Power amplifier temperature characteristic modeling method based on PSO-ELM

Abstract

The invention discloses a power amplifier temperature characteristic modeling method based on PSO-ELM. The method comprises six steps of temperature data acquisition, division of a training set and a test set, initialization of a PSO-ELM model, training of the PSO-ELM model, prediction of the PSO-ELM model and result comparison. According to the core of the method, the PSO-ELM method is adopted to achieve temperature characteristic modeling of the power amplifier, the performance change of the power amplifier can be accurately predicted in a complex and changeable temperature environment, and a prediction model with generalization ability can be constructed. According to the method, the performance index of the power amplifier at any temperature can be accurately predicted in combination with limited measurement data, so that the change of circuit performance is further mastered, and more effective and convenient design guidance is provided for engineers.

Description

Technical Field

[0001] The present invention relates to the field of power amplifier temperature reliability research, and more specifically, to a power amplifier temperature characteristic modeling method based on PSO-ELM. Background Art

[0002] At present, power amplifiers (PAs) play an important role in modern wireless communication systems. Their performance has a decisive influence on the signal transmission quality and the overall efficiency of the system. With the rapid development of wireless communication technology, the reliability and performance requirements of power amplifiers in various environments are also increasing. Especially in environments with sudden temperature changes, the performance of power amplifiers is particularly important. Therefore, it is of great significance to study the impact of temperature on the performance of power amplifiers and establish a model that can accurately reflect this impact.

[0003] In past studies, the impact of temperature on the performance of power amplifiers has been explored many times, and it has been found that an increase in temperature can cause varying degrees of performance degradation or even failure. From early studies to recent explorations, scholars have continuously tried to use various methods to model and analyze the temperature characteristics of power amplifiers. However, despite the remarkable results achieved by previous studies, how to accurately predict the performance changes of power amplifiers in complex and changeable temperature environments, and how to build a model with generalization capabilities, are still the challenges currently faced. In order to solve these problems, the core of the present invention adopts the PSO-ELM (Particle Swarm Optimization-Extreme Learning Machine) modeling method to characterize the nonlinear relationship between temperature and power amplifier performance, providing engineers with a more effective and convenient design guide. Summary of the invention

[0004] The purpose of the present invention is to study the temperature characteristics of a power amplifier and propose a power amplifier temperature characteristics modeling method based on PSO-ELM, which provides important guidance for the temperature reliability research of the power amplifier.

[0005] The technical method and solution for solving the above-mentioned temperature characteristic modeling of the present invention are as follows: a method for modeling the temperature characteristics of a power amplifier based on PSO-ELM, including six steps of acquiring temperature data, dividing a training set and a test set, initializing a PSO-ELM model, training a PSO-ELM model, predicting a PSO-ELM model, and comparing results;

[0006] Step 1: The acquisition of temperature measurement data mainly uses instruments such as a DC power supply, a vector network analyzer, a temperature test box, a driver amplifier and an attenuator to measure the small signal and large signal data of the power amplifier at different temperatures. Among them, the temperature test box, DC power supply and vector network analyzer are mainly used to measure the small signal data of the power amplifier. During the test, the operating voltage, input power and frequency of the power amplifier are set to predetermined values, the working mode is selected as Line, the temperature of the power amplifier is controlled to change within the maximum operating temperature range according to a certain gradient, and the mapping relationship between the small signal gain S21, input return loss S11, and output return loss S22 of the power amplifier at different temperatures and the frequency is recorded respectively to obtain the small signal parameters of the power amplifier. Moreover, the temperature test box, DC power supply, vector network analyzer, driver amplifier and attenuator are mainly used to obtain large signal output data. During the test, the operating voltage, input power and frequency of the power amplifier are mainly set to predetermined values, the working mode is selected as CW Power, the temperature of the power amplifier is controlled to change within the maximum operating temperature range according to a certain gradient, and the relationship between the output power, gain and frequency of the power amplifier at different temperatures is recorded respectively to obtain the large signal characteristic results of the power amplifier;

[0007] Step 2: The division of training set and test set is to divide the measured data of the power amplifier at different temperatures into training set and test set. The independent variables are temperature T, frequency Frequency, input power P in ; The dependent variables are small signal gain S21, input return loss S11, output return loss S22, output power P out and Gain;

[0008] Step 3: Initialization of the PSO-ELM model is to set the number of neuron nodes in the input and output layers, the number of hidden layers, the number of neuron nodes corresponding to each hidden layer, the population size, the maximum number of iterations of particles, the learning factor, the maximum and minimum inertia weights, the position of particles, the activation function of the hidden layer, the maximum number of iterations, the learning rate and the error accuracy;

[0009] Step 4: Training the PSO-ELM model involves importing the training set into the initialized PSO-ELM model for training;

[0010] Step 5: The prediction of the PSO-ELM model is to import the test set into the trained PSO-ELM model for prediction;

[0011] Step 6: Comparison of the prediction results of the PSO-ELM model is to compare the measured data with the prediction results of the PSO-ELM model. When the mean square error MSE between the measured data and the prediction results is less than 0.1, the modeling of the temperature characteristics of the power amplifier is completed; when the mean square error MSE between the measured data and the prediction results is greater than or equal to 0.1, adjust the parameters in the initialization process, and train and predict the PSO-ELM model again until the mean square error MSE between the measured data and the prediction results is less than 0.1, then the modeling of the temperature characteristics of the power amplifier is completed.

[0012] The mean square error MSE in step 6 is calculated as follows:

[0013]

[0014] Among them, P i represents the prediction result of the PSO-ELM model, T i represents the measurement data of the power amplifier at different temperatures, and N represents the number of samples.

[0015] Compared with the prior art, the technical solution of the present invention has the following beneficial effects:

[0016] The core of the present invention adopts the PSO-ELM modeling method, which can not only accurately predict the performance changes of power amplifiers under complex and changeable temperature environments, but also build a prediction model with generalization ability. The present invention can accurately predict the performance indicators of power amplifiers at any temperature by combining limited measurement data, so as to further grasp the changes in circuit performance, so as to provide engineers with a more effective and convenient design guidance for testing. BRIEF DESCRIPTION OF THE DRAWINGS

[0017] Figure 1 The modeling method flow chart of the present invention;

[0018] Figure 2 It is a schematic diagram of the process of obtaining the small signal and large signal measurement data of the power amplifier of the present invention;

[0019] Figure 3 The modeling result of the power amplifier small signal gain S21 of the present invention;

[0020] Figure 4 The modeling result of the input return loss S11 of the power amplifier of the present invention;

[0021] Figure 5 The output return loss S22 modeling result of the power amplifier of the present invention;

[0022] Figure 6 The power amplifier of the present invention is at an input power P in When the output power is 29dBmout Modeling results.

[0023] Figure 7 The power amplifier of the present invention is at an input power P in This is the gain modeling result at 29dBm. DETAILED DESCRIPTION

[0024] The principles and features of the present invention are described below in conjunction with the accompanying drawings. The examples given are only used to explain the present invention and are not used to limit the scope of the present invention.

[0025] like Figure 1 As shown, a power amplifier temperature characteristic modeling method based on PSO-ELM includes six steps: obtaining temperature data, dividing a training set and a test set, initializing a PSO-ELM model, training a PSO-ELM model, predicting a PSO-ELM model, and comparing results;

[0026] Step 1: The acquisition of temperature measurement data mainly uses instruments such as a DC power supply, a vector network analyzer, a temperature test box, a driver amplifier and an attenuator to measure the small signal and large signal data of the power amplifier at different temperatures. Among them, the temperature test box, DC power supply and vector network analyzer are mainly used to measure the small signal data of the power amplifier. During the test, the operating voltage, input power and frequency of the power amplifier are set to predetermined values, the working mode is selected as Line, the temperature of the power amplifier is controlled to change within the maximum operating temperature range according to a certain gradient, and the mapping relationship between the small signal gain S21, input return loss S11, and output return loss S22 of the power amplifier at different temperatures and the frequency is recorded respectively to obtain the small signal parameters of the power amplifier. Moreover, the temperature test box, DC power supply, vector network analyzer, driver amplifier and attenuator are mainly used to obtain large signal output data. During the test, the operating voltage, input power and frequency of the power amplifier are mainly set to predetermined values, the working mode is selected as CW Power, the temperature of the power amplifier is controlled to change within the maximum operating temperature range according to a certain gradient, and the relationship between the output power, gain and frequency of the power amplifier at different temperatures is recorded respectively to obtain the large signal characteristic results of the power amplifier;

[0027] Step 2: The division of training set and test set is to divide the measured data of the power amplifier at different temperatures into training set and test set. The independent variables are temperature T, frequency Frequency, input power P in ; The dependent variables are small signal gain S21, input return loss S11, output return loss S22, output power P out and Gain;

[0028] Step 3: Initialization of the PSO-ELM model is to set the number of neuron nodes in the input and output layers, the number of hidden layers, the number of neuron nodes corresponding to each hidden layer, the population size, the maximum number of iterations of particles, the learning factor, the maximum and minimum inertia weights, the position of particles, the activation function of the hidden layer, the maximum number of iterations, the learning rate and the error accuracy;

[0029] Step 4: Training the PSO-ELM model involves importing the training set into the initialized PSO-ELM model for training;

[0030] Step 5: The prediction of the PSO-ELM model is to import the test set into the trained PSO-ELM model for prediction;

[0031] Step 6: Comparison of the prediction results of the PSO-ELM model is to compare the measured data with the prediction results of the PSO-ELM model. When the mean square error MSE between the measured data and the prediction results is less than 0.1, the modeling of the temperature characteristics of the power amplifier is completed; when the mean square error MSE between the measured data and the prediction results is greater than or equal to 0.1, adjust the parameters in the initialization process, and train and predict the PSO-ELM model again until the mean square error MSE between the measured data and the prediction results is less than 0.1, then the modeling of the temperature characteristics of the power amplifier is completed.

[0032] The mean square error MSE in step 6 is calculated as follows:

[0033]

[0034] Among them, P i represents the prediction result of the PSO-ELM model, T i represents the measurement data of the power amplifier at different temperatures, and N represents the number of samples.

[0035] Embodiments of the present invention.

[0036] Taking a high-efficiency GaN HEMT power amplifier with a frequency band of 2.2-3.0 GHz as an example, at room temperature, its S21 is greater than 10 dB, S11 and S22 are both less than 0 dB, saturation Pout is 39.85 dBm, Gain is 11.85 dB, and OIP3 is 36.36 dBm. A power amplifier temperature characteristic modeling method based on PSO-ELM proposed in the present invention is described in detail.

[0037] Step 1: The acquisition of temperature measurement data mainly uses instruments such as a DC power supply, a vector network analyzer, a temperature test box, a driver amplifier and an attenuator to measure the small signal and large signal data of the power amplifier at different temperatures. Among them, the temperature test box, DC power supply and vector network analyzer are mainly used to measure the small signal data of the power amplifier. During the test, the operating voltage, input power and frequency of the power amplifier are set to predetermined values, the working mode is selected as Line, the temperature of the power amplifier is controlled to change within the maximum operating temperature range according to a certain gradient, and the mapping relationship between the small signal gain S21, input return loss S11, and output return loss S22 of the power amplifier at different temperatures and the frequency is recorded respectively to obtain the small signal parameters of the power amplifier. Moreover, the temperature test box, DC power supply, vector network analyzer, driver amplifier and attenuator are mainly used to obtain large signal output data. During the test, the operating voltage, input power and frequency of the power amplifier are mainly set to predetermined values, the working mode is selected as CW Power, the temperature of the power amplifier is controlled to change within the maximum operating temperature range according to a certain gradient, and the relationship between the output power, gain and frequency of the power amplifier at different temperatures is recorded respectively to obtain the large signal characteristic results of the power amplifier;

[0038] Step 2: The division of the training set and the test set is to divide the measurement data of the power amplifier at -40℃-120℃ into a training set and a test set. The independent variables are temperature, frequency, input power P in ; The dependent variables are small signal gain S21, input return loss S11, output return loss S22, output power P out and Gain;

[0039] Step 3: Initialization of the PSO-ELM model: set the number of input layer neuron nodes to 3, the number of output layer neuron nodes to 5, the number of hidden layers to 1, the number of neuron nodes corresponding to each hidden layer to 12, the population size to 25, the maximum number of particle iterations to 150, the learning factor to 1.5, the maximum inertia weight to 0.9, the minimum inertia weight to 0.4, the particle position to be between [0, 1], the activation function of the hidden layer to "sig", the maximum number of iterations to 1000, the learning rate to 0.0001, and the error precision to 0.0004;

[0040] Step 4: Training the PSO-ELM model involves importing the training set into the initialized PSO-ELM model for training;

[0041] Step 5: The prediction of the PSO-ELM model is to import the test set into the trained PSO-ELM model for prediction;

[0042] Step 6: The comparison of the prediction results of the PSO-ELM model shows that the mean square error (MSE) between the measured data and the prediction results of the PSO-ELM model is 0.0109, which is less than 0.1, and the temperature characteristic modeling of the power amplifier is completed. It can be seen that this method can well predict the change trend of the performance indicators of the power amplifier at any temperature.

[0043] In summary, the PSO-ELM-based power amplifier temperature characteristic modeling method provides an important reference for the temperature reliability research and circuit design of power amplifiers.

[0044] The above description is only a preferred embodiment of the present invention and is not intended to limit the present invention. Any modifications, equivalent substitutions, improvements, etc. made within the spirit and principle of the present invention should be included in the protection scope of the present invention.

Claims

1. A PSO-ELM-based power amplifier temperature characteristic modeling method, including six steps: temperature data acquisition, training set and test set division, PSO-ELM model initialization, PSO-ELM model training, PSO-ELM model prediction and result comparison; The acquisition of temperature data in step 1 is mainly to measure the small signal and large signal data of the power amplifier at different temperatures using instruments such as a DC power supply, a vector network analyzer, a temperature test box, a driver amplifier and an attenuator. The small signal data of the power amplifier are mainly measured by a temperature test box, a DC power supply and a vector network analyzer. During the test, the operating voltage, input power and frequency of the power amplifier are set to the preset values, and the working mode is selected as Line. The temperature of the power amplifier is controlled to change within the maximum operating temperature range according to a certain gradient, and the mapping relationship between the small signal gain S21, input return loss S11, output return loss S22 and frequency of the power amplifier at different temperatures is recorded respectively to obtain the small signal parameters of the power amplifier. Moreover, the large signal output data is mainly obtained by using a temperature test box, a DC power supply, a vector network analyzer, a driver amplifier and an attenuator. During the test, the operating voltage, input power and frequency of the power amplifier are mainly set to the preset values, and the working mode is selected as CW Power. The temperature of the power amplifier is controlled to change within the operating temperature range according to a certain gradient, and the relationship between the output power, gain and frequency of the power amplifier at different temperatures is recorded respectively to obtain the large signal characteristic results of the power amplifier; The second step of dividing the training set and the test set is to divide the measured data of the power amplifier at different temperatures into a training set and a test set. The independent variables are temperature T, frequency Frequency, and input power Pin; the dependent variables are small signal gain S21, input return loss S11, output return loss S22, output power Pout, and gain Gain; The initialization of the PSO-ELM model in step 3 is to set the number of neuron nodes in the input and output layers, the number of hidden layers, the number of neuron nodes corresponding to each hidden layer, the population size, the maximum number of particle iterations, the learning factor, the maximum and minimum inertia weights, the position of the particles, the activation function of the hidden layer, the maximum number of iterations, the learning rate and the error accuracy; The training of the PSO-ELM model in step 4 is to import the training set into the initialized PSO-ELM model for training; The prediction of the PSO-ELM model in step 5 is to import the test set into the trained PSO-ELM model for prediction; The step six of comparing the prediction results of the PSO-ELM model is to compare the measured data and the prediction results of the PSO-ELM model. When the mean square error MSE between the measured data and the prediction results is less than 0.1, the modeling of the temperature characteristics of the power amplifier is completed; when the mean square error MSE between the measured data and the prediction results is greater than or equal to 0.1, the parameters in the initialization process are adjusted, and the PSO-ELM model is trained and predicted again until the mean square error MSE between the measured data and the prediction results is less than 0.1, then the modeling of the temperature characteristics of the power amplifier is completed.

2. The method for modeling temperature characteristics of a power amplifier based on PSO-ELM according to claim 1, characterized in that: The mean square error MSE in step 6 is calculated as follows: Among them, P i represents the prediction result of the PSO-ELM model, T i represents the measurement data of the power amplifier at different temperatures, and N represents the number of samples.

Images