Attached Figure 2 With attached Fig. 9 The simulation fitting results of the reverse conduction capacitance-drain voltage (crss_vd) curve of the present invention and the prior art under the same size, same temperature conditions, same input parameters and output parameters are shown respectively. It can be seen intuitively that the average RMS error of the fitting in the present invention is 0.57%, and the average RMS error of the fitting in the prior art is 2.58%. The present invention reduces the average error value of the fitting.

[0125] Attached Figure 3 With attached Fig.10 The simulation fitting results of the output capacitance-drain voltage (coss_vd) curve of the present invention and the prior art under the same size, same temperature conditions, same input parameters and output parameters are shown respectively. It can be seen intuitively that the average RMS error of the fitting in the present invention is 0.35%, and the average RMS error of the fitting in the prior art is 1.52%. The present invention reduces the average error value of the fitting.

[0126] Attached Figure 4 With attached Fig.11 The simulation fitting results of the input capacitance-drain voltage (ciss_vd) curve of the present invention and the prior art under the same size, same temperature conditions, same input parameters and output parameters are shown respectively. It can be seen intuitively that the average RMS error of the fitting in the present invention is 0.69%, and the average RMS error of the fitting in the prior art is 0.87%. The present invention reduces the average error value of the fitting.

[0127] Attached Figure 5 With attached Fig.12The simulation fitting results of the gate capacitance-gate voltage (cgg_vg) curve of the present invention and the prior art under the same size, same temperature conditions, same input parameters and same output parameters are shown respectively. It can be seen intuitively that the average RMS error of the fitting in the present invention is 0.69%, and the average RMS error of the fitting in the prior art is 0.87%. The present invention reduces the average error value of the fitting.

[0128] Attached Figure 6 With attached Fig.13 The simulation fitting results of the drain current-gate voltage (id_vg) curve of the present invention and the prior art under the same size, same temperature conditions, same input parameters and same output parameters are shown respectively. It can be seen intuitively that the average RMS error of the fitting in the present invention is 7.65%, and the average RMS error of the fitting in the prior art is 25.42%. The present invention reduces the average error value of the fitting.

[0129] Attached Figure 7 With attached Fig.14 The simulation fitting results of the drain current-gate voltage (id_vg) curve of the present invention and the prior art under the same size, same temperature conditions, same input parameters and same output parameters are shown respectively. It can be seen intuitively that the average RMS error of the fitting in the present invention is 1%, and the average RMS error of the fitting in the prior art is 1.88%. The present invention reduces the average error value of the fitting.

[0130] Attached Figure 8 With attached Fig.15 The simulation fitting results of the leakage current-drain voltage (id_vd) curve of the present invention and the prior art under the same size, same temperature conditions, same input parameters and same output parameters are shown respectively. It can be seen intuitively that the average RMS error of the fitting in the present invention is 20.41%, and the average RMS error of the fitting in the prior art is 28.07%. The present invention reduces the average error value of the fitting.

[0131] In the present invention, steps 104 and 105 may be processed in parallel or sequentially, and the present invention does not impose any particular limitation on this.

[0132] Compared with the prior art, the GaN parameter extraction method of the present invention has the following advantages:

[0133] 1) Considering the influence of temperature of GaN HEMT, GaN HEMTs are grouped according to different sizes and temperatures, and parameters are extracted separately, which improves the accuracy of parameter extraction.

[0134] 2) The optimizer "Trust region" defined by the trust region algorithm in the present invention has greatly reduced the root mean square error (RMS) compared with the existing optimizer; it can also automatically expand the boundary and find the optimal value when the region cannot meet the conditions; at the same time, it has a default optimization iteration number with the best effect and the least time consumption. This makes the optimization process fast and efficient, and improves the efficiency and accuracy of parameter extraction.

[0135] 3) By calling functions to customize the optimization interval, it can meet the different needs of users, has strong flexibility and convenience, and has practical engineering significance.

[0136] 4) First extract the capacitance-voltage (cv) curve, then extract the current-voltage (iv) curve. At the same time, the current-voltage (iv) curve adopts the strategy of first local optimization and then global optimization, which makes the extraction more targeted and increases the efficiency and accuracy of the extraction;

[0137] 5) The model after parameter extraction can be saved and applied to devices of the same size but at different temperatures. In this way, only some parameters and curves need to be adjusted for the remaining devices to achieve the effect of parameter extraction for the entire model, which greatly saves the time for parameter extraction and increases the efficiency of parameter extraction, and has strong practical engineering significance.

[0138] The present invention also provides a GaN HEMT parameter extraction device based on an ASM model, which can be implemented by a general computer or computer system, wherein the computer or computer system has an input device, a display device, an external I / F, a communication I / F, a processor, and a memory. These hardware are connected to each other in a manner that allows communication via a bus.

[0139] The processor is a computer's internal CPU, or a dedicated CPU, DSP or other processing unit. The memory device includes a computer-readable storage medium, such as a computer read-only memory (ROM), a random access memory (RAM), a magnetic disk or an optical disk.

[0140] The memory stores a program running on the processor, and the processor executes the above-mentioned parameter extraction method of GaN HEMT based on the ASM model when running the program.

[0141] The present invention also provides a readable storage medium, on which a program or instruction is stored. When the program or instruction is executed by a processor, each process of the parameter extraction method of GaN HEMT based on the ASM model is implemented, and the same technical effect can be achieved. To avoid repetition, it will not be repeated here.

[0142] The embodiments of the present application are described above in conjunction with the accompanying drawings, but the present application is not limited to the above-mentioned specific implementation methods. The above-mentioned specific implementation methods are merely illustrative and not restrictive. Under the guidance of the present application, ordinary technicians in this field can also make many forms without departing from the purpose of the present application and the scope of protection of the claims, all of which are within the protection of the present application.

Claims

1. A parameter extraction method for GaN HEMT based on ASM model, characterized in that: The following steps are involved: Setting process parameters of the GaN HEMT, wherein the process parameters include one or more of size parameters, functional parameters, and working environment parameters; The more than one GaN HEMT devices are divided into one or more groups based on different sizes and temperatures, wherein the GaN HEMT devices in each group have the same size and different temperatures, respectively. For each group of GaN HEMT devices, the following processing is performed: Selecting a first GaN HEMT device having a first size S1 and a first temperature T1, adjusting different input parameter values ​​to obtain a plurality of measurement data output by the first GaN HEMT device, and generating a measurement curve based on the measurement data, wherein the measurement curve includes a capacitance-voltage relationship curve; Extract relevant parameters of capacitance-voltage relationship curve; Save the ASM model of the 1st GaN HEMT device; The ASM model is applied to GaN HEMT devices with the same first size S1 and other temperatures to extract relevant parameters of the capacitance-voltage relationship curve.

2. The parameter extraction method of GaN HEMT based on ASM model according to claim 1, characterized in that: The measurement curve also includes a current-voltage relationship curve, The parameter extraction method also includes extracting relevant parameters of the current-voltage relationship curve.

3. The parameter extraction method of GaN HEMT based on ASM model according to claim 1, characterized in that: The process parameters include: device gate length L, device gate width W, operating temperature T, device gate index NF, device gate-source length LSG, and device drain-source length LDG.

4. The parameter extraction method of GaN HEMT based on ASM model according to claim 1, characterized in that: Extracting parameters related to the capacitance-voltage relationship curve includes: extracting related parameters through a reverse conduction capacitance-drain voltage curve, an output capacitance-drain voltage curve, an input capacitance-drain voltage curve, and a gate capacitance-gate voltage curve.

5. The parameter extraction method of GaN HEMT based on ASM model according to claim 2, characterized in that: Extracting the relevant parameters of the current-voltage relationship curve includes first optimizing the parameters locally and then optimizing them globally. The local optimization includes: 1) For the drain current-gate voltage curve id_vg, extract the threshold voltage voff and subthreshold slope nfactor in the low current region of low source-drain voltage Vds; 2) For the drain current-gate voltage curve id_vg, extract the low-field mobility parameter u0, mobility attenuation coefficient ua, and mobility second-order attenuation coefficient ub in the low source-drain voltage Vds and high current region; 3) For the drain current-drain voltage curve id_vd, extract the DIBL parameters and subthreshold attenuation parameters in the high source-drain voltage Vds and low current region; 4) For the leakage current-drain voltage curve id_vd, extract the speed saturation parameter in the high source-drain voltage Vds and high current region. The parameters obtained through the above local optimization are fine-tuned in the global optimization.

6. The parameter extraction method of GaN HEMT based on ASM model according to claim 5, characterized in that: The above parameter extraction is performed locally and globally based on an optimizer using a Trust-region algorithm, wherein the optimizer directly performs a multi-dimensional search in a region to find the optimal value.

7. The parameter extraction method of GaN HEMT based on ASM model according to claim 6, characterized in that: During the optimization process, when the optimized parameter reaches the boundary value, the optimizer automatically expands the boundary.

8. The parameter extraction method of GaN HEMT based on ASM model according to claim 6, characterized in that: Use scripts to customize the optimization interval for parameter extraction. By calling the function used to specify the optimization interval range in the script, select the area where parameters need to be extracted on the measurement curve graph.

9. The parameter extraction method of GaN HEMT based on ASM model according to claim 2, characterized in that: Extracted capacitance-voltage curve parameters include: Extract drain-source capacitance parameters, edge capacitance parameters, drain saturation voltage parameters, and bias voltage parameters from the reverse conduction capacitance-drain voltage crss-vds curve; Extract the drain-source capacitance parameters and drain edge capacitance parameters from the output capacitance_drain voltage coss_vds curve; Extract gate-source overlap capacitance parameters from input capacitance_drain voltage ciss_vds curve; The AlGaN layer thickness parameters are extracted from the gate capacitance_gate voltage cgg_vgs curve.

10. A GaN HEMT parameter extraction device based on an ASM model, characterized in that: The invention comprises a memory and a processor, wherein the memory stores a program running on the processor, and the processor executes the parameter extraction method of GaN HEMT based on the ASM model as claimed in any one of claims 1 to 9 when running the program.

11. A computer-readable storage medium having computer instructions stored thereon, characterized in that: When the computer instructions are executed by a processor, the parameter extraction method of GaN HEMT based on the ASM model according to any one of claims 1 to 9 is executed.