CN116362180B - GaN HEMT large signal model parameter integrated extraction method - Google Patents
GaN HEMT large signal model parameter integrated extraction method Download PDFInfo
- Publication number
- CN116362180B CN116362180B CN202211595118.4A CN202211595118A CN116362180B CN 116362180 B CN116362180 B CN 116362180B CN 202211595118 A CN202211595118 A CN 202211595118A CN 116362180 B CN116362180 B CN 116362180B
- Authority
- CN
- China
- Prior art keywords
- parameters
- parameter
- equivalent circuit
- parasitic
- large signal
- Prior art date
- Legal status (The legal status is an assumption and is not a legal conclusion. Google has not performed a legal analysis and makes no representation as to the accuracy of the status listed.)
- Active
Links
- 238000000605 extraction Methods 0.000 title claims abstract description 26
- 230000003071 parasitic effect Effects 0.000 claims abstract description 57
- 230000002068 genetic effect Effects 0.000 claims abstract description 37
- 238000004422 calculation algorithm Methods 0.000 claims abstract description 29
- 238000012360 testing method Methods 0.000 claims abstract description 22
- 238000005457 optimization Methods 0.000 claims abstract description 16
- 238000004088 simulation Methods 0.000 claims abstract description 16
- 238000004364 calculation method Methods 0.000 claims abstract description 12
- 230000035772 mutation Effects 0.000 claims abstract description 8
- 238000003012 network analysis Methods 0.000 claims abstract description 6
- 238000000034 method Methods 0.000 claims description 37
- 210000000349 chromosome Anatomy 0.000 claims description 8
- 239000011159 matrix material Substances 0.000 claims description 8
- 239000003990 capacitor Substances 0.000 claims description 6
- 238000004458 analytical method Methods 0.000 claims description 5
- 239000008186 active pharmaceutical agent Substances 0.000 claims description 3
- 239000000284 extract Substances 0.000 claims description 3
- 230000001419 dependent effect Effects 0.000 claims description 2
- 230000007246 mechanism Effects 0.000 claims description 2
- 230000008569 process Effects 0.000 description 8
- 238000010586 diagram Methods 0.000 description 7
- 230000008859 change Effects 0.000 description 3
- 238000013461 design Methods 0.000 description 3
- 239000004065 semiconductor Substances 0.000 description 3
- 238000012937 correction Methods 0.000 description 2
- 230000000694 effects Effects 0.000 description 2
- 229910001218 Gallium arsenide Inorganic materials 0.000 description 1
- 206010064571 Gene mutation Diseases 0.000 description 1
- 206010063385 Intellectualisation Diseases 0.000 description 1
- 241001125929 Trisopterus luscus Species 0.000 description 1
- 238000013528 artificial neural network Methods 0.000 description 1
- 238000006243 chemical reaction Methods 0.000 description 1
- 230000007547 defect Effects 0.000 description 1
- 238000005516 engineering process Methods 0.000 description 1
- 230000014509 gene expression Effects 0.000 description 1
- 230000007614 genetic variation Effects 0.000 description 1
- 230000006872 improvement Effects 0.000 description 1
- 239000000463 material Substances 0.000 description 1
- XOFYZVNMUHMLCC-ZPOLXVRWSA-N prednisone Chemical compound O=C1C=C[C@]2(C)[C@H]3C(=O)C[C@](C)([C@@](CC4)(O)C(=O)CO)[C@@H]4[C@@H]3CCC2=C1 XOFYZVNMUHMLCC-ZPOLXVRWSA-N 0.000 description 1
- 108090000623 proteins and genes Proteins 0.000 description 1
- 238000012827 research and development Methods 0.000 description 1
- 230000035945 sensitivity Effects 0.000 description 1
- 238000012163 sequencing technique Methods 0.000 description 1
- 238000004904 shortening Methods 0.000 description 1
- 238000002922 simulated annealing Methods 0.000 description 1
- 238000012795 verification Methods 0.000 description 1
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/30—Circuit design
- G06F30/36—Circuit design at the analogue level
- G06F30/367—Design verification, e.g. using simulation, simulation program with integrated circuit emphasis [SPICE], direct methods or relaxation methods
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/12—Computing arrangements based on biological models using genetic models
- G06N3/126—Evolutionary algorithms, e.g. genetic algorithms or genetic programming
Landscapes
- Engineering & Computer Science (AREA)
- Physics & Mathematics (AREA)
- Theoretical Computer Science (AREA)
- Computer Hardware Design (AREA)
- Biophysics (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- General Engineering & Computer Science (AREA)
- General Physics & Mathematics (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Evolutionary Computation (AREA)
- Genetics & Genomics (AREA)
- Data Mining & Analysis (AREA)
- Physiology (AREA)
- Geometry (AREA)
- Artificial Intelligence (AREA)
- Biomedical Technology (AREA)
- Computational Linguistics (AREA)
- Microelectronics & Electronic Packaging (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- Software Systems (AREA)
- Junction Field-Effect Transistors (AREA)
Abstract
The invention discloses a GaNHEMT large signal model parameter integrated extraction method, which comprises the following steps: extracting parasitic parameter values by a small-signal equivalent circuit parasitic parameter extraction method, and establishing an initial population; stripping parasitic parameters in the initial population, obtaining partial intrinsic parameters through network analysis, and obtaining transconductance and admittance through direct current IV test data containing large signal characteristics to obtain S parameter data of a simulation equivalent circuit; calculating individual fitness according to the S parameter data; carrying out genetic calculation on individuals in the initial population according to set individual selection, crossing and mutation rules to obtain new genetic individuals; judging whether the genetic algorithm meets a preset target, outputting element parameters corresponding to the optimal individual as final parameters of the equivalent circuit, and establishing a large-signal equivalent circuit model. The invention realizes the automatic and efficient extraction of the optimization algorithm, greatly shortens the calculation time, reduces the calculation cost and improves the modeling efficiency of the large-signal equivalent circuit of the GaNHEMT device.
Description
Technical Field
The invention relates to a third-generation semiconductor material GaN HEMT equivalent circuit modeling technology, in particular to a GaN HEMT large signal model parameter integrated extraction method.
Background
The semiconductor device model has important influence on the improvement of the circuit design precision, and the more complex the circuit structure is, the higher the power of the working frequency band is, the higher the requirement on the device model is, and the establishment of an accurate semiconductor device model is very important for improving the success rate of the radio frequency microwave/millimeter wave monolithic integrated circuit design and shortening the circuit research and development period. The most important in establishing the device experience base equivalent model is to select a proper numerical algorithm to accurately and efficiently extract the element parameters in the equivalent circuit. The traditional compact model modeling method proposed in the past adopts a bottom-up (bottom-up) method, namely, after a complete small signal model is built, the integral large signal model is built by considering the nonlinearity of the eigenvalues on the basis of the small signal model. The method can accurately extract the circuit parameters, but repeated iterative correction of the small signal model and the large signal model is required to be established according to the test data, and parameter optimization steps can be carried out for a plurality of times in the parameter extraction process, so that the efficiency of parameter extraction is greatly weakened.
Disclosure of Invention
The invention aims to provide a GaN HEMT large signal model parameter integrated extraction method, which can realize automatic and efficient extraction of an optimization algorithm, greatly shorten the calculation time, reduce the calculation cost and improve the modeling efficiency of a GaN HEMT device large signal equivalent circuit.
The technical scheme for realizing the purpose of the invention is as follows: a GaN HEMT large signal model parameter integrated extraction method comprises the following steps:
step 1, extracting parasitic parameter values by a small-signal equivalent circuit parasitic parameter extraction method, and establishing an initial population for a genetic algorithm;
step 2: stripping parasitic parameters in the initial population, obtaining intrinsic parameters through network analysis, and obtaining transconductance g of leakage current related terms m And admittance G ds Obtaining S parameter data of a simulation equivalent circuit through direct current IV test data containing large signal characteristics;
step 3, calculating individual fitness of an individual according to the S parameter data of the simulation equivalent circuit;
step 4, carrying out genetic calculation on individuals in the initial population according to set individual selection, crossing and mutation rules to obtain new genetic individuals;
step 5, judging whether the genetic algorithm meets a preset target or not: if the termination condition is not met, updating the iterative initial population by the genetic individuals for optimization; if the parameters are satisfied, outputting an optimal individual according to the genetic individual, outputting element parameters corresponding to the optimal individual as final parameters of the equivalent circuit, and establishing a large-signal equivalent circuit model.
Compared with the prior art, the invention has the remarkable advantages that:
(1) According to the method, the large signal model and the small signal model are integrally modeled, and the small signal part element and the large signal part element are combined to obtain the optimal model parameters, so that the problem that the small signal model and the large signal model are required to be built according to test data and repeatedly iterated and corrected in the prior art is solved; in the prior art, the internal intrinsic parameters are obtained completely by relying on the values of external parasitic elements, and the parasitic parameters are often required to be manually adjusted in the parameter extraction process so as to obtain the intrinsic parameters irrelevant to the frequency;
(2) The invention is based on a genetic optimization algorithm and combines a cut-off condition method, can firstly acquire the initial value of the optimization algorithm, and substitutes the initial value as an initial population to iterate the algorithm, so that the problem of influence of initial value selection on the optimization result of the genetic algorithm is solved, the acquired initial value has a physical structural support, the condition that the optimal value does not accord with the actual condition is avoided, and the defects of difficult convergence of the numerical optimization method and incomplete direct extraction method are overcome.
Drawings
Fig. 1 is a topological structure diagram of a GaN HEMT large-signal equivalent circuit.
Fig. 2 is a low frequency equivalent circuit topology diagram of the pinch-off state.
Fig. 3 is an equivalent circuit diagram of inductance and resistance in the high frequency GaN HEMT "cold-forward" state.
Fig. 4 is a topological diagram of a small-signal equivalent circuit model in a pinch-off state.
Fig. 5 is a diagram showing initial simulation of parameters of the small signal model and comparison of measured S parameters in the pinch-off state.
FIG. 6 is a schematic diagram of an integrated modeling extraction element parameter flow.
FIG. 7 is a graph of the external parameters of the peeling Y INT A flow chart.
FIG. 8 shows the intrinsic element parameters with bias voltage V gs And V ds A graph of variation, wherein FIG. 8 (a) is C gs Trend graph, FIG. 8 (b) is C gd Trend graph, FIG. 8 (C) is C ds A change trend chart, FIG. 8 (d) is R gd A change trend chart, FIG. 8 (e) is R i Trend graph, fig. 8 (f) shows tau trend graph.
FIG. 9 is a comparison of S-parameter test and simulation results of a large signal model of a GaN HEMT device, wherein FIG. 9 (a) is V gs ＝-1.8V,V ds Comparison at 28V bias, fig. 9 (b) is V gs ＝-2.4V,V ds Comparison plot at =20v bias.
Fig. 10 is a schematic diagram of a load traction system test in ADS.
Fig. 11 is a graph comparing power scan simulation and test results of different frequency points GaN HEMTs, wherein graph (a) is an optimal power comparison graph and an optimal efficiency comparison graph at 1.1GHz, graph (b) is an optimal power comparison graph and an optimal efficiency comparison graph at 1.3GHz (c), and graph (c) is an optimal power comparison graph and an optimal efficiency comparison graph at 1.5 GHz.
Detailed Description
The method provided by the invention is a genetic algorithm-based GaN HEMT large-signal model parameter integrated extraction method, and the method comprises the steps of firstly extracting a parasitic parameter initial value through a small-signal equivalent circuit parasitic parameter extraction method, then optimizing a parasitic element value through a genetic algorithm, and obtaining partial intrinsic parameters and transconductance g through network analysis m Admittance G ds And obtaining through the relation with the leakage current Ids in the direct current IV test data, simulating the radio frequency characteristic S parameters under all bias states, calculating and experimental data errors, obtaining the optimal value of the circuit element, and forming the GaN HEMT large-signal equivalent circuit model integrated modeling method.
A genetic algorithm-based GaN HEMT large signal model parameter integrated extraction method specifically comprises the following steps:
and firstly, extracting small signal model parameters by combining a cut-off condition method and a numerical optimization method. Directly extracting parasitic capacitance, inductance and resistance element values in a low frequency band (f <5 GHz) and a high frequency band (f >20 GHz) respectively by using a cut-off condition method, and optimizing by combining a target error function to obtain an optimal element value;
and a second step of: stripping parasitic parameters in the initial population and obtaining gate capacitance C in intrinsic parameters through network analysis gs 、C gd Leakage capacitance C ds Resistance R i 、R gd Time tau, transconductance g for residual leakage current related term m And admittance G ds Obtaining S parameter data of the simulation equivalent circuit through direct current IV test data containing large signal characteristics;
and a third step of: calculating individual fitness of an individual according to the S parameter data of the simulation equivalent circuit;
fourth step: carrying out genetic calculation on individuals in the initial population according to set individual selection, crossing and mutation rules to obtain new genetic individuals;
fifth step: judging whether the genetic algorithm meets a preset target or not: if the termination condition is not met, repeating the fourth step to update the iterative initial population of the genetic individuals for optimization; if the parameters are satisfied, outputting an optimal individual according to the genetic individual, outputting element parameters corresponding to the optimal individual as final parameters of the equivalent circuit, and establishing a large-signal equivalent circuit model.
In the first step, the topology structure of the large-signal equivalent circuit adopted by the invention is shown in figure 1, and the selected equivalent circuit comprises 19 elements, including intrinsic parameters in a frame and an external parasitic parameter part. The parasitic parameter is generally considered to be an amount independent of the external bias voltage, while the internal intrinsic parameter is related to the device operating state, i.e., the amount of bias voltage provided by the gate-drain electrode. Under the general condition, the more the number of the equivalent circuit elements is, the more the physical characteristics of the device can be accurately expressed, but the reference lifting difficulty is gradually increased along with the increase of the number of the elements, so that the equivalent circuit topology is considered in a compromise way, and compared with the model topology proposed by the prior art, the selected model considers the contact capacitance C between flat plates pga 、C pda And C gda 。
First in the pinch-off state (V gs <V pinch-off ,V ds =0v), the impedance values of the parasitic inductance and the resistance are smaller than those of the capacitance when the frequency is lower than 5GHz, the capacitance isThe small signal equivalent circuit model can be simplified to the capacitive topology shown in fig. 2, with the characteristic impedance dominant. Thus, the Y parameter can be expressed as:
in the formula, nine unknown capacitance values exist, and the unknown capacitance values cannot be directly solved, and an assumption needs to be made according to the symmetry and the geometric dimension of the device structure: c (C) pga ＝C pda ,C gdi ＝2C gda ,C gs ＝C gd ,C pdi ＝3C pda The unknown quantity can be 5, and then the parameter C is scanned pda And C gda Equation 1.1 is continuously solved in the scanning process to obtain the corresponding capacitance value.
To extract the parasitic resistance and parasitic inductance values, the circuit external capacitance C is required to be set pda ,C gda ,C pga Stripping from the equivalent circuit, stripping the external capacitor, and then, at f>An equivalent circuit in the high frequency state of 20GHz can be equivalent to that shown in fig. 3.
For an equivalent circuit in this state, the equation for extracting parasitic resistance and parasitic inductance can be derived from circuit analysis theory:
in order to reduce the sensitivity of parasitic resistance and capacitance to frequency, the two sides of the Z parameter formula are multiplied by omega, so that the capacitance term is small and can be ignored. Inductance L g 、L d And L s Can be expressed as:
resistor R g 、R d 、R s Can be expressed as:
from formula (1.3), ω is on the ordinate of Im (ωZ) 2 The parasitic inductance value can be obtained by the slope of the curve with the abscissa, and the parasitic inductance value is similarly obtained by omega 2 Re (Z) is the ordinate, ω 2 The slope of the curve on the abscissa is the parasitic resistance value. Compared with GaAs HEMT, gaNHEMT has higher conduction band order delta E c There is a large built-in potential V bi The gate schottky junction can have larger capacitance, which is unfavorable for the extraction of parasitic resistance and inductance. While a gate diode can eliminate this effect under strong forward bias, selecting an excessive positive gate voltage will cause a current I to flow through the schottky junction g Too large, and the device is easily damaged. Therefore, the patent selects the positive bias of the grid voltage at V gs ＝2V，V ds =0v, and the gate capacitance is reduced as much as possible while ensuring the normal gate characteristics.
In scan C pda And C gda When it is found from the formulae (1.3) and (1.4), a value corresponding to each group C pda And C gda Scanning each parasitic resistance and parasitic inductance value of the value;
then the simulation S parameter corresponding to each group of element values is calculated by the pinch-off state small signal topology shown in fig. 4, so that the scanning quantity C can be continuously increased pda And C gda Searching to obtain a group of element values corresponding to the minimum residual error of the S parameter.
According to the method, a program for extracting the initial value of the small signal model parameter is written in Matlab, and the parasitic element value is extracted. Initial values of the elements are shown in table 1, and the corresponding pinch-off state S parameter fitting effect is shown in fig. 5.
TABLE 1 pinch-off state (V gs ＝-4V,V ds =0v) initial value of small signal model parameter
As can be seen from FIG. 5, the parasitic parameters of the small signal model can be accurately extracted by adopting the method, the S parameter fitting is good, and the extracted element value can be used as the initial value of the subsequent integrated modeling.
In the second step, the specific optimization flow is as shown in fig. 6, parasitic parameters in the initial population are stripped firstly, and then the grid capacitance C in the intrinsic parameters is obtained through network analysis gs 、C gd Leakage capacitance C ds Resistance R i 、R gd Time tau, for term transconductance g associated with leakage current m And admittance G ds And obtaining the S parameter of the equivalent circuit through simulation by using direct current IV test data with large signal characteristics.
The current mainstream optimization algorithms comprise a simulated annealing algorithm, a genetic algorithm, an artificial neural network algorithm and the like, wherein the genetic algorithm for searching an optimal solution based on the natural selection and gene mutation principle and the simulated natural evolution process is most widely applied, and the initial population of the genetic algorithm is generally a plurality of chromosomes which are randomly generated, so that the genetic algorithm is searched from a plurality of initial points instead of a single initial point, the earlier convergence speed of the genetic algorithm is higher, and a plurality of better areas can be found faster. And the mutation and crossover operation have randomness, so that the genetic algorithm has good capability of searching the global optimal solution. Therefore, the patent selects genetic algorithm to scan the parameters to obtain the optimal element parameter values.
The difference between the large signal model and the small signal model is that the intrinsic part in the circuit model can perform characteristic analysis on the device by adopting a linear method approximately when the small signal is generated, and the characteristic analysis is performed on the device when the circuit model works on the large signal as a nonlinear element related to bias voltage.
Taking parasitic capacitance inductance and resistance as genes in chromosome, firstly stripping external parasitic parameters to obtain intrinsic Y INT Network, calculating g-dividing under all bias states m And G ds The other six intrinsic parameter values are used for converting the actually measured S parameter of the device into the Y parameter and stripping the parasitic parameter from the Y parameter to obtain the actually measured intrinsic Y parameter. The specific stripping process is shown in fig. 7.
The small signal eigen element equivalent circuit can be represented by Y parameter and Z parameter:
wherein Y is DUT Y parameter matrix representing device under test, Y EXT Representing the parasitic capacitance of the outside:
Z RL is a Z-parameter matrix of parasitic resistance and inductance:
Y INT representing the internal intrinsic partial admittance matrix:
obtaining internal intrinsic Y according to the above method INT The network parameters, the analytical expressions for deriving each intrinsic parameter are as follows:
c(ω)＝(Y 21 (ω)-Y 12 (ω))(1+jd(ω)) (1.10)
wherein ω is the frequency bin under consideration, and the intrinsic Y parameters of formulae (1.9) to (1.16). The eigenvalue of the eigenvalue at each frequency point can be calculated at each bias according to the formula, and the average value under the full frequency band is used as the final value of the eigenvalue at the bias, because the eigenvalue varies with bias voltage and not with frequency.
Term g for current source Ids in the eigenelement m And G ds The test data are represented by formulas (1.17) and (1.18) using direct current IV:
in which I ds Representing the current between points d and s in FIG. 1, i.e., drain-source current value, V gs For the voltage value between g and s, the relation with the gate-source voltage can be expressed by the formula (1.19), V ds For the voltage between the d and s points, the relation between the d and s points and the drain-source voltage can be expressed by a formula (1.20), and parasitic and intrinsic parameters in a circuit model are obtained.
V gs ＝V GS -I ds ·R s (1.19)
V ds ＝V DS -I ds ·(R d +R s ) (1.20)
And in the third step, calculating the individual fitness of the individual according to the S parameter data of the simulation equivalent circuit and the measured radio frequency characteristics. The definition of the error function epsilon used in the search process is shown in formulas (1.21) - (1.25), wherein N is the number of frequencies considered epsilon ij Representing an objective error function, f n Is to measure the selected frequency value, deltaS ij Representing S parameter error, W of simulation and test ij And the reflection coefficient weight factor representing the high reflection coefficient region.
W ij ＝max|S ij |i,j＝1,2；i≠j (1.22)
W ij ＝1+|S ij |i＝j＝1,2 (1.23)
And in the fourth step, carrying out genetic calculation on individuals in the initial population according to set individual selection, crossing and mutation rules. The selection process adopts a selection mechanism of probability sequencing, and the method is that the chromosome fitness in the population is firstly sequenced from good to poor, the smaller the sequence number is, the better the corresponding chromosome is, and the larger the probability of being selected to the next generation is.
And then crossing with the parent chromosome according to a certain rule to generate offspring.
Parent 1 011111110000000000 parent 1 000000001111111111
Selecting the position of the crossing point, generating the child:
child 1 011111111111111111 child 2 000000000000000000
The new individuals formed after the cross operation have a certain probability of genetic variation, the individuals are mutated based on a certain probability, and the probability is mutation probability P m In general, P is set up m Less than or equal to 0.05. Thereby generating a new generation population by crossover and mutation.
In the fifth step, judging whether the genetic algorithm meets a preset target or not: if the termination condition is not met, repeating the second to fourth steps to update the iterative initial population of the genetic individuals for optimization; if the parameters are satisfied, outputting an optimal individual according to the genetic individual, outputting element parameters corresponding to the optimal individual as final parameters of the equivalent circuit, and establishing a large-signal equivalent circuit model.
According to the method, the calculation efficiency is improved, the process of manually adjusting parasitic parameters in the modeling process is avoided, repeated iterative correction of a small-signal model and a large-signal model is required to be established according to test data, the problem that internal intrinsic parameters are completely dependent on the values of external parasitic elements to obtain is solved, automatic and efficient extraction of an optimization algorithm can be realized, the calculation time is greatly shortened, the calculation cost is reduced, and the modeling efficiency of a large-signal equivalent circuit of a GaN HEMT device is improved.
The parasitic and intrinsic parameters were optimized according to this method, the values of the parasitic parameters obtained are shown in table 2, and the intrinsic parameter values with bias voltage are shown in fig. 8.
Table 2 integrated modeling parasitic parameter values
In order to verify whether the extracted large signal model can meet the precision requirements required by circuit analysis and design, the S parameter results of the simulation and test of the optimized device model are shown in fig. 9. Bias voltage V gs ＝-1.8V,V ds =28v and V gs ＝-2.4V,V ds =20v, frequency range 0.4-26 GHz, step 0.1GHz. It can be seen that the squareThe model established by the method can accurately represent the characteristics of the device.
In addition, large signal microwave property verification is also necessary for model testing. In the power amplifier, the S parameter of the GaN device is changed along with the change of an input signal, especially the S21 parameter is reduced along with the increase of the input signal, so that the conversion power gain is reduced along with the operation of the power element in the saturation region and is not different from a small signal state in which the output power and the input signal are in a direct proportion relation, the input/output ends of the GaN device are designed under the optimal condition of conjugate matching in the small signal operation state, the conjugate matching of the input/output ends is gradually not matched along with the entering of the power element into a nonlinear region, the maximum output power of the power element cannot be obtained, the optimal load impedance of the GaN device is changed along with the increase of the input signal power, and therefore, when the power amplifier works on a large signal, an equal output power curve at different load impedances is drawn for a given input power value, and the optimal load impedance at the maximum output power is found.
To verify the large signal output characteristics of the model, the output power P is tested by load impedance at the time of finding the maximum output power by load pulling (LoadPull) in ADS out And power added efficiency PAE characteristics, gate voltage V gs -2.55V, drain voltage V ds The device was operated in class AB at 28V, as shown in fig. 10, which is a load traction system test chart. By inputting different powers, corresponding output power and power additional efficiency are obtained, and the output power Pout and efficiency PAE of the optimal power circle and the output power PAE of the optimal efficiency circle obtained by the load traction at different working frequency points are basically identical with the data obtained by experimental tests, so that the accuracy of the model is verified.
The method for integrally extracting the parameters of the large-signal model of the GaN HEMT avoids the process of reciprocating iterative optimization of the small-signal model required by the establishment of the large-signal model in the traditional parameter extracting method, realizes the intellectualization of parameter extracting through programming, can accurately simulate the characteristics of direct current I-V, multi-bias S parameters and the like of a device, has the output power precision of the model in an L wave band of more than 95 percent, has the power additional efficiency PAE precision of more than 90 percent, and realizes the efficient integral extraction of the parameters of the large-signal empirical model of the GaN HEMT, thereby greatly improving the modeling efficiency of the device.
Claims (7)
1. The method for integrally extracting the GaN HEMT large signal model parameters is characterized by comprising the following steps of:
step 1: extracting parasitic parameter values by a small-signal equivalent circuit parasitic parameter extraction method, and establishing an initial population for a genetic algorithm;
step 2: stripping parasitic parameters in the initial population, obtaining intrinsic parameters through network analysis, and obtaining transconductance g of leakage current related terms m And admittance G ds Obtaining S parameter data of a simulation equivalent circuit through direct current IV test data containing large signal characteristics;
step 3: calculating individual fitness of an individual according to the S parameter data of the simulation equivalent circuit;
step 4: carrying out genetic calculation on individuals in the initial population according to set individual selection, crossing and mutation rules to obtain new genetic individuals;
step 5: judging whether the genetic algorithm meets a preset target or not: if the termination condition is not met, updating the iterative initial population by the genetic individuals for optimization; if the parameters are met, outputting an optimal individual according to the genetic individual, outputting element parameters corresponding to the optimal individual as final parameters of an equivalent circuit, and establishing a large-signal equivalent circuit model;
the step 2 specifically comprises the following steps:
step 2-1: stripping the extracted parasitic parameter value from the actually measured S parameter of the device to obtain an internal intrinsic portion admittance matrix Y INT Parameters;
step 2-2: deriving intrinsic parameters and Y INT Resolving the relation, and obtaining the grid source capacitance C in the intrinsic parameters by solving the equation gs Grid drain capacitor C gd Leakage capacitance C ds And gate-source capacitance C gs First resistor R connected in series i And gate-drain capacitance C gd A second resistor R connected in series gd Andtime delay tau;
step 2-3: leakage current dependent term transconductance g in intrinsic element m And admittance G ds Obtaining and obtaining direct current IV test data containing large signal characteristics, embedding large signal model parameters, and directly establishing a large signal equivalent circuit relationship;
step 2-4: simulating to obtain S parameters of the small-signal equivalent circuit;
the step 2-2 specifically comprises the following steps: the equivalent circuit is represented by Y parameter and Z parameter:
wherein Y is DUT Y parameter matrix representing device under test, Y EXT Representing the parasitic capacitance of the outside:
Z RL is a Z-parameter matrix of parasitic resistance and inductance:
Y INT representing the internal intrinsic partial admittance matrix:
based on an internal intrinsic partial admittance matrix Y INT Parameters, deriving each intrinsic parameter;
transconductance g in step 2-3 m And admittance G ds The method comprises the following steps:
V gs ＝V GS -I ds ·R s (1.7)
V ds ＝V DS -I ds ·(R d +R s ) (1.8)
in which I ds Representing the current between d and s points in the equivalent circuit topology, namely the drain-source current value, V gs For the voltage value between g and s, V GS Is the gate-source voltage V GS ，V ds For the voltage between d and s, V DS Is the drain-source voltage.
2. The integrated extraction method of the large signal model parameters of the GaN HEMT according to claim 1, wherein the extraction method of the parasitic parameters of the small signal equivalent circuit adopts a cut-off condition method.
3. The method for extracting large signal model parameters of GaN HEMT according to claim 2, wherein said step 1 extracts parasitic parameter values by a small signal equivalent circuit parasitic parameter extraction method, establishes an initial population for genetic algorithm, comprising the steps of:
step 1-1: first capacitor C in scanning small-signal equivalent circuit pda And a second capacitor C gda Obtaining a parasitic capacitance element in a pinch-off state and a low frequency state;
step 1-2: stripping the first parasitic capacitance C outside the circuit pda Second parasitic capacitance C gda And a third parasitic capacitance C pga Deducing and extracting parasitic inductance and parasitic resistance based on a circuit analysis theory in a high-frequency state;
step 1-3: after obtaining all external parasitic parameters, calculating pinch-off state circuit topology S parameters corresponding to each group of element values, and continuously changing the first capacitor C pda And a second capacitor C gda Scanning quantity searching to obtain a group with minimum corresponding S parameter residual errorAs the optimal element value.
4. The method for extracting large signal model parameters of GaN HEMT according to claim 3, wherein the low frequency is less than 5GHz and the high frequency is more than 20GHz.
5. The method for integrally extracting the large signal model parameters of the GaN HEMT according to claim 1, wherein the individual fitness is calculated by an error function epsilon, and the error function epsilon is:
W ij ＝max|S ij | i,j＝1,2；i≠j (1.12)
or W ij ＝1+|S ij | i＝j＝1,2 (1.13)
Where N is the number of frequencies under consideration ε ij Representing an objective error function, f n Is to measure the selected frequency value, deltaS ij Representing S parameter error, W of simulation and test ij And the reflection coefficient weight factor representing the high reflection coefficient region.
6. The method for integrally extracting the large signal model parameters of the GaN HEMT according to claim 1, wherein in the step 4, the individual selection adopts a selection mechanism of probability sorting, the chromosome fitness in the population is sorted from good to bad, the smaller the grade sequence number is, the better the corresponding chromosome is, and the greater the probability of being selected to the next generation is.
7. The integrated extraction method of GaN HEMT large signal model parameters according to claim 1, wherein in the step 4, chromosomes of parents are used for crossing according to a set rule to generate offspring, the positions of the crossing points are selected, and the generated offspring are obtained; setting variation probability P in cross-talk m The range is as follows: p (P) m ≤0.05。
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211595118.4A CN116362180B (en) | 2022-12-13 | 2022-12-13 | GaN HEMT large signal model parameter integrated extraction method |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202211595118.4A CN116362180B (en) | 2022-12-13 | 2022-12-13 | GaN HEMT large signal model parameter integrated extraction method |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN116362180A CN116362180A (en) | 2023-06-30 |
| CN116362180B true CN116362180B (en) | 2023-10-13 |
Family
ID=86911603
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202211595118.4A Active CN116362180B (en) | 2022-12-13 | 2022-12-13 | GaN HEMT large signal model parameter integrated extraction method |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN116362180B (en) |
Families Citing this family (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN117391024B (en) * | 2023-12-12 | 2024-02-23 | 浙江集迈科微电子有限公司 | Modeling method and device for GaN HEMT switching device, storage medium and terminal |
Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102542077A (en) * | 2010-12-15 | 2012-07-04 | 中国科学院微电子研究所 | Parameter extraction method of AlGaN/GaN HEMT small-signal model |
| CN105138730A (en) * | 2015-07-27 | 2015-12-09 | 电子科技大学 | Method for extracting small-signal model parameters of gallium nitride high-electron-mobility transistor |
| CN105426570A (en) * | 2015-10-28 | 2016-03-23 | 西安电子科技大学 | GaN HEMT large signal model improvement method based on active compensation sub-circuit |
| CN105787210A (en) * | 2016-03-25 | 2016-07-20 | 西安电子科技大学 | GaN HEMT (high electron mobility transistor) small signal circuit model parameter extraction method |
| CN106354947A (en) * | 2016-08-31 | 2017-01-25 | 徐跃杭 | Modeling method of microwave GaN power device |
| CN107818187A (en) * | 2016-09-14 | 2018-03-20 | 中航（重庆）微电子有限公司 | A kind of multiple-grid refers to AlGaN/GaN HETM small-signal models and its puies forward ginseng method |
| CN107844835A (en) * | 2017-11-03 | 2018-03-27 | 南京理工大学 | Multiple-objection optimization improved adaptive GA-IAGA based on changeable weight M TOPSIS multiple attribute decision making (MADM)s |
| CN107918708A (en) * | 2017-11-15 | 2018-04-17 | 成都海威华芯科技有限公司 | A kind of extracting method of GaN HEMT devices parasitic parameter |
| CN108062442A (en) * | 2017-12-12 | 2018-05-22 | 成都海威华芯科技有限公司 | A kind of AlGaN/GaN HEMT microwave power devices small-signal intrinsic parameters extracting method |
| CN114386347A (en) * | 2020-10-19 | 2022-04-22 | 苏州华太电子技术有限公司 | Modeling method of field effect transistor |
| CN114519275A (en) * | 2022-02-22 | 2022-05-20 | 西安电子科技大学芜湖研究院 | AlGaN/GaN HEMT device small signal model extraction method |
Family Cites Families (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| US8427241B2 (en) * | 2011-05-24 | 2013-04-23 | Amcom Communications, Inc. | High efficiency, high frequency amplifiers |
-
2022
- 2022-12-13 CN CN202211595118.4A patent/CN116362180B/en active Active
Patent Citations (11)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN102542077A (en) * | 2010-12-15 | 2012-07-04 | 中国科学院微电子研究所 | Parameter extraction method of AlGaN/GaN HEMT small-signal model |
| CN105138730A (en) * | 2015-07-27 | 2015-12-09 | 电子科技大学 | Method for extracting small-signal model parameters of gallium nitride high-electron-mobility transistor |
| CN105426570A (en) * | 2015-10-28 | 2016-03-23 | 西安电子科技大学 | GaN HEMT large signal model improvement method based on active compensation sub-circuit |
| CN105787210A (en) * | 2016-03-25 | 2016-07-20 | 西安电子科技大学 | GaN HEMT (high electron mobility transistor) small signal circuit model parameter extraction method |
| CN106354947A (en) * | 2016-08-31 | 2017-01-25 | 徐跃杭 | Modeling method of microwave GaN power device |
| CN107818187A (en) * | 2016-09-14 | 2018-03-20 | 中航（重庆）微电子有限公司 | A kind of multiple-grid refers to AlGaN/GaN HETM small-signal models and its puies forward ginseng method |
| CN107844835A (en) * | 2017-11-03 | 2018-03-27 | 南京理工大学 | Multiple-objection optimization improved adaptive GA-IAGA based on changeable weight M TOPSIS multiple attribute decision making (MADM)s |
| CN107918708A (en) * | 2017-11-15 | 2018-04-17 | 成都海威华芯科技有限公司 | A kind of extracting method of GaN HEMT devices parasitic parameter |
| CN108062442A (en) * | 2017-12-12 | 2018-05-22 | 成都海威华芯科技有限公司 | A kind of AlGaN/GaN HEMT microwave power devices small-signal intrinsic parameters extracting method |
| CN114386347A (en) * | 2020-10-19 | 2022-04-22 | 苏州华太电子技术有限公司 | Modeling method of field effect transistor |
| CN114519275A (en) * | 2022-02-22 | 2022-05-20 | 西安电子科技大学芜湖研究院 | AlGaN/GaN HEMT device small signal model extraction method |
Non-Patent Citations (2)
| Title |
|---|
| Research on Parameter Extraction of GaN HEMT Equivalent Circuit Model Based on Genetic Algorithm;Liming Zhang等;《Eighth Symposium on Novel Photoelectronic Detection Technology and Applications》;1-12 * |
| 氮化镓功率器件小信号模型参数提取算法研究;闻彰等;《电波科学学报》;第30卷(第4期);772-776 * |
Also Published As
| Publication number | Publication date |
|---|---|
| CN116362180A (en) | 2023-06-30 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106529102B (en) | AlGaN/GaN HEMT small signal model and parameter extraction method thereof | |
| CN116362180B (en) | GaN HEMT large signal model parameter integrated extraction method | |
| Kouhalvandi et al. | Automated deep neural learning-based optimization for high performance high power amplifier designs | |
| CN110717240B (en) | InP HEMT device noise equivalent circuit model establishment method | |
| CN107636656A (en) | GaN device technological parameter statistical analysis technique based on large signal equivalent circuit model | |
| CN108549767B (en) | Neural network space mapping modeling method for large-signal power transistor | |
| Xu et al. | Advances in artificial neural network models of active devices | |
| CN112733477B (en) | GaN HEMT modeling method based on error function precision compensation | |
| CN111539165B (en) | Chip design method and system based on yield load traction system | |
| CN107918708A (en) | A kind of extracting method of GaN HEMT devices parasitic parameter | |
| Li | A simulation-based evolutionary approach to LNA circuit design optimization | |
| Carneiro et al. | Doherty amplifier optimization using robust genetic algorithm and unscented transform | |
| Aldridge et al. | High performance SiGe HBT performance variability learning by utilizing neural networks and technology computer aided design | |
| Jarndal | Combined genetic algorithm and neural network technique for transistor modeling | |
| Liu et al. | A review of Wiener-type dynamic neural network for nonlinear device modeling | |
| Xie et al. | Design Space Exploration of Multi-Stage Op Amps by Symbolic Modeling and gm/ID Sampling | |
| CN114970419A (en) | InP terahertz HEMT positive and negative grid voltage small-signal model | |
| CN105653805B (en) | The method of road parasitism Interconnect model after correction | |
| De Jonghe et al. | Efficient analytical macromodeling of large analog circuits by transfer function trajectories | |
| CN110717242B (en) | InP HEMT device noise equivalent circuit model establishment method | |
| Jarndal | Neurogenetic small-signal modeling approach for microwave active devices | |
| Jarndal | Parasitic elements extraction of AlGaN/GaN HEMTs on SiC substrate using only pinch-off S-parameter measurements | |
| Cao et al. | Application of the genetic algorithm in MOSFET modeling and parameter extraction | |
| Chavez et al. | Deep learning-based ASM-HEMT high frequency parameter extraction | |
| Li et al. | An ANN-based small-signal equivalent circuit model for MOSFET device |
Legal Events
| Date | Code | Title | Description |
|---|---|---|---|
| PB01 | Publication | ||
| PB01 | Publication | ||
| SE01 | Entry into force of request for substantive examination | ||
| SE01 | Entry into force of request for substantive examination | ||
| GR01 | Patent grant | ||
| GR01 | Patent grant |