Artificial intelligence-based digital pre-distortion device and method for improving nonlinearity of power amplifier

Abstract

Disclosed are an artificial intelligence-based digital pre-distortion device and method for improving nonlinearity of a power amplifier. The artificial intelligence-based digital pre-distortion method for improving nonlinearity of a power amplifier, according to an embodiment of the present application, may comprise the steps of: (a) collecting training data including input signal characteristics and output signal characteristics of the power amplifier; (b) constructing an artificial intelligence model for predicting and correcting nonlinear distortion of the power amplifier by using the training data, and mounting, in an on-device manner, the artificial intelligence model on a communication device including the power amplifier; (c) inputting characteristic information of a real-time signal associated with the power amplifier into the artificial intelligence model; and (d) deriving a distortion correction value for the real-time signal by using the artificial intelligence model, and correcting the real-time signal on the basis of the derived distortion correction value.

Description

AI-based digital pre-distortion device and method for improving non-linearity of power amplifier

[0001] The present invention relates to an artificial intelligence-based digital pre-distortion device and method for improving the non-linearity of a power amplifier.

[0002] As mobile communication systems transition to the Orthogonal Frequency Division Multiplexing (OFDM) modulation method and data throughput increases significantly in 5G, it has led to an increase in the Peak-to-Average Power Ratio (PAPR) value. In particular, improving the efficiency of power amplifiers, which account for most of the power consumption of mobile communication systems, is becoming an increasingly important core technology.

[0003] In this regard, Figure 1 is a conceptual diagram illustrating the nonlinear operation characteristics of a power amplifier.

[0004] Referring to Figure 1, Digital Pre-Distortion (DPD) technology is widely used to improve the nonlinearity of power amplifiers, but current DPD technology is based on mathematical models and has limitations in accurately predicting and correcting complex and diverse distortion patterns. In addition, existing DPD technology compensates for signal distortion using look-up tables, which has the problem of being difficult to effectively respond to various operating conditions and environmental changes of power amplifiers.

[0005] In addition, in 5G communication, the nonlinearity of power amplifiers is further intensified due to the use of high frequency bands and broadband signals, and in Massive MIMO systems using multiple antennas, it is a very difficult task to manage and correct them integrally because each individual power amplifier has different characteristics.

[0006] Meanwhile, existing cloud-based AI systems face difficulties in applying nonlinearity correction to power amplifiers requiring real-time signal processing due to communication delays with the server. In particular, in environments with unstable network connections, not only is normal operation difficult, but there are also limitations such as high energy consumption for data transmission and reception.

[0007] The technology forming the background of the present invention is disclosed in Korean Registered Patent Publication No. 10-2055568.

[0008] The present invention aims to solve the problems of the aforementioned conventional technology by providing an AI-based digital pre-distortion device and method for improving the non-linearity of a power amplifier, which can effectively predict and correct the non-linearity of the power amplifier using AI technology.

[0009] The present invention aims to solve the problems of the aforementioned conventional technology by providing a lightweight AI model capable of analyzing and correcting various nonlinear distortion patterns occurring in a power amplifier in real time, and thereby providing an AI-based digital pre-distortion device and method for improving the nonlinearity of a power amplifier that can simultaneously improve the linearity and efficiency of the power amplifier.

[0010] However, the technical problems that the embodiments of the present invention aim to solve are not limited to the technical problems described above, and other technical problems may exist.

[0011] As a technical means for achieving the above-mentioned technical problem, an artificial intelligence-based digital pre-distortion method for improving the non-linearity of a power amplifier according to one embodiment of the present invention may include: (a) collecting training data including input signal characteristics and output signal characteristics of a power amplifier; (b) constructing an artificial intelligence model for predicting and correcting non-linear distortion of the power amplifier using the training data, and mounting the artificial intelligence model in an on-device form on a communication device equipped with the power amplifier; (c) inputting characteristic information of a real-time signal associated with the power amplifier into the artificial intelligence model; and (d) deriving a distortion correction value for the real-time signal through the artificial intelligence model and correcting the real-time signal based on the derived distortion correction value.

[0012] Additionally, the input signal characteristics may include at least one of input power level information corresponding to the power level of the input signal input to the power amplifier, frequency band information indicating frequency information of the band supported by the power amplifier, and modulation format information indicating the modulation method of the input signal.

[0013] Additionally, the output signal characteristics may include at least one of output power information indicating the power level of the output signal of the power amplifier, spectrum analysis information indicating spectrum changes within the frequency band of the output signal, and nonlinear distortion characteristic information including AM-AM conversion characteristics and AM-PM conversion characteristics.

[0014] Additionally, the learning data may include amplifier characteristics regarding at least one of a memory effect and temperature change indicating the influence of past input signals of the power amplifier on the current output, and environmental variables regarding at least one of power supply voltage change, frequency band limit, and power limit of the power amplifier.

[0015] In addition, the distortion correction value may include at least one of a DPD correction parameter including an AM-AM correction value and an AM-PM correction value, a feedback control parameter including a real-time correction control signal and an adaptive parameter adjustment value according to environmental changes of the power amplifier, and a signal quality indicator including Error Vector Magnitude (EVM) and Adjacent Channel Power Ratio (ACPR).

[0016] In addition, the communication device may be equipped with an array-type antenna comprising a plurality of individual antenna modules.

[0017] In addition, the artificial intelligence model may be applied independently to each of the power amplifiers provided in each of the individual antenna modules, or to the power amplifiers included in an antenna module group that is grouped to include at least two of the individual antenna modules.

[0018] In addition, the artificial intelligence model can be mounted on a beamforming IC equipped in the communication device.

[0019] Additionally, the above step (b) may include the step of applying a lightweighting technique to the artificial intelligence model, the lightweighting technique comprising at least one of pruning and quantization.

[0020] Additionally, the above step (b) may include a step in which the artificial intelligence model autonomously learns an optimal correction strategy to correct distortion occurring in the output of the power amplifier through reinforcement learning.

[0021] In addition, the above artificial intelligence model can be driven through a hardware accelerator or a Neural Processing Unit (NPU).

[0022] In addition, an artificial intelligence-based digital pre-distortion method for improving the non-linearity of a power amplifier according to one embodiment of the present invention may include the step of collecting real-time operation data of the power amplifier in response to a time-series change in the communication environment of the communication device, retraining the artificial intelligence model, and re-loading the retrained artificial intelligence model onto the communication device.

[0023] Meanwhile, an artificial intelligence-based digital pre-distortion device for improving the non-linearity of a power amplifier according to one embodiment of the present invention may include: a data collection unit that collects learning data including input signal characteristics and output signal characteristics of a power amplifier; a model learning unit that builds an artificial intelligence model for predicting and correcting non-linear distortion of the power amplifier using the learning data and mounts the artificial intelligence model in an on-device form on a communication device equipped with the power amplifier; and a correction execution unit that inputs characteristic information of a real-time signal associated with the power amplifier into the artificial intelligence model, derives a distortion correction value for the real-time signal through the artificial intelligence model, and corrects the real-time signal based on the derived distortion correction value.

[0024] In addition, the model learning unit can collect real-time operation data of the power amplifier in response to time-series changes in the communication environment of the communication device to retrain the artificial intelligence model, and re-install the retrained artificial intelligence model on the communication device.

[0025] The means for solving the problem described above are merely exemplary and should not be interpreted as intended to limit the present invention. In addition to the exemplary embodiments described above, additional embodiments may exist in the drawings and the detailed description of the invention.

[0026] According to the means for solving the problem of the present invention described above, it is possible to provide an AI-based digital pre-distortion device and method for improving the non-linearity of a power amplifier, which can effectively predict and correct the non-linearity of the power amplifier by utilizing AI technology.

[0027] According to the means for solving the problem of the present invention described above, a lightweight AI model capable of analyzing and correcting various nonlinear distortion patterns occurring in a power amplifier in real time is provided, and thereby, an AI-based digital pre-distortion device and method for improving the nonlinearity of a power amplifier can be provided, which can simultaneously improve the linearity and efficiency of the power amplifier.

[0028] According to the solution to the problem of the present invention described above, the nonlinearity of a power amplifier can be accurately predicted and corrected in real time through on-device AI-based DPD technology, and there is an advantage that high-accuracy distortion correction is possible with minimal computational resources through the lightweighting of the AI ​​model.

[0029] According to the solution to the problem of the present invention described above, by incorporating an AI model into the beamforming IC, independent distortion correction is possible for each power amplifier of an individual antenna module, thereby improving the overall performance of the Massive MIMO system.

[0030] According to the solution to the problem of the present invention described above, an on-device AI system that operates with very low latency compared to a cloud-based system is provided, which has the advantage of enabling stable operation regardless of the network connection status.

[0031] According to the solution to the problem described above, it is possible to effectively respond to changes in the communication environment over time through the establishment of autonomous correction strategies via reinforcement learning and periodic model retraining, thereby preventing long-term performance degradation of the power amplifier.

[0032] However, the effects obtainable from this invention are not limited to those described above, and other effects may exist.

[0033] Figure 1 is a conceptual diagram illustrating the nonlinear operation characteristics of a power amplifier.

[0034] FIG. 2 is a schematic diagram of an open LAN-based communication system comprising a communication device having an artificial intelligence-based digital pre-distortion device for improving the non-linearity of a power amplifier according to one embodiment of the present invention.

[0035] FIG. 3 is a conceptual diagram illustrating the operation flow of a communication device equipped with an artificial intelligence-based digital pre-distortion device for improving the non-linearity of a power amplifier according to one embodiment of the present invention.

[0036] FIG. 4 is a conceptual diagram showing a detailed circuit of a communication device equipped with an artificial intelligence-based digital pre-distortion device for improving the non-linearity of a power amplifier according to one embodiment of the present invention.

[0037] FIG. 5 is a detailed circuit diagram of an artificial intelligence-based digital pre-distortion device for improving the non-linearity of a power amplifier according to one embodiment of the present invention.

[0038] FIG. 6 is a schematic diagram of an artificial intelligence-based digital pre-distortion device for improving the non-linearity of a power amplifier according to one embodiment of the present invention.

[0039] FIG. 7 is an operation flowchart of an artificial intelligence-based digital pre-distortion method for improving the non-linearity of a power amplifier according to one embodiment of the present invention.

[0040] Figure 8 is a detailed flowchart of the retraining process of an artificial intelligence model.

[0041] Embodiments of the present invention are described below with reference to the attached drawings to enable those skilled in the art to easily implement the invention. However, the present invention may be embodied in various different forms and is not limited to the embodiments described herein. Furthermore, in order to clearly explain the present invention in the drawings, parts unrelated to the explanation have been omitted, and similar parts throughout the specification are denoted by similar reference numerals.

[0042] Throughout this specification, when a part is described as being "connected" to another part, this includes not only cases where they are "directly connected," but also cases where they are "electrically connected" or "indirectly connected" with other elements interposed between them.

[0043] Throughout the entire specification, when a component is described as being located "on," "on top," "on top," "under," "on bottom," or "on bottom" of another component, this includes not only cases where the component is in contact with the other component but also cases where another component exists between the two components.

[0044] Throughout this specification, when a part is described as "comprising" a certain component, this means that, unless specifically stated otherwise, it does not exclude other components but may include additional components.

[0045] The present invention relates to an artificial intelligence-based digital pre-distortion device and method for improving the non-linearity of a power amplifier.

[0046] FIG. 2 is a schematic diagram of an open LAN-based communication system comprising a communication device having an artificial intelligence-based digital pre-distortion device for improving the non-linearity of a power amplifier according to one embodiment of the present invention.

[0047] Referring to FIG. 2, an open RAN-based communication system (1000) according to one embodiment of the present invention may include an O-RU (O-RAN Distributed Unit; 10), an O-DU (O-RAN Radio Unit; 20), and an antenna module (30). Additionally, the artificial intelligence-based digital pre-distortion device (100) for improving the non-linearity of a power amplifier disclosed herein (hereinafter referred to as the 'digital pre-distortion device (100)') may be provided for the O-RU (10), which is a communication device forming the open RAN-based communication system (1000), but is not limited thereto. According to an embodiment of the present invention, the digital pre-distortion device (100) may be widely applied to various types of communication devices (equipment), such as base station devices and repeater devices.

[0048] In an O-RAN (Open-Radio Access Network), the physical layer is functionally separated, and the O-DU (20) included in the Open-RAN-based communication system (1000) may be responsible for the High PHY, which is option 7, and the O-RU (10) may be responsible for the Low PHY, which is option 8, and these O-RU (10) and O-DU (20) may be connected to each other via a fronthaul interface.

[0049] Communication devices (10), O-DU (20), and antenna modules (30) can communicate with each other through a network (not shown). A network (not shown) refers to a connection structure that enables information exchange between each node, such as terminals and servers. Examples of such a network (not shown) include, but are not limited to, a 3GPP (3rd Generation Partnership Project) network, an LTE (Long Term Evolution) network, a 5G network, a WIMAX (World Interoperability for Microwave Access) network, the Internet, a LAN (Local Area Network), a Wireless LAN (Wireless Local Area Network), a WAN (Wide Area Network), a PAN (Personal Area Network), a Wi-Fi network, a Bluetooth network, a satellite broadcasting network, an analog broadcasting network, and a DMB (Digital Multimedia Broadcasting) network.

[0050] In addition, a user terminal (not shown) equipped to mutually transmit and receive data (signals) with a communication device (10) may be, for example, a smartphone, a smartpad, a tablet PC, and any type of wireless communication device such as a PCS (Personal Communication System), GSM (Global System for Mobile communication), PDC (Personal Digital Cellular), PHS (Personal Handyphone System), PDA (Personal Digital Assistant), IMT (International Mobile Telecommunication)-2000, CDMA (Code Division Multiple Access)-2000, W-CDMA (W-Code Division Multiple Access), or Wibro (Wireless Broadband Internet) terminal.

[0051] Additionally, referring to FIG. 2, the open RAN-based communication system (1000) disclosed herein may be designed to build a broadband (e.g., 4.7 GHz level) base station system by applying envelope tracking technology and low-loss packaging matching technology to a high-power power amplifier (1) for a Sub-6 GHz band Massive MIMO base station, for example.

[0052] Specifically, referring to FIG. 2, the fronthole connection between the O-RU (10) and the O-DU (20) can be configured as eCPRI based on wired communication means such as optical fiber. Additionally, the lower PHY layer of the O-RU (10) can be implemented using an FPGA (Field Programmable Gate Array), and may be a layer that performs Fast Fourier Transform (FFT) / Inverse Fast Fourier Transform (IFFT) functions, Cyclic Prefix (CP) addition and removal functions, PRACH (Physical Random Access Channel) filtering functions, digital beamforming functions, etc.

[0053] Additionally, the digital front end (DFE) of the O-RU (10) may be a component that performs DUC (Digital Up-Conversion), DDC (Digital Down-Conversion), CFR (Crest Factor Reduction), DPD (Digital Pre-Distortion), ET (Envelope Tracking), etc., and the RF front end (RF FE) may be a component that includes an array antenna, a bandpass filter, PA (Power Amplifier), LNA (Low Noise Amplifier), DAC (Digital-to-Analog Converter), ADC (Analog-to-Digital Converter), etc.

[0054] FIG. 3 is a conceptual diagram illustrating the operation flow of a communication device equipped with an artificial intelligence-based digital pre-distortion device for improving the non-linearity of a power amplifier according to one embodiment of the present invention.

[0055] Additionally, referring to FIG. 3, the communication device (10) disclosed herein may have an array-type antenna comprising a plurality of individual antenna modules, and the artificial intelligence model (11) of the digital pre-distortion device (100) may be independently applied to each power amplifier (1) provided in each individual antenna module of the communication device (10).

[0056] As another example, the artificial intelligence model (11) of the digital pre-distortion device (100) may operate to perform digital pre-distortion (DPD) integrally on a power amplifier (1) included in an antenna module group that is grouped to include at least two of the plurality of individual antenna modules provided in the communication device (10), but is not limited thereto.

[0057] In this regard, the digital pre-distortion device (100) can perform grouping by considering the physical placement location of the antenna modules. Specifically, since antenna modules located adjacent to each other within an array antenna are likely to be exposed to similar operating environments (e.g., temperature, heat distribution, etc.), these adjacent antenna modules can be set as a single group to apply a common artificial intelligence model (11).

[0058] As another example, the digital pre-distortion device (100) can perform grouping based on the frequency band used by the antenna modules. For example, antenna modules using the same frequency band can be grouped together, or antenna modules having similar bandwidths can be grouped together. This can be understood as a grouping method that takes into account the characteristics that power amplifiers (1) operating in the same frequency band may exhibit similar non-linear distortion characteristics.

[0059] As another example, the digital pre-distortion device (100) can perform grouping by considering the beamforming pattern of the antenna modules. For example, antenna modules that operate together to form a beam in a specific direction can be set as a single group. This can be understood as a grouping method that takes into account the characteristic that antenna modules to which the same beamforming weight is applied are likely to operate at similar power levels.

[0060] In addition, the digital pre-distortion device (100) can perform grouping based on the output power level and modulation method of the antenna module. For example, by grouping antenna modules that operate in a similar output power range or use the same modulation method (e.g., QPSK, 16QAM, 64QAM, etc.) into one group, the accuracy of prediction and correction of the non-linear distortion characteristics of the power amplifier (1) can be improved.

[0061] In addition, the digital pre-distortion device (100) can analyze operation data of antenna modules collected in real time and dynamically group antenna modules that exhibit similar performance indicators (e.g., ACLR, EVM, etc.). Through this dynamic grouping, optimal DPD performance can be maintained even during system operation.

[0062] In addition, according to one embodiment of the present invention, the artificial intelligence model (11) of the digital pre-distortion device (100) may be mounted on a beamforming IC (Integrated Circuit) provided in the communication device (10). Specifically, when the artificial intelligence model (11) is mounted on the beamforming IC, the artificial intelligence model (11) may operate in conjunction with the basic phase and amplitude control functions of the beamforming IC. The beamforming IC can control the direction and shape of the beam by adjusting the phase and amplitude of the signal transmitted to each antenna element, and at this time, the artificial intelligence model (11) may operate to correct the nonlinearity of the power amplifier (1) that occurs during beamforming operation using the beamforming IC in real time.

[0063] More specifically, referring to FIG. 3, the communication device (10) may include a plurality of layers capable of transmitting and receiving signals through a plurality of RF channels. Each layer may perform precoding of an input signal through a pre-coder, and a beamforming IC may perform beamforming on the precoded signal.

[0064] For reference, in the description of the embodiments of the present invention, the beamforming IC may specifically include an RF switch section, a power amplifier (1), a control block, a gain amplifier, and a low-noise amplifier (LNA). The RF switch section may be a sub-module that performs the function of selectively switching between a transmission and a reception path, the control block may be a sub-module that controls each module within the beamforming IC, the gain amplifier may be a sub-module that adjusts the gain of a transmission signal, and the LNA may be a sub-module that performs the function of amplifying a reception signal.

[0065] In this regard, the artificial intelligence model (11) can be run within the beamforming IC using an NPU or edge computing. In particular, the artificial intelligence model (11) can be applied independently for each antenna module, each channel, or each beamforming IC. For example, according to an embodiment of the present invention, one artificial intelligence model (11) may correspond to one antenna module, multiple RF channels may be controlled by one artificial intelligence model (11), and all power amplifiers (1) included in one beamforming IC may be integrally controlled by one artificial intelligence model (11).

[0066] In other words, the artificial intelligence model (11) of the digital pre-distortion device (100) may be driven through a hardware accelerator or a Neural Processing Unit (NPU), but is not limited thereto.

[0067] FIG. 4 is a conceptual diagram showing a detailed circuit of a communication device equipped with an artificial intelligence-based digital pre-distortion device for improving the non-linearity of a power amplifier according to one embodiment of the present invention.

[0068] Referring to FIG. 4, the digital pre-distortion device (100) can be divided into an FPGA part (10a), an analog conversion part (10b), and an RF part (10c).

[0069] Specifically, the FPGA part (10a) receives and processes a frequency domain input signal (A) of the communication device (10), and this input signal (A) is a frequency domain signal prior to the IFFT, and the digital pre-distortion device (100) can analyze it to predict the output characteristics of the power amplifier (1) of the RF part (10c) in advance.

[0070] Meanwhile, the FPGA part (10a) can perform IFFT operations, CP insertion, and windowing processing on the input signal (A), and after processing the signal primarily through a pre-linear distortion compensation filter and a channel filter, perform Digital Pre-Distortion (DPD) processing using an artificial intelligence model (11) described later. In addition, the FPGA part (10a) can extract amplitude information of the signal using a CORDIC (COordinate Rotation Digital Computer) algorithm and generate envelope shaping information based thereon to determine the supply voltage of the power amplifier (1).

[0071] Additionally, the analog conversion part (10b) may be equipped with a DAC that converts digital signals processed in the FPGA part (10a) into analog signals, and an ADC that converts feedback signals from the RF part (10c) into digital signals.

[0072] Additionally, the RF part (10c) is a circuit part responsible for amplifying and transmitting the actual RF signal, and the supply voltage (V) for the power amplifier (1) through the supply modulator bott The power amplifier (1) can adjust the ) and, under the adjusted supply voltage, amplify the RF signal and transmit it through the antenna.

[0073] FIG. 5 is a detailed circuit diagram of an artificial intelligence-based digital pre-distortion device for improving the non-linearity of a power amplifier according to one embodiment of the present invention.

[0074] Referring to FIG. 5, the digital pre-distortion device (100) disclosed herein is mounted as a sub-module of a communication device (10) and may be configured to adjust the supply voltage applied to the power amplifier (1) through a supply modulator of the power amplifier (1) provided in the communication device (10).

[0075] Meanwhile, according to one embodiment of the present invention, the ‘DPD’ block among the submodules of the digital pre-distortion device (100) shown in FIG. 5 can operate to compensate for AM (Amplitude Modulation) and PM (Phase Modulation) errors of the power amplifier (1) using an artificial intelligence model (11) as described in detail below.

[0076] Additionally, the 'Pad out Delay' block among the submodules of the digital pre-distortion device (100) illustrated in FIG. 5 can operate to compensate for timing discrepancies between the ET path and the signal path and within the signal path.

[0077] Below, the specific functions and operations of the digital pre-distortion device (100) will be described in detail.

[0078] First, the digital pre-distortion device (100) can collect learning data including the input signal characteristics and output signal characteristics of the power amplifier (1).

[0079] Additionally, the digital pre-distortion device (100) can collect learning data including an amplifier (Amp) characteristic for at least one of a memory effect indicating the influence of a past input signal of the power amplifier (1) on the current output and a temperature change of the power amplifier (1), and an environmental variable for at least one of a power voltage change of the power amplifier (1), a frequency band limit, and a power limit.

[0080] More specifically, the digital pre-distortion device (100) can collect input signal characteristics of a power amplifier (1) as learning data, including at least one of input power level information corresponding to the power level of an input signal input to the power amplifier (1), frequency band information indicating frequency information of a band supported by the power amplifier (1), and modulation format information indicating a modulation method of the input signal.

[0081] Additionally, the digital pre-distortion device (100) can collect output signal characteristics of a power amplifier (1) as learning data, including at least one of output power information indicating the power level of the output signal of the power amplifier (1), spectrum analysis information indicating spectrum change within the frequency band of the output signal, and non-linear distortion characteristic information including AM-AM conversion characteristics and AM-PM conversion characteristics.

[0082] Meanwhile, regarding output power information, according to one embodiment of the present invention, the digital pre-distortion device (100) can obtain output level information for an output signal scheduled to be output through the communication device (100) in advance through analysis of the frequency domain input signal of the communication device (100).

[0083] In this regard, the spectrum of the envelope of a wireless communication signal has a bandwidth that is approximately 2.5 to 3 times wider than the bandwidth of the I / Q signal. For example, the I / Q baseband signal of a signal having an RF spectrum of 100 MHz of 5G NR can have a bandwidth of 50 MHz, and the envelope signal has a bandwidth of 125 MHz (2.5 times 50 MHz). As the bandwidth increases, it is very difficult to maintain high efficiency, and the open RAN-based communication system (1000) disclosed herein is designed to achieve a high level of supply conversion efficiency (e.g., 80% or more) even with high bandwidth signals.

[0084] According to one embodiment of the present invention, the supply modulator may be designed to support a 3-dB bandwidth of up to 130 MHz or more to track the envelope of a 5G NR 100 MHz signal.

[0085] Specifically, in the case of a supply modulator, large voltage changes cannot be applied within a single CP interval at the operating speed of existing ICs. To resolve this, symbol-unit I / Q data in the frequency domain prior to the IFFT can be acquired, and the system can operate to increase the voltage step by step in advance.

[0086] As another example, the digital pre-distortion device (100) can extract amplitude information using the CORDIC algorithm, but is not limited thereto.

[0087] In summary, the open RAN-based communication system (1000) disclosed herein can extract information on the symbol unit average power from the I / Q data extracted from the IFFT stage, raise the voltage in advance, and transmit a signal to the supply modulator IC to control the voltage in sampling units using the instantaneous power obtained from the CORDIC algorithm.

[0088] Additionally, the digital pre-distortion device (100) can build an artificial intelligence model (11) for predicting and correcting non-linear distortion of the power amplifier (1) using collected training data.

[0089] For example, the artificial intelligence model (11) is based on a multi-layer perceptron (MLP) structure and may be composed of an input layer, multiple hidden layers, and an output layer. Specifically, the input layer of the artificial intelligence model (11) may be configured to receive input signal characteristics, output signal characteristics, memory effects, and environment variables of the power amplifier (1). A Rectified Linear Unit (ReLU) may be used as the activation function of the hidden layer, and a linear activation function may be used in the output layer to output a distortion correction value.

[0090] As another example, the artificial intelligence model (11) may include a Recurrent Neural Network (RNN) structure and a Long Short-Term Memory (LSTM) layer specialized for time-series data processing. This allows for effective learning of the memory effect, which is the influence of past inputs of the power amplifier (1) on the current output.

[0091] Meanwhile, according to one embodiment of the present invention, a digital pre-distortion device (100) can train a lightweight artificial intelligence model (11) suitable for a communication device (10) by applying a lightweighting technique including at least one of pruning and quantization to an artificial intelligence model (11).

[0092] Specifically, when applying the pruning technique, the sparsity of the model can be increased by removing weights of the artificial intelligence model (11) whose absolute value is below a specific threshold. Specifically, pruning can be performed to remove a preset ratio (e.g., about 50%) from the total parameters, which can effectively reduce the model size while minimizing the degradation of accuracy. Additionally, when applying the quantization technique, memory usage can be reduced by converting weights and activation values ​​represented as 32-bit floating-point numbers into 8-bit integers, but is not limited to this.

[0093] At this time, considering that the power level information of the input signal is sufficient with a resolution of 0.1 dB, the digital pre-distortion device (100) quantizes the corresponding parameter into an 8-bit fixed-point number and quantizes the parameters representing the AM-AM conversion characteristic and the AM-PM conversion characteristic into a 16-bit fixed-point number, thereby improving computational efficiency while maintaining precision, but is not limited to this.

[0094] In addition, according to one embodiment of the present invention, the digital pre-distortion device (100) may allow an artificial intelligence model (11) to autonomously learn an optimal correction strategy for correcting distortion occurring in the output of a power amplifier (1) through reinforcement learning.

[0095] Here, the 'optimal correction strategy' may exemplarily refer to a combination of correction parameters that can achieve the best performance under given conditions by considering the trade-off between power efficiency and linearity.

[0096] In addition, when reinforcement learning is applied, the artificial intelligence model (11) acts as an agent, observing the current state (operating condition of the power amplifier) ​​at every time step and selecting an appropriate action (determination of correction value). At this time, the reward can be determined by comprehensively considering signal quality indicators such as ACLR and EVM and power efficiency. Through the application of such reinforcement learning, the artificial intelligence model (11) can autonomously derive the optimal correction strategy by going through trial and error under various operating conditions. This enables more flexible and adaptive correction compared to the existing lookup table-based DPD technique.

[0097] Additionally, the digital pre-distortion device (100) can mount the constructed artificial intelligence model (11) on the communication device (10) equipped with a power amplifier (1) in an on-device form. Here, mounting the artificial intelligence model (11) on the communication device (10) in an 'on-device form' may mean that the artificial intelligence model (11) is directly executed (operated) through hardware inside the communication device (10) rather than a separate cloud server.

[0098] Specifically, on-device mounting can be implemented by storing the weights and structure of a lightweight artificial intelligence model (11) in a dedicated hardware accelerator such as an NPU or FPGA and performing inference operations. Through this, the digital pre-distortion device (100) disclosed herein can correct non-linear distortion in real time without network delay and has the advantage of minimizing power consumption required for data transmission.

[0099] Meanwhile, according to one embodiment of the present invention, a digital pre-distortion device (100) can perform retraining on a pre-established artificial intelligence model (11) using real-time operation data of a power amplifier (1) collected in response to a time-series change in the communication environment of a communication device (10), and can re-install the retrained artificial intelligence model (11) into the communication device (10).

[0100] In this regard, 'time series change of the communication environment' may broadly include temperature changes in the installation environment where the communication device (10) is equipped, fluctuations in power supply voltage, and changes in device characteristics due to aging, and the retraining of the artificial intelligence model (11) may be triggered when specific conditions are satisfied. For example, retraining of the artificial intelligence model (11) may begin when the temperature of the power amplifier (1) changes above a preset threshold (e.g., 10°C, etc.), ACLR performance decreases below a set threshold (e.g., -45dBc, etc.), or EVM deteriorates above a specific level (e.g., 5%, etc.), but is not limited to these cases.

[0101] Meanwhile, according to one embodiment of the present invention, in order to determine a specific model requiring retraining among the artificial intelligence models (11) corresponding to a plurality of individual antenna modules provided in the communication device (10), the digital pre-distortion device (100) can monitor performance indicators for each antenna module. For example, if an abnormal temperature rise is detected in the power amplifier (1) of a specific antenna module or if the output signal quality of the module deteriorates, the digital pre-distortion device (100) can selectively retrain only the artificial intelligence model (11) corresponding to the antenna module.

[0102] Additionally, the digital pre-distortion device (100) can input characteristic information of a real-time signal associated with a power amplifier (1) into an artificial intelligence model (11). Additionally, the digital pre-distortion device (100) can derive a distortion correction value for the input real-time signal through the artificial intelligence model (11).

[0103] Specifically, the digital pre-distortion device (100) can derive a distortion correction value including at least one of a DPD correction parameter including an AM-AM correction value and an AM-PM correction value, a feedback control parameter including a real-time correction control signal and an adaptive parameter adjustment value according to environmental changes of the power amplifier (1), and a signal quality indicator including an EVM (Error Vector Magnitude) and an ACPR (Adjacent Channel Power Ratio) using output data of an artificial intelligence model (11).

[0104] In addition, the digital pre-distortion device (100) can perform correction on a real-time signal based on a distortion correction value output using an artificial intelligence model (11).

[0105] In this regard, a general DPD technique operates by applying inverse nonlinearity to an input signal to overcome the nonlinearity of a power amplifier (1), and the digital pre-distortion device (100) disclosed herein can operate to pre-distort the amplitude and phase of an input signal based on a distortion correction value derived by an artificial intelligence model (11). For example, if the power amplifier (1) exhibits a characteristic of compressing the signal, the artificial intelligence model (11) can derive a correction value that appropriately expands the amplitude of the input signal to compensate for this.

[0106] In addition, to correct AM-PM distortion, the digital pre-distortion device (100) can predict a phase change according to the amplitude change of the input signal and apply a phase correction value to offset it. This correction process is performed in real time in the DPD block within the FPGA, and precise correction can be performed on a sample basis.

[0107] FIG. 6 is a schematic diagram of an artificial intelligence-based digital pre-distortion device for improving the non-linearity of a power amplifier according to one embodiment of the present invention.

[0108] Referring to FIG. 6, the digital pre-distortion device (100) may include a data collection unit (110), a model learning unit (120), and a correction execution unit (130).

[0109] The data collection unit (110) can collect learning data including the input signal characteristics and output signal characteristics of the power amplifier (1).

[0110] Additionally, the data collection unit (110) can collect learning data that further includes an amplifier (Amp) characteristic for at least one of the memory effect indicating the influence of past input signals of the power amplifier (1) on the current output and the temperature change of the power amplifier (1), and an environmental variable for at least one of the power voltage change, frequency band limit, and power limit of the power amplifier (1).

[0111] More specifically, the data collection unit (110) can collect input signal characteristics of the power amplifier (1) as learning data, including at least one of input power level information corresponding to the power level of the input signal input to the power amplifier (1), frequency band information indicating the frequency information of the band supported by the power amplifier (1), and modulation format information indicating the modulation method of the input signal.

[0112] Additionally, the data collection unit (110) can collect output signal characteristics of the power amplifier (1) as learning data, including at least one of output power information indicating the power level of the output signal of the power amplifier (1), spectrum analysis information indicating spectrum change within the frequency band of the output signal, and non-linear distortion characteristic information including AM-AM conversion characteristics and AM-PM conversion characteristics.

[0113] The model learning unit (120) can build an artificial intelligence model (11) for predicting and correcting non-linear distortion of the power amplifier (1) using the collected learning data.

[0114] For example, the model learning unit (120) can train a lightweight artificial intelligence model (11) suitable for a communication device (10) by applying a lightweighting technique including at least one of pruning and quantization to the artificial intelligence model (11).

[0115] In addition, according to one embodiment of the present invention, the model learning unit (120) can enable the artificial intelligence model (11) to autonomously learn an optimal correction strategy for correcting distortion occurring in the output of the power amplifier (1) through reinforcement learning.

[0116] Additionally, the model learning unit (120) can mount the constructed artificial intelligence model (11) on a communication device (10) equipped with a power amplifier (1) in an on-device form.

[0117] Meanwhile, according to one embodiment of the present invention, the model learning unit (120) can perform retraining on a pre-established artificial intelligence model (11) using real-time operation data of a power amplifier (1) collected in response to a time-series change in the communication environment of a communication device (10), and can re-install the retrained artificial intelligence model (11) into the communication device (10).

[0118] The correction execution unit (130) can input characteristic information of a real-time signal associated with a power amplifier (1) into an artificial intelligence model (11). Additionally, the correction execution unit (130) can derive a distortion correction value for the input real-time signal through the artificial intelligence model (11).

[0119] Specifically, the correction execution unit (130) can derive a distortion correction value including at least one of a DPD correction parameter including an AM-AM correction value and an AM-PM correction value, a feedback control parameter including a real-time correction control signal and an adaptive parameter adjustment value according to environmental changes of the power amplifier (1), and a signal quality indicator including an EVM (Error Vector Magnitude) and an ACPR (Adjacent Channel Power Ratio) using the output data of the artificial intelligence model (11).

[0120] In addition, the correction execution unit (130) can perform correction on the real-time signal based on the distortion correction value output using the artificial intelligence model (11).

[0121] Below, based on the details described above, we will briefly examine the operation flow of the present invention.

[0122] FIG. 7 is an operation flowchart of an artificial intelligence-based digital pre-distortion method for improving the non-linearity of a power amplifier according to one embodiment of the present invention.

[0123] The artificial intelligence-based digital pre-distortion method for improving the non-linearity of a power amplifier illustrated in FIG. 7 can be performed by the digital pre-distortion device (100) described above. Therefore, even if the content is omitted below, the description of the digital pre-distortion device (100) can be equally applied to the description of the artificial intelligence-based digital pre-distortion method for improving the non-linearity of a power amplifier.

[0124] Referring to FIG. 7, in step S11, the data collection unit (110) can collect learning data including the input signal characteristics and output signal characteristics of the power amplifier (1).

[0125] Additionally, in step S11, the data collection unit (110) may collect learning data that further includes an amplifier (Amp) characteristic for at least one of a memory effect indicating the influence of past input signals of the power amplifier (1) on the current output and a temperature change of the power amplifier (1), and an environmental variable for at least one of a power voltage change of the power amplifier (1), a frequency band limit, and a power limit.

[0126] More specifically, in step S11, the data collection unit (110) can collect input signal characteristics of the power amplifier (1) as learning data, including at least one of input power level information corresponding to the power level of the input signal input to the power amplifier (1), frequency band information indicating the frequency information of the band supported by the power amplifier (1), and modulation format information indicating the modulation method of the input signal.

[0127] Additionally, in step S11, the data collection unit (110) can collect output signal characteristics of the power amplifier (1) as learning data, including at least one of output power information indicating the power level of the output signal of the power amplifier (1), spectrum analysis information indicating spectrum change within the frequency band of the output signal, and non-linear distortion characteristic information including AM-AM conversion characteristics and AM-PM conversion characteristics.

[0128] Next, in step S12, the model learning unit (120) can build an artificial intelligence model (11) for predicting and correcting non-linear distortion of the power amplifier (1) using the collected learning data.

[0129] For example, in step S12, the model learning unit (120) can train a lightweight artificial intelligence model (11) suitable for a communication device (10) by applying a lightweighting technique including at least one of pruning and quantization to the artificial intelligence model (11).

[0130] In addition, according to one embodiment of the present invention, in step S12, the model learning unit (120) may allow the artificial intelligence model (11) to autonomously learn an optimal correction strategy for correcting distortion occurring in the output of the power amplifier (1) through reinforcement learning.

[0131] Next, in step S13, the model learning unit (120) can mount the constructed artificial intelligence model (11) on a communication device (10) equipped with a power amplifier (1) in an on-device form.

[0132] Next, in step S14, the correction performing unit (130) can input characteristic information of the real-time signal associated with the power amplifier (1) into the artificial intelligence model (11).

[0133] Next, in step S15, the correction performing unit (130) can derive a distortion correction value for the input real-time signal through the artificial intelligence model (11).

[0134] Specifically, in step S15, the correction execution unit (130) can derive a distortion correction value including at least one of a DPD correction parameter including an AM-AM correction value and an AM-PM correction value, a feedback control parameter including a real-time correction control signal and an adaptive parameter adjustment value according to environmental changes of the power amplifier (1), and a signal quality indicator including an EVM (Error Vector Magnitude) and an ACPR (Adjacent Channel Power Ratio) using the output data of the artificial intelligence model (11).

[0135] Next, in step S16, the correction performing unit (130) can perform correction on the real-time signal based on the distortion correction value derived through step S15.

[0136] In the description above, steps S11 through S16 may be further divided into additional steps or combined into fewer steps according to an embodiment of the present invention. Additionally, some steps may be omitted as necessary, and the order of the steps may be changed.

[0137] Figure 8 is a detailed flowchart of the retraining process of an artificial intelligence model.

[0138] The retraining process of the artificial intelligence model illustrated in FIG. 8 can be performed by the digital pre-distortion device (100) described above. Therefore, even if the details are omitted below, the description of the digital pre-distortion device (100) can be applied equally to the description of FIG. 8.

[0139] Referring to FIG. 8, in step S13', the model learning unit (120) can perform retraining on the artificial intelligence model (11) that has already been built using real-time operation data of the power amplifier (1) collected in response to time-series changes in the communication environment of the communication device (10).

[0140] Next, in step S14', the model learning unit (120) can re-load the retrained artificial intelligence model (11) into the communication device (10).

[0141] In the description above, steps S13' through S14' may be further subdivided into additional steps or combined into fewer steps, depending on the embodiment of the present invention. Additionally, some steps may be omitted as necessary, and the order of the steps may be changed.

[0142] An artificial intelligence-based digital pre-distortion method for improving the nonlinearity of a power amplifier according to one embodiment of the present invention may be implemented in the form of program instructions that can be executed through various computer means and recorded on a computer-readable medium. The computer-readable medium may include program instructions, data files, data structures, etc., either individually or in combination. The program instructions recorded on the medium may be those specifically designed and configured for the present invention, or may be those known and available to those skilled in the art of computer software. Examples of computer-readable recording media include magnetic media such as hard disks, floppy disks, and magnetic tapes; optical recording media such as CD-ROMs and DVDs; magneto-optical media such as floptical disks; and hardware devices specifically configured to store and execute program instructions, such as ROM, RAM, and flash memory. Examples of program instructions include machine code, such as that generated by a compiler, as well as high-level language code that can be executed by a computer using an interpreter, etc. The above-described hardware device may be configured to operate as one or more software modules to perform the operation of the present invention, and vice versa.

[0143] In addition, the artificial intelligence-based digital pre-distortion method for improving the nonlinearity of the aforementioned power amplifier can also be implemented in the form of a computer program or application executed by a computer and stored on a recording medium.

[0144] The foregoing description of the present invention is for illustrative purposes only, and those skilled in the art will understand that other specific forms can be easily modified without altering the technical concept or essential features of the present invention. Therefore, the embodiments described above should be understood as illustrative in all respects and not restrictive. For example, each component described as a single unit may be implemented in a distributed manner, and components described as distributed may likewise be implemented in a combined form.

[0145] The scope of the present invention is defined by the claims set forth below rather than by the detailed description above, and all modifications or variations derived from the meaning and scope of the claims and the concept of equivalents thereof should be interpreted as being included within the scope of the present invention.

Claims

1. In an artificial intelligence-based digital pre-distortion method for improving the non-linearity of a power amplifier, (a) A step of collecting training data including input signal characteristics and output signal characteristics of a power amplifier; (b) constructing an artificial intelligence model for predicting and correcting nonlinear distortion of the power amplifier using the above training data, and mounting the artificial intelligence model on a communication device equipped with the power amplifier in an on-device form; (c) inputting characteristic information of a real-time signal associated with the power amplifier into the artificial intelligence model; and (d) a step of deriving a distortion correction value for the real-time signal through the artificial intelligence model and correcting the real-time signal based on the derived distortion correction value, A digital pre-distortion method including 2. In Paragraph 1, The above input signal characteristics are, It includes at least one of input power level information corresponding to the power level of an input signal input to the power amplifier, frequency band information indicating frequency information of a band supported by the power amplifier, and modulation format information indicating a modulation method of the input signal. The above output signal characteristics are, A digital pre-distortion method comprising at least one of output power information indicating the power level of the output signal of the power amplifier, spectrum analysis information indicating spectrum change within the frequency band of the output signal, and non-linear distortion characteristic information including AM-AM conversion characteristics and AM-PM conversion characteristics.

3. In Paragraph 2, The above training data is, Amplifier characteristics for at least one of memory effect and temperature change indicating the influence of past input signals of the power amplifier on the current output; and Environmental variables for at least one of the power supply voltage change, frequency band limit, and power limit of the above power amplifier, A digital pre-distortion method that further includes 4. In Paragraph 1, The above distortion correction value is, A digital pre-distortion method comprising at least one of a DPD correction parameter including an AM-AM correction value and an AM-PM correction value, a feedback control parameter including a real-time correction control signal and an adaptive parameter adjustment value according to environmental changes of the power amplifier, and a signal quality indicator including Error Vector Magnitude (EVM) and Adjacent Channel Power Ratio (ACPR).

5. In Paragraph 1, The above communication device is equipped with an array-type antenna comprising a plurality of individual antenna modules, and A digital pre-distortion method in which the artificial intelligence model is applied independently to each of the power amplifiers provided in each of the individual antenna modules, or is applied to the power amplifiers included in an antenna module group that is grouped to include at least two of the individual antenna modules.

6. In Paragraph 1, A digital pre-distortion method characterized in that the above artificial intelligence model is mounted on a beamforming IC provided in the above communication device.

7. In Paragraph 1, The above step (b) is, A step of applying a lightweighting technique comprising at least one of pruning and quantization to the artificial intelligence model, A digital pre-distortion method that includes 8. In Paragraph 1, The above step (b) is, A step in which the artificial intelligence model autonomously learns an optimal correction strategy to correct distortion occurring in the output of the power amplifier through reinforcement learning, Includes, The above artificial intelligence model is, A digital pre-distortion method characterized by being driven through a hardware accelerator or a Neural Processing Unit (NPU).

9. In Paragraph 1, A step of collecting real-time operation data of the power amplifier in response to time-series changes in the communication environment of the communication device, retraining the artificial intelligence model, and re-loading the retrained artificial intelligence model onto the communication device. A digital pre-distortion method that further includes 10. An artificial intelligence-based digital pre-distortion device for improving the non-linearity of a power amplifier, A data acquisition unit that collects learning data including input signal characteristics and output signal characteristics of a power amplifier; A model learning unit that constructs an artificial intelligence model for predicting and correcting nonlinear distortion of the power amplifier using the above training data, and mounts the artificial intelligence model in an on-device form on a communication device equipped with the power amplifier; and A correction execution unit that inputs characteristic information of a real-time signal associated with the power amplifier into the artificial intelligence model, derives a distortion correction value for the real-time signal through the artificial intelligence model, and corrects the real-time signal based on the derived distortion correction value. A digital pre-distortion device including 11. In Paragraph 10, The above input signal characteristics are, It includes at least one of input power level information corresponding to the power level of an input signal input to the power amplifier, frequency band information indicating frequency information of a band supported by the power amplifier, and modulation format information indicating a modulation method of the input signal. The above output signal characteristics are, A digital pre-distortion device comprising at least one of output power information indicating the power level of the output signal of the power amplifier, spectrum analysis information indicating spectrum change within the frequency band of the output signal, and non-linear distortion characteristic information including AM-AM conversion characteristics and AM-PM conversion characteristics.

12. In Paragraph 10, The above distortion correction value is, A digital pre-distortion device comprising at least one of a DPD correction parameter including an AM-AM correction value and an AM-PM correction value, a feedback control parameter including a real-time correction control signal and an adaptive parameter adjustment value according to environmental changes of the power amplifier, and a signal quality indicator including Error Vector Magnitude (EVM) and Adjacent Channel Power Ratio (ACPR).

13. In Paragraph 10, The above communication device is equipped with an array-type antenna comprising a plurality of individual antenna modules, and A digital pre-distortion device in which the artificial intelligence model is applied independently to each of the power amplifiers provided in each of the individual antenna modules, or is applied to the power amplifiers included in an antenna module group that is grouped to include at least two of the individual antenna modules.

14. In Paragraph 10, A digital pre-distortion device characterized in that the above artificial intelligence model is mounted on a beamforming IC provided in the above communication device.

15. In Paragraph 10, The above model learning unit is, A digital pre-distortion device that collects real-time operation data of the power amplifier in response to time-series changes in the communication environment of the communication device, retrains the artificial intelligence model, and re-installs the retrained artificial intelligence model into the communication device.

Images