CN114239389A - High-fidelity modeling method and system for microwave radio frequency process IP simulation model - Google Patents
High-fidelity modeling method and system for microwave radio frequency process IP simulation model Download PDFInfo
- Publication number
- CN114239389A CN114239389A CN202111484853.3A CN202111484853A CN114239389A CN 114239389 A CN114239389 A CN 114239389A CN 202111484853 A CN202111484853 A CN 202111484853A CN 114239389 A CN114239389 A CN 114239389A
- Authority
- CN
- China
- Prior art keywords
- neural network
- simulation
- radio frequency
- network model
- microwave radio
- 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.)
- Granted
Links
Images
Classifications
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F30/00—Computer-aided design [CAD]
- G06F30/20—Design optimisation, verification or simulation
- G06F30/27—Design optimisation, verification or simulation using machine learning, e.g. artificial intelligence, neural networks, support vector machines [SVM] or training a model
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/04—Architecture, e.g. interconnection topology
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06N—COMPUTING ARRANGEMENTS BASED ON SPECIFIC COMPUTATIONAL MODELS
- G06N3/00—Computing arrangements based on biological models
- G06N3/02—Neural networks
- G06N3/08—Learning methods
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Physics & Mathematics (AREA)
- Evolutionary Computation (AREA)
- Software Systems (AREA)
- Artificial Intelligence (AREA)
- General Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Computing Systems (AREA)
- Mathematical Physics (AREA)
- General Health & Medical Sciences (AREA)
- Molecular Biology (AREA)
- Computational Linguistics (AREA)
- Biophysics (AREA)
- Biomedical Technology (AREA)
- Data Mining & Analysis (AREA)
- Life Sciences & Earth Sciences (AREA)
- Health & Medical Sciences (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Medical Informatics (AREA)
- Computer Hardware Design (AREA)
- Geometry (AREA)
- Monitoring And Testing Of Transmission In General (AREA)
- Measurement Of Resistance Or Impedance (AREA)
Abstract
The invention discloses a high fidelity modeling method and a high fidelity modeling system of a microwave radio frequency process IP simulation model, relating to the technical field of radio frequency electrical property simulation, wherein the modeling method comprises the following steps: s1, selecting a microwave radio frequency transmission structure, and parameterizing the selected microwave radio frequency transmission structure; s2, establishing an electromagnetic simulation model of the microwave radio frequency transmission structure, establishing a pre-training neural network model based on simulation sample data, and then obtaining an accompanying neural network model of the pre-training neural network model; s3, establishing an actual measurement neural network model according to the real object sample; and S4, carrying out fusion prediction on the microwave radio frequency transmission structure by adopting the accompanying neural network model of the pre-trained neural network model and the actually measured neural network model. The invention solves the problems of large sample quantity requirement, high parameter extraction cost, difficulty in obtaining accurate and sufficient data required by process IP simulation modeling and the like in the prior art.
Description
Technical Field
The invention relates to the technical field of radio frequency electrical property simulation, in particular to a high-fidelity modeling method and system for a microwave radio frequency process IP simulation model.
Background
The interconnection transmission structure in the radio frequency microwave modeling simulation is crucial to the simulation design, and the influence rule of the process fluctuation level on the electrical performance is encapsulated into a reusable process IP (intellectual property) electrical performance simulation model by the interconnection transmission structure related in the radio frequency microwave product integration process and adopting a mathematical modeling and software engineering technology, so that the radio frequency electrical performance rapid simulation design based on the actual manufacturing process fluctuation level can be realized, the matching between the product design performance and the manufacturing process is improved, and the development and testing period of the radio frequency microwave product is shortened.
Patent ZL2018101404852 discloses a modeling and packaging method for radio frequency integration process tolerance and electrical performance coupling characteristics, which can model the radio frequency integration process tolerance and electrical performance coupling characteristics in an IP model manner, and realize the analysis of product performance influence caused by manufacturing fluctuation by adjusting the process parameter tolerance of the model in the simulation design of a radio frequency system; however, the method needs to rely on more sample points to obtain a simulation model with certain precision, and has strict requirements on the distribution characteristics of the sample points. Patent CN109657390A discloses a statistical modeling method for process IP in radio frequency integrated manufacturing, which can realize model development with statistical distribution rule reflecting process fluctuation through data processing, algorithm modeling and simulation design package construction, and realize analysis of product performance distribution caused by manufacturing level fluctuation through statistical analysis method in simulation design; patent ZL2012103626130 discloses a radio frequency model method of a through silicon via array structure, which establishes a through silicon via high-frequency characteristic model by considering parasitic inductance, parasitic resistance, skin effect and the like in the through silicon via array structure; but is limited to radio frequency modeling of the silicon through hole array structure under strict arrangement relation and length relation.
The process IP simulation model needs to be capable of representing the influence of process parameters on the electrical property under micron-sized fine fluctuation, and strong nonlinear coupling exists between the process parameters and the electrical property, so that the actual measurement and extraction of the electrical property parameters under high frequency are very easily influenced by test errors; meanwhile, the development of a process IP simulation model is greatly restricted due to the large requirement on the number of samples and high parameter extraction cost, so that accurate and sufficient data required by process IP simulation modeling is difficult to obtain only by means of actual measurement. The coupling influence rule between the process parameters and the electrical property can be obtained by establishing an electromagnetic field simulation model corresponding to the process IP structure, but a simulation result has larger deviation with actual measurement, and no applicable method is available at present for realizing the fusion correction of the electromagnetic coupling rule of the radio frequency transmission structure simulation model and the actual measurement data and the corresponding high-fidelity modeling and application.
Disclosure of Invention
In order to overcome the defects of the prior art, the invention provides a high-fidelity modeling method and system for a microwave radio frequency process IP simulation model, and solves the problems that the prior art is large in sample number requirement, high in parameter increasing cost, difficult to obtain accurate and sufficient data required by process IP simulation modeling and the like.
The technical scheme adopted by the invention for solving the problems is as follows:
a high-fidelity modeling method for an IP simulation model of a microwave radio frequency process comprises the following steps:
s1, selecting a microwave radio frequency transmission structure, and parameterizing the selected microwave radio frequency transmission structure;
s2, establishing a pre-training neural network model based on simulation sample data, and then obtaining an accompanying neural network model of the pre-training neural network model;
s3, establishing an actual measurement neural network model according to the real object sample;
and S4, carrying out fusion prediction on the microwave radio frequency transmission structure by adopting the accompanying neural network model of the pre-trained neural network model and the actually measured neural network model.
As a preferred technical solution, the step S2 includes the following steps:
s2-1, setting M simulation samples;
s2-2, setting simulation process parameters and simulation scattering parameters;
s2-3, training a neural network based on simulation sample data to obtain a pre-training neural network model A;
s2-4, solving an accompanying neural network model A' of A;
wherein M is an integer and M > 1.
As a preferred technical solution, the step S3 includes the following steps:
s3-1, selecting the N simulation samples in the step S2-1 to prepare N physical sample pieces;
s3-2, setting actually measured process parameters and actually measured scattering parameters;
s3-3, training a neural network based on the real object sample data to obtain an actually measured neural network model B;
wherein N is an integer, N > 1 and N < M.
As a preferred technical solution, the step S4 includes the following steps:
s4-1, setting a process parameter value of the radio frequency transmission structure to be predicted;
s4-2, calculating scattering parameter variation delta S1 relative to a reference process parameter value caused by process parameter value fluctuation by using a concomitant neural network model A';
s4-3, calculating an actually measured scattering parameter value S0 under the condition of the reference process parameter by using the actually measured neural network model B;
and S4-4, adding the S0 and the delta S1 to obtain a corrected predicted scattering parameter value.
As a preferred technical solution, in step S1, a microstrip transmission line, a coplanar waveguide, a strip transmission line, a multi-layer vertical interconnection via, a gold wire interconnection transmission structure, and/or a similar coaxial transmission structure is selected.
As a preferred technical solution, in step S2, an electromagnetic field full-wave simulation algorithm or a high-frequency algorithm is used to establish an electromagnetic simulation model, and simulation sample data is obtained.
As a preferred solution, in step S2, the sample points of different types of process parameters are combined to form an orthogonal space.
As a preferred solution, in step S2, the sample points of different types of process parameters are combined to form an orthogonal space and uniformly distributed.
As a preferable technical solution, in step S2, if there is a process parameter strongly non-linearly coupled to the electrical performance, the number of sample points of the process parameter in the orthogonal space is increased.
A microwave radio frequency process IP simulation model high fidelity modeling system is based on the microwave radio frequency process IP simulation model high fidelity modeling method and comprises the following modules:
a parameterization module to: selecting a microwave radio frequency transmission structure, and parameterizing the selected microwave radio frequency transmission structure;
a simulation companion modeling module to: establishing an electromagnetic simulation model of a microwave radio frequency transmission structure, establishing a pre-training neural network model based on simulation sample data, and then obtaining an accompanying neural network model of the pre-training neural network model;
an actual measurement model making module for: establishing an actually measured neural network model according to the real object sample;
a fusion prediction module to: and carrying out fusion prediction on the microwave radio frequency transmission structure by adopting an accompanying neural network model of the pre-trained neural network model and the actually measured neural network model.
Compared with the prior art, the invention has the following beneficial effects:
(1) the modeling method is used for realizing high-fidelity modeling of the process IP simulation model by fusing electromagnetic field simulation results under the condition of adopting less actually measured modeling sample data, is particularly suitable for modeling micron-sized process tolerance fluctuation and electrical property coupling rule of a broadband high-frequency radio frequency transmission structure related in radio frequency microwave product simulation design, and can remarkably improve the modeling effect of the process IP simulation model under the condition of small sample amount; the problems that in the prior art, the requirement on the number of samples is large, the parameter extraction cost is high, accurate and sufficient data required by process IP simulation modeling is difficult to obtain and the like are solved;
(2) the invention realizes the fusion and correction of the electromagnetic field simulation result and the actual measurement result, realizes the high-fidelity modeling of the process IP simulation model on micron-sized process parameter fluctuation, reduces the number of physical samples required by modeling, and can realize the rapid simulation calculation of the link-level simulation design of the radio frequency product system because the established process IP simulation model is closer to the actual result than the electromagnetic field simulation model.
Drawings
FIG. 1 is a schematic diagram of steps of a high-fidelity modeling method of a microwave radio frequency process IP simulation model of the invention;
FIG. 2 is a schematic structural diagram of a high-fidelity modeling system of a microwave radio-frequency process IP simulation model according to the invention;
FIG. 3 is a flow chart of a high fidelity modeling method for an IP simulation model of a microwave radio frequency process according to a preferred embodiment of the invention;
FIG. 4 is a high fidelity modeling calculation logic flow diagram of the microwave RF process IP simulation model in accordance with a preferred embodiment of the present invention;
FIG. 5 is a diagram illustrating a neural network algorithm according to a preferred embodiment of the present invention;
FIG. 6 is a schematic diagram of a companion network algorithm according to a preferred embodiment of the present invention;
FIG. 7 is a schematic diagram of the high fidelity modeling process of the microwave RF process IP simulation model according to a preferred embodiment of the present invention.
Detailed Description
The present invention will be described in further detail with reference to examples and drawings, but the present invention is not limited to these examples.
Example 1
As shown in fig. 1 to 7, the present invention provides a method and a system for modeling a high fidelity radio frequency process IP simulation model, which can be applied to the existing typical radio frequency transmission structure, so as to realize the high fidelity modeling of the process IP simulation model by fusing electromagnetic field simulation results under the condition of using less actual measurement modeling sample data, and is particularly suitable for modeling micron-scale process tolerance fluctuation and electrical property coupling rule of a broadband high frequency radio frequency transmission structure involved in radio frequency microwave product simulation design. The method can remarkably improve the modeling effect of the process IP simulation model under a small sample size.
The high-fidelity modeling method of the microwave radio frequency process IP simulation model comprises the following steps: selecting a microwave radio frequency transmission structure; setting M fine parameterized simulation samples; setting simulation process parameters and acquiring corresponding simulation scattering parameters; training a neural network to obtain a pre-training model A; acquiring an accompanying network model A' of a pre-training model A; preparing N accurate actual measurement sample pieces; actually measuring the technological parameters and scattering parameters of the sample; training a neural network B; and obtaining correction output according to the fusion prediction of A' and B.
Step one, selecting a microwave radio frequency transmission structure, and parameterizing the selected microwave radio frequency transmission structure.
A common typical passive transmission structure such as a microstrip transmission line, a coplanar waveguide, a strip transmission line, a multilayer vertical interconnection via hole, a gold wire interconnection transmission structure, a quasi-coaxial transmission structure and the like is designed for radio frequency product simulation, process parameters sensitive to the influence of process tolerance fluctuation on electrical performance in the selected structure are determined as modeling parameters, and a reusable part in the structure is determined as a modeling object.
And step two, designing a corresponding simulation model and simulation scanning process parameters according to the step one, establishing a pre-training model based on simulation sample data and solving an accompanying network model of the pre-training model.
Preferably, the simulation model design generally adopts a model supporting an electromagnetic field full-wave simulation algorithm or a high-frequency algorithm for solving; the process parameter simulation scanning setting ensures that sample points of different types of process parameters are combined to form an orthogonal space and are uniformly distributed; in particular, the number of parameter points within the scan range should be increased for process parameters that have strong nonlinear coupling with electrical performance.
Preferably, different sets of process Parameter sample data and corresponding scattering parameters (Scatter-Parameter) are obtained through simulation solving, a pre-training model with process parameters as input and scattering parameters as output is established by adopting a neural network algorithm, and the pre-training model has the capability of representing the influence rule of process Parameter fluctuation on the scattering parameters with high fidelity.
Further, a adjoint network model with the same network parameters as the pre-trained model is obtained by solving the partial derivatives of the scattering parameters to the process parameters based on the pre-trained model, so as to represent the variation of the scattering parameters caused by the process parameter variation.
The input of the accompanying network model is each process parameter, and the output is scattering parameter variation.
And step three, selecting a plurality of groups of different process parameter samples to prepare corresponding physical actual measurement sample pieces according to the simulation model designed in the step two, and training a neural network according to actual measurement data so as to evaluate the mapping relation between the measured process parameters and the scattering parameters with high precision.
And step four, adopting an accompanying network model and an actually measured neural network model to perform fusion prediction.
The fusion prediction mode is that aiming at a process tolerance parameter value needing to be predicted, a scattering parameter variation of the process tolerance relative to a reference process parameter is calculated by an accompanying network, an actually measured scattering parameter value under the reference process parameter is calculated by a neural network trained by actually measured data, and a predicted scattering parameter which is close to an actual value after being corrected is obtained by adding the actually measured scattering parameter value and the scattering parameter variation.
The invention provides a high-fidelity modeling and application storage medium and equipment for a microwave radio-frequency process IP simulation model.
The embodiment of the invention provides equipment, which comprises a memory, a processor and a computer program which is stored on the memory and can run on the processor, wherein when the computer program is executed by the processor, the high-fidelity modeling or application of an IP simulation model of a microwave radio frequency process is realized.
The invention realizes the fusion and correction of the electromagnetic field simulation result and the actual measurement result, realizes the high-fidelity modeling of the process IP simulation model on micron-sized process parameter fluctuation, reduces the number of physical samples required by modeling, and can realize the rapid simulation calculation of the link-level simulation design of the radio frequency product system because the established process IP simulation model is closer to the actual result than the electromagnetic field simulation model.
Example 2
As shown in fig. 1 to 7, as a further optimization of embodiment 1, this embodiment includes all the technical features of embodiment 1, and in addition, this embodiment further includes the following technical features:
the general flow of the scheme of the invention is shown in fig. 7, and is characterized in that: selecting a microwave radio frequency transmission structure; setting M fine parameterized simulation samples; setting simulation process parameters and acquiring corresponding simulation scattering parameters; training a neural network to obtain a pre-training model A; acquiring an accompanying network model A' of a pre-training model A; preparing N accurate actual measurement sample pieces; actually measuring the technological parameters and scattering parameters of the sample; training a neural network B; and obtaining correction output according to the fusion prediction of A' and B.
And S1, selecting the microwave radio frequency transmission structure, and parameterizing the selected microwave radio frequency transmission structure.
A common typical passive transmission structure such as a microstrip transmission line, a coplanar waveguide, a strip transmission line, a multilayer vertical interconnection via hole, a gold wire interconnection transmission structure, a quasi-coaxial transmission structure and the like is designed for radio frequency product simulation, process parameters sensitive to the influence of process tolerance fluctuation on electrical performance in the selected structure are determined as modeling parameters, and a reusable part in the structure is determined as a modeling object. As shown in fig. 7, the process parameters in the selected gold wire cascade radio frequency transmission process IP model are gold wire arc height, span, and gold wire pitch, and the reusable part is gold wire and the circuit chip at the connection of the bonding pads at both ends.
S2-1, designing a corresponding simulation model according to S1, setting simulation scanning process parameters GA in S2-2, obtaining a pre-training model A through training a neural network based on simulation sample data in S2-3, and solving an accompanying network A' of the pre-training model A in S2-4.
Preferably, the simulation model design generally adopts a model supporting an electromagnetic field full-wave simulation algorithm or a high-frequency algorithm for solving; in S2-2, setting simulation scanning of process parameters to ensure that sample points of different types of process parameters are combined to form an orthogonal space and are uniformly distributed; in particular, the number of parameter points within the scan range should be increased for process parameters that have strong nonlinear coupling with electrical performance. The gold wire cascade structure shown in fig. 7 increases the number of scanning points of the arch height parameter due to the strong nonlinear coupling of the arch height parameter and the electrical performance.
Preferably, different sets of process Parameter sample data and corresponding scattering parameters (Scatter-Parameter) are obtained through simulation solving, and a pre-training model A with process parameters as input and scattering parameters as output is established by adopting the neural network model shown in FIG. 7, wherein the pre-training model A has the capability of representing the influence rule of process Parameter fluctuation on the scattering parameters with high fidelity.
Further, in S2-4, based on the pre-trained model, by calculating the partial derivative of the scattering parameter to each process parameter, an accompanying network model A' having the same network parameter as the pre-trained model is obtained, and the model structure is as shown in FIG. 6
The method is used for representing the variation of the scattering parameters caused by the variation of the process parameters.
The input of the accompanying network model A' is each process parameter, and the output is the scattering parameter variation.
S3-1, selecting several groups of different process parameter samples to prepare corresponding physical actual measurement sample pieces according to a simulation model designed by S2-1, and establishing a neural network B by adopting a neural network algorithm structure shown in FIG. 7 according to actual measurement data obtained by S3-2 in S3-3 so as to represent the mapping relation between actual measurement process parameters and scattering parameters with high precision.
And S4, performing fusion prediction by adopting the adjoint network model A' and the measured neural network model B.
The fusion prediction mode is as shown in fig. 6, and includes setting a process parameter value of the radio frequency transmission structure to be predicted, calculating a scattering parameter variation of the process tolerance relative to a reference process parameter by an accompanying network model a', calculating an actual measurement scattering parameter value under the reference process parameter by a neural network B trained by actual measurement data, and adding the scattering parameter variation to the actual measurement scattering parameter value to obtain a predicted scattering parameter which is corrected to be close to the true value. As shown in fig. 7, the fluctuation Δ S1 of the scattering parameter of the process parameter G1 with respect to the reference process parameter G0 is calculated in sequence by the accompanying network model a', the scattering parameter S0 of the reference process parameter G0 is calculated by the neural network B, and finally the fluctuation Δ S1 is added to obtain the final prediction result.
In FIG. 6, GiRepresents the ith process parameter of the first process,
represents the jth neuron and G of the l-1 layer of the neural networkiConnection weight of bl-1Representing the bias value of the l-1 layer of the neural network,
represents the input of the jth neuron on the l-1 layer of the neural network,
representing the output, σ, of the jth neuron at layer l-1 of the neural networkl-1(. cndot.) represents the activation function of the l-1 layer of the neural network,
representing the k-th neuron of the l layer of the neural network and the output of the previous layer
Connection weight of blRepresents the bias value of the l layer of the neural network,
represents the input of the k-th neuron at the l layer of the neural network,
representing the output, σ, of the kth neuron of layer l of the neural networkl(. cndot.) represents the activation function of the l-layer of the neural network,
representsOutput of mth neuron of l +1 layer of neural network and previous layer
Connection weight of bl+1Representing the bias value, S, of the l +1 layer of the neural networkmRepresenting the mth output, Δ G, of the neural networkiRepresenting the fluctuation quantity, Δ S, of the ith process parametermiRepresenting the mth output factor Δ G of the neural networkiThe amount of fluctuation brought about.
As described above, the present invention can be preferably realized.
All features disclosed in all embodiments in this specification, or all methods or process steps implicitly disclosed, may be combined and/or expanded, or substituted, in any way, except for mutually exclusive features and/or steps.
The foregoing is only a preferred embodiment of the present invention, and the present invention is not limited thereto in any way, and any simple modification, equivalent replacement and improvement made to the above embodiment within the spirit and principle of the present invention still fall within the protection scope of the present invention.
Claims (10)
1. A high-fidelity modeling method for an IP simulation model of a microwave radio frequency process is characterized by comprising the following steps:
s1, selecting a microwave radio frequency transmission structure, and parameterizing the selected microwave radio frequency transmission structure;
s2, establishing an electromagnetic simulation model of the microwave radio frequency transmission structure, establishing a pre-training neural network model based on simulation sample data, and then obtaining an accompanying neural network model of the pre-training neural network model;
s3, establishing an actual measurement neural network model according to the real object sample;
and S4, carrying out fusion prediction on the microwave radio frequency transmission structure by adopting the accompanying neural network model of the pre-trained neural network model and the actually measured neural network model.
2. The microwave radio frequency process IP simulation model high fidelity modeling method according to claim 1, characterized in that the step S2 includes the following steps:
s2-1, setting M simulation samples;
s2-2, measuring the process parameters and the scattering parameters of the sample;
s2-3, training a neural network based on simulation sample data to obtain a pre-training neural network model A;
s2-4, solving an accompanying neural network model A' of A;
wherein M is an integer and M > 1.
3. The microwave radio frequency process IP simulation model high fidelity modeling method according to claim 2, characterized in that the step S3 includes the following steps:
s3-1, selecting the N simulation samples in the step S2-1 to prepare N physical sample pieces;
s3-2, setting actually measured process parameters and actually measured scattering parameters;
s3-3, training a neural network based on the real object sample data to obtain an actually measured neural network model B;
wherein N is an integer, N > 1 and N < M.
4. The microwave radio frequency process IP simulation model high fidelity modeling method according to claim 3, characterized in that the step S4 includes the following steps:
s4-1, setting a process parameter value of the radio frequency transmission structure to be predicted;
s4-2, calculating scattering parameter variation delta S1 relative to a reference process parameter value caused by process parameter value fluctuation by using a concomitant neural network model A';
s4-3, calculating an actually measured scattering parameter value S0 under the condition of the reference process parameter by using the actually measured neural network model B;
and S4-4, adding the S0 and the delta S1 to obtain a corrected predicted scattering parameter value.
5. The high-fidelity modeling method for the IP simulation model of the microwave radio-frequency process according to any one of claims 1 to 4, characterized in that in step S1, a microstrip transmission line, a coplanar waveguide, a strip transmission line, a multilayer vertical interconnection via, a gold wire interconnection transmission structure and/or a similar coaxial transmission structure is/are selected.
6. The microwave radio-frequency process IP simulation model high-fidelity modeling method according to claim 5, characterized in that in step S2, an electromagnetic field full-wave simulation algorithm or a high-frequency algorithm is used to build an electromagnetic simulation model to obtain simulation sample data.
7. The microwave radio frequency process IP simulation model high fidelity modeling method of claim 6, wherein in step S2, the sample points of different types of process parameters are combined to form an orthogonal space.
8. The microwave radio frequency process IP simulation model high fidelity modeling method of claim 7, wherein in step S2, the sample points of different types of process parameters are combined to form an orthogonal space and are distributed uniformly.
9. The microwave radio frequency process IP simulation model high fidelity modeling method of claim 8, wherein in step S2, if there is a process parameter strongly non-linearly coupled to the electrical property, the number of sample points of the process parameter in the orthogonal space is increased.
10. A high-fidelity modeling system of a microwave radio-frequency process IP simulation model, which is characterized in that based on any one of claims 1 to 9, the high-fidelity modeling method of the microwave radio-frequency process IP simulation model comprises the following modules:
a parameterization module to: selecting a microwave radio frequency transmission structure, and parameterizing the selected microwave radio frequency transmission structure;
a simulation companion modeling module to: establishing a pre-training neural network model based on simulation sample data, and then obtaining an accompanying neural network model of the pre-training neural network model;
an actual measurement model making module for: establishing an actually measured neural network model according to the real object sample;
a fusion prediction module to: and carrying out fusion prediction on the microwave radio frequency transmission structure by adopting an accompanying neural network model of the pre-trained neural network model and the actually measured neural network model.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111484853.3A CN114239389B (en) | 2021-12-07 | 2021-12-07 | High-fidelity modeling method and system for microwave radio frequency process IP simulation model |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202111484853.3A CN114239389B (en) | 2021-12-07 | 2021-12-07 | High-fidelity modeling method and system for microwave radio frequency process IP simulation model |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN114239389A true CN114239389A (en) | 2022-03-25 |
| CN114239389B CN114239389B (en) | 2023-04-25 |
Family
ID=80753679
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202111484853.3A Active CN114239389B (en) | 2021-12-07 | 2021-12-07 | High-fidelity modeling method and system for microwave radio frequency process IP simulation model |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN114239389B (en) |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2023215477A1 (en) * | 2022-05-06 | 2023-11-09 | Viasat, Inc. | Semiconductor package inspection with predictive model for wirebond radio frequency performance |
Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108345749A (en) * | 2018-02-11 | 2018-07-31 | 中国电子科技集团公司第二十九研究所 | Modeling and packaging method of the radio frequency integrated technique tolerance with electrical property coupled characteristic |
| CN111460617A (en) * | 2020-03-03 | 2020-07-28 | 华中科技大学 | IGBT junction temperature prediction method based on neural network |
| CN112632720A (en) * | 2020-12-16 | 2021-04-09 | 广东省科学院中乌焊接研究所 | Multidimensional data fusion and quantitative modeling method for metal additive manufacturing process system |
-
2021
- 2021-12-07 CN CN202111484853.3A patent/CN114239389B/en active Active
Patent Citations (3)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN108345749A (en) * | 2018-02-11 | 2018-07-31 | 中国电子科技集团公司第二十九研究所 | Modeling and packaging method of the radio frequency integrated technique tolerance with electrical property coupled characteristic |
| CN111460617A (en) * | 2020-03-03 | 2020-07-28 | 华中科技大学 | IGBT junction temperature prediction method based on neural network |
| CN112632720A (en) * | 2020-12-16 | 2021-04-09 | 广东省科学院中乌焊接研究所 | Multidimensional data fusion and quantitative modeling method for metal additive manufacturing process system |
Non-Patent Citations (2)
| Title |
|---|
| 李阳阳等: "宽带射频 SiP 集成工艺 IP 建模仿真技术及应用" * |
| 程加力: "射频微波MOS器件参数提取与建模技术研究" * |
Cited By (1)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| WO2023215477A1 (en) * | 2022-05-06 | 2023-11-09 | Viasat, Inc. | Semiconductor package inspection with predictive model for wirebond radio frequency performance |
Also Published As
| Publication number | Publication date |
|---|---|
| CN114239389B (en) | 2023-04-25 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| CN106844924B (en) | Method for optimizing PCB microstrip line structure based on response surface method and genetic algorithm | |
| US7157918B2 (en) | Method and system for calibrating a measurement device path and for measuring a device under test in the calibrated measurement device path | |
| CN110470966B (en) | Scattering parameter measuring method and device calibration method | |
| CN107832526B (en) | Method for optimizing return loss of BGA welding spot | |
| US9922888B2 (en) | General four-port on-wafer high frequency de-embedding method | |
| CN106777620A (en) | A kind of neutral net space reflection modeling method for power transistor | |
| CN114239389A (en) | High-fidelity modeling method and system for microwave radio frequency process IP simulation model | |
| CN116305522A (en) | Digital twin-based aircraft structure reliability simulation test model calibration method, device and medium | |
| CN106646193A (en) | Bonding wire parasitic parameter testing and extracting method | |
| US8185864B2 (en) | Circuit board analyzer and analysis method | |
| CN117607555B (en) | Electromagnetic parameter testing method, system and storage medium for microwave antenna | |
| Wang et al. | Multi‐harmonic sources location based on sparse component analysis and complex independent component analysis | |
| US7062424B2 (en) | Circuit modeling using topologically equivalent models | |
| CN101477582B (en) | Model modification method for a semiconductor device | |
| CN113655360A (en) | De-embedding method of on-chip test structure of RF MOS device | |
| Mubarak et al. | Calculating S-parameters and uncertainties of coaxial air-dielectric transmission lines | |
| Lin et al. | Fast and accurate yield rate prediction of PCB embedded common-mode filter with artificial neural network | |
| CN110688810A (en) | Method for modeling thermal effect of radio frequency LDMOS (laterally diffused metal oxide semiconductor) by utilizing neural network | |
| Pal et al. | Computation of Resonant Frequency and Gain from Inset Fed Rectangular Shaped Microstrip Patch Antenna Using Deep Neural Network | |
| Stewart et al. | Microstrip discontinuity modeling | |
| Bae et al. | Numerical verification of dielectric contactor as auxiliary loads for measuring the multi-port network parameter of vertical interconnection array | |
| Poddar et al. | Accurate, high speed modeling of integrated passive devices in multichip modules | |
| CN116449183B (en) | De-embedding structure and method for on-chip test of radio frequency chip, storage medium and terminal | |
| CN107729694B (en) | Multi-parameter electromagnetic field modeling simulation method based on neural network | |
| Koziel et al. | Fast tolerance-aware design optimization of miniaturized microstrip couplers using variable-fidelity EM simulations and response features |
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 |