Radio frequency integrated power amplifier linearity calibration method and system

Abstract

This application provides a method and system for calibrating the linearity of an integrated radio frequency (RF) power amplifier. The method involves acquiring the amplifier's input and output signals under different power and temperature conditions, and constructing a training sample set from all input and output signals. Based on the diminishing characteristics of the memory effect in each input signal, all input signals are divided into multiple input sample clusters. A pre-distortion model for each input sample cluster is pre-stored in the RF power amplifier's model library based on the training sample set. During the operation of the RF power amplifier, the input working signal is matched with each input sample cluster using time dimension weights to obtain signal labels for the working signal. The pre-distortion model corresponding to the signal label is then called from the model library to calibrate the linearity of the RF power amplifier. Using this method, the dynamic changes in the memory effect can be tracked to dynamically calibrate the amplifier's linearity.

Description

Technical Field

[0001] This application relates to the field of power amplifier technology, and more specifically, to a method and system for calibrating the linearity of an integrated radio frequency power amplifier. Background Technology

[0002] In modern wireless communication systems, complex modulation signals with high peak-to-average power ratio and large bandwidth are commonly used in pursuit of higher data rates and spectral efficiency. As a key component of the transmission link, the nonlinear distortion of the power amplifier can severely degrade signal quality and lead to spectrum regeneration and interference with adjacent channels. In order to balance efficiency and linearity, the power amplifier often operates in a near-saturated nonlinear region, which makes nonlinear distortion compensation a key technology. Digital predistortion technology has become the most mainstream linearization method due to its flexibility and high precision. Its core is to inject inverse distortion before the power amplifier to cancel the inherent nonlinearity of the amplifier.

[0003] In existing digital predistortion schemes, to handle the memory effect of power amplifiers, i.e., the current output is affected by historical inputs, models such as generalized memory polynomials are usually used for overall modeling and compensation. However, the memory effect of power amplifiers has obvious time-varying characteristics, that is, the influence of historical inputs on current nonlinearity decays over time. Most existing technologies fail to explicitly and precisely incorporate this decay law when building models or classifying signals. They often assign equal weights to input signals at different historical moments or use fixed memory depths for processing. This results in the model being unable to accurately track the dynamic changes of the memory effect under broadband signals and dynamic operating conditions, which limits the linearization accuracy in high-performance applications. Therefore, how to track the dynamic changes of the memory effect to dynamically calibrate the linearity of the amplifier has become a difficult problem for the industry. Summary of the Invention

[0004] This application provides a method and system for linearity calibration of an integrated radio frequency power amplifier, which can track the dynamic changes of memory effect to dynamically calibrate the linearity of the amplifier.

[0005] In a first aspect, this application provides a method for calibrating the linearity of an integrated radio frequency power amplifier, comprising: The input and output signals of the amplifier are collected under different power and temperature conditions, and all input and output signals are used to form a training sample set; Based on the diminishing characteristics of the power amplifier memory effect of the RF integrated power amplifier, all input signals are divided into multiple input sample clusters, and a pre-distortion model for each input sample cluster is pre-stored in the model library of the RF integrated power amplifier based on the training sample set. During the operation of the RF integrated power amplifier, the signal label of the working signal is obtained by matching the input working signal with the time dimension weight and each input sample cluster; The linearity of the RF integrated power amplifier is calibrated by calling the predistortion model corresponding to the signal tag from the model library.

[0006] In some embodiments, the input and output signals of the acquisition amplifier under different power and temperature conditions specifically include: An input signal with a preset modulation format and bandwidth is generated by a signal source; The power control unit allows the amplifier to operate at multiple different average output power levels in sequence. The temperature control device keeps the amplifier chip at multiple different junction temperature operating points in sequence. Under each fixed power and temperature combination, the input signal to the amplifier and the output signal from the amplifier are synchronously acquired via a data acquisition card.

[0007] In some embodiments, before dividing all input signals into multiple input sample clusters based on the diminishing characteristics of the power amplifier memory effect of the RF integrated power amplifier, the method further includes: determining the strength of the power amplifier memory effect of the RF integrated power amplifier in advance through a two-tone test.

[0008] In some embodiments, dividing all input signals into multiple input sample clusters based on the diminishing characteristics of the power amplifier memory effect of the RF integrated power amplifier specifically includes: Based on the power amplifier memory effect strength of the RF integrated power amplifier, an amplitude feature vector composed of the signal amplitude at the current moment and multiple consecutive historical moments is extracted from each input signal. Based on the diminishing characteristics of the memory effect, corresponding time dimension weights are assigned to the signal amplitude at different times; The memory decay sequence of each input signal is determined by weighting all time dimensions and using the amplitude feature vector of each input signal. For every two input signals, the weighted similarity between the two input signals is determined by the memory decay sequence of the two input signals; All input signals are divided into multiple input sample clusters by using all weighted similarity.

[0009] In some embodiments, the pre-distortion model for each input sample cluster pre-stored in the model library of the RF integrated power amplifier based on the training sample set specifically includes: Extract a subset of training data for each input sample cluster from the training sample set; The set of memory polynomial coefficients for each input sample cluster is determined based on each subset of training data; A predistortion model for each input sample cluster is generated based on the memory polynomial coefficient set for each input sample cluster. All predistortion models are pre-stored into the model library of the RF integrated power amplifier.

[0010] In some embodiments, the process of matching the input working signal with the time dimension weights and each input sample cluster to obtain the signal label of the working signal specifically includes: Extract the real-time amplitude sequence from the current input working signal, which consists of the amplitude of the signal at the current time and multiple consecutive historical time points; The matching similarity between the real-time amplitude sequence and each input sample cluster is determined by weighting all time dimensions. The signal label of the working signal is determined by all matching similarities.

[0011] In some embodiments, calibrating the linearity of the RF integrated power amplifier by calling a predistortion model corresponding to the signal tag from the model library specifically includes: Based on the signal label of the working signal, a predistortion model is retrieved from the pre-stored model library and uniquely corresponding to it is obtained; The current input signal is input into the predistortion model obtained by querying, and the output signal after predistortion processing is obtained. The output signal after pre-distortion processing is sent to the RF integrated power amplifier to compensate for nonlinear distortion and complete linearity calibration.

[0012] Secondly, this application provides a radio frequency integrated power amplifier linearity calibration system, comprising: The acquisition module is used to acquire the input and output signals of the amplifier under different power and temperature conditions, and to form a training sample set from all the input and output signals; The processing module is used to divide all input signals into multiple input sample clusters based on the diminishing characteristics of the power amplifier memory effect of the RF integrated power amplifier, and to pre-store the predistortion model of each input sample cluster in the model library of the RF integrated power amplifier based on the training sample set. The processing module is also used to match the input working signal with the time dimension weight and each input sample cluster during the operation of the RF integrated power amplifier to obtain the signal label of the working signal. The execution module is used to call the predistortion model corresponding to the signal tag from the model library to calibrate the linearity of the RF integrated power amplifier.

[0013] Thirdly, this application provides a computer device including a memory and a processor, the memory storing code, and the processor being configured to acquire the code and execute the above-described radio frequency integrated power amplifier linearity calibration method.

[0014] Fourthly, this application provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described radio frequency integrated power amplifier linearity calibration method.

[0015] The technical solutions provided by the embodiments disclosed in this application have the following beneficial effects: The RF integrated power amplifier linearity calibration method and system provided in this application first collects the input and output signals of the amplifier under different power and temperature conditions, and constructs a training sample set from all the input and output signals. Based on the diminishing characteristics of the RF integrated power amplifier's power amplifier memory effect, all input signals are divided into multiple input sample clusters, and a pre-distortion model for each input sample cluster is pre-stored in the RF integrated power amplifier's model library based on the training sample set. During the operation of the RF integrated power amplifier, the input working signal is matched with the time dimension weight and each input sample cluster to obtain the signal label of the working signal. The pre-distortion model corresponding to the signal label is called from the model library to calibrate the linearity of the RF integrated power amplifier.

[0016] Therefore, this application assigns time-dimensional weights to signal amplitudes at different historical moments based on the diminishing characteristics of the memory effect, and then performs weighted segmentation and matching of the input signal accordingly. First, signals under multiple operating conditions are collected to construct a training set, ensuring comprehensive model coverage. Then, the input signal is divided into multiple sample clusters using the set time weights, and a customized pre-distortion model is pre-stored for each cluster, thereby finely characterizing the nonlinear behavior under different memory decay modes. Finally, during operation, the same weighting mechanism is used to quickly match and call models for real-time signals, achieving dynamic calibration. This scheme, by introducing diminishing weights that conform to physical laws, makes the signal classification and model matching process closer to the actual memory behavior of the power amplifier, overcoming the shortcomings of traditional methods where the influence of historical signals is treated equally. Therefore, it can significantly improve the accuracy and adaptability of linearity calibration in broadband, high-power dynamic scenarios, achieving more effective compensation for nonlinear distortion. In summary, the scheme of this application can track the dynamic changes of the memory effect to dynamically calibrate the linearity of the amplifier. Attached Figure Description

[0017] Figure 1 This is an exemplary flowchart of a radio frequency integrated power amplifier linearity calibration method according to some embodiments of this application; Figure 2 This is an exemplary flowchart illustrating the acquisition of input and output signals according to some embodiments of this application; Figure 3 This is an exemplary flowchart illustrating the pre-distortion model pre-stored in a model library according to some embodiments of this application; Figure 4 This is a schematic diagram of the structure of a radio frequency integrated power amplifier linearity calibration system according to some embodiments of this application; Figure 5 This is a schematic diagram of the structure of a computer device that implements a linearity calibration method for an integrated radio frequency power amplifier according to some embodiments of this application. Detailed Implementation

[0018] To better understand the technical solution of this application, the technical solution of this application will be described in detail below with reference to the accompanying drawings and specific embodiments.

[0019] refer to Figure 1 The figure is an exemplary flowchart of a radio frequency integrated power amplifier linearity calibration method according to some embodiments of this application. The radio frequency integrated power amplifier linearity calibration method mainly includes the following steps: In step 101, the input and output signals of the amplifier under different power and temperature conditions are collected, and all the input and output signals are used to form a training sample set.

[0020] In some embodiments, reference Figure 2 This figure is an exemplary flowchart illustrating the acquisition of input and output signals according to some embodiments of this application. The acquisition of input and output signals of the amplifier under different power and temperature conditions in this application can be achieved by the following steps: In step 1011, an input signal with a preset modulation format and bandwidth is generated by a signal source; In step 1012, the amplifier is operated sequentially at multiple different average output power levels by the power control unit; In step 1013, the temperature control device is used to sequentially place the amplifier chip at multiple different junction temperature operating points; In step 1014, under each fixed power and temperature combination, the input signal input to the amplifier and the output signal output by the amplifier are synchronously acquired by the data acquisition card.

[0021] In practical implementation, the input signal with a preset modulation format and bandwidth generated by the signal source can be achieved in the following way: First, in the analog software, according to industry communication standards, such as Long Term Evolution (LTE) or New Radio (NR), a corresponding digital baseband in-phase / quadrature (I / Q) waveform is generated, and the modulation format, bandwidth, and peak-to-average power ratio (PAPR) are set. Then, the digital baseband I / Q waveform is converted into an analog baseband signal, and up-converted to the target RF frequency by a modulator. The generated signal is then used as the input signal. The analog software can be MATLAB or ADS, or other analog software; no limitation is made here. Furthermore, the modulation format, bandwidth, and PAPR can be set according to actual needs. For example, in this application, the modulation format is set to Quadrature Phase Shift Keying (QPS). Keying (QPSK), with a bandwidth set to 20MHz and a peak-to-average power ratio set to 8.52dB. Furthermore, the digital baseband I / Q waveform can be converted into an analog baseband signal by a digital-to-analog converter of a vector signal generator, which will not be elaborated here.

[0022] It should be noted that the power control unit in this application is a device capable of precisely controlling the signal power input to the RF integrated power amplifier, making it operate at a specific average output power level. As a preferred embodiment, the multiple different average output power levels can be preset in the following way: based on the gain compression characteristics of the RF integrated power amplifier, within its dynamic output power range, multiple discrete power points, including at least the small-signal linear region, the gain compression point (such as the 1dB compression point), and the saturation region, are selected as preset average output power levels. For example, multiple operating points that are 10dB, 6dB, 3dB, and 0dB lower than the saturation output power can be preset to systematically characterize its complete behavior from linear to strongly nonlinear. Alternatively, to obtain a more refined power behavior model, more operating points can be preset within this range at fixed power intervals (such as 2dB), which is not limited here.

[0023] It should be noted that the temperature control device in this application is a temperature control station. By setting the junction temperature operating point, the device can actively heat or cool the base that is thermally connected to the RF integrated power amplifier chip and monitor the temperature in real time, ultimately stabilizing the junction temperature of the RF integrated power amplifier chip at the preset operating point. As a preferred embodiment, multiple different junction temperature operating points in this application can be preset in the following way: According to the device datasheet of the RF integrated power amplifier or the standard of the target application environment, determine the temperature range that needs to be characterized (e.g., commercial grade 0°C to 70°C, or industrial grade -40°C to 85°C). Within this temperature range, select at least a lower limit temperature, room temperature (e.g., 25°C), and upper limit temperature as preset junction temperature operating points. For example, it can be preset as three operating points: -40°C, 25°C, and 85°C, to cover extreme low temperature, room temperature, and extreme high temperature conditions. Or, in order to obtain a more refined temperature behavior model, more operating points can be preset within this range at fixed intervals (e.g., 20°C). This is not limited here.

[0024] In specific implementation, under each combination of average output power level and junction temperature operating point, the synchronous acquisition of the input signal to the amplifier and the output signal from the amplifier via the data acquisition card can be achieved in the following way: First, the input signal is split into two paths by a power divider. One path is input to the RF integrated power amplifier, and the other path is directly connected to the first channel of the data acquisition card. Then, the output signal of the RF integrated power amplifier is connected to the second channel of the data acquisition card. Subsequently, under each combination of average output power level and junction temperature operating point, the input signal and the output signal are acquired via the data acquisition card. The sampling rate of the data acquisition card can be set to any value greater than five times the bandwidth of the input signal.

[0025] In practice, the training sample set of all input and output signals can be constructed in the following way: First, the input and output signals collected under each combination of average output power level and junction temperature operating point are used as a set of training samples, and the set of all training samples is used as the training sample set.

[0026] In step 102, based on the diminishing characteristics of the power amplifier memory effect of the RF integrated power amplifier, all input signals are divided into multiple input sample clusters, and a pre-distortion model for each input sample cluster is pre-stored in the model library of the RF integrated power amplifier based on the training sample set.

[0027] In some embodiments, before dividing all input signals into multiple input sample clusters based on the diminishing characteristics of the power amplifier memory effect of the RF integrated power amplifier, the strength of the power amplifier memory effect of the RF integrated power amplifier is determined in advance through a two-tone test. As a preferred embodiment, the determination of the power amplifier memory effect strength of the RF integrated power amplifier in advance through a two-tone test in this application can be implemented in the following manner: First, a vector signal generator is used to generate two continuous wave single-tone signals with similar frequencies (e.g., a frequency difference of 100kHz) and equal amplitudes as two-tone test signals, and these signals are combined and input to the... An integrated radio frequency (RF) power amplifier is then used. A spectrum analyzer is connected to the amplifier's output to measure its output spectrum. Specifically, the center frequencies of the two single-tone signals are fixed as the operating frequency of the RF power amplifier, and its total input power is kept constant at the typical operating back-off point of the RF power amplifier (e.g., always 6-10 dB below saturation power). The frequency interval between the two single-tone signals is then gradually changed from a smaller value (e.g., 100 kHz) to a larger value (e.g., 10 MHz). At each frequency interval, the power level of the third-order intermodulation distortion (IMD3) component displayed on the spectrum analyzer is recorded. Finally, the slope of the curve showing the IMD3 power level changing with the two-tone frequency interval is taken as the power amplifier memory effect strength of the RF power amplifier. The curve showing the IMD3 power level changing with the two-tone frequency interval is a first-order function curve, which can be obtained by fitting using the least squares method in existing technology, and will not be elaborated here.

[0028] It should be noted that, in this application, the power amplifier memory effect strength is a parameter value used to measure the sensitivity of the nonlinear distortion characteristics of the RF integrated power amplifier to the rate of change of its input signal.

[0029] In some embodiments, dividing all input signals into multiple input sample clusters based on the diminishing characteristics of the power amplifier memory effect of the RF integrated power amplifier can be achieved using the following steps: Based on the power amplifier memory effect strength of the RF integrated power amplifier, an amplitude feature vector composed of the signal amplitude at the current moment and multiple consecutive historical moments is extracted from each input signal. Based on the diminishing characteristics of the memory effect, corresponding time dimension weights are assigned to the signal amplitude at different times; The memory decay sequence of each input signal is determined by weighting all time dimensions and using the amplitude feature vector of each input signal. For every two input signals, the weighted similarity between the two input signals is determined by the memory decay sequence of the two input signals; All input signals are divided into multiple input sample clusters by using all weighted similarity.

[0030] In specific implementation, the amplitude feature vector composed of the signal amplitude at the current time and multiple consecutive historical times can be extracted from each input signal based on the power amplifier memory effect strength of the RF integrated power amplifier. This can be achieved in the following way: First, determine the maximum memory depth M based on the power amplifier memory effect strength of the RF integrated power amplifier. Specifically, take the integer obtained by rounding up five times the power amplifier memory effect strength as the maximum memory depth M. Then, for each input signal x(n) in the training sample set, extract the amplitude value of its signal envelope at time n and M consecutive times before n (i.e., n, n−1, …, n−M), and arrange these M+1 amplitude values ​​in chronological order into a vector. This vector is the amplitude feature vector of the input signal at time n.

[0031] In specific implementation, the time dimension weights for signal amplitudes at different times are set according to the diminishing characteristics of the memory effect in the following way: First, the time dimension weight corresponding to the signal amplitude at the current time is set to 1. Then, the time dimension weights corresponding to signal amplitudes at other times are set sequentially according to the memory depth from small to large, that is, the time difference between the historical time and the current time from small to large. The setting rule is to decrease according to the memory depth, that is, to set it as the product of the time dimension weight corresponding to the signal amplitude at the previous memory depth and the preset decreasing coefficient. The decreasing coefficient can be preset to a positive number less than 1 according to actual needs. For example, in this application, the decreasing coefficient is set to 0.8. In other embodiments, the decreasing coefficient can also be set to other positive numbers less than 1. This is not limited here.

[0032] It should be noted that the time dimension weight in this application is a coefficient used to quantify the difference in the degree of influence of the input signal at different historical moments on the current nonlinear distortion of the power amplifier.

[0033] In practice, the memory decay sequence of each input signal can be determined by using all time dimension weights and the amplitude feature vector of each input signal in the following way: multiply the signal amplitude at each time step in the amplitude feature vector of each input signal by the weight vector at each time step, and use the new sequence obtained after multiplication as the memory decay sequence of each input signal.

[0034] It should be noted that the memory decay sequence in this application is a sequence used to characterize the relationship between the memory effect and nonlinear coupling of the power amplifier.

[0035] In practice, the weighted similarity between two input signals can be determined by using the memory decay sequences of each pair of input signals in the following way: for each pair of input signals, determine the Euclidean distance between the memory decay sequences of each pair of input signals, and use the reciprocal of the Euclidean distance between the memory decay sequences of each pair of input signals as the weighted similarity between each pair of input signals.

[0036] It should be noted that the weighted similarity in this application is a parameter value used to quantify the similarity between two input signals in characterizing the nonlinear behavior of the power amplifier after considering their historical amplitude weighted attenuation.

[0037] In specific implementation, dividing all input signals into multiple input sample clusters using all weighted similarities can be achieved in the following way: The existing K-clustering algorithm can be used to cluster all input signals. Specifically, the weighted similarity between any two input signals is used as the similarity between the two input signals in the K-clustering algorithm process, and the final clustering result of the K-clustering algorithm is used as multiple input sample clusters. The K value in the K-clustering algorithm can be determined using the elbow rule. Specifically, the sum of squared errors within each cluster (i.e., the sum of similarities from each sample to its cluster center) under different candidate K values ​​(e.g., from 2 to 10) is calculated, and a curve showing the sum of squared errors changing with the K value is plotted. The K value corresponding to the inflection point of the curve is selected as the final number of clusters. The inflection point of the curve can be obtained by calculating the point corresponding to the maximum difference in the sum of squared errors between adjacent K values, which will not be elaborated further here.

[0038] It should be noted that the input sample cluster in this application is a collection of input signals that constitute the behavior mode of the RF integrated power amplifier.

[0039] In some embodiments, reference Figure 3 The figure is an exemplary flowchart illustrating the pre-stored predistortion model in a model library according to some embodiments of this application. The pre-stored predistortion model for each input sample cluster in the model library of the RF integrated power amplifier based on the training sample set in this application can be achieved using the following steps: In step 1021, a subset of training data for each input sample cluster is extracted from the training sample set; In step 1022, the set of memory polynomial coefficients for each input sample cluster is determined based on each subset of training data; In step 1023, a predistortion model for each input sample cluster is generated based on the memory polynomial coefficient set for each input sample cluster; In step 1024, all predistortion models are pre-stored into the model library of the RF integrated power amplifier.

[0040] In specific implementation, the extraction of training data subsets for each input sample cluster from the training sample set can be achieved in the following way: Select an input sample cluster as the selected sample cluster, obtain the output signal corresponding to each input signal in the selected sample cluster from the training sample set, and take the set of each input signal and the corresponding output signal as training data. Then, take the set of all training data corresponding to the selected sample cluster as the training data subset, continue to determine the training data subsets of the remaining input sample clusters, and finally, set a unique and corresponding signal label for each training data subset. For example, if there are 10 training data subsets, the signal labels of these 10 training data subsets can be set to integers between 1 and 10 respectively.

[0041] In specific implementation, determining the set of memory polynomial coefficients for each input sample cluster based on each training data subset can be achieved as follows: Select a training data subset as the chosen training data subset, and determine the basis function matrix corresponding to the chosen training data subset. Specifically, for all sampling points of all input signals in the chosen training data subset, determine the nonlinear basis function values ​​corresponding to each sampling point based on the preset model order and maximum memory depth M. Assuming that the training data subset contains N sampling points and P nonlinear basis functions are defined according to the model structure, arrange the P basis function values ​​of each sampling point in rows to form a basis function matrix X of size N rows × P columns. Subsequently, arrange the sampling values ​​of all output signals in the chosen training data subset that strictly correspond to the above N sampling points into a matrix of length... Let Y be a column vector of N. Finally, the linear equation system Xa=Y is solved by the least squares method in the prior art. The coefficient vector a of size P×1 is the memory polynomial coefficient set of the input sample cluster corresponding to the selected training data subset. The determination of the nonlinear basis function values ​​corresponding to each sampling point according to the preset model order and the maximum memory depth M means that for the input signal x(n) at each sampling time n, a series of basis function terms of the form x(nm)|x(nm)|^(k) are constructed according to the preset model order K and the maximum memory depth M. The memory index m takes from 0 to M and the nonlinear order k takes from 0 to K. The values ​​of the basis functions corresponding to all possible (m,k) combinations at time n are arranged in order of the nonlinear order k, which constitutes the data of one row in the basis function matrix X for that sampling point.

[0042] Further, it should be noted that the model order in this application can be preset according to the prior knowledge of the amplifier's nonlinear strength. A specific method is to test different candidate order values (such as K = 1, 3, 5, 7) on the training sample set, perform a complete model training for each candidate K value and evaluate its fitting accuracy on a set of independent validation data (such as calculating the normalized mean square error), and finally select the smallest K value that makes the accuracy meet the requirements (such as the error is lower than the threshold) or the accuracy improvement tends to be gentle as the preset model order. In other embodiments, the model order can also be preset by other methods, which is not limited here.

[0043] When specifically implemented, generating the predistortion model for each input sample cluster based on the memory polynomial coefficient set of each input sample cluster can be implemented in the following manner: bind the memory polynomial coefficient set of each input sample cluster to the standard mathematical expression of the memory polynomial model. The predistortion model is a mathematical function whose input is the complex baseband signal values at the current and historical moments, and the output is the predistorted signal value. The calculation rule of this function is completely defined by the memory polynomial coefficient set, that is, the input signal is weighted and summed using this memory polynomial coefficient set. Specifically, for a given input sample cluster, its predistortion model can be expressed as: for the input signal x(n), the output y(n) = ΣΣa(mk)x(n - m)·|x(n - m)|^k, where the two Σs are the summations over m from 0 to M and over k from 0 to K respectively. Here, M is the maximum memory depth, K is the model order, a(mk) is the memory polynomial coefficient corresponding to the memory depth m and the nonlinear order k in the memory polynomial coefficient set of the given input sample cluster, and x(n - m) is the sampling point of the input signal at the moment n - m, and n is the current moment.

[0044] When specifically implemented, pre-storing all the predistortion models into the model library of the radio frequency integrated power amplifier can be implemented in the following manner: First, establish a mapping relationship among the signal label, the clustering center, and the predistortion model of each input sample cluster, and establish an index structure, and write it into the model library of the radio frequency integrated power amplifier through a programming interface.

[0045] It should be noted that as a preferred embodiment, the model library in this application can be the non-volatile memory of the digital signal processor supporting the radio frequency integrated power amplifier.

[0046] In step 103, during the operation of the radio frequency integrated power amplifier, match the input working signal with the time dimension weight and each input sample cluster to obtain the signal label of the working signal.

[0047] In some embodiments, the signal label of the working signal is obtained by matching the input working signal with the time dimension weight and each input sample cluster, which can be achieved by the following steps: Extract the real-time amplitude sequence from the current input working signal, which consists of the amplitude of the signal at the current time and multiple consecutive historical time points; The matching similarity between the real-time amplitude sequence and each input sample cluster is determined by weighting all time dimensions. The signal label of the working signal is determined by all matching similarities.

[0048] In specific implementation, the real-time amplitude sequence composed of the signal amplitudes of the current time and multiple consecutive historical time points can be extracted from the current input working signal in the following way: For the current input working signal, according to the same maximum memory depth M as the offline stage, extract the amplitude values ​​of the signal envelope of the current time and the previous M consecutive historical time points. Specifically, starting from the current sampling point, backtrack to the number of historical sampling points determined by the maximum memory depth, read the signal amplitudes of all these time points in sequence, and arrange them in order from the most recent to the oldest time. The sequence obtained by arranging them is used as the real-time amplitude sequence.

[0049] In specific implementation, the matching similarity between the real-time amplitude sequence and each input sample cluster can be determined by all time dimension weights in the following way: First, obtain the pre-calculated and stored cluster centers of each input sample cluster from the model library, and then perform weighted processing on the real-time amplitude sequence and each cluster center by all time dimension weights respectively. That is, first multiply the amplitude value of each moment in the real-time amplitude sequence by the corresponding time dimension weight to obtain A, and then multiply the amplitude value of each moment in each cluster center by the corresponding time dimension weight and calculate the Euclidean distance with A. Finally, take the reciprocal of each Euclidean distance as the matching similarity between the real-time amplitude sequence and each input sample cluster respectively.

[0050] It should be noted that in this application, the matching similarity is a parameter value used to quantify the similarity between the real-time input working signal and the selected input sample cluster divided in the offline training stage in the weighted feature space.

[0051] In a specific implementation, the signal label of the working signal can be determined by all matching similarities in the following way: the signal label corresponding to the input sample cluster with the highest matching similarity is used as the signal label of the working signal.

[0052] In step 104, the linearity of the RF integrated power amplifier is calibrated by calling the predistortion model corresponding to the signal tag from the model library.

[0053] In some embodiments, calibrating the linearity of the RF integrated power amplifier by calling the predistortion model corresponding to the signal tag from the model library can be achieved using the following steps: Based on the signal label of the working signal, a predistortion model is retrieved from the pre-stored model library and uniquely corresponding to it is obtained; The current input signal is input into the predistortion model obtained by querying, and the output signal after predistortion processing is obtained. The output signal after pre-distortion processing is sent to the RF integrated power amplifier to compensate for nonlinear distortion and complete linearity calibration.

[0054] In practice, the current input signal is input into the predistortion model obtained by querying, and the output signal after predistortion processing can be obtained in the following way: the signal amplitude values ​​of the current time and the previous M consecutive historical time moments are substituted into the mathematical expression of the obtained predistortion model for calculation, where M is the maximum memory depth, and the output result is used as the output signal after predistortion processing.

[0055] In a specific implementation, the output signal after pre-distortion processing is sent to the RF integrated power amplifier to compensate for nonlinear distortion. The linearity calibration can be achieved in the following way: the output signal after pre-distortion processing is converted into an analog baseband signal through a digital-to-analog converter. Then, the analog baseband signal is up-converted to the operating frequency of the RF integrated power amplifier using an RF modulator to generate an RF pre-distortion signal. Finally, the RF pre-distortion signal is fed into the input terminal of the RF integrated power amplifier through an RF transmission link. The amplifier amplifies the input signal, which already contains inverse distortion components. Its inherent nonlinear distortion cancels out the inverse components in the pre-distortion signal, thereby linearizing and improving the final output signal of the amplifier, thus completing the calibration.

[0056] In another aspect, in some embodiments, this application provides a radio frequency integrated power amplifier linearity calibration system, with reference to... Figure 4 The figure is a schematic diagram of the structure of a radio frequency integrated power amplifier linearity calibration system according to some embodiments of this application. The radio frequency integrated power amplifier linearity calibration system 400 includes: a data acquisition module 401, a processing module 402, and an execution module 403, which are described below: The acquisition module 401 in this application is mainly used to acquire the input and output signals of the amplifier under different power and temperature conditions, and to form a training sample set of all the input and output signals. Processing module 402, in this application, is mainly used to divide all input signals into multiple input sample clusters according to the decreasing characteristics of the power amplifier memory effect of the RF integrated power amplifier, and to pre-store the predistortion model of each input sample cluster in the model library of the RF integrated power amplifier based on the training sample set. It should be noted that the processing module 402 in this application is also used to match the input working signal with the time dimension weight and each input sample cluster during the operation of the RF integrated power amplifier to obtain the signal label of the working signal; The execution module 403 in this application is mainly used to call the predistortion model corresponding to the signal tag from the model library to calibrate the linearity of the RF integrated power amplifier.

[0057] In addition, this application also provides a computer device, the computer device including a memory and a processor, the memory storing code, the processor being configured to acquire the code and execute the above-described radio frequency integrated power amplifier linearity calibration method.

[0058] In some embodiments, reference Figure 5 The figure is a schematic diagram of a computer device implementing a radio frequency integrated power amplifier linearity calibration method according to some embodiments of this application. The radio frequency integrated power amplifier linearity calibration method in the above embodiments can be achieved through... Figure 5 The computer device shown is used to implement this, and the computer device 500 includes at least one processor 501, a communication bus 502, a memory 503, and at least one communication interface 504.

[0059] Processor 501 can be a general-purpose central processing unit (CPU) or an application-specific integrated circuit (ASIC).

[0060] The communication bus 502 can be used to transmit information between the aforementioned components.

[0061] Memory 503 may be a read-only memory (ROM) or other type of static storage device capable of storing static information and instructions, random access memory (RAM) or other type of dynamic storage device capable of storing information and instructions, or electrically erasable programmable read-only memory (EEPROM), compact disc read-only memory (CDROM) or other optical disc storage, optical disc storage (including compressed optical discs, laser discs, optical discs, digital versatile optical discs, Blu-ray discs, etc.), magnetic disks or other magnetic storage devices, or any other medium capable of carrying or storing desired program code in the form of instructions or data structures and accessible by a computer, but not limited thereto. Memory 503 may exist independently and be connected to processor 501 via communication bus 502. Memory 503 may also be integrated with processor 501.

[0062] The memory 503 stores program code for executing the scheme of this application, and its execution is controlled by the processor 501. The processor 501 executes the program code stored in the memory 503. The program code may include one or more software modules. In the above embodiments, the RF integrated power amplifier linearity calibration method can be implemented by the processor 501 and one or more software modules in the program code in the memory 503.

[0063] Communication interface 504 uses any transceiver-like device to communicate with other devices or communication networks, such as Ethernet, radio access network (RAN), wireless local area networks (WLAN), etc.

[0064] In a specific implementation, as one example, a computer device may include multiple processors, each of which may be a single-core (single CPU) processor or a multi-core (multi CPU) processor. Here, a processor may refer to one or more devices, circuits, and / or processing cores used to process data (e.g., computer program instructions).

[0065] The aforementioned computer device can be a general-purpose computer device or a special-purpose computer device. In specific implementations, the computer device can be a desktop computer, a portable computer, a network server, a handheld digital assistant (PDA), a mobile phone, a tablet computer, a wireless terminal device, a communication device, or an embedded device. This application does not limit the type of computer device.

[0066] In addition, this application also provides a computer-readable storage medium storing a computer program that, when executed by a processor, implements the above-described radio frequency integrated power amplifier linearity calibration method.

[0067] In summary, the RF integrated power amplifier linearity calibration method and system disclosed in this application firstly acquires the input and output signals of the amplifier under different power and temperature conditions, and constructs a training sample set from all the input and output signals; based on the diminishing characteristics of the power amplifier memory effect, all input signals are divided into multiple input sample clusters, and a pre-distortion model for each input sample cluster is pre-stored in the model library of the RF integrated power amplifier based on the training sample set; during the operation of the RF integrated power amplifier, the input working signal is matched with the time dimension weight and each input sample cluster to obtain the signal label of the working signal; the pre-distortion model corresponding to the signal label is called from the model library to calibrate the linearity of the RF integrated power amplifier.

[0068] Therefore, this application assigns time-dimensional weights to signal amplitudes at different historical moments based on the diminishing characteristics of the memory effect, and then performs weighted segmentation and matching of the input signal accordingly. First, signals under multiple operating conditions are collected to construct a training set, ensuring comprehensive model coverage. Then, the input signal is divided into multiple sample clusters using the set time weights, and a customized pre-distortion model is pre-stored for each cluster, thereby finely characterizing the nonlinear behavior under different memory decay modes. Finally, during operation, the same weighting mechanism is used to quickly match and call models for real-time signals, achieving dynamic calibration. This scheme, by introducing diminishing weights that conform to physical laws, makes the signal classification and model matching process closer to the actual memory behavior of the power amplifier, overcoming the shortcomings of traditional methods where the influence of historical signals is treated equally. Therefore, it can significantly improve the accuracy and adaptability of linearity calibration in broadband, high-power dynamic scenarios, achieving more effective compensation for nonlinear distortion. In summary, the scheme of this application can track the dynamic changes of the memory effect to dynamically calibrate the linearity of the amplifier.

[0069] Although preferred embodiments of this application have been described, those skilled in the art, upon learning the basic inventive concept, can make other changes and modifications to these embodiments. Therefore, the appended claims are intended to be interpreted as including the preferred embodiments as well as all changes and modifications falling within the scope of this application.

[0070] Obviously, those skilled in the art can make various modifications and variations to this application without departing from the spirit and scope of this application. Therefore, if such modifications and variations fall within the scope of the claims of this application and their equivalents, this application also intends to include such modifications and variations.

Claims

1. A method of linearization calibration of a radio frequency integrated power amplifier, characterized by, include: The input and output signals of the amplifier are collected under different power and temperature conditions, and all input and output signals are used to form a training sample set; Based on the diminishing characteristics of the power amplifier memory effect of the RF integrated power amplifier, all input signals are divided into multiple input sample clusters, and a pre-distortion model for each input sample cluster is pre-stored in the model library of the RF integrated power amplifier based on the training sample set. During the operation of the RF integrated power amplifier, the signal label of the working signal is obtained by matching the input working signal with the time dimension weight and each input sample cluster; The linearity of the RF integrated power amplifier is calibrated by calling the predistortion model corresponding to the signal tag from the model library.

2. The method of claim 1, wherein, The input and output signals of the acquisition amplifier under different power and temperature conditions specifically include: An input signal with a preset modulation format and bandwidth is generated by a signal source; The power control unit allows the amplifier to operate at multiple different average output power levels in sequence. The temperature control device keeps the amplifier chip at multiple different junction temperature operating points in sequence. Under each fixed power and temperature combination, the input signal to the amplifier and the output signal from the amplifier are synchronously acquired via a data acquisition card.

3. The method of claim 1, wherein, Before dividing all input signals into multiple input sample clusters based on the diminishing characteristics of the power amplifier memory effect of the RF integrated power amplifier, the method further includes: determining the strength of the power amplifier memory effect of the RF integrated power amplifier in advance through a two-tone test.

4. The method of claim 1, wherein, Based on the diminishing characteristics of the memory effect of the RF integrated power amplifier, all input signals are divided into multiple input sample clusters, specifically including: Based on the power amplifier memory effect strength of the RF integrated power amplifier, an amplitude feature vector composed of the signal amplitude at the current moment and multiple consecutive historical moments is extracted from each input signal. Based on the diminishing characteristics of the memory effect, corresponding time dimension weights are assigned to the signal amplitude at different times; The memory decay sequence of each input signal is determined by weighting all time dimensions and using the amplitude feature vector of each input signal. For every two input signals, the weighted similarity between the two input signals is determined by the memory decay sequence of the two input signals; All input signals are divided into multiple input sample clusters by using all weighted similarity.

5. The method as described in claim 1, characterized in that, The predistortion model for each input sample cluster pre-stored in the model library of the RF integrated power amplifier based on the training sample set specifically includes: Extract a subset of training data for each input sample cluster from the training sample set; The set of memory polynomial coefficients for each input sample cluster is determined based on each subset of training data; A predistortion model for each input sample cluster is generated based on the memory polynomial coefficient set for each input sample cluster. All predistortion models are pre-stored into the model library of the RF integrated power amplifier.

6. The method as described in claim 1, characterized in that, The signal label of the working signal is obtained by matching the input working signal with the time dimension weight and each input sample cluster. Specifically, it includes: Extract the real-time amplitude sequence from the current input working signal, which consists of the amplitude of the signal at the current time and multiple consecutive historical time points; The matching similarity between the real-time amplitude sequence and each input sample cluster is determined by weighting all time dimensions. The signal label of the working signal is determined by all matching similarities.

7. The method as described in claim 1, characterized in that, The calibration of the linearity of the RF integrated power amplifier by calling the predistortion model corresponding to the signal tag from the model library specifically includes: Based on the signal label of the working signal, a predistortion model is retrieved from the pre-stored model library and uniquely corresponding to it is obtained; The current input signal is input into the predistortion model obtained by querying, and the output signal after predistortion processing is obtained. The output signal after pre-distortion processing is sent to the RF integrated power amplifier to compensate for nonlinear distortion and complete linearity calibration.

8. A linearity calibration system for an integrated radio frequency power amplifier, characterized in that, include: The acquisition module is used to acquire the input and output signals of the amplifier under different power and temperature conditions, and to form a training sample set from all the input and output signals; The processing module is used to divide all input signals into multiple input sample clusters based on the diminishing characteristics of the power amplifier memory effect of the RF integrated power amplifier, and to pre-store the predistortion model of each input sample cluster in the model library of the RF integrated power amplifier based on the training sample set. The processing module is also used to match the input working signal with the time dimension weight and each input sample cluster during the operation of the RF integrated power amplifier to obtain the signal label of the working signal. The execution module is used to call the predistortion model corresponding to the signal tag from the model library to calibrate the linearity of the RF integrated power amplifier.

9. A computer device, characterized in that, The computer device includes a memory and a processor, the memory storing code, and the processor being configured to retrieve the code and execute the radio frequency integrated power amplifier linearity calibration method as described in any one of claims 1 to 7.

10. A computer-readable storage medium storing a computer program, characterized in that, When the computer program is executed by the processor, it implements the radio frequency integrated power amplifier linearity calibration method as described in any one of claims 1 to 7.

Images