Chip RF calibration method and system based on artificial intelligence model, electronic equipment and storage medium

Abstract

The invention discloses a chip RF calibration method and system based on an artificial intelligence model, electronic equipment and a storage medium. The chip RF calibration method comprises the steps that S1, a mathematical model is constructed; constructing a chip RF calibration artificial intelligence AI mathematical model; the general-purpose tester is remotely controlled, the general-purpose tester extracts constellation diagram data of different DUTs in a non-air interference state by using VSA, and feature extraction is carried out through AI; s2, a model training step; a deep learning model with the minimum error is obtained through model training; taking the obtained deep learning model as a mathematical model for predicting RF correction parameters; step S3, an RF parameter prediction step; and predicting the RF parameters of the chip by using the constructed mathematical model. According to the chip RF calibration method and system based on the artificial intelligence model, the electronic equipment and the storage medium, RF calibration parameters of various chips can be accurately predicted in real time.

Description

Technical Field

[0001] The present invention belongs to the technical field of chip testing, and relates to an RF calibration method, in particular to a chip RF calibration method, system, electronic device and storage medium based on an artificial intelligence model. Background Art

[0002] An important quality check before mass production after chip manufacturing is to conduct complete tests internally to simulate all possible future scenarios. For example, a comprehensive test instrument (Radio Communication Integrated Tester) conducts cross-band 2G, 3G, 4G, 5G, WLAN (Wireless Local Area Network), and Bluetooth tests on multiple DUTs (Devices Under Test) and multiple antennas to confirm that the manufacturing function errors of the DUTs are within a reasonable range.

[0003] The comprehensive test instrument is connected to the DUT via lines such as a VSA (Vector Signal Analyzer) and a VSG (Vector Signal Generator) to measure the signal characteristics of the DUT. These characteristics are mainly used to analyze the gaps between various components or signals and the chip specifications, and can observe error signals.

[0004] However, the quality of different chips at the factory is not the same. Even if they pass the test items of the comprehensive test instrument, it is necessary to calibrate the RF (Radio Frequency) to make the chip signals more stable. However, the RF signal calibration is usually only carried out once in the chip test factory for all chips, and the calibration values are not adjusted for each individual chip.

[0005] Traditional RF parameter calibration relies on the test factory to calibrate the RF signals of all chips in the same batch at one time, and the parameters cannot be adjusted at the user end after leaving the factory. In the existing methods for signal-related analysis, the method of using the comprehensive test instrument to collect clean data is not utilized, and relevant AI applications cannot be carried out.

[0006] In view of this, there is an urgent need to design a new RF calibration method to overcome at least some of the above-mentioned defects of the existing RF calibration methods. Summary of the Invention

[0007] The present invention provides a chip RF calibration method, system, electronic device and storage medium based on an artificial intelligence model, which can accurately and real-time predict the RF calibration parameters of various chips.

[0008] To solve the above technical problems, according to one aspect of the present invention, the following technical solutions are adopted:

[0009] A chip RF calibration method based on an artificial intelligence model, the chip RF calibration method comprising:

[0010] Step S1, a mathematical model construction step; constructing an artificial intelligence (AI) mathematical model for chip RF calibration; remotely controlling a comprehensive tester, and the comprehensive tester uses a VSA to extract constellation diagram data of different DUTs in a state without air interference, and performs feature extraction through AI;

[0011] Step S2, a model training step; obtaining a deep learning model with the minimum error through model training; using the obtained deep learning model as a mathematical model for predicting RF correction parameters;

[0012] Step S3, an RF parameter prediction step; predicting chip RF parameters using the constructed mathematical model.

[0013] As an implementation manner of the present invention, in step S2, chip data is input into the chip RF calibration AI mathematical model, and the real and imaginary numbers of the constellation diagram are used as input features of the learning model. The target chip RF parameters are trained through the deep learning model, features are extracted through the neurons in the hidden layer of the deep learning model, and a set of appropriate weights is obtained through training; finally, in step S3, different chip RF parameters can be predicted through the trained model.

[0014] As an implementation manner of the present invention, the mathematical model construction step includes:

[0015] Step S11: Using a comprehensive tester to collect constellation diagrams without air interference and various RF signal modulation scenarios;

[0016] Step S12: Inferring ideal RF correction parameters using the collected constellation diagrams;

[0017] Step S13: Using a deep learning AI model to fit the ideal RF correction parameters, and constructing a deep learning mathematical model C(x) according to the neuron weights and structure, where x is the signal of the receiving or transmitting element;

[0018] Step S14: The user uses the signal S(t) of the receiving or transmitting element, where t is time, and through combination with the deep learning mathematical model C(x), it is used as the input data of the deep learning mathematical model C(x), and C(s(t)) predicts RF parameters such as radio frequency automatic gain control (AGC). The user can use this deep learning mathematical model C(x) at any time to predict the ideal RF correction parameters, so as to compensate for the signal power affected by different environments and reduce the constellation diagram error situation.

[0019] As an implementation manner of the present invention, the steps of constructing the mathematical model include: developing an AI feature integration and extraction module in the data stream, accessing the values of the comprehensive test instrument, and storing the set data in a variable data structure;

[0020] The steps of constructing the mathematical model further include: remotely controlling the comprehensive test instrument through the UI interface; controlling the comprehensive test instrument through the developed UI interface, integrating the chip data into a large AI model, and summarizing and training the features;

[0021] The steps of model training include: developing an AI training module in model construction. First, use the AI feature integration and extraction module to extract data from the variable data structure as the input data of the model; then, fit the ideal RF parameter data through a deep learning model, and finally generate accurate prediction results;

[0022] In the step of predicting RF parameters, predict the ideal RF parameters, and directly generate a model local file to allow users to perform model prediction at different times and locations.

[0023] According to another aspect of the present invention, the following technical solution is adopted: a chip RF calibration system based on an artificial intelligence model, the chip RF calibration system includes:

[0024] A mathematical model construction module for constructing a chip RF calibration artificial intelligence (AI) mathematical model; a remote control comprehensive test instrument, which uses a VSA to extract constellation diagram data of different DUTs in a state without air interference, and performs feature extraction through AI;

[0025] A model training module for obtaining a deep learning model with the minimum error through model training, and using the obtained deep learning model as a mathematical model for predicting RF correction parameters;

[0026] An RF parameter prediction module for predicting chip RF parameters using the constructed mathematical model.

[0027] As an implementation manner of the present invention, the model training module inputs chip data into the chip RF calibration AI mathematical model, uses the real and imaginary numbers of the constellation diagram as input features of the learning model, performs AI model training on the target chip RF parameters through a deep learning model, extracts features through the hidden layer neurons of the deep learning model, and obtains a set of appropriate weights through training; finally, the RF parameter prediction module can predict different chip RF parameters through the trained model.

[0028] As an implementation manner of the present invention, the manner in which the mathematical model construction module constructs the chip RF calibration artificial intelligence (AI) mathematical model includes:

[0029] Step S11: Use a comprehensive tester to collect constellation diagrams without air interference and various RF signal modulation scenarios;

[0030] Step S12: Deduce ideal RF calibration parameters using the collected constellation diagrams;

[0031] Step S13: Use a deep learning AI model to fit the ideal RF calibration parameters, and construct a deep learning mathematical model C(x) based on the neuron weights and structure, where x is the signal of the receiving or transmitting element;

[0032] Step S14: The user uses the signal S(t) of the receiving or transmitting element, where t is time, and through combination with the deep learning mathematical model C(x), as the input data of the deep learning mathematical model C(x). C(s(t)) predicts RF parameters such as radio frequency automatic gain control (AGC). The user can use this deep learning mathematical model C(x) at any time to predict the ideal RF calibration parameters, so as to compensate for the signal power affected by different environments and reduce the constellation diagram error situation.

[0033] As an implementation manner of the present invention, the mathematical model construction module includes: an AI feature integration and extraction module, which is used to integrate and extract AI features in the data stream, access the values of the comprehensive test instrument, and finally store them in the variable data structure;

[0034] The mathematical model construction module remotely controls the comprehensive tester through the UI interface, integrates the chip data into the large AI model, and summarizes and trains the features;

[0035] The model training module is used to train the model, take the data out of the variable data structure as the input data of the model; and fit the ideal RF parameter data through the deep learning model;

[0036] The RF parameter prediction module is used to predict the ideal RF parameters, and directly generate a model local file for the user to perform model prediction at different times and locations.

[0037] According to another aspect of the present invention, the following technical solution is adopted: An electronic device includes a memory, a processor, and a computer program stored on the memory and executable on the processor. When the processor executes the computer program, the steps of the above method are implemented.

[0038] According to another aspect of the present invention, the following technical solution is adopted: A storage medium stores computer program instructions, and when the computer program instructions are executed by a processor, the steps of the above method are implemented.

[0039] The beneficial effects of the present invention are as follows: The chip RF calibration method, system, electronic device and storage medium based on the artificial intelligence model proposed by the present invention can accurately and real-time predict the RF correction parameters of various chips.

[0040] This system will use a comprehensive test instrument to collect a large number of wireless communication signal information, and can extract the cleanest signal from the DUT transmitter end in the VSA, without the influence of air interference and other factors affecting the signal. Capture its IQ Constellation and develop and integrate an AI model to automatically perform RF calibration for each individual chip, so that after the user gets the chip, they can calibrate the RF signal at any time and place.

[0041] The present invention can use a comprehensive test instrument to summarize various chips, extract the advantages of various wireless communication signals, collect clean and complete transmitter IQ Constellation data, and use it as the input data of the AI model; the present invention can use the developed AI model to widely collect the ideal RF performance of multiple chips, and accurately predict the RF correction parameters (RF Calibration Parameters) of various chips; in addition, the present invention can use the developed AI model to allow users to calibrate RF parameters in real time at different times and places. Brief Description of the Drawings

[0042] Figure 1 It is a flowchart of the chip RF calibration method based on the artificial intelligence model in an embodiment of the present invention.

[0043] Figure 2 It is a schematic diagram of the composition of the chip RF calibration system based on the artificial intelligence model in an embodiment of the present invention.

[0044] Figure 3 It is a control flowchart of the AI model and the comprehensive test instrument in an embodiment of the present invention.

[0045] Figure 4 It is a control flowchart of the UI interface in an embodiment of the present invention.

[0046] Figure 5 It is a schematic diagram of the working principle of the AI feature integration and extraction module in an embodiment of the present invention.

[0047] Figure 6 It is a schematic diagram of the working principle of the AI training module in an embodiment of the present invention.

[0048] Figure 7 It is a flowchart of the user's real-time inference of the AI model in an embodiment of the present invention.

[0049] Figure 8 It is a schematic diagram of the composition of the electronic device in an embodiment of the present invention. Detailed implementation manners

[0050] The preferred embodiments of the present invention will be described in detail below with reference to the accompanying drawings.

[0051] To further understand the present invention, the preferred implementation manners of the present invention will be described below in conjunction with embodiments. However, it should be understood that these descriptions are only for further explaining the features and advantages of the present invention, rather than limiting the claims of the present invention.

[0052] The description of this part only focuses on several typical embodiments, and the present invention is not limited to the scope described in the embodiments. The mutual replacement of the same or similar prior art means and some technical features in the embodiments is also within the scope of the description and protection of the present invention.

[0053] The expressions of the steps in each embodiment in the specification are only for convenience of description, and the implementation manner of the present application is not limited by the order of step implementation.

[0054] "Coupling" or "connection" in the specification includes both direct connection and indirect connection.

[0055] The present invention discloses a chip RF calibration method based on an artificial intelligence model. Figure 1 is a flowchart of the chip RF calibration method based on the artificial intelligence model in an embodiment of the present invention; please refer to Figure 1 , the chip RF calibration method includes:

[0056]

Step S1

[0057] The receiver adjusts the transmission power according to the estimated signal-to-noise ratio, and the transmitter adjusts the appropriate power according to the environment. Whether it is a receiver or a transmitter, signals are extracted or transmitted through the constellation diagram. However, different environments, such as different signal powers, may affect the constellation diagram, causing the bits carried by the RF signal to deviate from their original positions. Therefore, RF correction parameters, such as radio frequency automatic gain control AGC (Automatic gain control), etc., are required to calibrate the original constellation diagram. The present invention predicts the RF correction parameters through an AI model, enabling users to perform RF signal calibration on the chip after leaving the factory.

[0058] The refined RF calibration AI model is as follows: The constellation diagram can infer the power curve in the time domain and predict the received signal strength RSSI (Received signal strength indicator) of the receiver; using the constellation diagram to infer many RF parameters, such as the radio frequency automatic gain control AGC (Automatic gain control), etc., so that the constellation diagram under different signal strengths can still maintain a certain level and will not cause bit error phenomena of the modulated signal carrying bits due to the signal strength.

[0059] In an embodiment of the present invention, the mathematical model construction step includes:

[0060] Step S11: Use a comprehensive tester to collect the transmitter IQ Constellation of the constellation diagram without air interference and various RF signal modulation scenarios;

[0061] Step S12: Infer ideal RF calibration parameters, such as signal power, using the collected transmitter IQ Constellation of the constellation diagram;

[0062] Step S13: Use a deep learning AI model to fit the ideal RF calibration parameters, such as signal power, and construct a deep learning mathematical model C(x) according to the neuron weights and structure, where x is the signal of the receiving or transmitting element;

[0063] Step S14: The user uses the signal S(t) of the receiving or transmitting element, where t is time, and through combination with the deep learning mathematical model C(x), it is used as the input data of the deep learning mathematical model C(x), and C(s(t)) predicts RF parameters, such as the radio frequency automatic gain control AGC (Automatic gain control). The user can use this deep learning mathematical model C(x) at any time to predict the ideal RF calibration parameters, such as signal power, so as to compensate for the signal power affected by different environments and reduce the constellation diagram bit error situation.

[0064] In an embodiment of the present invention, the step of constructing the chip RF calibration artificial intelligence AI mathematical model includes: developing an AI feature integration and extraction module in the data stream, accessing the values of the comprehensive test instrument, and storing the set data into the variable data structure;

[0065] In addition, the mathematical model step includes: remotely controlling the comprehensive tester using the UI interface; the step of constructing the chip RF calibration artificial intelligence AI mathematical model controls the comprehensive test instrument through the developed UI interface, integrates the chip data into the large AI model, and summarizes the features.

[0066]

Step S2

[0067] In an embodiment of the present invention, chip data is input into the chip RF calibration AI mathematical model, and the real and imaginary numbers of the constellation diagram are used as input features of the learning model. The AI model training for the target chip RF parameters is carried out through the deep learning model. The features are extracted by the hidden layer neurons of the deep learning model, and a set of appropriate weights is obtained through training. Finally, in step S3, different chip RF parameters can be predicted through the trained model.

[0068] In the model training step, an AI training module is developed during model construction. First, the AI feature integration and extraction module extracts data from the variable data structure as the input data of the model. Then, the ideal RF parameter data is fitted through the deep learning model, and finally, an accurate prediction result is generated.

[0069]

Step S3

[0070] In the RF parameter prediction step, chip data is input into the chip RF calibration AI mathematical model, and the real and imaginary numbers of the constellation diagram are used as input features of the learning model. The AI model training for the target chip RF parameters is carried out through the deep learning model. The features are extracted by the hidden layer neurons of the deep learning model, and a set of appropriate weights is obtained through training. Finally, different chip RF parameters can be predicted through this model.

[0071] In the RF parameter prediction step, ideal RF parameters are predicted, and the model local file is generated directly to allow users to perform model prediction at different times and locations.

[0072] The present invention also discloses a chip RF calibration system based on an artificial intelligence model. The chip RF calibration system includes: a mathematical model construction module 1, a model training module 2, and an RF parameter prediction module 3.

[0073] The mathematical model construction module 1 is used to construct a chip RF calibration artificial intelligence (AI) mathematical model; a remote control integrated tester. The integrated tester uses VSA to extract IQ Constellation data of different DUTs in a state without air interference, and performs feature extraction through AI.

[0074] The model training module 2 is used to obtain a deep learning model with the minimum error through model training; the obtained deep learning model is used as a mathematical model for predicting RF correction parameters. In an embodiment of the present invention, the RF parameter prediction module 2 is used to input chip data into the chip RF calibration AI mathematical model, use the real and imaginary numbers of the constellation diagram as the input features of the learning model, perform AI model training on the target chip RF parameters through the deep learning model, extract features through the neurons in the hidden layer of the deep learning model, and obtain a set of appropriate weights through training; finally, the RF parameter prediction module 3 can predict different chip RF parameters through this model.

[0075] The RF parameter prediction module 3 is used to predict the chip RF parameters by using the constructed mathematical model.

[0076] In an embodiment of the present invention, the mathematical model construction module 1 includes an AI feature integration and extraction module. The AI feature integration and extraction module is used to integrate and extract AI features in the data stream, access the values of the comprehensive test instrument, and finally store them in the variable data structure. The mathematical model construction module 1 can remotely control the comprehensive test instrument through the UI interface, integrate the chip data into the large AI model, and summarize the features.

[0077] The model training module 2 is used to train the model, take the data out of the variable data structure as the input data of the model; fit the ideal RF parameter data through the deep learning model, and finally generate accurate prediction results.

[0078] The RF parameter prediction module 3 is used to predict the ideal RF parameters, and directly generate a model local file to allow users to perform model predictions at different times and locations.

[0079] In a usage scenario of the present invention, the system of the present invention uses the python PyQt suite to build the UI interface, and uses TCP / IP to connect the python control interface to the comprehensive test instrument. This system uses an Ethernet cable as the physical connection method, uses the python VISA suite to connect to the IP location of the comprehensive test instrument, and transmits SCPI (Standard Commands for Programmable Instruments) to control the instrument.

[0080] In the AI feature integration and extraction module (Feature Extraction Database Module), the IQ Constellation of the comprehensive test instrument is accessed by using the python VISA query method, and the values are recorded in the data structure of the software.

[0081] In the AI Training Module, Python deep learning toolkits such as Keras and TensorFlow are used to develop an AI model, fit the ideal RF parameters, and store the established model in the local folder. Users can import the model file to predict the RF calibration parameters.

[0082] Figure 3 This is the control flow chart of the AI model and the comprehensive test instrument. In this system, the comprehensive test instrument is controlled at the PC (personal computer) end. An Ethernet cable is used to connect the instrument, and SPCI control instructions are transmitted via TCP / IP over the IEEE 488.2 (GPIB) standard protocol. The comprehensive test instrument uses VSA and VSG to connect to the DUT, and the PC end controls the DUT via UART.

[0083] Figure 4 This is the control flow chart of the UI interface. First, PyQt is used as the development toolkit for the UI interface, and the open-source PyVISA is used to remotely control the comprehensive test instrument. When a button on the UI interface is pressed, instrument connection starts, and the instrument can also be accessed or controlled through subsequent modules such as the AI Feature Integration and Extraction Module.

[0084] Figure 5 This is a schematic diagram of the working principle of the Feature Extraction Database Module. The measurement data of the comprehensive test instrument is accessed through the UI interface and stored in a Python data structure, providing the basis for the input of the entire AI model data.

[0085] Figure 6 This is a schematic diagram of the working principle of the AI Training Module. The previously mentioned Feature Extraction Database Module and the ideal RF calibration parameter prediction target are used as the input of the AI model training module. Python Keras and TensorFlow are used as the architecture of the deep learning classification model to enable the AI model to fit the ideal RF parameters. This allows users to use the calibration data predicted by the model to adjust the RF calibration parameters of the chip even in an environment with severe interference, improving the overall signal reception quality, just like observing the signal on the comprehensive test instrument without extra environmental interference.

[0086] Figure 7For the user, it is a real-time inference AI model flowchart. The user can apply the AI models stored under the AI Training Module at any time and in any space, and update the predicted RF parameters in the chip in real time, so as to improve the chip signal reception performance.

[0087] The present invention also discloses an electronic device. Figure 8 It is a schematic diagram of the composition of the electronic device in an embodiment of the present invention; please refer to Figure 8 , at the hardware level, the electronic device includes a memory, a processor, and at least one network interface; the processor can be a microprocessor, and the memory can include a memory, such as a random access memory (Random Access Memory, RAM), and can also include a non-volatile memory, etc. Of course, the electronic device can also be provided with other hardware according to needs.

[0088] The processor, network interface, and memory can be interconnected through an internal bus, and the internal bus can be an ISA (Industry Standard Architecture) bus, a PCI (Peripheral Component Interconnect Standard) bus, or an EISA (Extended Industry Standard Architecture) bus, etc.; the bus can include an address bus, a data bus, a control bus, etc. The memory is used to store programs (which can include an operating system program and application programs); the programs can include program codes, and the program codes can include computer operation instructions. The memory can include a memory and a non-volatile memory, and provide instructions and data to the processor.

[0089] In one embodiment, the processor can read the corresponding program from the non-volatile memory into the memory and then run it; the processor can execute the programs stored in the memory, and is specifically used to perform the following operations (as Figure 1 shown):

[0090]

Step S1

[0091]

Step S2

[0092]

Step S3

[0093] The present invention further discloses a storage medium, on which computer program instructions are stored. When the computer program instructions are executed by a processor, the following steps of the method of the present invention are implemented (as Figure 1 shown):

[0094]

Step S1

[0095]

Step S2

[0096]

Step S3

[0097] In summary, the chip RF calibration method, system, electronic device and storage medium based on the artificial intelligence model proposed by the present invention can accurately and real-time predict the RF correction parameters of various chips.

[0098] The technical features of the above-described embodiments can be combined arbitrarily. For the sake of brevity of description, not all possible combinations of the technical features in the above embodiments are described. However, as long as there is no contradiction in the combination of these technical features, it should be considered as the scope described in this specification.

[0099] The description and application of the present invention here are illustrative and do not intend to limit the scope of the present invention to the above embodiments. The effects or advantages involved in the embodiments may not be reflected in the embodiments due to various factors. The description of the effects or advantages is not used to limit the embodiments. The deformations and changes of the embodiments disclosed here are possible, and the substitutions and equivalent components of the embodiments are well known to those of ordinary skill in the art. Those skilled in the art should clearly understand that the present invention can be implemented in other forms, structures, arrangements, proportions, and with other components, materials and parts without departing from the spirit or essential characteristics of the present invention. Other deformations and changes can be made to the embodiments disclosed here without departing from the scope and spirit of the present invention.

Claims

1. A chip RF calibration method based on an artificial intelligence model, characterized in that, The chip RF calibration method includes: Step S1, mathematical model construction step: Construct an artificial intelligence (AI) mathematical model for chip RF calibration. Remotely control a comprehensive tester, and the comprehensive tester uses VSA to extract constellation diagram data of different DUTs in a state without air interference, and performs feature extraction through AI. Step S2, model training step: Obtain a deep learning model with the minimum error through model training. Use the obtained deep learning model as the mathematical model for predicting RF correction parameters. Step S3, RF parameter prediction step: Use the constructed mathematical model to predict the chip RF parameters.

2. The chip RF calibration method based on an artificial intelligence model according to claim 1, wherein: In step S2, input chip data into the chip RF calibration AI mathematical model, use the real and imaginary numbers of the constellation diagram as the input features of the learning model, perform AI model training on the target chip RF parameters through the deep learning model, extract features through the neurons in the hidden layer of the deep learning model, and obtain a set of appropriate weights through training. Finally, in step S3, different chip RF parameters can be predicted through the trained model.

3. The chip RF calibration method based on an artificial intelligence model according to claim 1, wherein: The mathematical model construction step includes: Step S11: Use a comprehensive tester to collect constellation diagrams without air interference and various RF signal modulation scenarios. Step S12: Infer ideal RF correction parameters using the collected constellation diagrams. Step S13: Use a deep learning AI model to fit the ideal RF correction parameters, and construct a deep learning mathematical model C(x) according to the neuron weights and structure, where x is the signal of the receiving or transmitting element. Step S14: The user uses the signal S(t) of the receiving or transmitting element, where t is time, combines it with the deep learning mathematical model C(x), and uses it as the input data of the deep learning mathematical model C(x). C(s(t)) predicts RF parameters such as radio frequency automatic gain control (AGC). The user can use this deep learning mathematical model C(x) at any time to predict the ideal RF correction parameters, thereby compensating for the signal power affected by different environments and reducing the constellation diagram error rate.

4. The chip RF calibration method based on an artificial intelligence model according to claim 1, wherein: The mathematical model construction step includes: Develop an AI feature integration and extraction module in the data stream, access the values of the comprehensive test instrument, and store the set data in a variable data structure. The mathematical model construction step further includes: Remotely control the comprehensive tester using the UI interface; control the comprehensive test instrument through the developed UI interface, integrate the chip data into a large AI model, and summarize and train the features. The model training step includes: Develop an AI training module during model construction. First, use the AI feature integration and extraction module to extract data from the variable data structure and use it as the input data of the model. Then, use the deep learning model to fit the ideal RF parameter data, and finally generate accurate prediction results. In the RF parameter prediction step, ideal RF parameters are predicted, and by generating a local model file, users can directly perform model prediction at different times and locations.

5. A chip RF calibration system based on an artificial intelligence model, characterized in that, The chip RF calibration system includes: A mathematical model construction module for constructing an artificial intelligence (AI) mathematical model for chip RF calibration; a remote control integrated tester that uses a VSA to extract constellation diagram data of different devices under test (DUTs) without air interference and performs feature extraction through AI. A model training module for obtaining a deep learning model with the minimum error through model training and using the obtained deep learning model as the mathematical model for predicting RF correction parameters. An RF parameter prediction module for predicting chip RF parameters using the constructed mathematical model.

6. The chip RF calibration system based on an artificial intelligence model according to claim 5, wherein: The model training module inputs chip data into the chip RF calibration AI mathematical model, uses the real and imaginary numbers of the constellation diagram as input features of the learning model, performs AI model training on the target chip RF parameters through the deep learning model, extracts features through the neurons in the hidden layer of the deep learning model, and obtains a set of appropriate weights through training. The final RF parameter prediction module can predict different chip RF parameters through the trained model.

7. The chip RF calibration system based on an artificial intelligence model according to claim 5, wherein: The method for the mathematical model construction module to construct the chip RF calibration AI mathematical model includes: Step S11: Use the integrated tester to collect constellation diagrams without air interference and various RF signal modulation scenarios. Step S12: Infer ideal RF correction parameters using the collected constellation diagrams. Step S13: Use the deep learning AI model to fit the ideal RF correction parameters and construct a deep learning mathematical model C(x) according to the neuron weights and structure, where x is the signal of the receiving or transmitting element. Step S14: The user uses the signal S(t) of the receiving or transmitting element, where t is time, combines it with the deep learning mathematical model C(x), and uses it as the input data of the deep learning mathematical model C(x). C(s(t)) predicts RF parameters such as radio frequency automatic gain control (AGC). The user can use this deep learning mathematical model C(x) at any time to predict the ideal RF correction parameters to compensate for the signal power affected by different environments and reduce the constellation diagram error rate.

8. The chip RF calibration system based on an artificial intelligence model according to claim 5, wherein: The mathematical model construction module includes: an AI feature integration and extraction module for integrating and extracting AI features in the data stream, accessing the values of the integrated test instrument, and finally storing them in a variable data structure. The mathematical model construction module remotely controls the integrated tester using the UI interface, integrates chip data into a large AI model, and summarizes and trains the features. The model training module is used to train the model, take the data out of the variable data structure as the input data of the model, and fit the ideal RF parameter data through the deep learning model. The RF parameter prediction module is used to predict ideal RF parameters, and directly enables users to perform model predictions at different times and locations by generating model local files.

9. An electronic device, comprising a memory, a processor, and a computer program stored on the memory and executable on the processor, characterized in that, When the processor executes the computer program, it implements the steps of the method according to any one of claims 1 to 4.

10. A storage medium having computer program instructions stored thereon, characterized in that, When the computer program instructions are executed by the processor, it implements the steps of the method according to any one of claims 1 to 4.

Images