Method and system for generating nonlinear simulation model

1. A method of generating a nonlinear simulation model of a device under test (DUT), the method comprising:

receiving large-signal nonlinear waveform data based on responses of the DUT to input signals;

determining a set of first dynamical variables and a set of second dynamical variables from the nonlinear waveform data;

calculating values of second dynamical variables;

providing nonlinear constitutive relations as functions of the first dynamical variables and the calculated values of the second dynamical variables; and

compiling the nonlinear simulation model of the DUT using the identified nonlinear constitutive relations.

2. The method of claim 1 , wherein the responses of the DUT comprise measured responses from the DUT or simulated responses from a model of the DUT.

3. The method of claim 1 , wherein providing the nonlinear constitutive relations comprises:

identifying respective nonlinear constitutive relations of a current source and a charge source of the DUT as respective functions of the first dynamical variables and the calculated values of the second dynamical variables; and

training artificial neural networks (ANNs) to learn the identified nonlinear constitutive relations of the current source and the charge source.

4. The method of claim 3 , wherein training the ANNs to learn the identified nonlinear constitutive relations of the current source and the charge source comprises:

defining neural network internal parameters corresponding to the nonlinear constitutive relations of the current source and the charge source, respectively; and

adjusting the neural network internal parameters to minimize an error function.

5. The method of claim 4 , wherein the error function comprises a weighted sum of a time domain error function and a frequency domain error function corresponding to the waveforms of the nonlinear waveform data.

6. The method of claim 3 , wherein the nonlinear constitutive relations comprise nonlinear computational models modeling relationships between each of the current source and the charge source and the responses of the nonlinear waveform data.

7. The method of claim 1 , wherein the first dynamical variables comprise instantaneous terminal voltages of the DUT, and the second dynamical variables comprise auxiliary variables.

8. The method of claim 7 , wherein the instantaneous terminal voltages comprise at least one input voltage and at least one output voltage of the DUT, and the auxiliary variables comprise junction temperature and a plurality of state variables of the DUT.

9. The method of claim 8 , wherein the at least one input voltage comprises a gate-source voltage (V gs ) and the at least one output voltage comprises a drain-source voltage (V ds ), and the plurality of state variables comprise state variables associated with different trapped charges of the DUT.

10. The method of claim 8 , wherein the at least one input voltage comprises a base-emitter voltage (V be ) and the at least one output voltage comprises a collector-emitter voltage (V ce ).

11. The method of claim 1 , wherein the nonlinear waveform data is received from a nonlinear vector network analyzer (NVNA).

12. A method of generating a nonlinear simulation model of a transistor, the method comprising:

providing large-signal nonlinear waveform data of the transistor from a nonlinear vector network analyzer;

identifying dependencies of a model current source and a model charge source on at least one instantaneous terminal voltage, temperature and at least one trap state of the transistor identified directly from the nonlinear waveform data;

representing the identified dependencies using corresponding artificial neural networks; and

compiling a model comprising the represented dependencies for implementation into a nonlinear circuit simulator to generate the nonlinear simulation model of the transistor.

13. The method of claim 12 , wherein the at least one instantaneous terminal voltage comprises instantaneous gate-source voltage (V gs ) and instantaneous drain-source voltage (V ds ) of the transistor.

14. The method of claim 13 , wherein the temperature comprises a dynamic junction temperature (T j ) of the transistor.

15. The method of claim 14 , wherein identifying the dependencies of the model current source and the model charge source comprises:

determining at least one instantaneous terminal voltage of the transistor based on the nonlinear waveform data; and

calculating steady state values of the temperature and the at least one trap state of the transistor based on the nonlinear waveform data.

16. The method of claim 15 , wherein representing the identified dependencies using the corresponding artificial neural networks comprises:

training the artificial neural networks to learn the identified dependencies using the determined at least one instantaneous terminal voltage and the calculated steady state values of the temperature and the at least one trap state of the FET for each of the dynamic trajectories.

17. The method of claim 13 , wherein the at least one trap state comprises a gate-lag trap state associated with a gate-lag of the transistor, and a drain-lag trap state associated with a drain-lag of the transistor.

18. The method of claim 12 , wherein providing the nonlinear waveform data of the transistor comprises:

performing calibrated waveform measurements of input signals at different radio frequency (RF) powers, DC biases, ambient temperatures and load conditions.

19. The method of claim 18 , wherein the different load conditions are provided by a load tuner or an active load-pull system.

20. A system for generating a nonlinear simulation model of a nonlinear electronic component, the system comprising:

a vector network analyzer configured to provide nonlinear waveform data of the electronic component;

a tuner configured to stimulate simultaneously input and output ports of the electronic component at multiple frequencies to generate responses of the electronic component for a range of stimulus signals;

a processing device configured to identify nonlinear constitutive relations of a model current source and a model charge source of the electronic component as functions of at least one instantaneous terminal voltage, temperature and at least one trap state of the electronic component based on the nonlinear waveform data; and

artificial neural networks configured to represent the identified nonlinear constitutive relations, the represented nonlinear constitutive relations being compiled for implementation into a nonlinear circuit simulator to generate the nonlinear simulation model of the electronic component.