CN113591390B - Model screening method and platform for nonlinear effect of receiver radio frequency link - Google Patents
Model screening method and platform for nonlinear effect of receiver radio frequency link Download PDFInfo
- Publication number
- CN113591390B CN113591390B CN202110908338.7A CN202110908338A CN113591390B CN 113591390 B CN113591390 B CN 113591390B CN 202110908338 A CN202110908338 A CN 202110908338A CN 113591390 B CN113591390 B CN 113591390B
- Authority
- CN
- China
- Prior art keywords
- model
- test
- receiver
- modeling
- radio frequency
- 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
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
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F18/00—Pattern recognition
- G06F18/20—Analysing
- G06F18/21—Design or setup of recognition systems or techniques; Extraction of features in feature space; Blind source separation
- G06F18/214—Generating training patterns; Bootstrap methods, e.g. bagging or boosting
-
- G—PHYSICS
- G06—COMPUTING; CALCULATING OR COUNTING
- G06F—ELECTRIC DIGITAL DATA PROCESSING
- G06F2119/00—Details relating to the type or aim of the analysis or the optimisation
- G06F2119/02—Reliability analysis or reliability optimisation; Failure analysis, e.g. worst case scenario performance, failure mode and effects analysis [FMEA]
-
- Y—GENERAL TAGGING OF NEW TECHNOLOGICAL DEVELOPMENTS; GENERAL TAGGING OF CROSS-SECTIONAL TECHNOLOGIES SPANNING OVER SEVERAL SECTIONS OF THE IPC; TECHNICAL SUBJECTS COVERED BY FORMER USPC CROSS-REFERENCE ART COLLECTIONS [XRACs] AND DIGESTS
- Y02—TECHNOLOGIES OR APPLICATIONS FOR MITIGATION OR ADAPTATION AGAINST CLIMATE CHANGE
- Y02D—CLIMATE CHANGE MITIGATION TECHNOLOGIES IN INFORMATION AND COMMUNICATION TECHNOLOGIES [ICT], I.E. INFORMATION AND COMMUNICATION TECHNOLOGIES AIMING AT THE REDUCTION OF THEIR OWN ENERGY USE
- Y02D30/00—Reducing energy consumption in communication networks
- Y02D30/70—Reducing energy consumption in communication networks in wireless communication networks
Landscapes
- Engineering & Computer Science (AREA)
- Theoretical Computer Science (AREA)
- Evolutionary Computation (AREA)
- Physics & Mathematics (AREA)
- General Engineering & Computer Science (AREA)
- Artificial Intelligence (AREA)
- Computer Vision & Pattern Recognition (AREA)
- Data Mining & Analysis (AREA)
- General Physics & Mathematics (AREA)
- Medical Informatics (AREA)
- Geometry (AREA)
- Computer Hardware Design (AREA)
- Software Systems (AREA)
- Life Sciences & Earth Sciences (AREA)
- Bioinformatics & Cheminformatics (AREA)
- Bioinformatics & Computational Biology (AREA)
- Evolutionary Biology (AREA)
- Monitoring And Testing Of Transmission In General (AREA)
Abstract
The invention provides a model screening method and a platform for nonlinear effects of a receiver radio frequency link. The method comprises the steps of obtaining a signal source and test output data of a receiver; selecting a pre-test assumption model by combining the existing prior knowledge with a receiver mechanism; identifying model parameters of the pre-test assumed model to obtain a pre-test assumed model after parameter identification; based on the signal source and the test output data, comparing the obtained result with the real system result by adopting a pre-test assumed model after parameter identification, judging whether the error reaches a set threshold value, if so, obtaining a modeled model, otherwise, reselecting the pre-test assumed model to carry out the parameter identification and verification process.
Description
Technical Field
The invention belongs to the technical field of modeling, and particularly relates to a model screening method and a platform for nonlinear effects of a receiver radio frequency link.
Background
The statements in this section merely provide background information related to the present disclosure and may not necessarily constitute prior art.
At present, modeling of a radio frequency link nonlinear effect of a receiver is mainly focused on research of a high-precision model, particularly an X-parameter model and a model based on machine learning are used as emphasis, but in the modeling process, model precision is one aspect of determining modeling effect, and nonlinear effect testing and model parameter identification are also emphasis of determining modeling effect. In the work of researching nonlinear effect modeling, a test platform is basically built temporarily for testing, so that certain differences exist in different test platforms of the same link, and objective comparison is difficult when model accuracy is compared. Therefore, the prior art cannot select and display a plurality of models on the same platform, and cannot select the most suitable model according to the input data.
Disclosure of Invention
In order to solve the technical problems in the background art, the invention provides a model screening method and a platform for nonlinear effects of a receiver radio frequency link, which can provide a platform for multi-type modeling, and particularly provide the same-platform comparison of model precision under the same data in the research of a new model. Each model has the corresponding modeling advantage, and the model with the highest precision is selected under certain bandwidth and environment, so that the performance of the receiver link is greatly improved. The establishment of a nonlinear effect multi-type modeling platform of a receiver radio frequency link is particularly important.
In order to achieve the above purpose, the present invention adopts the following technical scheme:
the first aspect of the invention provides a model screening method for nonlinear effects of a radio frequency link of a receiver.
A model screening method for nonlinear effect of a receiver radio frequency link comprises the following steps:
acquiring a signal source and test output data of a receiver;
selecting a pre-test assumption model by combining the existing prior knowledge with a receiver mechanism;
identifying model parameters of the pre-test assumed model to obtain a pre-test assumed model after parameter identification;
based on the signal source and the test output data, comparing the obtained result with the real system result by adopting a pre-test assumed model after parameter identification, judging whether the error reaches a set threshold value, if so, obtaining a modeled model, otherwise, reselecting the pre-test assumed model to carry out the parameter identification and verification process.
A second aspect of the invention provides a model screening platform for nonlinear effects of a receiver radio frequency link.
A model screening platform for nonlinear effects of a receiver radio frequency link, comprising:
a data acquisition module configured to: acquiring a signal source and test output data of a receiver;
A model selection module configured to: selecting a pre-test assumption model by combining the existing prior knowledge with a receiver mechanism;
a parameter identification module configured to: identifying model parameters of the pre-test assumed model to obtain a pre-test assumed model after parameter identification;
a verification model configured to: based on the signal source and the test output data, comparing the obtained result with the real system result by adopting a pre-test assumed model after parameter identification, judging whether the error reaches a set threshold value, if so, obtaining a modeled model, otherwise, reselecting the pre-test assumed model to carry out the parameter identification and verification process.
A third aspect of the present invention provides a computer-readable storage medium.
A computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps in a method for model screening for receiver radio frequency link nonlinear effects as described in the first aspect above.
A fourth aspect of the invention provides a computer device.
A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, the processor implementing the steps in a method of model screening for non-linear effects of a receiver radio frequency link as described in the first aspect above when the program is executed.
Compared with the prior art, the invention has the beneficial effects that:
according to the invention, multiple types of modeling is performed according to actual test data, multiple model performances are compared according to the same nonlinear effect test data in the same system platform, and meanwhile, a developed novel model can be imported into the system to perform performance comparison with the existing multiple models, so that the model can be developed into a receiver nonlinear effect modeling performance comparison platform instrument, and technical support is provided for receiver development.
The invention is applicable to multi-type modeling of nonlinear effects of various types and frequency bands of receivers between 150MHz and 40GHz, and can provide selection of various memoryless models and memoryless models.
Additional aspects of the invention will be set forth in part in the description which follows and, in part, will be obvious from the description, or may be learned by practice of the invention.
Drawings
The accompanying drawings, which are included to provide a further understanding of the invention and are incorporated in and constitute a part of this specification, illustrate embodiments of the invention and together with the description serve to explain the invention.
FIG. 1 is a flow chart of a method for model screening of nonlinear effects of a receiver radio frequency link of the present invention;
FIG. 2 is a system frame diagram of a second embodiment of the present invention;
FIG. 3 is a diagram illustrating a process for identifying model parameters according to a second embodiment of the present invention;
FIG. 4 is a block diagram of an LS-SVM regression algorithm in accordance with a second embodiment of the present invention;
FIG. 5 is a flow chart of a comparison of multi-model performance in a second embodiment of the present invention;
FIG. 6 is a flow chart of model screening in an embodiment of the invention.
Detailed Description
The invention will be further described with reference to the drawings and examples.
It should be noted that the following detailed description is illustrative and is intended to provide further explanation of the invention. Unless defined otherwise, all technical and scientific terms used herein have the same meaning as commonly understood by one of ordinary skill in the art to which this invention belongs.
It is noted that the terminology used herein is for the purpose of describing particular embodiments only and is not intended to be limiting of exemplary embodiments according to the present invention. As used herein, the singular is also intended to include the plural unless the context clearly indicates otherwise, and furthermore, it is to be understood that the terms "comprises" and/or "comprising" when used in this specification are taken to specify the presence of stated features, steps, operations, devices, components, and/or combinations thereof.
It is noted that the flowcharts and block diagrams in the figures illustrate the architecture, functionality, and operation of possible implementations of methods and systems according to various embodiments of the present disclosure. It should be noted that each block in the flowchart or block diagrams may represent a module, segment, or portion of code, which comprises one or more executable instructions for implementing the logical functions specified in the various embodiments. It should also be noted that, in some alternative implementations, the functions noted in the block may occur out of the order noted in the figures. For example, two blocks shown in succession may, in fact, be executed substantially concurrently, or the blocks may sometimes be executed in the reverse order, depending upon the functionality involved. It will also be noted that each block of the flowchart illustrations and/or block diagrams, and combinations of blocks in the flowchart illustrations and/or block diagrams, can be implemented by special purpose hardware-based systems which perform the specified functions or operations, or combinations of special purpose hardware and computer instructions.
The behavior model is also called a black box model, and modeling can be realized by only measuring input and output signals of the device and extracting characteristic parameters through a defined model structure. The method avoids complex operation processes of physical model simulation and circuit model simulation, and obtains characteristics which are irrelevant to internal circuit details and can accurately represent the exterior of the measured object. Its advantages are easy protection of intellectual property and no need of lengthy design period.
The behavior model has two main types: a memory-free model and a memory-equipped model. The model suitable for the memoryless power amplifier comprises: polynomial models, saleh models, some feedforward neural network models, etc. These models are not accurate enough to characterize the non-linear characteristics of memory. At present, there are mainly the following memory behavior models, namely a Wiener model, a Hammerstein model, a Volterra series model, a simplified-memory polynomial model of the Volterra series model, and some neural network models, including: radial basis function (Radical Basis Function, RBF) models, feed forward (BP) neural network models of different topologies, and the like.
Although the neural network model has learning ability and approximation ability, the generalization ability of the model is weak, so that the neural network model is difficult to realize in a nonlinear device. The X parameter is used as a behavior model capable of providing a power amplifier modeling scheme under a large-signal nonlinear condition, and the nonlinear performance of the radio frequency power amplifier device can be accurately measured and predicted under a real large-signal working state; carrying out standardized behavior model modeling test on the power amplifier model; as a strict mathematical expansion in the theory of an S parameter model, the method has a strict theoretical basis, can utilize advanced measurement technology to parameterize behavior model identification like the S parameter, and measures identification parameters by using a nonlinear vector network analyzer (Nonlinear Vector Network Analyzer, NVNA); software platform modeling may be implemented by advanced design system (Advanced Design System, ADS) simulation techniques; the hardware measurement can be realized for the system level design, and the cooperation of the hardware measurement and the software simulation can be realized for the cascading model between the devices. The X parameter is used for modeling the behavior model of the radio frequency power amplifier, so that the characterization of the power amplifier under the response of a large signal can be obtained, meanwhile, the accurate calibration can be carried out, the cost can be effectively saved, and the research and development period of a circuit and a system can be shortened.
The consideration of the theory and structure of the X parameter model and the accurate research of the measurement and identification technology of the X parameter model are important points of nonlinear modeling research under a large signal. Aiming at the field of the X parameter combined load traction technology and the high generalization capability characterization memory effect, the method is still a mainstream research hot spot in the future. How to describe the nonlinear characteristic and long-term memory effect of the large signal more accurately, build the modeling scheme with high precision and high simulation efficiency, form the professional modeling simulation and industrial standard of prediction as the important research direction of X parameter in the future of the radio frequency circuit modeling technology, effectively apply and popularize the measurement and identification parameters of the accurate X parameter model and exert far-reaching influence on the radio frequency front end design.
The application of machine learning, and in particular deep learning techniques, has significantly increased the level of research in many areas. Meanwhile, the deep learning technology proves to be very effective in the aspect of mining high-dimensional characteristics of data, is suitable for establishing a functional relation between input and output of a radio frequency device, and can fully utilize a large amount of measured data to improve the subsequent modeling precision.
At present, modeling of a radio frequency link nonlinear effect of a receiver is mainly focused on research of a high-precision model, particularly an X-parameter model and a model based on machine learning are used as emphasis, but in the modeling process, model precision is one aspect of determining modeling effect, and nonlinear effect testing and model parameter identification are also emphasis of determining modeling effect. In the work of researching nonlinear effect modeling, a test platform is basically built temporarily for testing, so that certain differences exist in different test platforms of the same link, and objective comparison is difficult when model accuracy is compared. Meanwhile, because the single modeling period is longer, in order to avoid unnecessary modeling, modeling resources and modeling efficiency are wasted, and model screening is performed before modeling is started. And (3) comprehensively screening the models by utilizing a lookup table mode according to the signal source data and the test system test data, determining the types and the quantity of the models to be modeled, and not modeling all the models. Therefore, the establishment of a model screening method, a platform, a storage medium and equipment for nonlinear effects of a receiver radio frequency link is particularly important.
Example 1
As shown in fig. 1, this embodiment provides a model screening method for nonlinear effects of a radio frequency link of a receiver, and this embodiment is applied to a server for illustration by using the method, and it can be understood that the method can also be applied to a terminal, and can also be applied to a system and a terminal, and implemented through interaction between the terminal and the server. The server can be an independent physical server, a server cluster or a distributed system formed by a plurality of physical servers, and can also be a cloud server for providing cloud services, cloud databases, cloud computing, cloud functions, cloud storage, network servers, cloud communication, middleware services, domain name services, security services CDNs, basic cloud computing services such as big data and artificial intelligent platforms and the like. The terminal may be, but is not limited to, a smart phone, a tablet computer, a notebook computer, a desktop computer, a smart speaker, a smart watch, etc. The terminal and the server may be directly or indirectly connected through wired or wireless communication, and the present application is not limited herein. In this embodiment, the method includes the steps of:
s101: acquiring a signal source and test output data of a receiver;
S102: selecting a pre-test assumption model by combining the existing prior knowledge with a receiver mechanism;
s103: identifying model parameters of the pre-test assumed model to obtain a pre-test assumed model after parameter identification;
s104: based on the signal source and the test output data, comparing the obtained result with the real system result by adopting a pre-test assumed model after parameter identification, judging whether the error reaches a set threshold value, if so, obtaining a modeled model, otherwise, reselecting the pre-test assumed model to carry out the parameter identification and verification process.
As one or more embodiments, the signal source and test output data include an excitation signal for identifying parameters of the pre-test hypothesis model and a verification signal for verifying output results of the pre-test hypothesis model after the parameter identification. The test output data includes: the system comprises a real-time frequency spectrum, a two-dimensional/three-dimensional dual-frequency diagram, an original signal waveform, an input/output power diagram under a fixed frequency point of a power compression test subsystem, a 1dB compression diagram under different frequencies, a constellation diagram of a power compression three-dimensional diagram, a constellation diagram of a amplitude-phase error test subsystem, a link gain, a noise coefficient and sensitivity of a link parameter test subsystem, a third-order cut-off point, an adjacent channel power ratio and a noise power ratio of a theoretical parameter calculation subsystem of the intermodulation test subsystem.
Specifically, input/output data selection: data is the basis for the establishment of a behavior model, and is often acquired from an established actual system, and is used for grasping some priori knowledge of the system, and one part of the data is an excitation signal for identifying the system, and the other part of the data is a verification signal for subsequently verifying the model. The accuracy of system modeling is related to the excitation signal adopted, the excitation data can cover the whole frequency band of the system, namely, the bandwidth range of the receiver is considered, the sample data volume is ensured to be large enough as far as possible, the information content is sufficient and complete, valuable information can be provided, and the extraction of model features is facilitated. The excitation signals commonly used include single-tone signals, multi-tone signals, digital modulation signals, random signals, etc., and the selection of the excitation signals is generally dependent on the specific situation.
As one or more embodiments, the pre-test hypothesis model employs a nonlinear model. In particular, determining the model type is a crucial step, and by using the existing prior knowledge, a suitable pre-test assumed model is selected by means of mechanism analysis and the like, and factors such as whether model parameters are easy to identify are considered, generally, the model is selected from a simple model structure, and when the result is unsatisfactory, a more complex model can be replaced for re-experiment until the requirement is met. The models are divided into linear models and nonlinear models, and the receiver selects the nonlinear models to characterize the nonlinearities of the receiver due to their significant nonlinearities. The nonlinear model includes: a memory-free model and a memory-equipped model. The memoryless model comprises: a memoryless polynomial model, a power series model, an orthogonal polynomial model, and a lookup table model. The memoryless model comprises: the memory model comprises: a memory polynomial model, a Volterra model, a Hammerstein model, a Wiener-Hammerstein model, a PLUME model, a NARMA model, an X-parameter model.
As one or more embodiments, identifying model parameters of the pre-test hypothesis model includes: and identifying model parameters of the pre-test assumed model by adopting a least square method, a recursive identification algorithm, a maximum likelihood identification method and a Fourier series method.
Specifically, after the model structure is determined, the model parameters are identified, and the main purpose of parameter identification is to analyze the characterization function equation of the nonlinear system according to the system input and output state variables, find the determined time domain characteristic function model with the minimum matching error, and find the parameters of the function model. The identification process of the model parameters is shown in fig. 3. The conventional recognition methods include a least square method (Least square method, LS), a recursive recognition algorithm, a maximum likelihood recognition method, a Fourier series method and the like, and some new recognition algorithms in recent years include: neural network-based recognition methods, genetic algorithm-based recognition methods, fuzzy logic-based recognition methods, and the like. The choice of the recognition algorithm directly affects the accuracy of the final model.
As one or more embodiments, the index for comparing the obtained result with the actual system result includes: normalized mean square error, root mean square error, mean absolute error.
Specifically, after modeling is completed, to check the generalization capability of the obtained model, the error between the output of the built model and the output of the actual system is checked to determine whether the required accuracy is achieved by using the data set which does not participate in recognition under the condition of the same input, and in terms of determining the accuracy of the model, generally, no unified method is available, and the model can be verified through relevant performance indexes, such as Normalized Mean Square Error (NMSE), root Mean Square Error (RMSE), mean Absolute Error (MAE), and the like. And the model can be intuitively verified according to whether the model characteristics are consistent with the actual system. When the model precision does not reach the standard, returning to the steps to check gradually to see whether the excitation signal is unreasonable to select, or whether the model structure is erroneously selected or the parameter identification algorithm is inaccurate.
Example two
The embodiment provides a model screening platform for nonlinear effects of a receiver radio frequency link.
A model screening platform for nonlinear effects of a receiver radio frequency link, comprising:
a data acquisition module configured to: acquiring a signal source and test output data of a receiver;
a model selection module configured to: selecting a pre-test assumption model by combining the existing prior knowledge with a receiver mechanism;
A parameter identification module configured to: identifying model parameters of the pre-test assumed model to obtain a pre-test assumed model after parameter identification;
a verification model configured to: based on the signal source and the test output data, comparing the obtained result with the real system result by adopting a pre-test assumed model after parameter identification, judging whether the error reaches a set threshold value, if so, obtaining a modeled model, otherwise, reselecting the pre-test assumed model to carry out the parameter identification and verification process.
As an embodiment, taking the integrated test of the nonlinear effect of the radio frequency link of the receiver as an example, the application of the present invention in the integrated test of the nonlinear effect of the radio frequency link of the receiver is described, as shown in fig. 2:
the overall modeling scheme provided by the embodiment consists of five parts, namely a test system sharing system software platform, a tested receiver link, a system hardware platform, a high-speed signal processing module and an industrial personal computer, so that an integrated software and hardware scheme is formed, and a comprehensive test and modeling platform for the nonlinear effect of the receiver radio frequency link is formed.
Based on real-time test data of the receiver radio frequency link nonlinear effect comprehensive test system, the modeling system provided by the embodiment analyzes the data, performs multiple types of modeling, including a memory-free model and a memory-free model, and supports comparison of a new model with an existing model. The overall modeling scheme realizes a multi-model comprehensive modeling platform, can be used for modeling nonlinear effects of multiple frequency bands between 150MHz and 40GHz and radio frequency links of various types of receivers, and can be used for performance comparison of selecting modeling types, new models and multiple models on an upper computer interface. In order to improve the working efficiency of a modeling system, a model screening method is provided for screening model types in advance, and according to test data of a test system, a model which is not suitable for modeling of the nonlinear effect of a current link does not operate.
The test system of the embodiment integrates various test subsystems into one set of equipment, shares one set of system software platform, system hardware platform, high-speed signal processing module and industrial personal computer, ensures the accuracy and unification of test results of various parameters of the nonlinear effect of the receiver link, and ensures the reliability and unification of input and output data selection in the modeling process. The modeling platform is designed according to the modeling function by dividing the modeling platform into four functional subsystems, as shown in fig. 2.
The hardware system is mainly composed of signal sources, microwave devices, signal processing boards, spectrum analyzers, power meters, industrial personal computers and other main equipment or instruments, and all the desk-top equipment is integrated into a standard cabinet, and meanwhile is convenient to disassemble, assemble, operate and maintain. The instrument equipment adopts a standardized case installation design with a compact structure, the daily operation habit of a test engineer is fully considered in the layout of the instrument equipment, and the whole instrument equipment is easy to use, attractive and elegant. The movable lockable roller is arranged on the cabinet, so that the movable lockable roller is convenient to move, the system is provided with labels at required positions, the labels are clear and correct, and the cabinet is provided with a residual space position for expansion. The signal source mainly generates high-power carrier signals with multiple paths and different frequencies. The power of the carrier signal can be adjusted or set according to the requirement of a user, signals in different modulation modes are generated, and meanwhile, each path of signal is respectively amplified in power and then is input to the combiner. The spectrometer and the power meter mainly analyze corresponding parameters of signals, such as real-time spectrum, spectrum distribution in a certain frequency band, related power values and the like, duplex data transmission is carried out through the Ethernet and the high-speed signal processing module, the system controls related instruments through the exchanger by adopting the Ethernet, and meanwhile, the related instruments transmit data to the signal processing module through the Ethernet.
The high-speed signal processing module adopts a signal processing board based on a Zynq UltraScale+RFSoC ZU28DR main chip (integrated with high-speed ADC and DAC inside IC) to support 8 paths of 12-bit ADC 4.096GSPS and 8 paths of 14-bit DAC 6.4GSPS. The complexity of the RF signal processing chain can be reduced, the input/output channel density maximized without sacrificing wide bandwidth and utilizing heterogeneous processing capabilities, and lower power consumption (eliminating ADC/DAC components, eliminating FPGA to analog interface power consumption). ARM Cortex-A53 processing subsystem, ultraScale+ programmable logic and highest signal processing bandwidth are provided in the Zynq UltraScale+ device, so that a comprehensive RF signal chain can be provided, and the high-performance RF application requirement can be met. The module receives the instructions from the upper computer and issues instructions for controlling the instruments under the test system, and receives the data input into the board card by the instruments. And carrying out data operation according to an upper computer instruction, obtaining a two-dimensional/three-dimensional dual-frequency diagram of a receiver radio frequency link, an input/output power diagram under a fixed frequency point, a 1dB compression point corresponding diagram under different frequencies, nonlinear effect related parameters (1 dB compression point, error vector amplitude and the like), a real-time frequency spectrum and the like, transmitting a processed data result to an industrial personal computer, and displaying and storing at an upper computer interface. Under the modeling system, the module receives the instructions from the upper computer and issues the instructions for controlling each instrument, simultaneously receives the signal source data and the test data of the nonlinear effect of the receiver in the test system, models the corresponding types in the module, and transmits the modeled data to the upper computer for display in real time.
The software system adopts a modularized design, introduces a comprehensive early warning mechanism, performs simple man-machine interaction with a test engineer, and comprehensively eliminates misoperation, so that the operation of the whole modeling system is efficiently controlled, and an upper computer displays corresponding modeling data (a memory-free model and a memory model, and a newly proposed novel model is compared with the existing model) and automatically generates a report.
According to the development trend of remote control in the test measurement field, an Ethernet control mode is adopted for all meters supporting Ethernet control, so that on one hand, the data throughput rate is improved, on the other hand, the system expansion and upgrading in the future are facilitated, and the maintenance is convenient.
The modeling system and the testing system share a hardware system, the upper computer performs function selection, modeling type selection and whether multi-model performance comparison is performed or not, an instruction is issued to the high-speed signal processing module, the high-speed signal processing module controls signal source output, signal source output parameters are uploaded to the high-speed signal processing module, the signal source output is carried out on nonlinear effect testing of a receiver radio frequency link in the testing system, test data are sent to the high-speed signal module, the high-speed signal module carries out identification, model verification and the like of model parameters according to the signal source input data and the test data of the testing system, and the modeling data are sent to the upper computer for display after a modeling task is completed.
Because the modeling period is long, in order to avoid unnecessary modeling, modeling resources and modeling efficiency are wasted, and model screening is performed before modeling is started.
The model types (no-memory model, new model) and the number of the model to be modeled are manually selected at the upper computer interface, and one or more of the memory-free model, the memory model, the new model and the number under a certain large class can be selected. After selection, judging: if the user uploads the self-grinding new model or the modeling algorithm, the user directly outputs the self-grinding new model or the modeling algorithm to a part for determining the model type and the quantity. Whether the user uploads the self-developed new model or the modeling algorithm or not, the signal source data and the test system test data are acquired.
And judging the input signal characteristics and the strong and weak nonlinearity of the link to be tested according to the signal source data and the test system test data, and comprehensively screening according to the input signal characteristics, the strong and weak nonlinearity of the link to be tested, the manually selected simulation model type and the number of the simulation modeling models. The method comprises the steps of automatically determining (recommending) the types and the quantity of models by using a lookup table mode, carrying out multistage classification on the existing models in advance by a system before model screening, carrying out more detailed classification according to the application range (input size signals, suitable narrowband broadband signals, suitable strong and weak nonlinearity of the models, and the like) of each model under the condition of large models, forming multistage classification, and storing the data into a multistage classification catalog. And after the types and the quantity of the models are automatically determined (recommended) by using a lookup table mode, the models are screened together with a new model or a modeling algorithm which is self-developed by a user, and the result of the model screening is sent to an upper computer for display, so that a modeling task is determined. As shown in fig. 6.
The model selection module is as follows:
(1) Memory-free model
The part is used for modeling a memory-free model and mainly comprises the following steps: a memoryless polynomial model, a power series model, an orthogonal polynomial model, and a lookup table model.
(2) Memory model
The part is used for modeling a memory model and mainly comprises the following steps: a memory polynomial model, a Volterra model, a Hammerstein model, a Wiener-Hammerstein model, a PLUME model, a NARMA model, an X-parameter model.
Because of the large number of models, the modeling process is not described in detail, and a digital intermediate frequency receiver (analog front end) and a power series model are selected as examples to introduce the modeling process.
1) Input/output data selection:
according to the characteristics of the power series model, a digital modulation signal is selected as an excitation signal x (t). The excitation signal is input into a front-end link of the receiver, the front-end link of the receiver is output into the test system, and the output data selects spectrum data in a certain bandwidth output by the test system.
2) Determining a behavior model structure:
the power series model is selected as a model structure, and has the advantages of simple form, easy parameter identification, visual description of each order of distortion and the like.
The input-output relationship is as follows:
Wherein a is 1 Is the linear gain of the system; a, a 2 ，a 3 ，a 4 ，…，a n Is the derivative of each order of the model function. The power series model must ensure that the derivatives of each order exist. If a is 2 ，a 3 ，a 4 ，…，α n Equal to 0, the system is a linear system.
The input signal is:
x(t)＝A(t)cos[2πf 0 t+φ(t)]
where the amplitude a (t) is a narrowband baseband signal.
Let the resolution of the signal x (t) be expressed as:
the input signal is:
wherein, the liquid crystal display device comprises a liquid crystal display device,is the complex envelope of x (t).
Using binomial expansion to obtain:
for the first region output we note that only n is odd and that the term 2 k-n= ±1 can contribute. The first zone output of the above equation is:
n is an odd number
The equivalent low pass of the first interval output of y (t) is:
the band pass output signal is:
thus, the envelope transfer characteristic is directly obtained in this case. The power series can also directly derive the harmonic levels in the form of coefficients, which allows the coefficient values that meet the conditions to be calculated in order to control these harmonics.
3) Model parameter identification:
when determining the model structure, the model parameters are identified, and the traditional identification methods include a least square method (Least square method, LS), a recursive identification algorithm, a maximum likelihood identification method, a least mean square error method (LMS algorithm), a recursive least square method (RLS algorithm) and the like, and some new identification algorithms which are emerging in recent years include: neural network-based recognition methods, genetic algorithm-based recognition methods, fuzzy logic-based recognition methods, particle swarm optimization algorithms (PSO algorithms), principal component analysis algorithms (PCA algorithms), and the like. The choice of the recognition algorithm directly affects the accuracy of the final model.
The power series model is represented by a matrix form, and if the sampling point number is M, the matrix form is as follows:
Y＝XP
wherein: p= [ a ] 1 ，a 2 ，…，a N ] T Y＝[y(1)，y(2)，…，y(M)] T
Coefficient a k Can be obtained by matching the measured characteristics using a polynomial of degree N, typically when a matrix X H When X has an inverse matrix, the model coefficients can be identified by using a least square method. The coefficient vector P can be estimated by the equation:
P＝[X H X] -1 X H Y
wherein: h represents the conjugate transpose of the matrix.
A larger value of N is required to better match the measurement characteristics, which reduces the efficiency of the model. In general, a balance point needs to be found between the approximation accuracy and the model efficiency.
4) Model verification:
the behavior model identification process is a process of obtaining the optimal coefficient according to the measured input and output data. Model identification typically uses nonlinear least squares to minimize the model output y mod And actually measured output data y meas Is obtained by mean square error epsilon (Mean square error (MSE)).
And measuring and recording input and output data at regular time intervals, and identifying model coefficients according to a given model structure. The identified model passes a group of inspection input data and records output, and then the error e (n) of the model is:
e(l)＝y meas (l)-y mod (l)
the performance parameters for model comparison are Normalized Mean Square Error (NMSE) and adjacent channel power ratio (Adjacent Channel Power Ratio, ACPR), expressed as follows:
The NMSE reflects the proximity of the model to the physical reality module. The NMSE may be used to determine how close the overall model output value approximates the ideal output value.
Wherein s (f) is a power spectral density function of the signal, [ f ] 1 ，f 2 ]For transmission channel, [ f ] 2 ，f 3 ]Is an adjacent channel. The power spectral density may be calculated by Fourier transform of the autocorrelation function of the signal, or by direct method.
Furthermore, the evaluation parameters representing the accuracy of the model are: error Vector Magnitude (EVM). Error vector magnitude (Error Vector Magnitude, EVM) is defined as the ratio of the root mean square of the average power of the error vector signal to the root mean square of the average power of the reference signal, expressed in percent. If the ideal signal output value is represented by X, e represents the error between the ideal output and the output signal of the integral model, the EVM can be used for measuring the amplitude distortion degree of the integral model to the signal.
If the data calculated by a model is more similar to the actual data, the values of NMSE and ACPR are smaller, indicating that the model is more accurate. But the model with the smallest NMSE is not necessarily the smallest ACPR.
(3) New model
The part provides a typical deep learning part algorithm, a neural network part algorithm for modeling and the introduction of other novel models, an interface is reserved in terms of hardware, and the implantation of the algorithm and the programming of a program can be performed at any time. The novel model based on deep learning and the novel model based on the neural network can provide partial models, and the upper computer interface can be used for selecting operation and comparing the precision with other models; the novel algorithm and model developed by the user provide interfaces, and the high-speed signal processing module is provided with various data interfaces, ARM chips and FPGA chips, so that the algorithm can be conveniently implanted and debugged, and the accuracy comparison can be carried out with the existing model after the debugging is successful.
Due to the numerous correlations of machine learning, such as least squares, support vector machines, singular value decomposition, etc., receiver link nonlinear effect modeling based on machine learning is modeled by taking a modified SVM algorithm, least squares support vector machine (Least Squares Support Vector Machine, LS-SVM), as an example.
The SVM is constructed by taking the principle of minimizing structural risk, and can obtain the optimal result under the condition of limited training sample sets, and has the advantages of simple structure, short training time, low algorithm complexity and strong popularization capability, and can also avoid the phenomena of local minimization and overlearning. LS-SVM adds the sum of squares term of the empirical error of the training set to the objective function and its constraint is transformed into a system of equation equations, which in turn transforms solving the quadratic programming problem (Quadratic Programming, QP) into solving the linear equation problem. On the premise of ensuring high identification accuracy, the LS-SVM simplifies the calculation complexity, thereby shortening the time of model identification and improving the training efficiency.
1) Input/output data selection:
a single tone signal is selected as the excitation signal x (t). The excitation signal is input into a front-end link of the receiver, the front-end link of the receiver is output into the test system, and the output data selects the data y (t) output by the test system. The training sample set for setting the input and output data x (t) and y (t) to be used for SVM regression is as follows:
D＝{(x 1 ，y 1 )，…，(x i ，y i )，…，(x N ，y N )}i＝1，2，…，N
Wherein x is i ＝[x i (1) ，x i (2) ，…，x i (m) ]∈R m For the ith m-dimensional input vector, y i E, R is the corresponding ith output value, and N is the number of training sample pairs.
2) Determining a behavior model structure:
the function of the SVM regression algorithm is to find an optimal prediction function f (x) such that the training input samples x= [ x ] 1 ，x 2 ，…，x N ] T The corresponding training output samples can be approximated as much as possible with f (x), introducing ε:
the fitting function of the model is:
Y＝f(x)＝W T ψ(x)+b
wherein W, b is the weight vector to be identified and the offset; ψ (x) is a nonlinear mapping function.
3) Model parameter identification:
the objective function of the LS-SVM regression algorithm is:
constraint conditions: y is i ＝W T ψ(x i )+b+e i
Wherein, (W, e i ) As a function of the object to be processed,for the predictive output of LS-SVM learning machine, < >>Is a prediction error, also known as a relaxation variable.
Wherein a is i Is Lagrange multiplication factor.
W was biased to zero according to Karush-Kuhn-Tucker (KKT) optimization conditions:
the LS-SVM regression function is available as:
from the above equation, the LS-SVM regression function can be structurally considered as a three-layer neural network: the first layer is an input vector set; the middle layer is composed of kernel functions, and each node corresponds to a support vector; and linearly adding the intermediate layer nodes to obtain a third layer output. As shown in fig. 4.
The kernel function adopts RBF kernel function:
k(x i ，x)＝exp(-g|x i -x| 2 )
Where g is the only parameter of the RBF kernel function.
4) Model verification:
the same as for the memoryless model verification.
(4) Multi-model performance comparison
The method comprises the steps of selecting modeling types and quantity by an upper computer, respectively carrying out multi-type modeling on the same test data, comparing the output data after modeling with actual measurement data, calculating the accuracy of different models, and displaying the accuracy in the same graph in an upper computer interface.
As shown in fig. 5, the test time of the receiver radio frequency link nonlinear effect test system is relatively long according to the test parameters, and the high-speed signal processing module provides a storage function to store test data into the hard disk. After the test is finished, the high-speed signal processing module calls related data to carry out modeling tasks according to modeling requirements, after a certain model type modeling task is finished, the next operation is selected according to the number of modeling types, if multiple types of modeling comparison is carried out, the modeling data of the certain model type are stored in the hard disk, and after all the modeling tasks are finished, the modeling data are displayed on an upper computer interface, so that copying of the modeling data is facilitated.
Aiming at the problem of modeling the nonlinear effect of the radio frequency link of the receiver, the embodiment provides a multi-type modeling platform scheme aiming at the nonlinear effect of the radio frequency link of the receiver. The scheme is suitable for multi-type modeling of nonlinear effects of various types and frequency bands of receivers between 150MHz and 40GHz, and can provide various memory-free models and memory-free models. Compared with the conventional modeling scheme, the provided modeling scheme carries out multi-type modeling according to actual test data, the performance of multiple models is compared according to the same nonlinear effect test data in the same system platform, meanwhile, the researched and developed novel model can be imported into the system to carry out performance comparison with the existing multiple models, and as the modeling period is longer, the model screening method is provided for screening the model in advance, the efficiency of the overall modeling system is improved, the modeling scheme can be developed into a receiver nonlinear effect modeling performance comparison platform instrument, and technical support is provided for receiver development.
Meanwhile, because the single modeling period is longer, in order to avoid unnecessary modeling, modeling resources and modeling efficiency are wasted, and model screening is performed before modeling is started. And (3) comprehensively screening the models by utilizing a lookup table mode according to the signal source data and the test system test data, determining the types and the quantity of the models to be modeled, and not modeling all the models.
Example III
The present embodiment provides a computer readable storage medium having stored thereon a computer program which when executed by a processor performs the steps in the method for model screening for receiver radio frequency link nonlinear effects as described in the above embodiment.
Example IV
The present embodiment provides a computer device, including a memory, a processor, and a computer program stored in the memory and executable on the processor, where the processor implements the steps in the method for model screening of nonlinear effects of a radio frequency link of a receiver according to the above embodiment when the processor executes the program.
It will be appreciated by those skilled in the art that embodiments of the present invention may be provided as a method, system, or computer program product. Accordingly, the present invention may take the form of a hardware embodiment, a software embodiment, or an embodiment combining software and hardware aspects. Furthermore, the present invention may take the form of a computer program product embodied on one or more computer-usable storage media (including, but not limited to, magnetic disk storage, optical storage, and the like) having computer-usable program code embodied therein.
The present invention is described with reference to flowchart illustrations and/or block diagrams of methods, apparatus (systems) and computer program products according to embodiments of the invention. It will be understood that each flow and/or block of the flowchart illustrations and/or block diagrams, and combinations of flows and/or blocks in the flowchart illustrations and/or block diagrams, can be implemented by computer program instructions. These computer program instructions may be provided to a processor of a general purpose computer, special purpose computer, embedded processor, or other programmable data processing apparatus to produce a machine, such that the instructions, which execute via the processor of the computer or other programmable data processing apparatus, create means for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be stored in a computer-readable memory that can direct a computer or other programmable data processing apparatus to function in a particular manner, such that the instructions stored in the computer-readable memory produce an article of manufacture including instruction means which implement the function specified in the flowchart flow or flows and/or block diagram block or blocks.
These computer program instructions may also be loaded onto a computer or other programmable data processing apparatus to cause a series of operational steps to be performed on the computer or other programmable apparatus to produce a computer implemented process such that the instructions which execute on the computer or other programmable apparatus provide steps for implementing the functions specified in the flowchart flow or flows and/or block diagram block or blocks.
Those skilled in the art will appreciate that implementing all or part of the above-described methods in accordance with the embodiments may be accomplished by way of a computer program stored on a computer readable storage medium, which when executed may comprise the steps of the embodiments of the methods described above. The storage medium may be a magnetic disk, an optical disk, a Read-Only Memory (ROM), a Random access Memory (Random AccessMemory, RAM), or the like.
The above description is only of the preferred embodiments of the present invention and is not intended to limit the present invention, but various modifications and variations can be made to the present invention by those skilled in the art. Any modification, equivalent replacement, improvement, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.
Claims (9)
1. A method for model screening of nonlinear effects of a radio frequency link of a receiver, comprising:
acquiring a signal source and test output data of a receiver;
the signal source and the test output data comprise an excitation signal and a verification signal, wherein the excitation signal is used for identifying parameters of the pre-test assumed model, and the verification signal is used for verifying output results of the pre-test assumed model after parameter identification;
selecting a pre-test assumption model by combining the existing prior knowledge with a receiver mechanism; the pre-test hypothesis model adopts a nonlinear model, and the nonlinear model comprises: a memory-free model and a memory-equipped model;
identifying model parameters of the pre-test assumed model to obtain a pre-test assumed model after parameter identification;
specifically, after a pre-test assumed model structure is determined, model parameters are identified, and the main purpose of parameter identification is to analyze a characterization function equation of a nonlinear system according to a system input/output state variable, find a determined time domain characteristic function model with the minimum matching error, and find parameters of the function model;
based on the signal source and the test output data, comparing the obtained result with a real system result by adopting a pre-verification assumed model after parameter identification, judging whether the error is smaller than a set threshold value, if so, obtaining a modeled model, otherwise, reselecting the pre-verification assumed model to carry out the parameter identification and verification process;
Specifically, after modeling is completed, the generalization capability of the obtained model is checked, and under the condition of the same input, the error between the output of the built model and the output of an actual system is checked by using a data set which does not participate in recognition to judge whether the required precision is reached;
judging the input signal characteristics and the strong and weak nonlinearity of the link to be tested according to the signal source data and the test system test data, and comprehensively screening according to the input signal characteristics, the strong and weak nonlinearity of the link to be tested, the manually selected simulation model type and the number of the simulation modeling models; the method comprises the steps of automatically determining the types and the quantity of models in a lookup table mode, carrying out multistage classification on the existing models in advance by a system before model screening, carrying out more detailed classification according to the application range of each model, namely, inputting a size signal, being suitable for a narrow-band broadband signal and being suitable for strong and weak nonlinearity of the model, forming multistage classification, and storing the data into a multistage classification catalog; after the types and the quantity of the models are automatically determined by using a lookup table mode, the model screening is completed together with a new model or a modeling algorithm which is automatically researched by a user, and the result of the model screening is sent to an upper computer for display, so that a modeling task is determined.
2. The method of model screening for receiver radio frequency link nonlinear effects of claim 1, wherein testing the output data comprises: the system comprises a real-time frequency spectrum, a two-dimensional/three-dimensional dual-frequency diagram, an original signal waveform, an input/output power diagram under a fixed frequency point of a power compression test subsystem, a 1dB compression diagram under different frequencies, a constellation diagram of a power compression three-dimensional diagram, a constellation diagram of a amplitude-phase error test subsystem, a link gain, a noise coefficient and sensitivity of a link parameter test subsystem, a third-order cut-off point, an adjacent channel power ratio and a noise power ratio of a theoretical parameter calculation subsystem of the intermodulation test subsystem.
3. The method for model screening of nonlinear effects of a receiver radio frequency link according to claim 1, wherein said memoryless model comprises: a memoryless polynomial model, a power series model, an orthogonal polynomial model, and a lookup table model.
4. The method for model screening of nonlinear effects of a receiver radio frequency link according to claim 1, wherein said memoryless model comprises: the memory model comprises: a memory polynomial model, a Volterra model, a Hammerstein model, a Wiener-Hammerstein model, a PLUME model, a NARMA model, an X-parameter model.
5. The method for model screening of nonlinear effects of a receiver radio frequency link according to claim 1, wherein said identifying model parameters of a pre-test hypothesis model comprises: and identifying model parameters of the pre-test assumed model by adopting a least square method, a recursive identification algorithm, a maximum likelihood identification method and a Fourier series method.
6. The method for model screening of nonlinear effects of a receiver radio frequency link according to claim 1, wherein the index for comparing the obtained result with the actual system result comprises: normalized mean square error, root mean square error, mean absolute error.
7. A model screening platform for nonlinear effects of a receiver radio frequency link, implemented by the model screening method for nonlinear effects of a receiver radio frequency link according to claim 1, comprising:
a data acquisition module configured to: acquiring a signal source and test output data of a receiver;
a model selection module configured to: selecting a pre-test assumption model by combining the existing prior knowledge with a receiver mechanism;
a parameter identification module configured to: identifying model parameters of the pre-test assumed model to obtain a pre-test assumed model after parameter identification;
A verification model configured to: based on the signal source and the test output data, comparing the obtained result with the real system result by adopting a pre-verification assumed model after parameter identification, judging whether the error is smaller than a set threshold value, if so, obtaining a modeled model, otherwise, reselecting the pre-verification assumed model to carry out the parameter identification and verification process.
8. A computer readable storage medium having stored thereon a computer program, which when executed by a processor performs the steps in a method for model screening of nonlinear effects of a radio frequency link of a receiver as claimed in any one of claims 1-6.
9. A computer device comprising a memory, a processor and a computer program stored on the memory and executable on the processor, characterized in that the processor implements the steps in the method for model screening of the nonlinear effects of a radio frequency link of a receiver as claimed in any one of claims 1-6 when said program is executed by said processor.
Priority Applications (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110908338.7A CN113591390B (en) | 2021-08-09 | 2021-08-09 | Model screening method and platform for nonlinear effect of receiver radio frequency link |
Applications Claiming Priority (1)
| Application Number | Priority Date | Filing Date | Title |
|---|---|---|---|
| CN202110908338.7A CN113591390B (en) | 2021-08-09 | 2021-08-09 | Model screening method and platform for nonlinear effect of receiver radio frequency link |
Publications (2)
| Publication Number | Publication Date |
|---|---|
| CN113591390A CN113591390A (en) | 2021-11-02 |
| CN113591390B true CN113591390B (en) | 2023-10-13 |
Family
ID=78256392
Family Applications (1)
| Application Number | Title | Priority Date | Filing Date |
|---|---|---|---|
| CN202110908338.7A Active CN113591390B (en) | 2021-08-09 | 2021-08-09 | Model screening method and platform for nonlinear effect of receiver radio frequency link |
Country Status (1)
| Country | Link |
|---|---|
| CN (1) | CN113591390B (en) |
Families Citing this family (2)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN114520750A (en) * | 2022-02-24 | 2022-05-20 | 哈尔滨工业大学 | Non-invasive unmanned equipment state monitoring and intelligent processing device |
| CN115242336B (en) * | 2022-08-01 | 2023-02-03 | 北京航空航天大学 | Nonlinear output frequency spectrum prediction method for receiver radio frequency front-end device |
Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101350597A (en) * | 2008-09-10 | 2009-01-21 | 北京北方烽火科技有限公司 | Method for modeling wideband radio-frequency power amplifier |
| CN106126804A (en) * | 2016-06-21 | 2016-11-16 | 上海无线电设备研究所 | The behavioral scaling modeling of a kind of power amplifier bottom circuit and verification method |
| CN106971077A (en) * | 2017-03-30 | 2017-07-21 | 中国人民解放军国防科学技术大学 | A kind of Dynamic Simulation Model verification method based on time slice parameter identification |
| CN110457854A (en) * | 2019-08-21 | 2019-11-15 | 青岛大学 | The Volterra digit pre-distortion parameters discrimination method of supported collection adjusted mean approximation message transmission |
-
2021
- 2021-08-09 CN CN202110908338.7A patent/CN113591390B/en active Active
Patent Citations (4)
| Publication number | Priority date | Publication date | Assignee | Title |
|---|---|---|---|---|
| CN101350597A (en) * | 2008-09-10 | 2009-01-21 | 北京北方烽火科技有限公司 | Method for modeling wideband radio-frequency power amplifier |
| CN106126804A (en) * | 2016-06-21 | 2016-11-16 | 上海无线电设备研究所 | The behavioral scaling modeling of a kind of power amplifier bottom circuit and verification method |
| CN106971077A (en) * | 2017-03-30 | 2017-07-21 | 中国人民解放军国防科学技术大学 | A kind of Dynamic Simulation Model verification method based on time slice parameter identification |
| CN110457854A (en) * | 2019-08-21 | 2019-11-15 | 青岛大学 | The Volterra digit pre-distortion parameters discrimination method of supported collection adjusted mean approximation message transmission |
Also Published As
| Publication number | Publication date |
|---|---|
| CN113591390A (en) | 2021-11-02 |
Similar Documents
| Publication | Publication Date | Title |
|---|---|---|
| Tehrani et al. | A comparative analysis of the complexity/accuracy tradeoff in power amplifier behavioral models | |
| CN113591390B (en) | Model screening method and platform for nonlinear effect of receiver radio frequency link | |
| Ma et al. | Partially‐coupled gradient‐based iterative algorithms for multivariable output‐error‐like systems with autoregressive moving average noises | |
| Rayas-Sanchez et al. | EM-based Monte Carlo analysis and yield prediction of microwave circuits using linear-input neural-output space mapping | |
| Liu et al. | Dynamic behavioral modeling of 3G power amplifiers using real-valued time-delay neural networks | |
| Barmuta et al. | Compact behavioral models of nonlinear active devices using response surface methodology | |
| CN108702136A (en) | Digital compensator | |
| CN105787219B (en) | A method of establishing Conduction Interference coupling channel multiple linear regression model using frequency point sampling is closed on | |
| Fellows et al. | Direct algorithm for the Pareto load‐pull optimisation of power‐added efficiency and adjacent‐channel power ratio | |
| CN114004162A (en) | Modeling method for smelting load harmonic emission level under multi-working-condition scene | |
| Louro et al. | New transistor behavioral model formulation suitable for Doherty PA design | |
| Mengozzi et al. | Joint dual-input digital predistortion of supply-modulated RF PA by surrogate-based multi-objective optimization | |
| CN101206727B (en) | Data processing apparatus, data processing method | |
| Huang et al. | Microscale searching algorithm for coupling matrix optimization of automated microwave filter tuning | |
| Enayati et al. | Real‐time harmonic estimation using a novel hybrid technique for embedded system implementation | |
| CN111880003B (en) | Method for extracting harmonic balance parameters of radio frequency power amplifier based on neural network | |
| Verbeyst et al. | The Volterra input-output map of a high-frequency amplifier as a practical alternative to load-pull measurements | |
| De Jonghe et al. | Efficient analytical macromodeling of large analog circuits by transfer function trajectories | |
| LU100964B1 (en) | A DNN-based method for modeling temperature characteristics of radio frequency power amplifier | |
| Giannini et al. | Neural networks and volterra series for time‐domain power amplifier behavioral models | |
| CN106817328A (en) | Digital pre-distortion table generating method, device and digital pre-distortion system | |
| De Tommasi et al. | Surrogate modeling of RF circuit blocks | |
| Su et al. | A Fast Two‐Tone Active Load‐Pull Algorithm for Assessing the Non‐linearity of RF Devices | |
| Liu et al. | Large-signal behavioral modeling of nonlinear amplifiers based on load-pull AM-AM and AM-PM measurements | |
| Stegmayer et al. | Volterra NN-based behavioral model for new wireless communications devices |
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 |